# Мудрецы и шляпы Задачу представляют Бурсиан О., Кохась K., Латышев А., Ретинский В. ## 1 Знакомьтесь: игра в шляпы Пусть дан неориентированный граф $G$, в каждой вершине которого находится один мудрец и один сундук со шляпами разного цвета. Мудрецы знакомы друг с другом, граф $G$, расположение мудрецов по вершинам графа и все цвета шляп в сундуках зафиксированы и известны всем. В частности, каждый мудрец понимает, в какой вершине находится каждый из его соседей. Судья проводит с мудрецами следующий тест. Он надевает каждому мудрецу шляпу из его сундука. Каждый мудрец видит только шляпы мудрецов, находящихся в соседних вершинах графа, но не видит своей шляпы и не знает ее цвета. Мудрецы не общаются. По команде судьи мудрецы одновременно называют цвет. Считается, что мудрецы успешно прошли тест = «выиграли», если хотя бы один из них угадал. Перед тестом мудрецам сообщили правила теста и дали возможность устроить совещание, на котором они должны определить публичную стратегию. Публичность означает, что все, включая судью, знают ее. Стратегия мудрецов должна быть детерминированной - каждый мудрец должен назвать цвет, исходя только из того, какие цвета он видит у соседей. Будем говорить, что стратегия выигрышная, если при любой раскладке шляп хотя бы один мудрец угадает цвет надетой на него шляпы. Также будем говорить, что мудрецы выигрывают, если они имеют выигрышную стратегию, и проигрывают, если не имеют. Вот пара задач на этот сюжет. 1.1. Каждому из двух мудрецов надевают шляпу белого, синего или красного цвета, и сообщают, что цвета их шляп разные. Каждый из них видит шляпу другого, но не видит свою. Они должны одновременно попытаться угадать цвет своей шляпы, написав его на бумажках. Докажите, что мудрецы могут заранее договориться действовать так, чтобы хотя бы один из них угадал. 1.2. Султан устроил экзамен 11 придворным мудрецам. По правилу экзамена султан размещает 10 мудрецов в 10 ям, расположенных по кругу, и еще одного мудреца сажает на вышку в центре круга. На лбу у каждого из первых 10 мудрецов султан пишет число 1 или 2 ; на лбу у центрального мудреца султан пишет число от 1 до 1024. Мудрец на вышке видит числа на всех остальных мудрецах, а те видят его число (но не видят друг друга). Все мудрецы должны одновременно попытаться угадать свои числа. Султан заранее объяснил мудрецам правила экзамена и дал им время посовещаться до начала экзамена. Могут ли мудрецы действовать так, чтобы хотя бы один из них заведомо угадал свое число? Мы будем отождествлять вершину графа $G$ и мудреца, который в ней находится. Будем считать, что цвета пронумерованы числами $0,1,2,3, \ldots$, и что в сундуке мудреца $X$ лежат шляпы с цветами от 0 до некоторого числа $h(X)-1$, множество (цветов) шляп в сундуке мудреца $X$ будем обозначать $\mathrm{Col} X$. Игрой в шляпы назовем пару $\langle G, h\rangle$, где $G=\langle V, E\rangle$ - граф, $h: V \rightarrow \mathbb{N}-$ функция, сопоставляющая вершине количество (цветов) шляп, лежащих в сундуке в этой вершине. Функцию $h$ мы будем называть «шляпностью». Вместо $h(A)$ мы будем иногда писать $\widehat{A}$. Если нам известно, что в игре $\langle G, h\rangle$ мудрецы выигрывают, будем говорить, что $\langle G, h\rangle$ выигрышный граф. Выигрышный граф будем называть простым, если он не содержит выигрышного подграфа $\left\langle G^{\prime}, h^{\prime}\right\rangle$, где $G^{\prime} \varsubsetneqq G$, $h^{\prime}=\left.h\right|_{V\left(G^{\prime}\right)}$. 1.3. Докажите, что на графе на рис. 1 мудрецы выигрывают. 1.4. Является ли выигрышный граф на рис. 1 простым? 1.5. Докажите, что на графе «путь $A_{1} A_{2} \ldots A_{n}$, где $n \geqslant 2$, а шляпности указаны на рис. 2 , мудрецы выигрывают. 1.6. Является ли выигрышный граф на рис. 2 простым? 1.7. Докажите, что на графе $K_{2,3}$ с указанными шляпностями (рис. 3) мудрецы выигрывают. 1.8. Могут ли мудрецы выиграть на графе $K_{1,3}$ с указанными шляпностями (рис. 4)? ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-002.jpg?height=163&width=177&top_left_y=133&top_left_x=177) Рис. 1. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-002.jpg?height=66&width=523&top_left_y=207&top_left_x=504) Рис. 2. Путь из $n$ вершин ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-002.jpg?height=160&width=182&top_left_y=137&top_left_x=1211) Рис. 3. Граф $K_{2,3}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-002.jpg?height=172&width=188&top_left_y=134&top_left_x=1665) Рис. 4. Граф $K_{1,3}$ 1.9. Докажите, что на цикле $C_{n}, n \geqslant 4$ с указанными шляпностями (рис. 5) мудрецы выигрывают. 1.10. На цикле $C_{n}, n \geqslant 4$ с указанными шляпностями (рис. 6) мудрецы выигрывают. 1.11. Существует ли простой выигрышный граф, на котором при любом раскладе шляп не менее двух мудрецов угадывают свой цвет? 1.12. Докажите, что на полном двудольном графе $K_{99,50}$ мудрецы проигрывают, если шляпность всех мудрецов равна 100 . 1.13. Мудрецы расположены на графе «путь $A B$ » (рис. $7, n \geqslant 4$ ). Перед совещанием судья пообещал мудрецам, что $A$ и $B$ получат шляпы одинаковых цветов. Смогут ли мудрецы выиграть? ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-002.jpg?height=177&width=391&top_left_y=894&top_left_x=204) Рис. 5. Цикл из $n$ вершин ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-002.jpg?height=177&width=391&top_left_y=894&top_left_x=844) Рис. 6. Цикл из $n$ вершин ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-002.jpg?height=103&width=532&top_left_y=965&top_left_x=1393) Рис. 7. Путь $A B$ из $n$ вершин 1.14. Пусть мудрецы играют на графе $G$ с функцией шляпности $h$, но судья заранее пообещал им, что в тот момент, когда он наденет шляпу на мудреца $A$, он шепнет ему на ухо истинное утверждение вида «я тебе только что надел шляпу одного из двух цветов: $c_{1}$ или $c_{2}$ ». Таким образом, мудрецы во время совещания знают о том, что судья сделает подсказку, но не знают, какие именно цвета он назовет. Поэтому мудрецы определяют стратегии всех, кроме $A$, как обычно, а мудрец $A$ получает набор из $C_{h(A)}^{2}$ стратегий - по одной на каждую возможную подсказку судьи. Докажите, что эта подсказка не повлияет на исход игры. ## 2 Игры на кликах 2.1. Существует ли выигрышный граф, не содержащий в качестве подграфа 4 -клики, на котором шляяпности всех мудрецов равны 4 ? 2.2. Мудрецы находятся в вершинах полного графа $K_{n}$, у $i$-го мудреца в сундуке $a_{i}$ шляп разных цветов. Докажите, что мудрецы выигрывают в том и только том случае, если $$ \frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}} \geqslant 1 $$ 2.3. Назовем стратегию мудрецов точной, если на каждом раскладе шляп угадывает ровно один мудрец. Найдите все игры, в которых возможна точная стратегия. 2.4. Граф $G$ - это полный граф на $n$ вершинах, из которого удалили одно ребро. Шляпности всех мудрецов равны $n$. Могут ли мудрецы выиграть на таком графе? 2.5. Граф $G$ - это полный граф на 4 вершинах $A, B, C, D$, из которого удалили ребро $C D$. При этом $h(A)=6, h(B)=6, h(C)=2, h(D)=3$. Могут ли мудрецы выиграть на таком графе? 2.6. Граф $G$ - это полный граф на $n$ вершинах $A_{1}, A_{2}, \ldots, A_{n}$, из которого удалили ребро $A_{n-1} A_{n}$. Шляпности вершин равны $a_{1}, \ldots, a_{n}$. Оказалось, что граф выигрышный и при этом $$ \frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}-\frac{1}{a_{n-1} a_{n}}=1 $$ Докажите, что $a_{1} a_{2} \ldots a_{n-2}$ делится на $a_{n-1} a_{n}$. 2.7. Могут ли мудрецы выиграть на графе «Средний бантик» (рис. 8)? 2.8. Могут ли мудрецы выиграть на графе «Большой бантик» (рис. 9)? ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-003.jpg?height=226&width=194&top_left_y=178&top_left_x=543) Рис. 8. Граф «Средний бантик» ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-003.jpg?height=268&width=417&top_left_y=137&top_left_x=1208) Рис. 9. Граф «Большой бантик» ## 3 Операции с графами Пусть $\left\langle G_{1}, h_{1}\right\rangle,\left\langle G_{2}, h_{2}\right\rangle$ - две игры такие, что $V_{1} \cap V_{2}=\{v\}$. Пусть $G=G_{1}+G_{v}$ - объединение графов $G_{1}$ и $G_{2}$, в котором обе вершины $v$ объединены в одну вершину. Пусть функция $h: V_{1} \cup V_{2} \rightarrow \mathbb{N}$ совпадает с $h_{i}$ на $V_{i} \backslash\{v\}$ $(i=1,2)$ и $h(v)=h_{1}(v) h_{2}(v)$. Игру $\langle G, h\rangle$ будем обозначать $G_{1} \times{ }_{v} G_{2}$ (рис. 10). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-003.jpg?height=263&width=1196&top_left_y=771&top_left_x=430) Рис. 10. Игра $G_{1} \times{ }_{v} G_{2}$ 3.1. Теорема о произведении. Если мудрецы выигрывают на графах $G_{1}$ и $G_{2}$, то они выигрывают и на графе $G_{1} \times{ }_{v} G_{2}$. Пусть $G_{1}$ и $G_{2}$ - два графа, не имеющие общих вершин. Подстановкой графа $G_{2}$ в граф $G_{1}$ на место вершини $v$ назовем граф, получающийся объединением графов $G_{1} \backslash v$ и $G_{2}$ с добавлением ребер, которые соединяют каждую вершину $G_{2}$ с каждым соседом $v$. Такую подстановку будем обозначать $G_{1}\left[v:=G_{2}\right]$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-003.jpg?height=188&width=1220&top_left_y=1482&top_left_x=428) Рис. 11. Подстановка графа на место вершины 3.2. Теорема о подстановке. Пусть мудрецы выигрывают в игры $\left\langle G_{1}, h_{1}\right\rangle$ и $\left\langle G_{2}, h_{2}\right\rangle$. Пусть $G-$ это граф подстановки $G_{1}\left[v:=G_{2}\right]$, где $v \in G_{1}$ - произвольная вершина. Тогда игра $\langle G, h\rangle$ выигрышная, где $$ h(u)= \begin{cases}h_{1}(u) & u \in G_{1} \\ h_{2}(u) \cdot h_{1}(v) & u \in G_{2}\end{cases} $$ 3.3. Пусть $\langle G, h\rangle$ - игра, в которой мудрецы выигрывают, $B C$ - ребро графа $G$. Рассмотрим граф $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, получающийся добавлением к графу $G$ новой вершины $A: V^{\prime}=V \cup\{A\}$, $E^{\prime}=E \cup\{A B, A C\}$. Тогда мудрецы выигрывают в игре $\left\langle G^{\prime}, h^{\prime}\right\rangle$ (см. рис. 12), где $$ h^{\prime}(u)= \begin{cases}2, & u=A \\ 2 h(B), & u=B \\ 2 h(C), & u=C \\ h(u), & \text { для остальных вершин } u \in V .\end{cases} $$ $\Rightarrow$ Рис. 12. Добавление вершины $A$ шляпности 2 к ребру $B C$ 3.4. Пусть $\left\langle G_{1}, h_{1}\right\rangle,\left\langle G_{2}, h_{2}\right\rangle$ - две игры, в которых мудрецы выигрывают. Пусть $A_{1}, A_{2}, \ldots$, $A_{k} \in V_{1} ; B_{1}, B_{2}, \ldots, B_{m} \in V_{2}$. Рассмотрим граф $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, получающийся добавлением к графу $G_{1} \cup G_{2}$ всех ребер $A_{i} B_{j}: V^{\prime}=V_{1} \cup V_{2}, E^{\prime}=E_{1} \cup E_{2} \cup\left\{A_{i} B_{j}, i=1, \ldots, k ; j=1, \ldots, m\right\}$ (рис. 13). Тогда мудрецы выигрывают в игре $\left\langle G^{\prime}, h^{\prime}\right\rangle$, где $$ h^{\prime}(u)= \begin{cases}h_{1}(u), & u \in G_{1} \backslash\left\{A_{1}, A_{2}, \ldots, A_{k}\right\} \\ h_{2}(u), & u \in G_{2} \backslash\left\{B_{1}, B_{2}, \ldots, B_{m}\right\} \\ h_{1}(u)+1, & u \in\left\{A_{1}, A_{2}, \ldots, A_{k}\right\} \\ h_{2}(u)+1, & u \in\left\{B_{1}, B_{2}, \ldots, B_{m}\right\}\end{cases} $$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-004.jpg?height=243&width=965&top_left_y=1035&top_left_x=543) Рис. 13. Склейка двух графов, $k=2, m=3$ 3.5. Пусть $\langle G, h\rangle$ - игра, в которой мудрецы выигрывают, $Z, C \in V$ - две вершины графа $G$. Рассмотрим граф $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, получающийся добавлением к графу $G$ нового пути $Z A B C: V^{\prime}=$ $V \cup\{A, B\}, E^{\prime}=E \cup\{Z A, A B, B C\}$. Тогда мудрецы выигрывают в игре $\left\langle G^{\prime}, h^{\prime}\right\rangle$, где $h^{\prime}(Z)=2 h(Z)$, $h^{\prime}(C)=h(C)+1, h^{\prime}(A)=2$ и $h^{\prime}(B)=3$, для остальных вершин $h^{\prime}$ совпадает с $h$ (рис. 14). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-004.jpg?height=249&width=742&top_left_y=1623&top_left_x=654) Рис. 14. Добавление двух вершин шляпностей 2 и 3 соединенных ребром 3.6. Пусть граф $\left\langle G, h^{\prime}\right\rangle$ выигрышный, $V(G)=\left\{A_{1}, \ldots, A_{n}\right\}$. Возьмем $n$ выигрышных графов $\left\langle G_{i}, h_{i}\right\rangle$ и отметим в каждом из них одну вершину $A_{i}$. Построим на отмеченных вершинах граф $G$. В этом «суперграфе» зададим шляпность $h\left(A_{i}\right)=h^{\prime}\left(A_{i}\right) h_{i}\left(A_{i}\right)$ (а у прочих вершин шляпность та же, что и в родных графах). Тогда мудрецы выигрывают. 3.7. Добавление к графу $G$ висячей вершины $A$, где $\widehat{A}>2$, не влияет на выигрышность. 3.8. а) Добавление к графу $G$ с вершиной $A$ шляпности 2 двух новых вершин $B$ и $C$ шляпности 5 и ребер $A B, A C, B C$ не влияет на выигрышность. b) Добавление к графу $G$ с вершиной $A$ шляпности 3 двух новых вершин $B$ и $C$ шляпности 7 и 5 и ребер $A B, A C, B C$ не влияет на выигрышность. 3.9. Пусть $G_{1}$ и $G_{2}$ - графы с вершиной $A$, на которых мудрецы проигрывают в случае, когда $\widehat{A}=2$. Тогда мудрецы проигрывают на графе $G=G_{1}+_{A} G_{2}$, в котором $\widehat{A}=2$. 3.10. Пусть $H=\langle G, h\rangle$ - проигрышная игра, $B$ - любая вершина графа $G$. Рассмотрим граф $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, получающийся добавлением к графу $G$ новой висячей вершины $A: V^{\prime}=V \cup\{A\}$, $E^{\prime}=E \cup\{A B\}$. Тогда мудрецы проигрывают в игре $\left\langle G^{\prime}, h^{\prime}\right\rangle$, где $h(A)=2, h^{\prime}(B)=2 h(B)-1$ и $h^{\prime}(u)=h(u)$ для остальных вершин $u \in V$. ## 4 Шахматы вслепую Игра «Шах ладъей». Два шахматиста $\mathcal{L}$ и $\mathcal{R}$ сидят напротив друг друга, на стене за спиной у каждого закреплена «его» шахматная доска так, что каждый шахматист не видит свою доску, но видит доску соперника. Судья выставляет на каждую из досок одного короля. После этого шахматисты независимо друг от друга должны назвать одну клетку на своей шахматной доске и судья ставит на эту клетку ладью. Если хотя бы один из королей оказался под боем ладьи (или ладья поставлена на ту же клетку, где стоит король), шахматисты выиграли, в противном случае они проиграли. Шахматные доски игроков могут быть произвольного и неодинакового размера. Как и в игре «Натs», шахматисты заранее вырабатывают публичную детерминированную стратегию. Судья знает эту стратегию и играет против шахматистов. Пусть граф $G$ - это цикл $A B C D A$, а шляпность задается функцией $h$. На самом деле граф $G$ - это полный двудольный граф $K_{2,2}$, из-за чего мудрецы $A$ и $C$ видят одно и то же, и мудрецы $B$ и $D$ также видят одно и то же. Тогда мы назовем пару игроков $A$ и $C$ шахматистом $\mathcal{L}$ и будем считать, что его доска имеет размеры $h(A) \times h(C)$, а пару $B$ и $D$ - шахматистом $\mathcal{R}$ и будем считать, что его доска имеет размеры $h(B) \times h(D)$. Таким образом игра «Hats» на цикле $A B C D A$ с функцией шляпности $h$ эквивалентна игре «Шах ладьей» на досках $L(h(A) \times h(C)$ ) и $R(h(B) \times h(D))$. Крестом в клетчатом прямоугольнике будем называть объединение любого вертикального и любого горизонтального ряда клеток. Крест однозначно задается клеткой, находящейся на пересечении этих рядов, она называется центром креста. Ладья, стоящая в центре креста, держит под боем в точности все клетки креста. В задаче 4.1 можно сдавать пункты по отдельности. 4.1. В игре «Шах ладьей» шахматисты выигрывают на следующих парах досок: $\mathrm{W} 1$ ) одна из досок имеет размер $1 \times k, k$ - любое натуральное число; W2) $L(2 \times k)$ и $R(2 \times m)$, где $k$ и $m$ - любые натуральные числа; W3) $L(3 \times 3), R(3 \times 3)$; W4) $L(2 \times 3), R(3 \times 4)$; W5) $L(2 \times 4), R(3 \times 3)$; W6) $L(2 \times 2), R(k \times m)$, где $\min (k, m) \leqslant 4$. На следующих парах досок шахматисты проигрывают: L1) $L(2 \times 3), R(4 \times 4)$; $\quad$ L2) $L(2 \times 3), R(3 \times 5)$; L3) $L(2 \times 4), R(3 \times 4)$; L4) $L(3 \times 3), R(3 \times 4)$; L5) $L(2 \times 2), R(5 \times 5)$; L6) $L(2 \times 5), R(3 \times 3)$. Для прочих размеров досок применима логика «мажоризации». Например, шахматисты проигрывают на досках $L(3 \times 4), R(3 \times 4)$, потому что они проигрывают даже в более простом случае $\mathrm{L} 3$. Шахматисты выигрывают на досках $L(2 \times 3), R(3 \times 3)$, потому что они выигрывают даже, когда одна из досок крупнее, как в W3. Шах ферзем. Рассмотрим вариант игры, в котором оба шахматиста выставляют на доску не ладью, а ферзя. Будем называть эту игру «Шах ферзем». 4.2. В игре «Шах ферзем» $L(4 \times 5), R(4 \times 5)$ шахматисты выигрывают. 4.3. В игре «Шах ферзем» $L(4 \times 4), R(5 \times 5)$ шахматисты выигрывают. 4.4. В игре «Шах ферзем» $L(7 \times 8), R(7 \times 7)$ шахматисты проигрывают. 4.5. В игре «Шах ферзем» $L(3 \times 4), R(7 \times 7)$ шахматисты проигрывают. 4.6. В игре «Шах ферзем» $L(4 \times 5), R(5 \times 5)$ шахматисты проигрывают. 4.7. Рассмотрим вариант игры «Шах ферзем», в котором 5 шахматистов расположены так, что каждый из них видит доски всех остальных, но не видит свою собственную. Все доски имеют размер $11 \times 11$ Как и в исходной игре, судья ставит на каждую доску короля, а шахматисты независимо указывают, куда поставить ферзя. Смогут ли шахматисты выиграть? ## 5 Еще несколько задач 5.1. Пусть $G$ - граф с вершинами $B$ и $C$. Пусть функция шляпности такова, что $\widehat{B}=\widehat{C}=2$ и при этом граф проигрышный. Добавим к графу новую вершину $A$, которая соединена только с $B$ и $C$. Тогда мудрецы проигрывают на полученном графе, если $\widehat{A}=2, \widehat{B}=3, \widehat{C}=7$, а шляпности остальных вершин не изменились. 5.2. Верно ли, что существует такое «большое» число $k$, что любой граф $G$, в котором степени всех вершин не превосходят 3 , а шляпности всех вершин равны $k$, проигрышный? 5.3. Верно ли, что существует такое «большое» число $d$, что любой граф $G$, в котором степени всех вершин равны $d$, а шляпности всех вершин равны 4 , выигрышный? ## РЕШЕНИЯ 1.1. Они могут разместить цвета по циклу, например, белый $\rightarrow$ синий $\rightarrow$ красный $\rightarrow$ белый, и далее пусть каждый из них, увидев цвет шляпы приятеля, запишет цвет, следующий за ним по циклу. ## 1.2. Ответ: да. Можно считать, что число на вышке записано в двоичной системе счисления, а мудрецам в ямах пишут на лбу числа 0 или 1 . Пусть $k$-й мудрец в яме отыгрывает стратегию «мой бит не совпадает с $k$-м битом на вышке», а мудрец на вышке просто составляет число из битов в ямах. 1.3. Считаем, что цвета - это числа от 0 до 3. Пусть верхний мудрец называет четность суммы цветов шляп двух других мудрецов. А двое остальных называют цвет, исходя из гипотез «сумма цветов всех шляп дает остаток 1 по модулю 4» и «сумма цветов всех шляп дает остаток 3 по модулю 4 ». ## 1.4. Ответ: да. Достаточно проверить, что при удалении одного ребра получается проигрышный граф, т.е. что граф «путь из двух звеньев» со шляпностями $2,4,4$ или $4,2,4$ проигрышный. Применим вероятностные соображения: так, для случая $2,4,4$, доли общего числа раскладов, на которых выигрывают мудрецы, равны соответственно $1 / 2,1 / 4,1 / 4$, что в сумме составляет как раз 1. Но случаи, в которых угадывают мудрецы в крайних вершинах, независимы: на $1 / 8$ доле всех раскладов угадывают оба крайних мудреца. Таким образом, суммарная доля всех раскладов, на которых угадывают мудрецы, равна $\frac{1}{2}+\frac{1}{4}+\frac{1}{4}-\frac{1}{8}<1$. Аналогично разбирается другой случай. 1.5. То, что мудрецы выигрывают при $n=2$, когда шляпы всего двух цветов, широко известно: первый мудрец отыгрывает гипотезу «у нас одинаковые шляпы», а второй - «у нас разные шляпы». Несколько раз применяя утверждение задачи 3.1, получаем отсюда утверждение для остальных $n$. 1.6. Ответ: да. Достаточно проверить, что граф «путь $P_{k}$ »о шляпностями $2,4, \ldots, 4$ проигрышный. С помощью конструктора задачи 3.7 это утверждение легко сводится к проверке проигрышности игры двух человек со шляпностями 2 и 4. Но можно затеять рассуждение по индукции. Пусть $A$ - мудрец шляпности $2, B$ - его сосед. По стратегии мудрец $A$ в зависимости от цвета $B$ (четыре варианта) называет один из двух цветов, скажем, красный и синий. Пусть в двух или более случаях мудрец $A$ называет синий цвет («плохие» расклады), а в остальных случаях - красный («хорошие» расклады). Пусть судья даст мудрецу $A$ красную шляпу, и объявит, что мудрецу $B$ будут выдаваться шляпы каких-то двух цветов, соответствующих плохому раскладу. Тогда мудрец $A$ заведомо не угадает, а мудрец $B$ может считать его шляпность равна 2. Игра свелась к более короткому пути того же вида. 1.7. Пусть в одной доле находятся две вершины $X$ и $Y$, а в другой доле - вершины $A, B, C$, причем $\widehat{Y}=2, \widehat{X}=\widehat{A}=\widehat{B}=\widehat{C}=4$. Опишем выигрышную стратегию мудрецов. Каждый из четырех цветов можно истолковывать как двухбитное двоичное число, состоящее из левого и правого бита. Мудрец, собирающийся назвать цвет определенной четности, уже определился с правым битом и должен выбрать одно из двух значений левого бита. Цвет шляпы $Y$ - один бит - мы также будем интерпретировать как четность. Пусть мудрец $Y$ назовет ту четность $y$, которая преобладает среди шляп $A, B, C$. Каждый из мудрецов $A, B, C$ видит шляпу $Y$. Поэтому они понимают, что если большинство шляп $A, B, C$ имеют ту же четность, что и шляпа $Y$, то мудрец $Y$ угадает цвет своей шляпы. Значит, им надо исходить из предположения, что большинство их шляп имеют четность $\bar{Y}$. Поэтому, называя свои ответы, каждый из них будет выбирать из двух возможных цветов, имеющих четность $\bar{Y}$. С другой стороны, мудрец $X$ так же, как и $Y$, видит, какая четность преобладает среди шляп $A, B, C$, и знает, чему равно $y$. Таким образом, мудрец $X$ знает, что если на мудреце $Y$ надета шляпа цвета $y$, то тот угадал. Поэтому мудрец $X$ должен исходить из предположения, что на мудреце $Y$ надета шляпа цвета $\bar{y}$. Но тогда он знает, что в этом случае мудрецы $A, B, C$ исходят из предположения, что большинство их шляп имеет четность $y$. Итак, мудрец $X$ видит, какая четность доминирует среди цветов $A, B, C$, и знает, что мудрецы $A, B, C$ исходят из предположения, что у них именно эта четность и доминирует. Пусть мудрец $X$ выберет среди мудрецов $A, B, C$ двоих со шляпами доминирующей четности, назовем этих мудрецов главными. Теперь покажем, как «предположения» превратить в стратегию. Столбцы следующей таблицы соответствуют цветам шляпы $X$, а строки - мудрецам $A, B, C$. Пусть каждый из мудрецов $A, B, C$ возьмет для своего ответа левый бит из своей строки таблицы (а правый бит определен «предположением»). | | 0 | 1 | 2 | 3 | | :--- | :--- | :--- | :--- | :--- | | $A$ | 0 | 0 | 1 | 1 | | $B$ | 0 | 1 | 0 | 1 | | $C$ | 0 | 1 | 1 | 0 | Мудрец $X$ видит, угадал ли кто-то из главных мудрецов среди $A, B, C$ цвет своей шляпы по указанной стратегии. Если оба главных мудреца не угадали, то это значит, что они оба указали неверный левый бит цвета своей шляпы. Пусть, например, главные мудрецы - это $A$ и $B$, пусть они указали соответственно биты 0 и 1 и не угадали. Мудрец $A$ указывает бит 0 , только если цвет шляпы $X$ равен 0 или 1 ; мудрец $B$ указывает бит 1 , только если цвет шляпы $X$ равен 1 или 3. Таким образом, оба события происходят только в случае, когда цвет шляпы $X$ равен 1 , именно этот цвет и должен назвать мудрец $X$. Аналогичные действия мудрец $X$ выполняет во всех остальных случаях. Его успех гарантируется следующим Волшебным свойством таблицы. Если в таблице выбрать две произвольные строки и в каждой из них (независимо от другой) закрасить две клетки, содержащие одинаковые символы, то ровно один столбеи, содержит две закрашенные клетки. Волшебное свойство проверяется (очевидным) перебором. Отметим, что для выигрышности стратегии хватило бы более слабого требования «не более чем один столбец таблицы содержит две закрашенные клетки». 1.8. Допустим, что у мудрецов есть выигрышная стратегия. Пусть $v-$ вершина степени $3, u_{1}$, $u_{2}, u_{3}$ - висячие вершины. Назначим вершине $v$ первый цвет. Пусть мудрецы $u_{1}, u_{2}, u_{3}$ согласно стратегии называют цвета $h_{1}, h_{2}, h_{3}$. Теперь проведем второй эксперимент: назначим вершине $v$ второй цвет. Пусть мудрецы $u_{1}$, $u_{2}, u_{3}$ согласно стратегии называют цвета $e_{1}, e_{2}, e_{3}$. Наконец, проведем финальный эксперимент. Для каждого $i=1,2,3$ обозначим через $d_{i}$ цвет, который не был назван мудрецом $u_{i}$ в первых двух экспериментах (если есть выбор берем любой цвет из двух возможных). Для каждого $i$ назначим висячей вершине $u_{i}$ цвет $d_{i}$. Цвета шляп у соседей мудреца $v$ уже заданы, значит, известен его ответ по стратегии. Назначим вершине $v$ тот из цветов - первый или второй, который не совпадает с этим ответом. Мудрецы проиграли. Это рассуждение работает, даже если $\widehat{v}=2$. 1.9. Это утверждение получается применением конструктора из задачи 3.4 , где в качестве $G_{2}$ берется граф, состоящий из одной вершины шляпности 1! Действительно, мудрецы выигрывают на пути, изображенном на рис. 2, а описанный в задаче граф получается добавлением к этому пути одновершинного графа $G_{2}$, при этом в месте склейки шляпности увеличиваются на 1. 1.10. Это утверждение получается применением конструктора из задачи 3.5. Действительно, мудрецы выигрывают на пути, изображенном на рис. 2, а описанный в задаче граф получается добавлением к этому пути нового ребра со шляпностями вершин 2 и 3 и подходящим изменением шляпностей вершин, к которым они прикрепляются. 1.11. Ответ: нет. Пусть $\langle G, a\rangle$ - простой выигрышный граф, $A \in V(G)$. Дадим мудрецу $A$ шляпу цвета 0 . Тогда для остальных мудрецов зафиксировалась стратегия на проигрышном графе $G \backslash A$. Значит, мудрецам на графе $G \backslash A$ можно раздать шляпы так, что никто из них не угадает. В полученном раскладе шляп на всем графе $G$ угадать может только мудрец $A$. 1.12. Приведем стратегию судьи. Сначала рассмотрим на доле $A$, состоящей из 50 мудрецов, 51 расклад шляп, когда все мудрецы из $A$ получают шляпы одного цвета, и этот цвет один из первых 51. На каждого из мудрецов второй доли $B$ наденем ту шляпу, которую он не назовет по своей стратегии ни для одного из 51 раскладов. Зафиксируем построенный на доле $B$ расклад и посмотрим, что по стратегии скажут для него мудрецы из $A$. Их 50, значит, какой-то цвет из первых 51 не назовет никто. Наденем шляпы этого цвета на всех мудрецов доли $A$. В результате никто из мудрецов не угадал. Приведенное рассуждение работает для всех двудольных графов, у которых в минимальной доле не более $k-2$ вершин, где $k$ - шляпность мудрецов. Если одна из долей содержит $k-1$ вершину, то мудрецы выигрывают в случае, когда размер второй доли очень велик. 1.13. Ответ: да, мудрецы выигрывают. Пусть мудрец $A$ играет по стратегии «Вижу 2 - говорю 2 , иначе говорю 0 », а мудрец $B$ - по стратегии «Вижу 2 - говорю 2 , иначе говорю $1 »$, Мудрецы $C$ и $D$, если видят на своем одиноком ( $A$ или $B$ ) соседе шляпу цвета 0 или 1 , говорят 2 . В противном случае они полагают, что у них не 2 , и играют с остальными мудрецами в игру из задачи 1.5. По этой стратегии мудрецы действительно выиграют. Потому что либо у мудрецов $A$ и $B$ шляпа цвета 0 или 1 и тогда кто-то из этих четырех выиграет сразу, либо у них шляпа цвета 2 и тогда у мудрецов $C$ и $D$ точно не 2 (иначе $A$ или $B$ угадали), и тогда $C$ и $D$ выигрывают на пути между ними. 1.14. Допустим, что мудрецы выигрывают с подсказкой. Зафиксируем для всех мудрецов, кроме $A$, стратегии, которые они используют при игре с подсказкой, и покажем, как можно задать стратегию мудреца $A$, чтобы с этим набором стратегий мудрецы выигрывали без подсказки. Допустим, что при назначении мудрецу $A$ шляпы цвета $x$ нашелся расклад шляп на всем гра$ф е$, в котором мудрец $A$ получил цвет $x$, соседи мудреца $A$ получили цвета $u, v, w, \ldots$, остальные мудрецы тоже получили какие-то цвета, и при этом никто из мудрецов (исключая $A$ ) не угадал. Тогда мы хотим, чтобы в этой ситуации мудрец $A$ угадал свой цвет, т. е. его стратегия должна удовлетворять требованию $f_{A}(u, v, w, \ldots)=x$. Эти требования, полученные для разных раскладов, не противоречат друг другу. Действительно, если бы существовал еще один расклад, где у соседей по-прежнему цвета $u, v, w, \ldots$, а мудрец $A$ получил другой цвет $y$, то мудрецы не могли бы выиграть с подсказкой $A^{*}$, поскольку, имея эти два расклада, можно сообщить мудрецу $A$, что цвет его шляпы $x$ или $y$, после чего реализовать тот из раскладов, для которого он не угадывает цвет своей шляпы. ## Игра на клике ## 2.1. Ответ: да. Например «бантик» (рис. 15). С помощью конструктора из задачи 3.4 можно строить подобные примеры для любого $n$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-008.jpg?height=129&width=214&top_left_y=2431&top_left_x=927) Рис. 15. Бантик 2.2. Так как $i$-й мудрец угадывает на $\frac{1}{a_{i}}$-й доле всех раскладов, то если сумма меньше 1 , найдется расклад, где никто не угадает. Докажем, что если сумма больше либо равна 1, мудрецы выигрывают. Мы предлагаем два решения. Решение 1 (теорема Холла). Зафиксируем $i$ (номер какого-то мудреца) и разобьем множество всех раскладов шляп на подмножества по $a_{i}$ элементов. В каждом раскладе шляп удалим цвет $c_{i}$ и для оставшегося набора $c=\left(c_{1}, \ldots, c_{i-1}, \widehat{c_{i}}, c_{i+1}, \ldots, c_{n}\right)$ (шляпка означает, что цвет пропущен) положим $$ A_{c}^{i}=\left\{\left(c_{1}, \ldots, c_{i-1}, x, c_{i+1}, \ldots, c_{n}\right): x \in \operatorname{Col}\left(A_{i}\right)\right\} $$ Имея в виду применение теоремы Холла о паросочетаниях, назовем множества $A_{c}^{i}$ «девушками», а сами расклады шляп - «юношами». Будем говорить, что юноша и девушка знакомы, если расклад шляп является элементом множества $A_{c}^{i}$. Каждый юноша знаком с $n$ девушками, при этом для каждого $i$ каждый юноша знаком ровно с одной девушкой вида $A_{c}^{i}$. Каждая девушка $A_{c}^{i}$ знакома ровно с $a_{i}$ юношами. Докажем, что существует паросочетание, сопоставляющее каждому юноше девушку. Для этого достаточно проверить условие теоремы, утверждающее, что каждые $m$ юношей знакомы вместе с не менее чем $m$ девушками. Рассмотрим произвольный набор из $m$ юношей. Так как при каждом $i$ девушка $A_{c}^{i}$ знакома ровно с $a_{i}$ юношами, $m$ юношей для каждого $i$ знакомы в сумме не менее чем с $m / a_{i}$ девушками вида $A_{c}^{i}$. Суммируя по $i$, получаем, что общее число знакомых девушек не меньше $\frac{m}{a_{1}}+\frac{m}{a_{2}}+\ldots+\frac{m}{a_{n}} \geqslant m$. Условие теоремы Холла выполнено. Итак, существует паросочетание, которое каждому раскладу шляп ставит в соответствие множество вида $A_{c}^{i}$. Отметим, что при выполнении равенства $\frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}=1$ это паросочетание фактически отмечает один элемент в каждом множестве $A_{c}^{i}$. Если же выполняется неравенство $\frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}>1$, то «останутся одинокие девушки», т. е. в некоторых множествах $A_{c}^{i}$ может не оказаться ни одного отмеченного элемента. Построенное паросочетание позволяет задать стратегию мудрецов. Пусть $j$-й мудрец действует по правилу: увидев шляпы других мудрецов, т. е. набор цветов $c=\left(c_{1}, \ldots, c_{j-1}, c_{j+1}, \ldots, c_{n}\right)$, он однозначно восстанавливает множество $A_{c}^{j}$, которое фактически состоит из всевозможных способов дополнить набор с до расклада шляп. При этом текущий расклад шляп является одним из элементов этого множества. Мудрец должен назвать тот цвет, который в множестве $A_{c}^{j}$ отмечен нашим паросочетанием (а если отмеченного элемента нет, называет цвет произвольно). Поскольку каждый расклад шляп отображается нашим паросочетанием как отмеченный элемент одного из множеств $A_{c}^{i}$, для этого расклада шляп $i$-й мудрец угадает цвет. Решение 2 (явная стратегия). Пусть $N=\operatorname{HOK}\left(a_{1}, a_{2}, \ldots, a_{n}\right)$, для $k$ от 1 до $n$ положим $d_{k}=$ $N / a_{k}$. Будем отождествлять множество возможных цветов шляпы $k$-го мудреца с множеством остатков $d_{k}, 2 d_{k}, \ldots, a_{k} d_{k}$ по модулю $N$. Пусть мудрецам даны шляпы: $k$-й мудрец получает шляпу цвета $x_{k} d_{k}$, где $x_{k} \in\left\{1,2, \ldots, a_{k}\right\}$. Положим $S=x_{1} d_{1}+x_{2} d_{2}+\ldots+x_{k} d_{k}(\bmod N)$. Каждый мудрец, видя окружающих, может написать все слагаемые этой суммы кроме своего собственного. Делая предположение о цвете своей шлляпы, он получает предполагаемое значение всей суммы. Пусть первый мудрец проверяет гипотезу $S \in\left\{1,2, \ldots, d_{1}\right\}$; второй мудрец проверяет гипотезу $S \in\left\{d_{1}+1, d_{1}+2, \ldots, d_{1}+d_{2}\right\}$ и т. д., $n$-й мудрец проверяет гипотезу $S \in\left\{d_{1}+d_{2}+\ldots+d_{n-1}+1, \ldots, d_{1}+d_{2}+\ldots+d_{n-1}+d_{n}\right\}$. Гипотеза $k$-го мудреца затрагивает $d_{k}$ последовательных остатков, среди которых ровно один делится на $d_{k}$. Именно этот остаток и определяет цвет шляпы, который должен назвать $k$-й мудрец. 2.3. Ответ: точные стратегии существуют только на кликах и только в том случае, когда выполняется равенство $$ \frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}=1 $$ Пусть в графе имеются две несмежные вершины $A$ и $B$. Тогда выдадим произвольный расклад шляп всем мудрецам, кроме $A$ и $B$. Теперь ответы мудрецов $A$ и $B$ определены по стратегии. Выдадим им такие шляпы, чтобы они угадали. На построенном раскладе шляп $A, B$ и, возможно, кто-то еще угадают. Следовательно, стратегия не является точной. То, что в случае клики существование точной стратегии эквивалентно равенству (1), следует из доказательства задачи 2.2 . 2.4. Ответ: нет, мудрецы проигрывают. Пару мудрецов $A, B$ можно интерпретировать как шахматиста с доской $n \times n$. Нетрудно видеть, что доля общего числа раскладов, на которых он выигрывает, равна $\frac{2}{n}-\frac{1}{n^{2}}$. Что касается остальных мудрецов, каждый выигрывает на $\frac{1}{n}$-й части всех раскладов. Таким образом, суммарная доля раскладов, на которых кто-то выигрывает, меньше 1. 2.5. Ответ: да, мудрецы выигрывают. Будем интерпретировать цвета шляп $A$ и $B$ как вычеты по модулю 6 , цвет $C$ - как вычет по модулю 2 , цвет $D$ - как вычет по модулю 3 . Обозначим цвета шляп мудрецов $A$ и $B$ через $a$ и $b$. Пусть $C$ называет цвет $c=(a+b) \bmod 2, D-$ цвет $d=(a+b) \bmod 3$. Если мудрецы $C$ и $D$ не угадали, выполняются равенство $a+b=c+1 \bmod 2$ и $(a+b=d+1 \bmod 3$ или $a+b=d+2 \bmod 3)$. Тогда пусть $A$ вычисляет свой цвет в предположениях $a+b=c+1 \bmod 2$ и $a+b=d+1 \bmod 3$; а $B-$ в предположениях $a+b=c+1 \bmod 2$ и $a+b=d+2 \bmod 3$. Предлагаем читателю в качестве упражнения получить этот же результат, используя конструктор задачи 3.3 . 2.6. Пусть $X$ - множество раскладов шляп для первых $n-2$ мудрецов, т. е., иначе говоря, это множество наборов из $n-2$ цветов, где первый цвет - это возможный цвет шляпы мудреца $A_{1}$, второй цвет - мудреца $A_{2}$ и т. д., $(n-2)$-й цвет - мудреца $A_{n-2}$. Пусть $\alpha=a_{1} a_{2} \ldots a_{n-2}$, тогда $|X|=\alpha$. Обозначим через $L_{i}\left(i=1,2, \ldots, a_{n-1}\right)$ - подмножества $X$, такие что если мудрец $A_{n-1}$ видит комбинацию из $L_{i}$, то он называет цвет $i$. Аналогично определим множества $R_{j}(j=1,2$, $\ldots, a_{n}$ ) для мудреца $A_{n}$. Пусть $L_{k}$ - множество $L_{i}$ минимальной мощности, $\left|L_{k}\right|=M \leqslant \frac{\alpha}{a_{n-1}}$. Теперь рассмотрим множества $R_{j} \backslash L_{k}\left(j=1,2, \ldots, a_{n}\right)$. В них суммарно $\alpha-M$ элементов, поэтому если $R_{m} \backslash L_{k}$ - минимальное по мощности, то $\left|R_{m} \backslash L_{k}\right| \leqslant \frac{\alpha-M}{a_{n}}$. Следовательно, $$ \begin{aligned} \left|L_{k} \cup R_{m}\right| \leqslant M+\frac{\alpha-M}{a_{n}} & =\frac{\alpha}{a_{n}}+M\left(1-\frac{1}{a_{n}}\right) \leqslant \frac{\alpha}{a_{n}}+\frac{\alpha}{a_{n-1}}\left(1-\frac{1}{a_{n}}\right)= \\ & =\alpha\left(\frac{1}{a_{n-1}}+\frac{1}{a_{n}}-\frac{1}{a_{n-1} a_{n}}\right)=\alpha\left(1-\sum_{i=1}^{n-2} \frac{1}{a_{i}}\right)=\alpha-\frac{\alpha}{a_{1}}-\ldots-\frac{\alpha}{a_{n-2}} \end{aligned} $$ Таким образом, когда на мудреце $A_{n-1}$ надета шляпа цвета $k$, а на мудреце $A_{n}$ - цвета $m$, выигрывать должен кто-то из мудрецов $A_{1}, A_{2}, \ldots A_{n-2}$, и число раскладов, для которых происходит это событие, составляет долю $\frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n-2}}$ от общего числа раскладов. Но как мы знаем, указанная доля ограничивает сверху число раскладов, на которых могут выиграть мудрецы $A_{1}, A_{2}$, $\ldots A_{n-2}$. Следовательно, оба неравенства (1) должны обращаться в равенство. Тогда $\left|L_{k}\right|=\frac{\alpha}{a_{n-1}}$ (и вообще $\left|L_{i}\right|=\frac{\alpha}{a_{n-1}}$ для всех $i$ ) и $\left|R_{m} \backslash L_{k}\right|=\frac{\alpha}{a_{n}}-\frac{\alpha}{a_{n-1} a_{n}}$. Аналогично $\left|R_{j}\right|=\frac{\alpha}{a_{n}}$, значит, $\left|R_{m} \cap L_{k}\right|=\frac{\alpha}{a_{n-1} a_{n}}$, и $\alpha$ делится на $a_{n-1} a_{n}$. 2.7. Ответ: да, мудрецы выигрывают. Пусть $G_{1}$ и $G_{2}-4$-клики, из которых склеен бантик, $A$ - общая вершина клик, цвета мудрецов - остатки по модулю 5 . Мудрецы $G_{1} \backslash A$ подсчитывают сумму остатков-цветов на $G_{1}$ и разыгрывают гипотезы для трех последовательных значений остатка этой суммы, скажем, для 2,3 и 4. Аналогично поступают мудрецы на $G_{2} \backslash A$. Что же делать мудрецу $A$ ? Он должен ухитриться отыграть сразу две гипотезы на $G_{1}$ и две гипотезы на $G_{2}$. Спасает его только то, что обе гипотезы - об одной и той же его несчастной шляпе! Чтобы стратегия сработала, надо по разному конвертировать цвета $A$ в остатки по модулю 5 , например, так: | Цвет | Красный | Синий | Белый | Желтый | Черный | | :---: | :---: | :---: | :---: | :---: | :---: | | Его код в $G_{1}$ | 0 | 1 | 2 | 3 | 4 | | Его код в $G_{2}$ | 0 | 2 | 4 | 1 | 3 | Указанная кодировка обладает тем свойством, что любая пара, состоящая из последовательных остатков в строке $G_{1}$, пересекается с любой парой последовательных остатков в строке $G_{2}$ не более чем по одному элементу. Мудрец $A$ вычисляет два последовательных остатка, соответствующих гипотезам «сумма всех остатков на $G_{1}$ равна 0 или 1 » и «сумма всех остатков на $G_{2}$ равна 0 или $1 »$, и называет цвет пересечения этих пар, если он есть. 2.8. Ответ: да, мудрецы и здесь выигрывают. Рассмотрим вычеты по модулю $740=4 \cdot 5 \cdot 37$. Пусть цвета шляп мудреца шляпности $k-$ вычеты, делящиеся на $\frac{740}{k}$. Тогда рассмотрим $S_{1}$ и $S_{2}$ - суммы вычетов-цветов в левой и правой 5 -кликах. Пусть левый мудрец шляпности 4 предполагает, что $S_{1}$ лежит в множестве $\{1,2, \ldots, 185\}$. Он получает 185 подряд идущих вычетов, среди которых ровно один делится на 185. Цвет, соответствующий этому вычету, мудрец и называет. Аналогично три левых мудреца шляпности 5 предполагают, что $S_{1}$ лежит в множествах $\{186, \ldots 333\},\{334, \ldots 481\},\{482, \ldots 629\}$ соответственно. Так же, но только работая с $S_{2}$, поступают правые мудрецы. Если еще никто не выиграл, то $A$ (мудрец со шляпностью 37) понимает, что $S_{1}$ лежит в $\{630, \ldots 720\}$, так что ему нужно выбирать из не более чем 6 подряд идущих цветов. То же можно сказать про $S_{2}$. Чтобы все сработало, давайте мудрецы левой и правой клики будут по-разному переводить цвета мудреца $A$ в вычеты. По нашему правилу цвета шляп мудреца $A$ - это вычеты по модулю 740 , делящиеся на 20, т.е. фактически это вычеты по модулю 37. Если мудрецы левой клики сопоставляют некоторому цвету вычет $x$, то пусть мудрецы правой клики сопоставляют этому же цвету вычет $6 x$ (mod 37) (отображение $x \mapsto 6 x$ взаимно однозначно на множестве остатков при делении на 37). Как нетрудно проверить, любые два множества вида $\{x, x+1, \ldots, x+5\}$ и $\{6 y, 6 y+6, \ldots, 6 y+30\}$ пересекаются не более чем по одному элементу. Значит, $A$ назовет цвет из пересечения этих множеств, либо назовет произвольный цвет, если пересечение пусто. ## Конструкторы 3.1. Цвет шляпы мудреца $v$ в графе $G_{1} \times_{v} G_{2}$ можно интерпретировать как пару цветов $\left(c_{1}, c_{2}\right)$, где $c_{i}$ - цвет шляпы $v$ в графе $G_{i}$. Пусть зафиксированы выигрышные стратегии для графов $G_{1}$ и $G_{2}$. Построим стратегию на $G_{1} \times_{v} G_{2}:$ пусть все мудрецы, кроме $v$, играют по выигрышной стратегии для соответствующего графа. Что касается мудреца $v$, он будет играть сразу по обеим стратегиям: пусть он независимо даст два ответа $c_{1}$ и $c_{2}$ - ответ $c_{i}$ соответствует его выигрышной стратегии для графа $G_{i}$ (для вычисления этого ответа мудрец $v$ смотрит только на своих соседей из графа $\left.G_{i}\right)$. Пара $\left(c_{1}, c_{2}\right)$ - это корректный цвет шляпы мудреца $v$ на графе $G_{1} \times_{v} G_{2}$, это и есть ответ мудреца $v$ в конструируемой стратегии. Построенная стратегия выигрышная, потому что либо кто-то из $G_{1}$ или из $G_{2}$ угадает цвет, либо $v$ угадает обе компоненты своего цвета. 3.2. Пусть $f_{1}$ и $f_{2}$ - выигрышные стратегии игр на графах $G_{1}$ и $G_{2}$ соответственно. Также пусть каждый мудрец $u$ из подграфа $G_{2}$ графа $G$ получает композитный цвет $\left(c_{1}, c_{2}\right)$, где $0 \leqslant c_{1} \leqslant h_{1}(v)-1,0 \leqslant c_{2} \leqslant h_{2}(u)-1$. Тогда все «левые половины» этих мудрецов отыгрывают стратегию $f_{1}(v)$, а «правые половины» этих мудрецов отыгрывают стратегию $f_{2}$. В частности это означает, что все мудрецы из подграфа $G_{2}$ будут называть цвета, у которых первая компонента одна и та же. Что касается остальных мудрецов из $G$, те из них, кто не является соседом $v$, играют по стратегии $f_{1}$. Мудрецы из $G_{1}$, являющиеся соседями мудреца $v$, после подстановки обнаружили, что вместо одного соседа $v$ у них теперь имеется $\left|V_{2}\right|$ соседей (причем, вообще говоря, с разными шляпами). Эти мудрецы поступают следующим образом: все они вместо одной шляпы $v$ видят все шляпы на подграфе $G_{2}$ и знают стратегии мудрецов на этом подграфе. Значит, они понимают, кто выигрывает в игре на подграфе $G_{2}$, обозначим этого игрока $v_{\text {new }}$ (если победителей несколько, они выбирают в качестве $v_{\text {new }}$ одного из победителей, например, того, кто идет первым в заранее составленном списке). В результате, каждый (бывший) сосед $v$ смотрит только на $v_{\text {new }}$, точнее, на первую компоненту его цвета, и тоже играет по стратегии $f_{1}$. В результате либо угадает кто-то из подграфа $G_{1} \backslash\{v\}$, либо $v_{\text {nеw }}$ угадает левую компоненту своего цвета, а правую компоненту он угадывает по своему определению. Таким образом, всегда кто-нибудь угадает свой цвет. 3.3. Цвета шляп игроков $B$ и $C$ можно интерпретировать как пары вида $(c, \epsilon)$, где $c$ - возможный цвет шляпы в игре $\langle G, h\rangle, \epsilon \in\{0,1\}$. Пусть мудрец $A$ отыгрывает гипотезу $c(A)=\epsilon_{B}+\epsilon_{C}(\bmod 2)$. Мудрецы $B$ и $C$ видят своих соседей в графе $G$ и знают, какие цвета $c(B), c(C)$ они должны назвать по выигрышной стратегии в игре $\langle G, h\rangle$. Видя шляпу мудреца $A$, а также шляпы друг друга, мудрецы $B$ и $C$ могут вычислить, какие значения $\epsilon_{B}$ и $\epsilon_{C}$ им следует взять в дополнение к $c(B), c(C)$. 3.4. По сравнению с исходными графами у мудрецов $A_{i}$ и $B_{j}$ добавился один новый цвет, будем считать, что этот цвет - красный. Пусть «мегамудрец» $A$ говорит, что у него все шляпы красные, если видит хоть одну красную шляпу у мудреца $B$, в противном случае пусть $A$ играет по обычной стратегии в графе $G_{1}$. Если «мегамудрец» $B$ видит хоть одну красную шляпу у $A$, то он понимает, что $A$ выиграл, если у $B$ есть хоть одна красная шляпа. Тогда $B$ должен позаботиться о тех раскладах, где у него нет ни одной красной шляпы, и просто отыгрывать свою стратегию на графе $G_{2}$. Если же «мегамудрец» $B$ не видит ни одной красной шляпы у $A$, то он понимает, что $A$ выиграл, только если у $B$ тоже нет красных шляп, тогда, чтобы проконтролировать оставшиеся расклады, $B$ говорит, что у него все шляпы красные. 3.5. Мы укажем стратегию для новых вершин и частично для старых, затем разберем несколько раскладов, на которых они выиграют «сразу», а затем скажем, что на остальных они выиграют при помощи стратегии для графа $G$. Опишем выигрышную стратегию. Будем обозначать $c_{x}$ цвет шляпы, которую получил мудрец $x$. Цвет вершины $Z$ будем считать «композитным»: $c_{Z}=(\epsilon, C)$, где первый «бит» $\epsilon$ может принимать значения 0 и 1 , а второй цвет $C$ - те $h(Z)$ цветов, которые были изначально в графе $G$. - Мудрец $A$, в случае если видит на мудреце $B$ шляпу цвета 0 или 1 , говорит то, что видит на нем, иначе говорит первый бит цвета $c_{Z}$. - Мудрец $B$, в случае если видит на мудреце $C$ шляпу нового цвета, говорит «2», а иначе он говорит значение разности $1-c_{A}$. Разберем все варианты пар $\left(c_{A}, C_{B}\right)$ и опишем остальную часть стратегии, заодно доказав, что она выигрышная. В случаях $(0,0)$ и $(1,1)$ мудрец $A$ сразу угадывает. В случаях $(0,1)$ и $(1,0)$, если не дать мудрецу $C$ шляпу нового цвета, то угадает мудрец $B$. Поэтому мудрец $C$, видя на мудреце $B$ шляпу цвета отличного от 2 , может смело называть шляпу нового цвета и кто-то из $A$, $B, C$ угадает. Остались варианты $(0,2)$ и $(1,2)$. В них мудрец $A$ угадает, если его цвет совпадет с первым битом $c_{Z}$, поэтому $Z$ может считать, что у него первый бит отличается от цвета $c_{A}$, то есть ему нужно угадывать только второй «бит», то есть $h(Z)$ возможных значений. В то же время мудрец $B$ в этих двух случаях угадает только в случае, когда у мудреца $C$ новый цвет. То есть $C$, видя на мудреце $B$ шляпу цвета 2 , может считать, что нового цвета у него нет. Таким образом, нужно выиграть в игру $\langle G, h\rangle$, что мы умеем делать по условию. 3.6. Утверждение непосредственно следует из теоремы о произведении (задача 3.1). 3.7. Пусть $A$ - новая висячая вершина, $B$ - соседняя с ней вершина графа $G$, обозначим граф с добавленной вершиной через $G_{1}$. В одну сторону утверждение очевидно. Если игра на $G$ выигрышная, то игра на $G_{1}$ тоже выигрышная. Докажем теперь, что если игра на $G_{1}$ выигрышная, то и на $G_{1}$ тоже выигрышная. Пусть мудрецы выбрали выигрышную стратегию на $G_{1}$. Напомним, что в задаче 1.14 доказано, что если во время теста судья дает одному из мудрецов, скажем $B$, подсказку вида «у тебя шляпа одного из двух цветов $c_{1}$ или $c_{2}$, то эта подсказка не повлияет на исход игры. Назовем эту подсказку $B^{*}$. Докажем, что если мудрецы выигрывают в игру на $G_{1}$, то они смогут выиграть на графе $G$ с подсказкой $B^{*}$. Пусть фиксирована выигрышная стратегия $f$ на графе $G_{1}$. Построим выигрышную стратегию на $G$ с подсказкой $B^{*}$. Пусть все мудрецы из $V(G) \backslash B$ пользуются стратегией $f$. Для каждой пары цветов $\left(b_{1}, b_{2}\right), b_{1} \neq b_{2}$, которые можно выдать мудрецу $B$, найдем такой цвет $a \in \operatorname{Col} A$, что $f_{A}\left(b_{1}\right) \neq a, f_{A}\left(b_{2}\right) \neq a$, такой цвет найдется, поскольку $\widehat{A} \geqslant 3$. Теперь зададим стратегию мудреца $B$ на графе $G$ с учетом подсказки, а именно, пусть в случае, когда $B$ видит у соседей на графе $G$ комплект цветов $c$ и получает подсказку «твоя шляпа цвета $b_{1}$ или $b_{2}$ », он выдает ответ $f_{B}(a, c)$, т. е. отвечает по стратегии $f$, как если бы он видел на $G_{1}$ цвет $a$ у мудреца $A$ и комплект цветов $c$ у остальных соседей. При этом может оказаться, что цвет, который назовет $B$, не совпадет ни с $b_{1}$, ни с $b_{2}$. Эта стратегия выигрышная, потому что когда в игре на $G_{1}$ мы даем во время теста мудрецу $A$ шляпу цвета $a$ и следим, чтобы он не угадал (не выдавая $B$ какие-то шляпы) то для мудреца $B$ остаются возможными помимо прочего цвета $b_{1}$ и $b_{2}$, и на графе $G_{1}$ кто-то угадывает. Значит, с подсказкой $B^{*}$ угадывают. 3.8. Докажем более общий факт, из которого следуют сразу оба утверждения. Добавление к графу $G$ с вершиной $A$ двух новых вершин $B$ и $C$ и ребер $A B, B C, C A$ не влияет на выигрышность, если шляпности новых вершин удовлетворяют условию $2(\widehat{B}+\widehat{C})<\widehat{B} \cdot \widehat{C}$ (шляпности вершин графа $G$ не изменились). Пусть $G^{\prime}$ - полученный граф. Если мудрецы выигрывают на графе $G$, то они выигрывают и на $G^{\prime}$. Проверим, что из выигрышности $G^{\prime}$ следует выигрышность $G$. Пусть $f^{\prime}$ - выигрышная стратегия на $G^{\prime}$. Построим $f$ - выигрышную стратегию на $G$. Для мудрецов из $G \backslash A$ стратегия остается прежней. Зададим стратегию мудреца $A$. Пусть дан расклад шляп на графе $G$. Мудрец $A$ видит цвета всех своих соседей в графе $G$ и рассматривает $\widehat{B} \cdot \widehat{C}$ способов выбрать цвета для $B$ и $C$. Далее мудрец $A$ смотрит, что он должен отвечать на все эти варианты согласно стратегии $f^{\prime}$, и называет цвет, встречающийся в его ответах чаще всего (точнее, не реже всего). Покажем, что полученная стратегия выигрывает на графе $G$. Зафиксируем расположение шляп на графе $G$ и предположим, что ни один мудрец из $G \backslash A$ не угадал. Если на мудрецов $B$ и $C$ надеть шляпы произвольным образом, то мы получим расстановку шляп на $G^{\prime}$, в которой все мудрецы из $G \backslash A$ не угадают свой цвет по стратегии $f^{\prime}$. Значит, его угадает $A, B$ или $C$. Но из имеющихся $\widehat{B} \cdot \widehat{C}$ случаев расстановки шляп на $B$ и $C$ мудрец $B$ угадает в $\widehat{C}$ случаях, а мудрец $C-$ в $\widehat{B}$ случаях. Значит, $A$ должен угадать свой цвет не менее чем в $\widehat{B} \cdot \widehat{C}-(\widehat{B}+\widehat{C})>\frac{\widehat{B} \cdot \widehat{C}}{2}$ случаях. Таким образом, мудрец $A$ угадывает свой цвет в большинстве вариантов, и как раз этот цвет он угадает по стратегии $f$. 3.9. Обозначим через $N_{i}$ множество соседей мудреца $A$ в графе $G_{i}$, а через $S$ множество всевозможных раскладов шляп у мудрецов из $N_{2}$. Если $x$ - один из двух возможных цветов шляпы мудреца $A$, то второй цвет мы будем обозначать $\bar{x}$. Зафиксируем для мудрецов на графе $G$ произвольную стратегию $f$ и опишем стратегию чертика, по которой он сможет обыграть мудрецов. Выберем произвольный расклад шляп $s \in S$ у соседей $A$ в подграфе $G_{2}$. Этот расклад однозначно задает стратегию $f^{s}$ мудрецов в подграфе $G_{1}$, которая, как мы знаем, проигрышная на этом подграфе. Выберем произвольный опровергающий расклад шляп $\varphi_{s}$ на графе $G_{1}$ для стратегии $f^{s}$. Тогда мудрец $A$ получает шляпу $\varphi_{s}(A)$ и не угадывает, это значит, что выбранный расклад шляп $\varphi_{s}$ определяет такой расклад шляп $t=\left.\varphi_{s}\right|_{N_{1}}$ мудрецам из $N_{1}$, что $f_{A}(t, s)=f_{A}^{s}(t)=\overline{\varphi_{s}(A)}$. Пусть чертик строит опровергающие расклады шляп на всем графе $G$, применяя следующий принцип: если в конструируемом опровергающем раскладе шляп предполагается дать мудрецам из $N_{2}$ расклад шляп $s$, то при этом всем мудрецам в подграфе $G_{1}$ будет выдан расклад $\varphi_{s}$. При соблюдении этого принципа, во-первых, мудрец $A$ и все остальные мудрецы на подграфе $G_{1}$ заведомо не угадают цвет своей шляпы, а во-вторых, стратегия мудреца $A$ теперь полностью определяется лишь раскладами шляп из $S$ (потому что на компоненте $G_{1}$ мы сразу же задаем расклад $h_{s}$ и никаких других вариантов не рассматриваем). Итак, чертик видит, что мудрецы из $V\left(G_{2}\right) \backslash A$ применяют стратегию $f$, а мудрец $A$ факти- ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-014.jpg?height=60&width=1811&top_left_y=1483&top_left_x=128) ку граф $G_{2}$ проигрышный, для этой стратегии существует опровергающий расклад шляп $\psi$ на подграфе $G_{2}$. Этот расклад шляп позволяет чертику корректно задать опровергающий расклад шляп на всем графе $G$. Действительно расклад шляп $\psi$ определяет набор $s=\left.\psi\right|_{N_{2}}$ цветов шляп у соседей $A$ в подграфе $G_{2}$, а набор $s$ задает опровергающий расклад шляп $\varphi_{s}$ на подграфе $G_{1}$, при этом расклады $\psi$ и $\varphi_{s}$ совместимы: оба расклада назначают мудрецу $A$ цвет $\varphi_{s}(A)$, тогда как стратегия $f$ требует назвать цвет $\overline{\varphi_{s}(A)}$. 3.10. Пусть мудрецы зафиксировали стратегию $f$ на графе $G^{\prime}$. Построим проигрышный расклад для этой стратегии. Для этого посмотрим на стратегию мудреца $A$. Он какой-то из двух цветов говорит реже, а именно не более, чем $h(B)-1$ раз. Выдадим ему шляпу этого цвета. Теперь, чтобы он не угадал, мы обязаны давать $B$ шляпу одного из оставшихся $h(B)$ цветов (если их осталось больше, оставим ровно $h(B)$ ). Ну и отлично, мы как раз умеем выдавать расклад на оставшемся графе, где у $B$ только $h(B)$ цветов. Это и будет опровергающий расклад на всем графе. ## Шахматы вслепую ## 4.1. W1) Тривиально. На языке шляп это игра, где какому-то из мудрецов всегда дают шляпу одного и того же цвета, конечно, он его угадает, даже не глядя на других. W2) На языке шляп в соответствующем 4-цикле шляпность двух соседних мудрецов равна двум, эти два мудреца и обеспечат выигрыш, не глядя на остальных. W3) Это утверждение - пересказ на язык игры «Шах ладьей» утверждения о том, что мудрецы выигрывают на 4 -цикле, если всем дают шляпы трех цветов. Например, стратегия мудрецов, описанная в [1], на шахматном языке выглядит следующим образом. Если шахматист видит, что король напарника стоит в центре, он ставит ладью в центр. В противном случае он ставит ладью на ту клетку, куда показывает стрелка, ведущая от короля (на вспомогательной диаграмме для этого шахматиста), см. рис. 16. Координаты клеток на рисунке соответствуют номерам цветов шляп из [1]. Так, шахматист $\mathcal{L}$, увидев, что король напарника стоит на клетке $(2,2)$, ставит свою ладью на клетку $(1,0)$ (этот случай соответствует более жирной стрелке на левом рис. 16). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-015.jpg?height=208&width=229&top_left_y=1118&top_left_x=571) Стратегия $\mathcal{L}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-015.jpg?height=211&width=217&top_left_y=1114&top_left_x=1234) Стратегия $\mathcal{R}$ Рис. 16 . W4) Пронумеруем клетки доски $L(2 \times 3)$ слева направо сверху вниз, рис. $17 a$. Пусть стратегия шахматиста $\mathcal{R}$ задается таблицей на рис. $17 b$. Здесь в клетках доски $R(3 \times 4)$ поставлено 6 меток. Метка $r_{i}$ означает, что шахматист $\mathcal{R}$, увидев, что король напарника стоит на $i$-й клетке доски $L(2 \times 3)$, поставит ладью на клетку доски $R(3 \times 4)$ с меткой $r_{i}$. Стратегию шахматиста $\mathcal{L}$ тоже зададим с помощью доски $3 \times 4$, см. рис. 17 c. Здесь в каждой клетке доски $R(3 \times 4)$ поставлено число от 1 до 6 - номер какой-то клетки доски $L(2 \times 3)$. Когда шахматист $\mathcal{L}$ видит, что король стоит на доске $R(3 \times 4)$ в клетке с меткой $k$, он поставит ладью на клетку номер $k$ доски $L(2 \times 3)$. Чтобы избежать недоразумений в обозначениях, для шахматиста $\mathcal{R}$ мы используем метки вида «буква $r$ с индексом», а для шахматиста $\mathcal{L}$ - метки вида «число». | 1 | 2 | 3 | | :--- | :--- | :--- | | 4 | 5 | 6 | a) Разметка доски $L$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-015.jpg?height=137&width=174&top_left_y=2053&top_left_x=924) b) Стратегия шахматиста $\mathcal{R}$ | 1 | 3 | 3 | 5 | | :--- | :--- | :--- | :--- | | 2 | 1 | 4 | 5 | | 2 | 6 | 6 | 4 | c) Стратегия шахматиста $\mathcal{L}$ Рис. 17. Выигрышная стратегия для игры $L(2 \times 3), R(3 \times 4)$ Опишем, каким образом задаются выигрышные стратегии с помощью введенных обозначений. Утверждение. Стратегия шахматистов выигрышная в том и только том случае, когда для любых трех различных клеток $a, b, c$ доски $L(2 \times 3)$, таких что клетки $b$ и $c$ не лежат в кресте клетки $a$, выполняется следующее свойство на доске $R$ : все клетки доски $R(3 \times 4)$, помеченные числом а, лежат в пересечении крестов $r_{b}$ и $r_{c}$. Например, при $a=1, b=5, c=6$ клетки с меткой 1 на доске $R(3 \times 4)$ лежат в пересечении крестов $r_{5}$ и $r_{6}$. (Пересечение крестов $r_{5}$ и $r_{6}$ подкрашено на рис. $17 b$.) Доказательство утверждения. Пусть судья поставил короля на клетку доски $R(3 \times 4)$, где стоит метка 1 (для других меток рассуждения аналогичны), этот случай задает нам тройку клеток $a=1, b=5, c=6$. Тогда шахматист $L$ по своей стратегии поставит ладью на клетку 1 доски $L(2 \times 3)$. Далее, пусть судья поставит короля на 5 -ю или на 6 -ю клетку доски $L(2 \times 3)-$ только в этих случаях король на доске $L$ окажется не под шахом. Значит, в этих ситуациях шах обеспечивает шахматист $\mathcal{R}$. Это значит, что клетка с меткой 1 должна лежать в кресте клетки $r_{5}$ и аналогично клетка с меткой 1 должна лежать в кресте клетки $r_{6}$. Утверждение доказано. Осталось заметить (непосредственным перебором), что в приведенном примере утверждение выполнено. W5) Стратегия задается аналогично случаю W4) (см. рис. 18). | 1 | 2 | 3 | 4 | | :--- | :--- | :--- | :--- | | 5 | 6 | 7 | 8 | a) Разметка доски $L$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-016.jpg?height=206&width=217&top_left_y=751&top_left_x=797) b) Стратегия шахматиста $\mathcal{R}$ | 3 | 5 | 4 | | :--- | :--- | :--- | | 8 | 6 | 8 | | 2 | 7 | 1 | c) Стратегия шахматиста $\mathcal{L}$ Рис. 18. Выигрышная стратегия для игры $L(2 \times 4), R(3 \times 3)$ W6) На языке шляп этот случай означает, что цикл содержит путь $P_{3}$ со шляпностями вершин $2, x, 2$, где $x \leqslant 4$. На таком пути мудрецы выигрывают. L1) Как в п. W4), пронумеруем клетки доски $L(2 \times 3)$ слева направо сверху вниз, рис. 19a. Тогда стратегия шахматиста $\mathcal{R}$ задается таблицей, наподобие той, что показана на рис. 19b. Отметим, что для этого способа задания стратегии допускается, что на одной клетке может стоять сразу несколько меток $r_{i}$. Стратегия шахматиста $\mathcal{L}$ тоже может быть задана с помощью доски $4 \times 4$, например, как это сделано на рис. 19 c. | 1 | 2 | 3 | | :--- | :--- | :--- | | 4 | 5 | 6 | a) Разметка доски $L$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-016.jpg?height=177&width=171&top_left_y=1645&top_left_x=971) b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-016.jpg?height=177&width=172&top_left_y=1645&top_left_x=1453) c) Рис. 19. Два способа задавать стратегии Как и в п. W4, имеет место следующее утверждение. Утверждение. Стратегия шахматистов выигрышная в том и только том случае, когда для любых трех различных клеток $a, b, c$ доски $L(2 \times 3)$, таких что клетки $b$ и $c$ не лежат в кресте клетки $a$, выполняется следующее свойство на доске $R$ : все клетки доски $R(4 \times 4)$, помеченные числом а, лежат в пересечении крестов $r_{b}$ и $r_{c}$. Например, для стратегии, показанной на рис. $19 c$, тройка клеток $a=1, b=5, c=6$ не удовлетворяет утверждению: из трех клеток с меткой 1 лишь одна клетка лежит в пересечении крестов $r_{5}$ и $r_{6}$. (Пересечение крестов $r_{5}$ и $r_{6}$ подкрашено на рис. 19b.) Докажем, что у шахматистов нет выигрышной стратегии. Пусть зафиксирована стратегия шахматиста $\mathcal{R}$. Пользуясь утверждением, попробуем понять, где на доске $R(4 \times 4)$ могут находиться клетки с метками 1,2 , и 3 . По утверждению клетки с меткой 1 лежат в пересечении крестов $r_{5}$ и $r_{6}$, клетки с меткой 2 - в пересечении крестов $r_{4}$ и $r_{6}$, а клетки с меткой $3-$ в пересечении $r_{4}$ и $r_{5}$. Заметим, что объединение попарных пересечений любых трех крестов (возможно, совпадающих) на доске $R(4 \times 4)$ содержит не более 8 клеток. Действительно, разберем случаи. 1. Если центры крестов лежат в разных вертикалях и горизонталях, то каждое попарное пересечение состоит из двух клеток - см. пример на рис. $19 b$, где закрашено пересечение крестов $r_{5}$ и $r_{6}$ - итого не более 6 клеток. 2. Если центры никаких двух крестов не совпадают и при этом два центра лежат на одной горизонтали или вертикали (как например $r_{4}$ и $r_{5}$ на рис. $19 b$, то пересечение этих двух крестов содержит 4 клетки и добавление третьего креста может дать в объединение попарных пересечений еще 4 клетки, только если центр этого креста лежит на одной линии с одним из первых двух центров (как $r_{4}$ и $r_{6}$ на рис. $19 b$. В этом случае получается 8 клеток, при этом 7 из них лежат в одном кресте (в рассматриваемом примере - в кресте $r_{4}$ ). 3. Если центры двух крестов совпадают, то пересечение крестов содержит 7 клеток. При любом положении третьего центра множество попарных пересечений не увеличится. Итак, для клеток с метками $1,2,3$ на доске $R(4 \times 4)$ имеется не более 8 позиций, аналогично для клеток с метками $4,5,6$ тоже не более 8 позиций. Поскольку доска $R(4 \times 4)$ содержит 16 клеток, мы имеем 8 позиций для меток $1,2,3$ и 8 позиций для меток $4,5,6$. Но как установлено в переборе, 8 позиций реализуются только в виде множества «целый крест» плюс одна клетка. Осталось заметить, что двумя крестами и двумя дополнительными клетками невозможно накрыть всю доску $R(4 \times 4)$. L2) Как и в п. L1 доска $L(2 \times 3)$ здесь та же самая, а правая доска тоже «довольно большая». Аналогично убеждаемся, что объединение попарных пересечений любых трех крестов (возможно, совпадающих) на доске $R(3 \times 5)$ содержит не более 8 клеток; случаи, в которых это пересечение содержит 7 или 8 клеток, показаны на рис. 20 , - это случаи, когда центры двух крестов лежат в одной строке или столбце (в том числе, когда они лежат в одной клетке). Во всех вариантах объединение попарных пересечений трех крестов занимает одну горизонталь доски полностью, а в каждой из двух других горизонталей оно занимает меньше половины клеток. Это значит, что объединение двух таких множеств не может составлять всю доску. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-017.jpg?height=229&width=368&top_left_y=1573&top_left_x=113) 8 клеток ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-017.jpg?height=229&width=368&top_left_y=1573&top_left_x=844) 8 клеток ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-017.jpg?height=225&width=362&top_left_y=1575&top_left_x=1575) 7 клеток $r_{b}$ - в любом месте Рис. 20. Объединение попарных пересечений трех крестов на доске $R(3 \times 5)$ L3) Это рассуждение сообщил нам Олег Чемокос. Зафиксируем какие-нибудь стратегии шахматистов $\mathcal{L}$ и $\mathcal{R}$ и проверим, что найдутся положения королей, в которых оба короля избегают шаха. Заметим, что если король стоит в какой-либо клетке доски $L(2 \times 4)$, то на доске имеется три «слабые» клетки (в другой строке), - клетки, с которых ладья не может побить этого короля. Для выигрышной стратегии все положения короля на доске $R(3 \times 4)$, вызывающие ответный ход шахматиста $L$ в слабую клетку, должны находиться в одном кресте. При этом в роли слабых клеток могут оказаться любые три клетки одной строки - это произойдет, если король находится во второй строке в столбце, содержащем четвертую клетку. Стратегия шахматиста $\mathcal{L}$ - это отображение $\phi: R(3 \times 4) \rightarrow L(2 \times 4)$, которое определяет, на какую клетку $\phi\left(K_{R}\right)$ шахматист $\mathcal{L}$ поставит ладью, если увидит короля на клетке $K_{R}$ доски $R(3 \times 4)$. Покрасим клетки доски $R(3 \times 4)$ в два цвета: клетку $x$, для которой клетка $\phi(x)$ находится в первой строке доски $L(2 \times 4)$ красим в белый цвет, остальные в черный. Не умаляя общности можем считать, что клеток белого цвета на доске не меньше, чем черного. Разберем два случая, охватывающих все возможности, которыми может быть реализовано это «не меньше». 1. Одна из строк доски $R(3 \times 4)$ содержит три белые клетки $r_{1}, r_{2}, r_{3}$ (будем называть эту строку первой) и еще одна строка (вторая) - две белые клетки $r_{4}$ и $r_{5}$. Тогда в первой строке доски $L(2 \times 4)$ можно выбрать клетку $\ell$, для которой прообраз $\phi^{-1}(\ell)$ - либо пустое множество, либо одна клетка первой строки, скажем, $r_{1}$. В этом случае остальные клетки первой строки являются слабым множеством, а его прообраз содержит клетки $r_{2}, r_{3}, r_{4}$ и $r_{5}$, не лежащие в одном кресте. 2. Каждая строка доски $R(3 \times 4)$ содержит две белые клетки. Тогда выберем в первой строке доски $L(2 \times 4)$ клетку $\ell$, для которой прообраз $\phi^{-1}(\ell)$ состоит не более чем из одной клетки (для определенности она находится в третьей строке). В этом случае остальные клетки первой строки доски $L(2 \times 4)$ являются слабым множеством, а их четыре прообраза, лежащие в первых двух строках доски $R(3 \times 4)$, образуют множество, не лежащее в одном кресте. L4) Допустим, что у шахматистов существует выигрышная стратегия. Воспользуемся обозначениями для описания выигрышных стратегий из п. W4). Пронумеруем клетки доски $L(3 \times 3)$ числами от 1 до 9 . Тогда стратегия шахматиста $\mathcal{R}$ задается выставлением на доску $R(3 \times 4)$ девяти символов $-r_{1}, r_{2}, \ldots, r_{9}$. При этом стратегия шахматиста $\mathcal{L}$ задается выписыванием в каждой клетке доски $R(3 \times 4)$ чисел от 1 до 9 , которые мы будем называть метками. Как и в предыдущем пункте, при выставлении короля в клетку $i$ на доске $L(3 \times 3)$ имеется 4 клетки, с которых ладья не может бить этого короля. Эти клетки и их номера будем называть $i$-слабыми. Для выигрышности стратегии необходимо, чтобы при всех $i$ метки на доске $R(3 \times 4)$, совпадающие с $i$-слабыми номерами, находились в кресте с центром в $r_{i}$. Заметим, что символы $r_{1}, r_{5}$ и $r_{9}$ должны находиться в разных строках доски $R(3 \times 4)$. Действительно, нетрудно видеть, что каждая клетка доски $L(3 \times 3)$ находится в слабой позиции по отношению к одной из клеток с номерами 1,5 или 9. (Например, 1 и 2 в слабой позиции к 9,3 - в слабой позиции к 5 и т. д.) Следовательно, каждая метка на доске $R(3 \times 4)$ лежит в $r_{1}$, $r_{5}-$ или $r_{9}$-кресте. Это может быть только если символы $r_{1}, r_{5}$ и $r_{9}$ находятся в разных строках. Аналогично, символы $r_{i}, r_{j}$ и $r_{k}$ находятся в разных строках, если клетки $i, j, k$ занимают три разные строки и три столбца доски $L(3 \times 3)$. Следствие. Для расположения символов $r_{1}, r_{2}, \ldots, r_{9}$ на доске $R(3 \times 4)$ возможны два случая: 1) либо символы $r_{1}, r_{2}, r_{3}$ стоят в одной строке доски $R(3 \times 4)$, символы $r_{4}, r_{5}, r_{6}$ стоят в другой строке, а символы $r_{7}, r_{8}, r_{9}$ - в третьей; 2) либо символы $r_{1}, r_{4}, r_{7}$ стоят в одной строке доски $R(3 \times 4)$, символы $r_{2}, r_{5}, r_{8}$ стоят в другой строке, а символы $r_{3}, r_{6}, r_{9}$ - в третьей. Следствие доказывается умеренно противным перебором. Докажем теперь, что выигрышных стратегий с таким обширным набором свойств не существует. Поставим ладей на все клетки $r_{i}$ доски $R(3 \times 4)$ (в клетку ставим столько ладей, сколько в ней символов $r_{i}$ ). По следствию в первой строке доски $R(3 \times 4)$ имеется «пустая» клетка, т. е. клетка, не содержащая ни одного символа $r_{i}$, но содержащая некоторую метку $a$. Пусть она для определенности находится в четвертом столбце (рис. 21). По утверждению на эту клетку направлено 4 ладейных удара, причем две из этих четырех ладей находятся в одной строке, и еще две - в другой. Это значит, что в одной из клеток четвертого столбца заведомо стоят две ладьи. Пусть для определенности это клетка находится во второй строке. Теперь мы знаем, что всего во второй строке поставлено 3 ладьи, причем две из них находятся в одной клетке. Значит, во второй строке есть две «пустые» клетки. Выберем ту из них, над которой в первой строке стоит не более одной ладьи. Пусть эта клетка для определенности находится в первом ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-018.jpg?height=206&width=266&top_left_y=2487&top_left_x=358) Рис. 21. Стратегия для случая L3 | 1 | 2 | 3 | 4 | 5 | | :---: | :---: | :---: | :---: | :---: | | 6 | 7 | 8 | 9 | 10 | a) Разметка доски $L$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-018.jpg?height=208&width=208&top_left_y=2420&top_left_x=1615) b) Стратегия шахматиста $\mathcal{L}$ Рис. 22. Ищем стратегию для игры $L(2 \times 5), R(3 \times 3)$ столбце и содержит метку $b$. На выбранную клетку тоже направлено 4 ладейных удара от двух пар ладей, стоящих в двух строках. Одна пара ладей расположена, очевидно, во второй же строке, а другая пара - в третьей строке (в первой строке над клеткой $b$ не более одной ладьи). Теперь мы видим, что одна из клеток третьей строки - во втором или третьем столбце - не может собрать на себя 4 ладейных удара из двух разных строк. Противоречие. L5) Допустим, что у шахматистов существует выигрышная стратегия. Воспользуемся обозначениями для описания выигрышных стратегий из п. W4). Пронумеруем клетки доски $L(2 \times 2)$ числами от 1 до 4 . Тогда стратегия шахматиста $\mathcal{R}$ задается выставлением на доску $R(5 \times 5)$ четырех символов $-r_{1}, r_{2}, r_{3}, r_{4}$. На доске $R(5 \times 5)$ найдется хотя бы одна клетка, назовем ее $Q$, не лежащая ни в одном из четырех крестов, определяемых этими символами. Стратегия шахматиста $\mathcal{L}$ задается выписыванием в каждой клетке доски $R(5 \times 5)$ числа от 1 до 4 . Рассмотрим число, написанное в клетке $Q$, не умаляя общности это 1. Рассмотрим число на доске $L(2 \times 2)$, стоящее по диагонали от 1 , не умаляя общности, это 4 . Пусть судья поставит королей: на доске $R(5 \times 5)$ в клетку $Q$, а на доске $L(2 \times 2)$ - в клетку 4 . Тогда игрок $\mathcal{L}$ поставит ладью на клетку 1 доски $L(2 \times 2)$, а игрок $\mathcal{R}$ поставит ладью на клетку $r_{4}$ доски $R(5 \times 5)$. Ни одна из ладей не ставит шах. Шахматисты проиграли. L6) Разметим доску $L$, как на рис. 22 a). Как и W4), стратегия шахматиста $\mathcal{L}$ задается выписыванием в каждую клетку доски $R(3 \times 3)$ числа от 1 до 10 - номера клетки на доске $L(2 \times 5)$. Поскольку на доске $L(2 \times 5)$ всего две горизонтали, найдется две строки доски $R(3 \times 3)$, в каждой из которых написаны номера двух клеток, таких что все эти четыре клетки (возможно, среди них есть совпадающие) лежат в одной горизонтали доски $L(2 \times 5)$. Пусть $j$ - номер клетки из второй горизонтали, являющийся $i$-слабым сразу по отношению ко всем этим клеткам. Например, пусть на доске $R(3 \times 3)$ стоят метки $1,2,3,4$, как на рис. $22 \mathrm{~b})$. Тогда $1-, 2$-, 3 и 4 -слабым одновременно является номер 10. Это значит, то ладья на клетке $r_{10}$ доски $R(3 \times 3)$ держит под боем клетки с метками $1,2,3$ и 4 . Это невозможно: чтобы бить метки 1 и 2 , она должна находиться в верхней строке доски $R(3 \times 3)$, а чтобы бить 3 и 4 - в нижней. По той же причине невозможен и общий случай: клетка $r_{j}$ должна оказаться сразу в двух строках $R(3 \times 3)$. 4.2. Покрасим клетки обеих досок, как показано на рис. 23, а). Пусть оба шахматиста выставляют своих ферзей только на клетки, помеченные ферзями, причем пусть первый шахматист действует из предположения «Короли стоят на клетках одинакового цвета», а второй - из предположения «Короли стоят на клетках разного цвета». Впрочем, на столь небольшой доске работает несколько неожиданный эффект: ферзь, стоящий на клетке $c 2$, держит под боем все клетки того же цвета в шахматной раскраске! Поэтому вместо «экзотической» раскраски можно использовать обычную шахматную, рис. 23, b). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-019.jpg?height=177&width=209&top_left_y=2082&top_left_x=775) a) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-019.jpg?height=180&width=211&top_left_y=2077&top_left_x=1088) b) Рис. 23. «Шах ферзем» на досках $4 \times 5$ 4.3. Утверждение задачи было найдено с помощью компьютера. Участники конференции предложили симпатичные логические стратегии. Решение 1 (Кононенко Николай). Зададим стратегию шахматистов. Разметим доску $R(5 \times 5)$ как показано на рис. 24 a). Увидев короля на клетке с меткой $j$, шахматист $\mathcal{L}$ ставит ладью на клетку доски $L(4 \times 4)$, помеченную числом $j$, рис. $24 \mathrm{~b}$ ). Таким образом, шахматист $\mathcal{L}$ использует всего четыре позиции для своего ферзя. Для каждой клетки доски $L(4 \times 4)$ на рис. 24 с) показано, с каких позиций ферзь шахматиста $\mathcal{L}$ не бьет эту клетку. Например, числа 1 и 2 в левом нижнем углу означают, что левая нижняя угловая клетка доски $L(4 \times 4)$ не пробивается ферзем с 1-й и со 2 -й позиций, показанных на рис. 24 b), а прочерк означает, что клетка пробивается со всех позиций. | 3 | 1 | 3 | 1 | 3 | | :--- | :--- | :--- | :--- | :--- | | 1 | 3 | 3 | 3 | 4 | | 3 | 3 | 3 | 3 | 3 | | 2 | 3 | 3 | 3 | 4 | | 3 | 2 | 3 | 2 | 3 | | $a$ | $b$ | $c$ | $d$ | $e$ | a) Стратегия шахматиста $\mathcal{L}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-020.jpg?height=231&width=237&top_left_y=433&top_left_x=949) b) Куда ставим ферзя | 3 | 2,4 | 4 | 1,2 | | :---: | :---: | :---: | :---: | | 2 | 4 | - | 4 | | 3 | - | 4 | 1,4 | | 1,2 | 3 | 1 | 3 | c) Инструкция для $\mathcal{R}$ Рис. 24. Увидев короля на доске $L(4 \times 4)$, шахматист $\mathcal{R}$ с помощью рис. 24 с) сразу понимает, с каких «невыгодных» позиций ферзь его напарника не может поставить шах. Поэтому он должен разместить своего ферзя на доске $R(5 \times 5)$ так, чтобы тот пробивал все клетки, отправляющие ферзя шахматиста $\mathcal{L}$ на невыгодную позицию. Для невыгодных позиций 1,2 можно поставить ферзя на клетку $b 3$, для 1,4 - на клетку $c 4$, для 2,4 - на клетку $c 2$, для 3 - на клетку $c 3$. Решение 2 (Преснова Екатерина, Рацеева Ольга). Разметим доску $L(4 \times 4)$ как показано на рис. 25 а). Наблюдение 1: для этой разметки выполняется свойство - для любой пары меток все клетки, помеченные этой парой меток пробиваются одним ферзем. Например, все клетки, помеченные 0 или 1 , пробиваются с клетки $a 4$; все клетки, помеченные 0 или 2 , пробиваются с клетки $c 2$ и т. д. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-020.jpg?height=55&width=1759&top_left_y=1406&top_left_x=180) ки $R(5 \times 5)$, рис. $25 \mathrm{~b})$. Наблюдение 2 : любая клетка доски $R(5 \times 5)$ пробивается как минимум с двух разных мест из этого набора. Тогда шахматист $\mathcal{L}$, увидев короля на доске $R(5 \times 5)$ сразу понимает, с каких позиций (среди указанных четырех) ферзь игрока $\mathcal{R}$ не может объявить шах. В силу наблюдения 2 таких «невыгодных» позиций не более двух, и шахматист $\mathcal{L}$ должен поставить своего ферзя на доске $L(4 \times 4)$ так, чтобы тот пробивал все клетки, отправляющие ферзя шахматиста $\mathcal{L}$ на невыгодную позицию. Это возможно, в силу наблюдения 1. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-020.jpg?height=248&width=237&top_left_y=1846&top_left_x=316) a) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-020.jpg?height=308&width=303&top_left_y=1788&top_left_x=728) b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-020.jpg?height=249&width=368&top_left_y=1897&top_left_x=1392) Рис. 26 . Рис. 25. 4.4. Решение 1 (Костина Екатерина, Миргалимова Розалина, Хамикова Марина). Приведем доказательство более сильного факта - что в игре «Шах ферзем» на досках $L(4 \times 6), R(7 \times 7)$ мудрецы проигрывают. Пусть судья планирует поставить короля на одну из клеток $A, B, C$ доски $L(4 \times 6)$ (рис. 26). Для каждого из этих положений игрок $\mathcal{R}$ готов поставить ферзя на доску $R(7 \times 7)$. Поскольку три ферзя не могут держать под боем все клетки доски $7 \times 7$, на доске $R(7 \times 7)$ найдется клетка, которую не бьет ни один из ферзей. Если судья поставит короля на эту клетку, игрок $\mathcal{L}$ поставит в ответ ферзя на доску $L(4 \times 6)$. Так как клетки $A, B, C$ не могут быть побиты одним ферзем, хотя бы одна из них окажется не под боем, и судья, окончательно сформулирует свои планы по поводу доски $B$ : он поставит короля именно на эту клетку. Шахматисты проиграли. Решение 2 (обычные вероятностные соображения). Но поданы они так, что мозг не сразу готов их воспринимать. Стратегия шахматиста $\mathcal{L}$ - это отображение $\phi: R(7 \times 7) \rightarrow L(7 \times 8)$, которое определяет, на какую клетку $\phi\left(K_{R}\right)$ шахматист $\mathcal{L}$ поставит ферзя, если увидит короля на клетке $K_{R}$ доски $R(7 \times$ 7). Зафиксируем какую-нибудь стратегию шахматиста $\mathcal{L}$ и проверим, что для любой стратегии шахматиста $\mathcal{R}$, найдутся положения королей, для которых они не окажутся под шахом. Для каждого положения $K_{L}$ короля на доске $L$ вычислим сумму $$ S\left(K_{L}\right)=\sum_{\substack{Q_{L}: \\ Q_{L} \notin C\left(K_{L}\right)}}\left|\phi^{-1}\left(Q_{L}\right)\right| $$ Здесь индекс суммирования $Q_{L}$ пробегает клетки доски $L$, находясь на которых ферзь шахматиста $\mathcal{L}$ не дает шах королю, стоящему на клетке $K_{L}$. Таким образом, сумма подсчитывает количество возможных положений короля на доске $R$, для которых игрок $\mathcal{L}$, пользуясь фиксированной стратегией $\phi$, не сможет сделать шах. Теперь просуммируем все эти величины по всем клеткам доски $L$ : $$ S=\sum_{\substack{K_{L}: \\ K_{L} \in L(7,8)}} S\left(K_{L}\right)=\sum_{\substack{K_{L}: \\ K_{L} \in L(7,8)}} \sum_{\substack{Q_{L}: \\ Q_{L} \notin C\left(K_{L}\right)}}\left|\phi^{-1}\left(Q_{L}\right)\right| $$ Величина $S$ равна числу способов поставить двух королей на доски так, чтобы при этом шахматист $\mathcal{L}$ не поставил шах своему королю. Отметим, что индекс $K_{L}$ в первой сумме пробегает 56 значений - все клетки доски $L$. Оценим сумму $S$ другим способом: поставим короля на любую из клеток $K_{R}$ доски $R(7 \times 7)$ (49 вариантов), тогда шахматист $\mathcal{L}$ поставит ферзя на клетку $\phi\left(K_{R}\right)$. На доске $L(7 \times 8)$ максимальное число клеток, которое может бить ферзь, равно 26 , поэтому для любой клетки $\phi\left(K_{R}\right)$ найдется не меньше $56-26=30$ клеток $K_{L}$, находящихся вне ее креста. Таким образом, $S \geqslant 49 \cdot 30$. Возвращаясь к формуле (2), мы можем сделать вывод, что существует клетка $K_{L}$, для которой $S\left(K_{L}\right) \geqslant \frac{49.30}{56}=26,25$. Иными словами, если на доске $L$ король поставлен на клетку $K_{L}$, то для этой клетки найдется 27 положений короля на доске $R$, для которого игрок $\mathcal{L}$ не ставит шах своему королю. Во всех этих случаях шах должен поставить игрок $\mathcal{R}$, причем его ход однозначно задается королем на клетке $K_{L}$. Но любой ферзь на доске $R(7 \times 7)$ бьет не более 25 клеток, значит, найдется положение короля на доске $R$ не под шахом. Таким образом, всегда можно выбрать позиции королей на обеих досках, для которых оба шахматиста не поставят шах. 4.5. Надо бы это доказать 4.6. Надо бы это доказать 4.7. Ответ: да. На доске $11 \times 11$ можно поставить 5 ферзей, которые держат под боем все клетки (например, $b 4, d 10, f 6, h 2$ и $j 8$ ). Тогда скомбинируем идею решения задачи 4.2 и стандартную идею игры в шляпы пяти цветов на 5 -клике. 5.1. Пусть мудрецы зафиксировали какую-то стратегию на новом графе. Стратегию мудреца $A$ можно задать в виде таблицы $3 \times 7$ : строки соответствуют цветам шляп мудреца $B$, столбцы цветам шляп мудреца $A$. В клетке таблицы записывают номер цвета ( 0 или 1 ), который называет мудрец $A$, когда видит у $B$ и $C$ соответствующие цвета шляп. В каждом столбце таблицы один из символов - 0 или 1 - встречается два раза. Отметим клетки, содержащие повторяющийся символ. (Если символ встретился во всех трех клетках столбца, отметим любые две из них.) Отмеченные клетки могут находится либо в первой и второй строках, либо во второй и третьей, либо в первой и третьей. Поскольку столбцов 7, по принципу Дирихле найдутся две строки, в которых отмеченные клетки занимают три столбца. В отмеченных клетках одного столбца может находиться два нуля или две единицы, следовательно, можно выбрать два столбца из трех, так, что в указанных столбцах в отмеченных клетках стоит одно и то же число. Итак, мы выбрали в таблице две строки (для определенности $i$-ю и $j$-ю) и два столбца (для определенности $k$-й и $\ell$-й), на пересечении которых стоит одно и то же число, для определенности 0 . Теперь мы без проблем построим опровергающий расклад шляп на все графе. Дадим мудрецу $A$ шляпу цвета 1 , мудрецу $B$ будем подбирать шляпу $i$-го и $j$-го цвета, а мудрецу $C-$ $k$-го или $\ell$-го. При таком подходе к делу мудрец $A$ заведомо не угадает свой цвет, так как в соответствии с таблицей назовет цвет 0 . Что же касается назначения конкретных цветов мудрецам $B$ и $C$, да и всем остальным, запустим игру на графе $G$ : после фиксации цвета мудреца $A$ стратегия остальных мудрецов на графе $G$ определена однозначно, принятые ограничения на цвета шляп $B$ и $C$ позволяют полагать, что их шляпности теперь равны 2 . Поскольку граф $G$ проигрышный, мы сумеем предъявить на нем опровергающий расклад шляп. 5.2. Докажем следующее чуть более сильное утверждение. Существует натуральное число $N$ такое, что на любом графе $G$, у которого степени всех вершин не превосходят 3 , а функции шляпности задается формулой $$ \widehat{a}= \begin{cases}3, & \text { если } \operatorname{deg} a=1 \\ 41, & \text { если } \operatorname{deg} a=2 \\ N, & \text { если } \operatorname{deg} a=3\end{cases} $$ мудрецы проигрывают. Доказательство. Обозначим $m=80\binom{81}{41}+1$. Покажем, что $N=80\binom{m}{41}\binom{81}{41}+1$ сгодится. Индукция по числу вершин. База (граф из двух вершин) очевидна. Переход. Рассмотрим вершину наименьшей степени, обозначим ее $A$. Случай $1 . \operatorname{deg} A=1$. По утверждению задачи 3.7 добавление вершины степени 3 не влияет на выигрышность графа, а если в проигрышном графе еще увеличить шляпность одной из вершин, то он так и останется проигрышным. Случай 2. $\operatorname{deg} A=2$. Обозначим соседей $A$ через $B$ и $C, \operatorname{deg} B \leqslant \operatorname{deg} C$. Ле м ма. Таблица $(2 k-1) \times\left(2(\ell-1)\binom{2 k-1}{k}+1\right)$ покрашена в 2 цвета. Тогда можно выбрать такие $k$ строк и $\ell$ столбцов, что все клетки на пересечении этих строк и столбцов покрашены в один и тот же цвет. Доказательство. Рассмотрим произвольный столбец, в нем $2 k-1$ клетка. Тогда по принципу Дирихле найдется $k$ клеток одного цвета. Отметим эти $k$ клеток. Сделаем этого для каждого столбца. Число способов отметить $k$ клеток из $2 k-1$ равно $\binom{2 k-1}{k}$, эти клетки могут быть одного из двух цветов. Значит, по принципу Дирихле найдутся $\ell$ столбцов таких, что отмеченные клетки находятся в одном и том же наборе строк и они одного и того же цвета. Это и требовалось. Случай 2.1. $\operatorname{deg} B=\operatorname{deg} C=2$. Покажем, что мудрецы проиграют в графе $G$, даже если $\widehat{A}=2$. Предположим, что существует выигрышная стратегия мудрецов в $G$. Стратегия мудреца $A$ - это таблица $41 \times 41$, в которой каждая клетка покрашена в один из двух цветов. Оставив от нее только 5 строк, воспользуемся леммой для $k=\ell=3$. Из таблицы можно выбрать 3 строки и 3 столбца так, что их пересечения покрашены в один цвет. Тогда судья наденет на $A$ шляпу другого цвета, а для $B$ и $C$ будет выбирать только трех цветов, соответствующих трем строкам и трем столбцам соответственно. Мы получим выигрышную стратегию для $G \backslash A, \widehat{B}=\widehat{C}=3$, что противоречит индукционному предположению. Случай 2.2. $\operatorname{deg} B=2, \operatorname{deg} C=3$. Покажем, что и здесь мудрецы проиграют в графе $G$, даже если $\widehat{A}=2$. Для $k=3, \ell=41$ таблица из леммы имеет размеры $5 \times 801$, а шляпности мудрецов $B$ и $C$ существенно больше: они равны 41 и $N$. Применяя лемму аналогично случаю 2.1 , опять получаем противоречие. Случай 2.3. $\operatorname{deg} B=\operatorname{deg} C=3$. Действуем аналогично случаю 2.1, используем лемму для $k=\ell=41$. Ей можно пользоваться, поскольку при $k=\ell=41$ таблица имеет размеры $81 \times m$, оба из которых меньше шляпности мудрецов $N$. Случай $3 . \operatorname{deg} A=3$. Пусть $B, C$ и $D$ - соседи мудреца $A$. Тогда $\operatorname{deg} B=\operatorname{deg} C=\operatorname{deg} D=3$. Зафиксируем какой-нибудь цвет мудреца $D$. Тогда стратегия мудреца $A-$ табличка $N \times N$. Оставив 81 строку и $m$ столбов, используем лемму, находим 41 строку и 41 столбец, пересечения которых покрашены в один цвет. Число способов выбрать 41 строку и 41 столбец равно $\binom{m}{41}\left({ }_{41}^{81}\right)$. Проделаем эти действия для каждого из $N$ возможных цветов мудреца $D$. Так как $N=80\binom{m}{41}\binom{81}{41}+1$, то найдется цвет мудреца $D$, для которого наборы строк, столбцов и цвета пересечений совпадают. Наденем на мудреца $A$ противоположный цвет, оставим по 41 варианту цветов мудрецам $B, C, D$, и получим противоречие с индукционным предположением. ## Список литературы [1] Задачи Санкт-Петербургской олимпиады школьников по математике 2016 года. М.: МЦНМО, 2017 . # Sages and hats Bursian O., Kokhas K., Latyshev A., Retinskiy V. ## 1 Get acquaintance: "Hats" game Let undirected graph $G$ be given, there is one sage and one chest with hats of different colors in each vertex of $G$. The sages know each other. Graph $G$, location of the sages in graph vertices and colors of hats in the chests are fixed and known to all. In particular, each sage understands, in which vertex each of his neighbours is situated. Judge conducts the following test with the sages. Judge places a hat on head of each sage from the corresponding chest. Each sage sees only the hats of sages in neighbour graph vertices, but he does not see his own hat and does not know its color. The sages are not allowed to communicate. By judge signal the sages name a color simultaneously. We say that the sages have passed the test successfully $=$ «have win», if at least one of them has named correct color of his own hat. Rules were explained to the sages before the test and they had possibility to hold a discussion, in which they have to determine a public strategy. Publicity means that everybody, including the judge, knows this strategy. The strategy of sages should be deterministic, i.e. each sage decision is determined uniquely by the hats of other sages. We call a strategy winning if for any hat placement at least one sage names correctly color of his own hat. Also we say that the sages win if they have a winning strategy and loose if they have not. Here is a pair of problems on this topic. 1.1. Each of two sages obtains a hat of white, blue or red color, placed on his head. The sages received the message before the test that the colors of their hats are different. Each of them sees the hat of the other sage but does not see his own hat. They must try to guess simultaneously colors of their hats (each of them writes a color on piece of paper). Prove that the sages can come to agreement how to act in advance so that at least one of them will guess the color correctly. 1.2. Sultan gives examination for 11 court sages (viziers). By examination rules sultan puts 10 sages on 10 pits, located around a circle, and he imprisons one more sage to the tower in the centre of the circle. On forehead of each of the first 10 sages sultan writes number 1 or 2 ; on forehead of the central sage sultan writes a number from 1 to 1024. The sage in the tower sees numbers of all the others, and they see his number (but do not see each other). All the sages must try to guess their own numbers simultaneously. Sultan explained to the sages the rules of examination in advance and gave them time for discussion before the examination starts. Can the sages act so that at least one of them certainly guesses his number? We will identify the vertex of graph $G$ and the sage in it. We mean that colors are numbered by $0,1,2,3, \ldots$, and that hats with colors from 0 to some number $h(X)-1$ lie in the chest of sage $X$, the set of hat colors in the chest of sage $X$ is denoted by $\mathrm{Col} X$. We call a hats game a pair $\langle G, h\rangle$, where $G=\langle V, E\rangle$ is a graph, $h: V \rightarrow \mathbb{N}$ is a function that shows how many hats are in the chest in each vertex. We call function $h$ «hatness». We will sometimes write $\widehat{A}$ instead of $h(A)$. If the sages win in game $\langle G, h\rangle$, we say that $\langle G, h\rangle$ is a winning graph. We call a winning graph simple, if it does not contain winning subgraph $\left\langle G^{\prime}, h^{\prime}\right\rangle$, where $G^{\prime} \varsubsetneqq G, h^{\prime}=\left.h\right|_{V\left(G^{\prime}\right)}$. 1.3. Prove that the sages win on the graph in fig. 1 . 1.4. Is winning graph in fig. 1 simple? 1.5. Prove that the sages win on the graph «path $A_{1} A_{2} \ldots A_{n} »$, where $n \geqslant 2$ and hatnesses are given in fig. 2 . 1.6. Is winning graph in fig. 2 simple? 1.7. Prove that the sages win on graph $K_{2,3}$ with given hatnesses (fig. 3). 1.8. Do the sages win on graph $K_{1,3}$ with given hatnesses (fig. 4)? 1.9. Prove that the sages win on the cycle $C_{n}, n \geqslant 4$, with given hatnesses (fig. 5 ). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-025.jpg?height=160&width=185&top_left_y=137&top_left_x=176) Figure 1. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-025.jpg?height=66&width=523&top_left_y=207&top_left_x=504) Figure 2. Path with $n$ vertices ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-025.jpg?height=162&width=188&top_left_y=136&top_left_x=1208) Figure 3. Graph $K_{2,3}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-025.jpg?height=166&width=188&top_left_y=134&top_left_x=1665) Figure 4. Graph $K_{1,3}$ 1.10. Prove that the sages win on the cycle $C_{n}, n \geqslant 4$ with given hatnesses (fig. 6). 1.11. Does there exist a simple winning graph, on which at least two sages guess their own colors correctly for any hats placement? 1.12. Prove that on complete bipartite graph $K_{99,50}$ sages loose, if hatness of each sage equals 100 . 1.13. Sages are on the graph «path $A B »$ (fig. $7, n \geqslant 4$ ). Before the discussion the judge promised to the sages that $A$ and $B$ will get hats of the same color. Do the sages win? ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-025.jpg?height=166&width=379&top_left_y=822&top_left_x=216) Figure 5. Cycle with $n$ vertices ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-025.jpg?height=163&width=389&top_left_y=821&top_left_x=845) Figure 6. Cycle with $n$ vertices ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-025.jpg?height=95&width=532&top_left_y=889&top_left_x=1393) Figure 7. Path $A B$ with $n$ vertices 1.14. Let sages play on graph $G$ with hatness function $h$. The judge promised them in advance that during the test when he will put a hat on the sage $A$ head he will whisper into his ear a true statement of the form «I has placed on your head a hat of one of two colors $c_{1}$ or $c_{2} »$. Therefore, in the discussion the sages know that during the test the judge will give to a sage $A$ a hint, but they do not know what colors he will name. So the sages determine strategies for all the sages except $A$, as usual, and sage $A$ gets the set of $\binom{h(A)}{2}$ strategies, one for each possible judge's hint. Prove that this hint does not affect the game result. ## 2 Games on cliques 2.1. Does there exist a winning graph not containing a 4-clique as subgraph, for which hatnesses of all sages equal 4 ? 2.2. Sages are in vertices of complete graph $K_{n}$, in the chest of $i$-th sage there are $a_{i}$ hats of different colors. Prove that the sages win if and only if $$ \frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}} \geqslant 1 $$ 2.3. We call a strategy of sages exact, if for each hats placement exactly one sage guesses his color correctly. Find all games for which an exact strategy exists. 2.4. Graph $G$ is obtained from complete graph on $n$ vertices by removing one edge. The hatnesses of all sages equal $n$. Do the sages win on such graph? 2.5. Graph $G$ is complete graph with 4 vertices $A, B, C, D$, in which edge $C D$ has been removed. And $h(A)=6, h(B)=6, h(C)=2, h(D)=3$. Do the sages win on such graph? 2.6. Graph $G$ is complete graph with $n$ vertices $A_{1}, A_{2}, \ldots, A_{n}$, in which edge $A_{n-1} A_{n}$ has been removed. The hatnesses of vertices equal $a_{1}, \ldots, a_{n}$. It turned out that the graph is winning and $$ \frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}-\frac{1}{a_{n-1} a_{n}}=1 $$ Prove that $a_{1} a_{2} \ldots a_{n-2}$ is divisible by $a_{n-1} a_{n}$. 2.7. Do the sages win on the graph «Medium bow» (fig. 8)? 2.8. Do the sages win on the graph «Big bow» (fig. 9)? ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-026.jpg?height=226&width=191&top_left_y=178&top_left_x=544) Figure 8. Graph «Medium bow» ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-026.jpg?height=268&width=420&top_left_y=137&top_left_x=1206) Figure 9. Graph «Big bow» ## 3 Graph operations Let $\left\langle G_{1}, h_{1}\right\rangle,\left\langle G_{2}, h_{2}\right\rangle$ be two games such that $V_{1} \cap V_{2}=\{v\}$. Let $G=G_{1}+G_{2}$ be the union of graphs $G_{1}$ and $G_{2}$, in which both vertices $v$ are united into one vertex. Let function $h: V_{1} \cup V_{2} \rightarrow \mathbb{N}$ coincide with $h_{i}$ on $V_{i} \backslash\{v\}(i=1,2)$ and $h(v)=h_{1}(v) h_{2}(v)$. Denote the game $\langle G, h\rangle$ by $G_{1} \times_{v} G_{2}$ (fig. 10). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-026.jpg?height=254&width=1196&top_left_y=770&top_left_x=430) Figure 10. Game $G_{1} \times{ }_{v} G_{2}$ 3.1. Theorem about product. If sages win on graphs $G_{1}$ and $G_{2}$, then they win also on graph $G_{1} \times{ }_{v} G_{2}$. Let $G_{1}$ and $G_{2}$ be two graphs without common vertices. By the substitution of graph $G_{2}$ to graph $G_{1}$ on the place of vertex $v$ we call the graph obtained by union of graphs $G_{1} \backslash v$ and $G_{2}$ with adding of all edges, that connect each vertex of $G_{2}$ with each neighbour of $v$. We denote the substitution by $G_{1}\left[v:=G_{2}\right]$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-026.jpg?height=186&width=1210&top_left_y=1436&top_left_x=432) Figure 11. Substitution of graph on the place of vertex 3.2. Theorem about substitution. Let sages win in games $\left\langle G_{1}, h_{1}\right\rangle$ and $\left\langle G_{2}, h_{2}\right\rangle$. Let $G$ be the graph of substitution $G_{1}\left[v:=G_{2}\right]$, where $v \in G_{1}$ is an arbitrary vertex. Then game $\langle G, h\rangle$ is winning, where $$ h(u)= \begin{cases}h_{1}(u) & u \in G_{1} \\ h_{2}(u) \cdot h_{1}(v) & u \in G_{2}\end{cases} $$ 3.3. Let $\langle G, h\rangle$ be a game, in which sages win, let $B C$ be an edge of graph $G$. Consider graph $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, obtained by adding new vertex $A$ to graph $G: V^{\prime}=V \cup\{A\}$, and two edges: $E^{\prime}=$ $E \cup\{A B, A C\}$. Then the sages win in the game $\left\langle G^{\prime}, h^{\prime}\right\rangle$ (see fig. 12), where $$ h^{\prime}(u)= \begin{cases}2, & u=A \\ 2 h(B), & u=B \\ 2 h(C), & u=C \\ h(u), & \text { for other vertices } u \in V\end{cases} $$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-027.jpg?height=270&width=754&top_left_y=128&top_left_x=648) Figure 12. Addition of vertex $A$ of hatness 2 to edge $B C$ 3.4. Let $\left\langle G_{1}, h_{1}\right\rangle,\left\langle G_{2}, h_{2}\right\rangle$ be two games, in which sages win. Let $A_{1}, A_{2}, \ldots, A_{k} \in V_{1} ; B_{1}, B_{2}, \ldots$, $B_{m} \in V_{2}$. Consider graph $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, obtained by adding all the edges $A_{i} B_{j}$ to graph $G_{1} \cup G_{2}$ : $V^{\prime}=V_{1} \cup V_{2}, E^{\prime}=E_{1} \cup E_{2} \cup\left\{A_{i} B_{j}, i=1, \ldots, k ; j=1, \ldots, m\right\}$ (fig. 13). Then the sages win in game $\left\langle G^{\prime}, h^{\prime}\right\rangle$, where $$ h^{\prime}(u)= \begin{cases}h_{1}(u), & u \in G_{1} \backslash\left\{A_{1}, A_{2}, \ldots, A_{k}\right\} \\ h_{2}(u), & u \in G_{2} \backslash\left\{B_{1}, B_{2}, \ldots, B_{m}\right\} \\ h_{1}(u)+1, & u \in\left\{A_{1}, A_{2}, \ldots, A_{k}\right\} \\ h_{2}(u)+1, & u \in\left\{B_{1}, B_{2}, \ldots, B_{m}\right\}\end{cases} $$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-027.jpg?height=246&width=968&top_left_y=982&top_left_x=538) Figure 13. Gluing of two graphs, $k=2, m=3$ 3.5. Let $\langle G, h\rangle$ be a game in which sages win, let $Z, C \in V$ be two vertices of graph $G$. Consider graph $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, obtained by adding new path $Z A B C$ to graph $G: V^{\prime}=V \cup\{A, B\}, E^{\prime}=$ $E \cup\{Z A, A B, B C\}$. Then the sages win in game $\left\langle G^{\prime}, h^{\prime}\right\rangle$, where $h^{\prime}(Z)=2 h(Z), h^{\prime}(C)=h(C)+1$, $h^{\prime}(A)=2$ and $h^{\prime}(B)=3$ and $h^{\prime}(u)=h(u)$ for other vertices $u \in V$ (fig. 14). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-027.jpg?height=248&width=742&top_left_y=1572&top_left_x=657) Figure 14. Addition of two vertices with hatnesses 2 and 3 connected by edge 3.6. Let graph $\left\langle G, h^{\prime}\right\rangle$ be winning, $V(G)=\left\{A_{1}, \ldots, A_{n}\right\}$. Take $n$ winning graphs $\left\langle G_{i}, h_{i}\right\rangle$ and mark in each of them one vertex $A_{i}$. Construct graph $G$ on the set of marked vertices. For this «supergraph» define hatness function as $h\left(A_{i}\right)=h^{\prime}\left(A_{i}\right) h_{i}\left(A_{i}\right)$ (and for the other vertices hatness is the same as in the initial graphs). Then the sages win. 3.7. The addition to graph $G$ a pendant vertex $A$, where $\widehat{A}>2$, does not affect the result of the game. 3.8. a) Addition of two new vertices $B$ and $C$ of hatness 5 and edges $A B, A C, B C$ to graph $G$ with vertex $A$ of hatness 2 does not affect the result of the game. b) Addition of two new vertices $B$ and $C$ of hatnesses 7 and 5 and edges $A B, A C, B C$ to graph $G$ with vertex $A$ of hatness 3 does not affect the result of the game. 3.9. Let $G_{1}$ and $G_{2}$ be graphs with vertex $A$ in which sages loose if $\widehat{A}=2$. Then the sages loose on graph $G=G_{1}+_{A} G_{2}$, in which $\widehat{A}=2$. 3.10. Let $H=\langle G, h\rangle$ be a loosing game, $B$ be an arbitrary vertex of graph $G$. Consider graph $G^{\prime}=\left\langle V^{\prime}, E^{\prime}\right\rangle$, obtained by addition new pendant vertex $A$ to graph $G: V^{\prime}=V \cup\{A\}, E^{\prime}=E \cup\{A B\}$. Then the sages loose in game $\left\langle G^{\prime}, h^{\prime}\right\rangle$, where $h(A)=2, h^{\prime}(B)=2 h(B)-1$ and $h^{\prime}(u)=h(u)$ for other vertices $u \in V$. ## 4 Blind chess Game «Check by rook». Two chess players $\mathcal{L}$ and $\mathcal{R}$ are sitting opposite each other, there is a chessboard on the wall behind of each of them. Each chess player does not see his own board but sees the board of the other chess player. The judge puts one king on each of these boards. After that the chess players must to name one cell of his chessboard independently of each other and the judge puts rook on this square. If at least one of the kings is in rook check (under attack of rook) (or the rook has been put on the same cell, where king is), then the chess players win, otherwise they lose. Chessboards of the players can be different and of arbitrary sizes. As in the game «Hats», the chess players determine public deterministic strategy in advance. The judge knows this strategy and plays against the chess players. Let graph $G$ be the cycle $A B C D A$ with hatness function $h$. In fact, graph $G$ is the complete bipartite graph $K_{2,2}$, so sages $A$ and $C$ see the same, and sages $B$ and $D$ also see the same. Then a pair of players $A$ and $C$ we call chess player $\mathcal{L}$ and we will mean that his board has size $h(A) \times h(C)$, and a pair $B$ and $D$ we call chess player $\mathcal{R}$ and we will mean that his board has size $h(B) \times h(D)$. Therefore game «Hats» on the cycle $A B C D A$ with hatness function $h$ is equivalent to game «Rook check» on boards $L(h(A) \times h(C))$ and $R(h(B) \times h(D))$. Union of any vertical and any horizontal row of chessboard we call cross. Cross is uniquely defined by the cell, situated in the intersection of these rows, it is called the centre of the cross. The rook located in the centre of the cross, holds in check exactly all cells of the cross. In problem 4.1 you can submit items solutions separetaly. 4.1. In the game «Check by rook» the chess players win on the following pairs of board: W1) one of the boards has size $1 \times k$, where $k$ is an arbitrary positive integer; W2) $L(2 \times k)$ and $R(2 \times m)$, where $k$ and $m$ are arbitrary positive integers; W3) $L(3 \times 3), R(3 \times 3)$; W4) $L(2 \times 3), R(3 \times 4)$; W5) $L(2 \times 4), R(3 \times 3)$; W6) $L(2 \times 2), R(k \times m)$, where $\min (k, m) \leqslant 4$. The chess players lose on the following pairs of boards: L1) $L(2 \times 3), R(4 \times 4)$; L2) $L(2 \times 3), R(3 \times 5)$; L3) $L(2 \times 4), R(3 \times 4)$; L4) $L(3 \times 3), R(3 \times 4)$; L5) $L(2 \times 2), R(5 \times 5)$; L6) $L(2 \times 5), R(3 \times 3)$. For boards of other sizes the question if the sages win can be solved by comparing with these cases. For example, the chess players lose on the boards $L(3 \times 4), R(3 \times 4)$ because they lose even in simpler case L3. The chess players win on the boards $L(2 \times 3), R(3 \times 3)$ because they win even when one of the boards is larger (as in case W3). Check by queen. Consider variant of the game when both chess players put on the board a queen instead of a rook. We call this game «Check by queen». 4.2. In the game «Check by queen» $L(4 \times 5), R(4 \times 5)$ the chess players win. 4.3. In the game «Check by queen» $L(4 \times 4), R(5 \times 5)$ the chess players win. 4.4. In the game «Check by queen» $L(7 \times 8), R(7 \times 7)$ the chess players lose. 4.5. In the game «Check by queen» $L(3 \times 4), R(7 \times 7)$ the chess players lose. 4.6. In the game «Check by queen» $L(4 \times 5), R(5 \times 5)$ the chess players lose. 4.7. Consider variant of the game «Check by queen», in which 5 chess players are situated so that each of them sees the boards of the others but does not see his own board. All boards have size $11 \times 11$ As in the initial game, the judge puts one king on each board, and the chess players simultaneously point to the cell, where the queen has to be put. Can the chess players win? ## 5 Several more problems 5.1. Let $G$ be graph with vertices $B$ and $C$. Let hatness function is such that $\widehat{B}=\widehat{C}=2$ and the graph is losing. Add to the graph new vertex $A$, which is connected only with $B$ and $C$. Then the sages lose on the obtained graph, if $\widehat{A}=2, \widehat{B}=3, \widehat{C}=7$, and hatnesses of other vertices have not changed. 5.2. Is it true that there exists such «large» number $k$ that any graph $G$, in which degrees of all vertices do not exceed 3 , and hatnesses of all vertices equal $k$, is losing? 5.3. Is it true that there exists such «large» number $d$, that any graph $G$, in which degrees of all vertices equal $d$, and hatnesses of all vertices equal 4 , is winning? ## SOLUTIONS 1.1. They can arrange colors on cycle, for example, white $\rightarrow$ blue $\rightarrow$ red $\rightarrow$ white, and further let each of them will write next color in order of the cycle, looking at the color of fellow's hat. ### 1.2. Answer: yes. Let the number on the tower be written in binary notation, and the numbers 0 or 1 be written on the forehead of each of the sages in pits (instead of 1 and 2). Let the $k$-th sage in the pit play according the strategy «my bit does not coincide with the $k$-th bit on the tower», and the sage in the tower only compose his number from the bits in pits. 1.3. Let colors be numbers from 0 to 3 . Let the upper sage call the parity of sum of hat colors on two other sages. And the other two sages, since they see the third one and each other, can understand what two colors they must exclude, and play as if they are on the graph-edge with two colors. ### 1.4. Answer: yes. It is enough to check that after removing of one edge losing graph is obtained, i.e. that graph «path with two edges» with hatnesses $2,4,4$ or $4,2,4$ is losing. Apply probabilistic arguments: so, for the case 2, 4, 4, the fractions of the number of placements, for which the sages win, equals $1 / 2,1 / 4,1 / 4$ correspondingly, that is exactly 1 in total. But the cases, in which the sages in outermost vertices guess correctly, are independent: on $1 / 8$ part of all placements both outermost sages guess correctly. Therefore, the total part of all placements, on which the sages guess correctly, equals $\frac{1}{2}+\frac{1}{4}+\frac{1}{4}-\frac{1}{8}<1$. Another case is considered similarly. 1.5. It is well known that the sages win for $n=2$, when hats can be one of two colors: the first plays according the hypothesis «we have equal hats», and the second «we have different hats». Applying the problem 3.1 several times, we obtain the statement for the other $n$. ### 1.6. Answer: yes. It is enough to check that graph «path $P_{k} »$ with hatnesses $2,4, \ldots, 4$ is losing. Using the constructor from problem 3.7 this statement is easily reduced to the checking whether the game of two persons with hatnesses 2 and 4 is wining or not. But it can be done also by induction. Let $A$ be the sage with hatness $2, B$ be his neighbour. By the strategy sage $A$ depending on the color of $B$ (four variants) says one of two colors, say, red or blue. Let sage $A$ say blue color in two or more cases («bad» placements), and say red in the other cases («good» placements). Let the judge give red hat to sage $A$ and declares that sage $B$ will get hats of some two colors corresponding to bad placement. Then sage $A$ will certainly guess incorrectly, and sage $B$ can consider that his hatness equals 2 . The game is reduced to shorter path of the same kind. 1.7. Let $X$ and $Y$ be two vertices in the first part, and $A, B, C$ be vertices in the second part, and $\widehat{Y}=2, \widehat{X}=\widehat{A}=\widehat{B}=\widehat{C}=4$. Describe the winning strategy of the sages. Each of four colors can be interpreted as 2-bit binary number consisting of the left and the right bits. If a sage is ready to say color of defined parity, he already determined with right bit and must to choose one of two values of left bit. The color of $Y$ 's hat, one bit, we also will interpret as parity. Let sage $Y$ say the parity $y$ that is dominant among hats $A, B, C$. Each of sages $A, B, C$ sees hat of $Y$. So they understand that if the most of hats $A, B, C$ have the same parity as $Y$ 's hat, then sage $Y$ guesses correctly color of his hat. Therefore, they act assuming that the most of hats has parity $\bar{Y}$. Hence, in order to say their guesses each of them must choose a color between two possible colors with parity $\bar{Y}$. From the other hand, sage $X$ ( and sage $Y$ too) sees, what parity is dominant among hats $A, B$, $C$, and knows the value of $y$. Therefore, sage $X$ knows that if the hat of color $y$ is on sage $Y$, then $Y$ has guessed correctly. Hence sage $X$ assumes that hat of color $\bar{y}$ is on sage $Y$. But then he knows that in this case sages $A, B, C$ proceed from the assumption that most of their hats have parity $y$. Thus, sage $X$ sees what is dominant parity among colors $A, B, C$, and he knows that sages $A, B, C$ proceed from the assumption that this parity is really dominant among their hats. Let sage $X$ choose the two sages with hats of dominant parity among $A, B, C$, we call these sages main. Now we show how the «assumptions» can be transformed into a strategy. Columns of the following table correspond to colors of hat $X$, and rows correspond to sages $A, B, C$. Let each of sages $A, B$, $C$ take left bit from his own row of table for his guess (and right bit is defined by the «assumption»). | | 0 | 1 | 2 | 3 | | :--- | :--- | :--- | :--- | :--- | | $A$ | 0 | 0 | 1 | 1 | | $B$ | 0 | 1 | 0 | 1 | | $C$ | 0 | 1 | 1 | 0 | Sage $X$ sees whether somebody of main sages among $A, B, C$ has guessed correctly the color of his own hat by the given strategy or not. If two main sages have guessed incorrectly then it means that they both pointed wrong left bit of color of their own hats. Let, for example, main sages be $A$ and $B$, let they point to bits 0 and 1 correspondingly and have not guessed correctly. Sage $A$ points to bit 0 , only if the color of hat $X$ equals 0 or 1 ; sage $B$ points to bit 1 , only if the color of hat $X$ equals 1 or 3 . Therefore, both events happen only in the case when the color of hat $X$ equals 1 , and sage $X$ must say exactly this color. Sage $X$ performs similar actions in all other cases. His success is guaranteed by the following magic property of the table. If two arbitrary rows are chosen in the table and two cells in each of them (independently of the other), containing equal symbols are painted over, then exactly one column contains two painted over cells. Magic property is checked by (evident) considering of all the possible cases. Note that weaker requirement «no more than one column of the table contains two painted over cells» is enough to provide the strategy to be winning. 1.8. Suppose that the sages have a winning strategy. Let $v$ be the vertex of degree $3, u_{1}, u_{2}, u_{3}$ be pendant vertices. Will assign the first color to vertex $v$. Let sages $u_{1}, u_{2}, u_{3}$ say colors $h_{1}, h_{2}, h_{3}$ according to the strategy. Now conduct the second experiment: assign second color to vertex $v$. Let sages $u_{1}, u_{2}, u_{3}$ say colors $e_{1}, e_{2}, e_{3}$ according to the strategy. Finally, conduct the last experiment. For each $i=1,2,3$ denote by $d_{i}$ the color that has not been said by sage $u_{i}$ in the first two experiments (if we have a choice, then we take any color of two possible ones). For each $i$ assign color $d_{i}$ to pendant vertex $u_{i}$. The hat colors for neighbours of sage $v$ are already given, so, his answer by the strategy is known. Assign to vertex $v$ first or second color that does not coincide with this answer. The sages have lost. 1.9. This statement is obtained by applying the constructor from problem 3.4, where graph consisting of one vertex with hatness 1 is taken as $G_{2}$ ! Indeed, the sages win on the path drawn in fig. 2 , and the graph, described in the problem, is obtained by the addition of one-vertex graph $G_{2}$ to this path, and in the place of gluing the hatnesses increase by 1. 1.10. This statement is obtained by applying the constructor from problem 3.5, Indeed, the sages win on the path drawn in fig. 2, and the graph, described in the problem, is obtained by addition new edge with hatnesses of vertices 2 and 3 to this path and appropriate change of vertex hatnesses, to which they are fastened. 1.11. Answer: no. Let $\langle G, a\rangle$ be simple winning graph, $A \in V(G)$. Give hat of color 0 to sage $A$. Then for the other sages strategy on a losing graph $G \backslash A$ has been fixed. So there is a losing hat placement for the sages on the graph $G \backslash A$, give it to them. In the obtained hats placement on the whole graph $G$ only sage $A$ can guess correctly. 1.12. Present the judge's strategy. First consider 51 hats placements on the part $A$, consisting of 50 sages, where all sages of $A$ will obtain hats of the same color, and this color is one of the first 51 colors. Will put on each of the sages of the second part $B$ the hat, that he will say by his strategy for no one of 51 placements. Fix constructed on the part $B$ placement and look what the sages of $A$ will say by their strategy. There are 50 sages, therefore, some color from the first 51 ones nobody says. Put hats of this color on all the sages of part $A$. As a result nobody of the sages has guessed correctly. The presented reasoning works for all bipartite graphs, in the smaller part of which there are no more than $k-2$ vertices, where $k$ is hatness of the sages. If one of the parts contains $k-1$ vertices, then the sages win in the case, when the size of the second part is very large. 1.13. Answer: yes, the sages win. Let sage $A$ plays by the strategy «If I see 2 , then I say 2 , otherwise I say 0 », and sage $B$ by the strategy «If I see 2 , then I say 2 , otherwise I say $1 »$, Sages $C$ and $D$ say 2, if they see hat of color 0 or 1 on their lonely neighbour $(A$ or $B$ ). In the opposite case they assume that they have hats not of color 2 , and play the game from problem 1.5 with the other sages. By this strategy the sages really win. Because either sages $A$ и $B$ have hat of color 0 or 1 and then somebody of these four wins, or they have hat of color 2 and then sages $C$ and $D$ have exactly not 2 (otherwise $A$ or $B$ have guessed correctly), and therefore $C$ and $D$ will win on the path between them. 1.14. Suppose that the sages win with the hint. Fix strategies for all sages except $A$, that they use in game with hint, and will show, how to give the strategy of sage $A$ in order to the sages win without hint. Assume that if we assign hat of color $x$ to sage $A$ there exists the hats placement on the whole graph, in which sage $A$ has obtained color $x$, the neighbours of sage $A$ have obtained colors $u, v, w, \ldots$, the other sages have also obtained some colors, and nobody of the sages (excluding $A$ ) has guessed correctly. Then we want sage $A$ to guess his color correctly in this situation, i.e. his strategy must satisfy the requirement $f_{A}(u, v, w, \ldots)=x$. These requirements obtained for different placements, do not contradict to each other. Indeed, if there exists another placement, where neighbours still have colors $u, v, w, \ldots$, and sage $A$ obtained another color $y$, then the sages could not win with hint $A^{*}$, since, having these two placements in mind, the judge can tell sage $A$, that the color of his hat is either $x$ or $y$, and after that realize the placement for which $A$ 's guess is incorrect. ## Game on clique ### 2.1. Answer: yes. For example, «bow» (fig. 15). By the constructor from problem 3.4 one can to construct similar examples for any $n$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-031.jpg?height=131&width=214&top_left_y=1985&top_left_x=927) Figure 15. Bow 2.2. Since $i$-th sage guesses correctly on $\frac{1}{a_{i}}$-th part of all placements, if the sum is less than 1 , there exists placement, where nobody guesses correctly. Prove that if the sum is greater or equal to 1 , then the sages win. We suggest two solutions. Solution 1 (Hall's theorem). Fix $i$ (the number of some sage) and partition the set of all hats placements into subsets of $a_{i}$ elements. In each hats placement we delete color $c_{i}$ and for the remaining set $c=\left(c_{1}, \ldots, c_{i-1}, \widehat{c_{i}}, c_{i+1}, \ldots, c_{n}\right)$ (symbol "hat" means that this color is omitted) let $$ A_{c}^{i}=\left\{\left(c_{1}, \ldots, c_{i-1}, x, c_{i+1}, \ldots, c_{n}\right): x \in \operatorname{Col}\left(A_{i}\right)\right\} $$ Keeping in mind application of Hall's marriage theorem, call the sets $A_{c}^{i}$ «girls», and placements themselves call «boys». Will say that boy $s$ and girl $A_{c}^{i}$ know each other, if the hats placement $s$ is an element of the set $A_{c}^{i}$. Each boy knows $n$ girls, and for each $i$ each boy knows exactly one girl of type $A_{c}^{i}$. Each girl $A_{c}^{i}$ knows exactly $a_{i}$ boys. Prove that there exists matching sending each boy to a girl. For this it is enough to check the theorem condition that each $m$ boys know together at least $m$ girls. Consider an arbitrary set of $m$ boys. Since for each $i$ girl $A_{c}^{i}$ knows exactly $a_{i}$ boys, then for each $i \quad m$ boys know in total at least $m / a_{i}$ girls of kind $A_{c}^{i}$. Summing over $i$, obtain that the total number of girls which are familiar with these $m$ boys is not less than $\frac{m}{a_{1}}+\frac{m}{a_{2}}+\ldots+\frac{m}{a_{n}} \geqslant m$. The condition of Hall's theorem holds. Thus, there exists a matching that put into correspondence to every hat placement a set of kind $A_{c}^{i}$. Note that, when equality $\frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}=1$ holds, this matching in fact selects one element in each set $A_{c}^{i}$. Otherwise, if inequality $\frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}>1$ holds, then «there will be lonely girls», i. e. no elements are selected in some sets $A_{c}^{i}$. The constructed matching allows to define sages' strategy. Let $j$-th sage acts by the rule: looking at hats of the other sages, i.e. the set of colors $c=\left(c_{1}, \ldots, c_{j-1}, c_{j+1}, \ldots, c_{n}\right)$, he reconstructs the set $A_{c}^{j}$, which is really consists of all possible ways to supplement set $c$ to the hats placement on the whole graph. Observe that the current hats placement is one of the elements of this set. The sage must say the color that is marked in set $A_{c}^{j}$ by our matching (if there is no marked element, he says color arbitrarily). Since each hats placement is mapped by our matching to the selected element of one of sets $A_{c}^{i}$, for this hats placement $i$-th sage will guess correctly his own color. Solution 2 (explicit strategy). Let $N=\operatorname{LCM}\left(a_{1}, a_{2}, \ldots, a_{n}\right)$ and for $k$ from 1 to $n$ let $d_{k}=N / a_{k}$. We identify the set of possible hat colors of $k$-th sage and the set of remainders $d_{k}, 2 d_{k}, \ldots, a_{k} d_{k}$ modulo $N$. Let hats be given to the sages: $k$-th sage obtains hat of color $x_{k} d_{k}$, where $x_{k} \in\left\{1,2, \ldots, a_{k}\right\}$. Let $S=x_{1} d_{1}+x_{2} d_{2}+\ldots+x_{k} d_{k}(\bmod N)$. Each sage looking around can write all the summands of this sum except his own one. Making assumption about the value of the sum, he can calculate the color of his own hat. Let the first sage check hypothesis $S \in\left\{1,2, \ldots, d_{1}\right\}$; the second sage check hypothesis $S \in\left\{d_{1}+1, d_{1}+2, \ldots, d_{1}+d_{2}\right\}$ and so on, the $n$-th sage check hypothesis $S \in$ $\left\{d_{1}+d_{2}+\ldots+d_{n-1}+1, \ldots, d_{1}+d_{2}+\ldots+d_{n-1}+d_{n}\right\}$. The hypothesis of $k$-th sage concerns $d_{k}$ consecutive remainders, exactly one among them is divisible by $d_{k}$. This remainder defines the color of hat that the $k$-th sage should say. 2.3. Answer: exact strategies exist only on cliques and only when the equality $$ \frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n}}=1 $$ holds. If the graph contains two non adjacent vertices $A$ and $B$, then put arbitrary hats to all sages except $A$ and $B$. Now the answers of $A$ and $B$ are defined by the strategy. Give them hats for which their guesses are correctly. On the constructed hats placement $A, B$ and, possibly, somebody else guesses correctly. Therefore, the strategy is not exact. Hence the graph is a clique. It follows from the proof of problem 2.2 that for cliques the existence of an exact strategy is equivalent to equality (1). ### 2.4. Answer: no, the sages lose. A pair of sages $A, B$ can be interpreted as a chess player with $n \times n$ board. It is easy to see that the fraction of total number of placements, for which he wins, equals $\frac{2}{n}-\frac{1}{n^{2}}$. As for the other sages, each of them wins on $\frac{1}{n}$-th part of all placements. Therefore, the total fraction of all placements, on which somebody wins, is not less than 1 . 2.5. Answer: yes, the sages win. We interpret hat colors of sages $A$ and $B$ as residues modulo 6 , a color $C$ as a residue modulo 2 , a color $D$ as a residue modulo 3. Denote hat colors of sages $A$ and $B$ by $a$ and $b$. Let sage $C$ say color $c=(a+b) \bmod 2$, sage $D$ say color $d=(a+b)$ mod 3. If sages $C$ and $D$ have not guessed correctly, the equality $a+b=c+1 \bmod 2$ holds and also one of the following equalities holds: $a+b=d+1 \bmod 3$ or $a+b=d+2 \bmod 3$. Then let $A$ compute his own color assuming that $a+b=c+1 \bmod 2$ and $a+b=d+1 \bmod 3$; and $B$ compute assuming that $a+b=c+1 \bmod 2$ and $a+b=d+2 \bmod 3$. We offer to the reader obtain the same result as an exercise using the constructor from problem 3.3 . 2.6. Let $X$ be the set of hats placements for the first $n-2$ sages, i.e., in other words, that is the collection of sets of $n-2$ colors, where the first color is a possible hat color of sage $A_{1}$, the second color is a possible hat color of sage $A_{2}$ and so on, the $(n-2)$-th color is a possible hat color of sage $A_{n-2}$. Let $\alpha=a_{1} a_{2} \ldots a_{n-2}$, then $|X|=\alpha$. Denote by $L_{i}\left(i=1,2, \ldots, a_{n-1}\right)$ the subsets of $X$, such that if sage $A_{n-1}$ sees on his neighbours the set of colors from $L_{i}$, then he says color $i$. Similarly define sets $R_{j}\left(j=1,2, \ldots, a_{n}\right)$ for sage $A_{n}$. Let $L_{k}$ be the set $L_{i}$ of minimum cardinality, $\left|L_{k}\right|=M \leqslant \frac{\alpha}{a_{n-1}}$. Now consider sets $R_{j} \backslash L_{k}\left(j=1,2, \ldots, a_{n}\right)$. In these sets there are $\alpha-M$ elements in total, so if $R_{m} \backslash L_{k}$ is the set of minimum cardinality, then $\left|R_{m} \backslash L_{k}\right| \leqslant \frac{\alpha-M}{a_{n}}$. Therefore, $$ \begin{aligned} \left|L_{k} \cup R_{m}\right| \leqslant M+\frac{\alpha-M}{a_{n}} & =\frac{\alpha}{a_{n}}+M\left(1-\frac{1}{a_{n}}\right) \leqslant \frac{\alpha}{a_{n}}+\frac{\alpha}{a_{n-1}}\left(1-\frac{1}{a_{n}}\right)= \\ & =\alpha\left(\frac{1}{a_{n-1}}+\frac{1}{a_{n}}-\frac{1}{a_{n-1} a_{n}}\right)=\alpha\left(1-\sum_{i=1}^{n-2} \frac{1}{a_{i}}\right)=\alpha-\frac{\alpha}{a_{1}}-\ldots-\frac{\alpha}{a_{n-2}} \end{aligned} $$ Thus, when sage $A_{n-1}$ has hat of color $k$, and sage $A_{n}$ has hat of color $m$, somebody of sages $A_{1}, A_{2}$, $\ldots A_{n-2}$ will win, and the number of placements, for which this event happens, equals the fraction $\frac{1}{a_{1}}+\frac{1}{a_{2}}+\ldots+\frac{1}{a_{n-2}}$ of total number of placements. But as we know, the indicated fraction bounds from above the number of placements, for which sages $A_{1}, A_{2}, \ldots A_{n-2}$ win. Therefore, both inequalities (1) must be equalities. Then $\left|L_{k}\right|=\frac{\alpha}{a_{n-1}}$ (and in general $\left|L_{i}\right|=\frac{\alpha}{a_{n-1}}$ for all $i$ ), and $\left|R_{m} \backslash L_{k}\right|=\frac{\alpha}{a_{n}}-\frac{\alpha}{a_{n-1} a_{n}}$. Analogously $\left|R_{j}\right|=\frac{\alpha}{a_{n}}$. Therefore $\left|R_{m} \cap L_{k}\right|=\frac{\alpha}{a_{n-1} a_{n}}$, and $\alpha$ is divisible by $a_{n-1} a_{n}$. 2.7. Answer: yes, the sages win. Let $G_{1}$ and $G_{2}$ be 4 -cliques in the bow, $A$ be their common vertex, colors of the sages be the remainders modulo 5 . Let the sages on $G_{1} \backslash A$ compute the sum of remainders-colors on $G_{1}$ and play according hypotheses that the sum has reminder, say, for 2,3 and 4 modulo 5 . The sages on $G_{2} \backslash A$ play similarly. What sage $A$ have do? He should play checking two hypotheses on $G_{1}$ and two hypotheses on $G_{2}$ at the same moment. The only that rescues him is that both hypotheses are about his unhappy hat! The strategy will work if we could convert colors of $A$ 's hat to remainders modulo 5 in different ways: | Color | Red | Blue | White | Yellow | Black | | :---: | :---: | :---: | :---: | :---: | :---: | | His code in $G_{1}$ | 0 | 1 | 2 | 3 | 4 | | His code in $G_{2}$ | 0 | 2 | 4 | 1 | 3 | The presented coding has the property that any pair consisting of successive remainders in row $G_{1}$, intersects with any pair of successive remainders in row $G_{2}$ in no more than one element. Sage $A$ computes two successive remainders corresponding to the hypotheses "the sum of all remainders on $G_{1}$ equals 0 or 1 » and «the sum of all remainders on $G_{2}$ equals 0 or 1 ", and says color of the intersection of these pairs, if it exists. 2.8. Answer: yes, the sages win here too. Consider residues modulo $740=4 \cdot 5 \cdot 37$. Let hat colors of sage with hatness $k$ be the residues divisible by $\frac{740}{k}$. Denote by $S_{1}$ and $S_{2}$ the sums of residues-colors in the left and the right 5 -cliques. Let the left sage of hatness 4 suppose that $S_{1}$ belongs to the set $\{1,2, \ldots, 185\}$. He obtains 185 consecutive residues, exactly one of them is divisible by 185 . The sage says the color, corresponding to this residue. Similarly three left sages with hatness 5 suppose that $S_{1}$ belongs to sets $\{186, \ldots 333\},\{334, \ldots 481\}$, $\{482, \ldots 629\}$ correspondingly. The right sages acts similarly, but working with $S_{2}$. If now nobody has won, then $A$ (the sage with hatness 37 ) understands that $S_{1}$ belongs to $\{630, \ldots 720\}$, so he needs to choose from at most 6 consecutive colors. The same can be said about $S_{2}$. Let the sages of left and right cliques convert the colors of sage $A$ to the residues by different ways. By our rule the hat colors of sage $A$ are the residues modulo 740 , divisible by 20 , i.e. really this is the residues modulo 37 . If the sages of left clique convert a color to residue $x$, then the sages of right clique convert the same color to residue $6 x(\bmod 37)$ (the map $x \mapsto 6 x$ is 1-to-1 on the sets of residues modulo 37). As it is easy to see, any two sets of the form $\{x, x+1, \ldots, x+5\}$ and $\{6 y, 6 y+6, \ldots, 6 y+30\}$ intersect in at most one one element. Then $A$ calls the intersection color of these sets (or calls an arbitrary color if the intersection is empty). ## Constructors 3.1. Hat color of sage $v$ in the graph $G_{1} \times{ }_{v} G_{2}$ can be interpreted as a pair of colors $\left(c_{1}, c_{2}\right)$, where $c_{i}$ is the hat color of $v$ in graph $G_{i}$. Fix winning strategies for graphs $G_{1}$ and $G_{2}$. Construct strategy on $G_{1} \times{ }_{v} G_{2}$ : let all sages except $v$, play by the winning strategy for the corresponding graph (the neighbours of $v$ in $G_{i}$ look only at the component $c_{i}$ of $v$ 's composite color). As to sage $v$, he plays by both strategies giving two answers $c_{1}$ and $c_{2}$ independently; answer $c_{i}$ corresponds to his winning strategy for graph $G_{i}$ (for computing this answer sage $v$ looks only on his neighbours from graph $G_{i}$ ). Pair $\left(c_{1}, c_{2}\right)$ is concrete hat color of sage $v$ on the graph $G_{1} \times{ }_{v} G_{2}$, that is the answer of sage $v$ in the constructing strategy. The constructed strategy is winning because either somebody from $G_{1}$ or from $G_{2}$ will guess correctly his own color, or $v$ will guess correctly both components of his own color. 3.2. Let $f_{1}$ and $f_{2}$ be winning strategies from games on graphs $G_{1}$ and $G_{2}$ correspondingly. Let also each sage $u$ of subgraph $G_{2}$ of graph $G$ obtain composite color $\left(c_{1}, c_{2}\right)$, where $0 \leqslant c_{1} \leqslant$ $h_{1}(v)-1,0 \leqslant c_{2} \leqslant h_{2}(u)-1$. Then all «left halves» of these sages play strategy $f_{1}(v)$, and «right halves» of these sages play strategy $f_{2}$. In particular it means that all sages from subgraph $G_{2}$ name colors, that have the same first component. As for the other sages from $G$, those of them, who are not the neighbours of $v$, play by the strategy $f_{1}$. The sages from $G_{1}$, that are the neighbours of sage $v$, after the substitution discovered[found out] that instead of one neighbour of $v$ they had now $\left|V_{2}\right|$ neighbours (and, generally speaking, with different hats). These sages act by the following way: all they instead of one hat of $v$ see all the hats on subgraph $G_{2}$ and know the strategies of sages on that[this] graph. Therefore, they understand who wins in the game on subgraph $G_{2}$, denote this player by $v_{\text {new }}$ (if there are several winners, then they choose one of the winners as $v_{\text {new }}$, for example, the winner that is in the first place in the list, drawn up before). As a result, each former neighbour of $v$ looks only at $v_{\text {new }}$, precisely, on the first component of his color, and also plays by the strategy $f_{1}$. As a result either somebody guesses correctly from subgraph $G_{1} \backslash\{v\}$, or $v_{\text {new }}$ guesses correctly the left component of his own color, and the right component he guesses by his definition. Thus, somebody will guess correctly his own color. 3.3. Hat colors of players $B$ and $C$ can be interpreted as pairs in the form $(c, \epsilon)$, where $c$ is possible hat color in the game $\langle G, h\rangle, \epsilon \in\{0,1\}$. Let sage $A$ play according hypothesis $c(A)=\epsilon_{B}+\epsilon_{C}(\bmod 2)$. Sages $B$ and $C$ see their neighbours in graph $G$ and know what colors $c(B), c(C)$ they must name by winning strategy in game $\langle G, h\rangle$. Looking at hat of sage $A$, and also the hats of each other, sages $B$ and $C$ can compute what values $\epsilon_{B}$ and $\epsilon_{C}$ they must take in addition to $c(B), c(C)$. 3.4. By comparison with initial graphs one new color has been added for sages $A_{i}$ and $B_{j}$, we mean that this color is red. Let «megasage» $A$ says that he has all hats red, if he sees at least one red hat on sage $B$, in the opposite case let $A$ play by usual strategy on graph $G_{1}$. If «megasage» $B$ sees at least one red hat on $A$, then he understands that $A$ has won, if $B$ has at least one red hat. Then $B$ must care about the placements, where he has no red hats, and simply plays his strategy on graph $G_{2}$. If «megasage» $B$ does not see red hats on $A$, then he understands that $A$ has won, only if $B$ does not have red hats too, then, for checking the remained placements, $B$ says that all his hats are red. 3.5. We show strategy for new vertices and partially for old ones, then consider several placements, on which they win «immediately», and then say that on the other ones they win with the help of strategy for graph $G$. Describe a winning strategy. Denote by $c_{x}$ the hat color which sage $x$ has obtained. The color of vertex $Z$ we will consider as «composite»: $c_{Z}=(\epsilon, C)$, where the first «bit» $\epsilon$ can take on values 0 and 1 , and the second color $C$ can take on those $h(Z)$ colors that were in graph $G$ initially. - Sage $A$, in the case when he sees hat of color 0 or 1 on sage $B$, says what he sees, otherwise he says the first bit of color $c_{Z}$. - Sage $B$, in the case when he sees hat of new color on sage $C$, says 2 , otherwise he says the value of $1-c_{A}$. Consider all variants of pairs $\left(c_{A}, C_{B}\right)$ and describe the other part of strategy, also proving that it is winning. In the cases $(0,0)$ and $(1,1)$ sage $A$ immediately guesses correctly. In the cases $(0,1)$ and $(1,0)$, if we will not give hat of new color to sage $C$, then sage $B$ will guess correctly. So sage $C$, seeing on sage $B$ hat of color different from 2 , can fearlessly name hat of new color and somebody of $A, B$, $C$ will guess correctly. It remains to consider the cases $(0,2)$ and $(1,2)$. Sage $A$ will guess correctly in them, if his color coincides with the first bit of $c_{Z}$, therefore $Z$ can mean that his first bit differs from color $c_{A}$, that is he needs to guess only the second «bit», i.e. $h(Z)$ possible values. At the same time sage $B$ in these two cases will guess correctly only in the case, when sage $C$ has new color. That is $C$, seeing hat of color 2 on sage $B$, can assume that his color is not new color. Thus, the sages need to win in the game $\langle G, h\rangle$, that we are able to do by the condition. 3.6. The statement immediately follows from the theorem about product (product 3.1). 3.7. Let $A$ be new pendant vertex, $B$ be a neighbour to it vertex of graph $G$, denote the graph with added vertex by $G_{1}$. In one direction the statement is evident. If the game on $G$ is winning, then game on $G_{1}$ is also winning. Prove now that if the game on $G_{1}$ is winning, then the game on $G_{1}$ is also winning. Let the sages choose a winning strategy on $G_{1}$. Remind that in problem 1.14 it has been proved that if during the test the judge gives to one of the sages, say $B$, hint in the form «you have hat of one of two colors $c_{1}$ or $c_{2} »$, then this hint does not affect the game result. Denote this hint by $B^{*}$. Prove that if the sages win in the game on $G_{1}$, then they can win on graph $G$ with hint $B^{*}$. Fix a winning strategy $f$ on graph $G_{1}$. Construct a winning strategy on $G$ with hint $B^{*}$. Let all sages from $V(G) \backslash B$ use strategy $f$. For each pair of colors $\left(b_{1}, b_{2}\right), b_{1} \neq b_{2}$, that can be given to sage $B$, find such a color $a \in \operatorname{Col} A$ that $f_{A}\left(b_{1}\right) \neq a, f_{A}\left(b_{2}\right) \neq a$, there exists such color, since $\widehat{A} \geqslant 3$. Now specify the strategy of sage $B$ on graph $G$ taking into account the hint, namely, let in the case, when $B$ sees the set of colors $c$ on the heads of his neighbours on graph $G$ and obtains the hint «your hat is of color $b_{1}$ or $b_{2}$ », he gives the answer $f_{B}(a, c)$, i. e. answers by strategy $f$, as he saw on $G_{1}$ color $a$ on the head of sage $A$ and set of colors $c$ on the heads of other neighbours. And it can be turned out that the color that $B$ says, does not coincide with $b_{1}$ and with $b_{2}$. This strategy is winning because when in the game on $G_{1}$ we give hat of color $a$ to sage $A$ during the test and observe that he does not guess correctly (not giving some hats to $B$ ), then for sage $B$ colors $b_{1}$ and $b_{2}$ are remained possible, and on graph $G_{1}$ somebody guesses correctly. Thus, the sages win with hint $B^{*}$. 3.8. Prove the more general case, from which both statements immediately follow. The addition of two new vertices $B$ and $C$ and edges $A B, B C, C A$ to graph $G$ with vertex $A$ does not affect whether graph is winning or not, if the hatnesses of new vertices satisfy to the condition $2(\widehat{B}+\widehat{C})<\widehat{B} \cdot \widehat{C}$ (the hatnesses of vertices of graph $G$ have not been changed). Let $G^{\prime}$ be the obtained graph. If the sages win on graph $G$, then they win on $G^{\prime}$ too. Verify that if $G^{\prime}$ is winning, then $G$ is winning. Let $f^{\prime}$ be a winning strategy on $G^{\prime}$. Construct $f$, the winning strategy on $G$. For the sages from $G \backslash A$ the strategy remains the same. Set the strategy of sage $A$. Let a hats placement on graph $G$ be given. Sage $A$ sees the colors of all his neighbours in graph $G$ and considers $\widehat{B} \cdot \widehat{C}$ ways to choose colors for $B$ and $C$. Further sage $A$ examines what he has to answer in all these cases according to strategy $f^{\prime}$, and calls the color that is the most common in his answers. Show that the obtained strategy wins on graph $G$. Fix a hats placement on graph $G$ and suppose that no one of the sages from $G \backslash A$ has guessed correctly. If put hats on the heads of sages $B$ and $C$ arbitrarily, then we will obtain a hats placement on $G^{\prime}$, in which all the sages from $G \backslash A$ has not guessed correctly their colors by strategy $f^{\prime}$. Therefore, $A, B$ or $C$ will guess correctly. But from available $\widehat{B} \cdot \widehat{C}$ ways of hats placement on the heads of $B$ and $C$ sage $B$ will guess correctly in $\widehat{C}$ cases, and sage $C$ in $\widehat{B}$ cases. Therefore, $A$ has to guess correctly his color in at least $\widehat{B} \cdot \widehat{C}-(\widehat{B}+\widehat{C})>\frac{\widehat{B} \cdot \widehat{C}}{2}$ cases. Hence, sage $A$ guesses his color in the most of cases, and he will guess just this color by strategy $f$. 3.9. Denote by $N_{i}$ the set of neighbours of sage $A$ in graph $G_{i}$, and by $S$ the set of all possible hats placements of the sages from $N_{2}$. If $x$ is one of two possible hat colors of sage $A$, then the second color we will denote by $\bar{x}$. Fix arbitrary strategy $f$ for the sages on graph $G$ and describe devil's(sometimes we call the judge so) strategy, by which he will can to outplay the sages. Choose arbitrary hats placement $s \in S$ for the neighbours of $A$ in subgraph $G_{2}$. This placement is uniquely sets strategy $f^{s}$ of the sages in subgraph $G_{1}$, which is, as we know, losing on this subgraph. Choose arbitrary disproving hats placement $\varphi_{s}$ on graph $G_{1}$ for strategy $f^{s}$. Then sage $A$ obtains hat $\varphi_{s}(A)$ and has not guessed correctly, it means that the chosen hats $\varphi_{s}$ defines such a hats placement $t=\left.\varphi_{s}\right|_{N_{1}}$ for the sages from $N_{1}$, that $f_{A}(t, s)=f_{A}^{s}(t)=\overline{\varphi_{s}(A)}$. Let the devil construct disproving hats placements on all graph $G$ applying the following principle: if in the constructing disproving hats placement it is assumed to give hats placement $s$ to the sages from $N_{2}$, then hats placement $\varphi_{s}$ will be given to all the sages in subgraph $G_{1}$. In the observance of this principle, first, sage $A$ and all the other sages on subgraph $G_{1}$ have not certainly guessed correctly colors of their own hats, and second, the strategy of sage $A$ is now completely defined only by hats placements from $S$ (since on component $G_{1}$ we immediately set placement $h_{s}$ and do not consider any other variants[cases]). So, the devil sees that the sages from $V\left(G_{2}\right) \backslash A$ apply strategy $f$, and sage $\underline{A \text { really use the strategy }}$ «if I see placement $s$ on the heads of my neighbours from $N_{2}$, then I say color $\overline{\varphi_{s}(A)}$. Since graph $G_{2}$ is losing, then there exists a disproving hats placement $\psi$ on subgraph $G_{2}$ for this strategy. This hats placement allows to the devil to set correctly a disproving hats placement on all graph $G$. Indeed, hats placement $\psi$ defines the set $s=\left.\psi\right|_{N_{2}}$ of hat colors for the neighbours of $A$ in subgraph $G_{2}$, and set $s$ defines a disproving hats placement $\varphi_{s}$ on subgraph $G_{1}$, and placements $\psi$ and $\varphi_{s}$ are compatible: both placements assign color $\varphi_{s}(A)$ to sage $A$, whereas strategy $f$ requires to call color $\overline{\varphi_{s}(A)}$. 3.10. Let the sages fix strategy $f$ on graph $G^{\prime}$. Construct a losing placement for this strategy. Look at the strategy of sage $A$ for this. He says one of two colors less often, namely, no more than $h(B)-1$ times. Give the hat of this color to him. Now, in order to prevent him to guess correctly, we must give to $B$ the hat of one of the remaining $h(B)$ colors (if more colors remain, select $h(b)$ colors). That is good, we just know how to give a placement on the remained graph, where $B$ has only $h(B)$ colors. That is a disproving hats placement on the whole graph. ## Blind chess ## 4.1. W1) It is trivial. In the language of hats it is a game, where some of the sages always obtains hat of the same color, certainly, he will guess it correctly, even not looking at the others. W2) In the language of hats in the corresponding 4-cycle the hatness of two neighbour sages equals two, these two sages will provide win, even not looking at the others. W3) This statement is a retelling to the language of the game «Check by rook» of the statement that the sages win on 4 -cycle, if they all obtain hats of three colors. For example, the strategy of the sages, described in [1], in the chess language looks by the following way. If chess player sees that the king of the fellow is located in the centre, he puts the rook to the centre. In the opposite case he puts the rook to the cell, where the arrow points, leading from the king (on the auxiliary diagram for this chess player), see fig. 16. The coordinates of cells in the figure correspond to the numbers of hat colors from [1]. So, chess player $\mathcal{L}$, seeing that the king of the fellow is located in cell $(2,2)$, puts his rook on cell $(1,0)$ (this case corresponds to bold arrow in left fig. 16). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-038.jpg?height=208&width=208&top_left_y=1118&top_left_x=587) Strategy of $\mathcal{L}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-038.jpg?height=208&width=200&top_left_y=1118&top_left_x=1248) Strategy of $\mathcal{R}$ Figure 16 . W4) Number the cells of board $L(2 \times 3)$ from left to right from top to bottom, fig. $17 a$. Let the strategy of chess player $\mathcal{R}$ is given by the table in fig. 17b. Here six labels have been put in the cells of board $R(3 \times 4)$. Label $r_{i}$ means that chess player $\mathcal{R}$, seeing that the king of the fellow is located in $i$-th cell of board $L(2 \times 3)$, puts the rook on the cell of board $R(3 \times 4)$ with label $r_{i}$. The strategy of chess player $\mathcal{L}$ we also set with the help of board $3 \times 4$, see fig. $17 c$. Here there is a number from 1 to 6 in each cell of board $R(3 \times 4)$, this number denotes some cell of board $L(2 \times 3)$. When chess player $\mathcal{L}$ sees that the king is located on board $R(3 \times 4)$ in the cell with label $k$, he puts the rook on the cell with number $k$ of board $L(2 \times 3)$. To avoid misunderstandings in the notations, we use the labels of type «letter $r$ with index» for chess player $\mathcal{R}$, and the labels of type «number» for chess player $\mathcal{L}$. | 1 | 2 | 3 | | :--- | :--- | :--- | | 4 | 5 | 6 | a) Labelling of board $L$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-038.jpg?height=120&width=163&top_left_y=2142&top_left_x=929) b) The strategy of chess player $\mathcal{R}$ | 1 | 3 | 3 | 5 | | :--- | :--- | :--- | :--- | | 2 | 1 | 4 | 5 | | 2 | 6 | 6 | 4 | c) The strategy of chess player $\mathcal{L}$ Figure 17. The winning strategy for game $L(2 \times 3), R(3 \times 4)$ Describe how winning strategies can be set with the help of the introduced notations. Statement. A strategy of chess players is winning if and only if for any three different cells $a, b$, $c$ of board $L(2 \times 3)$ such that cells $b$ and $c$ do not belong to the cross of cell $a$, the following property on board $R$ holds: all the cells of board $R(3 \times 4)$, marked by number a, belong to the intersection of the crosses $r_{b}$ and $r_{c}$. For example for $a=1, b=5, c=6$ the cells with label 1 on board $R(3 \times 4)$ are located in the intersection of crosses $r_{5}$ and $r_{6}$. (The intersection of crosses $r_{5}$ and $r_{6}$ is painted in fig. 17b.) Proof of the statement. Let the judge put the king on the cell of board $R(3 \times 4)$ labeled by 1 (for the other labels the reasoning are similar), this case specifies to us the triple of cells $a=1, b=5$, $c=6$. Then chess player $L$ puts the rook on cell 1 of board $L(2 \times 3)$ by his strategy. Further, let the judge put the king on 5 -th or on 6-th cell of board $L(2 \times 3)$, only in these cases the king on board $L$ will not be in check. Therefore, in these situations check is provided by chess player $\mathcal{R}$. It means that the cell with label 1 must belong to the cross of cell $r_{5}$ and similarly the cell with label 1 must belong to the cross of cell $r_{6}$. The statement is proved. It remains to note (trying all possible cases) that the statement holds for the given example. W5) The strategy is specified similarly to the case W4), see fig. 18 . | 1 | 2 | 3 | 4 | | :--- | :--- | :--- | :--- | | 5 | 6 | 7 | 8 | a) Labelling of board $L$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-039.jpg?height=209&width=203&top_left_y=792&top_left_x=812) b) The strategy of chess player $\mathcal{R}$ | 3 | 5 | 4 | | :--- | :--- | :--- | | 8 | 6 | 8 | | 2 | 7 | 1 | c) The strategy of chess player $\mathcal{L}$ Figure 18. The winning strategy for the game $L(2 \times 4), R(3 \times 3)$ W6) In the language of hats this case means that the cycle contains path $P_{3}$ with hatnesses of vertices $2, x, 2$, where $x \leqslant 4$. The sages win on such path. L1) As in item W4), number the cells of board $L(2 \times 3)$ from left to right from top to bottom, fig. $19 a$. Then the strategy of chess player $\mathcal{R}$ is specified by the table, similar to the one shown in fig. $19 b$. Note that for this way to specify the strategy it is allowed that several labels $r_{i}$ are on the same cell. The strategy of chess player $\mathcal{L}$ can also be specified with the help of board $4 \times 4$, for example, as it is done in fig. 19 . ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-039.jpg?height=112&width=208&top_left_y=1686&top_left_x=433) a) Labelling of board $L$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-039.jpg?height=177&width=183&top_left_y=1670&top_left_x=959) b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-039.jpg?height=177&width=172&top_left_y=1670&top_left_x=1453) c) Figure 19. Two ways to specify the strategies As in item W4, the following statement holds. Statement. Strategy of chess players is winning if and only if for any three different cells $a, b, c$ of board $L(2 \times 3)$, such that cells $b$ and $c$ do not belong to the cross of cell $a$, the following property on board $R$ holds: all cells of board $R(4 \times 4)$, marked by number a, belong to the intersection of crosses $r_{b}$ and $r_{c}$. For example, for the strategy shown in fig. $19 c$, the triple of cells $a=1, b=5, c=6$ does not satisfy to the statement: only one of the three cells with label 1 belongs to the intersection of crosses $r_{5}$ and $r_{6}$. (The intersection of crosses $r_{5}$ and $r_{6}$ is painted in fig. 19b.) Prove that the chess players have not a winning strategy. Fix a strategy of chess player $\mathcal{R}$. Using the statement, try to understand, where the cells with labels 1,2 , and 3 can be located on board $R(4 \times 4)$. By the statement, the cells with label 1 belong to the intersection of crosses $r_{5}$ and $r_{6}$, the cells with label 2 to the intersection of crosses $r_{4}$ and $r_{6}$, and the cells with label 3 to the intersection of $r_{4}$ and $r_{5}$. Note that the union of pairwise intersection of any three crosses (possibly, coinciding) on board $R(4 \times 4)$ contains at most 8 cells. Indeed, consider the cases. 1. If the centres of the crosses belong to different verticals and horizontals, then each pairwise intersection consists of two cells, see the example in fig. $19 b$, where the intersection of crosses $r_{5}$ and $r_{6}$ is painted; there are not more than 6 cells. 2. If the centres of any two crosses do not coincide and two centres belong to one horizontal or vertical (as for example $r_{4}$ and $r_{5}$ in fig. 19b, then the intersection of these two crosses contains 4 cells and adding of the third cross can give another 4 cells to the union of pairwise intersections, only if the centre of this cross belongs to the same line as one of the first two centres (as $r_{4}$ and $r_{6}$ in fig. $19 b$. In this case there are 8 cells, and 7 of them belong to one cross (in the considered example in the cross $r_{4}$ ). 3. If the centres of two crosses coincide, then the intersection of crosses contains 7 cells. For any location of the third centre the set of pairwise intersection does not increase. Thus, for cells with labels $1,2,3$ on board $R(4 \times 4)$ there are at most 8 positions, similarly for the cells with labels $4,5,6$ there are at most 8 positions too. Since board $R(4 \times 4)$ contains 16 cells, we have 8 positions for labels $1,2,3$ and 8 positions for labels $4,5,6$. But as it was established by trying all possible cases, 8 positions can be realized only as the set «whole cross» plus one cell. It remains to note that it is impossible to cover all the board $R(4 \times 4)$ by two crosses and two additional cells. L2) As in item L1 board $L(2 \times 3)$ here is the same, and the right board is also «large enough». Similarly we make sure that the union of pairwise intersections of any three crosses (possibly, coinciding) on board $R(3 \times 5)$ contains no more than 8 cells; the cases, in which this intersection contains 7 or 8 cells, are shown in fig. 20, this are the cases, when the centres of two crosses belong to one row or one column (including the case, when they are in one cell). In all the cases the union of pairwise intersections of three crosses occupies one whole horizontal of the board, and in each of two other horizontals it occupies less than a half of cells. It means that the union of two such sets cannot cover the whole board. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-040.jpg?height=228&width=371&top_left_y=1508&top_left_x=114) 8 cells ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-040.jpg?height=232&width=368&top_left_y=1503&top_left_x=844) 8 клеток ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-040.jpg?height=232&width=368&top_left_y=1506&top_left_x=1572) 7 cells $r_{b}$, in any place Figure 20. The union of pairwise intersections of three crosses on board $R(3 \times 5)$ L3) This reasoning is offered by Oleg Chemokos. Fix some strategies of chess players $\mathcal{L}$ and $\mathcal{R}$ and check that such positions of kings can be found that both kings will avoid a check. Note that if a king is located in some cell of board $L(2 \times 4)$, then there are three «weak» cells (in another row) on the board, the cells, from which rook cannot attack this king. For winning strategy all the positions of king on board $R(3 \times 4)$, for which chess player $L$ puts a rook to a weak cell, must be located in one cross. And any three cells of one row can form the set of weak cells; it happens, if a king is located in the second row in the column containing the fourth cell. The strategy of chess player $\mathcal{L}$ is a mapping $\phi: R(3 \times 4) \rightarrow L(2 \times 4)$ that defines what cell $\phi\left(K_{R}\right)$ for putting the rook will be chosen by chess player $\mathcal{L}$, if he will see the king in the cell $K_{R}$ of board $R(3 \times 4)$. Paint the cells of board $R(3 \times 4)$ in two colors: cell $x$, for which cell $\phi(x)$ is located in the first row of board $L(2 \times 4)$ we paint in white color, the others into black. Without loss of generality we may think that the number of white cells on the board is not less than the number of black cells. Consider two cases, covered all the possibilities, describing how this «not less» can be realized. 1. One of the rows of the board $R(3 \times 4)$ contains three white cells $r_{1}, r_{2}, r_{3}$ (will call this row the first) and another one row (the second) contains two white cells $r_{4}$ and $r_{5}$. Then cell $\ell$ can be chosen in the first row of the board $L(2 \times 4)$ such that preimage $\phi^{-1}(\ell)$ is either empty set, or just one cell of the first row, say, $r_{1}$. In this case the other cells of the first row form a weak set, and its preimage contains cells $r_{2}, r_{3}, r_{4}$ and $r_{5}$, not belonging to one cross. 2. Each row of board $R(3 \times 4)$ contains two white cells. Then choose in the first row of the board $L(2 \times 4)$ a cell $\ell$, for which preimage $\phi^{-1}(\ell)$ consists of at most one cell (for specificity, it is located in the third row). In this case the other cells of the first row of the board $L(2 \times 4)$ form a weak set, and their four preimages, belonging to the first two rows of board $R(3 \times 4)$, form a set that cannot be covered by one cross. L4) Suppose that the chess players have a winning strategy. Using the notations for describing winning strategies from item W4). Number the cells of board $L(3 \times 3)$ by numbers from 1 to 9 . Then the strategy of chess player $\mathcal{R}$ is specified by a placement of nine symbols: $r_{1}, r_{2}, \ldots, r_{9}$ on board $R(3 \times 4)$. And the strategy of chess player $\mathcal{L}$ is specified by writing the numbers from 1 to 9 in each cell of board $R(3 \times 4)$, which we call labels. As in the previous item, when we put the king on cell $i$ on board $L(3 \times 3)$ there are 4 cells, from which the rook cannot attack this king. These cells and their numbers we call $i$-weak. If the strategy is winning, then it is necessary that for all $i$ the labels on board $R(3 \times 4)$, coinciding with $i$-weak numbers, are located in the cross with centre in $r_{i}$. Note that symbols $r_{1}, r_{5}$ and $r_{9}$ must be located in different rows of board $R(3 \times 4)$. Indeed, it is easy to see that each cell of board $L(3 \times 3)$ is located in weak position with respect to one of the cells with numbers 1,5 or 9 . (For example, 1 and 2 are in weak position with respect to 9,3 is in weak position with respect to 5 and so on). Therefore, each label on board $R(3 \times 4)$ is located in $r_{1^{-}}, r_{5}$ - or $r_{9}$-cross. It can be only if symbols $r_{1}, r_{5}$ and $r_{9}$ are located in different rows. Similarly, symbols $r_{i}, r_{j}$ and $r_{k}$ are located in different rows, if cells $i, j, k$ occupy three different rows and three columns of board $L(3 \times 3)$. Corollary. Two possible cases of placement of symbols $r_{1}, r_{2}, \ldots, r_{9}$ on board $R(3 \times 4) 7$ are possible: 1) either symbols $r_{1}, r_{2}, r_{3}$ are located in one row of board $R(3 \times 4)$, symbols $r_{4}, r_{5}, r_{6}$ are located in another row, and symbols $r_{7}, r_{8}, r_{9}$ are in the third row; 2) or symbols $r_{1}, r_{4}, r_{7}$ are located in one row of board $R(3 \times 4)$, symbols $r_{2}, r_{5}, r_{8}$ are located in another row, and symbols $r_{3}, r_{6}, r_{9}$ are in the third row. The corollary is proved by gently nasty looking all the possible cases. Prove that there are no winning strategies that have these properties. Put rooks on all cells $r_{i}$ of board $R(3 \times 4)$ (we put on cell as many rooks as there are symbols $r_{i}$ in it). By the corollary in the first row of board $R(3 \times 4)$ there is an «empty» cell, i.e. cell, containing no symbols $r_{i}$, but containing some label $a$. Let for the specificity it be located in the fourth column (fig. 21). By the statement, 4 rook's attacks are directed to this cell, and two of these four rooks are located in one row, and another two are in another row. It means that two rooks are certainly located in one of the cells of fourth column. Let for the specificity this cell be located in the second row. Now we know that in the second row 3 rooks have been put in total, and two of them are located in one cell. Therefore, there are two «empty» cells in the second row. We choose one of them, above which in the first row no more than one rook is located. Let this cell be located in the first column for the specificity and contains label b. 4 rook's attacks from two pairs of rooks, located in two rows, are directed to the chosen cell. One pair of rooks is located, evidently, in the second row, and another pair is located in the third row (there is no more than one rook in the first row above cell $b$ ). Now we see that one of ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-041.jpg?height=206&width=280&top_left_y=2487&top_left_x=357) Figure 21. The strategy for the case L3 | 1 | 2 | 3 | 4 | 5 | | :---: | :---: | :---: | :---: | :---: | | 6 | 7 | 8 | 9 | 10 | a) Labelling of board L ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-041.jpg?height=208&width=214&top_left_y=2420&top_left_x=1612) b) The strategy of chess player $\mathcal{L}$ Figure 22. We find a strategy for the game $L(2 \times 5), R(3 \times 3)$ the cells in the third row, in the second or in the third column, cannot gather 4 rook's attacks from two different rows. It is contradiction. L5) Suppose that the chess players have a winning strategy. Use the notations for describing winning strategies from item W4). Number the cells of board $L(2 \times 2)$ by numbers from 1 to 4 . Then the strategy of chess player $\mathcal{R}$ is specified by placement of four symbols $r_{1}, r_{2}, r_{3}, r_{4}$ on the board $R(5 \times 5)$. On board $R(5 \times 5)$ at least one cell can be found, not belonging to any of four crosses, defined by these symbols; call this cell $Q$. The strategy of chess player $\mathcal{L}$ is specified by writing in each cell of board $R(5 \times 5)$ the numbers from 1 to 4 . Consider the number, written in cell $Q$, without loss of generality, it is 1 . Consider the number on board $L(2 \times 2)$, located on the same diagonal as 1 , without loss of generality, it is 4 . Let the judge put the kings: on cell $Q$ on board $R(5 \times 5)$ and on cell 4 on board $L(2 \times 2)$. Then player $\mathcal{L}$ puts the rook on cell 1 of board $L(2 \times 2)$, and player $\mathcal{R}$ puts the rook on cell $r_{4}$ of board $R(5 \times 5)$. None of the rooks has a king in check. The chess players lost. L6) Label board $L$, as in fig. 22 a). As in W4), the strategy of chess player $\mathcal{L}$ is specified by writing in each cell of board $R(3 \times 3)$ the numbers from 1 to 10 , the numbers of cells on board $L(2 \times 5)$. Since there are only two horizontals in board $L(2 \times 5)$, there exist two rows of board $R(3 \times 3)$, in each of them the numbers of two cells are written, such that all these four cells (possibly, there are coinciding among them) belong to one horizontal of board $L(2 \times 5)$. Let $j$ be the number of the cell from the second horizontal, that is $i$-weak with respect to all these cells. For example, let labels 1, 2, 3, 4 are located on board $R(3 \times 3)$, as in fig. 22 b). Then the number 10 is 1-, 2-, 3- and 4 -weak simultaneously. It means that the rook on cell $r_{10}$ of board $R(3 \times 3)$ attacks the cells with labels $1,2,3$ and 4 . It is impossible: it must be located in the upper row of board $R(3 \times 3)$ to attack labels 1 and 2, and in the bottom row to attack 3 and 4 . By the same reason the general case is also impossible: cell $r_{j}$ must be located in two rows of $R(3 \times 3)$ simultaneously. 4.2. Paint the cells of both boards as shown in fig. 23, a). Let both chess players put their queens only on the cells that occupied by the queens, and let the first chess player act according the assumption «Kings are located in cells of the same color», and the second from the assumption «The kings are located in cells of different colors». We can use the usual chess coloring instead of «exotic» coloring as above. Indeed, the queen, located in cell $c 2$, holds under attack all the cells of the same color in chessboard coloring! And the same for $c 3$, fig. 23, b). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-042.jpg?height=175&width=209&top_left_y=1877&top_left_x=775) a) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-042.jpg?height=163&width=226&top_left_y=1889&top_left_x=1072) b) Figure 23. «Check by queen» on boards $4 \times 5$ 4.3. Statement of the problem has been found with help of computer. The participants of our conference suggest beautiful logical strategies. Solution 1 (Kononenko Nukolay). Specify strategy of the chess players. Label board $R(5 \times 5)$ as shown in fig. $24 \mathrm{a}$ ). Seeing the king on the cell with label $j$, chess player $\mathcal{L}$ puts the rook on the cell of board $L(4 \times 4)$, labelled by number $j$, fig. $24 \mathrm{~b}$ ). Therefore, chess player $\mathcal{L}$ uses only four positions for his queen. For each cell of board $L(4 \times 4)$ in fig. 24 c) it is shown, from which positions the queen of chess player $\mathcal{L}$ does not attack this cell. For example, the numbers 1 and 2 in the lower left corner mean that the lower left corner cell of board $L(4 \times 4)$ is not under attack by the queen located at 1 -th and in 2-th positions, shown in fig. $24 \mathrm{~b}$ ), and "-" means that the cell is under attack from all positions. | 3 | 1 | 3 | 1 | 3 | | :--- | :--- | :--- | :--- | :--- | | 1 | 3 | 3 | 3 | 4 | | 3 | 3 | 3 | 3 | 3 | | 2 | 3 | 3 | 3 | 4 | | 3 | 2 | 3 | 2 | 3 | | $a$ | $b$ | $c$ | $d$ | $e$ | a) The strategy of chess player $\mathcal{L}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-043.jpg?height=232&width=234&top_left_y=181&top_left_x=954) | 3 | 2,4 | 4 | 1,2 | | :---: | :---: | :---: | :---: | | 2 | 4 | - | 4 | | 3 | - | 4 | 1,4 | | 1,2 | 3 | 1 | 3 | b) On which cell we put the queer) Instruction for $\mathcal{R}$ Figure 24. Seeing the king on board $L(4 \times 4)$, chess player $\mathcal{R}$ with help of fig. 24 c) immediately understands, from which «unfavourable» positions the queen of his fellow cannot put the king in check. Therefore he must to locate his queen on board $R(5 \times 5)$ so that it attacks all the cells, sending the queen of chess player $\mathcal{L}$ to a unfavourable position. For unfavourable positions 1,2 it is possible to put the queen on cell $b 3$, for 1,4 on cell $c 4$, for 2 , 4 on cell $c 2$, for 3 on cell $c 3$. Solution 2 (Presnova Ekaterina, Raceeva Olga). Label board $L(4 \times 4)$ as shown in fig. 25 a). Observation 1: for this labelling the property holds: for any pair of labels all the cells marked by this pair of labels, can be attacked by one queen. For example, all the cells, marked by 0 or 1 , can be attacked from cell $a 4$; all the cells marked by 0 or 2 , can be attacked from cell $c 2$ and so on. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-043.jpg?height=57&width=1759&top_left_y=1185&top_left_x=183) board $R(5 \times 5)$, fig. $25 \mathrm{~b})$. Observation 2: any cell of board $R(5 \times 5)$ can be attacked from at least two different positions from this set. Then chess player $\mathcal{L}$, seeing the king on board $R(5 \times 5)$ immediately understands, from which positions (among the indicated four positions) the queen of player $\mathcal{R}$ cannot put the king in check. In view of observation 2 there are no more than 2 such «unfavourable» positions, and chess player $\mathcal{L}$ should put his queen on board $L(4 \times 4)$ so that he attacks all the cells, sending the queen of chess player $\mathcal{L}$ on unfavourable position. It is possible, in view of observation 1 . ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-043.jpg?height=249&width=237&top_left_y=1623&top_left_x=316) a) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-043.jpg?height=297&width=299&top_left_y=1573&top_left_x=730) b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-043.jpg?height=235&width=365&top_left_y=1687&top_left_x=1394) Figure 26 . Figure 25 . 4.4. Solution 1 (Kostina Ekaterina, Mirgalimova Rozalina, Hamikova Marina). We will prove the stronger fact, that in the game «Check by queen» on boards $L(4 \times 6), R(7 \times 7)$ the sages lose. Let the judge plans to put the king on one of cells $A, B, C$ of board $L(4 \times 6)$ (fig. 26). For each of these positions player $\mathcal{R}$ is ready to put the queen on board $R(7 \times 7)$. Since three queens cannot hold in check all the cells of board $7 \times 7$, there exists a cell on board $R(7 \times 7)$, which is attacked by no one of the queens. If the judge puts the king on this cell, player $\mathcal{L}$ answers putting the queen on board $L(4 \times 6)$. Since cells $A, B, C$ cannot be attacked by one queen, at least one of them will be not under attack, and the judge, finally will formulates his plans regards board $B$ : he will pat the king exactly on this cell. The chess players have lost. 4.5. The result is obtained by computer. 4.6. The result is obtained by computer. 4.7. Answer: yes. 5 queens can be placed on $11 \times 11$ board, that attack all the cells (for example, $b 4, d 10, f 6, h 2$ and $j 8)$. Then combine idea of the solution of problem 4.2 and standard idea of the game with hats of five colors on 5-clique. 5.1. Let the sage fix some strategy on new graph. The strategy of sage $A$ can be given in the form of $3 \times 7$ table: the rows correspond to the hat colors of sage $B$, the columns to the hat colors of sage $A$. In cell of table number of color (0 or 1) will be written, that sage $A$ calls, when he sees on $B$ and $C$ the corresponding hat colors. In each column of the table one of the symbols, 0 or 1 , occurs two times. Mark the cells containing repeated symbol. (If symbol occurs in all three cells of column, mark any two of them.) The marked cells can be located either in the first and the second rows, or in the first and the third. Since there are 7 columns, by the pigeonhole principle there exist two rows, in which the marked cells occupy three columns. In the marked cells of one column two zeroes and two ones can be, therefore, it is possible to choose two columns of three so that in the indicated columns in the marked cells there is the same number. Thus, we have chosen in the table two rows (for the specificity $i$-th and $j$-th) and two columns (for the specificity $k$-th and $\ell$-th), in the intersection of which there is the same number, for the specificity it is 0 . Now we without problem construct a disproving hats placement on all the graph. Give hat of color 1 to sage $A$, will choose hat of $i$-th and $j$-th color to sage $B, k$-th or $\ell$-th to sage $C$. For this approach sage $A$ will certainly guessed incorrectly his own color, since in according to the table he will call color 0 . As to assignment of concrete colors to sages $B$ and $C$, and also for the others, consider game on graph $G$ : after fixation of color of sage $A$ the strategy of other sages on graph $G$ is uniquely defined, the accepted restrictions of hat colors of $B$ and $C$ allow to suppose that their hatnesses now equal 2. Since graph $G$ is losing, we will can to present a disproving hats placement on it. 5.2. Prove the following a few more strong statement. There exists positive integer $N$ such as on any graph $G$, degrees of all vertices of which do not exceed 3 , and hatness function is given by formula $$ \widehat{a}= \begin{cases}3, & \text { if } \operatorname{deg} a=1 \\ 41, & \text { if } \operatorname{deg} a=2 \\ N, & \text { if } \operatorname{deg} a=3\end{cases} $$ the sages lose. Proof. Denote $m=80\binom{81}{41}+1$. Show that $N=80\binom{m}{41}\binom{81}{41}+1$ is suitable. Induction by the number of vertices. Basis (graph with two vertices) is evident. Induction step. Consider the vertex of the least degree, denote it by $A$. Case 1. $\operatorname{deg} A=1$. By the statement of problem 3.7 addition of vertex of degree 3 does not affect the property of graph be winning or not, and if increase hatness of one of vertices much in losing graph, then it remains losing. Case 2. $\operatorname{deg} A=2$. Denote the neighbours of $A$ by $B$ and $C, \operatorname{deg} B \leqslant \operatorname{deg} C$. Le m ma. Table $(2 k-1) \times\left(2(\ell-1)\binom{2 k-1}{k}+1\right)$ is painted in 2 colors. Then it is possible to choose such $k$ rows and $\ell$ columns that all these cells in the intersection of these rows and columns have been painted in the same color. Proof. Consider arbitrary column, there are $2 k-1$ cells in it. Then by the pigeonhole principle there exist $k$ cells of one color. Mark these $k$ cells. Make this for each column. The number of ways to mark $k$ cells of $2 k-1$ equals $\binom{2 k-1}{k}+1$, these cells can be of one of two colors. Therefore, by the pigeonhole principle there exist $\ell$ columns such that the marked cells are located in the same set of rows and they are of the same color. Q.E.D. Case 2.1. $\operatorname{deg} B=\operatorname{deg} C=2$. Show that the sages lose in graph $G$, even if $\widehat{A}=2$. Suppose that there exists a winning strategy of the sages on $G$. The strategy of sage $A$ is a $41 \times 41$ table, in which each cell is painted in one of two colors. Оставив от нее only 5 rows, apply the lemma for $k=\ell=3$. 3 rows and 3 columns can be chosen from the table so that their intersections are painted in one color. Then the judge will put hat of another color on $A$, and for $B$ and $C$ will choose hats of only three colors corresponding to three rows and three columns correspondingly[?]. We will obtain the winning strategy for $G \backslash A, \widehat{B}=\widehat{C}=3$ that contradict the induction hypothesis. Case 2.2. deg $B=2$, $\operatorname{deg} C=3$. Show that here the sages lose too on graph $G$, even if $\widehat{A}=2$. For $k=3, \ell=41$ the table from the lemma has size $5 \times 801$, and the hatnesses of sages $B$ and $C$ are much larger: they equal 41 and $N$. Applying the lemma similarly to case 2.1, we obtain contradiction again. Case 2.3. $\operatorname{deg} B=\operatorname{deg} C=3$. Similarly to case 2.1 , use the lemma for $k=\ell=41$. We can use it because for $k=\ell=41$ the table has size $81 \times N$, both dimensions are less than hatnesses of sages of $N$. Case 3. $\operatorname{deg} A=3$. Let $B, C$ and $D$ be the neighbours of sage $A$. Then $\operatorname{deg} B=\operatorname{deg} C=\operatorname{deg} D=3$. Fix some color of sage $D$. Then the strategy of sage $A$ is a $N \times N$ table. Taking 81 rows and $m$ columns, apply the lemma, find 41 rows and 41 columns, the intersections of which are painted in one color. The number of ways to choose 41 rows and 41 columns equals $\binom{m}{41}\binom{81}{41}$. Perform these actions for each of $N$ possible colors of sage $D$. Since $N=80\binom{m}{41}\binom{81}{41}+1$, there exists a color of sage $D$, for which the sets of rows, columns and colors of intersections coincide. Put on sage $A$ hat of the opposite color, take 41 colors for each of the sages $B, C, D$, and obtain the contradiction to the induction hypothesis. ## References [1] Задачи Санкт-Петербургской олипмпиады школьников по математике 2016 года. М.: МЦНМО, 2017. # Шашки Фейнмана ## Введение в ‘алгоритмическую’ квантовую теорию поля Е. Ахмедова, Р. Валиева, А. Воропаев, М. Скопенков, А. Устинов Резюме. Оказывается, радужные разводы на мыльных пузырях и невероятные законы движения электронов можно объяснить с помощью одной очень незатейливой модели. Это игра, в которой по простым правилам по клетчатой доске движется шашка, а мы следим за ее поворотами. Эти “шаштки Фейнмана", с некоторыми серьезными оговорками, могут описать все на свете явления, кроме атомного ядра и гравитации. Мы будем решать задачи по математике и обсуждать их физический смысл; предварительных знаний по физике не требуется. ## Основные результаты. - явная формула для доли света данного цвета, отражаемого стеклянной пластиной данной ширины (задача 24; - явная формула для вероятности обнаружения в данной точке электрона, испущенного из $(0,0)$ (задача 15 см. рис. 11). Хотя эти результаты сформулированы в терминах физики, они являются математическими теоремами, так как далее приводятся математические модели интересующих нас явлений, со строгими определениями. Точнее, последовательность моделей, описывающих явления со всё большей точностью. План. Мы начнём с базовой модели и будем уточнять её в каждом последующем разделе. Перед каждым уточнением мы укажем, для ответа на какой физический вопрос оно нужно, какие упрощающие предположения в нём используются и какие экспериментальные резу.льтаты оно позволяет объяснить. Наша цель - получить так называемую двумерную квантовую электродинамику, однако финальные шаги в этом направлении (намеченные в разделах 8 10) пока не сделаны. (Её 4 -мерный аналог способен, с некоторыми ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-047.jpg?height=506&width=603&top_left_y=541&top_left_x=1366) Базовая модель из \$1 ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-047.jpg?height=491&width=588&top_left_y=1091&top_left_x=1379) Модель с массой из $\$ 3$ Рис. 1: Вероятность нахождения электрона в данной точке (белым показаны сильные колебания) серьезными оговорками и исключениями, описать все явления, но мы не будем рассматривать его.) Схема уточнений может помочь вам в выборе порядка решения задач: ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-047.jpg?height=368&width=1976&top_left_y=1992&top_left_x=17) Соглашения. Если в условии задачи сформулировано утверждение, то его надо доказать. Загадка - это задача, в которой требуется и сформулировать утверждение, и доказать его. Трудные задачи отмечены звёздочками; за решение одной из них вы получите первый разряд по шашкам Фейнмана, а за решение трёх из трёх разных разделов - станете мастером. Решения принимаются в письменном виде, но вы можете потратить решённую звёздочку на 10 попыток сдать задачу устно (успешных или нет). Если не получается решить задачу, попробуйте следующие: в них могут быть подсказки. Даже если вы не докажете основные результаты, вы многое узнаете. Приветствуются попытки сформулировать и доказать свои собственные гипотезы; вы получите ранг гроссмейстера за нетривиальное открытие (а может быть, даже напишете собственную научную статью). ## 1 Базовая модель Вопрос: какова вероятность обнаружения электрона в клетке $(x, y)$, если он был испущен из клетки $(0,0)$ ? Предположения: электрон равномерно движется по оси $y$ и не движется по оси $z$, точечный источник, единичная масса и шаг решётки, нет взаимодействий, нет рождения электрон-позитронных пар. Результаты: эксперимент на двух щелях, сохранение заряда. На бесконечной шахматной доске шашка ходит на соседнюю по диагонали клетку, влево-вверх или вправо-вверх. Каждому пути $s$ шашки сопоставим вектор $\vec{a}(s)$ на плоскости следующим образом. В начале движения этот вектор направлен вверх и имеет длину 1. Пока шашка движется вдоль прямой, вектор не меняется, а после каждого поворота шашки он поворачивается на $90^{\circ}$ по часовой стрелке (независимо от того, в какую сторону повернула шашка). В конце движения вектор сжимается $2^{(y-1) / 2}$ раз, где $y$ - общее число ходов шашки (т.е. заменяется на вектор такого же направления, но длины $1 / 2^{(y-1) / 2}$ ). Полученный в итоге вектор и есть ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-048.jpg?height=465&width=474&top_left_y=316&top_left_x=1499) $\vec{a}(s)$. Например, для верхнего пути на рис. 2 вектор $\vec{a}(s)=(1 / 8,0)$. Обозначим $\vec{a}(x, y):=\sum_{s} \vec{a}(s)$, где суммирование ведется по всем путям шашки из клетки $(0,0)$ в клетку $(x, y)$, начинающихся с хода вправовверх. Если таких путей нет, то будем считать $\vec{a}(x, y):=\overrightarrow{0}$. Например, $\vec{a}(1,3)=(0,-1 / 2)+(1 / 2,0)=(1 / 2,-1 / 2)$. Квадрат длины вектора $\vec{a}(x, y)$ называется вероятностья ${ }^{1}$ обнаружения в точке $(x, y)$ электрона, испущенного из точки $(0,0)$. Она обозначается $P(x, y):=|\vec{a}(x, y)|^{2}$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-048.jpg?height=275&width=500&top_left_y=802&top_left_x=1492) Рис. 2: Пути шашек На рис. 1 цвет точки $(x, y)$, где $x+y$ чётно, показывает значение $P(x, y)$. Стороны «угла» на этом рисунке не параллельны прямым $y= \pm x$ (и никто не знает, почему!). Будем называть клетки $(x, y)$ с чётными и нечётными $x+y$ чёрными и белыми соответственно. 1. Наблюдения для малых $y$. Ответьте на вопросы для $y=1,2,3,4$ (и сформулируйте свои собственные вопросы и предположения для произвольного $y)$ : найдите $\vec{a}(x, y)$ и $P(x, y)$ для всех $x$; когда $P(x, y)=0$ ? Чему равна $\sum_{x \in \mathbb{Z}} P(x, y)$ для фиксированного $y$ ? Как направлены $\vec{a}(1, y)$ и $\vec{a}(0, y)$ ? Вероятност» ${ }^{2}$ обнаружения электрона а клетке $(x, y)$ при поглощении в клетке $\left(x^{\prime}, y^{\prime}\right)$ определяется аналогично $P(x, y)$, только суммирование производится по путям $s$, не проходящим через $\left(x^{\prime}, y^{\prime}\right)$. Обозначим ее $P\left(x, y\right.$ минуя $\left.x^{\prime}, y^{\prime}\right)$. 2. Эксперимент на двух щелях. Верно ли, что $P(x, y)=$ $P(x, y$ минуя 0,2$)+P(x, y$ минуя 2,2$)$ ? А что $P\left(x, y\right.$ минуя $\left.x^{\prime}, y^{\prime}\right) \leq P(x, y)$ ? 3. Найдите $P(0,12)$. Придумайте быстрый способ составления таблицы значений $\vec{a}(x, y)$, не требующий перебора всех путей. (Команда, первой при- ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-048.jpg?height=292&width=497&top_left_y=1433&top_left_x=1522) Рис. 3: $a_{2}(x, 1000)$ нёсшая правильное решение, получает первый разряд.) Обозначим через $a_{1}(x, y)$ и $a_{2}(x, y)$ координаты вектора $\vec{a}(x, y)$; см. рис. 3 . 4. Уравнение Дирака. Выразите $a_{1}(x, y)$ и $a_{2}(x, y)$ через $a_{1}(x \pm 1, y-1)$ и $a_{2}(x \pm 1, y-1)$. 5. Сохранение вероятности/заряда. Для каждого целого $y>0$ выполнено $\sum_{x \in \mathbb{Z}} P(x, y)=1$. 6. Симметрия. Найдите формулу, которая связывает между собой $a_{1}(x, 100)$ при $x<0$ и при $x \geq 0$. Тот же вопрос про $a_{1}(x, 100)+a_{2}(x, 100)$. 7. Принцип Гюйгенса. Как быстро найти $\vec{a}(x, 199)$, зная $\vec{a}(x, 100)$ при всех целых $x$ ? 8. Используя компьютер, постройте графики функций $f_{y}(x)=P(x, y)$ для разных $y$ (соедините каждую пару точек $\left(x, f_{y}(x)\right)$ и $\left(x+2, f_{y}(x+2)\right)$ отрезком); ср. с рис. 3 То же для функции $a_{1}(x, y)$. 9.* Найдите явную формулу для вектора $\vec{a}(x, y)$ и вероятности $P(x, y)$ (в ответе можно использовать сумму не более $y$ слагаемых). 10.** (Загадка) Угадайте простую "приближенную формулу" для $\vec{a}(x, y)$ и $P(x, y)$, работающую при $|x|<0$ найдётся такое $N$, что для любого натурального $n>N$ верно $\left|a_{n}-a\right|<\varepsilon$. Обозначение: $a=\lim _{n \rightarrow \infty} a_{n}$. В частности, $\lim _{n \rightarrow \infty}\left(2-\frac{1}{2^{n}}\right)=2$. Определим бесконечную сумму как $\sum_{k=0}^{\infty} a_{k}:=\lim _{n \rightarrow \infty} \sum_{k=0}^{n} a_{k}$. Тогда в самом деле $\sum_{k=0}^{\infty} \frac{1}{2^{k}}=2$. Следующее обобщение, называемое биномом Ньютона, разрешается использовать без доказательства: $$ (1+x)^{r}=\sum_{k=0}^{\infty} \frac{r(r-1) \cdots(r-k+1)}{k(k-1) \cdots 1} x^{k} $$ для всех комплексных $x$, для которых $|x|<1$, и для всех действительных $r$, или для $x=1$ и $r>-1$ В частности, для $r=-1$ и $r=-\frac{1}{2}$ получаем: $$ \frac{1}{1-x}=\sum_{k=0}^{\infty} x^{k} \quad \text { и } \quad \frac{1}{\sqrt{1-x}}=\sum_{k=0}^{\infty} \frac{2 k(2 k-1) \cdots(k+1)}{k(k-1) \cdots 1} \frac{x^{k}}{4^{k}} $$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-049.jpg?height=54&width=1396&top_left_y=2063&top_left_x=153) $$ \sqrt{2 \pi} k^{k+1 / 2} e^{-k} \leq k!\leq e k^{k+1 / 2} e^{-k} $$ где $e$ обозначает $\lim _{n \rightarrow \infty}(1+1 / n)^{n}$. Это иррациональное число между 2.71 и 2.72 .[^1] ## 3 Macca Вопрос: какова вероятность обнаружения правого электрона массой $m$ в клетке $(x, y)$, если он был испущен из клетки $(0,0)$ ? Предположения: масса и шаг решётки теперь произвольные. Результаты: формула для вероятности для маленького шага решётки. Чтобы провести эксперимент, проверяющий нашу модель, нам нужно следующее обобщение. Зафиксируем $\varepsilon, m>0$, называемые шагом решётки и массой частици соответственно. Каждому пути $s$ сопоставим вектор $\vec{a}(s, m \varepsilon)$ следующим образом. В начале движения у нас будет вектор $(0,1)$. Пока шашка движется вдоль прямой, вектор не меняется, а после каждого поворота он поворачивается на $90^{\circ}$ по часовой стрелке и умножается на $m \varepsilon$. В конце движения вектор сжимается в $\left(1+m^{2} \varepsilon^{2}\right)^{(y-1) / 2}$ раз, где $y$ - общее число ходов шашки. Полученный в итоге вектор и есть $\vec{a}(s, m \varepsilon)$. Векторы $\vec{a}(x, y, m \varepsilon, \pm)$ и числа $P(x, y, m \varepsilon, \pm)$ определяются аналогично $\vec{a}(x, y, \pm)$ и $P(x, y, \pm)$, только вместо $\vec{a}(s)$ используется $\vec{a}(s, m \varepsilon)$. В частности, $P(x, y, 1,+)=P(x, y,+)$. На нижнем графике рисунка 1 цвет точки $(x, y)$ для чётных $x+y$ показывает значение $P(x, y, 0.02,+)+P(x, y, 0.02,-)$. 13. (Загадка) Безмассовые и тяжёлые частицы. Найдите $P(x, y, 0,+)$ и придумайте определение числа $P(x, y, \infty,+)$ для всех $x, y$. 14. Решите аналоги задач 4, 5, и 9 для $m \varepsilon \neq 1$. В эксперименте мы измеряем вероятность найти электрон не в отдельных точках, а на интервалах $x_{0} \leq x \leq x_{0}+\Delta x, y=y_{0}$. Здесь $x_{0}, y_{0}, \Delta x$ - это не целые числа, а длины, измеряемые в метрах. Если все клетки маленькие и имеют размеры $\frac{1}{n} \times \frac{1}{n}$, то этот интервал можно приблизить набором чёрных клеток $\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor\right)$ с $x, y$, удовлетворяющими (не)равенствам выше. Это приводит нас к следующей задаче. 15. * (Первая основная задача) Непрерывный предел. Найдите $\lim _{n \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-\right)$ и $\lim _{n \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},+\right)$ для всех $x, y, m$. В ответе разрешается использовать следующие выражения 7 . $$ J_{0}(z):=\sum_{k=0}^{\infty}(-1)^{k} \frac{(z / 2)^{2 k}}{(k!)^{2}} \quad \text { и } \quad J_{1}(z):=\sum_{k=0}^{\infty}(-1)^{k} \frac{(z / 2)^{2 k+1}}{k!(k+1)!} $$ ## 4 Источник Вопрос: какова вероятность обнаружения правого электрона в клетке $(x, y)$, если он был испущен источником с длиной волны $\lambda$ ? Предположения: теперь источник реалистичный. Результаты: распространение волн. Реалистичный источник испускает электроны не точечно из $x=0$ (как в нашей игре), а широким волновым импульсом. В терминах нашей игры это означает, что шашка может начинать из любой чёрной клетки на горизонтали $y=0$ (не слишком далеко от начала координат), но исходное направление вектора $\vec{a}(s)$ повёрнуто на угол, пропорциональный расстоянию от стартовой клетки до $(0,0)$; см. рис. 4 Более строго, зафиксируем действительные $\varepsilon, \lambda>0$ и нечётное $\Delta$, называемые шагом решётки, длиной волны, и шириной импулъса соответственно. Обозначим через $R^{\alpha} \vec{a}$ поворот вектора $\vec{a}$ на угол $|\alpha|$ против часовой стрелки, если $\alpha \geq 0$, и по часовой, если $\alpha<0$. Определим вектор ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-050.jpg?height=365&width=574&top_left_y=1959&top_left_x=1392) Рис. 4: Пути шашек теперь могут начинаться с разных клеток. $$ \vec{a}(x, y, \lambda / \varepsilon, \Delta,+):=\frac{1}{\sqrt{\Delta}} \sum_{\substack{x_{0}=1-\Delta \\ x_{0} \text { чётно }}}^{\Delta-1} \sum_{s} R^{2 \pi x_{0} \varepsilon / \lambda} \vec{a}(s) $$[^2]где вторая сумма берётся по всем путям $s$ из клетки $\left(x_{0}, 0\right)$ в клетку $(x, y)$, которые начинаются и заканчиваются ходом вправо-вверх. Квадрат длины этого вектора назовём вероятностью обнаружить правый электрон в клетке $(x, y)$, если он был испущен источником с длиной волны $\lambda$ и шириной импулъса $\Delta$. Эта вероятность обозначается $P(x, y, \lambda / \varepsilon, \Delta,+)$. Определим $\vec{a}(x, y, \lambda / \varepsilon, \Delta,-)$ и $P(x, y, \lambda / \varepsilon, \Delta,-)$ аналогично. В частности, $P(x, y, \lambda / \varepsilon, 1,+)=P(x, y,+)$ для всех $\lambda, \varepsilon$, и $\vec{a}(x+1,1, \lambda / \varepsilon, \Delta,+)=\frac{1}{\sqrt{\Delta}}\left(-\sin \frac{2 \pi x \varepsilon}{\lambda}, \cos \frac{2 \pi x \varepsilon}{\lambda}\right)$ для чётных $|x|<\Delta$. 16. Пусть $\Delta=3, \lambda / \varepsilon=4$. Найдите вектор $\vec{a}(x, y, 4,3,+)$ и вероятность $P(x, y, 4,3,+)$ для $y=1,2,3$ и всех $x$. Чему равна $\sum_{x \in \mathbb{Z}}(P(x, y, 4,3,+)+P(x, y, 4,3,-))$ для фиксированного $y=1,2,3$ ? Когда $P(x, 3,4,3,+)=0$ ? 17. Сохранение вероятности/заряда. Решите аналоги задач 4 и 5 для $\vec{a}(x, y, \lambda / \varepsilon, \Delta,-)$, $\vec{a}(x, y, \lambda / \varepsilon, \Delta,+)$, и $P(x, y, \lambda / \varepsilon, \Delta,+)+P(x, y, \lambda / \varepsilon, \Delta,-)$ вместо $a_{1}(x, y), a_{2}(x, y)$ и $P(x, y)$. 18. Причинность. И $\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)$, и $\Delta P(x, y, \lambda / \varepsilon, \Delta,+)$ не зависят от $\Delta$ для $\Delta>y+|x|$. Обозначим $8^{8} \vec{a}(x, y, \lambda / \varepsilon, \pm)=\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta, \pm)$ и $P(x, y, \lambda / \varepsilon, \pm)=\Delta P(x, y, \lambda / \varepsilon, \Delta, \pm)$ для $\Delta>y+|x|$. 19. Волна. Как найти $\vec{a}(x, 100, \lambda / \varepsilon,+)$ для всех чётных $x$, если нам известен этот вектор для одного чётного $x$ ? 20. Распространение волн. Для всех $x, y, \lambda, \varepsilon$ найдите $\vec{a}(x, y, \lambda / \varepsilon,-), P(x, y, \lambda / \varepsilon,-)$ и $P(x, y, \lambda / \varepsilon,-)+$ $P(x, y, \lambda / \varepsilon,+)$. ## 5 Среда Вопрос: какая доля света данного цвета отражается стеклянной пластиной данной ширины? Предположения: угол падения равен $90^{\circ}$, нет поляризации света; масса частицы теперь зависит от $x$, но не зависит от цвета. Результаты: отражение от тонкой плёнки. Нашу модель также можно применить для описания распространения света в прозрачной среде, например в стекле 9 Свет распространяется так, как будто он имеет ненулевую массу в среде и нулевую вне её 10. Длина волны определяет цвет. 21. (Загадка) Дайте такое определение аналога $\vec{a}(s, m \varepsilon)$ для случая, когда масса $m=m(x)$ зависит от $x$, чтобы аналоги задач 4 и 5 остались верными. Для заданной массы $m=m(x)$ определим вектор $\vec{a}(x, y, m(x) \varepsilon, \lambda / \varepsilon, \pm)$ и число $P(x, y, m(x) \varepsilon, \lambda / \varepsilon, \pm)$ аналогично $\vec{a}(x, y, \lambda / \varepsilon, \pm)$ и $P(x, y, \lambda / \varepsilon, \pm)$, только заменим в определении $\vec{a}(s)$ на $\vec{a}(s, m \varepsilon)$. Пока что будем считать, что $\varepsilon=1$. 22. Отражение от одной поверхности. Найдите $P(x, y, m(x), \lambda,+)$ и $P(x, y, m(x), \lambda,-)$ для $$ m(x)=m_{0}(x) \equiv 0 \quad \text { и } \quad m(x)=m_{1}(x)= \begin{cases}0.2, & \text { при } x=0 \\ 0, & \text { при } x \neq 0\end{cases} $$ Фиксируем нечётное $L>1$, называемое шириной стеклянной пластины. Сначала предположим для простоты, что свет отражается только от двух поверхностей пластины, и для этого рассмотрим 11 $$ m_{2}(x)= \begin{cases}-0.2, & \text { при } x=1 \\ +0.2, & \text { при } x=L \\ 0, & \text { иначе. }\end{cases} $$[^3] Вероятность отражения/прохождения[12] света с длиной волны $\lambda$ от стеклянной пластины ширины $L$ - это соответственно $$ \begin{aligned} & P(\lambda, L,-)=\lim _{\substack{y \rightarrow+\infty \\ y \text { чётно }}} P\left(0, y, m_{2}(x), \lambda,-\right) \\ & P(\lambda, L,+)=\lim _{\substack{y \rightarrow+\infty \\ y \text { чётно }}} P\left(L+1, y, m_{2}(x), \lambda,+\right) \end{aligned} $$ 23. Постройте график функции $f(L)=P\left(0,2 L, m_{2}(x), 16,-\right)$. 24.* (Вторая основная задача) Отражение от двух поверхностей. Найдите $P(\lambda, L,-), P(\lambda, L,+)$, $P(\lambda, L,-)+P(\lambda, L,+)$ и $\max _{L} P(\lambda, L,-)$. На самом деле, свет отражается внутри пластины; этим нельзя пренебрегать, если мы хотим точнее вычислить вероятность отражения (вычислить её точно, без упрощающих предположений, видимо, невозможно для таких сложных веществ, как стекло). Поэтому мы существенно модифицируем нашу модель. Возьмём произвольное $\varepsilon>0$. Зафиксируем $m>0$ и положим $m_{3}(x)=m$ при $0x$ выразите $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ через $P(x, y,+)$ и $P\left(x^{\prime}-x_{0}, y,+\right)$. 33. Принцип запрета (для промежуточных состояний). Докажите, что $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ не изменится, если суммы в определении будут только по тем парам путей $s, s^{\prime}$, у которых нет общих ходов. Определим $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ аналогично для $E=(x \pm 1, y-1), E^{\prime}=\left(x^{\prime} \pm 1, y-1\right)$. 34. Сохранение вероятности/заряда. Для каждого фиксированного $у \geq 1$ выполняется равенство $\sum_{E, E^{\prime}, F, F^{\prime}} P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=1$, где сумма берётся по всем четвёркам $F=(x, y)$, $F^{\prime}=\left(x^{\prime}, y\right), E \stackrel{=}{=}(x \pm 1, y-1), E^{\prime}=\left(x^{\prime} \pm 1, y-1\right)$. У электрона есть античастица, называемая позитрон. Можно представлять себе античастицу как частицу, движущуюся назад во времени. Движение можно описать следующей похожей моделью. Поставим на доску ещё одну шашку, которая сможет двигаться только вправо-вниз и влево-вниз. Будем называть новую шашку чёрной, а старую белой. Для каждой пары путей $s, s^{\prime}$ белой и чёрной шашек соответственно, определим вектор $\vec{a}\left(s, s^{\prime}\right)$ аналогично написанному выше, только для каждого поворота чёрной шашки будем поворачивать вектор против часовой стрелки, а не по (независимо от направления поворота). В частности, $\vec{a}\left(s, s^{\prime}\right)=$ $\left(0,2^{1-y}\right)$, если $s^{\prime}$ - это то же путь, что и $s$, но проходимый в обратном направлении. Обозначим $\square^{18}$ через $$ \vec{a}\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right):=\sum_{s, s^{\prime}} \vec{a}\left(s, s^{\prime}\right) $$ сумму по всем парам, состоящим из пути $s$ белой шашки, начинающимся ходом $A B$ и заканчивающимся ходом $E F$, и пути $s^{\prime}$ чёрной шашки, начинающимся ходом $F^{\prime} E^{\prime}$ и заканчивающимся ходом $B^{\prime} A^{\prime}$. Квадрат длины $P\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right)=\left|\vec{a}\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right)\right|^{2}$ будем называть вероятностью обнаружить электрон в клетке $F$, а позитрон в $F^{\prime}$, если они были испущены из $A$ $u A^{\prime}$. 35. Независимость. Решите аналог задачи 32 для $P\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right)$.[^6] ## 8 Взаимодействие *** Вопрос: тот же, что и в базовой модели, но теперь координата по оси $O$ у интерпретируется, как время. Предположения: электрон создаёт магнитное поле, влияющее на движение; ограждение волнонепроницаемыми стенками. Результаты: взаимодействие со стенками влияет на движение электрона. Мы дошли до неизученной области: физически корректное строгое определение следующего уточнения пока неизвестно. В этом разделе мы даём НЕПРАВИЛЬНОЕ определение, приводящее ## к парадоксальным результатам: - уточнение описывает взаимодействие электрона со стенками, а не с собственным полем; - взаимодействие (со стенками) передаётся с бесконечной скоростью, а не со скоростью света. Но мы надеемся, что предложенные здесь идеи всё равно могут представлять интерес. Движение электрона само по себе создаёт магнитное поле, которое в свою очередь влияет на движение. Созданное поле случайно, с вероятностью, получающейся сложением по всем возможным промежуточным полям. Фиксируем $g \geq 0$, называемое константой взаимодействия 19 (она связана с зарядом электрона). Фиксируем чёрную клетку $(x, y)$, называемую конечным положением. Фиксируем прямоугольник $R$, составленный из всех таких клеток $\left(x^{\prime}, y^{\prime}\right)$, что $1-y0$, называемые массой и шагом решётки. Начальная конфигурация - это расстановк2 ${ }^{22}$ одного из знаков "+" или “-" в некоторых вершинах внутри $R$, лежащих между линиями $y=0$ и $y=1$. С физической точки зрения эти знаки означают начальное положение позитронов и электронов соответственно (а точки без знака свободны). Для нашей игры это означает, что шашка проходит вершины со знаком " -", двигаясь влево-вверх или вправо-вверх, а вершины со знаком "+" - влево-вниз или вправо-вниз (и не проходит через вершины без знака). Аналогично, конечной конфигураиией называется расстановка знаков на вершинах, лежащих между линиями $y=y_{\max }$ и $y=y_{\max }-1$. Промежуточная конфигурация - это расстановка одного из знаков "+" и "-" в некоторых вершинах внутри $R$ так, что в каждой чёрной клетке полосы $1 \leq y \leq y_{\max }-1$ разность между количеством "+" и “-" в 2 верних вершинах равна соответствующей разности в 2 нижних вершинах. Для нашей игры это означает, что шашки начинают и заканчивают движение только на линиях $y=0$ и $y=y_{\max }$. Другими словами, в каждой чёрной клетке знаки в вершинах могут быть расставлены одним из 19 способов 23, показанных в левой части рисунка 6. Эти 19 способов назовём базовыми конфигурациями.[^8]![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-057.jpg?height=648&width=1202&top_left_y=47&top_left_x=427) Рис. 6: Базовые конфигурации Предположим, что каждой базовой конфигурации сопоставили комплексное число (зависящее от $m \varepsilon$ ). "Правильный" способ выбрать эти числа неизвестен; считайте, что они - фиксированные параметры нашей модели. Рассмотрим произвольную промежуточную конфигурацию. В каждую чёрную клетку внутри $R$, не имеющую общих точек с границей, напишем число, соответствующее базовой конфигурации этой клетки. Сопоставим промежуточной конфигурации произведение всех написанных чисел. Теперь для каждой пары из начальной и конечной конфигураций сложим комплексные числа, сопоставленные всем возможным промежуточным конфигурациям между ними. Сопоставим нашей паре полученное комплексное число. 39. (Загадка) Рассмотрим конфигурации совсем без знаков “+”. Возьмём $A, A^{\prime}, B, B^{\prime}, E, E^{\prime}, F, F^{\prime}-$ чёрные клетки, описанные в $\$ 7$. и произвольные $x_{\min }<-y$ и $x_{\max }>x_{0}+y$. Фиксируем начальную и конечную конфигурации с ровно двумя знаками "-", расположенными в правых верхних углах клеток $A, A^{\prime}$ и $E, E^{\prime}$ соответственно. Сопоставьте каждой из 6 базовых конфигураций без знаков "+" комплексное число так, чтобы сумма чисел, сопоставленных всем промежуточным конфигурациям без знаков "+", была равна вектору $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$, определённому в $\$ 7$ Особый интерес представляет случай, когда начальная конфигурация состоит из одного знака “ -" в правой верхней вершине клетки $(0,0)$, а конечная конфигурация - из одного знака " -" в левой нижней или правой нижней вершине чёрной клетки ( $x, y_{\max }$ ). Комплексные числа, соответствующие этим парам, обозначаются соответственно $\vec{a}\left(x, y_{\max }, m \varepsilon, x_{\min }, x_{\max },+\right)$ и $\vec{a}\left(x, y_{\max }, m \varepsilon, x_{\min }, x_{\max },-\right)$. Непрерывный предел, который мы ожидаем получить в этой модели, формулируется в терминах модифицированных функиий Бесселя и функиий Ганкеля: $$ \begin{array}{rlrl} K_{0}(x) & =\int_{1}^{\infty} \frac{e^{-x t}}{\sqrt{t^{2}-1}} d t & H_{0}^{(1)}(x) & =\frac{2}{i \pi} \int_{1}^{\infty} \frac{e^{i x t}}{\sqrt{t^{2}-1}} d t \\ K_{1}(x) & =x \int_{1}^{\infty} e^{-x t} \sqrt{t^{2}-1} d t & H_{1}^{(1)}(x)=-\frac{2 x}{i \pi} \int_{1}^{\infty} e^{i x t} \sqrt{t^{2}-1} d t \end{array} $$ 40. ${ }^{* *}$ Непрерывный предел. Сопоставьте такие комплексные числа 19 базовым конфигурациям, чтобы для всех $|y|<|x|$ было верно $$ \begin{aligned} & \lim _{n \rightarrow \infty} \lim _{x_{\max } \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },-\right)=m K_{0}\left(m \sqrt{x^{2}-y^{2}}\right) \\ & \lim _{n \rightarrow \infty} \lim _{x_{\max } \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },+\right)=-i m \frac{x+y}{\sqrt{x^{2}-y^{2}}} K_{1}\left(m \sqrt{x^{2}-y^{2}}\right) \end{aligned} $$ а для всех $|y|>|x|$ было верно $$ \begin{aligned} & \lim _{n \rightarrow \infty} \lim _{\max \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },-\right)=i m \frac{\pi}{2} H_{0}^{(1)}\left(m \sqrt{y^{2}-x^{2}}\right) \\ & \lim _{n \rightarrow \infty} \lim _{\max \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },+\right)=m \frac{\pi}{2} \cdot \frac{x+y}{\sqrt{y^{2}-x^{2}}} H_{1}^{(1)}\left(m \sqrt{y^{2}-x^{2}}\right) \end{aligned} $$ Обсудим физический смысл этого уточнения. Для $y \gg|x|$ правые части двух последних равенств очень близки к правым частям ответа к задаче 10. Поэтому может возникнуть желание интерпретировать уточнение, как более точное приближение для вероятности обнаружить электрон в клетке $(x, y)$. Но против этого есть серьёзные возражения. Во-первых, приниипиально невозможно точно измерить координаты электрона 24. Эта априорная неопределённость имеет тот же порядок, что и исправление, вносимое уточнением. Таким образом, уточнение не добавляет ничего нового к описанию движения электрона. Во-вторых, для фиксированной начальной конфигурации сумма квадратов модулей чисел, присвоенных всем возможным конечным конфигурациям, не равна 1 (даже в непрерывном пределе). Причиной этого является то, что конфигурации не являются взаимоисключающими: для данной конфигурации поля существует положительная вероятность обнаружить при измерении, что оно находится в другой конфигурации. Мы не знаем ясного объяснения этому. В общем, это уточнение не имеет прямой физической интерпретации, и его следует рассматривать, как составную часть будущих уточнений. ## 10 (1+1)-мерная квантовая электродинамика*** Вопрос: какова вероятность обнаружить электроны (или электрон+позитрон) с импульсами $q$ и $q^{\prime}$ в далёком будущем, если они были испущены с импульсами $p$ и $p^{\prime}$ в далёком прошлом? Предположения: взаимодействие теперь есть; остались только базовые упрощающие предположения: нет ядерных сил и гравитации, электрон движется только по оси $O x$ (а ось $O y$ интерпретируется, как время). Результаты: квантовые поправки. Объединение (пока неизвестных) уточнений, обсуждаемых в двух предыдущих разделах даст элементарное определение $(1+1)$-мерной КЭД. ## Будущие исследования Алгоритмическая квантовая теория поля для каждой экспериментально наблюдаемой величины и положительного числа $\varepsilon$ даёт точную формулировку алгоритма, выдающего предсказанное значение величины с точностью $\varepsilon$. (Разумеется, предсказанное значение не обязано совпадать с экспериментом для $\varepsilon$ меньшего, чем точность самой теории.) Это является продолжением конструктивной квантовой теории поля, которая ещё далека от того, чтобы считаться алгоритмической. ## Эпилог (подводные камни) Мы надеемся, что по крайней мере некоторые из наших читателей заинтересовались элементарными частицами и хотят узнать о них больше. В качестве эпилога дадим несколько предостережений таким читателям. В научно-популярной литературе теорию элементарных частиц часто переупрощают. Наш материал не является исключением. Простые модели, представленные здесь, очень грубы, и слепо доверять им нельзя. Простота - их единственное достоинство; если относиться к ним слишком серьёзно, эти модели даже могут дать неправильную физическую интуицию. Настоящее понимание теории частиц требует серьёзных познаний в физике и математике. Также мы хотим отметить, что сейчас практически нет математических результатов в решёточной квантовой теории поля; почти всё, что имеется - всего лишь компьютерные симуляции. Наконец, есть "теории Новой Физики", которые разрабатываются безо всяких объективных критериев истины: такие теории не поддерживаются ни экспериментальными, ни математическими доказательствами (и некоторые из них имеют экспериментальное опровержение).[^9] ## Указания, ответы, решения Для каждого вектора $\vec{a} \in \mathbb{R}^{2}$ обозначим через $a_{1}, a_{2}$ координаты вектора $\vec{a}$. Например, $\vec{a}(x, y, m \varepsilon,-)=$ $\left(a_{1}(x, y, m \varepsilon,-), a_{2}(x, y, m \varepsilon,-)\right)$. Иногда мы будем смотреть на вектор $\vec{a}$, как на комплексное число $a_{1}+i a_{2}$ (при этом для решение большинства задач не требуется использование комплексных чисел). В дальнейшем, если не сказано иначе, $x$ и $y$ одной чётности. 1. Ответ. Вектор в клетке $(x, y)$ таблицы - это $\vec{a}(x, y)$, а если клетка $(x, y)$ пуста, то $\vec{a}(x, y)=(0,0)$. | 4 | $\left(\frac{1}{2 \sqrt{2}}, 0\right)$ | | $\left(0,-\frac{1}{2 \sqrt{2}}\right)$ | | $\left(\frac{1}{2 \sqrt{2}},-\frac{1}{\sqrt{2}}\right)$ | | $\left(0, \frac{1}{2 \sqrt{2}}\right)$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | $\left(\frac{1}{2}, 0\right)$ | | $\left(\frac{1}{2},-\frac{1}{2}\right)$ | | $\left(0, \frac{1}{2}\right)$ | | | 2 | | | $\left(\frac{1}{\sqrt{2}}, 0\right)$ | | $\left(0, \frac{1}{\sqrt{2}}\right)$ | | | | 1 | | | | $(0,1)$ | | | | | $\mathrm{y} \mathrm{x}$ | -2 | -1 | 0 | 1 | 2 | 3 | 4 | В таблице ниже число в клетке $(x, y)$ равно $P(x, y)$, а если клетка $(x, y)$ пуста, то $P(x, y)=0$. | 4 | $1 / 8$ | | $1 / 8$ | | $5 / 8$ | | $1 / 8$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | $1 / 4$ | | $1 / 2$ | | $1 / 4$ | | | 2 | | | $1 / 2$ | | $1 / 2$ | | | | 1 | | | | 1 | | | | | $\mathrm{y} x \mathrm{x}$ | -2 | -1 | 0 | 1 | 2 | 3 | 4 | (Векторы $\vec{a}(x, y)$ можно легко вычислить последовательно с помощью задачи 4.) См. также следующую таблицу. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-059.jpg?height=648&width=948&top_left_y=1184&top_left_x=151) Для всех натуральных $y$ верно $\sum_{x \in \mathbb{Z}} P(x, y)=1$. (Это требуется доказать в задаче 5 ) Для $-4 \leq-yy$ и $x<-y+1$ верно $P(x, y)=0$. (В общем случае неизвестно, возможно ли равенство $P(x, y)=0$ для каких-то $x, y$ одной чётности при $-y+1 \leq x \leq y$.) Для всех нечётных $y$ верно $\vec{a}(0, y)=0$. Для всех $n \in \mathbb{N} \cup\{0\}$ верно: $$ \begin{aligned} & \text { вектор } \vec{a}(0,8 n+2) \text { направлен вправо, } \\ & \text { вектор } \vec{a}(0,8 n+4) \text { направлен вниз, } \\ & \text { вектор } \vec{a}(0,8 n+6) \text { направлен влево, } \\ & \text { вектор } \vec{a}(0,8 n+8) \text { направлен вверх. } \end{aligned} $$ Аналогично, для всех чётных $y$ верно $\vec{a}(1, y)=0$. Для всех $n \in \mathbb{N} \cup\{0\}$ верно: $$ \begin{aligned} & \text { вектор } \vec{a}(1,8 n+3) \text { направлен вправо-вниз, } \\ & \text { вектор } \vec{a}(1,8 n+5) \text { направлен влево-вниз, } \\ & \text { вектор } \vec{a}(1,8 n+7) \text { направлен влево-вверх, } \\ & \text { вектор } \vec{a}(1,8 n+9) \text { направлен вправо-вверх. } \end{aligned} $$ (См. замечание после ответа к задаче 9.) Замечание. В таблице на пересечении $k$-ого столбца и $n$-ой строки стоит $2^{(n+k-2) / 2} \vec{a}(k-n+1, n+k-1)$ : | $(0,1)$ | $(0,1)$ | $(0,1)$ | $(0,1)$ | $(0,1)$ | $(0,1)$ | $(0,1)$ | $(0,1)$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $(1,0)$ | $(1,-1)$ | $(1,-2)$ | $(1,-3)$ | $(1,-4)$ | $(1,-5)$ | $(1,-6)$ | $(1,-7)$ | | $(1,0)$ | $(0,-1)$ | $(-1,-1)$ | $(-2,0)$ | $(-3,2)$ | $(-4,5)$ | $(-5,9)$ | $(-6,14)$ | | $(1,0)$ | $(-1,-1)$ | $(-2,0)$ | $(-2,2)$ | $(-1,4)$ | $(1,5)$ | $(4,4)$ | $(8,0)$ | | $(1,0)$ | $(-2,-1)$ | $(-2,1)$ | $(0,3)$ | $(3,3)$ | $(6,0)$ | $(8,-6)$ | $(8,-14)$ | | $(1,0)$ | $(-3,-1)$ | $(-1,2)$ | $(3,3)$ | $(6,0)$ | $(6,-6)$ | $(2,-12)$ | $(-6,-14)$ | | $(1,0)$ | $(-4,-1)$ | $(1,3)$ | $(6,2)$ | $(6,-4)$ | $(0,-10)$ | $(-10,-10)$ | $(-20,0)$ | | $(1,0)$ | $(-5,-1)$ | $(4,4)$ | $(8,0)$ | $(2,-8)$ | $(-10,-10)$ | $(-20,0)$ | $(-20,20)$ | 2. Ответ: нет на оба вопроса. Например, $P(2,4)=5 / 8 \neq 1 / 8+1 / 4=P(2,4$ минуя 2,2$)+P(2,4$ минуя 0,2$)$ и $P(0,4)=1 / 8<1 / 4=P(0,4$ минуя 2,2$)$. 3. Ответ: $P(0,12)=25 / 512$. Подсказка: ответ можно получит сразу с помощью замечания после ответа к задаче 9, или его можно быстро вычислить рекурсивно с помощью ответа к задаче 4 . 4. Oтвет: $\left\{\begin{array}{l}a_{1}(x, y)=\frac{1}{\sqrt{2}} a_{2}(x+1, y-1)+\frac{1}{\sqrt{2}} a_{1}(x+1, y-1) ; \\ a_{2}(x, y)=\frac{1}{\sqrt{2}} a_{2}(x-1, y-1)-\frac{1}{\sqrt{2}} a_{1}(x-1, y-1) .\end{array}\right.$ Решение. Выведем формулу для $a_{2}(x, y)$; формула для $a_{1}(x, y)$ выводится аналогично. Рассмотрим произвольный путь $s$ из $(0,0)$ в $(x, y)$. Обозначим через $t(s)$ количество поворотов у $s$. Обозначим $\vec{a}(s)=$ $\left(a_{1}(s), a_{2}(s)\right)$. Заметим, что $a_{2}(s) \neq 0$ тогда и только тогда, когда $t(s)$ чётно, и $a_{1}(s) \neq 0$ тогда и только тогда, когда $t(s)$ нечётно. Таким образом, $a_{2}(x, y)=\sum_{s: t(s) \text { чётно }} a_{2}(s)$ и $a_{1}(x, y)=\sum_{s: t(s) \text { нечётно }} a_{2}(s)$. Последний ход пути $s$ сделан либо из $(x-1, y)$, либо из $(x+1, y)$. Очевидно, что если $t(s)$ чётно, то последний ход направлен вправо-вверх, а если нечётно - то влево-вверх. Так как нас интересует $a_{2}(x, y)$, то будем предполагать, что последний ход сделан вправо-вверх. Обозначим через $s^{\prime}$ путь $s$ без последнего хода. Если направления последних ходов $s$ и $s^{\prime}$ совпадают, то $\vec{a}(s)=\frac{1}{\sqrt{2}} \vec{a}\left(s^{\prime}\right)$, иначе $\vec{a}(s)=\frac{1}{\sqrt{2}}\left(a_{2}\left(s^{\prime}\right),-a_{1}\left(s^{\prime}\right)\right)$. Таким образом $$ a_{2}(x, y)=\sum_{s: t(s) \text { even }} a_{2}(s)=\frac{1}{\sqrt{2}}\left(\sum_{s^{\prime}: t\left(s^{\prime}\right) \text { even }} a_{2}\left(s^{\prime}\right)-\sum_{s^{\prime}: t\left(s^{\prime}\right) \text { odd }} a_{1}\left(s^{\prime}\right)\right)=\frac{1}{\sqrt{2}}\left(a_{2}(x-1, y-1)-a_{1}(x-1, y-1)\right) $$ 5. Докажем индукцией по $y$, что $\sum_{x \in \mathbb{Z}} P(x, y)=1$ для всех $y>0$. Очевидно, $\sum_{x \in \mathbb{Z}} P(x, 1)=1$. Шаг индукции непосредственно следует из следующих вычислений: $$ \begin{gathered} \sum_{x \in \mathbb{Z}} P(x, y+1)=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y+1)^{2}+a_{2}(x, y+1)^{2}\right]=\sum_{x \in \mathbb{Z}} a_{1}(x, y+1)^{2}+\sum_{x \in \mathbb{Z}} a_{2}(x, y+1)^{2}= \\ =\frac{1}{2} \sum_{x \in \mathbb{Z}}\left(a_{1}(x+1, y)+a_{2}(x+1, y)\right)^{2}+\frac{1}{2} \sum_{x \in \mathbb{Z}}\left(a_{2}(x-1, y)-a_{1}(x-1, y)\right)^{2}= \\ =\frac{1}{2} \sum_{x \in \mathbb{Z}}\left(a_{1}(x, y)+a_{2}(x, y)\right)^{2}+\frac{1}{2} \sum_{x \in \mathbb{Z}}\left(a_{2}(x, y)-a_{1}(x, y)\right)^{2}=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y)^{2}+a_{2}(x, y)^{2}\right]=\sum_{x \in \mathbb{Z}} P(x, y) \end{gathered} $$ ## Обобщение закона сохранения, найденное Глебом Минаевым и Иваном Русских. Сделаем замену координат $(x, y) \mapsto(t, u)=\left(\frac{x+y}{2}-1, \frac{y-x}{2}\right)$, то есть повернём картинку на $45^{\circ}$ по часовой стрелке и сдвинем на 1 по горизонтали (чтобы не учитывать первый шаг, который всегда один и тот же). Обозначим $\vec{b}(t, u):=\vec{a}(t-u+1, t+u+1), Q(t, u):=P(t-u+1, t+u+1), \vec{B}(t, u):=\left(1+m^{2} \varepsilon^{2}\right)^{(t+u) / 2} \vec{b}(t, u)$. Координаты векторов $\vec{b}(t, u)$ и $\vec{B}(t, u)$ будем обозначать через $b_{1}(t, u), b_{2}(t, u), B_{1}(t, u), B_{2}(t, u)$. Заметим, что в новой системе координат мы сдвигаем шашку не по диагонали между чёрными клетками доски, а в соседнюю точку на целочисленной решётке. Тогда можно считать, что мы начинаем в $(0,0)$ и можем попасть в любую точку $(\mathbb{N} \cup\{0\})^{2}$, а «нулевой шаг» из $(-1,0)$ в $(0,0)$ учитывается лишь при подсчёте количества поворотов. Для подмножества точек нашей решётки $M \subset(\mathbb{N} \cup\{0\})^{2}$ определим $\vec{b}(t, u$ минуя $M):=\sum_{s} \vec{b}(s)$, где суммирование ведётся по всем путям $s$ из $(0,0)$ в $(t, u)$, не проходящим через точки множества $M$. Аналогичным образом определим $Q(t, u$ минуя $M), \vec{b}(t, u$ минуя $M, \pm), Q(t, u$ минуя $M, \pm)$. Также для множества $M$ определим $$ Q(M):=\sum_{p \in M} Q(p \text { минуя } M \backslash\{p\}) \text {. } $$ Следует отметить, что для бесконечного множества $M$ сумма становится бесконечной. Порядок суммирования для бесконечной суммы в данном случае не важен, поскольку все слагаемые положительны. Теорема 1. Пусть дано такое конечное мно»ество $M \subset(\mathbb{N} \cup\{0\})^{2}$, что не существует бесконечного пути шашки из $(0 ; 0)$, не проходящего через точки множества $M$. Тогда $Q(M)=1$. Доказательство. Докажем утверждение индукцией по $m:=\max _{(t, u) \in M}(t+u)$, то есть по максимальному номеру диагонали, идущей вправо-вниз, на которой лежит хотя бы одна точка множества $M$ (будем называть эту диагональ максималъной). База: $m=0$. В этом случае $M=(0,0)$, а $Q(M)=Q(0,0)=1$. Шаг. Докажем лемму: Лемма 1. Пусть даны точка $(t, u) \in(\mathbb{N} \cap\{0\})^{2}$ и множсество $A \subset(\mathbb{N} \cap\{0\})^{2}$, не содержащее точек $(t, u)$, $(t, u+1) u(t+1, u)$. Тогда $$ Q(t, \text { минуя } A)=Q(t, u+1 \text { минуя } A,-)+Q(t+1, \text { и минуя } A,+) $$ ## Доказательство. $$ \begin{aligned} & Q(t, u+1 \text { минуя } A,-)+Q(t+1, u \text { минуя } A,+)=b_{1}(t, u+1 \text { минуя } A)^{2}+b_{2}(t+1, u \text { минуя } A)^{2}= \\ & \begin{aligned} =\frac{1}{2}\left(\left(b_{1}(t, u \text { минуя } A)\right.\right. & \left.\left.+b_{2}(t, u \text { минуя } A)\right)^{2}+\left(b_{2}(t, u \text { минуя } A)-b_{1}(t, u \text { минуя } A)\right)^{2}\right)= \\ = & b_{1}(t, u \text { минуя } A)^{2}+b_{2}(t, u \text { минуя } A)^{2}=Q(t, u \text { минуя } A) \end{aligned} \end{aligned} $$ Рассмотрим какую-нибудь точку $(t, u) \in M$ с максимальным $t+u$. Пусть в $(t, u)$ снизу приходит какойлибо путь шашки из $(0,0)$, не проходящий через $M \backslash\{(t, u)\}$. Тогда есть путь в $(t, u-1)$, не проходящий через $M \backslash\{(t, u-1)\}$. Заметим, что тогда $(t, u-1) \notin M$, так как путь в $(t ; u)$ приходит снизу. Значит, есть путь и в $(t+1, u-1)$, не проходящий через $M \backslash\{(t+1, u-1)\}$. Если $(t+1, u-1) \notin M$, то существует путь, приходящий в эту точку и уходящий на бесконечность, при этом не проходящий через точки множества $M$, например, путь, который после точки $(t, u)$ идёт только вправо (он действительно не проходит через точки множества $M$, так как выбранная диагональ максимальна). Значит, $(t+1, u-1) \in M$. Заметим, что $Q(a, b$ минуя $A)=Q(a, b$ минуя $A,+)+Q(a, b$ минуя $A,-)$ для любого множества $A \subset(\mathbb{N} \cap\{0\})^{2}$. Пусть $K:=M \backslash\{(t, u) ;(t+1, u-1)\}$. Тогда можно записать следующую цепочку равенств для множества $A=K$ (в ней используется лемма 11 : $$ \begin{aligned} Q(M)= & Q(K)+Q(t, u \text { минуя } M \backslash\{(t, u)\})+Q(t+1, u-1 \text { минуя } M \backslash\{(t+1, u-1)\})= \\ = & Q(K)+Q(t, u \text { минуя } M \backslash\{(t, u)\},+)+Q(t, u \text { минуя } M \backslash\{(t, u)\},-)+ \\ & +Q(t+1, u-1 \text { минуя } M \backslash\{(t+1, u-1)\},+)+Q(t+1, u-1 \text { минуя } M \backslash\{(t+1, u-1)\},-)= \\ = & Q(K)+Q(t, u \text { минуя } M \backslash\{(t, u)\},+)+ \\ & +Q(t, u-1 \text { минуя } M \backslash\{(t, u-1)\})+Q(t+1, u-1 \text { минуя } M \backslash\{(t+1, u-1)\},-)= \\ = & Q(M \cup\{(t, u-1)\}) . \end{aligned} $$ Поэтому, если мы добавим точку $(t, u-1)$ в множество $M$, вероятность $Q(M)$ не изменится. Аналогичное верно и для путей, приходящих в $(t, u)$ слева. Заметим, что каждой операцией мы ставим точки на диагональ $t+u=m-1$, а значит, максимальная диагональ, занятая точками множества $M$, не менялась. Тем самым, добавив несколько точек в множество $M$, мы сделали так, чтобы в точки из множества $M$, лежащие на максимальной диагонали, больше не приходило путей из $(0,0)$, не проходящих через другие точки множества $M$. Следовательно, при удалении всех точек с максимальной диагонали не меняется $Q(M)$ и не появляется бесконечного пути, не проходящего через точки $M$. Заметим, что мы изменили $M$ таким образом, что $Q(M)$ не изменилось, а номер максимальной занятой диагонали уменьшился. По предположению индукции новое $Q(M)$ равно 1, значит, и старое равнялось 1 . 6. Ответ. Для всех $x$ верно $a_{1}(x, y)=a_{1}(-x, y)$ и $a_{2}(x, y)+a_{1}(x, y)=a_{2}(2-x, y)+a_{1}(2-x, y)$. Решение. Для каждого пути $s$ обозначим через $f(s)$ отражение $s$ относительно оси $O y$. Тогда если $s$ ведёт в $(x, y)$, то $f(s)$ ведёт в $(-x, y)$. Для каждого пути $s$ обозначим через $g(s)$ путь, состоящий из тех же ходов, что $s$, но сделанных в обратном порядке. Заметим, что изменение порядка ходов не влияет на то, где путь закончится. Теперь рассмотрим такой путь $s$ в $(x, y)$, что $t(s)$ нечётно. Последний ход $s$ направлен влево-вверх. Таким образом, последний ход $f(s)$ направлен вправо-вверх, значит первый ход $g(f(s))$ направлен вправо-вверх. Таким образом, $f \circ g$ - это биекция между путями в $(x, y)$ с нечётным числом поворотов, начинающихся ходом вправо-вверх, и путями в $(-x, y)$ с нечётным числом поворотов, начинающихся ходом вправо-вверх. Второе равенство можно доказать с помощью результата из задачи 4 : $$ a_{1}(x, y)+a_{2}(x, y)=\sqrt{2} \cdot a_{1}(x-1, y+1)=\sqrt{2} \cdot a_{1}(1-x, y+1)=a_{1}(2-x, y)+a_{2}(2-x, y) $$ Замечание. Эти равенства также можно доказать индукцией по $y$, используя задачу 4 Также из формулы для $a_{2}(x, y)$ из задачи 9 следует, что для всех $x, y$ верно: $(y-x) a_{2}(x, y)=(y+x-2) a_{2}(2-x, y)$. Если вы найдёте комбинаторное доказательство последней формулы, то, пожалуйста, сообщите нам. 7. Ответ: для всех $0|x|$ верно: $$ \begin{aligned} & a_{1}(x, y)=2^{(1-y) / 2} \sum_{w \equiv 2 \frac{y-x-2}{2}}(-1)^{\frac{y-x-2+2 w \cdot \operatorname{sign}(x)}{4}}\binom{|x|}{w}\binom{\frac{y-|x|-2}{2}}{\frac{y-x-2-2 w}{4}} \\ & a_{2}(x, y)=2^{(1-y) / 2} \sum_{w \equiv_{2} \frac{y-x}{2}}(-1)^{\frac{y-x+2 w \cdot \operatorname{sign}(x)}{4}}\binom{|x|}{w}\binom{\frac{y-|x|-2}{2}}{\frac{y-x-2 w}{4}} \end{aligned} $$ для всех $y=x>0$ верно: $$ \begin{aligned} & a_{1}(y, y)=0 \\ & a_{2}(y, y)=2^{(1-y) / 2} \end{aligned} $$ для всех $00$ верно: $$ \begin{aligned} & a_{1}(x, y)=0 \\ & a_{2}(x, y)=0 \end{aligned} $$ Oтвет 2: для всех $y>|x|$ верно: $$ \begin{aligned} & a_{1}(x, y)=2^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r}= \\ & a_{2}(x, y)=2^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r-1} \end{aligned} $$ Для прочих $y$ всё аналогично ответу 1. Ответ 3. При чётном $p+q$ и $2-q \leq p \leq q$ выполняется ${ }^{25}$ $$ \begin{aligned} & a_{1}(p, q)=(-1)^{q} 2^{(q-3) / 2} \cdot \sum_{k=\frac{q+|p|-2}{2}}^{q-2}(-2)^{-k} \cdot \frac{k!}{(q-2-k)!\left(\frac{p-q}{2}+1+k\right)!\left(k-\frac{p+q}{2}+1\right)!} \\ & a_{2}(p, q)=2^{(1-q) / 2} \cdot\binom{q-1}{\frac{q-p}{2}}-(-1)^{q} 2^{(q-3) / 2} \sum_{k=\frac{q+|p|-2}{2}}^{q-2}(-2)^{-k} \cdot \frac{k!(k+p+1)}{(q-1-k)!\left(\frac{p-q}{2}+1+k\right)!\left(k-\frac{p+q}{2}+1\right)!} \end{aligned} $$ ${ }^{25}$ Для знатоков: дроби в этих выражениях равны соответственно $$ \binom{k}{q-2-k, \frac{p-q}{2}+1+k, k-\frac{p+q}{2}+1} \text { и }\binom{k}{q-2-k, \frac{p-q}{2}+1+k, k-\frac{p+q}{2}+1} \cdot \frac{k+p+1}{q-1-k} . $$ Замечание. Из ответа 2 видно, что для всех $1 \leq k \leq y-1$ числа $a_{1}(-y+2 k, y)$ и $a_{2}(-y+2 k, y)$ равны коэффициентам при $t^{y-k-1}$ и $t^{y-k}$ в разложении $2^{(1-y) / 2}(1+t)^{y-k-1}(1-t)^{k-1}$. В частности, $$ \begin{aligned} & a_{1}(0,4 n+2)=\frac{(-1)^{n}}{2^{(4 n+1) / 2}}\binom{2 n}{n}, \quad a_{1}(0,4 n)=0, \\ & a_{2}(0,4 n+2)=0 \\ & \text { и } \quad a_{2}(0,4 n)=\frac{(-1)^{n}}{2^{(4 n-1) / 2}}\binom{2 n-1}{n} . \end{aligned} $$ Суммы в формулах для $a_{1}(x, y)$ и $a_{2}(x, y)$ для $y>|x|$ являются частными случаями хорошо известных гипергеометрических функиий (но знакомство с ними не требуется для дальнейшего): $$ \begin{aligned} & a_{1}(x, y)=2^{(1-y) / 2}{ }_{2} F_{1}\left(1-\frac{x+y}{2}, 1+\frac{x-y}{2}, 1 ;-1\right) \\ & a_{2}(x, y)=2^{(1-y) / 2}\left(1-\frac{x+y}{2}\right){ }_{2} F_{1}\left(2-\frac{x+y}{2}, 1+\frac{x-y}{2}, 2 ;-1\right) \end{aligned} $$ Указание. Сначала докажем формулы из ответа 2. Вычислим $a_{1}(x, y)$ для $y>|x|$. Нам достаточно рассматривать только пути с нечётным числом поворотов (остальные не влияют на $a_{1}(x, y)$ ). Обозначим через $R$ и $L$ количество ходов вправо-вверх и влево-вверх соответственно. Эти числа однозначно определяются условиями $L+R=y$ и $R-L=x$. Обозначим через $2 r+1$ количество поворотов в пути. Обозначим через $x_{1}, x_{2}, \ldots, x_{r+1}$ количество ходов вправо-вверх перед первым, третьим, ..., последним поворотом соответственно. Обозначим через $y_{1}, y_{2}, \ldots, y_{r+1}$ количество ходов влево-вверх после первого, третьего, ..., последнего поворота соответственно. Тогда $x_{k}, y_{k} \geq 1$ для всех $1 \leq k \leq r+1$ и $$ \begin{aligned} & R=x_{1}+\cdots+x_{r+1} \\ & L=y_{1}+\cdots+y_{r+1} \end{aligned} $$ С другой стороны, каждый набор натуральных чисел ( $\left.x_{1}, x_{2}, \ldots, x_{r+1}, y_{1}, y_{2}, \ldots, y_{r+1}\right)$, удовлетворяющий полученным уравнениям, однозначно определяет путь из $(0,0)$ в $(x, y)$ с нечётным числом поворотов. Мы свели задачу к комбинаторной: найти количество натуральных решений двух полученных уравнений. Для первого уравнения это количество равно количеству способов поставить $r$ перегородок между $R$ шарами, то есть, $\binom{R-1}{r}$. Тогда $$ a_{1}(x, y)=2^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{R-1}{r}\binom{L-1}{r} $$ Так как $L+R=y$ и $R-L=x$, мы вывели искомую формулу из ответа 2 . Формулу для $a_{1}(x, y)$ из ответа 1 можно вывести, пользуясь многочленом из замечания после ответа. В зависимости от знака $x$ мы получаем: $$ 2^{(1-y) / 2}(1+t)^{y-k-1}(1-t)^{k-1}=\left\{\begin{array}{l} 2^{(1-y) / 2}\left(1-t^{2}\right)^{\frac{y-x-2}{2-x}}(1-t)^{x}, \text { при } x \geq 0 \\ 2^{(1-y) / 2}\left(1-t^{2}\right)^{\frac{y+2}{2}}(1+t)^{-x}, \text { при } x<0 \end{array}\right. $$ Для всех $|x||x|$ аналогично. Случай $y \leq|x|$ очевиден. Решение Александра Кудрявчева, участника Летней конференции Турнира городов. Докажем ответ 3. Обозначим $c_{p, q}=a_{1}(2 p-q+1, q+1) \cdot 2^{q / 2}, d_{p, q}=a_{2}(2 p-q+1, q+1) \cdot 2^{q / 2}$ (они определены корректно при $0 \leq p \leq q-1)$. Обозначим через $C(x, y), D(x, y)$ производящие функции для $c_{p, q}$ и $d_{p, q}$. Из ответа к задаче 4 получаем следующие уравнения: $$ \left\{\begin{array}{l} \frac{C(x, y)-C(x, 0)}{y}=D(x, y)+C(x, y) \\ \frac{D(x, y)-D(x, 0)}{y}=x \cdot(D(x, y)-C(x, y)) \end{array}\right. $$ Так как $C(x, 0)=0$ и $D(x, 0)=1$, то $$ \left\{\begin{array}{l} C(x, y)=y \cdot(D(x, y)+C(x, y)) \\ D(x, y)-1=x y \cdot(D(x, y)-C(x, y)) \end{array}\right. $$ Решив систему, получаем $C(x, y)=\frac{y}{1-y-x y+2 x y^{2}}$ и $D(x, y)=\frac{1-y}{1-y-x y+2 x y^{2}}$. Тогда $$ \frac{C(x, y)}{y}=\frac{D(x, y)}{1-y}=1+y(1+x-2 x y)+y^{2}(1+x-2 x y)^{2}+\ldots $$ Обозначим это выражение через $E(x, y)$ и запишем его коэффициент при $x^{n} \cdot y^{m}$ (при $n \leq m$, иначе он равен 0$)$ : $$ \sum_{k=\max (n, m-n)}^{m}(-2)^{m-k} \cdot \frac{k!}{(m-k)!(n-m+k)!(k-n)!} $$ действительно, из каждого слагаемого вида $y^{k}(1+x-2 x y)^{k}$ можно взять ровно одну комбинацию множителей: - чтобы степень при $y$ была равна $m$, множителей $-2 x y$ должно быть $m-k$; - чтобы степень при $x$ была равна $n$, множителей $x$ должно быть $n-(m-k)$; - чтобы всего множителей было $k$, множителей 1 должно быть $k-(n-(m-k))-(m-k)=k-n$. Количество же способов выбрать $m-k, k-n$ и $n-(m-k)$ скобок равно дроби справа. Теперь из соотношений $C(x, y)=y \cdot E(x, y), D(x, y)=E(x, y)-C(x, y)$ несложно вывести ответ 3 . 10. Oтвem: $$ \vec{a}(x, y)=\frac{\sqrt{2}\left(\sin \left(\frac{\pi y}{4}-\frac{x^{2}}{2 y}\right), e^{x / y} \cos \left(\frac{\pi y}{4}+\frac{x(2-x)}{2 y}\right)\right)}{\sqrt{\pi y}}+\mathrm{O}\left(\frac{\log ^{2} y}{y}\right) \text { при }|x|=\mathrm{O}(\sqrt{y}) $$ Замечание. Сравните графики $P(x, 400), a_{1}(x, 400)$ и $a_{2}(x, 400)$ (синие) с их приближениями в правой части (оранжевые) для $y=400$ : ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-066.jpg?height=412&width=1882&top_left_y=1250&top_left_x=150) Решение: Заметим, что по задаче 9 $$ \begin{aligned} & a_{2}(-n+2 k+1, n+1)=2^{-n / 2}\left[x^{k-1}\right](-1)^{k}(1+x)^{n-k-1}(1-x)^{k} \\ & a_{1}(-n+2 k+1, n+1)=2^{-n / 2}\left[x^{k}\right](-1)^{k}(1+x)^{n-k-1}(1-x)^{k} \end{aligned} $$ Поэтому $$ a_{2}(-n+2 k+1, n+1)=2^{-n / 2}(-1)^{k} \int_{0}^{1}\left(1+e^{2 \pi i t}\right)^{n-k-1}\left(1-e^{2 \pi i t}\right)^{k} e^{-2 \pi i(k-1) t} d t $$ Для нечетного $n$ определим $l=k-\frac{n-1}{2}$. Тогда $$ a_{2}(-n+2 k+1, n+1)=\frac{i^{k}}{\sqrt{2}} \int_{0}^{1}(\sin (2 \pi t))^{\frac{n-1}{2}}(\operatorname{tg}(\pi t))^{l} e^{-2 \pi i(l-1) t} d t $$ Поскольку $l<0$ верно: $$ S_{1}(y)+S_{2}(y)=1 $$ Из ответов к задачам 4,6 и 7 следует: $$ a_{1}(0,2 y)=\frac{1}{\sqrt{2}} \sum_{x} a_{1}(x, y)\left(a_{2}(x, y)-a_{1}(x, y)\right)+a_{2}(x, y)\left(a_{2}(x, y)+a_{1}(x, y)\right)=\frac{1}{\sqrt{2}}\left(S_{2}(y)+2 S_{12}(y)-S_{1}(y)\right) $$ По определению и ответу к задаче 4 получаем: $$ S_{1}(y+1)-S_{2}(y+1)=2 S_{12}(y) $$ Тогда $$ S_{1}(y+1)-S_{2}(y+1)=S_{1}(y)-S_{2}(y)+a_{1}(0,2 y) \sqrt{2} $$ Так как $S_{1}(y)+S_{2}(y)=1$, мы получили рекуррентную формулу $S_{1}(y+1)=S_{1}(y)+\frac{1}{\sqrt{2}} a_{1}(0,2 y)$. Из замечания к решению задачи 9 по индукции следует, что $$ S_{1}(y)=\frac{1}{2} \sum_{k=0}^{\lfloor y / 2\rfloor-1} \frac{1}{(-4)^{k}}\binom{2 k}{k} $$ Осталось заметить, что $P(y,-)=S_{1}(y)$ по задаче 11 . Так как последовательность $\binom{2 j}{j} \frac{1}{4^{j}}$ убывающая, мы получаем: $P(y,-)3$, так что $\max _{y} P(y,-)=P(2,-)=P(3,-)=1 / 2$. По биному Ньютона мы получаем: $\sum_{k=0}^{\infty}\binom{2 k}{k} x^{k}=\frac{1}{\sqrt{1-4 x}}$ для всех $x \in\left[-\frac{1}{4}, \frac{1}{4}\right)$. Подставляя $x=-\frac{1}{4}$, мы получаем $\lim _{y \rightarrow \infty} P(y,-)=\lim _{y \rightarrow \infty} \frac{1}{2} \sum_{k=0}^{\infty}\binom{2 k}{k}\left(-\frac{1}{4}\right)^{k}=\frac{1}{2 \sqrt{2}}$. Замечание. С помощью формулы Стирлинга (см. \$22 можно оценить скорость сходимости: $$ \left|P(k,-)-\lim _{y \rightarrow \infty} P(y,-)\right|<\frac{1}{2 \cdot 4\lfloor k / 2\rfloor}\binom{2\lfloor k / 2\rfloor}{\lfloor k / 2\rfloor} \leq \frac{e}{2 \pi \sqrt{2\lfloor k / 2\rfloor}}<\frac{1}{2 \sqrt{k}} $$ ## 13. Oтвет: $$ P(x, y, 0,+)= \begin{cases}1, & \text { при } x=y \\ 0, & \text { при } x \neq y\end{cases} $$ $$ P(x, y, \infty,+)= \begin{cases}1, & \text { при } x=1, y \equiv 1 \quad \bmod 2 \\ 0, & \text { иначе }\end{cases} $$ Решение. Если количество поворотов ну пути $s$ ненулевое, то по определению $\vec{a}(s, 0)=0$. Тогда $\vec{a}(x, y, 0,+) \neq$ 0 если и только если $x=y$. Случай $x=y$ очевиден. Ответ для $P(x, y, \infty,+)$ нельзя доказать, поскольку он является определением. Но мы можем объяснить, почему было выбрано именно такое определение. Если $m \varepsilon$ 'очень большой', то $\vec{a}(s, m \varepsilon)$ будет 'маленьким', если только шашка не поворачивает после каждого хода. Замечание. $P(x, y, \infty,+)=\lim _{m \varepsilon \rightarrow \infty} P(x, y, m \varepsilon,+)$. В самом деле, $$ \lim _{m \varepsilon \rightarrow \infty}|\vec{a}(s, m \varepsilon)|=\lim _{m \varepsilon \rightarrow \infty} \frac{(m \varepsilon)^{t(s)}}{\left(\sqrt{1+(m \varepsilon)^{2}}\right)^{y-1}}=\lim _{m \varepsilon \rightarrow \infty} \frac{1}{\left(\sqrt{1+(m \varepsilon)^{2}}\right)^{y-1-t(s)}}= \begin{cases}1, & \text { при } t(s)=y-1 \\ 0, & \text { иначе }\end{cases} $$ где $t(s)$ равно количеству поворотов на пути $s$. Для вычисления $P(x, y, m \varepsilon,+)$ достаточно рассматривать только пути $s$ с $t(s) \equiv 0 \bmod 2$, откуда мы и получаем ответ. 14. Ответ. Обозначим через $a_{1}(x, y, m \varepsilon,-), a_{2}(x, y, m \varepsilon,-)$ и $a_{1}(x, y, m \varepsilon,+), a_{2}(x, y, m \varepsilon,+)$ первую и вторую координаты векторов $\vec{a}(x, y, m \varepsilon,-)$ и $\vec{a}(x, y, m \varepsilon,+$ ) соответственно (см. соглашение в начале раздела «Указания, ответы, решения»). Тогда: $$ \begin{aligned} & a_{1}(x, y, m \varepsilon,-)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{1}(x+1, y-1, m \varepsilon,-)+m \varepsilon a_{2}(x+1, y-1, m \varepsilon,+)\right) \\ & a_{2}(x, y, m \varepsilon,+)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{2}(x-1, y-1, m \varepsilon,+)-m \varepsilon a_{1}(x-1, y-1, m \varepsilon,-)\right) \\ & a_{2}(x, y, m \varepsilon,-)=a_{1}(x, y, m \varepsilon,+)=0 \end{aligned} $$ Для всех $y>|x|$ верно: $$ \begin{aligned} & a_{1}(x, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r}(m \varepsilon)^{2 r+1} \\ & a_{2}(x, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r-1}(m \varepsilon)^{2 r} \end{aligned} $$ Для всех $y=x>0$ верно: $$ \begin{aligned} & a_{1}(y, y, m \varepsilon)=0 \\ & a_{2}(y, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \end{aligned} $$ Для всех $00$ верно $$ \begin{aligned} & a_{1}(x, y, m \varepsilon)=0 \\ & a_{2}(x, y, m \varepsilon)=0 \end{aligned} $$ Решение. Сначала решим аналог задачи 4 для произвольного $m \varepsilon$. Выведем формулу для $a_{2}(x, y, m \varepsilon,+)$. Рассмотрим произвольный путь $s$ из $(0,0)$ в $(x, y)$. Последний ход пути $s$ сделан либо из $(x-1, y)$, либо из $(x+1, y)$. Обозначим через $s^{\prime}$ путь $s$ без последнего хода. Если направления последних ходов у $s$ и $s^{\prime}$ совпадают, то $\vec{a}(s, m \varepsilon)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}} \vec{a}\left(s^{\prime}, m \varepsilon\right)$, иначе $\vec{a}(s, m \varepsilon)=$ $\frac{m \varepsilon}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{2}\left(s^{\prime}, m \varepsilon\right),-a_{1}\left(s^{\prime}, m \varepsilon\right)\right)$. Таким образом, $a_{2}(x, y, m \varepsilon,+)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{2}(x-1, y-1, m \varepsilon,+)-m \varepsilon a_{1}(x-1, y-1, m \varepsilon,-)\right)$. Формула для $a_{1}(x, y, m \varepsilon,-)$ доказывается аналогично. То, что $a_{2}(x, y, m \varepsilon,-)=a_{1}(x, y, m \varepsilon,+)=0$, очевидно. Теперь решим аналог задачи 5 мы докажем, что $\sum_{x \in \mathbb{Z}} P(x, y, m \varepsilon)=1$ для всех $y \geq 1$ индукцией по $y$, используя полученные формулы. Очевидно, $\sum_{x \in \mathbb{Z}} P(x, 1, m \varepsilon)=1$. Шаг индукции сразу следует из следующих вычислений: $$ \begin{aligned} & \sum_{x \in \mathbb{Z}} P(x, y+1, m \varepsilon)= \\ &=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y+1, m \varepsilon,-)^{2}+a_{2}(x, y+1, m \varepsilon,+)^{2}\right]= \\ &= \sum_{x \in \mathbb{Z}} a_{1}(x, y+1, m \varepsilon,-)^{2}+\sum_{x \in \mathbb{Z}} a_{2}(x, y+1, m \varepsilon,+)^{2}= \\ &= \frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{1}(x+1, y, m \varepsilon,-)+m \varepsilon a_{2}(x+1, y, m \varepsilon,+)\right)^{2}+ \\ &+\frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{2}(x-1, y, m \varepsilon,+)-m \varepsilon a_{1}(x-1, y, m \varepsilon,-)\right)^{2}= \\ &=\frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{1}(x, y, m \varepsilon,-)+m \varepsilon a_{2}(x, y, m \varepsilon,+)\right)^{2}+\frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{2}(x, y, m \varepsilon,+)-m \varepsilon a_{1}(x, y, m \varepsilon,+)\right)^{2}= \\ &=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y, m \varepsilon,-)^{2}+a_{2}(x, y, m \varepsilon,+)^{2}\right]=\sum_{x \in \mathbb{Z}} P(x, y, m \varepsilon) \end{aligned} $$ Решим аналог задачи 9 для произвольного $m \varepsilon$. Найдём $a_{1}(x, y)$ для всех $y>|x|$. Нам достаточно рассмотреть пути с нечётным количеством поворотов. Обозначим через $2 r+1$ количество поворотов в пути. Обозначим через $R$ и $L$ количество ходов вправо-вверх и влево-вверх соответственно. Обозначим через $x_{1}, x_{2}, \ldots, x_{r+1}$ количество ходов вправо-вверх перед первым, третьим, ..., последним поворотом соответственно. Обозначим через $y_{1}, y_{2}, \ldots, y_{r+1}$ количество ходов влево-вверх после первого, третьего, ..., последнего поворота соответственно. Тогда $x_{k}, y_{k} \geq 1$ для всех $1 \leq k \leq r+1$ и $$ \begin{aligned} R & =x_{1}+\cdots+x_{r+1} \\ L & =y_{1}+\cdots+y_{r+1} \end{aligned} $$ Мы свели задачу к комбинаторной: найти количество натуральных решений двух полученных уравнений. Для первого уравнения это количество равно количеству способов поставить $r$ перегородок между $R$ шарами, то есть, $\binom{R-1}{r}$. Тогда: $$ a_{1}(x, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{R-1}{r}\binom{L-1}{r}(m \varepsilon)^{2 r+1} $$ Так как $L+R=y$ и $R-L=x$, мы получили искомую формулу. Доказательство формулы для $a_{2}(x, y)$ при $y>|x|$ аналогично. Случай $y \leq|x|$ очевиден. 15. Oтвет: при $|x|0$. При $r \geq T$ отношение последовательных слагаемых в выражениях для $a_{1}$ и $a_{2}$ равны соответственно $$ \begin{aligned} &\left(\frac{m}{n}\right)^{2} \frac{\left(\left(y_{n}+x_{n}-2\right) / 2-r\right)\left(\left(y_{n}-x_{n}-2\right) / 2-r\right)}{(r+1)^{2}}<\left(\frac{m}{n}\right)^{2} \cdot \frac{n(y+x)}{2 T} \cdot \frac{n(y-x)}{2 T}=\frac{z^{2}}{4 T^{2}} \\ &\left(\frac{m}{n}\right)^{2} \frac{\left(\left(y_{n}+x_{n}-2\right) / 2-r\right)\left(\left(y_{n}-x_{n}-2\right) / 2-r+1\right)}{(r-1) r}<\left(\frac{m}{n}\right)^{2} \cdot \frac{n(y+x)}{2 T} \cdot \frac{n(y-x)}{2 T-2}=\frac{z^{2}}{4 T(T-1)} \end{aligned} $$ Из $n>e^{3 z+1}$ следует, что $T=\lfloor\log n\rfloor+1 \geq 3 z+1$ и $\frac{z^{2}}{4 T^{2}}<\frac{z^{2}}{4 T(T-1)}<\frac{1}{2}$. Значит, в обоих случаях остаточный член (то есть сумма по всем $r \geq T$ ) меньше суммы геометрической прогрессии со знаменателем $\frac{1}{2}$. Значит, $$ \begin{aligned} a_{1}\left(n_{x}, n_{y}, \frac{m}{n},-\right) & =\frac{m}{n}\left(1+O\left(\frac{m^{2} y}{n}\right)\right) \\ & \cdot\left[\sum_{r=0}^{T-1}(-1)^{r}\binom{\left(y_{n}+x_{n}-2\right) / 2}{r}\binom{\left(y_{n}-x_{n}-2\right) / 2}{r}\left(\frac{m}{n}\right)^{2 r}+O\left(\frac{\left(y^{2}-x^{2}\right)^{T}}{(T!)^{2}} \cdot\left(\frac{m}{2}\right)^{2 T}\right)\right] \end{aligned} $$ $$ a_{2}\left(n_{x}, n_{y}, \frac{m}{n},+\right)=\frac{m}{n}\left(1+O\left(\frac{m^{2} y}{n}\right)\right) $$ $$ \cdot\left[\sum_{r=1}^{T-1}(-1)^{r}\binom{\left(y_{n}+x_{n}-2\right) / 2}{r}\binom{\left(y_{n}-x_{n}-2\right) / 2}{r-1}\left(\frac{m}{n}\right)^{2 r-1}+O\left(\frac{(y+x)^{T}(y-x)^{T-1}}{(T-1)!T!} \cdot\left(\frac{m}{2}\right)^{2 T-1}\right)\right] $$ Остаточный член в последней формуле можно оценить следующим образом. Так как $T!\geq(T / 3)^{T}$ и $T \geq \sqrt{y^{2}-x^{2}} \frac{3 m \sqrt{e}}{2}=\frac{3 \sqrt{e}}{2} z$, то $$ \frac{\left(y^{2}-x^{2}\right)^{T}}{(T!)^{2}} \cdot\left(\frac{m}{2}\right)^{2 T} \leq \frac{\left(y^{2}-x^{2}\right)^{T}}{(T)^{2 T}} \cdot\left(\frac{3 m}{2}\right)^{2 T} \leq e^{-T} \leq \frac{1}{n} $$ Теперь преобразуем биномиальные коэффициенты следующим образом: $$ \binom{\alpha n-1}{r}=\frac{(\alpha n-1) \cdots(\alpha n-r)}{r!}=\frac{(\alpha n)^{r}}{r!}\left(1-\frac{1}{\alpha n}\right) \cdots\left(1-\frac{r}{\alpha n}\right) $$ Если $\alpha \geq \varepsilon_{0}$ и $r0$. Рассмотрим $N$ такое, что $\sum_{i=0}^{N} b_{i}>b-\varepsilon$. Пусть $M_{i}$ для $0 \leq i \leq N$ таковы, что $\forall t \geq M_{i}: a_{i}(t)>b_{i}-\frac{\varepsilon}{2^{2+1}}$. Тогда для всех $n>\max \left(M_{0}, M_{1}, \ldots, M_{N}\right)$ выполнено $\sum_{i=0}^{\infty} a_{i}(n)>b-\varepsilon$. Значит, $\lim _{n \rightarrow \infty} \sum_{i=0}^{\infty} a_{i}(n)=b$. По лемме и утверждению 3 получаем $\lim _{n \rightarrow \infty} n \cdot \sum_{l=0 ; 2 \mid l}^{\infty} C_{A-1}^{l} C_{B-1}^{l}\left(\frac{m}{n}\right)^{2 l+1}=\sum_{l=0 ; 2 \mid l}^{\infty} \frac{\left(\frac{x+y}{2}\right)^{l}\left(\frac{y-x}{2}\right)^{l} \cdot m^{2 l+1}}{(l!)^{2}}$ Значит, по задаче 14 и утверждению 2 $$ \begin{aligned} \lim _{n \rightarrow \infty} n \vec{a}\left(x_{n}, y_{n}, \frac{m}{n},-\right) & =\lim _{n \rightarrow \infty} n \cdot \sum_{l=0 ; 2 \mid l}^{\infty}(-1)^{l} C_{A-1}^{l} C_{B-1}^{l}\left(\frac{m}{n}\right)^{2 l+1}= \\ & =\left(\sum_{l=0 ; 2 \mid l}^{\infty} \frac{\left(\frac{x+y}{2}\right)^{l}\left(\frac{y-x}{2}\right)^{l} \cdot m^{2 l+1}}{(l!)^{2}}\right)-\left(\sum_{l=0 ; 2 \nmid l}^{\infty} \frac{\left(\frac{x+y}{2}\right)^{l}\left(\frac{y-x}{2}\right)^{l} \cdot m^{2 l+1}}{(l!)^{2}}\right)=m \cdot J_{0}\left(m \sqrt{y^{2}-x^{2}}\right) \end{aligned} $$ | 3 | $(0,0)_{+}$
$(-1,0)_{-}$ | | $(0,1)_{+}$
$(0,0)_{-}$ | | $(0,-2)_{+}$
$(0,0)_{-}$ | | $(0,2)_{+}$
$(-1,0)_{-}$ | | $(0,-1)_{+}$
$(0,0)_{-}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 2 | | $(0,0)_{+}$
$(-1,0)_{-}$ | | $(0,-1)_{+}$
$(1,0)_{-}$ | | $(0,1)_{+}$
$(-1,0)_{-}$ | | $(0,-1)_{+}$
$(0,0)_{-}$ | | | 1 | | | $(0,-1)_{+}$
$(0,0)_{-}$ | | $(0,1)_{+}$
$(0,0)_{-}$ | | $(0,-1)_{+}$
$(0,0)_{-}$ | | | | $y$ x | -3 | -2 | -1 | 0 | 1 | 2 | 3 | 4 | 5 | Аналогично получаем $\lim _{n \rightarrow \infty} n \vec{a}\left(x_{n}, y_{n}, \frac{m}{n},+\right)=-i m \frac{x+y}{\sqrt{y^{2}-x^{2}}} \cdot J_{1}\left(m \sqrt{y^{2}-x^{2}}\right)$. 16. Oтвет: см. таблицу ниже; пара $(c, d)_{+}$в клетке $(x, y)$ означает, что $\vec{a}(x, y, 4,3,+)=\frac{1}{\sqrt{3}}(c, d)$, а $(c, d)_{-}$ означает, что $\vec{a}(x, y, 4,3,-)=\frac{1}{\sqrt{3}}(c, d)$; если клетка $(x, y)$ пуста, то $\vec{a}(x, y, 4,3,+)=\vec{a}(x, y, 4,3,-)=(0,0)$. Указание. Заметим, что нас интересуют только чётные $x_{0}$, так что $2 \pi x_{0} \varepsilon / \lambda=\frac{\pi x_{0}}{2}$ делится нацело на $\pi$. По аналогии с задачей 1, отсюда следует, что $\vec{a}(x, y, 4,3,+)$ параллелен $O y$, а $\vec{a}(x, y, 4,3,-)$ параллелен $O x$ (будьте внимательны: это неверно для произвольного $\varepsilon / \lambda$ ). 17. Найдём более общую формулу; решением этой задачи будет частный случай $m \varepsilon=1$. Обозначим $$ \begin{aligned} & a_{-}(x, y):=\vec{a}(x, y, m \varepsilon, \lambda / \varepsilon, \Delta,-) \\ & a_{+}(x, y):=\vec{a}(x, y, m \varepsilon, \lambda / \varepsilon, \Delta,+) \end{aligned} $$ где векторы в правой части определяются аналогично $\vec{a}(x, y, \lambda / \varepsilon, \Delta, \pm)$ и $\vec{a}(x, y, m \varepsilon, \pm)$. (Обратите внимание, что $a_{-}(x, y)$ и $a_{+}(x, y)$ это вектор-значные функции; в отличии от $a_{1}(x, y)$ и $a_{2}(x, y)$ они не являются координатами какого-то вектора.) Тогда мы получаем уравнение Дирака (где $i:=R^{\pi / 2}$ - это поворот на $90^{\circ}$ против часовой стрелки) $$ \left\{\begin{array}{l} a_{-}(x, y)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{-}(x+1, y-1)+\frac{-i m \varepsilon}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{+}(x+1, y-1) \\ a_{+}(x, y)=\frac{-i m \varepsilon}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{-}(x-1, y-1)+\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{+}(x-1, y-1) \end{array}\right. $$ Из задачи 14 мы знаем, что эти равенства верны при $\Delta=1$. Они остаются верными и для произвольного $\Delta$, потому что не важно, в каком порядке применять $R^{2 \pi x_{0} \varepsilon / \lambda}$ и $i=R^{\pi / 2}$. С помощью этих равенств мы по аналогии с задачей 14 получаем $\sum_{x \in \mathbb{Z}} P(x, y, \lambda / \varepsilon, \Delta)=1$ для всех $y \geq 1$ индукцией по $y$. 18. Так как $\Delta P(x, y, \lambda / \varepsilon, \Delta,+)=|\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)|^{2}$, достаточно доказать, что $\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)$ не зависит от $\Delta$ при $\Delta>y+|x|$. По определению $\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)=\sum_{\substack{x_{0}=1-\Delta \\ x_{0} \text { чётно }}}^{x_{0}=\Delta-1} \sum_{s} R^{2 \pi x_{0} \varepsilon / \lambda} \vec{a}(s)$. Но путей $s$, начинающихся в $\left(x_{0}, 0\right)$ и заканчивающихся в $(x, y)$, не существует при $\left|x-x_{0}\right|>y$, так что если число $\Delta \geq y+|x|$, то его увеличение не меняет сумму. 19. Oтвет: если $x_{1}$ и $x_{2}$ чётны, то $$ \begin{aligned} & \vec{a}\left(x_{1}, y, \lambda / \varepsilon,+\right)=R^{\left(x_{1}-x_{2}\right) 2 \pi \varepsilon / \lambda} \vec{a}\left(x_{2}, y, \lambda / \varepsilon,+\right) \\ & \vec{a}\left(x_{1}, y, \lambda / \varepsilon,-\right)=R^{\left(x_{1}-x_{2}\right) 2 \pi \varepsilon / \lambda} \vec{a}\left(x_{2}, y, \lambda / \varepsilon,-\right) \end{aligned} $$ Решение. Из решения задачи 18 следует, что $\vec{a}(x, y, \lambda / \varepsilon,+)=\sum_{x_{0} \text { чётно }} \sum_{s} R^{2 \pi x_{0} \varepsilon / \lambda} \vec{a}(s)$, где вторая сумма по всем путям $s$ из $\left(x_{0}, 0\right)$ в $(x, y)$, которые начинаются и заканчиваются ходом вправо-вверх. Каждому пути $s_{1}$ из $\left(x_{0}, 0\right)$ в $\left(x_{1}, y\right)$ сопоставим единственный путь $s_{2}$ из $\left(x_{0}+\left(x_{2}-x_{1}\right), 0\right)$ в $\left(x_{2}, y\right)$, полученный из $s_{1}$ параллельным переносом на $\left(x_{1}-x_{2}, 0\right)$. Таким образом, суммы для $x=x_{1}$ и $x=x_{2}$ берутся по одним и тем же векторам, только при $x=x_{1}$ они повёрнуты на $2 \pi x_{0} \varepsilon / \lambda$, а при $x=x_{2}$ они повёрнуты на $2 \pi\left(x_{0}+\left(x_{2}-x_{1}\right)\right) \varepsilon / \lambda$, откуда следует ответ. 20. Найдём более общую формулу; решением этой задачи будет частный случай $m \varepsilon=1$. Ответ. Обозначим $E=\arccos \left(\frac{\cos (2 \pi \varepsilon / \lambda)}{\sqrt{1+m^{2} \varepsilon^{2}}}\right)$. Тогда для чётного $x+y$ верно: $$ \begin{aligned} & \vec{a}(x, y, \lambda / \varepsilon,-)=\left(\frac{m \varepsilon \cos (2 \pi \varepsilon x / \lambda) \sin ((y-1) E)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E}, \frac{m \varepsilon \sin (2 \pi \varepsilon x / \lambda) \sin ((y-1) E)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E}\right) \\ & \vec{a}(x, y, \lambda / \varepsilon,+)=\left(\frac{\cos (2 \pi \varepsilon(x-2) / \lambda) \sin (y E)-\cos (2 \pi \varepsilon x / \lambda) \sin ((y-2) E)}{2 \cos (2 \pi \varepsilon / \lambda) \sin E}\right. \\ &\left.\frac{-\sin (2 \pi \varepsilon(x-2) / \lambda) \sin (y E)+\sin (2 \pi \varepsilon x / \lambda) \sin ((y-2) E)}{2 \cos (2 \pi \varepsilon / \lambda) \sin E}\right) \\ & P(x, y, \lambda / \varepsilon,-)= \frac{m^{2} \varepsilon^{2} \sin ^{2}((y-1) E)}{m^{2} \varepsilon^{2}+\sin ^{2}(2 \pi \varepsilon / \lambda)} \\ & P(x, y, \lambda / \varepsilon,+)= \frac{m^{2} \varepsilon^{2} \cos ^{2}((y-1) E)+\sin ^{2}(2 \pi \varepsilon / \lambda)}{m^{2} \varepsilon^{2}+\sin ^{2}(2 \pi \varepsilon / \lambda)} \\ & P(x, y, \lambda / \varepsilon,-)+P(x, y, \lambda / \varepsilon,+)=1 \end{aligned} $$ Замечание. В комплексной форме, $$ \begin{aligned} & \vec{a}(x, y, \lambda / \varepsilon,-)=r^{x} \frac{m \varepsilon \sin ((y-1) E)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E} \\ & \vec{a}(x, y, \lambda / \varepsilon,+)=r^{x-1}\left(\frac{\sin ((y-1) E) \sin (2 \pi \varepsilon / \lambda)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E}+i \cos ((y-1) E)\right) \end{aligned} $$ где $r=\cos (2 \pi \varepsilon / \lambda)+i \sin (2 \pi \varepsilon / \lambda)$. Это решение уравнения Дирака (см. решение задачи 17) с (квази) периодическим начальным условием $a_{-}(x, 0)=0$ и $a_{+}(x, 0)=\left(-\sin \frac{2 \pi x \varepsilon}{\lambda}, \cos \frac{2 \pi x \varepsilon}{\lambda}\right)$. Число $E$ имеет физический смысл: это энергия. Указание. Формулы для $\vec{a}(x, y, \lambda / \varepsilon,+)$ и $\vec{a}(x, y, \lambda / \varepsilon,-)$ могут быть доказаны индукцией по $y$. База очевидна; шаг следует из задачи 17. После этого формулы для $P(x, y, \lambda / \varepsilon,+)$ и $P(x, y, \lambda / \varepsilon,-)$ можно вычислить напрямую. Путь к решению. Будем воспринимать векторы, как комплексные числа. Тогда $R^{\alpha} \vec{a}=(\cos \alpha+i \sin \alpha) \vec{a}$. Обозначим $r=\cos (2 \pi \varepsilon / \lambda)+i \sin (2 \pi \varepsilon / \lambda)$; тогда $r^{-1}=\cos (2 \pi \varepsilon / \lambda)-i \sin (2 \pi \varepsilon / \lambda)$. Обозначим $f(y)=r^{-y} \vec{a}(y, y, \lambda / \varepsilon,+)$ и $g(y)=r^{-y} \vec{a}(y, y, \lambda / \varepsilon,-)$. Обозначим $M=\frac{m \varepsilon}{\sqrt{1+m^{2} \varepsilon^{2}}}$ и $U=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}$. Из задачи 19 следует, что при чётном $x+y$ вектор $\vec{a}(x, y, \lambda / \varepsilon,+)=r^{x} f(y)$ и $\vec{a}(x, y, \lambda / \varepsilon,-)=r^{x} g(y)$. Из задачи 17 следует $r^{x} f(y)=$ $U r^{x-1} f(y-1)-i M r^{x-1} g(y-1)$ и $r^{x} g(y)=U r^{x+1} g(y-1)-i M r^{x+1} f(y-1)$. Так как $r \neq 0$, то мы можем поделить обе части на $r^{x}$ и получить следующую систему: $$ \left\{\begin{array}{l} f(y)=U r^{-1} f(y-1)-i M r^{-1} g(y-1) \\ g(y)=\operatorname{Urg}(y-1)-i M r f(y-1) \\ f(1)=i r^{-1} \\ g(1)=0 \end{array}\right. $$ Есть несколько способов решить эту систему (мы рассматривали варианты решений через производящие функции и через диагонализацию 'матрицы перехода'), но наиболее доступным (для матшкольников), видимо, является преобразований этой системы к виду линейной рекурренты порядка 2. Из (1) мы получаем $g(y-1)=\frac{1}{M} \operatorname{irf}(y)-\frac{U}{M} i f(y-1)$. Отсюда и из (2) получаем $g(y)=\frac{U}{M} \operatorname{ir} f(y)-\frac{1}{M} \operatorname{irf}(y-1)$. Отсюда и из (1) получаем $f(y)=U\left(r+r^{-1}\right) f(y-1)-f(y-2)$. Решением этой рекурренты является $f(n)=c_{1} d_{+}^{n}+c_{2} d_{-}^{n}$, где $$ \begin{aligned} & d_{+}=\frac{U\left(r+r^{-1}\right)+\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}}{2}=U \cos (2 \pi \varepsilon / \lambda)+i \sqrt{U^{2} \sin ^{2}(2 \pi \varepsilon / \lambda)+M^{2}}, \\ & d_{-}=\frac{U\left(r+r^{-1}\right)-\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}}{2}=U \cos (2 \pi \varepsilon / \lambda)-i \sqrt{U^{2} \sin ^{2}(2 \pi \varepsilon / \lambda)+M^{2}} . \end{aligned} $$ Это корни уравнения $t^{2}-U\left(r+r^{-1}\right) t+1=0$, а $c_{1}$ и $c_{2}$ - это константы, которые можно найти, используя (3) и (4). Так как мы работаем с комплексными числами, запись $\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}$ неоднозначна. Но мы можем выбрать любой из двух корней; $\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}$ обозначает $2 i \sqrt{U^{2} \sin ^{2}(2 \pi \varepsilon / \lambda)+M^{2}}$. Мы получаем $$ c_{1}=i \frac{r^{-1}-U d_{-}}{d_{+}-d_{-}} \text {и } c_{2}=i \frac{r^{-1}-U d_{+}}{d_{-}-d_{+}} . $$ Заметим, что $E=\arccos (U \cos (2 \pi \varepsilon / \lambda))$. Мы получаем $d_{+}=\cos E+i \sin E$ и $d_{-}=\cos E-i \sin E$. Тогда $\vec{a}(x, y, \lambda / \varepsilon,+)=r^{x} f(y)=i r^{x} \frac{\left(r^{-1}-U d_{-}\right) d_{+}^{y}-\left(r^{-1}-U d_{+}\right) d_{-}^{y}}{d_{+}-d_{-}}=i r^{x-1} \frac{\sin (y E)}{\sin E}-i U r^{x} \frac{\sin ((y-1) E)}{\sin E}$. Теперь вычислим $P(x, y, \lambda / \varepsilon,+)=\vec{a}(x, y, \lambda / \varepsilon,+) \cdot \overrightarrow{\vec{a}}(x, y, \lambda / \varepsilon,+)$. Из формул выше следует, что $\bar{r}=r^{-1}$, $\overline{d_{+}}=d_{-}, \overline{d_{-}}=d_{+}, \overline{c_{1}}=i \frac{r-U d_{+}}{d_{+}-d_{-}}$и $\overline{c_{2}}=i \frac{r-U d_{-}}{d_{-}-d_{-}}$. Отсюда (с помощью некоторых упрощений) мы наконец получаем $P(x, y, \lambda / \varepsilon,+)=\frac{\left(d_{+}-d_{-}\right)^{2}-M^{2}\left(d_{-}^{y-1}-d_{-}^{y-1}\right)^{2}}{\left(d_{+}-d_{-}\right)^{2}}=1-\frac{M^{2} \sin ^{2}((y-1) E)}{\sin ^{2} E}$. Аналогично мы получаем, что $\vec{a}(x, y, \lambda / \varepsilon,-)=r^{x} M \frac{\sin ((y-1) E)}{\sin E}$ и $P(x, y, \lambda / \varepsilon,-)=\frac{M^{2} \sin ^{2}((y-1) E)}{\sin ^{2} E}$. 21. Решение. Обозначим через $x_{i}$ координату по оси $O x$ у клетки, в которую шашка попадает на $i$-м ходе пути $s$ (то есть, шашка идёт по клеткам $(0,0),(1,1),\left(x_{2}, 2\right), \cdots,\left(x_{y}, y\right)$ ). Чтобы дать определение вектора $\vec{a}(s, m(x) \varepsilon)$, начнём с вектора $(0,1)$. Пока шашка двигается прямо, вектор не меняется, но если шашка меняет направление после $n$-ого хода, вектор поворачивается на $90^{\circ}$ по часовой стрелке и умножается на $m\left(x_{n}\right) \varepsilon$. Кроме того, в конце мы делим вектор на $\prod_{n=1}^{y-1} \sqrt{1+m^{2}\left(x_{n}\right) \varepsilon^{2}}$, где $y$ - это количество ходов пути $s$. Полученный вектор мы обозначаем через $\vec{a}(s, m(x) \varepsilon)$. Числа $a_{1}(x, y, m(x) \varepsilon)$ и $a_{2}(x, y, m(x) \varepsilon)$ определяются аналогично $a_{1}(x, y)$ и $a_{2}(x, y)$, только $\vec{a}(s)$ заменяется на $\vec{a}(s, m(x) \varepsilon)$. По аналогии с задачей 4 мы получаем формулы $$ \begin{aligned} & a_{1}\left(x_{0}, y, m(x) \varepsilon\right)=\frac{1}{\sqrt{1+m^{2}\left(x_{0}+1\right) \varepsilon^{2}}}\left(a_{1}\left(x_{0}+1, y, m(x) \varepsilon\right)+m\left(x_{0}+1\right) \varepsilon \cdot a_{2}\left(x_{0}+1, y, m(x) \varepsilon\right)\right) \\ & a_{2}\left(x_{0}, y, m(x) \varepsilon\right)=\frac{1}{\sqrt{1+m^{2}\left(x_{0}-1\right) \varepsilon^{2}}}\left(a_{2}\left(x_{0}-1, y, m(x) \varepsilon\right)-m\left(x_{0}-1\right) \varepsilon \cdot a_{1}\left(x_{0}-1, y, m(x) \varepsilon\right)\right) \end{aligned} $$ 22. Oтвет: Для всех $x, y$ одной чётности: $P\left(x, y, m_{0}(x), \lambda / \varepsilon,+\right)=1$; $P\left(x, y, m_{0}(x), \lambda / \varepsilon,-\right)=0$. $P\left(x, y, m_{1}(x), \lambda / \varepsilon,+\right)= \begin{cases}\frac{1}{1.04}=\frac{25}{26}, & \text { при } 01$ верно неравенство $y=2 L<3(L-1)$, то количество поворотов у таких путей равно 1. Тогда либо путь начинается в клетке $(0,0)$ и поворачивает в клетке $(L, L)$, либо он начинается в клетке $(2-2 L, 0)$ и поворачивает в $(1,2 L-1)$. Векторы этих двух путей отличаются поворотом на $\pi-\frac{\pi(L-1)}{4}$ и делением на 1.04. Так как квадрат длины вектора первого пути равен $\frac{1}{26}$, а квадрат длины второго равен $\frac{1}{26 \cdot 1.04^{2}}$, то по теореме косинусов $$ P\left(0,2 L, m_{2}(x), 16,-\right)=\frac{1}{26}\left(1+\frac{1}{1.04^{2}}-\frac{2}{1.04} \cos \frac{\pi(L-1)}{4}\right) \approx 0.074\left(1-\cos \frac{\pi(L-1)}{4}\right) $$ График этой функции изображён на следующем рисунке (любезно предоставленном Punnawith Thuwajit, участником Летней конференции Турнира городов). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-076.jpg?height=594&width=916&top_left_y=1268&top_left_x=570) 24. Oтвет: $$ \begin{array}{ll} P(\lambda, L,-)=\frac{0.1664}{1.1664+\operatorname{ctg}^{2} \frac{2 \pi(L-1)}{\lambda}} ; & \max _{L} P(\lambda, L,-)=\frac{0.1664}{1.1664} \approx 0.14 \\ P(\lambda, L,+)=\frac{1}{1+0.1664 \sin ^{2} \frac{2 \pi(L-1)}{\lambda}} ; & P(\lambda, L,-)+P(\lambda, L,+)=1 \end{array} $$ Указание. Обозначим $m=0.2$ и $\phi=\frac{2 \pi}{\lambda}(L-1)$. При чётном $y>2 L-2$ мы получаем $$ \begin{aligned} & P\left(0, y, m_{2}(x), \lambda,-\right)= \\ & =\left|e^{\frac{2 \pi i}{\lambda}(2-y)} \frac{m}{\sqrt{1+m^{2}}}\left(-1+\frac{1}{1+m^{2}} e^{2 i \phi}+\frac{m^{2}}{\left(1+m^{2}\right)^{2}}\left(e^{2 i \phi}\right)^{2}+\cdots+\frac{1}{m^{2}}\left(\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right)^{\left[\frac{y-2}{2(L-1)}\right] 2(L-1)}\right)\right|^{2} . \end{aligned} $$ Тогда, так как $\cos 2 \phi=1-2 \sin ^{2} \phi$, мы получаем $$ \begin{gathered} P(\lambda, L,-)=\frac{m^{2}}{1+m^{2}} \cdot\left|1-\frac{1}{1+m^{2}} e^{2 i \phi} \cdot \sum_{k=0}^{\infty}\left(\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right)^{k}\right|^{2}= \\ =\frac{m^{2}}{1+m^{2}} \cdot\left|1-\frac{1}{1+m^{2}} e^{2 i \phi} \frac{1}{1-\frac{m^{2}}{1+m^{2}} e^{2 i \phi}}\right|^{2}=\frac{m^{2}}{1+m^{2}} \cdot \frac{\left|1-e^{2 i \phi}\right|^{2}}{\left|1-\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right|^{2}}= \\ =\frac{m^{2}\left(1+m^{2}\right)(2-2 \cos (2 \phi))}{\left(1+m^{2}\right)^{2}+m^{4}-2 m^{2}\left(1+m^{2}\right) \cos (2 \phi)}=\frac{4 m^{2}\left(1+m^{2}\right)}{\left(1+2 m^{2}\right)^{2}+\operatorname{ctg}^{2} \phi}=\frac{0.1664}{1.1664+\operatorname{ctg}^{2} \frac{2 \pi}{\lambda}(L-1)} \end{gathered} $$ Таким образом, $\max _{L} P(\lambda, L,-)=\frac{0.1664}{1.1664} \approx 0.14$. Аналогично, $$ \begin{aligned} P(\lambda, L,+) & =\left|\frac{1}{1+m^{2}} \cdot \sum_{k=0}^{\infty}\left(\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right)^{k}\right|^{2}=\left|\frac{1}{1+m^{2}} \cdot \frac{1}{1-\frac{m^{2}}{1+m^{2}} e^{2 i \phi}}\right|^{2}= \\ & =\frac{1}{\left(1+m^{2}\right)^{2}+m^{4}-2\left(1+m^{2}\right) m^{2} \cos (2 \phi)}=\frac{1}{1+4 m^{2}\left(1+m^{2}\right) \sin ^{2} \phi}=\frac{1}{1+0.1664 \sin ^{2} \frac{2 \pi}{\lambda}(L-1)} \end{aligned} $$ Заметим, что $$ 1-\frac{1}{1+0.1664 \sin ^{2} \frac{2 \pi}{\lambda}(L-1)}=\frac{0.1664}{\frac{1}{\sin ^{2} \frac{2 \pi}{\lambda}(L-1)}+0.1664}=\frac{0.1664}{1.1664+\operatorname{ctg}^{2} \frac{2 \pi}{\lambda}(L-1)} $$ Тогда $P(\lambda, L,-)+P(\lambda, L,+)=1$. 25. Omвem: $$ \frac{\left(n^{2}-1\right)^{2}}{\left(n^{2}+1\right)^{2}+4 n^{2} \cot ^{2} \frac{2 \pi L n}{\lambda}} \quad \text { и } \quad m=\frac{(0.3328+0.864 \sqrt{1.04}) \cdot \pi}{\lambda} $$ где $n=\sqrt{1+m \lambda / \pi}$ ( коэффициент преломления; заметьте, что связь между $m$ и $n$ изменилась, когда мы перешли к отражению внутри стекла). Для указанного $m$ предел равен $$ \frac{0.1664}{1.1664+\cot ^{2} \frac{2 \pi L(1.08+0.08 \sqrt{1.04})}{\lambda}} $$ Замечание. Первая формула из ответа хорошо известна и подтверждена экспериментами. Примечательное научно-популярное обсуждение задач 23 можно найти в Feуnman, а вывод уравнений на физическом уровне строгости - в Landafshitz, глава Х, §86, задача 4]. Путь к решению. Рассмотрим уравнение Дирака из решения задачи 17. Будем искать решение вида $$ \left(a_{+}(x, y), a_{-}(x, y)\right)=\left(a_{+}(x), a_{-}(x)\right) e^{-2 \pi i \varepsilon y / \lambda} $$ Используем задачу 20, чтобы угадать $a_{ \pm}(x)$ по отдельности для $x \leq 0,0 \leq x \leq L / \varepsilon$ и $x \geq L / \varepsilon$. Предположим для простоты, что $1 / \varepsilon-$ нечётное число. Для краткости будем писать $m(x):=m_{3}(x)$. Сначала рассмотрим полуплоскость $x \leq 0$ (слева от стекла). Сформулируем, что такое решение уравнения Дирака в полуплоскости. По аналогии с задачей 17 , под уравнением Дирака в чёрной клетке $(x, y)$ мы понимаем систему $$ \left\{\begin{array}{l} a_{-}(x-1, y+1)=\frac{1}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{-}(x, y)+\frac{-i m(x) \varepsilon}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{+}(x, y) \\ a_{+}(x+1, y+1)=\frac{-i m(x) \varepsilon}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{-}(x, y)+\frac{1}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{+}(x, y) \end{array}\right. $$ Решением уравнения Дирака в полуплоскости $x \leq 0$ называется пара $\left(a_{+}(x, y), a_{-}(x, y)\right)$, удовлетворяющая уравнениям в каждой чёрной клетке полуплоскости. В частности, $a_{-}(x, y)$ определено при $x \leq 0$, в то время как $a_{+}(x, y)$ определено при $x \leq 1$. Будем искать решение вида $\left(a_{+}(x), a_{-}(x)\right)=\left(i e^{2 \pi i x \varepsilon / \lambda}, r(\varepsilon) e^{-2 \pi i x \varepsilon / \lambda}\right)$ для некоторого $r(\varepsilon) \in \mathbb{C}$, который будет найден позднее, где формулы для $a_{+}(x)$ и $a_{-}(x)$ верны для $x \leq 1$ и $x \leq 0$ соответственно. (В терминах физики, $a_{+}(x, y)$ - это падающая волна, $a_{-}(x, y)$ - это отражённая волна, а $\left|\lim _{\varepsilon \rightarrow 0} r(\varepsilon)\right|^{2}$ - это вероятность отражения.) Для полуплоскости $x \geq L / \varepsilon$ возьмём $\left(a_{+}(x), a_{-}(x)\right)=\left(t(\varepsilon) e^{2 \pi i x \varepsilon / \lambda}, 0\right)$ для некоторого $t(\varepsilon) \in \mathbb{C}$, который будет найден позднее, где формулы для $a_{+}(x)$ и $a_{-}(x)$ верны для $x \geq L / \varepsilon+1$ и $x \geq L / \varepsilon$ соответственно. (В терминах физики, $a_{+}(x)$ - это прошедшая волна, а $\left|\lim _{\varepsilon \rightarrow 0} t(\varepsilon)\right|^{2}-$ вероятность прохождения.) Для $0 \leq x \leq L / \varepsilon$ будем искать решение вида $$ \begin{aligned} & a_{+}(x)=a(\varepsilon) e^{2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)}+b(\varepsilon) e^{-2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)} \\ & a_{-}(x)=c(\varepsilon) e^{2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)}+d(\varepsilon) e^{-2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)} \end{aligned} $$ для некоторых коэффициентов $a, b, c, d \in \mathbb{C}$ и $\lambda^{\prime}>0$, зависящих от $\varepsilon$ (это комбинация преломлённой и отражённой волн). В дальнейшем $a, b, c, d, \lambda^{\prime}, r, t$ обозначают пределы соответствующих функций при $\varepsilon \rightarrow 0$. Предел $\lambda^{\prime}$ (длины преломлённой волни) можно найти с помощью уравнения Дирака. Подставив выражения для $a_{ \pm}(x)$ в уравнение Дирака, сократив общие множители, и взяв предел при $\varepsilon \rightarrow 0$, мы получим $$ \begin{aligned} m a & =\left(m-\frac{2 \pi}{\lambda^{\prime}}+\frac{2 \pi}{\lambda}\right) c \\ m c & =\left(m+\frac{2 \pi}{\lambda^{\prime}}+\frac{2 \pi}{\lambda}\right) a \end{aligned} $$ Также мы получим аналогичное равенство с $b$ и $d$ вместо $c$ и $a$. Отсюдда следует, что $\lambda^{\prime}=\lambda / n$, где $n=$ $\sqrt{1+m \lambda / \pi}$ (коэффициент преломления). Сшивание решений $\left(a_{+}(x)\right.$ в $x=1$ и $x=L / \varepsilon+1$, а $a_{-}(x)$ в $x=0$ и $\left.x=L / \varepsilon\right)$ даёт нам систему линейных уравнений от $a, b, c, d, r, t$ : $a_{-}(0)=r(\varepsilon)=c(\varepsilon)+d(\varepsilon), \quad a_{-}(L / \varepsilon)=0=c(\varepsilon) e^{2 \pi i L / \lambda^{\prime}}+d(\varepsilon) e^{-2 \pi i L / \lambda^{\prime}}$, $a_{+}(1)=i e^{2 \pi i L / \lambda}=a(\varepsilon) e^{2 \pi i L / \lambda^{\prime}}+b(\varepsilon) e^{-2 \pi i L / \lambda^{\prime}}, \quad a_{+}(L / \varepsilon+1)=t(\varepsilon) e^{2 \pi i(L+\varepsilon) / \lambda}=a(\varepsilon) e^{2 \pi i(L+\varepsilon) / \lambda^{\prime}}+b(\varepsilon) e^{-2 \pi i(L+\varepsilon) / \lambda^{\prime}}$. Переходя к пределу при $\varepsilon \rightarrow 0$ и решаю полученную систему линейных уравнений, мы находим $a, b, c, d, r, t$. Разумеется, надо проверить, что прежде, чем мы перешли к пределу, наши выражения для $a_{ \pm}(x)$, полученные из этих уравнений, удовлетворяли уравнению Дирака в каждой чёрной клетке. Мы опускаем это. Осталось доказать, что $\left(a_{+}(x), a_{-}(x)\right)=\lim _{\substack{y \rightarrow+\infty \\ y+x \text { even }}} R^{2 \pi y \varepsilon / \lambda} \vec{a}\left(x, y, m_{3}(x) \varepsilon, \frac{\lambda}{\varepsilon},-\right)$; Тогда число $|r|^{2}$ является искомым пределом. Мы не знаем элементарного доказательства этого, и поэтому опускаем детали. Нестрого, идея заключается в том, чтобы показать, что предел не зависит от начальных значений $a_{+}(x, 0)$ при $x>0$ и $a_{-}(x, 0)$ при $xL / \varepsilon$ и от $a_{-}(x, 0)$ при $x<0$ очевидна. Чтобы доказать независимость от $a_{ \pm}(x, 0)$ при $0 \leq x \leq L / \varepsilon$, предположим, что $a_{ \pm}(x, 0)$ существует только на отрезке $0 \leq x \leq L / \varepsilon$, и рассмотрим линейное преобразование $$ \begin{aligned} & U:\left(a_{-}(0,0), a_{-}(2,0), \ldots, a_{-}(L / \varepsilon-1,0), a_{+}(0,0), \ldots, a_{+}(L / \varepsilon-1,0)\right) \\ & \mapsto\left(a_{-}(0,2), a_{-}(2,2), \ldots, a_{-}(L / \varepsilon-1,2), a_{+}(0,2), \ldots, a_{+}(L / \varepsilon-1,2)\right) \end{aligned} $$ Покажем, что модули собственных значений $U$ меньше 1. Рассмотрим собственный вектор $v$ с координатами $v_{1}=a_{-}(0,0), \ldots, v_{L / \varepsilon+1}=a_{+}(L / \varepsilon-1,0)$. Без нарушения общности $v_{k}-$ самая левая ненулевая координата. Если $k>1$, то $U v$ имеет ненулевую $(k-1)$-ую координату; но $U v$ пропорционален $v$, получаем противоречие. Так что $k=1$. Отсюда $a_{-}(-2,2) \neq 0$ из уравнения Дирака. Значит, по сохранению вероятности $$ \begin{aligned} \|v\| & =\left|a_{-}(0,0)\right|^{2}+\left|a_{-}(2,0)\right|^{2}+\cdots+\left|a_{+}(L / \varepsilon-1,0)\right|^{2} \\ & =\left|a_{-}(-2,2)\right|^{2}+\left|a_{-}(0,2)\right|^{2}+\cdots+\left|a_{+}(L / \varepsilon-1,2)\right|^{2} \\ & >\left|a_{-}(0,2)\right|^{2}+\cdots+\left|a_{+}(L / \varepsilon-1,2)\right|^{2} \\ & =\|U v\| \end{aligned} $$ Отсюда модуль собственного значения $v$ строго меньше 1 . Поэтому $\left\|U^{y}\right\| \rightarrow 0$ при $y \rightarrow \infty$, откуда следует, что значения $a_{ \pm}(x, 0)$ при $0 \leq x \leq L$ не влияют на предел. 26. Ответ. Вектор в клетке $(x, y)$ таблицы - это $\vec{a}(x, y, u,+)$, а если клетка $(x, y)$ пуста, то $\vec{a}(x, y, u,+)=$ $(0,0)$. | 4 | | | $\left(0, \frac{1}{2 \sqrt{2}}\right)$ | | $(0,0)$ | | $\left(0, \frac{1}{2 \sqrt{2}}\right)$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | | | $\left(0,-\frac{1}{2}\right)$ | | $\left(0, \frac{1}{2}\right)$ | | | 2 | | | | $\left(0,-\frac{1}{\sqrt{2}}\right)$ | | | | | 1 | | | | $(0,-1)$ | | | | | $\mathrm{y}$ | | | | | | | | В следующей таблице число в клетке $(x, y)$ - это $P(x, y, u,+)$, а пустая клетка $(x, y)$ означает, что $P(x, y, u,+)=0$. | 4 | | | $1 / 8$ | | 0 | | $1 / 8$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | | | $1 / 4$ | | $1 / 4$ | | | 2 | | | | | $1 / 2$ | | | | 1 | | | | 1 | | | | | $\mathrm{y}$ | | | | | | | | (Векторы $\vec{a}(x, y, u,+)$ можно легко вычислить, используя задачу 31.) Для всех $y>1$ верно $\sum_{x \in \mathbb{Z}}(P(x, y, u,+)+P(x, y, u,-))=1$. (Это одно из утверждений задачи 31 ) 27. Ответ (любезно предоставленный Глебом Минаевым и Иваном Русских, участниками Летней конференции Турнира городов): ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-079.jpg?height=760&width=1008&top_left_y=768&top_left_x=178) 28. Решение. Обозначим через $u^{\prime}$ поле, получающееся из $u$ заменой знаков в вершинах чёрной клетки $(a, b)$. Заметим, что если путь $s$ не начинается и не заканчивается в клетке $(a, b)$, то он либо проходит через две вершины, значение $u$ в которых изменилось, либо не проходит ни через одну. Таким образом, $\vec{a}(s, u)=\vec{a}\left(s, u^{\prime}\right)$. Если $(a, b) \neq(0,0)$ и $(a, b) \neq(x, y)$, то равенство $P\left(x, y, u^{\prime},+\right)=P(x, y, u,+)$ следует из этого напрямую. Осталось рассмотреть только случаи $(a, b)=(0,0)$ и $(a, b)=(x, y)$. Каждый путь $s$ из $(0,0)$ в $(x, y)$ будет проходить ровно через одну вершину, значение $u$ в которой изменилось, поэтому $\vec{a}\left(s, u^{\prime}\right)=-\vec{a}(s, u)$. Таким образом, в этих случаях мы также получаем $P\left(x, y, u^{\prime},+\right)=P(x, y, u,+)$. 29. Решение. Назовём обращением чёрной клетки $(x, y)$ изменение значений $u$ во всех вершинах этой клетки. Обозначим через вершину $(x, y)$ верхнюю правую вершину клетки $(x, y)$, а через $u(x, y)$ - значение $u$ в этой вершине. Заметим, что обращение чёрной клетки не меняет знак ни одной белой клетки, так как в каждой белой клетке количество изменённых вершин чётно. Если все значения поля $u$ в прямоугольнике положительны, то и все белые клетки в прямоугольнике положительны. Таким образом, если в $и$ изначально были отрицательные белые клетки, то нельзя сделать $u$ тождественно равным +1 , обращая чёрные клетки. Предположим, в прямоугольнике $M \times N$, где $N$ - это высота, нет отрицательных белых клеток. Без ограничения общности будем считать вершину $(0,0)$ левым нижним углом прямоугольника. Докажем индукцией по $N$, что $u$ можно сделать тождественно равным +1 с помощью обращения чёрных клеток. База: $N=0$. Приведём алгоритм, позволяющий сделать $u$ тождественно равным +1 в прямоугольнике $M \times 0$ (то есть, в вершинах $(0,0),(1,0), \ldots,(M, 0)$ ). Заметим, что для любого $k$, обратив либо клетку $(k+1,0)$, либо клетку $(k+1,1)$, мы можем изменить значение $u$ в вершинах $(k, 0)$ и $(k+1,0)$. Будем называть эту операцию обращением $k$. Алгоритм такой. Найдём наименьшее такое $0 \leq k \leq M$, что $u(k, 0)=-1$, и применим обращение $k$. Если после применения этой операции ещё остались такие $0 \leq k \leq M$, что $u(k, 0)=-1$, то наименьшее такое $k$ хотя бы на 1 больше, чем до применения операции. Значит, не больше чем за $M+1$ применений операции мы получим $u$, тождественно равное +1 в прямоугольнике $M \times 0$. Шаг: Пусть дан прямоугольник $M \times(N+1)$. Сначала сделаем $и$ тождественно равным +1 в нижнем прямоугольнике $M \times N$ (это возможно по предположению индукции). После этого для каждой белой клетки $(k, N+1)$ обратим клетку $(k, N+2)$, если в обеих общих вершинах этих клеток поле $u$ имело значение -1 . Так как все белые клетки были положительными, то после этого поле $u$ может иметь отрицательное значение только в одной или двух вершинах в углах прямоугольника $M \times(N+1)$ (и только если в соответственном углу стоит чёрная клетка). Но тогда значения в любой из этих вершин можно исправить, обратив другую чёрную клетку, которая к ней прилегает. Сделав это, мы получим искомое поле $u$, тождественно равное +1 во всех вершинах прямоугольника $M \times(N+1)$. 30. Oтвет: Если ширина или длина прямоугольника равна 1, то таких клеток ноль. Если же они обе больше 1, то любое чётное неотрицательное число, не большее общего количества белых клеток в прямоугольнике. Указание. Построим граф: вершинами будут белые клетки прямоугольника, а ребро между двумя вершинами графа означает, что соответствующие клетки соприкасаются по вершине, и в этой вершине значение $u$ равно +1 . Тогда отрицательными белыми клетками будут в точности те, которым соответствуют вершины нечётной степени в графе. По лемме о рукопожатиях количество таких вершин чётно. Покажем, что любое чётное неотрицательное число, не большее общего количества белых клеток, можно реализовать. Для поля $u_{0}$, тождественно равного +1 , отрицательных белых клеток не будет. Поле $u_{2 k+2}$ с $2 k+2$ отрицательными белыми клетками получается из поля $u_{2 n}$ заменой значений $u_{2 n}$ на -1 в любой обведённой области на рисунке, в которой это ещё не сделано. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-080.jpg?height=731&width=940&top_left_y=874&top_left_x=158) 31. Oтвет. Напомним, что $u(x, y)$ - это значение $u$ в правой верхней вершине клетки $(x, y)$. Тогда $$ \begin{aligned} & a_{1}(x, y, u)=\frac{1}{\sqrt{2}} u(x, y-1)\left(a_{1}(x+1, y-1, u)+a_{2}(x+1, y-1, u)\right) \\ & a_{2}(x, y, u)=\frac{1}{\sqrt{2}} u(x-1, y-1)\left(-a_{1}(x-1, y-1, u)+a_{2}(x-1, y-1, u)\right) \end{aligned} $$ Указание. Решение аналога задачи 4 отличается от решения оригинала только добавлением множителей $u(x, y-1)$ и $u(x-1, y-1)$, так как последний ход пути $s$ проходит через правую верхнюю вершину либо клетки $(x, y-1)$, либо $(x-1, y-1)$. Аналог задачи 5 можно доказать индукцией по $y$. Шаг индукции следует из следующих вычислений: $$ \begin{gathered} \sum_{x \in \mathbb{Z}} P(x, y+1, u)=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y+1, u)^{2}+a_{2}(x, y+1, u)^{2}\right]=\sum_{x \in \mathbb{Z}} a_{1}(x, y+1, u)^{2}+\sum_{x \in \mathbb{Z}} a_{2}(x, y+1, u)^{2}= \\ =\frac{1}{2} \sum_{x \in \mathbb{Z}} u(x, y)^{2}\left(a_{1}(x+1, y, u)+a_{2}(x+1, y, u)\right)^{2}+\frac{1}{2} \sum_{x \in \mathbb{Z}} u(x-1, y)^{2}\left(a_{2}(x-1, y, u)-a_{1}(x-1, y, u)\right)^{2}= \\ =\sum_{x \in \mathbb{Z}} \frac{\left(a_{1}(x, y, u)+a_{2}(x, y, u)\right)^{2}}{2}+\sum_{x \in \mathbb{Z}} \frac{\left(a_{2}(x, y, u)-a_{1}(x, y, u)\right)^{2}}{2}=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y, u)^{2}+a_{2}(x, y, u)^{2}\right]=\sum_{x \in \mathbb{Z}} P(x, y, u) . \\ \hline \mathbf{3 2} \text { Oтвет: } P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)= \begin{cases}P(x, y,-) P\left(x^{\prime}-x_{0}, y,-\right), & \text { если } E=(x+1, y), E^{\prime}=\left(x^{\prime}+1, y\right), \\ P(x, y,-) P\left(x^{\prime}-x_{0}, y,+\right), & \text { если } E=(x+1, y), E^{\prime}=\left(x^{\prime}-1, y\right), \\ P(x, y,+) P\left(x^{\prime}-x_{0}, y,-\right), & \text { если } E=(x-1, y), E^{\prime}=\left(x^{\prime}+1, y\right), \\ P(x, y,+) P\left(x^{\prime}-x_{0}, y,+\right), & \text { если } E=(x-1, y), E^{\prime}=\left(x^{\prime}-1, y\right)\end{cases} \end{gathered} $$ Решение. Заметим, что $\vec{a}\left(s, s^{\prime}\right)=-i \vec{a}(s) \vec{a}\left(s^{\prime}\right)$, где $\overrightarrow{c_{1}} \overrightarrow{c_{2}}$ обозначает произведение комплексных чисел $\overrightarrow{c_{1}}$ и $\overrightarrow{c_{2}}$. Рассмотрим случай $x^{\prime}>x$. Тогда из-за того, что $x_{0} \geq 2 y$, если существует путь из $A$ в $F^{\prime}$, то из $A^{\prime}$ нет пути ни в $F$, ни в $F^{\prime}$. Аналогично про существование пути из $A^{\prime}$ в $F$. Значит, нас интересуют только пары путей $s$ и $s^{\prime}$, где $s$ ведёт из $A$ в $F$, а $s^{\prime}$ из $A^{\prime}$ в $F^{\prime}$. $\underset{s: A B \rightarrow E^{\prime} F^{\prime}}{\text { Таким образом, }} \overrightarrow{\left(s, s^{\prime}\right)}=0$ и $$ \begin{aligned} & s: A B \rightarrow E^{\prime} F^{\prime} \\ & s^{\prime}: A^{\prime} B^{\prime} \rightarrow E F \end{aligned} $$ $$ \vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=\sum_{\substack{s: A B, \rightarrow E F \\ s^{\prime}: A B^{\prime} \rightarrow E^{\prime} F^{\prime}}} \vec{a}\left(s, s^{\prime}\right)=\sum_{\substack{s: A B \rightarrow E F \\ s^{\prime}:: A^{\prime} B^{\prime} \rightarrow E^{\prime} F^{\prime}}}-i \vec{a}(s) \vec{a}\left(s^{\prime}\right)=-i\left(\sum_{s: A B \rightarrow E F} \vec{a}(s)\right)\left(\sum_{s: A^{\prime} B^{\prime} \rightarrow E^{\prime} F^{\prime}} \vec{a}\left(s^{\prime}\right)\right) $$ откуда сразу следует ответ. Случай $x^{\prime}0$ there is $N$ such that for each integer $n>N$ we have $\left|a_{n}-a\right|<\varepsilon$. Notation: $a=\lim _{n \rightarrow \infty} a_{n}$. For instance, $\lim _{n \rightarrow \infty}\left(2-\frac{1}{2^{n}}\right)=2$. By definition, put $\sum_{k=0}^{\infty} a_{k}=\lim _{n \rightarrow \infty} \sum_{k=0}^{n} a_{k}$. Then indeed $\sum_{k=0}^{\infty} \frac{1}{2^{k}}=2$. The following generalization is called Newton's binomial theorem (allowed to use without proof): $$ (1+x)^{r}=\sum_{k=0}^{\infty} \frac{r(r-1) \cdots(r-k+1)}{k(k-1) \cdots 1} x^{k} $$ for each complex $x$ with $|x|<1$ and each real $r$, or for $x=1$ and $r>-1$. In particular, for $r=-1$ and $-\frac{1}{2}$ we get $$ \frac{1}{1-x}=\sum_{k=0}^{\infty} x^{k} \quad \text { and } \quad \frac{1}{\sqrt{1-x}}=\sum_{k=0}^{\infty} \frac{2 k(2 k-1) \cdots(k+1)}{k(k-1) \cdots 1} \frac{x^{k}}{4^{k}} $$ The following Stirling formula allows to estimate the summands in Newton's binomial theorem: $$ \sqrt{2 \pi} k^{k+1 / 2} e^{-k} \leq k(k-1) \cdots 1 \leq e k^{k+1 / 2} e^{-k} $$ Here $e$ denotes $\lim _{n \rightarrow \infty}(1+1 / n)^{n}$. It is an irrational number between 2.71 and 2.72 . ## 3 Mass Question: what is the probability to find a right electron of mass $m$ at $(x, y)$, if it was emitted from $(0,0)$ ? Assumptions: the mass and the lattice step are now arbitrary. Results: a formula for the probability for small lattice step. To check our model against experiment we need the following generalization. Fix $\varepsilon, m>0$ called lattice step and particle mass respectively. To each path $s$ of the checker, assign a vector $\vec{a}(s, m \varepsilon)$ as follows. Start with the vector $(0,1)$. While the checker moves straightly, the vector is not changed, but each time when the checker changes the direction, the vector is rotated through $90^{\circ}$[^11]clockwise and multiplied by $m \varepsilon$. In addition, at the very end the vector is divided by $\left(1+m^{2} \varepsilon^{2}\right)^{(y-1) / 2}$, where $y$ is the total number of moves. The final position of the vector is what we denote by $\vec{a}(s, m \varepsilon)$. The vectors $\vec{a}(x, y, m \varepsilon, \pm)$ and the numbers $P(x, y, m \varepsilon, \pm)$ are defined analogously to $\vec{a}(x, y, \pm)$ and $P(x, y, \pm)$, only $\vec{a}(s)$ is replaced by $\vec{a}(s, m \varepsilon)$. For instance, $P(x, y, 1,+)=P(x, y,+)$. In Figure 1 to the bottom, the color of a point $(x, y)$ with even $x+y$ depicts the value $P(x, y, 0.02,+)+P(x, y, 0.02,-)$. 13. (Puzzle) Massless and heavy particles. Find $P(x, y, 0,+)$ and define $P(x, y, \infty,+)$ for each $x, y$. 14. Solve analogues of Problems 4, 5, and 9 for $m \varepsilon \neq 1$. In an experiment, we measure the probabilities to find the electron in intervals $x_{0} \leq x \leq x_{0}+\Delta x$, $y=y_{0}$ rather than at particular points. Here $x_{0}, y_{0}, \Delta x$ are not integers but actual lengths measured in meters. If all squares have small size $\frac{1}{n} \times \frac{1}{n}$, then the interval is approximated by the collection of black squares $\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor\right)$ with $x, y$ satisfying the above (in)equalities. This leads to the following problem. ${ }^{6}$ 15. * (First main problem) Continuum limit. For each $x, y, m$ find $\lim _{n \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-\right)$ and $\lim _{n \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},+\right)$ In the answer, it is allowed to use the following expressions ${ }^{7}$ $$ J_{0}(z):=\sum_{k=0}^{\infty}(-1)^{k} \frac{(z / 2)^{2 k}}{(k!)^{2}} \quad \text { and } \quad J_{1}(z):=\sum_{k=0}^{\infty}(-1)^{k} \frac{(z / 2)^{2 k+1}}{k!(k+1)!} $$ ## 4 Source Question: what is the probability to find a right electron at $(x, y)$, if it was emitted by a source of wave length $\lambda$ ? Assumptions: the source is now realistic. ## Results: wave propagation. A realistic source does not produce electrons localized at $x=0$ (as in our game) but a rather wide wave impulse instead. For our game, this means that the checker can start from an arbitrary black square on the horizontal line $y=0$ (not too far from the origin), but the initial direction of the vector $\vec{a}(s)$ is rotated through an angle proportional to the distance from the starting square to the origin; see Figure 4. Formally, fix real $\varepsilon, \lambda>0$ and odd $\Delta$ called lattice step, wave length, and impulse width respectively. Denote by $R^{\alpha} \vec{a}$ the rotation of a vector $\vec{a}$ through the angle $|\alpha|$, which is counterclockwise for $\alpha \geq 0$ and clockwise for $\alpha<0$. Define the vector ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-088.jpg?height=369&width=576&top_left_y=1226&top_left_x=1391) Figure 4: Checker paths may now start at distinct squares $$ \vec{a}(x, y, \lambda / \varepsilon, \Delta,+):=\frac{1}{\sqrt{\Delta}} \sum_{\substack{x_{0}=1-\Delta \\ x_{0} \text { even }}}^{\Delta-1} \sum_{s} R^{2 \pi x_{0} \varepsilon / \lambda} \vec{a}(s) $$ where the second sum is over all checker paths $s$ from the square $\left(x_{0}, 0\right)$ to the square $(x, y)$, starting and ending with an upwards-right move. The length square of the vector is the probability to find a right electron at $(x, y)$, emitted by a source of wave length $\lambda$ and impulse width $\Delta$. It is denoted by $P(x, y, \lambda / \varepsilon, \Delta,+)$. Define $\vec{a}(x, y, \lambda / \varepsilon, \Delta,-)$ and $P(x, y, \lambda / \varepsilon, \Delta,-)$ analogously. For instance, $P(x, y, \lambda / \varepsilon, 1,+)=P(x, y,+)$ for each $\lambda, \varepsilon$, and $\vec{a}(x+1,1, \lambda / \varepsilon, \Delta,+)=\frac{1}{\sqrt{\Delta}}\left(-\sin \frac{2 \pi x \varepsilon}{\lambda}, \cos \frac{2 \pi x \varepsilon}{\lambda}\right)$ for even $|x|<\Delta$. 16. Let $\Delta=3, \lambda / \varepsilon=4$. Find the vector $\vec{a}(x, y, 4,3,+)$ and the probability $P(x, y, 4,3,+)$ for $y=1,2,3$ and each $x$. What is $\sum_{x \in \mathbb{Z}}(P(x, y, 4,3,+)+P(x, y, 4,3,-))$ for fixed $y=1,2,3$ ? When $P(x, 3,4,3,+)=0$ ? 17. Probability/charge conservation. Solve analogues of Problems 4 and 5 for $\vec{a}(x, y, \lambda / \varepsilon, \Delta,-)$, $\vec{a}(x, y, \lambda / \varepsilon, \Delta,+)$, and $P(x, y, \lambda / \varepsilon, \Delta,+)+P(x, y, \lambda / \varepsilon, \Delta,-)$ instead of $a_{1}(x, y), a_{2}(x, y)$, and $P(x, y)$. 18. Causality. Both $\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)$ and $\Delta P(x, y, \lambda / \varepsilon, \Delta,+)$ do not depend on $\Delta$ for $\Delta>y+|x|$. Denote $^{8} \vec{a}(x, y, \lambda / \varepsilon, \pm)=\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta, \pm)$ and $P(x, y, \lambda / \varepsilon, \pm)=\Delta P(x, y, \lambda / \varepsilon, \Delta, \pm)$ for any $\Delta>y+|x|$. 19. Wave. How to find $\vec{a}(x, 100, \lambda / \varepsilon,+)$ for all even $x$, if we know it for just one even $x$ ? 20.* Wave propagation. For each $x, y, \lambda, \varepsilon$ find $\vec{a}(x, y, \lambda / \varepsilon,-), P(x, y, \lambda / \varepsilon,-), P(x, y, \lambda / \varepsilon,-)+P(x, y, \lambda / \varepsilon,+)$.[^12] ## 5 Medium Question: which part of light of given color reflects from a glass plate of given width? Assumptions: right angle of incidence, no polarization of light; the mass now depends on $x$ but not on the color. Results: thin-film reflection. Our model can be applied also to describe propagation of light in transparent media such as glass ${ }^{9}$. Light propagates as if it had some nonzero mass inside the media, while the mass remains zero outside ${ }^{10}$. The wavelength determines the color of light. 21. (Puzzle) Define an analogue of $\vec{a}(s, m \varepsilon)$ in the case when the mass $m=m(x)$ depends on $x$ so that analogues of Problems 4 and 5 remain true. Given a mass $m=m(x)$, define $\vec{a}(x, y, m(x) \varepsilon, \lambda / \varepsilon, \pm)$ and $P(x, y, m(x) \varepsilon, \lambda / \varepsilon, \pm)$ analogously to $\vec{a}(x, y, \lambda / \varepsilon, \pm)$ and $P(x, y, \lambda / \varepsilon, \pm)$, only replace $\vec{a}(s)$ by $\vec{a}(s, m \varepsilon)$ in the definition. First we take $\varepsilon=1$. 22. One-surface reflection. Find $P(x, y, m(x), \lambda,+)$ and $P(x, y, m(x), \lambda,-)$ for $$ m(x)=m_{0}(x) \equiv 0 \quad \text { and } \quad m(x)=m_{1}(x)= \begin{cases}0.2, & \text { for } x=0 \\ 0, & \text { for } x \neq 0\end{cases} $$ Now fix odd $L>1$ called the width of a glass plate. First assume for simplicity that light is reflected only by the two surfaces of the plate and thus take ${ }^{11}$ $$ m_{2}(x)= \begin{cases}-0.2, & \text { for } x=1 \\ +0.2, & \text { for } x=L \\ 0, & \text { otherwise }\end{cases} $$ The reflection/transmission probabilities ${ }^{12}$ of light of wavelength $\lambda$ by glass plate of width $L$ are respectively $$ \begin{aligned} & P(\lambda, L,-)=\lim _{\substack{y \rightarrow+\infty \\ y \text { even }}} P\left(0, y, m_{2}(x), \lambda,-\right) \\ & P(\lambda, L,+)=\lim _{\substack{y \rightarrow+\infty \\ y \text { even }}} P\left(L+1, y, m_{2}(x), \lambda,+\right) \end{aligned} $$ 23. Plot the graph of the function $f(L)=P\left(0,2 L, m_{2}(x), 16,-\right)$. 24.* (Second main problem) Two-surface reflection. Find $P(\lambda, L,-), P(\lambda, L,+), P(\lambda, L,-)+P(\lambda, L,+)$, and $\max _{L} P(\lambda, L,-)$. In fact the light is reflected inside the plate; this should be taken into account for a more accurate computation of the reflection probability (there is no hope for an exact solution for complicated matter such as glass). For this purpose we need to modify the model essentially. Take arbitrary $\varepsilon>0$. Fix $m>0$ and let $m_{3}(x)=m$ for $0x$ express $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ through $P(x, y,+)$ and $P\left(x^{\prime}-x_{0}, y,+\right)$. 33. Exclusion principle (for intermediate states). Prove that $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ is not changed, if the sums in the definition are over only those pairs of paths $s, s^{\prime}$ which have no common moves. Define $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ analogously also for $E=(x \pm 1, y-1), E^{\prime}=\left(x^{\prime} \pm 1, y-1\right)$. 34. Probability/charge conservation. We have $\sum_{E, E^{\prime} F, F^{\prime}} P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=1$, where the sum is over all quadruples $F=(x, y), F^{\prime}=\left(x^{\prime}, y\right), E=(x \pm 1, y-1), E^{\prime}=\left(x^{\prime} \pm 1, y-1\right)$ with fixed $y \geq 1$. An electron has an antiparticle called positron. One can think of an antiparticle as a particle moving backwards in time. The motion is described by a similar model as follows. Add one more checker to the checkerboard, which now moves either downwards-right or downwards-left to the diagonal neighbors. The added checker is called black, while the one studied before is called white. For each pair of paths $s, s^{\prime}$ of the white and black checker respectively, define a vector $\vec{a}\left(s, s^{\prime}\right)$ analogously to the above, only for each turn of the black checker rotate the vector counterclockwise rather than clockwise (independently of the direction the checker turns). For instance, $\vec{a}\left(s, s^{\prime}\right)=\left(0,2^{1-y}\right)$, if $s$ and $s^{\prime}$ is the same path with opposite directions. Denote ${ }^{18}$ by $$ \vec{a}\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right):=\sum_{s, s^{\prime}} \vec{a}\left(s, s^{\prime}\right) $$ the sum is over all the pairs of a white-checker path $s$ starting with the move $A B$ and ending with the move $E F$, and a black-checker path $s^{\prime}$ starting with the move $F^{\prime} E^{\prime}$ and ending with the move $B^{\prime} A^{\prime}$. The length square $P\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right)=\left|\vec{a}\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right)\right|^{2}$ is called the probability to find an electron at $F$ and a positron at $F^{\prime}$, if they are emitted from $A$ and $A^{\prime}$. 35. Independence. Solve analogue of Problem 32 for $P\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right)$. ## 8 Interaction*** Question: the same as in the basic model but the $y$-coordinate is now interpreted as time. Assumptions: the electron now generates a magnetic field affecting the motion; encircling by waveproof walls. Results: interaction with the walls changes the motion of an electron. We have reached an unexplored area: a physically correct rigorous definition of the next upgrade is not yet known. In this section we give a WRONG definition leading to paradoxical results: - the upgrades measures the interaction of the electron with the walls rather than self-interaction; - the interaction (with the walls) propagates at infinite speed rather than the speed of light. But we hope that the introduced ideas still might be of interest. The moving electron itself generates an magnetic field, which in turn affects the motion. The generated field is random, with the probability resulting from summation over all possible intermediate fields. Fix $g \geq 0$ called the interaction constant ${ }^{19}$ (it is related to the electron charge). Fix a black square $(x, y)$ called the final position. Fix the rectangle $R$ formed by all the squares $\left(x^{\prime}, y^{\prime}\right)$ such that $1-y0$ called mass and lattice step. An initial configuration is any assignment ${ }^{22}$ of one of the signs " + " or " - " to some vertices inside $R$ lying between the lines $y=0$ and $y=1$. Physically the signs mean the initial positions of positrons and electrons respectively (and points without a sign are vacant). For our game, this means that the checkers pass the vertices with the "-" sign upwards-left or -right, and the vertices with the " + " sign - downwardsleft or -right (and do not pass through vertices without a sign). Analogously, a final configuration is an assignment to vertices between the lines $y=y_{\max }$ and $y=y_{\max }-1$. An intermediate configuration is any assignment of one of the signs " + " or "-" to some vertices inside $R$ such that the difference between the number of " + " and "-" signs on the 2 top vertices of each black square in the strip $1 \leq y \leq y_{\max }-1$ equals the difference on the 2 bottom vertices. For our game, this means that the checkers start and finish motion in the lines $y=0$ or $y=y_{\max }$ only. In other words, the signs at the vertices of each black square are in one of the 19 positions ${ }^{23}$ shown in Figure 6 to the left. These 19 positions are called basic configurations. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-093.jpg?height=634&width=1196&top_left_y=1136&top_left_x=430) Figure 6: Basic configurations Suppose that one has assigned a complex number (depending on $m \varepsilon$ ) to each of the 19 basic configurations. A "right" choice of the numbers is unknown; think of them as fixed parameters of our model. Then take any intermediate configuration. In each black square inside $R$ not having common points with the boundary, write the complex number assigned to the basic configuration in the black square. Assign the product of all the written numbers to the intermediate configuration. Now to any pair of initial and final configurations, sum the complex numbers assigned to all possible intermediate configurations between them. Assign the resulting complex number to the pair. 39. (Puzzle) Restrict to configurations without " + " signs. Let $A, A^{\prime}, B, B^{\prime}, E, E^{\prime}, F, F^{\prime}$ be the black squares defined in $\S 7$. Take any $x_{\min }<-y$ and $x_{\max }>x_{0}+y$. Fix the initial and final configurations with exactly two "-" signs, located at the top-right vertices of the squares $A, A^{\prime}$, and $E, E^{\prime}$ respectively. Assign complex numbers to the 6 basic configurations without " + " signs so that the sum of the numbers assigned to all intermediate configurations without " + " signs equals to the vector $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ from $\S 7$. A case of particular interest is when the initial configuration consists of just one " -" sign at the topright vertex of the square $(0,0)$ and the final configuration consists of just one "-" sign at the bottom-right[^17]or bottom-left vertex of a black square $\left(x, y_{\max }\right)$. The complex numbers assigned to the pairs in question are denoted by $\vec{a}\left(x, y_{\max }, m \varepsilon, x_{\min }, x_{\max },-\right)$ and $\vec{a}\left(x, y_{\max }, m \varepsilon, x_{\min }, x_{\max },+\right)$ respectively. The desired continuum limit of these complex numbers involves the following modified Bessel functions and Hankel functions: $$ \begin{aligned} K_{0}(x)=\int_{1}^{\infty} \frac{e^{-x t}}{\sqrt{t^{2}-1}} d t & H_{0}^{(1)}(x)=\frac{2}{i \pi} \int_{1}^{\infty} \frac{e^{i x t}}{\sqrt{t^{2}-1}} d t \\ K_{1}(x)=x \int_{1}^{\infty} e^{-x t} \sqrt{t^{2}-1} d t & H_{1}^{(1)}(x)=-\frac{2 x}{i \pi} \int_{1}^{\infty} e^{i x t} \sqrt{t^{2}-1} d t \end{aligned} $$ 40. ${ }^{* *}$ Continuum limit. Assign complex numbers to the 19 basic configurations so that for each $|y|<|x|$ $$ \begin{aligned} & \lim _{n \rightarrow \infty} \lim _{x_{\max } \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },-\right)=m K_{0}\left(m \sqrt{x^{2}-y^{2}}\right) \\ & \lim _{n \rightarrow \infty} \lim _{x_{\max } \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },+\right)=-i m \frac{x+y}{\sqrt{x^{2}-y^{2}}} K_{1}\left(m \sqrt{x^{2}-y^{2}}\right) \end{aligned} $$ and for each $|y|>|x|$ we have $$ \begin{aligned} & \lim _{n \rightarrow \infty} \lim _{x_{\max } \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },-\right)=i m \frac{\pi}{2} H_{0}^{(1)}\left(m \sqrt{y^{2}-x^{2}}\right) \\ & \lim _{n \rightarrow \infty} \lim _{x_{\max } \rightarrow \infty} n \vec{a}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-x_{\max }, x_{\max },+\right)=m \frac{\pi}{2} \cdot \frac{x+y}{\sqrt{y^{2}-x^{2}}} H_{1}^{(1)}\left(m \sqrt{y^{2}-x^{2}}\right) \end{aligned} $$ Let us discuss the physical meaning of the upgrade. For $y \gg|x|$ the right-hand sides of the latter two equations are very close to the right-hand sides of the answer to Problem 10. Thus one may wish to interpret the upgrade as a more accurate approximation for the probability to find the electron in a square $(x, y)$. But this faces serious objections. First, it is in principle impossible to measure the coordinates of an electron exactly ${ }^{24}$. Such a priori uncertainty has the same order of magnitude as the correction introduced by the upgrade. Thus the upgrade does not actually add anything to description of the electron motion. Second, for fixed initial configuration, the squares of the absolute values of the numbers assigned to all the possible final configurations do not sum up to 1 (even in the continuum limit). The reason is that distinct configurations are not mutually exclusive: for a field in a given configuration, there is a positive probability to find it in a different configuration. We do not know any clear explanation of that. To summarize, the upgrade lacks a direct physical interpretation, and should be considered as an ingredient for further upgrades. ## 10 (1+1)-dimensional quantum electrodynamics*** Question: what is the probability to find electrons (or electron+positron) with momenta $q$ and $q^{\prime}$ in the far future, if they were emitted with momenta $p$ and $p^{\prime}$ in the far past? Assumptions: interaction now switched on; all simplifying assumptions removed except the default ones: no nuclear forces, no gravitation, electron moves only along the $x$-axis (and $y$-coordinate is interpreted as time). Results: quantum corrections. Unifying the (so far unknown) upgrades discussed in the previous two sections would give an elementary definition of $(1+1)$-dimensional QED. ## Future research An algorithmic quantum field theory is a one which for each experimentally observable quantity and a positive number $\varepsilon$ provides a precise statement of an algorithm giving the predicted value of the quantity within accuracy $\varepsilon$. (Surely, the predicted value does not have to agree with the experiment for $\varepsilon$ less than accuracy of theory itself.) This is an extension of constructive quantum field theory, the latter currently being far away from algorithmic one.[^18] ## Epilogue (underwater rocks) We hope that at least some of our readers have become interested in elementary particles and want to learn more about them. As an epilogue, let us give a few warnings to such readers. In popular science, theory of elementary particles is usually oversimplified. This sequence of problems is not an exception. The toy models introduced here are very rough and should be considered with a grain of salt. Simplicity is their only advantage; if taken too seriously, the models could even give a wrong physical intuition. Real understanding of particles theory requires excellent knowledge of both physics and mathematics. We should also remark that nowadays there are almost no mathematical results in lattice quantum field theory; what we have is usually just a numeric simulation. Finally, there are "theories of New Physics" which are developed without any objective truth criterion: such theories are supported by neither experimental nor mathematical proofs (and some of them have experimental disproofs). ## Hints, solutions, answers For any vector $\vec{a} \in \mathbb{R}^{2}$ denote by $a_{1}, a_{2}$ the coordinates of $\vec{a}$. E.g., $\vec{a}(x, y, m \varepsilon,-)=\left(a_{1}(x, y, m \varepsilon,-), a_{2}(x, y, m \varepsilon,-)\right)$. Sometimes the vector $\vec{a}$ is considered as a complex number $a_{1}+i a_{2}$ (although complex numbers are not required for the solution of most problems). In what follows assume that $x$ and $y$ have the same parity unless the opposite is indicated. 1. Answer. In the table, the vector in the cell $(x, y)$ is $\vec{a}(x, y)$, and an empty cell $(x, y)$ means that $\vec{a}(x, y)=(0,0)$. | 4 | $\left(\frac{1}{2 \sqrt{2}}, 0\right)$ | | $\left(0,-\frac{1}{2 \sqrt{2}}\right)$ | | $\left(\frac{1}{2 \sqrt{2}},-\frac{1}{\sqrt{2}}\right)$ | | $\left(0, \frac{1}{2 \sqrt{2}}\right)$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | $\left(\frac{1}{2}, 0\right)$ | | $\left(\frac{1}{2},-\frac{1}{2}\right)$ | | $\left(0, \frac{1}{2}\right)$ | | | 2 | | | $\left(\frac{1}{\sqrt{2}}, 0\right)$ | | $\left(0, \frac{1}{\sqrt{2}}\right)$ | | | | 1 | | | | $(0,1)$ | | | | | $y / x$ | -2 | -1 | 0 | 1 | 2 | 3 | 4 | In the following table, the number in the cell $(x, y)$ is $P(x, y)$, and an empty cell $(x, y)$ means that $P(x, y)=0$. | 4 | $1 / 8$ | | $1 / 8$ | | $5 / 8$ | | $1 / 8$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | $1 / 4$ | | $1 / 2$ | | $1 / 4$ | | | 2 | | | $1 / 2$ | | $1 / 2$ | | | | 1 | | | | 1 | | | | | $\mathrm{y} / \mathrm{x}$ | -2 | -1 | 0 | 1 | 2 | 3 | 4 | (The vectors $\vec{a}(x, y)$ can be easily computed consecutively using Problem 4.) See also the following figure. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-096.jpg?height=651&width=942&top_left_y=1208&top_left_x=154) For any positive $y$ we have $\sum_{x \in \mathbb{Z}} P(x, y)=1$. (This is Problem 5.) For $-4 \leq-yy$ and $x<-y+1$ we have $P(x, y)=0$. (In general it is not known, if $P(x, y)=0$ can vanish for $x$ and $y$ of the same parity satisfying $-y|x|$ we have $$ \begin{aligned} a_{1}(x, y)=2^{(1-y) / 2} \sum_{w \equiv 2 \frac{y-x-2}{2}}(-1)^{\frac{y-x-2+2 w \cdot \operatorname{sign}(x)}{4}}\binom{|x|}{w}\binom{\frac{y-|x|-2}{2}}{\frac{y-x-2-2 w}{4}} \\ a_{2}(x, y)=2^{(1-y) / 2} \sum_{w \equiv_{2} \frac{y-x}{2}}(-1)^{\frac{y-x+2 w \cdot \operatorname{sign}(x)}{4}}\binom{|x|}{w}\binom{\frac{y-|x|-2}{2}}{\frac{y-x-2 w}{4}} \end{aligned} $$ for each $y=x>0$ we have $$ \begin{aligned} & a_{1}(y, y)=0 \\ & a_{2}(y, y)=2^{(1-y) / 2} \end{aligned} $$ for each $00$ we have $$ \begin{aligned} & a_{1}(x, y)=0 \\ & a_{2}(x, y)=0 \end{aligned} $$ Answer 2: for each $y>|x|$ we have $$ \begin{aligned} & a_{1}(x, y)=2^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r} \\ & a_{2}(x, y)=2^{(1-y) / 2} \sum_{r=1}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r-1} \end{aligned} $$ For other $y$ the answer is the same as Answer 1. Answer 3. For even $p+q$ and $2-q \leq p \leq q$ we have ${ }^{25}$ $$ \begin{aligned} & a_{1}(p, q)=(-1)^{q} 2^{(q-3) / 2} \cdot \sum_{k=\frac{q+|p|-2}{2}}^{q-2}(-2)^{-k} \cdot \frac{k!}{(q-2-k)!\left(\frac{p-q}{2}+1+k\right)!\left(k-\frac{p+q}{2}+1\right)!} \\ & a_{2}(p, q)=2^{(1-q) / 2} \cdot\binom{q-1}{\frac{q-p}{2}}-(-1)^{q} 2^{(q-3) / 2} \sum_{k=\frac{q+|p|-2}{2}}^{q-2}(-2)^{-k} \cdot \frac{k!(k+p+1)}{(q-1-k)!\left(\frac{p-q}{2}+1+k\right)!\left(k-\frac{p+q}{2}+1\right)!} \end{aligned} $$ Remark. Answer 2 shows that $a_{1}(-y+2 k, y)$ and $a_{2}(-y+2 k, y)$ are the coefficients before $t^{y-k-1}$ and $t^{y-k}$ in the expansion of $2^{(1-y) / 2}(1+t)^{y-k-1}(1-t)^{k-1}$ for $1 \leq k \leq y-1$. In particular, $$ \begin{aligned} & a_{1}(0,4 n+2)=\frac{(-1)^{n}}{2^{(4 n+1) / 2}}\binom{2 n}{n}, \quad \text { and } \quad a_{1}(0,4 n)=0 \\ & a_{2}(0,4 n+2)=0 \\ & a_{2}(0,4 n)=\frac{(-1)^{n}}{2^{(4 n-1) / 2}}\binom{2 n-1}{n} \end{aligned} $$ The sums in the expressions for $a_{1}(x, y)$ and $a_{2}(x, y)$ for $y>|x|$ are particular cases of well-known Gaussian hypergeometric functions (but familiarity with them is not required for what follows): $$ \begin{aligned} & a_{1}(x, y)=2^{(1-y) / 2}{ }_{2} F_{1}\left(1-\frac{x+y}{2}, 1+\frac{x-y}{2}, 1 ;-1\right) \\ & a_{2}(x, y)=2^{(1-y) / 2}\left(1-\frac{x+y}{2}\right){ }_{2} F_{1}\left(2-\frac{x+y}{2}, 1+\frac{x-y}{2}, 2 ;-1\right) \end{aligned} $$ Hint. Let us find $a_{1}(x, y)$ for $y>|x|$. Consider a path from the square $(0,0)$ to the square $(x, y)$ with an odd number of turns (the others do not contribute to $a_{1}(x, y)$ ). Denote by $R$ and $L$ the number of upwards-right[^19] В следующем абзаце мы докажем, что если узел лежит в плоскости, то он изотопен тривиалъному узлу. Обозначим узел, лежащий в плоскости, через $M_{1} M_{2} \ldots M_{n}$. Возьмем точку $Z$ вне этой плоскости. Переведем $M_{1} M_{2} \ldots M_{n}$ в тривиальный узел $M_{1} Z M_{n}$ следующей последовательностью элементарных движений: $$ M_{1} M_{2} \rightarrow M_{1} Z M_{2}, \quad Z M_{2} M_{3} \rightarrow Z M_{3}, \quad Z M_{3} M_{4} \rightarrow Z M_{4}, \quad \ldots, \quad Z M_{n-1} M_{n} \rightarrow Z M_{n} $$ Следуюший результат показывает, что промежуточные узлы изотопии, переводящей узел, лежащий в плоскости, в тривиальный узел, можно выбрать также лежащими в плоскости. Теорема Шёнфлиса. Любая ломаная без самопересечений в плоскости изотопна (в этой плокости) треугольнику. Это усиленная версия следующего выдающегося результата. Теорема Жордана. Любая замкнутая несамопересекающаяся ломаная $L$ в плоскости $\mathbb{R}^{2}$ разбивает эту плоскость в точности на две части, т.е. $\mathbb{R}^{2}-L-$ не связное множество, являющееся объединением двух связных множеств. Подмножество плоскости называется связным, если любые две точки этого подмножества можно соединить ломаной, в нем лежащей. Алгоритмическое объяснение почему теорема Жордана (а значит и теорема Шёнфлиса) нетривиальна, и доказательство теоремы Жордана см. в $\S 1.3$ 'Intersection number for polygonal lines in the plane' [Sk18], [Sk]. Задача 1.1.4. Предположим, что существует такая точка узла, что если мы идем вдоль узла, начиная с этой точки, то на некоторой диаграмме мы сначала встречаем только переходы, а затем только проходы. Тогда этот узел изотопен тривиальному узлу. ${ }_{4}^{4}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-124.jpg?height=236&width=569&top_left_y=1281&top_left_x=732) Рис. 4: Изменение перекрестка Изменение перекрестка - такое изменение некоторого перекрестка, что проходы становятся переходами и наоборот, см. рис. 4 . Очевидно, после любого изменения перекрестка в диаграммах трилистника и восьмерки, показанных на рис. 1 мы получим диаграмму узла, изотопного тривиальному. Лемма 1.1.5. Любая диаграмма узла может быть преобразована изменениями перекрестков в диаграмму узла, изотопного тривиальному. ${ }^{5}$ В этом тексте вместо изучения узлов с точностью до изотопии, мы будем изучать диаграммы узлов с точностью до эквивалентности, порожденной движениями Рейдемейстера (см. рис. $5^{6}$ ) и движениями изотопии плоскости (см. рис. 6 в середине и справа). Т.е. мы будем использовать без доказательства следующий результат.[^20]![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-125.jpg?height=206&width=1636&top_left_y=77&top_left_x=206) Рис. 5: Движения Рейдемейстера. Даграммы узлов совпадают вне кругов, ограниченных пунктиром. Никакие другие части диаграмм, кроме нарисованных, не пересекаются с этими кругами. (То же верно для рис. $6,4,9$ и 10.) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-125.jpg?height=210&width=1676&top_left_y=538&top_left_x=186) Рис. 6: (Слева) K строгому определению первого движения Рейдемейстера (В середине, справа) Движения изотопии плоскости Теорема 1.1.6 (Рейдемейстера). * Два узла изотопны тогда и только тогда, когда некоторая диаграмма первого узла может быть получена из некоторой диаграммы второго узла с помощью движений Рейдемейстера и движений изотопии плоскости. ## 1.2 Основные определения и результаты о зацеплениях Зацепление - это набор попарно непересекающихся узлов. Упорядоченные наборы называются упорядоченными или раскрашенными зацеплениями, в то время как неупорядоченные наборы называются неупорядоченными или нераскрашенными зацеплениями. В этом тексте под словом «зацепление» понимается «упорядоченное зацепление». ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-125.jpg?height=154&width=225&top_left_y=1625&top_left_x=495) (a) (b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-125.jpg?height=836&width=484&top_left_y=1495&top_left_x=992) Рис. 7: Зацепление Хопфа, тривиальное зацепление и еще три зацепления Тривиальное зацепление (с любым числом компонент) - зацепление, состоящее из треугольников, лежащих в параллельных плоскостях. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-126.jpg?height=238&width=263&top_left_y=106&top_left_x=411) (b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-126.jpg?height=244&width=230&top_left_y=84&top_left_x=904) (w) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-126.jpg?height=222&width=239&top_left_y=106&top_left_x=1366) (t) Рис. 8: Кольца Борромео, зацепление Уайтхеда и трилистник Диаграммы, изотопия и т.д. для зацеплений определяются аналогично узлам. Аналоги лемм 1.1.6 и 1.1.5 для зацеплений верны. Задача 1.2.1. (а) Зацепление Хопфа изотопно зацеплению, полученному из него перестановкой компонент. (b) Зацепление Хопфа изотопно некоторому зацеплению, компоненты которого симметричны относительно некоторой прямой. (c) Четвертое зацепление на рис. 7 изотопно зацеплению Уайтхеда на рис. 8.w. (d,e*) Тоже, что и в пунктах (a,b) для зацепления Уайтхеда. (f)* Кольца Борромео изотопны зацеплению, компоненты которого переставляются по циклу при повороте на угол $2 \pi / 3$ относительно некоторой прямой. Теорема 1.2.2. (а) Зачепление Хопфа не изотопно тривиальному зачеплению. (b) Зацепление Уайтхеда не изотопно тривиальному зацеплению. (c) Зацепление Хопфа не изотопно зацеплению Уайтхеда. (d) Кольца Борромео не изотопны тривиальному зачеплению. Пункты (а) и (с) доказываются с использованием коэффиииента зацепления по модулю 2, придумайте его самостоятельно или см. §1.3. Пункты (b) и (d) доказываются с использованием либо полиномов Александера-Конвея, см. §2.3, либо с помощью «тройного коэффициента зацепления» (инварианта Масси-Милнора) и «высшего коэффициента зацепления» (инварианта Сато-Левина) [Sk, §4.4-§4.6]. Пункт (d) также может быть доказан с использованием правилъных раскрасок, см. 82.2 . ## 1.3 Гауссов коэффициент зацепления по модулю 2 для диаграмм Задача 1.3.1. Пусть $A, B, C, D, E, F, O$ - это точки в пространстве, никакие четыре из которых не лежат в одной плоскости. Следующие три условия эквивалентны. (i) Контур треугольника $D E F$ пересекает часть плоскости $A B C$, ограниченную треугольником $A B C$, ровно в одной точке. (ii) Отрезок $B C$ проходит ниже (в смысле конструкции диаграммы зацепления) ровно одной из сторон $D E F$ при взгляде из точки $A$. (iii) Контур треугольника $A B C$ проходит ниже нечетного числа сторон $D E F$ при взгляде из точки $O$. Предположим, что существует изотопия между двумя двухкомпонентными зацеплениями, и вторая компонента фиксирована в процессе изотопии. Тогда след первой компоненты - самопересекающийся цилиндр, не имеющий общих точек со второй компонентой. Если после изотопии компоненты расцеплены, то цилиндр можно дополнить до самопересекающегося диска, не имеющего общих точек со второй компонентой. Это наблюдение, вместе с задачей 1.3.1 и [Sk, Projection lemma 4.2.4] мотивируют следующее определение. Коэффициентом зацепления по модулю $2 \mathrm{kk}_{2}$ плоской диаграммы двухкомпонентного зацепления называется взятое по модулю 2 число таких перекрестков диаграммы, для которых первая компонента проходит над второй компонентой. Задача 1.3.2. (a) Найдите коэффициент зацепления по модулю 2 для диаграмм на рис. 7 , для пар колец Борромео и для зацепления Уайтхеда (рис. 8). (b) Коэффициент зацепления по модулю 2 сохраняется при движениях Рейдемейстера. Согласно (b) коэффициент зацепления по модулю 2 двухкомпонентного зацепления (а также его класса изотопии) можно корректно определить как коэффициент зацепления по модулю 2 любой диаграммы этого зацепления. Будем использовать без доказательства следующую лемму о четности: любые две замкнутые ломаные на плоскости, вершины которых находятся в общем положении, пересекаются в четном числе точек. Обсуждение и доказательство этой леммы можно найти в §1.3 'Intersection number for polygonal lines in the plane' [Sk18], [Sk]. Задача 1.3.3. (a) Коэффициент зацепления по модулю 2 не изменяется при перестановке компонент зацепления. (b) Существует двухкомпонентное зацепление, не изотопное тривиальному, у которого коэффициент зацепления по модулю 2 равен нулю. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-127.jpg?height=282&width=910&top_left_y=905&top_left_x=564) Рис. 9: Узлы $K_{+}, K_{-}, K_{0}$ Обозначим через $D_{+}, D_{-}, D_{0}$ любые три диаграммы ориентированных (узлов или) зацеплений, отличающиеся, как показано на рис. 9 (см. соглашения о рисунках в примечаниии к рис. 5). Также обозначим через $K_{+}, K_{-}, K_{0}$ любые три зацепления с диаграммами $D_{+}, D_{-}, D_{0}$. Если некоторый инвариант (например, $\mathrm{lk}_{2}$ ) определен для неориентированных зацеплений (или узлов), тогда его значение для зацепления присваивается этому зацеплению с любой ориентацией. Теорема 1.3.4. Существует единственный изотопический инвариант $\mathrm{lk}_{2}$ со значениями 0 и 1 (неориентированных) двухкомпонентных зачеплений, принимающий значение 0 на тривиальном зацеплении и такой, что (для дюобъх двух зачеплений $K_{+}$и $K_{-}$, именищих диаграмми, отличаншиеся как показано на рис. 9) $$ \mathrm{lk}_{2} K_{+}-\mathrm{lk}_{2} K_{-}=\left\{\begin{array}{cc} 1 & \text { в точке пересчения двух различных компонент } \\ 0 & \text { в точке самопересечения одной из компонет } \end{array}\right. $$ Задача 1.3.5. * Если коэффициент зацепления по модулю 2 двух (непересекающихся контуров) треугольников в пространстве равен нулю, то зацепление, образованное этими треугольниками, изотопно тривиальному. Теорема 1.3.6 (Conway-Gordon-Sachs). * Если никакие 4 из 6 точек в трехмерном пространстве не леэжат в одной плоскости, то сущесвует пара зачепленных треуголъников с вершинами в этих 6 точках. Это означает, что внутренноств первого треуголъника пересекает контур второго треуголъника ровно в одной точке. ## 1.4 Арф-инвариант Рассмотрим плоскую диаграмму некоторого узла и ее точку $P$, не являющуюся перекрестком. Назовем $P$ отмеченной точкой. Неупорядоченная пара перекрестков $A, B$ называется скре- щивающейся (или $P$-скрещивающейся), если, обходя диаграмму в каком-нибудь направлении, начиная с $P$, и отмечая только перекрестки в $A$ и в $B$, мы сначала встретим переход в $A$, затем проход в $B$, затем проход в $A$ и, наконец, переход в $B$. $P$-Ap $\delta$-инвариантом $\operatorname{arf}_{P}$ плоской диаграммы называется четность числа ее $P$-скрещивающихся пар перекрестков. Задача 1.4.1. (а) Если $P$-Арф-инвариант плоской диаграммы не равен нулю, то $P$ не может быть точкой, рассмотренной в задаче 1.1.4. (b,c,d) Найдите $P$-Арф-инвариант (какой-нибудь плоской диаграммы) тривиального узла, трилистника и восьмерки (при вашем выборе отмеченной точки $P$ ). (e) $P$-Арф-инвариант не зависит от выбора отмеченной точки $P$. Согласно (е) $A p \phi$-инвариант плоской диаграммы можно корректно определить как ее $P$ Арф-инвариант для произвольной отмеченной точки $P$. (f) Арф-инвариант плоской диаграммы сохраняется при движениях Рейдемейстера. Согласно (f) $А р \oint$-инвариант arf узла (или даже изотопического класса узлов) можно корректно определить как Арф-инвариант произвольной плоской диаграммы этого узла. Задача 1.4.2. (а) Если на рис. $9 K_{+}$и $K_{-}$- это плоские диаграммы узлов, то $K_{0}$ - это плоская диаграмма двухкомпонентного зацепления и $\operatorname{arf} K_{+}-\operatorname{arf} K_{-}=\mathrm{lk}_{2} K_{0}$. (b) Существует узел, не изотопный тривиальному, Арф-инвариант которого равен нулю. Теорема 1.4.3. Для (неориентированных) узлов существует единственный изотопический инвариант arf со значениями 0 и 1, принимающий значение 0 на тривиальном узле и такой что $$ \operatorname{arf} K_{+}-\operatorname{arf} K_{-}=\mathrm{lk}_{2} K_{0} $$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-128.jpg?height=366&width=855&top_left_y=1283&top_left_x=597) Рис. 10: Пронос Задача 1.4.4. Два узла называются $А р ф$-эквивалентными, если какая-нибудь диаграмма первого узла (с некоторой ориентацией) может быть преобразована в какую-нибудь диаграмму второго узла (с некоторой ориентацией) движениями Рейдемейстера и с помощью проносов, показанными на рис. 10 . (a) Если два узла Арф-эквивалентны, то их Арф-инварианты совпадают. (b)* Восьмерка Арф-эквивалентна трилистнику. (c)* Если у двух узлов Арф-инварианты совпадают, то эти узлы Арф-эквивалентны. Теорема 1.4.5. * Рассмотрим 7 точек в пространстве, никакие 4 из которых не лежат в одной плоскости, и $\binom{7}{2}=21$ отрезков, их соединяющих. Тогда существует замкнутая ломаная, составленная из этих отрезков и не изотопная контуру треуголъника. ## 1.5 Ориентированные узлы и зацепления Вы знаете, что такое ориентированная ломаная, так что Вы также знаете, что такое ориентированный узел (рис. 11). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-129.jpg?height=246&width=566&top_left_y=83&top_left_x=739) Рис. 11: Два трилистника с противоположными ориентациями Неформальное понятие и строгое определение изотопии ориентированных узлов даются аналогично изотопии узлов. Задача 1.5.1. Изотопные ориентированные ломаные без самопересечений на плоскости и на сфере определяются аналогично изотопным ориентированным узлам в пространстве. (a) Ориентированный сферический треугольник изотопен на сфере тому же треугольнику с противоположной ориентацией. (b) Аналог пункта (a) для плоскости неверен. Задача 1.5.2. (a) Два тривиальных узла; (b) Два трилистника; (c) Две восьмерки с противоположными ориентациями изотопны. Теорема 1.5.3. (Н. Trotter, 1964) Существует ориентированный узел, не изотопный такому же узлу с противоположной ориентацией. Этот факт доказывается с использованием полиномов Джонса [PS96], [CDM12]; доказательство выходит за рамки данного текста. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-129.jpg?height=764&width=1683&top_left_y=1212&top_left_x=183) Рис. 12: Связная сумма узлов Связная сумма \# ориентированных узлов определена на рис. $12 .{ }^{7}$ Эта операция не является корректно определенной для ориентированных узлов. Так что мы обозначим через $K \# L$ любую связную сумму $K$ и $L$.[^21] Задача 1.5.4. Для любых ориентированных узлов $K, L, M$ и тривиального ориентированного узла $O$ имеем (a) $K \# O=K$. (b) $K \# L=L \# K$. (c) $(K \# L) \# M=K \#(L \# M)$. (d) $\operatorname{arf}(K \# L)=\operatorname{arf} K+\operatorname{arf} L$ (здесь узлы $K, L$ неориентированные). (Строгий смысл пункта (а) - «существует связная сумма $K$ и $O$, изотопная $K$ ». Строгое смысл пунктов (b) и (c) аналогичен. См. замечание ниже.) Замечание. Класс изотопии узла - это множество узлов, ему изотопных. Ориентированный класс изотопии $[K \# L]$ связной суммы двух ориентированных классов изотопии $[K],[L]$ ориентированных узлов $K, L$ не зависит от выборов, сделанных в процессе построения, и от выбора представителей $K, L$ классов $[K],[L]$. Следовательно, связная сумма ориентированных классов изотопии ориентированных узлов корректно определена как $[K] \#[L]:=$ $[K \# L]$, см. [Sk15p, Remark 2.3.a]. Для классов изотопии неориентированных узлов связная сумма не является корректно определенной операцией. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-130.jpg?height=604&width=990&top_left_y=777&top_left_x=535) Рис. 13: Связная сумма зацеплений Связная сумма \# зацеплений (упорядоченных или нет, ориентированных или нет) определяется аналогично связной сумме узлов, см. рис. 13. Это операция не является корректно определенной для зацеплений и задача 1.5.6 показывает, что операция также не является корректно определенной для классов изотопии зацеплений. Так что мы обозначим через $K \# L$ любую связную сумму $K$ и $L$. Задача 1.5.5. (a,b,c,d) Докажите аналоги утверждений из задачи 1.5.4.a,b,c,d для зацеплений. Задача 1.5.6. Существует две изотопные пары $(K, L)$ и ( $\left.K^{\prime}, L^{\prime}\right)$ (a) неупорядоченных; (b)* упорядоченных двухкомпонентных зацеплений (ориентированных или нет) таких, что некоторые связные суммы $K \# L$ и $K^{\prime} \# L^{\prime}$ не изотопны. ## 1.6 Гауссов коэффициент зацепления для плоских диаграмм Пусть $(\overrightarrow{A B}, \overrightarrow{C D})$ - упорядоченная пара векторов (ориентированных отрезков) на плоскости, пересекающихся в точке $P$. Определим знак точки $P$ как +1 если $A B C$ ориентирован по часовой стрелке и как -1 иначе (рис. 14). Коэффициент зацепления $\mathrm{lk}$ диаграммы ориентированного двухкомпонентного зацепления - сумма знаков всех точек пересечения на диаграмме, в которых первая компонента проходит над второй компонентой. В каждом перекрестке первый (второй) вектор - это звено первой (второй) компоненты. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-131.jpg?height=220&width=461&top_left_y=80&top_left_x=797) Рис. 14: Знак точки пересечения Задача 1.6.1. (a,b) Найдите коэффициент зацепления для (некоторой диаграммы) зацепления Хопфа и пары колец Борромео, выбрав ориентацию компонент самостоятельно. (c) Коэффициент зацепления сохраняется при движениях Рейдемейстера. Из (с) следует, что коэффициент зацепления ориентированного двухкомпонентного зацепления (или его класса изотопии) - корректно определяется как коэффициент зацепления любой плоской диаграммы зацепления. Модуль коэффиииента зацепления (неориентированного) двухкомпонентного зацепления (или его класса изотопии) корректно определен не зависимо от ориентации компонент. Будем использовать без доказательства следующую лемму о тривиальности: для любых двух замкнутых ломаных в плоскости, вершины которых находятся в общем положении, сумма знаков их точек пересечения равна нулю. Обсуждение и доказательство см. §1.3 'Intersection number for polygonal lines in the plane' в [Sk18], [Sk]. Задача 1.6.2. (а) Перестановка компонент сохраняет коэффициент зацепления или меняет его знак на противоположный? (b) Изменение ориентации любой компоненты меняет знак коэффициента зацепления на противоположный. (c) Нарисуйте ориентированное двухкомпонентное зацепление с коэффициентом зацепления -5 . (d) Для любой связной суммы $K \# L$ ориентированных двухкомпонентных зацеплений $K, L$ имеем $\operatorname{lk}(K \# L)=\mathrm{lk} K+\mathrm{lk} L$. (е) Существует двухкомпонентное зацепление, не изотопное тривиальному зацеплению, коэффициент зацепления которого равен нулю. Теорема 1.6.3. Существует единственный целочисленный изотопический инвариант $\mathrm{lk}$ ориентированных двухкомпонентных зачеплений, который равен 0 для тривиального зачепления, такой, что $$ \mathrm{lk} K_{+}-\mathrm{lk} K_{-}= \begin{cases}1 & \text { для перекрестков, в которых пересекаются разные компоненты; } \\ 0 & \text { для перекрестков, в которых одна компонента пересекает себя. }\end{cases} $$ Задача 1.6.4. Два зацепления (ориентированные или нет) называются сингулярно гомотопными, если одно зацепление можно непрерывно продеформировать в другое так, что во время деформации разные компоненты не пересекаются (но могут появлятся самопересечения компонент). Или, эквивалентно, некоторая плоская диаграмма первого зацепления может быть преобразована в некоторую плоскую диаграмму второго зацепления с использованием движений Рейдемейстера и изменений перекрестка для одной компоненты. (a) Зацепление Уайтхеда сингулярно гомотопно тривиальному зацеплению. (b) Если два ориентированных двухкомпонентных зацепления сингулярно гомотопны, тогда их коэффициенты зацепления равны. (c)* Если коэффициенты зацеплений двух двухкомпонентных зацеплений равны, то эти зацепления сингулярно гомотопны. (d)* Кольца Борромео сингулярно не гомотопны тривиальному зацеплению. ## 2 Задачи после промежуточного финиша ## 2.1 Инвариант Кэссона Знак перекрестка ориентированной плоской диаграммы узла определен после рисунка 14; первый (второй) вектор - это вектор перехода (прохода). Очевидно, знак не зависит от ориентации диаграммы и таким образом определен для неориентированных диаграмм. Знак $P$-скрещивающейся пары перекрестков на плоской диаграмме узла (для любой отмеченной точки $P$ ) - это произведение знаков соответствующих перекрестков. $P$-инвариант Кэссона плоской диаграммы - это сумма знаков всех $P$-скрещивающихся пар перекрестков. Задача 2.1.1. (a) Нарисуйте плоскую диаграмму узла и отмеченную точку $P$ такие, что $P$-инвариант Кэссона плоской диаграммы равен -5 . (b,c,d,e,f) То же, что и в задачах 1.4.1.b,c,d,e,f для инварианта Кэссона. (g) Сформулируйте и докажите аналог задачи 1.4.2.а для инварианта Кэссона. $(\mathrm{h}, \mathrm{i})$ То же, что и в задачах 1.4.2.b и 1.5.4.d для инварианта Кэссона. Из (e,f) следует, что инвариант Кэссона (число Кэссона) с п плоской диаграммы узла (или даже изотопического класса узла) является корректно определенным инвариантом, если положить его равным $P$-инварианту Кэссона любой плоской диаграммы этого узла с любой отмеченной точкой $P$. Теорема 2.1.2. Существует единственный целочисленный изотопический инвариант $c_{2}$ для (неориентированных) узлов, который равен 0 для тривиалъного узла и для которого $$ c_{2}\left(K_{+}\right)-c_{2}\left(K_{-}\right)=\operatorname{lk} K_{0} $$ ( Число lk $K_{0}$ определено, поскольку изменение ориентации обеих компонент ориентированного зацепления не меняет коэффиииент зац,еления.) ## 2.2 Правильные раскраски Дугой на плоской диаграме (узла или зацепления) называется связный фрагмент, идущий от одного прохода до другого. Правильная раскраска плоской диаграммы (узла или зацепления) - это такая раскраска ее дуг в три цвета, что по крайней мере два цвета использовано, и в каждом перекрестке встречаются либо все три цвета, либо только один цвет. Плоская диаграмма (узла или зацепления) называется раскрашиваемой в три цвета если ее можно правильно раскрасить. Задача 2.2.1. Для каждого из следующих узлов или зацеплений возьмите любую плоскую диаграмму и определите, является ли она раскрашиваемой в три цвета: (a) тривиальный узел; (b) трилистник; (c) восьмерка. (d-j) зацепления на рис. 7 и 8 . Задача 2.2.2. (a) Раскрашиваемость в три цвета плоской диаграммы сохраняется при движениях Рейдемейстера. (b) Ни одно из зацеплений на рис. 7 и 8 (кроме тривиального зацепления) не изотопно тривиальному зацеплению. Задача 2.2.3. Узел 51 не изотопен тривиальному узлу. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-133.jpg?height=324&width=323&top_left_y=63&top_left_x=858) Рис. 15: Узел 51 ## 2.3 Полиномы Александера-Конвея Задача 2.3.1. * (a) Существует единственный изотопический инвариант трехкомпонентных ориентированных зацеплений arf со значениями 0 и 1 , принимающий значение 0 на тривиальном зацеплении и такой, что $$ \operatorname{arf}\left(K_{+}\right)-\operatorname{arf}\left(K_{-}\right)= \begin{cases}\mathrm{kk}_{2} K_{0} & \text { в точке пересечения различных компонент; } \\ 0 & \text { в точке самопересечения одной из компонент. }\end{cases} $$ (Здесь $\mathrm{lk}_{2} K_{0}$ определено, так как $K_{0}$ - двухкомпонентное зацепление.) $)^{8}$ (b) Существует единственный изотопический инвариант двухкомпонентных ориентированных зацеплений $a_{3}$ со значениями 0 и 1 , принимающий значение 0 на тривиальном зацеплении и такой, что $$ a_{3}\left(K_{+}\right)-a_{3}\left(K_{-}\right)=\operatorname{arf} K_{0} $$ (Здесь arf $K_{0}$ определено, так как $K_{0}$ либо узел, либо трехкомпонентное зацепленине.) (c) Существует единственный изотопический инвариант четырехкомпонентных ориентированных зацеплений $a_{3}$ со значениями 0 и 1 , принимающий значение 0 на тривиальном зацеплении и такой, что $$ a_{3}\left(K_{+}\right)-a_{3}\left(K_{-}\right)= \begin{cases}\operatorname{arf} K_{0} & \text { в точке пересечения различных компонент; } \\ 0 & \text { в точке самопересечения одной из компонент. }\end{cases} $$ (Здесь arf $K_{0}$ определено, так как $K_{0}$ - трехкомпонентное зацепленине.) Доказательство существования в задачах 2.3.1 и теореме 2.3.3 не приводится в этом тексте. Простое доказательство можно найти в [Ka06', §2-§5], [Ka06]. Вы можете заработать плюс, доказав единственность, и решать другие задачи, предполагая существование доказанным. Связь с правильными раскрасками см. в $\S 6$ [Ka06']. Задача 2.3.2. Найдите (сами выбрав ориентацию компонент) (a) arf инвариант колец Борромео; $\left(\mathrm{b}, \mathrm{c}, \mathrm{d}^{*}\right) a_{3}$ инвариант зацепления Хопфа, зацепления Уайтхеда и четырехкомпонентного зацепления Борромео, т.е. такого четырехкомпонентного зацепления, для которого любое трехкомпонентное подзацепление изотопно тривиальному зацеплению, но само зацепление не изотопно тривиальному. Теорема 2.3.3. * (а) Существует единственная бесконечная последовательность $c_{-1}=$ $0, c_{0}, c_{1}, c_{2}, \ldots$ целочисленных изотопических инвариантов ориентированных неупорядоченных зацеплений, принимающая значения $c_{0}=1$ и $c_{1}=c_{2}=\ldots=0$ на тривиальном зацеплении, $и$ такая, что $$ c_{n}\left(K_{+}\right)-c_{n}\left(K_{-}\right)=c_{n-1}\left(K_{0}\right) $$[^22]для любого $n \geq 0$. (b) Существует единственная бесконечная последовательность $c_{-1}=0, c_{0}, c_{1}, c_{2}, \ldots$ целочисленных изотопических инвариантов ориентированных упорядоченных зацеплений, принимающая значения $c_{0}=1$ и $c_{1}=c_{2}=\ldots=0$ на тривиальном зацеплении, и такая, что для любого $n \geq 0$ имеем $$ c_{n}\left(K_{+}\right)-c_{n}\left(K_{-}\right)=c_{n-1}\left(K_{0}\right) $$ где $K_{0}$ - это $K_{0}$ с рис. 9 с некоторым упорядочением компонент. Вообще-то две версии теоремы 2.3.3 эквивалентны. Вы можете использовать теорему 2.3.3.b без доказательства. ${ }^{9}$ Многочлен $C(K)(t):=c_{0}(K)+c_{1}(K) t+c_{2}(K) t^{2}+\ldots$ называется многочленом Конвея, см. задачу 2.3.5.е. Введение этого многочлена позволяет находить все инварианты $c_{n}$ так же быстро, как и один из них. Формула из теоремы 2.3.3 эквивалентна следующему равенству $$ C\left(K_{+}\right)-C\left(K_{-}\right)=t C\left(K_{0}\right) $$ Задача 2.3.4. Найдите многочлены Конвея следующих зацеплений (выберите сами ориентацию для каждой компоненты). (a) тривиальное зацепление с двумя компонентами; (b) тривиальное зацепление с $n$ компонентами; (c) зацепление Хопфа; (d) трилистник; (e) восьмерка; (f) зацепление Уайтхеда; (g) кольца Борромео; (h) узел $5_{1}$. Задача 2.3.5. (a) Имеет место равенство $c_{0}(K)=1$ если $K$ узел, и $c_{0}(K)=0$ в противном случае (т.е. если $K$ имеет более одной компоненты). (b) Если $K$ - узел, то $c_{2 j+1}(K)=0$, и $c_{2}$ это инвариант Кэссона. (c) Если $K$ - двухкомпонентное зацепление, то $c_{2 j}(K)=0$, и $c_{1}$ это коэффициент зацепления. (d) Если $K-k$-компонентное зацепление, то $c_{j}(K)=0$, если или $j \leq k-2$, или $j-k$ чётно. (е) Для каждого узла или зацепления все инварианты $c_{n}$, кроме конечного числа, нулевые. Задача 2.3.6. (а) Изменение ориентации всех компонент зацепления (в частности, изменение ориентации узла) сохраняет многочлен Конвея. (b) Существует двухкомпонентное зацепление такое, что изменение ориентации одной его компоненты меняет степень многочлена Конвея (в частности, такое изменение ни сохраняет многочлен Конвея, ни меняет его знак). (с) Для любой связной суммы $K \# L$ узлов $K, L$ верно $C(K \# L)=C(K) C(L)$. Зацепление называется разделяемым, если оно изотопно зацеплению, компоненты которого содержатся в непересекающихся шарах. Задача 2.3.7. (a,b,c) Ни одно зацепление из теоремы 1.2.2 не является разделяемым. (d) Многочлен Конвея разделяемого зацепления тривиален.[^23] ### 2.4 Vassiliev-Goussarov invariants (sketch)* Denote by - $\Sigma$ the set of isotopy classes of singular knots [PS96, 4.1], - $\delta_{n}$ the set of all chord diagrams that have $n$ chords [PS96, 4.8]; - $\sigma(K)$ the chord diagram of a singular knot $K$ [PS96, 4.8], [CDM12, 3.4.1] (not to be confused with Gauss diagrams for a non-singular knot $K$ [CDM12, 1.8.4]). Теорема 2.4.1 (Vassiliev-Kontsevich, [PS96], [CDM12]). For any map $\lambda: \delta_{n} \rightarrow \mathbb{R}$ there exists a map $v: \Sigma \rightarrow \mathbb{R}$ having properties (1)-(3) below if and only if $\lambda$ satisfies to the 1-term and the 4 -term relations $[P S 96,(4.5),(4.6)]$. (1) For any singular knots $K_{+}, K_{-}$and $K^{0}$ from [PS96, (4.1)] (notice the difference with fig. 9) we have $$ v\left(K_{+}\right)-v\left(K_{-}\right)=v\left(K^{0}\right) $$ $\left(2_{n}\right) v(K)=0$ for each singular knot that has more than $n$ double points, and (3) $v(K)=\lambda(\sigma(K))$ for each singular knot that has exactly $n$ double points. A map $v: \Sigma \rightarrow \mathbb{R}$ such that (1) holds is called a Vassiliev-Goussarov invariant. A map $v: \Sigma \rightarrow \mathbb{R}$ such that $\left(2_{n}\right)$ holds is called a map of order at most $n$. Задача 2.4.2. (a) The map $v$ of theorem 2.4.1 is unique up to Vassiliev-Goussarov invariant of order at most $n-1$. More precisely, the difference between maps $v, v^{\prime}: \Sigma \rightarrow \mathbb{R}$ satisfying to (1), (2n) and (3), satisfies to (1) and $\left(2_{n-1}\right)$. (b) Prove the 'only if' part of theorem 2.4.1. $(0),(1),(2),(3)^{*}$ Prove the 'if' part of theorem 2.4 .1 for $n=0,1,2,3$. Hint: for $n=2$ use theorem 2.1.2, for $n=3$ use the coefficient of $h^{3}$ in $J\left(e^{h}\right)$, where $J$ is the Jones polynomial in $t$-parametrization [CDM12, 2.4.2, 2.4.3]. In the remaining problems theorem 2.4.1 can be used without proof. Assertion ' $v(K)=x$ for any singular knot $K$ whose chord diagram is $a$ ' is shortened to ' $v(a)=x^{\prime}$. Задача 2.4.3. (a) There exists a unique Vassiliev-Goussarov invariant $v_{2}: \Sigma \rightarrow \mathbb{R}$ of order at most 2 such that - $v_{2}(O)=0$ for the trivial knot $O$, and - $v_{2}(1212)=1$ ((1212) is the 'non-trivial diagram with 2 chords' [PS96, Figure 4.4], 3rd diagram of the first line). Warning: in this problem it is allowed to use theorem 2.4.1 but not theorem 2.1.2. (b,b',c,d) Calculate $v_{2}$ for the right trefoil, left trefoil, figure eight knot and the $5_{1}$ knot. Задача 2.4.4. (a) There exists a unique Vassiliev-Goussarov invariant $v_{3}: \Sigma \rightarrow \mathbb{R}$ of order at most 3 such that - $v_{3}(O)=0$ for the trivial knot $O$ and for the left trefoil $O$, and - $v_{3}(123123)=1$ ((123123) is the 'non-trivial most symmetric diagram with 3 chords', [PS96, Figure 4.4], 5th diagram of the second line). $\left(\mathrm{b}, \mathrm{c}, \mathrm{d}^{*}\right)$ Calculate $v_{3}$ for the right trefoil, figure eight knot and the $5_{1}$ knot. Hints: Problems 2, 3, 4ab, Results/Theorems 11, 13, 14 from [PS96, §4]. Задача 2.4.5. (a) There exists a unique Vassiliev-Goussarov invariant $v_{4}: \Sigma \rightarrow \mathbb{R}$ of order at most 4 such that - $v_{4}(O)=0$ for the trivial knot $O$, for the left trefoil $O$, and for the right trefoil $O$, - $v_{4}(12341234)=2, v_{4}(12341432)=3$ and $v_{4}(12341423)=5$ [PS96, Problem 4.4.b]. $\left(c^{*}, \mathrm{~d}^{*}\right)$ Calculate $v_{4}$ for the figure eight knot and the $5_{1}$ knot. ## Указания и решения к задачам до промежуточного финиша 1.1.1. (a,b,c,d) См. [Pr95]. Проще всего изготовить трилистник и восьмерку из веревки или шнурка, а затем попытаться получить из этих узлов все узлы, изображенные на рисунке 1 . Выполнить некоторые преобразования узла восьмерка вам поможет рис. 16 . ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-136.jpg?height=416&width=720&top_left_y=358&top_left_x=665) Рис. 16: Изотопия восьмерки (е) Рассмотрим два узла с совпадающими плоскими диаграммами в горизонтальной плоскости $\pi$. Для каждой точки $X$ в пространстве пусть $p(X)$ - прямая, проходящая через $X$ и перпендикулярная $\pi$. Пусть $h(X)$ - высота $X$ относительно $\pi$, положительная ( $h(X)>0$ ) если $X$ находится в верхнем полупространстве и отрицательная $(h(X)<0$ ) если $X$ находится в нижнем полупространстве. Каждой точке $A$ первого узла поставим в соответствие точку $A^{\prime}$ второго узла с помощью следующей процедуры. Рассмотрим два случая: Случай 1: Проекиия точки А на те является перекрестком на плоской диаграмме. В этом случае прямая $p(A)$ пересекает первый узел только в точке $A$. Так как плоские диаграммы узлов совпадают, то прямая $p(A)$ пересекает второй узел также только в одной точке. Обозначим эту точку через $A^{\prime}$. Случай 2: проекиия точки А на $\pi$ является перекрестком на плоской диаграмме. В этом случае прямая $p(A)$ пересекает первый узел еще в одной точке $B \neq A$. Так как плоские диаграммы узлов совпадают, то прямая $p(A)$ пересекает второй узел тоже в двух точках: $C$ и $D$. Без ограничения общности можно считать что $h(C)>h(D)$. Тогда если $h(A)>h(B)$, то положим $A^{\prime}=C$, иначе $A^{\prime}=D$. Для каждой точки $A$ первого узла и каждого числа $t \in[0,1]$ пусть $A(t)$ - точка на прямой $p(A)$ на высоте $h(A(t))=(1-t) h(A)+t h\left(A^{\prime}\right)$. По построению $A(0)=A, A(1)=A^{\prime}$, и преобразование первого узла, двигающее точку $A(0)$ к точке $A(1)$ с постоянной скоростью так, что в момент времени $t$ она находится в положении $A(t)$, является искомой изотопией. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-136.jpg?height=317&width=528&top_left_y=1869&top_left_x=755) Рис. 17: Мостик через перекресток 1.1.2. Для каждого перекрестка на плоской диаграме выберем на участке перекрестка, являющемся переходом, две точки, близкие к пересечению и находящиеся по разные стороны от него. Заменим отрезок между двумя выбранными точками «мостиком», соединяющим выбранные точки и «подымающимся» над плоской диаграмой (см. рис. 17.). После замены всех перекрестков такими «мостиками» мы получим искомый узел. 1.1.3. (а) Используйте результаты задач 1.4.1, 2.1.1, или 2.2.1. (b) Используйте результаты задач 1.4.1 или 2.1.1. (c) Используйте результат задачи 2.2.1. (d) Рассмотрим любую связную сумму $n$ копий трилистника. Согласно утверждениям 2.1.1.c,i инвариант Кэссона такого узла равен $n$. Следовательно, для различных значений $n$ соответствующие узлы не изотопны. 1.1.4. Выберем узел, проецирующийся на данную плоскую диаграмму тем же способом, что и в утверждении 1.1.2. Будем считать, что все «мостики» лежат в верхнем полупространстве относительно плоскости проекции. По предположению существуют точки $X$ и $Y$ узла, которые разбивают узел на две ломаные $p$ и $q$ такие, что - $q$ лежит на плоскости проекции и проходит только через проходы; - $p$ проецируется на ломаную $p^{\prime}$, которая проходит только через переходы. Возьмем точку $Z$ в верхнем полупространстве и точку $T$ в нижнем полупространстве. Построим изотопию между заданным узлом и замкнутой ломаной $X Z Y T$, которая изотопна тривиальному узлу. Построение изотопии состоит из трех шагов, в каждом из которых точки $X, Y$ остаются фиксированными. Шаг 1. Изотопия между q и $X T Y$. Предположим, что $q=A_{0} A_{1} \ldots A_{n}$, где $A_{0}=X$ и $A_{n}=Y$. Тогда изотопия получается последовательностью элементарных движений $$ A_{0} A_{1} \rightarrow A_{0} T A_{1}, \quad T A_{1} A_{2} \rightarrow T A_{2}, \quad T A_{2} A_{3} \rightarrow T A_{3}, \quad \ldots T A_{n-1} A_{n} \rightarrow T A_{n} $$ Шаг 2. Изотопия между p и p'. Уберем все «мостики» с помощью элементарных движений. Шаг 3. Изотопия между $p^{\prime}$ и XZY. Строится аналогично шагу 1. 1.1.5. Следует из утверждения 1.1.4. Другая идея доказательства (см. теорему 3.8 в [PS96]). Обозначим через $\pi$ горизонтальную плоскость, содержащую диаграмму. Для каждой точки $X$ в пространстве $p(X)$ и $h(X)$ определеяются как в решении задачи 1.1.1.е. Пусть $l$ - прямая на плоскости, которая проходит через вершину $A_{0}$ плоской диаграмы, и пусть вся диаграмма лежит по одну сторону от прямой $l$. Пусть $A_{0}, A_{1}, \ldots, A_{n}$ - все вершины плоской диаграмы в порядке их появления при движении вдоль диаграммы в некотором направлении. Выберем точки $B_{0}, \ldots, B_{n}$ так, что $A_{i} \in p\left(B_{i}\right)$ для $i=1, \ldots, n$, и $h\left(B_{i}\right)h\left(B_{n}\right)$. Тогда узел $B_{0} \ldots B_{n} B_{n+1}$ изотопен тривиальному узлу. Действительно, по построению прямой $l$, проекция узла на любую плоскость, перпендикулярную $l$, является замкнутой ломаной без самопересечений. Остается только изменить перекрестки плоской диаграммы так, чтобы они согласовывались с проекцией построенного узла на плоскость $\pi$. 1.1.6. См. $\S 1.7$ в [PS96]. Замечание. Так как в $\S 1.6$ [PS96] нет такого же строгого определения для движений Рейдемейстера как для движений изотопии плоскости, ${ }^{10}$ рассуждения в $\S 1.7$ [PS96] не являются[^24]строгим доказательством. По-видимому, строгое доказательство может быть получено с использованием строгого определения движений Рейдемейстера. 1.2.1. (a) Следует из (b) (или может быть доказано независимо). (d) Следует из (e) (или может быть доказано независимо). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-138.jpg?height=466&width=1178&top_left_y=339&top_left_x=431) Рис. 18: Изотопия зацепления Уайтхеда (e) См. рис. 18 . (f) Возьмем три эллипса, заданных системами уравнений $$ \left\{\begin{array}{c} x=0 \\ y^{2}+2 z^{2}=1 \end{array}, \quad\left\{\begin{array} { c } { y = 0 } \\ { z ^ { 2 } + 2 x ^ { 2 } = 1 } \end{array} \quad \text { и } \quad \left\{\begin{array}{c} z=0 \\ x^{2}+2 y^{2}=1 \end{array}\right.\right.\right. $$ См. рис. 19. Возьмем четырехугольники, описанные около этих эллипсов и симметричные относительно координатных осей. Искомая прямая задается уравнением $x=y=z$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-138.jpg?height=758&width=1590&top_left_y=1342&top_left_x=232) Рис. 19: Кольца Борромео ### 1.2.2. (а,с) Используйте результаты задачи 1.3.2. $(\mathrm{b}, \mathrm{d})$ Используйте, например, результаты задачи 2.2.1 и 2.2.2. 1.3.1. Обозначим через $\langle A B C\rangle$ внутренность треугольника $A B C$. (i $\Leftrightarrow$ ii) Отрезок $X Y$ проходит над отрезком $B C$ при взгляде из точки $A$, если и только если $X Y$ пересекает $\langle A B C\rangle$. Контур треугольника $D E F$ - это объединение отрезков $D E, E F$ и $D F$. Поэтому і $\Leftrightarrow \mathrm{ii}$. (i $\Leftrightarrow$ iii) Контур треугольника $D E F$ либо не пересекается с поверхностью тетраэдра $O A B C$, либо пересекается с ней ровно в двух точках. Это влечет эквивалентность следующих трех свойств: - контур треугольника $D E F$ пересекает $\langle A B C\rangle$ ровно в одной точке; - контур треугольника $D E F$ пересекает в точности один из треугольников $\langle O A B\rangle,\langle O B C\rangle$ или $\langle O A C\rangle$. - контур треугольника $A B C$ проходит в точности под одной из сторон $D E F$ при взгляде из точки $O$. 1.3.2. (а) Ответы (слева направо): $1,0,0,0$. (b) Докажите утверждение отдельно для каждого движения Рейдемейстера. При движениях I и III число перекрестков, в которых первая компонента проходит над второй, не изменяется. При движении II это число изменяется на 0 или $\pm 2$. 1.3.3. (а) Рассмотрим плоскую диаграмму зацепления. По лемме о четности (сформулированной перед задачей 1.3.3) число таких перекрестков, где первая компонента проходит над второй, имеет ту же четность что и число перекрестков, где вторая компонента проходит над первой. Это завершает доказательство. (b) Примером является четвертое зацепление на рис. 7. Используя коэффициент зацепления, можно доказать, что это зацепление не изотопно тривиальному, см. §1.6. 1.3.4. Предположим, что $\mathrm{f}$ - другой инвариант, отличный от $\mathrm{lk}_{2}$ и удовлетворяющий предположениям. Тогда $\mathrm{f}-\mathrm{lk}_{2}$ - изотопический инвариант, равный нулю на тривиальном зацеплении и инвариантный относительно изменений перекрестков. Согласно аналогу леммы 1.1.5 для зацеплений любая плоская диаграмма зацепления может быть получена из диаграммы зацепления, изотопного тривиальному, при помощи изменений перекрестков. Следовательно, $\mathrm{f}-\mathrm{lk}_{2}=0$. 1.3.5. Доказательство не должно представлять трудности, и мы рекомендуем вам восстановить его детали самостоятельно. 1.3.6. См. Теорему 1.1 в $\S 1[\operatorname{Sk} 14]$. 1.4.1. (а) Если $P$ - такая точка на плоской диаграмме, как описано в задаче 1.1.4, то не существует $P$-скрещивающихся пар перекрестков. Следовательно, $P$-Арф-инвариант равен нулю. (b) Oтвет 0. Тривиальный узел не имеет перекрестков и скрещивающихся пар перекрестков. Поэтому Арф-инвариант этого узла равен нулю при любом выборе отмеченной точки. (c) Oтвет: 1. Трилистник имеет 3 перекрестка. Для любой отмеченной точки $P$ ровно одна пара перекрестков является $P$-скрещивающейся. Следовательно, $P$-Арф-инвариант трилистника равен 1 . (d) Oтвem: 1 . (е) Достаточно показать, что Арф-инвариант не изменяется, когда отмеченная точка, двигаясь вдоль плоской диаграммы, пересекает один перекресток. Пусть $P_{1}$ и $P_{2}$ - две отмеченные точки такие, что отрезок $P_{1} P_{2}$ содержит в точности один перекресток $X$. Рассмотрим два случая. Случай 1: $P_{1} P_{2}$ участвует в проходе. Тогда $X$ не образует ни $P_{1}$-скрещивающейся, ни $P_{2}$ скрещивающейся пары ни с одним другим перекрестком. Поэтому $P_{1}$ - и $P_{2}$-Арф-инварианты диаграммы равны. Случай 2: $P_{1} P_{2}$ участвует в переходе. Тогда $X$ делит диаграмму на две замкнутые ломаные $q_{1}$ и $q_{2}$ такие, что $P_{1}$ лежит на $q_{1}$, а $P_{2}$ лежит на $q_{2}$. Обозначим через $n_{1}$ (соответственно, $n_{2}$ ) число всех таких перекрестков на $q_{1}$ и $q_{2}$, в которых $q_{1}$ проходит над $q_{2}$ (соответственно, $q_{2}$ проходит над $q_{1}$ ). Обозначим через $N_{1}$ (соответственно, $N_{2}$ ) количество $P_{1}$-скрещивающихся (соответственно, $P_{2}$-скрещивающихся) пар, образованных $X$ и некоторым пересечением $q_{1}$ и $q_{2}$. Тогда $$ \operatorname{arf}_{P_{1}} D-\operatorname{arf}_{P_{2}} D=N_{1}-N_{2}=n_{1}-n_{2} \overline{\overline{2}} n_{1}+n_{2} \overline{\overline{2}}= $$ где $D$ - данная плоская диаграмма. В последней формуле - первое равенство выполнено поскольку пара перекрестков в $D$ является $P_{1}$-скрещивающейся или $P_{2}$-скрещивающейся (но не одновременно!) тогда и только тогда когда она образована $X$ и некоторым пересечением $q_{1}$ и $q_{2} ;$ - второе равенство выполнено поскольку $N_{1}=n_{1}$ и $N_{2}=n_{2}$; действительно, точка пересечения $q_{1}$ и $q_{2}$ образует $P_{1}$-скрещивающуюся (соответственно, $P_{2}$-скрещивающуюся) пару с $X$ тогда и только тогда когда в этой точке $q_{1}$ проходит над (соответственно, под) $q_{2}$; - 三 обозначает сравнение по модулю 2 ; - последнее сравнение выполнено ввиду леммы о четности, примененной к ломаным $q_{1}$ и $q_{2}$. (f) Докажем утверждение отдельно для каждого движения Рейдемейстера. Основная идея состоит в удачном выборе отмеченной точки. Движжение типа I. Выберем отмеченные точки до и после движения, как показано на рис. 20 (слева). Проверьте, что перекресток $A$ не образует $P$-скрещивающуюся пару ни с одним другим перекрестком. Движение muna II. Выберем отмеченные точки до и после движения, как показано на рис. 20 (в центре). Проверьте, что ни один из перекрестков $A$ и $B$ не образует $P$-скрещивающуюся пару ни с одним другим перекрестком. Движение muna III. Выберем отмеченные точки до и после движения, как показано на рис. 20 (справа). Проверьте, что ни один из перекрестков $A, B$ не образует $P$-скрещивающуюся пару ни с одним другим перекрестком, и что ни один из перекрестков $A^{\prime}, B^{\prime}$ не образует $P^{\prime}$ скрещивающуюся пару ни с одним другим перекрестком. Затем проверьте, что перекресток $X$, отличный от $A, B, C$, образует $P$-скрещивающуюся пару с $C$, если и только если $X$ образует $P^{\prime}$-скрещивающуюся пару с $C^{\prime}$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-140.jpg?height=231&width=1635&top_left_y=1408&top_left_x=207) Рис. 20: Arf-инвариант не меняется при движениях Рейдемейстера 1.4.2. (а) Выберем отмеченные точки $P_{+}, P_{-}$, как показано на рис. 21. Проверьте, что перекресток $A_{-}$не образует $P_{-}$-скрещивающуюся пару ни с одним другим перекрестком в $K_{-}$. Затем проверьте, что число таких перекрестков, которые образуют $P_{+}$-скрещивающуюся пару с $A_{+}$в $K_{+}$, сравнимо с $\mathrm{lk}_{2} K_{0}$ по модулю 2 . ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-140.jpg?height=274&width=591&top_left_y=1999&top_left_x=732) Рис. 21: Скайн-соотношения для Arf-инварианта (b) Рассмотрим любую связную сумму двух трилистников. Согласно утверждениям 1.4.1.c и 1.5.4.d Арф-инвариант этого узла равен нулю. В то же время по утверждениям 2.1.1.c,і инвариант Кэссона этого узла равен 2. Следовательно, этот узел не изотопен тривиальному узлу. 1.4.3. Существование. Следует из утверждений 1.4.1.b и 1.4.2.a. Единственность. Доказательство основано на тех же идеях, что и доказательство теоремы 1.3.4. Используйте саму лемму 1.1.5, а не ее аналог для зацеплений. 1.4.4. См. стр. $75-78$ в [Kа87]. 1.4.5. См. теорему 2 в [CG83]. 1.5.1. (b) Первое решение. Назовем ориентированную ломаную положительной, если ограниченная ею часть плоскости лежит справа от любого ее звена (здесь используется теорема Жордана, сформулированная в замечании после теоремы 1.1.3). Покажите, что положительность ломаной сохраняется при элементарных движениях. Указание ко второму решению. Положительность ломаной из предыдущего решения можно эквивалентно определить следующим образом. Будем говорить, что ориентированная ломаная $A_{1} \ldots A_{n}$ положительна, если для любой ее внутренней точки $O$ сумма ориентированных углов $\angle A_{1} O A_{2}+\angle A_{2} O A_{3}+\ldots+\angle A_{n-1} O A_{n}+\angle A_{n} O A_{1}$ положительна (т. е. индекс всякой внутренней точки ломаной относительно самой ломаной положителен). 1.5.2. (a), (b), (c) Каждый из трех указанных ориентированных узлов переводится в ориентированный узел с противоположной ориентацией вращением на угол $\pi$ относительно «вертикальной» оси, проходящей через «самую верхнюю» точку узла (см. самую левую диаграмму на рис. 1 в первом и втором ряду для трилистника и восьмерки соответственно). Это вращение входит в непрерывное семейство вращений на угол $\pi t, t \in[0,1]$ относительно той же прямой. Это и есть требуемая изотопия. 1.5.4. (а) См. рис. 22 . ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-141.jpg?height=306&width=1743&top_left_y=1224&top_left_x=148) Рис. 22: К доказательству изотопности $K \# O$ и $K$ (b) Достаточно взять «маленький» узел из класса изотопии $[L]$ и «провести» его через узел класса изотопии $[K]$, см. рис. 23. (c) Классы изотопии узлов из левой и правой частей равенства имеют общего представителя, показанного на рис. 24 . (d) Выберите отмеченную точку $P$ близко к «месту соединения» узлов. Затем проверьте, что всякая $P$-скрещивающаяся пара перекрестков в $K \# L$ получается из скрещивающейся пары в $K$ или из скрещивающейся пары в $L$. 1.5.5. (d) Проверьте, что всякая точка пересечения разных компонент в $K \# L$ получается из точки пересечения разных компонент в $K$ или точки пересечения разных компонент в $L$. 1.5.6. (а) В качестве примера можно взять одинаковые зацепления состоящие из трилистника и тривиального узла в непересекающихся кубах. См. рис. 3.16 в [PS96]. (b) Cм. [As]. На рис. 25 альтернативный пример, предложенный А. Рябичевым. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-142.jpg?height=1606&width=658&top_left_y=80&top_left_x=690) Рис. 23: К доказательству изотопности $K \# L$ и $L \# K$ 1.6.1. Ответы: (a) 1 ; (b) 0 . (c) Доказательство аналогично 1.3.2. Достаточно проверить, что знаки всех точек пересечения не меняются при движениях Редемейстера. 1.6.2. (a) Доказательство аналогично 1.3.2.b. Рассмотрим плоскую диаграмму зацепления. Из леммы о тривиальности (сформулированной после задачи 1.6.2) следует, что сумма знаков точек пересечения, в которых первая компонента выше второй, противоположна сумме знаков точек пересечения, в которых вторая компонента компонента выше первой. Поскольку перестановка компонент зацепления меняет знаки всех точек пересечения компонент, то коэффициент зацепления не меняется. (b) Смена ориентации любой компоненты меняет знаки всех точек пересечения. (с) Достаточно рассмотреть любую связную сумму пяти зацеплений Хопфа, у которых ориентации на компонентах выбраны таким образом, чтобы их коэффициент зацепления был равен -1 . (d) Доказательство аналогично 1.5.4. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-143.jpg?height=406&width=1090&top_left_y=68&top_left_x=474) Рис. 24: К доказательству изотопности $(K \# L) \# M$ и $K \#(L \# M)$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-143.jpg?height=964&width=1044&top_left_y=584&top_left_x=497) Рис. 25: Связная сумма изотопических классов упорядоченных зацеплений не является корректно определенной операцией (е) Примером является зацепление Уайтхеда (оно не изотопно тривиальному по теореме 1.2.2). 1.6.3. Доказательство аналогично теореме 1.3.4. ## Указания и решения к задачам после промежуточного финиша 2.1.1. Oтветы: (b) 0; (c) 1 ; (d) -1 . (a) Рассмотрим любую связную сумму пяти восьмерок. Из п. (d) и задачи 2.1.1.j следует, что инвариант Кэссона данного узла равен -5 . (b) Тривиальный узел не имеет перекрестков и потому скрещивающихся пар перекрестков. Поэтому инвариант Кэссона этого узла равен нулю. (c) Все три перекрестка трилистника имеют одинаковый знак. Поскольку трилистник имеет только одну скрещивающуюся пару перекрестков (вне зависимости от выбора отмеченной точки), то инвариант Кэссона этого узла равен 1. (е) Доказательство аналогично задаче 1.4.1.е. Используйте лемму о тривиальности, сформулированную после задачи 1.6.1. Здесь и далее важно следить за знаками точек пересечения! (f) Доказательство аналогично утверждению 1.4.1.f. (g) Доказательство аналогично утверждению 1.4.2.а. (h) Рассмотрим любую связную сумму трилистника и восьмерки. Из задач 2.1.1.c,d и 2.1.1.i следует, что инвариант Кэссона этого узла равен 0 . С другой стороны, из задач 2.3.4.d,е и 2.3.6.b следует, что многочлен Конвея этого узла равен $\left(t^{2}+1\right)\left(t^{2}-1\right) \neq 1$. Следовательно, этот узел не изотопен тривиальному. (i) Доказательство аналогично утверждению 1.5.4.d. 2.1.2. Доказательство аналогично теореме 1.4.3. 2.2.1. Ответы: b,e - раскрашиваемы, a,c,d,f,g,h,i,j - не раскрашиваемы. Пример раскрашенного в три цвета трилистника см. в прикрепленной копии [Pr95, стр. 35, Рис. 4.3]. Правильную раскраску последней диаграммы с рис. 7 см. на рис. 26 слева. (Эта диаграмма была ошибочно названа нераскрашиваемой в три цвета в $[\operatorname{Pr} 95, \S 4]$. Эта небольшая ошибка была найдена Л. Баннором, С. Зотовой и Л. Кравцовой.) 2.2.2. (a) См. прикрепленную копию $[\operatorname{Pr} 95$, стр. 34, Теорема 4.1] (b) Следует из (a) и утверждений 2.2.1.d-j (см. также [Pr95, стр. 30]). Последняя диаграмма с рис. 7 отлична от тривиального зацепления числом правильных раскрасок плоской диаграммы. Докажите, что это число сохраняется при движениях Рейдемейстера. Плоская диаграмма называется раскрашиваемой в 5 цветов, если существует раскраска ее дуг в цвета $0,1,2,3,4$ при которой - хотя бы два цвета использованы; - в каждой точке пересечения если верхняя дуга имеет цвет $a$, а две нижние дуги - цвета $b$ и $c$, то $2 a \equiv b+c(\bmod 5)$. Аналогично задаче 2.2.2.а можно показать, что раскрашиваемость в 5 цветов сохраняется при движениях Рейдемейстера. Для завершения доказательства достаточно заметить, что узел $5_{1}$ является раскрашиваемым в 5 цветов (см. рис. 26 справа), а тривиальный узел - нет. 2.3.1. Это частные случаи теоремы 2.3.3 по модулю 2 . 2.3.2. Ответы: (a, b) 0; (c) 1 (независимо от выбора ориентации). Замечание. Инварианты arf и $a_{3}$ для зацеплений могут зависеть от выбора ориентации компонент (для $a_{3}$ это можно посмотреть в прикрепленной копии [CDM12, 2.3.4]). 2.3.3. Едиственность доказывается аналогично теоремам 2.1.2 и 1.4.3; сначала решите задачу 2.3.4. Bывод (a) из (b). Достаточно показать, что все инварианты $c_{n}$, определенные в (b), сохраняются при изменении порядка компонент. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-145.jpg?height=616&width=1148&top_left_y=72&top_left_x=446) Рис. 26: Раскраска в 3 цвета зацепления и раскраска в 5 цветов узла 51 ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-145.jpg?height=350&width=320&top_left_y=801&top_left_x=859) Рис. 27: Четырехкомпонентное зацепление Борромео Пусть $D$ - плоская диаграмма некоторого зацепления с двумя или более компонентами и пусть $D^{\prime}$ - плоская диаграмма, полученная из $D$ изменением порядка компонент. Через сг $D$ обозначим число перекрестков в $D$. Через $u(D)$ обозначим минимально возможное число изменений перекрестков, необходимое для получения диаграммы зацепления, изотопного тривиальному, из $D$ (такая последовательность изменений перекрестков существует по аналогу леммы 1.1.5 для зацеплений). Докажем индукцией по сr $D$. Если сr $D=0$, тогда $D$ - диаграмма зацепления, изотопного тривиальному, и по утверждению 2.3.4.b имеем $C(D)=0$ для любого порядка компонент. Предположим, что сr $D>0$; тогда продолжим доказательство индукцией по $u(D)$. Если $u(D)=0$, то $D$ - диаграмма зацепления, изотопного тривиальному; этот случай рассмотрен выше. Предположим, что $u(D)>0$ и пусть $D_{*}$ - зацепление, полученное из $D$ изменением перекрестка таким, что $u\left(D_{*}\right)0$; тогда продолжим доказательство индукцией по $u(D)$. Если $u(D)=0$, то $K$ изотопно тривиальному зацеплению; этот случай рассмотрен выше. Предположим, что $u(D)>0$ и пусть $D_{*}$ - зацепление, полученное из $D$ изменениями перекрестков, и такое, что $u\left(D_{*}\right)N. Khoroshavkina, E. Morozov, A. Skopenkov, R. Živaljević ${ }^{\dagger}$ #### Abstract . We define simple invariants of knots or links (linking number, Arf-Casson invariants and Alexander-Conway polynomials) motivated by interesting results whose statements are accessible to a non-specialist or a student (e.g. Theorems 1.1.3 and 1.2.2). We show how the simplest invariants naturally appear in an attempt to unknot a knot or unlink a link. Then we present certain 'skein' recursive relations for the simplest invariants, which allow to introduce stronger invariants. We state the Vassiliev-Kontsevich theorem in a way convenient for calculating the invariants themselves, not only the dimension of the space of the invariants. We also present coloring invariants although we cannot explain in an elementary way how they appear. We give rigorous definitions of the main notions in a way not obstructing intuitive understanding. No prerequisites are required for this text. ## Recommendations for participants. If a mathematical statement is formulated as a problem, then the objective is to prove this statement. If such a problem is labeled 'theorem' ('lemma', 'corollary', etc.), then this statement is considered more important. Usually we formulate as a problem a beautiful or important statement before giving a sequence of problems which constitute its proof. In this case, in order to prove this statement, one may need to solve some of the subsequent problems. We do not want to deprive you of the pleasure of finding the right moment when you finally are ready to prove the statement. In general, if you are stuck on a certain problem, try looking at the next ones. They may turn out to be helpful. Remarks are formally not used later. When an important notion is defined, it is marked in bold letters so that it would be easier to find its definition later. We suggest to all students working on the project to consult the jury on any questions on the project. Students who successfully work or complete a part of the project will get interesting extra problems (subsections marked by stars; they are not yet elaborated). For every solution which has been written down and marked with either ' + ' or ' + .' a student (or a group of students) gets a 'bean'. The jury may also award extra beans for beautiful solutions, solutions of hard problems, or solutions typeset in $\mathrm{T}_{\mathrm{E}} \mathrm{X}$. The jury has infinitely many beans. One may submit a solution in oral form, but one loses a bean with each 5 attempts (successful or not). Please notify us if you already know solutions of some of the problems. If you confirm your knowledge by presenting some of them, then you will be allowed to use them in solutions of other problems, without earning before a plus-mark (plus-sign) for their solutions. ${ }^{*}$ We are grateful to S. Chmutov, A. Ryabichev and A. Sossinsky for useful discussions, to V. Prasolov and the MCCME publishing house for allowing us to use some problems and figures from [Pr95]. This text is based on lectures by A. Skopenkov at Independent University of Moscow (including Math in Moscow Program) and Moscow Institute of Physics and Technology. ${ }^{\dagger}$ D. Eliseev: Samara University. A. Enne: Petrozavodsk State University. M. Fedorov, N. Khoroshavkina, E. Morozov: Higher School of Economics, Moscow. A. Glebov: Novosibirsk State University. A. Skopenkov: https://users.mccme.ru/skopenko, Moscow Institute of Physics and Technology, Independent University of Moscow. R. Živaljević: http://www.rade-zivaljevic.appspot.com/, Mathematical Institute SANU, Belgrade. ## 1 Problems before the semifinal ### 1.1 Main definitions and results on knots We start with informal description of the main notions (rigorous definitions are given after Problem 1.1.1). You can imagine a knot as a thin elastic string whose ends have been glued together, see fig. 1. As in this figure, knots are usually represented by their 'nice' plane projections called knot diagrams. Imagine laying down the rope on a table and carefully recording how it crosses itself (i.e. which part lies on top of the other). It should be kept in mind that the projections of the same knot on different planes can look quite dissimilar. A trivial knot is the outline (the boundary) of a triangle. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-149.jpg?height=442&width=1748&top_left_y=687&top_left_x=206) o Figure 1: Knots isotopic to the trefoil knot (top row) and to the figure eight knot (bottom row) By an isotopy of a knot we mean its continuous deformation in space as a thin elastic string; no self-intersections are allowed throughout the deformation. Two knots are isotopic if one can be transformed to the other by an isotopy. As a proof that specific knots are isotopic we accept a big clear picture, or an experiment with rope that can be repeated by a Jury member. Problem 1.1.1. (a) Some two knots represented in the top row of fig. 1 are isotopic to the leftmost knot in this row. For one of these two knots decompose your isotopy into Reidemeister moves shown in fig. 5 . (b)* All the knots represented in the top row of fig. 1 are isotopic to each other. $\left(\mathrm{c}, \mathrm{d}^{*}\right)$ The same is true for the knots represented in the bottom row of fig. 1 . (e) All knots with the same knot diagram are isotopic. Remark. Here we justify the necessity of a rigorous definition of isotopy. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-149.jpg?height=230&width=1722&top_left_y=1896&top_left_x=188) Figure 2: A (non-ambient) isotopy between the trefoil knot and the trivial knot In fig. 2 we see an isotopy between the trefoil knot and the trivial knot. Is it indeed an isotopy? This is the so called 'piecewise linear non-ambient isotopy' which is different from the 'piecewise linear ambient isotopy' defined and used later. (The first notion better reflects the idea of continuous deformation without self-intersections, but is hardly accessible to high school students, cf. [Sk16i].) In fact, any two knots are piecewise linear non-ambient isotopic! The usual problem with intuitive definitions is not that it is hard to make them rigorous, but that this can be done in several ways. A knot is a spatial closed non-self-intersecting polygonal line. A plane diagram of a knot is its generic ${ }^{1}$ projection onto a plane ${ }^{2}$, together with the information which part of the knot 'goes under' and which part 'goes over' at any given crossing. Problem 1.1.2. For any knot diagram there is a knot projected to this diagram. (Such a knot need not be unique; see though problem 1.1.1.e.) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-150.jpg?height=228&width=897&top_left_y=401&top_left_x=598) Figure 3: Elementary move Suppose that two sides $A C$ and $C B$ of a triangle $A B C$ are edges of a knot. Moreover, assume that the knot and (the part of the plane bounded by) the triangle $A B C$ do not intersect at any other points. An elementary move $A C B \rightarrow A B$ is the replacement of the two edges $A C$ and $C B$ by the edge $A B$, or the inverse operation $A B \rightarrow A C B$ (fig. 3 ). ${ }^{3}$ Two knots $K, L$ are called (piecewise linearly ambiently) isotopic if there is a sequence of knots $K_{1}, \ldots, K_{n}$ such that $K_{1}=K, K_{n}=L$ and every subsequent knot $K_{j+1}$ is obtained from the previous one $K_{j}$ by an elementary move. Theorem 1.1.3. (a) The trivial knot is not isotopic to the trefoil knot. (b) The trivial knot is not isotopic to the figure eight knot. (c) The trefoil knot is not isotopic to the figure eight knot. (d) There is an infinite number of pairwise non-isotopic knots. This is proved using Arf and Casson invariants, see $\S 1.4$ and $\S 2.1$, or using proper colorings, see $\S 2.2$ (so you need not spend much time on proving this result right now). Remark. This remark might be useful as both a hint and a warning to problems 1.1.4 and 1.1.5. In the following paragraph we prove that if a knot lies in a plane, then the knot is isotopic to the trivial knot. Denote the knot in a plane by $M_{1} M_{2} \ldots M_{n}$. Take a point $Z$ outside the plane. Then $M_{1} M_{2} \ldots M_{n}$ is transformed to the trivial knot $M_{1} Z M_{n}$ by the following sequence of elementary moves: $$ M_{1} M_{2} \rightarrow M_{1} Z M_{2}, \quad Z M_{2} M_{3} \rightarrow Z M_{3}, \quad Z M_{3} M_{4} \rightarrow Z M_{4}, \quad \ldots, \quad Z M_{n-1} M_{n} \rightarrow Z M_{n} $$ The following result shows that intermediate knots of an isotopy from a knot lying in a plane to the trivial knot can be chosen also to lie in this plane. Schoenflies Theorem. Any closed polygonal line without self-intersections in the plane is isotopic (in the plane) to a triangle. This is a stronger version of the following celebrated result. Jordan Theorem. Every closed non-self-intersecting polygonal line $L$ in the plane $\mathbb{R}^{2}$ splits the plane into exactly two parts, i.e. $\mathbb{R}^{2}-L$ is not connected and is a union of two connected sets. A subset of the plane is called connected, if every two points of this subset can be connected by a polygonal line lying in this subset.[^25] For an algorithmic explanation why the Jordan Theorem (and so the Schoenflies Theorem) is non-trivial, and for a proof of the Jordan Theorem, see $\S 1.3$ 'Intersection number for polygonal lines in the plane' of [Sk18], [Sk]. Problem 1.1.4. Suppose that there is a point on a knot such that if we go around the knot starting from this point, then on some plane diagram we first meet only overcrossings, and then only undercrossings. Then the knot is isotopic to the trivial knot. ${ }^{4}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-151.jpg?height=247&width=572&top_left_y=435&top_left_x=755) Figure 4: Crossing change A crossing change is change of overcrossing to undercrossing or vise versa, see fig. 4 . Clearly, after any crossing change on the diagrams of the trefoil knot and the figure eight shown in fig. 1 we obtain a diagram of a knot isotopic to the trivial knot. Lemma 1.1.5. Every plane diagram of a knot can be transformed by crossing changes to a plane diagram of a knot isotopic to the trivial knot. ${ }^{5}$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-151.jpg?height=212&width=1634&top_left_y=1110&top_left_x=228) Figure 5: Reidemeister moves. The plane diagrams are identical outside the disks bounded by dashed circles. No other sides of the plane diagrams except for the pictured ones intersect the disks. (Same for fig. 6, 4, 9 and 10.) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-151.jpg?height=214&width=1672&top_left_y=1586&top_left_x=210) Figure 6: (Left) To a rigorous definition of the first Reidemeister move (Middle, right) Plane isotopy moves In this text instead of knots up to isotopy we shall study plane diagrams of knots up to (equivalence generated by) Reidemeister moves shown in fig. $5^{6}$ and plane isotopy moves shown in fig. 6 (middle, right). I.e. we shall use without proof the following result. Theorem 1.1.6 (Reidemeister). * Two knots are isotopic if and only if some plane diagram of the first knot can be obtained from some plane diagram of the second one by Reidemeister moves and plane isotopy moves.[^26] ### 1.2 Main definitions and results on links A link is a collection of pairwise disjoint knots, which are called the components of the link. Ordered collections are called ordered or colored links, while non-ordered collections are called non-ordered or non-colored links. In this text we abbreviate 'ordered link' to just 'link'. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-152.jpg?height=173&width=241&top_left_y=496&top_left_x=519) (a) (b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-152.jpg?height=850&width=922&top_left_y=370&top_left_x=581) Figure 7: The Hopf link, the trivial link and another three links ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-152.jpg?height=241&width=228&top_left_y=1392&top_left_x=450) (b) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-152.jpg?height=239&width=228&top_left_y=1385&top_left_x=927) (w) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-152.jpg?height=236&width=233&top_left_y=1389&top_left_x=1399) (t) Figure 8: The Borromean rings, the Whitehead link and the trefoil knot A trivial link (with any number of components) is a link formed by triangles in parallel planes. Plane diagram, isotopy, etc. for links are defined analogously to knots. The analogues of Lemmas 1.1.6 and 1.1.5 for links are correct. Problem 1.2.1. (a) The Hopf link is isotopic to the link obtained from the Hopf link by switching the components. (b) The Hopf link is isotopic to some link whose components are symmetric with respect to some straight line. (c) The fourth link in fig. 7 is isotopic to the Whitehead link in fig. 8.w. (d,e*) The same as in (a,b) for the Whitehead link. (f)* The Borromean rings link is isotopic to a link whose components are permuted in a cyclic way under the rotation by angle $2 \pi / 3$ with respect to some straight line. Theorem 1.2.2. (a) The Hopf link is not isotopic to the trivial link. (b) The Whitehead link is not isotopic to the trivial link. (c) The Hopf link is not isotopic to the Whitehead link. (d) The Borromean rings link is not isotopic to the trivial link. Parts (a) and (c) are proved using linking number modulo 2, invent it yourself or see $\S 1.3$. Parts (b) and (d) are proved using either the Alexander-Conway polynomials, see $\$ 2.3$, or 'triple linking' (Massey-Milnor) number and 'higher linking' (Sato-Levine) number [Sk, §4.4-§4.6]. Part (d) can also be proved using proper colorings, see $\S 2.2$. ### 1.3 The Gauss linking number modulo 2 via plane diagrams Problem 1.3.1. Let $A, B, C, D, E, F, O$ be points in space, no four of which lie in one plane. The following three conditions are equivalent. (i) The outline of $D E F$ intersects the part of the plane $A B C$ bounded by the triangle $A B C$ at exactly one point. (ii) The segment $B C$ passes below (like in the construction of link diagram) exactly one side of $D E F$ as seen from $A$. (iii) The outline of $A B C$ passes below an odd number of sides of $D E F$ as seen from $O$. Suppose that there is an isotopy between two 2-component links, and the second component is fixed throughout the isotopy. Then the trace of the first component is a self-intersecting cylinder disjoint from the second component. If after the isotopy the components are unlinked, then the cylinder can be completed to a self-intersecting disk disjoint from the second component. This observation, together with problem 1.3.1 and [Sk, the Projection lemma 4.2.4], motivates the following definition. The linking number modulo $2 \mathrm{lk}_{2}$ of the plane diagram of a 2-component link is the number modulo 2 of crossing points on the diagram at which the first component passes above the second component. Problem 1.3.2. (a) Find the linking number modulo 2 for the plane diagrams in fig. 7 , for pairs of Borromean rings and for the Whitehead link (fig. 8). (b) The linking number modulo 2 is preserved under Reidemeister moves. By (b) the linking number modulo 2 of a 2-component link (or even of its isotopy class) is well-defined by setting it to be the linking number modulo 2 of any plane diagram of the link. We shall use without proof the following Parity lemma: any two closed polygonal lines in the plane whose vertices are in general position intersect at an even number of points. For a discussion and a proof see $\S 1.3$ 'Intersection number for polygonal lines in the plane' of [Sk18], [Sk]. Problem 1.3.3. (a) Switching the components of a link preserves the linking number modulo 2 . (b) There is a 2-component link which is not isotopic to the trivial link but which has zero linking number modulo 2 . ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-153.jpg?height=278&width=890&top_left_y=1926&top_left_x=602) Figure 9: Knots $K_{+}, K_{-}, K_{0}$ Denote by $D_{+}, D_{-}, D_{0}$ any three diagrams of oriented (knots or) links differing as shown in fig. 9 (for a convention on figures see caption to fig. 5). We also denote by $K_{+}, K_{-}, K_{0}$ any three links who have diagrams $D_{+}, D_{-}, D_{0}$. If an invariant (like $\mathrm{lk}_{2}$ ) is defined for non-oriented links (or knots), then its value on a link is assigned to the link with any orientation. Theorem 1.3.4. There is a unique mod2-valued isotopy invariant $\mathrm{lk}_{2}$ of (non-oriented) 2component links that assumes value 0 on the trivial link and such that (for any links $K_{+}$and $K_{-}$ having plane diagrams differing as shown in fig. 9) $$ \mathrm{lk}_{2} K_{+}-\mathrm{lk}_{2} K_{-}= \begin{cases}1 & \text { if at the crossing point different components cross each other; } \\ 0 & \text { if at the crossing point one component crosses itself. }\end{cases} $$ Problem 1.3.5. * If the linking number modulo 2 of two (disjoint outlines of) triangles in space is zero, then the link formed by the triangles is isotopic to the trivial link. Theorem 1.3.6 (Conway-Gordon-Sachs). * If no 4 of 6 points in 3-space lie in the same plane, then there are two linked triangles with vertices at these 6 points. That is, the part of the plane bounded by the first triangle intersects the outline of the second triangle exactly at one point. ### 1.4 The Arf invariant Take a plane diagram of a knot and a point $P$ on the diagram different from crossing points. Call $P$ a basepoint. A non-ordered pair of crossing points $A$ and $B$ is called skew (or $P$-skew) if going around the diagram in some direction starting from $P$ and marking only crossings at $A$ and $B$, we first mark overcrossing at $A$, then undercrossing at $B$, then undercrossing at $A$, and at last overcrossing at $B$. The $P$-Arf invariant $\operatorname{arf}_{P}$ of the plane diagram is the parity of the number of all skew pairs of crossing points. Problem 1.4.1. (a) If the $P$-Arf invariant of a plane diagram is non-zero, then $P$ is not a point as in problem 1.1.4. (b,c,d) Find the $P$-Arf invariant (of some plane diagram) of the trivial, the trefoil and the figure eight knots (for your choice of a basepoint $P$ ). (e) The $P$-Arf invariant is independent of the choice of a basepoint $P$. By (e) the Arf invariant of a plane diagram is well-defined by setting it to be the $P$-Arf invariant for any basepoint $P$. (f) The Arf invariant of a plane diagram is preserved under Reidemeister moves. By (f) the Arf invariant (Arf number) arf of a knot (or even of isotopy class of a knot) is well-defined by setting it to be the Arf invariant of any plane diagram of the knot. Problem 1.4.2. (a) If in fig. $9 K_{+}, K_{-}$are plane diagrams of knots, then $K_{0}$ is a plane diagram of 2-component link and arf $K_{+}-\operatorname{arf} K_{-}=\mathrm{lk}_{2} K_{0}$. (b) There is a knot which is not isotopic to the trivial knot but which has zero Arf invariant. Theorem 1.4.3. There is a unique mod2-valued isotopy invariant arf of (non-oriented) knots that assumes value 0 on the trivial knot and such that $$ \operatorname{arf} K_{+}-\operatorname{arf} K_{-}=\mathrm{lk}_{2} K_{0} $$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-154.jpg?height=380&width=858&top_left_y=2114&top_left_x=620) Figure 10: Pass move Problem 1.4.4. Two knots are called pass equivalent if some plane diagram of the first knot (with some orientation) can be transformed to some plane diagram of the second knot (with some orientation) using Reidemeister moves and pass moves of fig. 10 . (a) If two knots are pass equivalent, then their Arf invariants are equal. (b)* The eight figure knot is pass equivalent to the trefoil knot. (c)* If the Arf invariants of two knots are equal, then the knots are pass equivalent. Theorem 1.4.5. * Take any 7 points in space, no four of which belong to any plane. Take $\binom{7}{2}=21$ segments joining them. Then there is a closed polygonal line formed by taken segments and non-isotopic to the boundary of a triangle. ### 1.5 Oriented knots and links You know what is oriented polygonal line, so you know what is oriented knot (fig. 11). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-155.jpg?height=254&width=570&top_left_y=786&top_left_x=755) Figure 11: Two trefoil knots with the opposite orientations Both the informal notion and rigorous definition of isotopic oriented knots are given analogously to isotopic knots. Problem 1.5.1. Isotopic oriented polygonal lines without self-intersections on the plane and on the sphere are defined analogously to isotopic oriented knots in space. (a) An oriented spherical triangle is isotopic on the sphere to the same triangle with the opposite orientation. (b) The analogue of (a) for the plane is false. Problem 1.5.2. (a) Two trivial; (b) Two trefoil; (c) Two figure eight knots with the opposite orientations are isotopic. Theorem 1.5.3. (H. Trotter, 1964) There exists an oriented knot which is not isotopic to the same knot with the opposite orientation. This is proved using the Jones poynomial [PS96], [CDM12]; the proof is outside the scope of this text. The connected sum \# of oriented knots is defined in fig. $12 .{ }^{7}$ This is not a well-defined operation on oriented knots. So we denote by $K \# L$ any of the connected sums of $K$ and $L$. Problem 1.5.4. For any oriented knots $K, L, M$ and the trivial oriented knot $O$ we have (a) $K \# O=K$. (b) $K \# L=L \# K$. (c) $(K \# L) \# M=K \#(L \# M)$. (d) $\operatorname{arf}(K \# L)=\operatorname{arf} K+\operatorname{arf} L$ (here knots $K, L$ are non-oriented). (The rigorous meaning of (a) is 'there is a connected sum of $K$ and $O$ isotopic to $K$ '. Analogous rigorous meanings have (b) and (c). See though Remark below.)[^27] ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-156.jpg?height=484&width=1696&top_left_y=344&top_left_x=192) Figure 12: Connected sum of knots Remark. An isotopy class of a knot is the set of knots isotopic to this knot. The oriented isotopy class $[K \# L]$ of the connected sum of two oriented isotopy classes $[K],[L]$ of oriented knots $K, L$ is independent of the choices used in the construction, and of the representatives $K, L$ of $[K],[L]$. Hence the connected sum of oriented isotopy classes of oriented knots is well-defined by $[K] \#[L]:=[K \# L]$, see $[$ Sk15p, Remark 2.3.a]. For isotopy classes of non-oriented knots the connected sum is not well-defined. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-156.jpg?height=605&width=989&top_left_y=1313&top_left_x=557) Figure 13: Connected sum of links The connected sum \# of links (ordered or not, oriented or not) is defined analogously to the connected sum of knots, see fig. 13. This is not a well-defined operation on links, and problem 1.5.6 shows that this does not give a well-defined operation on their isotopy classes. So we denote by $K \# L$ any of the connected sums of $K$ and $L$. Problem 1.5.5. (a,b,c,d) Prove the analogues of problem 1.5.4.a,b,c,d for links. Problem 1.5.6. There are two isotopic pairs $(K, L)$ and $\left(K^{\prime}, L^{\prime}\right)$ of (a) non-ordered; (b)* ordered 2-component links (oriented or not) such that some connected sums $K \# L$ and $K^{\prime} \# L^{\prime}$ are not isotopic. ### 1.6 The Gauss linking number via plane diagrams Let $(\overrightarrow{A B}, \overrightarrow{C D})$ be ordered pair of vectors (oriented segments) in the plane intersecting at a point $P$. Define the sign of $P$ to be +1 if $A B C$ is oriented clockwise and to be -1 otherwise (fig. 14). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-157.jpg?height=220&width=458&top_left_y=300&top_left_x=820) Figure 14: The sign of intersection point The linking number $\mathrm{lk}$ of the plane diagram of an oriented 2-component link is the sum of signs of all those crossing points on the diagram at which the first component passes above the second component. At every crossing point the first (the second) vector is the oriented edge of the first (the second) component. Problem 1.6.1. (a,b) Find the linking number for (some plane diagram of) the Hopf link and pairs of Borromean rings, for your choice of orientation on the components. (c) The linking number is preserved under Reidemeister moves. By (c) the linking number of an oriented 2-component link (or of its isotopy class) is welldefined by setting it to be the linking number of any plane diagram of the link. The absolute value of the linking number of a (non-oriented) 2-component link (or of its isotopy class) is well-defined by taking any orientations on the components. We shall use without proof the following Triviality lemma: for any two closed oriented polygonal lines in the plane whose vertices are in general position the sum of signs of their intersection points is zero. For a discussion and a proof see $\S 1.3$ 'Intersection number for polygonal lines in the plane' of [Sk18], [Sk]. Problem 1.6.2. (a) Does switching the components of a link preserve or negate the linking number? (b) Reversing the orientation of either of the components negates the linking number. (c) Draw an oriented 2-component link whose linking number is -5 . (d) For any of the connected sums $K \# L$ of oriented 2-component links $K, L$ we have $\operatorname{lk}(K \# L)=$ $\mathrm{lk} K+\mathrm{lk} L$. (e) There is a 2-component link which is not isotopic to the trivial link but which has zero linking number. Theorem 1.6.3. There is a unique integer-valued isotopy invariant $\mathrm{lk}$ of oriented 2-component links that assumes value 0 on the trivial link and and such that $$ \mathrm{lk} K_{+}-\mathrm{lk} K_{-}= \begin{cases}1 & \text { if at the crossing point different components cross each other; } \\ 0 & \text { if at the crossing point one component crosses itself. }\end{cases} $$ Problem 1.6.4. Two links (oriented or not) are called link homotopic if one link can be continuously deformed to the other so that no intersections of different components appear throughout the deformation (but self-intersections of the components may appear). Or, equivalently, some plane diagram of the first link can be transformed to some plane diagram of the second link using Reidemeister moves and crossing changes of one component. (a) The Whitehead link is link homotopic to the trivial link. (b) If two oriented 2-component links are link homotopic, then their linking numbers are equal. (c)* If the linking numbers of two 2-component links are equal, then the links are link homotopic. (d)* The Borromean rings link is not link homotopic to the trivial link. ## 2 Problems after the semifinal Subsection 2.2 only use the material of $\S \S 1.1$ and 1.2. Subsection 2.3 only use the material of $\S \S 1.1,1.2$ and 1.5. ### 2.1 The Casson invariant The sign of a crossing point of an oriented plane diagram of a knot is defined after figure 14; the first (the second) vector is the vector of overcrossing (of undercrossing). Clearly, the sign is independent of the orientation of the diagram, and so is defined for non-oriented diagram. The sign of a $P$-skew pair of crossing points in a plane diagram of a knot (for any basepoint $P$ ) is the product of the signs of the two crossing points. The $P$-Casson invariant of a plane diagram is the sum of signs over all $P$-skew pairs of crossing points. Problem 2.1.1. (a) Draw a plane diagram of a knot and a basepoint $P$ such that $P$-Casson invariant is -5 . (b,c,d,e,f) Same as problems 1.4.1.b,c,d,e,f for the Casson invariant. (g) Find and prove the analogue of problem 1.4.2.a for the Casson invariant. (h,i) Same as problems 1.4.2.b and 1.5.4.d for the Casson invariant. By (e,f) the Casson invariant (Casson number) $c_{2}$ of a plane diagram, of a knot, or even of isotopy class of a knot, is well-defined by setting it to be the $P$-Casson invariant of any plane diagram of the knot for any basepoint $P$. Theorem 2.1.2. There is a unique integer-valued isotopy invariant $c_{2}$ of (non-oriented) knots that assumes value 0 on the trivial knot and for which $$ c_{2}\left(K_{+}\right)-c_{2}\left(K_{-}\right)=\operatorname{lk} K_{0} $$ (The number $\mathrm{lk} K_{0}$ is well-defined because change of the orientation on both components of an oriented link does not change the linking number.) ### 2.2 Proper colorings A strand in a plane diagram (of a knot or link) is a connected piece that goes from one undercrossing to the next. A proper coloring of a plane diagram (of a knot or link) is a coloring of its strands in one of three colors so that at least two colors are used, and at each crossing, either all three colors are present or only one color is present. A plane diagram (of a knot or link) is 3-colorable if it has a proper coloring. Problem 2.2.1. For each of the following knots or links take any diagram and decide if it is 3-colorable. (a) the trivial knot. (b) the trefoil knot. (c) the figure eight knot. (d-j) links in fig. 7 and 8 . Problem 2.2.2. (a) The 3-colorability of a plane diagram is preserved under the Reidemeister moves. (b) Neither of links in fig. 7 and 8 (except the trivial link) is isotopic to the trivial link. Problem 2.2.3. * The 51 knot is not isotopic to the trivial knot. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-159.jpg?height=292&width=317&top_left_y=79&top_left_x=885) Figure 15: The $5_{1}$ knot ### 2.3 Alexander-Conway polynomials Problem 2.3.1. * (a) There is a unique mod2-valued isotopy invariant arf of oriented 3component links that assumes value 0 on the trivial link and for which $$ \operatorname{arf}\left(K_{+}\right)-\operatorname{arf}\left(K_{-}\right)= \begin{cases}\operatorname{lk}_{2} K_{0} & \text { at the crossing point different components cross each other; } \\ 0 & \text { at the crossing point one component crosses itself. }\end{cases} $$ (Here $\mathrm{lk}_{2} K_{0}$ is defined because $K_{0}$ is a 2-component link.) ${ }^{8}$ (b) There is a unique mod2-valued isotopy invariant $a_{3}$ of oriented 2-component links that assumes value 0 on the trivial link and for which $$ a_{3}\left(K_{+}\right)-a_{3}\left(K_{-}\right)=\operatorname{arf} K_{0} $$ (Here arf $K_{0}$ is defined because $K_{0}$ is either a knot or a 3-component link.) (c) There is a unique mod2-valued isotopy invariant $a_{3}$ of oriented 4-component links that assumes value 0 on the trivial link and for which $$ a_{3}\left(K_{+}\right)-a_{3}\left(K_{-}\right)= \begin{cases}\operatorname{arf} K_{0} & \text { at the crossing point different components cross each other; } \\ 0 & \text { at the crossing point one component crosses itself. }\end{cases} $$ (Here arf $K_{0}$ is defined because $K_{0}$ is a 3-component link.) Proof of the existence in problems 2.3.1 and theorem 2.3.3 is outside the scope of this text. See an elementary proof in $[K a 06$, , §2-§5], [Ka06]. You can earn a plus-mark (plus-sign) for proving the uniqueness, and solve other problems assuming the existence. For a relation to proper colorings see $[\mathrm{Ka06}, \S 6$. Problem 2.3.2. Calculate (for your choice of orientation on the components) (a) the arf invariant of the Borromean rings; $\left(\mathrm{b}, \mathrm{c}, \mathrm{d}^{*}\right)$ the $a_{3}$ invariant of the Hopf link, of the Whitehead link, and of 4-Borromean rings, i.e. of any link of your choice for which every 3 -component sublinks are isotopic to the trivial link, but the entire link is not isotopic to the trivial link. Theorem 2.3.3. * (a) There is a unique infinite sequence $c_{-1}=0, c_{0}, c_{1}, c_{2}, \ldots$ of $\mathbb{Z}$-valued isotopy invariants of oriented non-ordered links that assume values $c_{0}=1$ and $c_{1}=c_{2}=\ldots=0$ on the trivial knot and for which $$ c_{n}\left(K_{+}\right)-c_{n}\left(K_{-}\right)=c_{n-1}\left(K_{0}\right) $$ whenever $n \geq 0$.[^28](b) There is a unique infinite sequence $c_{-1}=0, c_{0}, c_{1}, c_{2}, \ldots$ of $\mathbb{Z}$-valued isotopy invariants of oriented ordered links that assume values $c_{0}=1$ and $c_{1}=c_{2}=\ldots=0$ on the trivial knot and such that for any $n \geq 0$ we have $$ c_{n}\left(K_{+}\right)-c_{n}\left(K_{-}\right)=c_{n-1}\left(K_{0}\right) $$ where $K_{0}$ is $K_{0}$ from fig. 9 with some ordering of the components. Actually two versions of theorem 2.3.3 are equivalent. You can use theorem 2.3.3.b without proof. ${ }^{9}$ The polynomial $C(K)(t):=c_{0}(K)+c_{1}(K) t+c_{2}(K) t^{2}+\ldots$ is called the Conway polynomial, see problem 2.3.5.e. Introduction of this polynomial allows to calculate all the invariants $c_{n}$ as quickly as one of them. The formula in theorem 2.3.3 is equivalent to $$ C\left(K_{+}\right)-C\left(K_{-}\right)=t C\left(K_{0}\right) $$ The polynomial $C(K)(t):=c_{0}(K)+c_{1}(K) t+c_{2}(K) t^{2}+\ldots$ is called the Conway polynomial, see problem 2.3.5.e. Introduction of this polynomial allows to calculate all the invariants $c_{n}$ as quickly as one of them. The formula in theorem 2.3.3 is equivalent to $$ C\left(K_{+}\right)-C\left(K_{-}\right)=t C\left(K_{0}\right) $$ Problem 2.3.4. Calculate the Conway polynomial of the following links (for your choice of orientation on the components). (a) the trivial link with 2 components; (b) the trivial link with $n$ components; (c) the Hopf link; (d) the trefoil knot; (e) the figure eight knot; (f) the Whitehead link; (g) the Borromean rings; (h) the $5_{1}$ knot. Problem 2.3.5. (a) We have $c_{0}(K)=1$ if $K$ is a knot and $c_{0}(K)=0$ otherwise (i.e. if $K$ has more than one component). (b) For a knot $K$ we have $c_{2 j+1}(K)=0$ and $c_{2}$ is the Casson invariant. (c) For a 2-component link $K$ we have $c_{2 j}(K)=0$ and $c_{1}$ is the linking coefficient. (d) For a $k$-component link $K$ we have $c_{j}(K)=0$ if either $j \leq k-2$ or $j-k$ is even. (e) For every knot or link all but a finitely many of the invariants $c_{n}$ are zeroes. Problem 2.3.6. (a) Change of the orientations of all components of a link (in particular, change of the orientation of a knot) preserves the Conway polynomial. (b) There is a 2-component link such that change of the orientation of its one component changes the degree of the Conway polynomial (so this change neither preserves nor negates the Conway polynomial). (c) For any of the connected sums $K \# L$ of knots $K, L$ we have $C(K \# L)=C(K) C(L)$. A link is split if it is isotopic to a link whose components are contained in disjoint balls. Problem 2.3.7. (a,b,c) No link of theorem 1.2.2 is split. (d) The Conway polynomial of a split link is trivial.[^29] ### 2.4 Vassiliev-Goussarov invariants (sketch) * Denote by - $\Sigma$ the set of isotopy classes of singular knots [PS96, 4.1], - $\delta_{n}$ the set of all chord diagrams that have $n$ chords [PS96, 4.8]; - $\sigma(K)$ the chord diagram of a singular knot $K$ [PS96, 4.8], [CDM12, 3.4.1] (not to be confused with Gauss diagrams for a non-singular knot $K$ [CDM12, 1.8.4]). Theorem 2.4.1 (Vassiliev-Kontsevich, [PS96], [CDM12]). For any map $\lambda: \delta_{n} \rightarrow \mathbb{R}$ there exists a map $v: \Sigma \rightarrow \mathbb{R}$ having properties (1)-(3) below if and only if $\lambda$ satisfies to the 1-term and the 4 -term relations [PS96, (4.5), (4.6)]. (1) For any singular knots $K_{+}, K_{-}$and $K^{0}$ from [PS96, (4.1)] (notice the difference with fig. 9) we have $$ v\left(K_{+}\right)-v\left(K_{-}\right)=v\left(K^{0}\right) $$ $\left(2_{n}\right) v(K)=0$ for each singular knot that has more than $n$ double points, and (3) $v(K)=\lambda(\sigma(K))$ for each singular knot that has exactly $n$ double points. A map $v: \Sigma \rightarrow \mathbb{R}$ such that (1) holds is called a Vassiliev-Goussarov invariant. A map $v: \Sigma \rightarrow \mathbb{R}$ such that $\left(2_{n}\right)$ holds is called a map of order at most $n$. Problem 2.4.2. (a) The map $v$ of theorem 2.4.1 is unique up to Vassiliev-Goussarov invariant of order at most $n-1$. More precisely, the difference between maps $v, v^{\prime}: \Sigma \rightarrow \mathbb{R}$ satisfying to (1), (2n) and (3), satisfies to (1) and $\left(2_{n-1}\right)$. (b) Prove the 'only if' part of theorem 2.4.1. $(0),(1),(2),(3)^{*}$ Prove the 'if' part of theorem 2.4 .1 for $n=0,1,2,3$. Hint: for $n=2$ use theorem 2.1.2, for $n=3$ use the coefficient of $h^{3}$ in $J\left(e^{h}\right)$, where $J$ is the Jones polynomial in $t$-parametrization [CDM12, 2.4.2, 2.4.3]. In the remaining problems theorem 2.4.1 can be used without proof. Assertion ' $v(K)=x$ for any singular knot $K$ whose chord diagram is $a$ ' is shortened to ' $v(a)=x$ '. Problem 2.4.3. (a) There exists a unique Vassiliev-Goussarov invariant $v_{2}: \Sigma \rightarrow \mathbb{R}$ of order at most 2 such that - $v_{2}(O)=0$ for the trivial knot $O$, and - $v_{2}(1212)=1$ ((1212) is the 'non-trivial diagram with 2 chords' [PS96, Figure 4.4], 3rd diagram of the first line). Warning: in this problem it is allowed to use theorem 2.4.1 but not theorem 2.1.2. (b,b',c,d) Calculate $v_{2}$ for the right trefoil, left trefoil, figure eight knot and the $5_{1}$ knot. Problem 2.4.4. (a) There exists a unique Vassiliev-Goussarov invariant $v_{3}: \Sigma \rightarrow \mathbb{R}$ of order at most 3 such that - $v_{3}(O)=0$ for the trivial knot $O$ and for the left trefoil $O$, and - $v_{3}(123123)=1$ ((123123) is the 'non-trivial most symmetric diagram with 3 chords', [PS96, Figure 4.4], 5th diagram of the second line). $\left(\mathrm{b}, \mathrm{c}, \mathrm{d}^{*}\right)$ Calculate $v_{3}$ for the right trefoil, figure eight knot and the $5_{1}$ knot. Hints: Problems 2, 3, 4ab, Results/Theorems 11, 13, 14 from [PS96, §4]. Problem 2.4.5. (a) There exists a unique Vassiliev-Goussarov invariant $v_{4}: \Sigma \rightarrow \mathbb{R}$ of order at most 4 such that - $v_{4}(O)=0$ for the trivial knot $O$, for the left trefoil $O$, and for the right trefoil $O$, - $v_{4}(12341234)=2, v_{4}(12341432)=3$ and $v_{4}(12341423)=5$ [PS96, Problem 4.4.b]. $\left(\mathrm{c}^{*}, \mathrm{~d}^{*}\right)$ Calculate $v_{4}$ for the figure eight knot and the $5_{1}$ knot. ## Hints and solutions for problems before the semifinal 1.1.1. (a,b,c,d) 'Probably the best way of solving this problem is to make a model of the trefoil knot and the figure eight knot by using a shoelace and then move it around from one position to the other. Fig. 16 gives some hints concerning transformations of the figure eight knot.' $[\operatorname{Pr} 95$, $\S 2]$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-162.jpg?height=426&width=724&top_left_y=410&top_left_x=687) Figure 16: Isotopy of the figure eight knot (e) Let us consider two knots with coinciding plane diagrams, in a horizontal plane $\pi$. For each point $X$ in the space let $p(X)$ be the line containing $X$, perpendicular to $\pi$. Let $h(X)$ be the height of $X$, relative to $\pi$, which is positive $(h(X)>0)$ if $X$ is in the upper half-space and negative $(h(X)<0)$ if $X$ is in the lower half-space. To each point $A$ of the first knot associate a point $A$ of the second knot by the following procedure. There are two cases: Case 1: The projection of the point $A$ on $\pi$ is not a crossing point on the plane diagram. In this case $p(A)$ intersects the first knot only at the point $A$. Since the plane diagrams coincide, the line $p(A)$ intersects the second knot also at a single point. Define $A^{\prime}$ to be this point. Case 2: The projection of the point $A$ on $\pi$ is a crossing point of the plane diagram. In this case the line $p(A)$ intersects the first knot in an additional point $B$. Since the plane diagrams coincide, the line $p(A)$ intersects the second knot in two points $C$ and $D$, where we assume that $h(C)>h(D)$. If $h(A)>h(B)$, we define $A^{\prime}=C$, and in the opposite case $A^{\prime}=D$. For each point $A$ of the first knot and each number $t \in[0,1]$ let $A(t)$ be the point on the line $p(A)$ with the height $h(A(t))=(1-t) h(A)+t h\left(A^{\prime}\right)$. By construction $A(0)=A, A(1)=A^{\prime}$ and the transformation of the first knot, which moves $A(0)$ in the direction of $A(1)$ with constant speed, so that at the time $t$ it occupies the position $A(t)$, is the required isotopy. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-162.jpg?height=318&width=515&top_left_y=1798&top_left_x=791) Figure 17: The bridge over some crossing point 1.1.2. See fig. 17. For each crossing point of the plane diagram, on the upper edge of the crossing, choose two points, close to the intersection and on the opposite sides of the intersection. Replace the line segment between the two chosen points by a 'bridge' rising above the plane diagram, which connects these two points. After replacing all crossing points by the corresponding bridges, we obtain the required knot. 1.1.3. (a) Use the results of problems 1.4.1, 2.1.1, or 2.2.1. (b) Use the results of problems 1.4.1 or 2.1.1. (c) Use the result of problem 2.2.1. (d) Take any of the connected sums of $n$ trefoil knots. By assertions 2.1.1.c,i the Casson invariant of this knot is $n$. It follows that for different values of $n$ the corresponding knots are not isotopic. 1.1.4. Choose a knot projected to the given plane diagram in the same way as in assertion 1.1.2. Suppose that all 'bridges' lie in the upper half-space w.r.t. the projection plane. By the assumption there are points $X$ and $Y$ on the knot which divide the knot into two polygonal lines $p$ and $q$ such that - $q$ lies in the projection plane and passes only through undercrossings; - $p$ is projected to polygonal line $p^{\prime}$ which passes only through overcrossings. Take a point $Z$ in the upper half-space, and a point $T$ in the lower half-space. Let us construct an isotopy between the given knot and the closed polygonal line $X Z Y T$, which is isotopic to the trivial knot. The isotopy consists of 3 steps, all of them keeping $X, Y$ fixed. Step 1. An isotopy between $q$ and $X T Y$. Suppose that $q=A_{0} A_{1} \ldots A_{n}$, where $A_{0}=X$ and $A_{n}=Y$. Then the isotopy is given by $$ A_{0} A_{1} \rightarrow A_{0} T A_{1}, \quad T A_{1} A_{2} \rightarrow T A_{2}, \quad T A_{2} A_{3} \rightarrow T A_{3}, \quad \ldots T A_{n-1} A_{n} \rightarrow T A_{n} $$ Step 2. An isotopy between $p$ and $p^{\prime}$. Remove all the 'bridges' by elementary moves. Step 3. An isotopy between $p^{\prime}$ and $X Z Y$. This is done analogously to step 1 . 1.1.5. Follows by assertion 1.1.4. Another idea of the proof (cf. [PS96, Theorem 3.8]). Denote by $\pi$ the horizontal plane containing the plane diagram. For each point $X$ in the space, $p(X)$ and $h(X)$ are defined in the solution of the problem 1.1.1.e. Let $l$ be a line in the plane, which passes through a vertex $A_{0}$ of the plane diagram, while the whole diagram is contained in one of the two half-planes determined by $l$. Let $A_{0}, A_{1}, \ldots A_{n}$ be all vertices of the plane diagram, in the order of their appearance, while we move along the diagram in some direction. Choose points $B_{0}, \ldots, B_{n}$ so that $A_{i} \in p\left(B_{i}\right)$ for $i=1, \ldots, n$, and $h\left(B_{i}\right)h\left(B_{n}\right)$. We claim that the knot $B_{0} \ldots B_{n} B_{n+1}$ is isotopic to the trivial knot. Indeed, by the choice of the line $l$, the projection of the knot onto any plane, perpendicular to the line $l$, is a closed polygonal line without self-intersections. It remains to modify crossing in the plane diagram so that they are in agreement with the projection of the constructed knot to the plane $\pi$. 1.1.6. See $[\mathrm{PS} 96, \S 1.7]$. Remark. Since [PS96, §1.6] does not contain as rigorous definition of Reidemeister moves as that of plane isotopies, ${ }^{10}$ the argument in [PS96, §1.7] does not constitute a rigorous proof. We believe that a rigorous proof can be recovered using rigorous definition of Reidemeister moves. 1.2.1. (a) This follows by (b) (or can be proved independently). (d) This follows by (e) (or can be proved independently). (e) See figure 18 .[^30]![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-164.jpg?height=472&width=1186&top_left_y=69&top_left_x=450) Figure 18: Isotopy of the Whitehead link (f) Take three ellipses given by the following three systems of equations: $$ \left\{\begin{array}{c} x=0 \\ y^{2}+2 z^{2}=1 \end{array}, \quad\left\{\begin{array} { c } { y = 0 } \\ { z ^ { 2 } + 2 x ^ { 2 } = 1 } \end{array} \quad \text { and } \quad \left\{\begin{array}{c} z=0 \\ x^{2}+2 y^{2}=1 \end{array}\right.\right.\right. $$ See figure 19. Take the quadrilaterals circumscribed around these ellipses and symmetric w.r.t. the coordinate axes. Then the straight line is given by $x=y=z$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-164.jpg?height=752&width=1592&top_left_y=1048&top_left_x=255) Figure 19: Borromean rings 1.2.2. (a,c) Use results of problem 1.3.2. (b,d) Use, for example, results of problems 2.2.1 and 2.2.2. 1.3.1. In the plane passing through $A, B, C$ denote by $\langle A B C\rangle$ the part bounded by the triangle $A B C$. $(\mathrm{i} \Leftrightarrow$ ii) A segment $X Y$ passes above the segment $B C$ as seen from $A$ if and only if $X Y$ intersects $\langle A B C\rangle$. The outline of $D E F$ is the union of segments $D E, E F$ and $D F$. Hence $\mathrm{i} \Leftrightarrow \mathrm{ii}$. (i $\Leftrightarrow$ iii) The outline of $D E F$ is either disjoint from the surface of tetrahedron $O A B C$ or intersects the surface at exactly two points. Hence the following three properties are equivalent: - the outline of $D E F$ intersects $\langle A B C\rangle$ at exactly one point; - the outline of $D E F$ intersects exactly one of the triangles $\langle O A B\rangle,\langle O B C\rangle$ and $\langle O A C\rangle$. - the outline of $A B C$ passes below exactly one side of $D E F$ as seen from $O$. 1.3.2. (a) Answers (from left to right): $1,0,0,0$. (b) Prove the statement separately for every Reidemeister move. For moves I and III the number of crossing points where the first component passes above the second one does not change. For move II this number changes by 0 or $\pm 2$. 1.3.3. (a) Take a plane diagram of a link. By the Parity lemma (stated before problem 1.3.3) the number of crossing points where the first component passes above the second one has the same parity as the number of crossing points where the second component passes above the first one. This completes the proof. (b) An example is the fourth link in fig. 7. One can prove that this link is not isotopic to the trivial one using linking number, see $\S 1.6$. 1.3.4. Suppose that $\mathrm{f}$ is another invariant aside from $\mathrm{lk}_{2}$ satisfying the assumptions. Then $\mathrm{f}-\mathrm{lk}_{2}$ is an isotopy invariant assuming zero value on the trivial link and invariant under crossing changes. By the analogue of lemma 1.1.5 for links any plane diagram of a link can be obtained from the diagram of a link isotopic to the trivial link by some crossing changes. Hence $\mathrm{f}-\mathrm{lk}_{2}=0$. 1.3.5. The proof should not be hard, and we encourage you to supply the details. 1.3.6. See $[$ Sk14, §1, Theorem 1.1]. 1.4.1. (a) If $P$ is a point on the plane diagram as in problem 1.1.4, then there are no $P$-skew pairs of crossings. Hence the $P$-Arf invariant is zero. (b) Answer: 0. The trivial knot has no crossings and no skew pairs of crossings. Therefore the Arf invariant of this knot is 0 for any choice of the basepoint. (c) Answer: 1. The trefoil knot has 3 crossings. For any basepoint $P$ exactly one pair of crossings is $P$-skew. Hence $P$-Arf invariant of the trefoil knot is 1 . (d) Answer: 1 . (e) It suffices to show that the Arf invariant remains unchanged when the basepoint moves through one crossing on the plane diagram. Let $P_{1}$ and $P_{2}$ be two basepoints such that the segment $P_{1} P_{2}$ contains exactly one crossing point $X$. Consider two cases. Case 1: $P_{1} P_{2}$ passes through undercrossing. Then $X$ does not form either $P_{1}$-skew or $P_{2}$-skew pair with any other crossing. Hence $P_{1^{-}}$and $P_{2}$-Arf invariants of the diagram are equal. Case 2: $P_{1} P_{2}$ passes through overcrossing. Then $X$ divides the diagram into two closed polygonal lines $q_{1}$ and $q_{2}$ such that $P_{1}$ lies on $q_{1}$ and $P_{2}$ lies on $q_{2}$. Denote by $n_{1}$ (respectively, $n_{2}$ ) the number of intersections of $q_{1}$ and $q_{2}$ for which $q_{1}$ passes above $q_{2}$ (respectively, $q_{2}$ passes above $q_{1}$ ). Denote by $N_{1}$ (respectively, $N_{2}$ ) the number of $P_{1}$-skew (respectively, $P_{2}$-skew) pairs formed by $X$ and some intersection of $q_{1}$ and $q_{2}$. Then $$ \operatorname{arf}_{P_{1}} D-\operatorname{arf}_{P_{2}} D=N_{1}-N_{2}=n_{1}-n_{2} \underset{\overline{2}}{\overline{2}} n_{1}+n_{2} \underset{\overline{2}}{\overline{2}} 0 $$ where $D$ is the given plane diagram. Here - the first equality holds because a pair of crossings is either $P_{1}$-skew or $P_{2}$-skew (but not both) if and only if the pair is formed by $X$ and some intersection of $q_{1}$ and $q_{2}$; - the second equality holds because $N_{1}=n_{1}$ and $N_{2}=n_{2}$; indeed, an intersection of $q_{1}$ and $q_{2}$ forms a $P_{1}$-skew (respectively, $P_{2}$-skew) pair with $X$ if and only if at this intersection $q_{1}$ passes above (respectively, below) $q_{2}$; - $\overline{\text { 2 }}$ is congruence modulo 2 ; - the last congruence follows by Parity lemma for $q_{1}$ and $q_{2}$. (f) Prove the statement for each Reidemeister move separately. The common idea is to cleverly choose a basepoint. Type I move. Take basepoints before and after the move as in fig. 20 (left). Check that the crossing $A$ does not form a $P$-skew pair with any other crossing. Type II move. Take basepoints before and after the move as in fig. 20 (middle). Check that neither of the crossings $A$ and $B$ forms a $P$-skew pair with any other crossing. Type III move. Take basepoints before and after the move as in fig. 20 (right). Check that neither of the crossings $A, B$ forms a $P$-skew pair with any other crossing and that neither of the crossings $A^{\prime}, B^{\prime}$ forms a $P^{\prime}$-skew pair with any other crossing. Then check that a crossing $X$ distinct from $A, B, C$ forms $P$-skew pair with $C$ if and only if $X$ forms $P^{\prime}$-skew pair with $C^{\prime}$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-166.jpg?height=206&width=1626&top_left_y=323&top_left_x=236) Figure 20: Arf-invariant does not change under Reidemeister moves 1.4.2. (a) Take basepoints $P_{+}, P_{-}$as in fig. 21. Check that the crossing $A_{-}$does not form a $P_{-}$-skew pair with any other crossing in $K_{-}$. Then check that the number of crossings which form a $P_{+}$-skew pair with $A_{+}$in $K_{+}$equals $\mathrm{lk}_{2} K_{0}$ modulo 2 . ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-166.jpg?height=268&width=588&top_left_y=858&top_left_x=752) Figure 21: To the proof of skein relation for Arf invariant (b) Take any of the connected sums of the two trefoil knots. By assertions 1.4.1.c and 1.5.4.d the Arf invariant of this knot is 0 . However by assertions 2.1.1.c,i the Casson invariant of this knot is 2, hence this knot is not isotopic to the trivial knot. 1.4.3. Existence. Follows from assertions 1.4.1.b and 1.4.2.a. Uniqueness. The proof goes along the lines of the proof of theorem 1.3.4. Use lemma 1.1.5 itself instead of its analogue for links. 1.4.4. See $[K a 87$, pp. 75-78]. 1.4.5. See [CG83, Theorem 2]. 1.5.1. (b) First solution. An oriented polygonal line is called positive if the bounded part of the plane is always on the right side of each of its oriented segments (see the Jordan theorem in remark after theorem 1.1.3). Prove that the positivity of an oriented polygonal line is preserved by elementary moves. Hint to the second solution. The positivity can be equivalently defined as follows. We say that an oriented polygonal line $A_{1} \ldots A_{n}$ is positive if for each of its inner (interior) points $O$ the sum of oriented angles $\angle A_{1} O A_{2}+\angle A_{2} O A_{3}+\ldots+\angle A_{n-1} O A_{n}+\angle A_{n} O A_{1}$ is always positive (i.e. the winding number of the oriented polygonal line around any interior point is positive). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-166.jpg?height=301&width=1743&top_left_y=2137&top_left_x=169) Figure 22: Proof of $K \# O=K$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-167.jpg?height=268&width=726&top_left_y=83&top_left_x=686) Figure 23: Proof of $(K \# L) \# M=K \#(L \# M)$ 1.5.2. (a), (b), (c) Each of the three indicated oriented knots is transformed into the oriented knot with the opposite orientation by the rotation through the angle $\pi$ around the 'vertical' axis passing through the 'upper' point of the knot (see the leftmost diagram in fig. 1 , the first and the second row for the trefoil and the figure eight knot, respectively). This rotation is included into a continuous family of rotations through the angle $\pi t, t \in[0,1]$, with respect to the same line. This is the required isotopy. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-167.jpg?height=1141&width=566&top_left_y=836&top_left_x=755) Figure 24: Proof of $K \# L=L \# K$ 1.5.4. (a) See fig. 22 . (b) It is sufficient to take a small knot of the class $L$ and push it through a knot from the class $K$, see fig. 24 . (c) Isotopic classes of both the left hand and the right hand side of the equality have a common representative exhibited in fig. 23 . (d) Choose basepoint close to the 'place of connection'. Check that all skew pairs of crossings in $K \# L$ are obtained from the skew pairs of crossings in $K$ and in $L$. 1.5.5. (d) Check that all crossings of different components in $K \# L$ are obtained from such crossings in $K$ and in $L$. 1.5.6. (a) As an example we can take equal links consisting of a trefoil and an unknot in disjoint cubes. Cf. [PS96, Figure 3.16]. (b) See $[$ As]. Fig. 25 presents an alternative example suggested by A. Ryabichev. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-168.jpg?height=952&width=1030&top_left_y=281&top_left_x=534) Figure 25: Connected sum of isotopy classes of ordered links is not well-defined 1.6.1. Answers: (a) 1 ; (b) 0 . (c) The proof is analogous to assertion 1.3.2. It suffices to check that the signs of all crossing points does not change. 1.6.2. (a) The proof is analogous to assertion 1.3.2.b. Take a plane diagram of a link. By the Triviality lemma (stated before problem 1.6.2) the sum of signs of crossing points where the first component passes above the second one has opposite sign to the sum of signs of crossing points where the second component passes above the first one. Switching the components negates the sign of every crossing point. This completes the proof. (b) Reversing the orientation of either of the components negates the sign of every crossing point. (c) Take the connected sum of 5 Hopf links oriented so that their linking numbers equal to -1 . (d) The proof is analogous to assertion 1.5.4. The signed set of crossing points of plane diagram of $K \# L$ is the union of the signed sets of crossing points of plane diagrams of links $K$ and $L$. (e) An example is the Whitehead link. The Whitehead link is not isotopic to the trivial link by theorem 1.2.2. 1.6.3. The proof is analogous to theorem 1.3.4. ## Hints and solutions for some problems after the semifinal 2.1.1. Answers: (b) 0 ; (c) 1 ; (d) -1 . (a) Take any connected sum of five figure eight knots. By (d) and assertion 2.1.1.i below the Casson invariant of this knot is -5 . (b) The trivial knot has no crossings, and so no skew pairs of crossings. Therefore the Casson invariant of this knot is 0 . (c) All three crossings of the trefoil knot have the same sign. Since the trefoil knot has exactly one linked pair of crossings (regardless the choice of the base-point), we obtain that the Casson invariant of this knot is 1 . (e) The proof is analogous to assertion 1.4.1.e. Use the Triviality lemma stated after problem 1.6.1. Here and below keep in mind the signs of intersection points! (f) The proof is analogous to assertion 1.4.1.f. (g) The proof is analogous to assertion 1.4.2.a. (h) Take any connected sum of the trefoil knot and the figure eight knot. By assertions 2.1.1.c,d and 2.1.1.i the Casson invariant of this knot is 0 . However, by assertions 2.3.4.d,e and 2.3.6.b the Conway polynomial of this knot is $\left(t^{2}+1\right)\left(t^{2}-1\right) \neq 1$. Hence this knot is not isotopic to the trivial knot. (i) The proof is analogous to assertion 1.5.4.d. 2.1.2. The proof is analogous to theorem 1.4.3. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-169.jpg?height=480&width=980&top_left_y=1125&top_left_x=558) Figure 26: A 3-coloring of a link and 5-coloring of the 51 knot 2.2.1. Answers: b,e,h - 3-colorable, a,c,d,f,g,i,j - not 3-colorable. For a proper coloring of a diagram of trefoil knot see the attached copy of [Pr95, p. 30, figure 4.3]. For a proper coloring of the last diagram from fig. 7 see fig. 26 left. (This diagram was erroneously stated to be not 3-colorable in $[\operatorname{Pr} 95, \S 4]$. This minor mistake was found by L.M. Bannöhr, S. Zotova and L. Kravtsova.) 2.2.2. (a) See the attached copy of [Pr95, pp. 29-30, Theorem 4.1]. (b) Follows from (a) and assertions 2.2.1.d-j (see the attached copy of [Pr95, p. 30]). The last diagram from fig. 7 is distinguished from the trivial link by the number of proper colorings of a plane diagram. Prove that this number is preserved under the Reidemeister moves. 2.2.3. A plane diagram is 5 -colorable if there exists a coloring of its strands in five colors $0,1,2,3,4$ so that - at least two colors are used; - at each crossing if the upper strand has color $a$ and two lower strands have colors $b$ and $c$, then $2 a \equiv b+c(\bmod 5)$. Similarly to assertion 2.2.2.a one can prove that the 5 -colorability of a plane diagram is preserved under Redemeister moves. The $5_{1}$ knot is 5 -colorable, see fig. 26 , right. The trivial knot is not. This completes the proof. 2.3.1. These are particular cases of mod2 version of theorem 2.3.3. 2.3.2. Answers: (a, b) 0 ; (c) 1 (independently of the choice of orientation). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-170.jpg?height=347&width=322&top_left_y=236&top_left_x=880) Figure 27: 4-Borromean rings Remark. The invariants arf, $a_{3}$ for links may depend on the orientation on the components (for $a_{3}$ see the attached copy of [CDM12, 2.3.4]). 2.3.3. The uniqueness is analogous to theorems 2.1 .2 and 1.4.3; solve first problem 2.3.4. Deduction of (a) from (b). It suffices to show that all invariants $c_{n}$ defined in (b) are preserved under changes of the order of the components. Let $D$ be a plane diagram of some link with two or more components and let $D^{\prime}$ be a plane diagram obtained from $D$ by a change of the components' order. By cr $D$ denote the number of crossings in $D$. By $u(D)$ denote the minimal possible number of crossing changes needed to obtain a diagram of a link which is isotopic to the trivial one from $D$ (such sequence of crossing changes exists by the analogue of lemma 1.1.5 for links). The proof is by induction on cr $D$. If cr $D=0$, then $D$ is a diagram of a link which is isotopic to the trivial one and by assertion 2.3.4.b we have $C(D)=0$ for any ordering of the components. Suppose that cr $D>0$; then continue the proof by induction on $u(D)$. If $u(D)=0$, then $D$ is a diagram of a link which is isotopic to the trivial one; this case is considered above. Suppose that $u(D)>0$ and let $D_{*}$ be a link obtained from $D$ by a crossing change and such that $u\left(D_{*}\right)0$; then continue the proof by induction on $u(D)$. If $u(D)=0$, then $K$ is isotopic to the trivial link; this case is considered above. Suppose that $u(D)>0$ and let $D_{*}$ be a link obtained from $D$ by a crossing change and such that $u\left(D_{*}\right)Мельников, Борис Френкин, Алексей Чиликов ## А. ВВОДНАЯ СЕРИЯ. КЛАССИЧЕСКОЕ УРАвНЕНИЕ ПЕЛЛЯ. Вводная серия А относится к классическим фактам об уравнении Пелля $x^{2}-D y^{2}=$ 1 для целых чисел. Решение $( \pm 1,0)$ называется тривиальным, а остальные - нетривиалъными. Задача А.1. Пусть $\left(x_{1}, y_{1}\right)$ - решение уравнения Пелля, т.е. $x_{1}^{2}-D y_{1}^{2}=1$. Пусть $\left(x_{1}+\sqrt{D} y_{1}\right)^{n}=x_{n}+\sqrt{D} y_{n}$. Докажите, что тогда $\left(x_{n}, y_{n}\right)$ - тоже решение уравнения Пелля (отметим, что $x_{-n}=x_{n}, y_{-n}=-y_{n}$ ). Решение $\left(x_{n}, y_{n}\right.$ ) называется степенью решения $\left(x_{1}, y_{1}\right)$. Введите понятие произведения решений. Задача А.2. Дожажите, что все нетривиалъные решения (если существуют) являются степенъю одного решения $\left(x_{1}, y_{1}\right.$ ) (с точностью до знака при $x$ ). Задача А.3. Докажите, что если D есть квадрат целого, то нетривиалъных решений у уравнения Пелля нет. В следующих пунктах серии считаем, что $D \neq m^{2} \forall m \in \mathbb{Z}, D \in \mathbb{Z}$. Задача А.4. Покажите, что существуют $M>0$ и точка $(x, y)$ с ненулевыми цельми координатами, что $\left|x^{2}-D y^{2}\right| \leq M$. Задача А.5. С помощъю лемми Минковского докажите, что существует $M>0$, для которого таких точек бесконечно много. Задача А.6. Докажите, что существует такое целое положительное число $k<$ $M$, что уравнение $\left|x^{2}-D y^{2}\right|=k$ имеет бесконечно много решений. Задача А.7. С помощью предъдущих пунктов докажите, что нетривиальное решение уравнения Пелля существует. Задача А.8. Onишите все раииональные решения уравнения Пелля в общем виде. ## В. УРАвНЕНИЯ ПЕЛЛЯ ДЛЯ МНОГОЧЛЕНОв. Задача В.1. Найдите все пары $(P(x), Q(x))$ многочленов над $\mathbb{R}$ такие, что $$ P^{2}(x)-\left(x^{2}-1\right) Q^{2}(x)=1 $$ Задача В.2. Решите уравнение Пелля для рационалъных функиий. Задача В.3. Решите уравнения а) $P^{2}+Q^{2}=R^{2}$, б) $P^{2}+Q^{2}=1$ для многочленов с комплексными коэффициентами. Задача B.4. Докажите, что для $n>2$ и для многочленов $P, Q$ существуют комплексные числа $\psi_{1}, \psi_{2}, \psi_{3}$, такие, что $\psi_{i}^{n}=-1$ и $P^{n}+Q^{n}$ делится на $(P-$ $\left.\psi_{1} Q\right)\left(P-\psi_{2} Q\right)\left(P-\psi_{3} Q\right)$ Задача В.5. Пусть $P^{n}+Q^{n}=R^{n}$. В условиях n. В4 докажите, что существуют такие многочлены $R_{1}, R_{2}, R_{3}$, что $P-\psi_{i} Q=R_{i}^{n}$. Задача В.6. Докажите, что для $n>2$ уравнение Ферма $x^{n}+y^{n}=z^{n}$ для многочленов и для раиионалъных функиий не имеет нетривиалъных решений. ## С. ПРИМЕНЕНИЯ ТЕОРЕМЫ КОЛЛАРА. Под нетривиалъным решением здесь понимается неконстантное решение. Задача С.1. Решите уравнение Пелля, когда D есть квадрат многочлена. Задача С.2. Рассмотрим уравнение над множеством многочленов из $\mathbb{Z}[x]$ : $$ P^{2}(x)-\left(R^{2}(x)-1\right) Q^{2}(x)=1 $$ Здесъ $R$ - уже произвольный неконстантный многочлен, а не просто переменнал. Докажите, что множество решений состоит из степеней одного нетривиалъного: $(R, 1)$. Задача С.3. Теорема Коллара *** Аналогичное утверждение верно для многочленов из $\mathbb{C}[x]$ Задача С.4. Решите уравнение (1) для случая, когда $R$ - константа. Отделъно рассмотреть случай, когда $R^{2}=1$ и когда $R^{2} \neq 1$. Далее в серии С предполагается, что $R \neq$ const. Кроме того, можно пользоваться теоремой Коллара. Задача С.5. Докажите, что $Q$ имеет вид $Q_{n}=\sum_{k=0}^{[n / 2]}\binom{n}{2 k+1}\left(R^{2}-1\right)^{k} R^{n-1-2 k}$ для некоторого целого п. Выпишите аналогичную формулу для $P$. Задача С.6. Докажите, что $Q_{n} \equiv n \bmod R-1$, mo есть $Q_{n}-n=(R-1) S$, для некоторого многочлена $S$. Задача С.7. Решите систему уравнений для многочленов из $\mathbb{C}[x]$ $$ \left\{\begin{array}{c} X^{2}-\left(R^{2}-1\right) Y^{2}=1 \\ Y-(R-1) Z=V \\ V \cdot U=1 \end{array}\right. $$ Задача С.8. Основная задача. Докажите, что для произволъного уравнения $W$ в челых числах можно построить систему уравнений для многочленов из $\mathbb{C}[x]$, имеющую нетривиальное решение тогда и только тогда, когда имеет решение уравнение $W$. ## D. АФФиннЫЕ МнОгОБРАЗИЯ. Задача D.1. Задайте следующие множества уравнениями или наборами уравнений: (1) эллипс; (2) пару прямвх на плоскости; (3) окружность и эллипс; (4) окружность и параболу, проходящие через начало координат и касающиеся в этой точке, причем осъ ординат - осъ параболъ; (5) экваториалъную окружсность на единичной сфере в трехмерном пространстве; (6) п-мерный тор (то есть декартово произведение $n$ окружностей). Определение 1. Идеалом в кольце $R$ называется такое множество $I$, что (1) $I \cdot R \subset I$; (2) $I+I \subset I$. Идеал называется главным, если он порождается одним элементом. Определение 2. Пусть $R$ некоторое колъцо. Будем называть афинным многообразием над $R$ множество $X$ решений конечного числа алгебраических уравнений от нескольких переменных, то есть тех, которые можно записать в виде $P\left(x_{1}, \ldots, x_{n}\right)=0$, где $P-$ многочлен. Определение 3. Пусть $X$ задано несколъкими уравнениями в $R\left[x_{1}, \ldots, x_{n}\right]$. Левые части этих уравнений порождают идеал в кольце $R\left[x_{1}, \ldots, x_{n}\right]$, который мы будем обозначать $I(X)$. Кольцом функций $R[X]$ на $X$ будем называть кольцо $R\left[x_{1}, \ldots, x_{n}\right] / I(X)$. Задача D.2. Убедитесъ в том, что для каждого элемента из $R[X]$ в каждой точке многообразия $X$ можно однозначно определить значение. Задача D.3. Покажите, что колъцо функиий полукубической параболы, заданной уравнением $y^{2}=x^{3}$ в $\mathbb{C}^{2}$, изоморфно кольиу $\mathbb{C}\left[t^{2}, t^{3}\right] \subset \mathbb{C}[t]$. Задача D.4. Пусть $X$ и Y аффинные многообразия и задано отображение $\varphi: R[Y] \rightarrow$ $R[X]$. Постройте по $\varphi$ отображение $f: X \rightarrow Y$, которое бы индуцировало $\varphi$. Замечание 4. Нас будут интересовать толъко отображения между многообразия.ми, которые происходят из отображения соответствующих колец. Такие отображения будем называть алгебраическими. Задача D.5. Рассмотрим на $\mathbb{C}^{2}$ действие группы $\mathbb{Z} / 2 \mathbb{Z}$ такое, что нетривиалъный элемент группы переводит $(x, y)$ в $(-x,-y)$. Это действие индуиирует автомор$ф и з м ~ к о л ь ц а ~ \mathbb{C}[x, y]$. Покажите, что подколъцо, инвариантное относительно этого действия, изоморфно $\mathbb{C}[u, v, w] /\left(u v-w^{2}\right)$. Постройте изоморфизм между аффинным многообразием, заданным этим колъцом, и фактором $\mathbb{C}^{2}$ по действию $\mathbb{Z} / 2$ (как топологическими пространствами). Пусть $I$ некоторый идеал в кольце $R\left[x_{1}, \ldots, x_{n}\right]$. Многообразие, соответствующее кольцу $R\left[x_{1}, \ldots, x_{n}\right] / I$ будем обозначать $V(I)$. Радикалом идеала $I$ будем называть множество $\operatorname{rad}(I)$ всех таких элементов кольца $R\left[x_{1}, \ldots, x_{n}\right]$, что некоторая их степень содержится в $I$. Теорема 5 (Теорема Гильберта о нулях). Пусть $k$ некоторое алгебраически замкнутое поле. Тогда для любого идеала $I \subset k\left[x_{1}, \ldots, x_{n}\right]$ въполнено $$ I(V(I))=\operatorname{rad}(I) $$ Задача D.6. Исполъзуя теорему, покажите, что точки аффинного многообразия $X$ над $\mathbb{C}$ взаимнооднозначно соответствуют максималъным идеалам колвца $\mathbb{C}[X]$. Задача D.7. Пусть $X$ и $Y$ - аффинные многообразия над $\mathbb{C}$. Пусть задано отображение $\varphi: \mathbb{C}[Y] \rightarrow \mathbb{C}[X]$. Постройте по нему отображение из множества максимальных идеалов кольца $\mathbb{C}[X]$ в множество максимальных идеалов кольца $\mathbb{C}[Y]$, не используя теорему Гилъберта о нулях. Убедитесъ, что это отображение совпадает с отображением, построенным в задаче 4. Определение 6. Отображение $\iota: X \rightarrow Y$ между аффинными многообразиями будем называть замкнутым вложением, если $R[X]=R[Y] / I$ и отображение колеи, функиий, соответствующее ь является отображением факторизации $R[Y] \rightarrow$ $R[Y] / I$. Замечание 7. Легко заметить, что любое афинное многообразие над $\mathbb{C}$ канонически замкнуто вложено в некоторое $\mathbb{C}^{n}$. Задача D.8. Задайте двукратную обмотку как замкнутое вложение окруэности в mop. Задача D.9. Рассмотрим естественное вложение $(\mathbb{C} \backslash 0)^{n}$ в $\mathbb{C}^{n}$. Покажите, что это вложение является алгебраическим отображением. Докажите, что оно не является замкнутым вложением. Задача D.10. Покажите, что стандартное вложение $\varphi: \mathbb{C}\left[t^{2}, t^{3}\right] \hookrightarrow \mathbb{C}[t]$ не задает замкнутое вложение прямой в полукубическую параболу. ## Е. ОТОБРАЖЕНИЯ АФФИННЫХ МНОГООБРАЗИЙ. Не забывайте подходить с вопросами и за подсказками! Определение 8. Множество $R$ с двумя ассоциативными коммутативными бинарными операциями $\cdot u+$ (умножением и сложением) называется кольцом, если выполнены следующие свойства: (1) $\exists 0 \in R: 0+a=a+0=a \forall a \in R$; (2) $\forall a \in R \exists b \in R: a+b=b+a=0$; (3) $\exists 1 \in R: 1 \cdot a=a \cdot 1=a \quad \forall a \in R$; (4) $a \cdot(b+c)=a \cdot b+a \cdot c$; (5) $(a+b) \cdot c=a \cdot c+b \cdot c$. Замечание 9. Основными кольиами (кроме множеств целых, рациональных, вещественных и комплексных чисел), которые нас интересуют, являются колъца многочленов, а так же кольца функций на аффинных многообразиях. Замечание 10. Многочлени $P_{1}\left(x_{1}, \ldots, x_{n}\right), \ldots P_{k}\left(x_{1}, \ldots, x_{n}\right)$, которые задают аффинное многообразие $X$, порождают такэже некоторое подмножество $I=P_{1}$. $R\left[x_{1}, \ldots, x_{n}\right]+\ldots+P_{k} \cdot R\left[x_{1}, \ldots, x_{n}\right] \subset R\left[x_{1}, \ldots, x_{n}\right]$. Все элементы этого множсества зануляются на точках многообразия. Такое множество является идеалом в кольце функиий на $X$. Задача Е.1. Пусть I некоторый идеал в кольце многочленов $\mathbb{C}\left[x_{1}, \ldots, x_{n}\right]$. Через $X_{I}$ обозначим многообразие, соответствующее данному идеалу. Докажите следующие свойства: (1) $I \subseteq J \Rightarrow X_{J} \subseteq X_{I}$ (2) $X_{I} \cup X_{J}=X_{I \cdot J}=X_{I \cap J}$; (3) если $I+J=R$, mo $I \cdot J=I \cap J$; (4) идеалъ $I_{1}=\langle x\rangle$ и $I_{2}=\left\langle x^{2}\right\rangle$ в кольце $\mathcal{R}[x]$ задают одинаковые множества нулей. Задача E.2. (1) Докажите, что все идеалы кольца многочленов от одной переменной $\mathcal{R}[x]$ являются главными. Такие колъиа называются кольцами главных идеалов (КГИ). (2) Приведите пример колъца, не являющегося КГИ. Определение 11. Дан идеал $I \subseteq R$. Факторкольцом $R / I$ называется множество классов эквивалентности элементов колъца $R$ по модулю $I: a \sim b \Leftrightarrow a-b \in I$. Умножение и сложение на факторкольце задаются следующим образом: (1) $(a+I)+(b+I)=(a+b)+I$; (2) $(a+I) \cdot(b+I)=(a \cdot b+I)$. Задача Е.3. Проверъте корректность данного определения. Докажите, что факторкольио является кольцом. Замечание 12. В задаче 2 из листочка "Многообразия и уравнения"мы уже неявно работали с факторкольцом. А именно, мы показали, что для многообразия $X_{I}$ над $\mathbb{C}$, заданного идеалом $I$ ( $X_{I}$ - это множество общих нулей всех элементов из I), все значения полиномиальных функиий на $X_{I}$ определяются факторкольиом $\mathbb{C}\left[x_{1}, \ldots, x_{n}\right] / I$ Определение 13. Пусть многообразие $X$ над $R$ задано идеалом $I(X)$ (здесъ мъ подразумеваем, что набор уравнений порождает некоторый идеал). Тогда кольцом функций $R[X]$ многообразия $X$ называется факторколъцо $R\left[x_{1}, \ldots, x_{n}\right] / I(X)$. Замечание 14. В задаче 5 из предыдущей серии было показано, что любой идеал $I \subset \mathbb{C}\left[x_{1}, \ldots, x_{n}\right]$ имеет конечный набор порождающих $\Gamma=\left\{g_{1}, \ldots, g_{k}\right\}$. Рассмотрим идеал $G$, порожденный наибольшими мономами всех элементов из I. В задаче 5 мы доказали, что старшие мономы элементов $\Gamma$ также порождажт идеал G. Такой набор образующих идеала I принято называть базисом Грёбнера. Рассмотрим на $\mathbb{C}^{2}$ действие инволюции $\varphi:(x, y) \rightarrow(-x,-y)\left(\right.$ т.е. $\left.\varphi^{2}=I d\right)$. Эта инволюция задает отображение $\varphi^{*}$ из кольца $\mathbb{C}[x, y]$ в себя. Покажите, что множество инвариантных элементов относительно $\varphi^{*}$ образует подкольцо изоморфное $\mathbb{C}[u, v, w] /\left(u v-w^{2}\right)$. Определение 15. Отображение $\iota: X \rightarrow Y$ между аффинными многообразиями будем называть замкнутым вложением, если $R[X]=R[Y] / I$ и отображение, соответствующее ь является отображением факторизации $R[Y] \rightarrow R[Y] / I$. Замечание 16. Легко заметить, что любое афинное многообразие над $\mathbb{C}$ канонически замкнуто вложено в некоторое $\mathbb{C}^{n}$. Задача Е.4. Рассмотрим естественное вложение $(\mathbb{C} \backslash 0)^{n}$ в $\mathbb{C}^{n}$. Покажите, что это вложение является алгебраическим отображением. Докажите, что оно не является замкнутым вложением. Задача Е.5. Покажите, что стандартное вложение $\varphi: \mathbb{C}\left[t^{2}, t^{3}\right] \hookrightarrow \mathbb{C}[t]$ не задает замкнутое вложение прямой в полукубическую параболу. Задача E.6. В этой задаче мы построим нетривиалъное алгебраическое вложение прямой $\mathcal{R}$ в $\mathcal{R}^{3}$. Для начала рассмотрим отображение $$ \varphi: \mathcal{R} \rightarrow \mathcal{R}^{2}: t \mapsto\left(t^{3}-3 t, t^{4}-4 t^{2}\right) $$ Убедитесъ, что образ этого отображения в $\mathcal{R}^{2}$ является проекиией на плоскость узла трилистника, из которого выкинули одну точку (она лежит на бесконечности). Придумайте такой многочлен $h(t)$, что отображение $$ t \mapsto(\varphi(t), h(t)) $$ переводит прямую в трилистник в $\mathcal{R}^{3}$, из которого выкинули одну точку (она так же будет лежать на бесконечности). Замечание 17. Очевидно, что не у любой системы полиномиальных уравнений есть решение. Например, у системы $\{x-y=0, x-y+1=0\}$ нет решений. Заметим, что идеал, который порождают эти два многочлена совпадает со всем колъиом $R[x, y]$ так как $1 \in I$. Таким образом, если идеал I порождается элементами $g_{1}, \ldots, g_{k}$ и существуют элементъ $h_{1}, \ldots, h_{k} \in R\left[x_{1}, \ldots, x_{n}\right]$, такие, что $$ g_{1} h_{1}+\ldots g_{k} h_{k}=1 $$ то I совпадает со всем $R\left[x_{1} \ldots, x_{n}\right]$. В следующей задаче мы доказываем, что если $1 \notin I$, то найдется общее решение для всех элементов $I$. Задача Е. 7 (Слабая теорема Гильберта о нулях). Любой собственный идеал $I \subsetneq$ $\mathbb{C}\left[x_{1}, \ldots, x_{n}\right]$ задает непустое мно»сетво решений. Докажем эту теорему в несколько шагов. Введем отображение подстановки числа $a \in \mathbb{C}$ на место первой координать: $e v_{a}: \mathbb{C}\left[x_{1}, \ldots, x_{n}\right] \rightarrow \mathbb{C}\left[x_{2}, \ldots, x_{n}\right]: f\left(x_{1}, \ldots, x_{n}\right) \mapsto f\left(a, x_{2}, \ldots, x_{n}\right)$. (1) Пусть $I \cap \mathbb{C}\left[x_{1}\right]=\left\langle p\left(x_{1}\right)\right\rangle$, deg $p>0$. Тогда существует $a \in \mathbb{C}$ такое, что $e v_{a}(I) \subsetneq \mathbb{C}\left[x_{2}, \ldots, x_{n}\right]$. (2) Предположим, что $I \cap \mathbb{C}\left[x_{1}\right]=\emptyset$. Докажите, что найдется такое $a \in \mathbb{C}$ что, выполнено еv $v_{a}(I) \subsetneq \mathbb{C}\left[x_{2}\right]$. (3) Предположим, что $I \cap \mathbb{C}\left[x_{1}\right]=\emptyset$. Докажите, что найдется такое $a \in \mathbb{C}$, что выполнено еv $v_{a}(I) \subsetneq \mathbb{C}\left[x_{2}, \ldots, x_{n}\right]$ (4) Докажите теорему. ## F. КОМПЛЕКСНЫЙ СЛУчАЙ. Нам потребуется следующая теорема, которой был посвящен проект https://www.turgor.ru/lktg/2007/2/index.php . Теорема Матиясевича. Не существует алгоритма, позволяющего по набору коэффициентов многочлена от нескольких переменных $H\left(x_{1}, \ldots, x_{m}\right)$ установить, имеет ли он целочисленное решение или нет. ${ }^{1}$ Мы рассмотрим комплексный случай, из которого вытекает и случай произвольного поля. Серия F1. Разложение многочленов и подстановки.[^31] Задача F.1. Докажите, что $P(x+P(x))$ делится на $P(x)$. Задача F.2. Дан квадратный трехчлен $P(x) \neq$ const с челыми коэффиииентами. Докажите, что существует целое п такое, что все простые делители числа $P(n)$ меньше $n / 2019$. Замечание. Аналогичный факт даже для кубического многочлена нам неизвестно. Мы можем доказать это для двучленов $a x^{n}+b$. Задача F.3. Пусть $P\left(x_{1}, \ldots, x_{n}\right), Q\left(x_{1}, \ldots, x_{n}\right)$ - произвольные полиномы от $n$ переменных. Пусть $$ \widehat{P}\left(x_{1}, \ldots, x_{n}, u\right)=P\left(x_{1}+u Q\left(x_{1}, \ldots, x_{n}\right), \ldots, x_{n}+u Q\left(x_{1}, \ldots, x_{n}\right)\right) $$ Тогда существует $R\left(x_{1}, \ldots, x_{n}\right.$, , , такой, что $$ \widehat{P}\left(x_{1}, \ldots, x_{n}, u\right)=P\left(x_{1}, \ldots, x_{n}\right)+Q\left(x_{1}, \ldots, x_{n}\right) R\left(x_{1}, \ldots, x_{n}, u\right) $$ Задача F.4. Существует многочлен $H\left(x_{1}, \ldots, x_{n}\right)$ такой, что при всех целых $k$ таких, что $1 \leq k \leq 2019$ все многочлены $H\left(x_{1}, \ldots, x_{n}\right)-k$ имеют нетривиалвное разложение на множители. Задача F.5. В предъдущей задаче можно выбрать $H\left(x_{1}, \ldots, x_{n}\right)$ в виде $P_{k} Q_{k}+k$, где $Q_{i}$ алгебраически независимы (т.е. нет ненулевого многочлена $R$ такого, что $\left.R\left(Q_{1}, \ldots, Q_{2019}\right) \equiv 0\right)$. Задача F.6. Существуют такие семейства полиномов $H_{m}\left(x_{1}, \ldots, x_{m}\right), P_{m}\left(x_{1}, \ldots, x_{m}\right)$ для которых одновременно выполняются условия делимости $$ H_{k}\left(x_{1}, \ldots, x_{k}\right) \mid\left(P_{m}\left(x_{1}, \ldots, x_{m}\right)-k\right) $$ при всех $k \in\{1, \ldots, m\}$, причем $H_{m}$ и $P_{m}$ существенно зависят от $x_{m}$ и при этом $H_{s}$ при $sRoman Krutovsky, Igor Melnikov ## A. Introductory cycle. The classical Pell equation. The introductory cycle A deals with some classical facts regarding the Pell equation $x^{2}-D y^{2}=1$ for integers. The solution $( \pm 1,0)$ is called trivial, and all the others nontrivial. Problem A.1. Let $\left(x_{1}, y_{1}\right)$ be a solution of a Pell equation: $x_{1}^{2}-D y_{1}^{2}=1$. Let $\left(x_{1}+\right.$ $\left.\sqrt{D} y_{1}\right)^{n}=x_{n}+\sqrt{D} y_{n}$. Prove that $\left(x_{n}, y_{n}\right)$ also is a solution of the Pell equation (observe that $\left.x_{-n}=x_{n}, y_{-n}=-y_{n}\right)$. The solution $\left(x_{n}, y_{n}\right)$ is called a power of the solution $\left(x_{1}, y_{1}\right)$. Define the notion of the product of solutions. Problem A.2. Prove that all nontrivial solutions (if any) are powers of a single solution $\left(x_{1}, y_{1}\right)$ (up to the sign of $x$ ). Problem A.3. Prove that if $D$ is a square of an integer then the Pell equation has no nontrivial solutions. Up to the end of the cycle we suppose $D \neq m^{2} \forall m \in \mathbb{Z}, D \in \mathbb{Z}$. Problem A.4. Prove that there exist $M>0$ and a point with integer nonzero coordinates $(x, y)$ such that $\left|x^{2}-D y^{2}\right| \leq M$. Problem A.5. Using Minkowski lemma, prove that there exists $M>0$ such that there is an infinite number of such points. Problem A.6. Prove that there exists a positive integer $k2$ and for polynomials $P, Q$ there exist complex numbers $\psi_{1}, \psi_{2}, \psi_{3}$ such that $\psi_{i}^{n}=-1$ and $P^{n}+Q^{n}$ is a multiple of $\left(P-\psi_{1} Q\right)\left(P-\psi_{2} Q\right)\left(P-\psi_{3} Q\right)$. Problem B.5. Suppose $P^{n}+Q^{n}=R^{n}$. In the conditions of item B4, prove that there exist polynomials $R_{1}, R_{2}, R_{3}$ such that $P-\psi_{i} Q=R_{i}^{n}$. Problem B.6. Prove that for $n>2$ the Fermat equation for polynomials and for rational functions $x^{n}+y^{n}=z^{n}$ has no nontrivial solutions. ## C. Applications of Kollar theorem Here a nontrivial solution means a non-constant solution. Problem C.1. Solve the Pell equation if $D$ is the square of a polynomial. Problem C.2. Consider the following equation over polynomials from $\mathbb{Z}[x]$ : $$ P^{2}(x)-\left(R^{2}-1\right) Q^{2}(x)=1 $$ Here $R$ is an arbitrary non-constant polynomial, not necessarily a variable. Prove that the set of the solutions consists of the powers of a single nontrivial solution $(R, 1)$. Problem C.3. The theorem of J.Kollar***. The similar statement is valid for polynomials from $\mathbb{C}[x]$. Problem C.4. Solve the equation (1) for a constant $R$. Consider cases $R^{2}=1$ and $R^{2} \neq 1$ separately. Up to the end of cycle $\mathrm{C}$ we assume $R \neq$ const. Furthermore Kollar theorem may be used. Problem C.5. Prove that $Q$ is of the form $Q_{n}=\sum_{k=0}^{[n / 2]}\binom{n}{2 k+1}\left(T^{2}-1\right)^{k} T^{n-1-2 k}$ for some integer $n$. Find the similar formula for $P$. Problem C.6. Prove that $Q_{n} \equiv n$ modulo $R^{2}-1$, that is, $Q_{n}-n=\left(R^{2}-1\right) S$ for some polynomial $S$. Problem C.7. Solve the following system of equations for polynomials from $\mathbb{C}[x]$ : $$ \left\{\begin{array}{c} X^{2}-\left(R^{2}-1\right) Y^{2}=1 \\ Y-\left(R^{2}-1\right) Z=V \\ V \cdot U=1 \end{array}\right. $$ Problem C.8. Main problem. Prove that for an arbitrary equation $W$ for integers there exists a system of equations for polynomials from $\mathbb{C}[x]$, which has a solution iff $W$ has a solution. ## D. AFFINE VARIETIES Problem D.1. Define the following sets by equations or systems of equations: (1) an ellips; (2) a pair of lines in the plane; (3) a circle and an ellips; (4) a circle and a parabola both containing the origin and touching in this point, $Y$-axis being the axis of the parabola; (5) the equatorial circle on the unit sphere in the space; (6) the n-dimensional torus (the Cartesian product of $n$ circles). Definition 1. Let $S$ be a ring. An affine variety over $R$ is a set $X$ of the solutions of a finite system of algebraic equations in several variables, that is, the equations of the form $P\left(x_{1}, \ldots, x_{n}\right)=0$ for some polynomial $P$. Definition 2. Let $X$ be determined by some equations over $R\left[x_{1}, \ldots, x_{n}\right]$. The left parts of these equations generate an ideal in $R\left[x_{1}, \ldots, x_{n}\right]$, in the sequel denoted by $I(X)$. The ring of functions $R[X]$ on $X$ will mean the ring $R\left[x_{1}, \ldots, x_{n}\right] / I(X)$. Problem D.2. Ensure that in every point of $X$ the value of every element from $R[X]$ can be uniquely determined. Problem D.3. Show that the ring of functions on the semicubic parabola determined by the equation $y^{2}=x^{3}$ over $\mathbb{C}$ is isomorphic to $\mathbb{C}\left[t^{2}, t^{3}\right] \subset \mathbb{C}[t]$. Problem D.4. Let $X$ and $Y$ be affine varieties, and let $\varphi: R[Y] \rightarrow R[X]$ be an arbitrary map. Construct the corresponding map $f: X \rightarrow Y$ that induces $\varphi$. Remark 1. We are interested only in those maps between varieties that stem from maps of the corresponding rings. Such maps will be called algebraic. Investigation of these maps enables us to get information about varieties from the algebraic structure of rings of functions. Problem D.5. Consider the action of $\mathbb{Z} / 2 \mathbb{Z}$ on $\mathbb{C}^{2}$ such that the nontrivial element of the group takes $(x, y)$ to $(-x,-y)$. This action induces an automorphism of the ring $\mathbb{C}[x, y]$. Show that the subring invariant under this action is isomorphic to $\mathbb{C}[u, v, w] /\left(u v-w^{2}\right)$. Construct an isomorphism (of topological spaces) between the affine variety determined by this ring, and the quotient space of $\mathbb{C}^{2} \operatorname{modulo} \mathbb{Z} / 2$. Let $I$ be an ideal in $R\left[x_{1}, \ldots, x_{n}\right]$. The variety corresponding to $R\left[x_{1}, \ldots, x_{n}\right] / I$ will be denoted by $V(I)$. The radical of $I$ is the set rad(I) consisting of all elements from $R\left[x_{1}, \ldots, x_{n}\right]$ such that some their power belongs to $I$. Theorem 2 (Hilbert Nullstellensatz). Let $k$ be an algebraically closed field. Then for every ideal $I \subset k\left[x_{1}, \ldots, x_{n}\right]$ we have $$ I(V(I))=\operatorname{rad}(I) $$ Problem D.6. Using the above theorem, show that there is a bijection between the points of an affine variety $X$ over $\mathbb{C}$ and the maximal ideals of the ring $\mathbb{C}[X]$. Problem D.7. Let $X$ and $Y$ be two affine varieties over $\mathbb{C}$. Given a map $\varphi: \mathbb{C}[Y] \rightarrow \mathbb{C}[X]$, construct a map from the set of maximal ideals of $\mathbb{C}[X]$ to the set of maximal ideals of $\mathbb{C}[Y]$ not using Nullstellensatz. Ensure that this map is identical to the map constructed in Problem 4. Definition 3. A map $\iota: X \rightarrow Y$ between affine varieties will be called a closed embedding if $R[X]=R[Y] / I$ and the map of the rings of functions, corresponding to $\iota$ is the factorization map $R[Y] \rightarrow R[Y] / I$. Remark 4. Clearly every affine variety over $\mathbb{C}$ is canonically embedded into some $\mathbb{C}^{n}$. Problem D.8. Define a twofold cable as a closed embedding of a circle into a torus. Problem D.9. Consider the natural embedding $(\mathbb{C} \backslash 0)^{n}$ в $\mathbb{C}^{n}$. Show that this is an algebraic map. Prove that it is not a closed embedding. Problem D.10. Show that the standard embedding $\varphi: \mathbb{C}\left[t^{2}, t^{3}\right] \hookrightarrow \mathbb{C}[t]$ does not determine a closed embedding of a line into a semicubic parabola. ## E. MAPS OF AFFINE VARIETIES Don't forget to apply with questions and for hints! Definition 5. A set $R$ with two associative commutative binary operations $\cdot$ and + (multiplication and addition) is called a (commutative associative) ring if the following conditions are fulfilled: (1) $\exists 0 \in R: 0+a=a+0=a \forall a \in R$; (2) $\forall a \in R \exists b \in R: a+b=b+a=0$; (3) $\exists 1 \in R: 1 \cdot a=a \cdot 1=a \quad \forall a \in R$; (4) $a \cdot(b+c)=a \cdot b+a \cdot c$; (5) $(a+b) \cdot c=a \cdot c+b \cdot c$. Remark 6. We will be interested mainly (except the rings of integers, rational, real and complex numbers) in rings of polynomials as well as rings of functions on affine varieties. Remark 7. Polynomials $P_{1}\left(x_{1}, \ldots, x_{n}\right), \ldots P_{k}\left(x_{1}, \ldots, x_{n}\right)$ that determine an affine variety $X$, also generate the set of functions $I=P_{1} \cdot R\left[x_{1}, \ldots, x_{n}\right]+\ldots+P_{k} \cdot R\left[x_{1}, \ldots, x_{n}\right] \subset$ $R\left[x_{1}, \ldots, x_{n}\right]$. All these functions vanish at the points of the variety. Such a subset is an ideal in the ring of functions on $X$. Definition 8. An ideal of a ring $R$ is a set $I$ such that (1) $I \cdot R \subset I$; (2) $I+I \subset I$. An ideal is called principal if it is generated by a single element. Problem E.1. Let $I$ be an ideal of the polynomial ring $\mathbb{C}\left[x_{1}, \ldots, x_{n}\right]$. Let $X_{I}$ be the variety corresponding to this ideal. Prove the following properties: (1) $I \subseteq J \Rightarrow X_{J} \subseteq X_{I}$ (2) $X_{I} \cup X_{J}=X_{I \cdot J}=X_{I \cap J}$; (3) if $I+J=R$ then $I \cdot J=I \cap J$; (4) the ideals $I_{1}=\langle x\rangle$ and $I_{2}=\left\langle x^{2}\right\rangle$ of the ring $\mathcal{R}[x]$ determine the same set of zeroes. Problem E.2. (1) Prove that in the ring $\mathcal{R}[x]$ of polynomials in a single variable, all ideals are principal. Such rings are called rings of principal ideals (RPI). (2) Give an example of a ring which is not a RPI. Definition 9. Given an ideal $I \subseteq R$. The quotient ring $R / I$ is the set of equivalence classes of elements from $R$ modulo $I: a \sim b \Leftrightarrow a-b \in I$. Multiplication and addition in the quotient ring are defined by the following formulas: (1) $(a+I)+(b+I)=(a+b)+I$; (2) $(a+I) \cdot(b+I)=(a \cdot b+I)$. Problem E.3. Check correctness of the above definition. Prove that the quotient ring is a ring. Remark 10. In Problem 2 from the cycle "Varieties and equations" we already did implicitly use a quotient ring. Namely we have shown that for a variety $X_{I}$ over $\mathbb{C}$ determined by the ideal $I$ ( $X_{I}$ is the set of common zeroes for all elements from $I$ ) the values of the polynomial functions over $X_{I}$ are determined by the quotient ring $\mathbb{C}\left[x_{1}, \ldots, x_{n}\right] / I$. Definition 11. Let a variety $X$ over $R$ be determined by an ideal $I(X)$ (here we assume that a set of equations determines an ideal). Then the ring of (polynomial) functions $R[X]$ over $X$ is the quotient ring $R\left[x_{1}, \ldots, x_{n}\right] / I(X)$. Remark 12. In Problem 5 of the preceding cycle it was shown that any ideal $I \subset$ $\mathbb{C}\left[x_{1}, \ldots, x_{n}\right]$ is generated by some finite set of elements $\Gamma=\left\{g_{1}, \ldots, g_{k}\right\}$. Consider the ideal $G$ generated by the leading monomials of all elements from I. In Problem 5 we have proved that the leading monomials of the elements from $\Gamma$ also generate G. Such a set of generators for $I$ is called a Groebner basis. Definition 13. A map $\iota: X \rightarrow Y$ between affine varieties is called a closed embedding if $R[X]=R[Y] / I$ and the map corresponding to $\iota$ is the map of factorization $R[Y] \rightarrow$ $R[Y] / I$. Remark 14. Clearly any affine variety over $\mathbb{C}$ has a canonical closed embedding to some $\mathbb{C}^{n}$. Problem E.4. Here we construct a nontrivial algebraic embedding of a line $\mathcal{R}$ into $\mathcal{R}^{3}$. To begin with, consider the map $$ \varphi: \mathcal{R} \rightarrow \mathcal{R}^{2}: t \mapsto\left(t^{3}-3 t, t^{4}-4 t^{2}\right) $$ Ensure that its image in $\mathcal{R}^{2}$ is the projection to the plane from a trefoil with a single point removed (to infinity). Find a polynomial $h(t)$ such that the map $$ t \mapsto(\varphi(t), h(t)) $$ transforms a line to a trefoil in $\mathcal{R}^{3}$ with a single point removed (also to infinity). Remark 15. Obviously not every system of polynomial equations has a solution. For instance, the system $\{x-y=0, x-y+1=0\}$ has no solution. Observe that the ideal generated by these two polynomials coincides with the whole ring $R[x, y]$ because $1 \in I$. So if an ideal I is generated by elements $g_{1}, \ldots, g_{k}$ and for some $h_{1}, \ldots, h_{k} \in R\left[x_{1}, \ldots, x_{n}\right]$ we have $$ g_{1} h_{1}+\ldots g_{k} h_{k}=1 $$ then $I$ coincides with $R\left[x_{1} \ldots, x_{n}\right]$. In the next problem we prove that if $1 \notin I$ then there exists a solution common for all elements of $I$. Problem E. 5 (Weak Hilbert Nullstellensatz). Every proper ideal $I \subsetneq \mathbb{C}\left[x_{1}, \ldots, x_{n}\right]$ determines a nonempty set of solutions. The proof consists of several steps. Consider the map which replaces the first coordinate by some number $a \in \mathbb{C}$ : $$ e v_{a}: \mathbb{C}\left[x_{1}, \ldots, x_{n}\right] \rightarrow \mathbb{C}\left[x_{2}, \ldots, x_{n}\right]: f\left(x_{1}, \ldots, x_{n}\right) \mapsto f\left(a, x_{2}, \ldots, x_{n}\right) $$ (1) Suppose $I \cap \mathbb{C}\left[x_{1}\right]=\left\langle p\left(x_{1}\right)\right\rangle$, deg $p>0$. Then there exists $a \in \mathbb{C}$ such that ev ${ }_{a}(I) \subsetneq$ $\mathbb{C}\left[x_{2}, \ldots, x_{n}\right]$ (2) Suppose $I \cap \mathbb{C}\left[x_{1}\right]=\emptyset$. Prove that for some $a \in \mathbb{C}$ we have ev $(I) \subsetneq \mathbb{C}\left[x_{2}\right]$. (3) Suppose $I \cap \mathbb{C}\left[x_{1}\right]=\emptyset$. Prove that for some $a \in \mathbb{C}$ we have ev $v_{a}(I) \subsetneq \mathbb{C}\left[x_{2}, \ldots, x_{n}\right]$ (4) Prove the theorem. ## F. THE COMPLEX CASE. We require the following theorem that was the subject of the project https://www.turgor.ru/lktg/2007/2/index.php . Matiyasevich theorem. There is no algorithm to check existence of an integer root for a polynomial $H\left(x_{1}, \ldots, x_{m}\right)$ in several variables, using the set of its coefficients. ${ }^{1}$ We will consider the complex case which implies the case of an arbitrary field. Cycle F1. Decomposition of polynomials, and substitutions. Problem F.1. Prove that $P(x+P(x))$ is a multiple of $P(x)$. Problem F.2. Given a square trinomial $P(x) \neq$ const with integer coefficients. Prove that there exists an integer $n$ such that all prime divisors of $P(n)$ are less than $n / 2019$. Remark. The similar fact is unknown even for cubic polynomials. We can prove it for binomials ax ${ }^{n}+b$. Problem F.3. Let $P\left(x_{1}, \ldots, x_{n}\right), Q\left(x_{1}, \ldots, x_{n}\right)$ be arbitrary polynomials in $n$ variables. Suppose $$ \widehat{P}\left(x_{1}, \ldots, x_{n}, u\right)=P\left(x_{1}+u Q\left(x_{1}, \ldots, x_{n}\right), \ldots, x_{n}+u Q\left(x_{1}, \ldots, x_{n}\right)\right) $$ Then there exists $R\left(x_{1}, \ldots, x_{n}, u\right)$ such that $$ \widehat{P}\left(x_{1}, \ldots, x_{n}, u\right)=P\left(x_{1}, \ldots, x_{n}\right)+Q\left(x_{1}, \ldots, x_{n}\right) R\left(x_{1}, \ldots, x_{n}, u\right) $$ Problem F.4. There exists a polynomial $H\left(x_{1}, \ldots, x_{n}\right)$ such that for all integer $k$ with $1 \leq k \leq 2019$ all polynomials $H\left(x_{1}, \ldots, x_{n}\right)-k$ have a nontrivial factor decomposition. Problem F.5. In the above problem we can choose $H\left(x_{1}, \ldots, x_{n}\right)$ in the form $P_{k} Q_{k}+$ $k$ where $Q_{i}$ are algebraically independent (there is no nonzero polynomial $R$ such that $\left.R\left(Q_{1}, \ldots, Q_{2019}\right) \equiv 0\right)$. ## Problem F.6. There exist families of polynomials $H_{m}\left(x_{1}, \ldots, x_{m}\right), P_{m}\left(x_{1}, \ldots, x_{m}\right)$ which both fulfil the conditions of divisibility $$ H_{k}\left(x_{1}, \ldots, x_{k}\right) \mid\left(P_{m}\left(x_{1}, \ldots, x_{m}\right)-k\right) $$ for all $k \in\{1, \ldots, m\}$, and furthermore $H_{m}$ and $P_{m}$ essentially depend on $x_{m}$ but $H_{s}$ for $s1 июля 2019 г. #### Abstract Аннотация Мы показываем, что для двух аффинных многообразий над произвольным полем характеристики ноль не существует в общем виде алгоритма проверки наличия вложения одного алгебраического многообразия в другое. Более того, мы устанавливаем это для аффинных многообразий, чьи координатные кольца заданы образующими и определяющими соотношениями. Более того, одно из этих многообразий можно взять аффинным пространством, а в случае поля вещественных чисел - аффинной прямой. ## 1 Введение Задача классификации алгебраических многообразий с точностью до изоморфизма представляется одной из центральных задач алгебраической геометрии. Чрезвычайно интересной и фундаментальной является[^33]задача об алгоритмической разрешимости проверки наличия изоморфизма. Близкой задачей является задача о вложимости многообразий. В общем виде эта задача формулируется так: Пусть $\mathcal{A}$ и $\mathcal{B}$ - два алгебраических многообразия. Определить, существует ли вложение $\mathcal{A}$ в $\mathcal{B}$. В случае, когда многообразия заданы каким-либо конструктивным способом (например, системами уравнений и образующими), эта задача естественным образом приводит к следующей - придумать алгоритм, позволяющий по заданным многообразиям установить существование вложения (даже без его явного построения) или же отсутствие такового. Иными словами, к вопросу об алгоритмической разрешимости проблемы вложимости. Вопросы об алгоритмической разрешимости различных классов задач в неявной форме ставились еще в XIX веке (например, Десятая проблема Гильберта о диофантовых уравнениях). Формализация понятия алгоритма, проведенная в работах Тьюринга, Черча, Геделя и иных авторов в 30-е годы XX века, дала толчок к активному исследованию данных вопросов, но к сожалению это не отразилось в должной мере на алгебраической геометрии. Одним из наиболее примечательных результатов, достигнутых в этой области, стало полученное Ю. В. Матиясевичем отрицательное решение Десятой проблемы Гильберта ([1). Исследования в этой области активно продолжаются и в наши дни (см. например [2], [4], [5], 6], 7], 8], 79]). В данной работе мы рассмотрим вопрос об алгоритмической разрешимости задачи вложимости многообразий над полями $\mathbb{R}$ и $\mathbb{C}$. Иными словами, о существовании алгоритма, позволяющего определить по уравнениям, задающим $\mathcal{A}$ и $\mathcal{B}$, существует ли искомое вложение. Полученные результаты естественным образом обобщаются на многие другие поля нулевой характеристики. Мы показываем, что для двух аффинных многообразий над произвольным полем характеристики ноль не существует в общем виде алгоритма проверки наличия вложения. Мы это проверяем для аффинных многообразий, чьи координатные кольца заданы образующими и определяющими соотношениями. В частном случае, когда $\mathcal{A}=A$ - аффинная прямая, соответствующее вложение задается полиномами от одной переменной. Эти полиномы должны удовлетворять уравнениям, задающим $\mathcal{B}$. При этом отображения не должны быть константными (иначе все $\mathcal{A}$ отображается в одну точку $\mathcal{B}$, что не является вложением). Соответственно, вопрос о существовании вложения эквивалентен вопросу о существовании у системы уравнений, задающих $\mathcal{B}$, неконстантных решений в кольце многочленов $\mathbb{F}[t]$ над основным полем $\mathbb{F}$. В более общем случае, когда $\mathcal{A}=A^{m}-m$-мерная аффинная плоскость, отображение задается полиномами от $m$ переменных. Для того, чтобы оно было вложением, необходимо чтобы эти полиномы были алгебраически независимыми (т.е. соответствующее расширение имело степень трансцендентности $m$ ). В нашей работе установлено, что проблема вложимости алгоритмически неразрешима для случая, когда основным полем является поле вещественных чисел $\mathbb{R}$ или поле комплексных чисел $\mathbb{C}$. Авторы выражают благодарность J. Kollar за предоставленные им новые результаты об уравнениях Пелля и ценные замечания. Данная работа была проведена с помощью Российского Научного Фонда Грант N 17-11-01377. ## 2 Предварительные сведения ## 2.1 Уравнения Пелля для многочленов Уравнением Пелля над кольцом $\mathcal{R}$ называется уравнение вида $$ X^{2}-\lambda Y^{2}=1 $$ где $\lambda \in \mathcal{R}$ - параметр, а $X$ и $Y$ - неизвестные. Решения также ищутся в кольце $\mathcal{R}$. Описание множества решений уравнения Пелля над различными кольцами является интересной и сложной задачей. В данном разделе мы приведем некоторые сведения об уравнениях Пелля, поскольку они будут играть центральную роль в наших дальнейших конструкциях. Для наших целей будут интересны уравнения Пелля над кольцом многочленов от одной или нескольких переменных над полем $\mathbb{F}$, а также один из частных случаев уравнения Пелля - уравнение $$ X^{2}-\left(T^{2}-1\right) Y^{2}=1 $$ где $T \in \mathcal{R}$ - некоторый заданный параметр. Решения уравнения $X^{2}-\lambda Y^{2}=1$ для произвольного $\lambda$ образуют абелеву группу относительно операции $$ \circ:(X, Y) \circ\left(X^{\prime}, Y^{\prime}\right)=\left(X X^{\prime}+\lambda Y Y^{\prime}, X Y^{\prime}+X^{\prime} Y\right) $$ В любом решении можно поменять знак как при $X$, так и при $Y$. Смена знака при $Y$ приводит к получению обратного элемента (относительно операции о), а одновременная смена знака при $X$ и $Y$ - к умножению на решение $(-1,0)$, которое само является элементом порядка 2 в этой группе. Единичным элементом в этой группе является $(1,0)$. Для случая $\lambda=T^{2}-1$, где $T$ - многочлен над $\mathbb{F}$, существует также очевидное решение $(T, 1)$. Применяя к нему операцию о, получаем целую серию решений. Если $T$ - многочлен, отличный от константы, то все эти решения будут различными. Значительно менее тривиальным является вопрос о точной структуре группы решений. Оказывается, что в интересных нам случаях вся она (с точностью до упомянутой выше замены знаков) порождается одним единственным элементом («примитивным» решением). Похожие утверждения для различных случаев доказаны в [2], 3]. Окончательное решение для важного нам случая $\mathcal{R}=\mathbb{F}[t]$ и $\lambda=T^{2}-1$, где $T \in \mathbb{F}[t] \backslash \mathbb{F}$ было получено J. Kollar и сообщено авторам в частной переписке (результат готовится к публикации). Имеет место весьма нетривиальная Теорема 1 (J. Kollar) Если $T$ - многочлен над F, не являющийся константой, то группа (относительно операции о решений (⿴) (как и в классическом целочисленном случае) порождена элементом $(T, 1)$. Иными словами, множество решений уравнения (2) представимо в виде $\{ \pm 1,0\} \cup\left\{\left( \pm X_{N}, \pm Y_{N}\right)\right\}$, где $X_{N}=\sum_{k=0}^{[N / 2]}\binom{N}{2 k}\left(T^{2}-1\right)^{k} T^{N-2 k}, Y_{N}=$ $\sum_{k=0}^{[N / 2]}\binom{N}{2 k+1}\left(T^{2}-1\right)^{k} T^{N-1-2 k}$ для некоторого целого положительного $N$. ## 2.2 Многочлены и делимость Нам потребуется еще несколько вспомогательных утверждений, связанных с делимостью многочленов от многих переменных. Основным результатом данного раздела будет построение полинома $\hat{P}$ и семейства полиномов $W_{k}$ от нескольких переменных, удовлетворяющих условиям делимости $W_{k} \mid(\hat{P}-3 k)$. Эти полиномы будут существенно использованы в построениях из раздела 4 (и не требуются для случая вещественных чисел). Приводимые здесь рассуждения в целом основаны на идеях, обсуждавшихся в 10, задача 2. Проект на конференции школьников был основан на задаче, поставленной С. В. Конягиным: Дан многочлен $P(x)$ с целочисленными коэффициентами. верно ли что найдется такое $n$, что все простые делители $P(n)$ меньше $n$. С. В. Конягин решил задачу для многочлена вида $a x^{n}+b$. Идея, изложенная ниже позволяет решить ее для квадратного трехчлена. Общее решение нам не известно. Начнем с наблюдения: Лемма 1 Пусть $P\left(x_{1}, \ldots, x_{n}\right), Q\left(x_{1}, \ldots, x_{n}\right)$ - произвольные полиномы oт $n$ переменных над полем $\mathbb{F}$. Пусть $$ P^{[1]}\left(x_{1}, \ldots, x_{n}, u\right)=P\left(x_{1}+u Q\left(x_{1}, \ldots, x_{n}\right), \ldots, x_{n}+u Q\left(x_{1}, \ldots, x_{n}\right)\right) $$ Тогда существует $R\left(x_{1}, \ldots, x_{n}, u\right)$, такой, что $$ P^{[1]}\left(x_{1}, \ldots, x_{n}, u\right)=P\left(x_{1}, \ldots, x_{n}\right)+Q\left(x_{1}, \ldots, x_{n}\right) R\left(x_{1}, \ldots, x_{n}, u\right) $$ Из данной леммы вытекает Теорема 2 Пусть $C_{1}, \ldots, C_{m}$ - некоторые константы из произвольного поля $\mathbb{F}$. Тогда существуют такие семейства полиномов $H_{m}\left(x_{1}, \ldots, x_{m}\right), \quad P_{m}\left(x_{1}, \ldots, x_{m}\right)$ с коэффициентами из $\mathbb{F}$, для которых одновременно выполняются условия делимости $H_{k}\left(x_{1}, \ldots, x_{k}\right)$ | $\left(P_{m}\left(x_{1}, \ldots, x_{m}\right)-C_{k}\right)$ при всех $k \in\{1, \ldots, m\}$, причем $H_{m}$ и $P_{m}$ существенно зависят от $x_{m}$ и при этом $H_{s} n p u s0$, то $V=Y \bmod (T-1)=N$ для некоторого целого $N$ и $Z=(Y-N) /(T-1)$. $\triangleright П$ Пи $\operatorname{deg} T>0$ уравнение $Y-(T-1) Z=V$ в совокупности с ранее показанным условием $\operatorname{deg} V=0$ означает, что остаток от деления многочлена $Y$ на многочлен $T$ есть в точности константа $V \in \mathbb{F}$. Легко видеть, что в выражении для $Y$ все слагаемые делятся на $T-1$ без остатка, за исключением слагаемого, отвечающего $k=0$. Оно равно $\binom{N}{1} T^{N-1}=N \bmod (T-1)$. Таким образом, значение $V$ в любом решении исходной системы должно быть целым числом $N$. Прямой подсчет показывает, что $Z=(Y-N) /(T-1)$. Таким образом, слои построенного множества $\mathcal{B}_{(d)}$ обладают понятной структурой. Возможны три случая. 1. При $\operatorname{deg} T>0$ каждому набору целых чисел $N_{i}$ соответствуют решения $Y_{i}$ и $X_{i}$, являющиеся многочленами, и определеные с точностью до знака, а также константы $V_{i}=N_{i}$ и $U_{i}=1 / V_{i}$ и $Z_{i}=\left(Y_{i}-V_{i}\right) /(T-1)$. 2. При $\operatorname{deg} T=0$ и $T \neq \pm 1$ имеем константные решения для $Y_{i}$, выбираемые из заданной последовательности. Значения $X_{i}, Z_{i}, V_{i}$ и $U_{i}$ также являются константами и определяются выбранными значениями $Y_{i}$. 3. При $T= \pm 1$ получаем $X_{i}= \pm 1$, для произвольно выбранных констант $V_{i}$ и полиномов $Z_{i}$ определяем $U_{i}=1 / V_{i}$ и $Y_{i}=(T-1) Z_{i}+V_{i}$. Все вышеприведенные рассуждения справедливы для произвольного основного поля $\mathbb{F}$. Теперь используем специфику случая $\mathbb{F}=\mathbb{R}$. Перейдем к пространству размерности $5 d+2$, обозначим новую координату через $S$, и дополним основную систему (12) уравнением $$ T=S^{2}+2 $$ Далее заметим, что из $T=S^{2}+2$ следует, что случай $T= \pm 1$ невозможен. Таким образом, все общие решения систем (12) и (14) либо являются константами (случай $\operatorname{deg} T=0, T \neq \pm 1$ ), либо отвечают некоторому целочисленному набору параметров $\left(N_{1}, \ldots, N_{d}\right)$. Назовем первые решения «плохими», а вторые - «хорошими». Рассмотрим семейство полиномов Матиясевича $Q\left(\sigma_{1}, \ldots, \sigma_{\tau}, x_{1}, \ldots, x_{s}\right)$. Пусть $d \leq s$. Тогда, добавив новое уравнение $$ Q_{\sigma}\left(V_{1}, \ldots, V_{s}\right)=0 $$ к системам (12) и (14), мы получаем систему, задающую новое многообразие. Обозначим его $\mathcal{B}_{(d), \sigma}^{\prime}$. Если $Q_{\sigma}=0$ не имеет целочисленных решений, то у исходной системы нет «хорших» решений. Тогда многообразие $\mathcal{B}_{(d), \sigma}^{\prime}$ нульмерно, и вложений из $A$ в $\mathcal{B}_{(d)}^{\prime}$ не существует. В противном случае для каждого решения $N_{1}, \ldots N_{s}$ строятся явно неконстанные полиномы $S(t), T(t), Y_{i}(t), X_{i}(t)$ и $Z_{i}(t)$, являющиеся решениями. Например, можно положить $S(t)=t, T(t)=t^{2}+2$, и определить $X_{i}, Y_{i}$ и $Z_{i}$ по формулам для случая 22 из леммы 2. Также однозначно определяются константы $V_{i}=N_{i}$ и $U_{i}=1 / N_{i}$. В совокупности эти функции задают вложение прямой в многообразие $\mathcal{B}_{(d), \sigma}^{\prime}$. Поскольку задача о существовании целочисленных решений для $Q_{\sigma}$ алгоритмически неразрешима, то алгоритмически неразрешима и задача о вложимости $A$ в $\mathcal{B}_{(d), \sigma}^{\prime}$ (в частности, в $\left.\mathcal{B}_{(s), \sigma}^{\prime}\right)$. Входом при этом являются уравнения, задающие $\mathcal{B}_{(d), \sigma}^{\prime}$. Таким образом, нами доказана Теорема 3 Задача о вложимости аффинной прямой над $\mathbb{R}$ в произволъное алгебраические многобразие $\mathcal{B}$ (заданное образующими и соотношениями) алгоритмически неразрешима. Отсюда сразу следует: Теорема 4 Задача о вложимости произвольного алгебраического многобразия $\mathcal{A}$ над $\mathbb{R}$ в произвольное алгебраические многобразие $\mathcal{B}$ алгоритмически неразрешима. ## 4 СЛучай комплексных чисел В комплексном случае ситуация сложнее, чем в вещественном. В самом деле, рассуждение для $\mathbb{R}$ существенно опирается на тот факт, что уравнение $S^{2}+a$ не имеет решений при $a>0$ (отюда следует, что $T=S^{2}+2 \neq \pm 1$ ). Однако в поле комплексных чисел это неверно. Таким образом, мы не можем исключить случай $T= \pm 1$ и $X_{i}= \pm 1$. В этом случае $Y_{i}$ может быть задано, вообще говоря, произвольным образом. Поэтому в случае основного поля $\mathbb{C}$ мы немного усложним рассуждение. А именно, рассмотрим вопрос о вложимости аффинного пространства $A^{m}$ в заданное многообразие $\mathcal{B}$. Нашей целью будет, как и раньше, сконструировать такой класс многообразий, для представителей которого (при некотором подходящем $m$ ) невозможно установить существование искомого вложения по определяющим соотношениям. Результаты данного параграфа сформулированы для случая основного поля $\mathbb{C}$, но проходят и для произвольного поля нулевой характеристики, поскольку все коэффициенты у наших полиномов целые. Зададим алгебраическое многообразие $\mathcal{B}_{(d, e)}$ при помощи системы образующих и соотношений. $$ \left\{\begin{array}{c} X_{i j}^{2}-\left(T_{j}^{2}-1\right) Y_{i j}^{2}=1 \\ Y_{i j}-\left(T_{j}-1\right) Z_{i j}=V_{i j} \\ V_{i j} U_{i j}=1 \\ T_{j+1}=\prod_{k=1}^{j}\left(\left(T_{k}^{2}-1\right) W_{k}\right) W_{j+1} \\ T_{1}=\hat{P}\left(W_{1}, \ldots, W_{n}\right) \end{array}\right. $$ где $1 \leq i \leq d, 1 \leq j \leq e$.При этом полином $\hat{P}$ выберем таким образом, чтобы $W_{j}$ был делителем $\hat{P}\left(W_{1}, \ldots, W_{n}\right)-3 j$. Полиномы с нужными свойствами были построены в разделе 2.2 (см. лемму 2). Иными словами, мы «клонируем» основную систему (12) из предыдущего параграфа в большом числе экземпляров, и дополняем ее «свя- зующими» соотношениями между параметрами $T_{j}$. Исследуем решения этой системы в $\mathbb{C}\left[t_{1}, \ldots, t_{m}\right]$. Совершенно очевидно, что соотношения для $X_{i j}, Y_{i j}, Z_{i j}, U_{i j}, V_{i j}$ при фиксированном $T_{j}$ аналогичны ранее рассмотренным. Иными словами, при фиксированном наборе $T_{j}$ множество решений есть прямая сумма слоев $\mathcal{B}_{(d)}$, ранее уже изученных. Однако при различных $T_{j}$ соответствующие слои ведут себя по-разному. При этом наличие связей между $T_{j}$ говорит о том, что возможно, не все варианты допустимы. Как показывают результаты предыдущего раздела, для каждого $j$ имеют место следующие важные случаи: 1. $T_{j}= \pm 1$; 2. $\operatorname{deg} T_{j}=0, T_{j} \neq \pm 1$; 3. $\operatorname{deg} T_{j}>0$. Далее в тексте мы будем обозначать через $C_{i}$ константы из основного поля $\mathbb{C}$ (в частности, полиномы степени 0). Для дальнейших рассуждений удобно сформулировать и доказать еще одно вспомогательное утверждение: Лемма 4 Пусть для некоторого $N$ выполнено $T_{N}=C_{N} \neq 0$. Тогда все $W_{k}$ при $k \leq N$ и все $T_{k}$ при при $k \leq N-1$ являются константами. $\triangleright$ Заметим, что произведение нескольких многочленов из $\mathbb{F}[t]$ является ненулевой константой в том и только в том случае, когда все сомножители являются также ненулевыми константами. Теперь докажем утверждение по индукции. База индукции. При $N=0$ имеем $T_{1}=W_{1}=C_{1} \neq 0$ и утверждение очевидно. Индуктивный переход. Пусть для $N-1$ утверждение верно, и пусть $T_{N}=C_{N} \neq 0$. Поскольку $T_{N}=\prod_{k=1}^{N-1}\left(\left(T_{k}^{2}-1\right) W_{k}\right) W_{N}=C_{N} \neq 0$, сразу получаем что $W_{N}$ - ненулевая константа. Также все $\left(T_{k}^{2}-1\right)$ являются ненулевыми константами. Следовательно, и все $T_{k}$ - тоже константы. В частности, $T_{N-1}$ - константа. Если при этом $T_{N-1} \neq 0$, то утверждение леммы для $N$ следует из утверждения леммы для $N-1$. Таким образом, осталось показать, что $T_{N-1} \neq 0$. Но если $T_{N-1}=0$, то поскольку $T_{N}=T_{N-1}\left(T_{N-1}^{2}-1\right) W_{N}$, то и $T_{N}=0$, что невозможно. Таким образом, лемма 4 доказана. Теперь необходимо рассмотреть несколько случаев. Первый случай: среди $T_{j}$ нет констант. В этом случае, все компоненты слоя соответствуют ранее изученным множествам $\mathcal{B}_{(d)}$ (для первого случая из предыдущего параграфа). Они параметризуются целочисленными наборами $N_{i j}$ независимо друг от друга. Слой является прямой суммой соответствующих компонент. Второй случай: среди $T_{j}$ есть ненулевые константы, ни одна из которых не равна $\pm 1$. Пусть $\hat{j}$ - максимальный из индексов этих констант. Тогда в силу леммы 4 при $j<\hat{j}$ мы имеем $T_{j}=C_{j}$. Иными словами, все предыдущие $T_{j}$ - также константы. Одновременно с этим и $W_{j}$ при $j \leq \hat{j}$ также будут константами. В частности, $T_{1}=\hat{P}\left(W_{1}, \ldots, W_{n}\right)$ также будет ненулевой константой $C_{1}$. Поскольку при этом все $W_{k}$ являются делителями $\hat{P}\left(W_{1}, \ldots, W_{n}\right)-$ $3 k=C_{1}-3 k$, то все они, за исключением, возможно, одного значения, также являются ненулевыми константами. Этим единственным значением, очевидно, должно быть $\hat{j}+1$ (разумеется, при $\hat{j}\hat{j}$ получаем $T_{j}=\prod_{k=1}^{j-1}\left(\left(T_{k}^{2}-1\right) W_{k}\right) W_{j}=0$. В силу леммы 4 при $j<\hat{j}$ мы имеем $T_{j}=C_{j}$. При этом $C_{j} \neq \pm 1$ (иначе $C_{j+1}=0$ ). Аналогично и все $W_{j}$, за исключением, возможно, одного значения, также являются ненулевыми константами. Рассмотрим семейство полиномов Матиясевича $Q\left(\sigma, x_{1}, \ldots, x_{s}\right)$. Проблема существования решения диофантова уравнения $Q\left(\sigma, V_{1 j}, \ldots, V_{s j}\right)=$ 0 алгоритмически неразрешима. Пусть $d \leq s$. Тогда, добавив новые уравнения $Q\left(\sigma, V_{i 1}, \ldots, V_{i s}\right)=0$ к системе (16), мы получаем систему, задаю- щую новое многообразие. Обозначим его $\mathcal{B}_{(d, e), \sigma}^{\prime}$. Лемма 5 Если $Q_{\sigma}=0$ не имеет целочисленных решений, то при $m \geq$ $d+2$ не существует вложения $\mathcal{A}=A^{m}$ в $\mathcal{B}_{(d, e), \sigma}^{\prime}$. $\triangleright$ Пусть искомое вложение существует. Тогда оно задается некоторой системой полиномов от $m$ переменных. Соответствующие полиномы будут решениями системы (16). Значения координат $U_{i j}, V_{i j}$ будут константами. Поскольку $Q_{\sigma}=0$ не имеет целочисленных решений, то у исходной системы нет решений для которых $\operatorname{deg} T_{1}>0$. Тогда возможные решения соответствуют второму либо третьему из расмотренных выше случаев. Далее рассмотрим их отдельно. Если имеет место второй случай, то все $W_{i}$, за исключением, быть может, одного, являются константами. Значения $W_{i}$ однозначно парамеризуют все значения $T_{j}$. Значения $T_{j}$ в этом случае однозначно определяют все остальные кординаты. Таким образом, образ соответствующего отображения имеет размерность 1. Следовательно, при $m \geq 2$ оно не может быть вложением. В третьем случае также все $W_{i}$, за исключением, быть может, одного, являются константами. Значения $W_{i}$ снова однозначно парамеризуют все значения $T_{j}$. Однако при $T_{\hat{j}}= \pm 1$ для компоненты с номером $\hat{j}$ мы находимся в условиях третьего случая из раздела 3, В этом случае $Z_{i \hat{j}}$ могут быть выбраны любыми, остальные же переменные однозначно ими определяются. Еще одна свободная переменная $W_{\hat{j}+1}$ может возникнуть по тем же соображениям, что и в предыдущем случае. Таким образом, образ соответствующего отображения имеет размерность $\leq d+1$. Следовательно, при $m \geq d+2$ оно не может быть вложением. Лемма 6 Если $Q_{\sigma}$ имеет целочисленние решения, то существует вложение аффинного пространства $\mathcal{A}=A^{s}$ в многообразие $\mathcal{B}_{(d, s), \sigma}^{\prime}$. $\triangleright$ Поскольку $Q_{\sigma}$ имеет целочисленные решения, то для каждого такого решения $N_{1}, \ldots, N_{s}$ по любому заданному набору $T_{j}, 1 \leq j \leq s$, могут быть построены явно функции $U_{i j}\left(T_{j}\right), V_{i j}\left(T_{j}\right), Y_{i j}\left(T_{j}\right), X_{i j}\left(T_{j}\right)$ и $Z_{i j}\left(T_{j}\right)$, являющиеся решениями. Для окончания построения нужно осталось показать, что можно подобрать подходящие $T_{j}$ (и $W_{j}$ ), чтобы выполнялись связывающие их уравнения. Заметим, что $T_{j}$ однозначно определяются по $W_{j}$. Сами же $W_{j}$ нами уже построены при помощи полиномов $H_{j}$ из теоремы 2. При этом каждый из $H_{j}$ зависит от переменных $x_{1}, \ldots, x_{j}$. Таким образом, задавая произвольную параметризацию для $x_{j}$ мы получим подходящий полином $H_{j}$ (а значит и $W_{j}$ ). Таким образом, $W_{j}$ алгебраически независимы. Следовательно, в совокупности с остальными координатными функциями, они задают некоторое вложение аффинного пространства $\mathcal{A}=A^{s}$ в многообразие $\mathcal{B}_{(d, s), \sigma}^{\prime}$. Заметим, что при доказательстве отсутствия вложения в лемме 5 мы наложили некоторое дополнительное условие на $d$, но не на $e$. Напротив, для существования вложения (при наличии решений у $Q_{\sigma}$ ) в лемме 6 ограничения накладываются на $e$, но не на $d$. Это дает нам возможность согласовать параметры. Для этого достаточно положить $m=e=s$ и $d=$ $s-2$. В этом случае $A^{s}$ вложимо в $\mathcal{B}_{(s-2, s), \sigma}^{\prime}$, если $Q_{\sigma}$ имеет целочисленные решения, и не вложимо в противном случае. Поскольку задача о существовании целочисленных решений для $Q_{\sigma}$ алгоритмически неразрешима, то алгоритмически неразрешима и задача о вложимости $A^{s}$ в $\mathcal{B}_{(s-2, s), \sigma}^{\prime}$ (входом при этом являются уравнения, задающие $\mathcal{B}_{(s-2, s), \sigma}^{\prime}$. Таким образом, нами доказана Теорема 5 Существует целое положительное s, для которого задача о вложимости аффинного пространства $A^{s}$ над $\mathbb{C}$ в алгебраические многобразие $\mathcal{B}_{(s-2, s), \sigma}^{\prime}$ алгоритмически неразрешима. Отсюда сразу следуют: Теорема 6 Существует целое положительное s, для которого задача о вложимости аффинного пространства $A^{s}$ над $\mathbb{C}$ в произволъное алгебраические многобразие $\mathcal{B}$ (заданное образующими и соотношениями) алгоритмически неразрешима. и Теорема 7 Задача о вложимости произволъного алгебраического многобразия $\mathcal{A}$ над $\mathbb{C}$ в произволъное алгебраические многобразие $\mathcal{B}$ алгоритмически неразрешима. ## 5 Заключение В работе показана алгоритмическая неразрешимость проблемы вложимости двух алгебраических многообразий над полями вещественных чисел $\mathbb{R}$ и комплексных чисел $\mathbb{C}$. Построен класс многообразий, для представителей которого алгоритмически неразрешима проблема существования вложения аффинной прямой над $\mathbb{R}$. Построен класс многообразий, для представителей которого алгоритмически неразрешима проблема существования вложения аффинного пространства размерности $m$ над $\mathbb{C}$ для достаточно большого $m$. Результаты для $\mathbb{C}$ обобщены на случай произвольного поля нулевой характеристики, а для $\mathbb{R}$ - на случай произвольного упорядоченного поля нулевой характеристики. ## Список литературы [1] Матиясевич Ю. В. Десятая проблема Гильберта. Москва, "Наука", 1993 . [2] Denef J. The diophantine problem for polynomial rings and fields of rational functions. Trans. Amer. Math. Soc. 242 (1978), p. 391-399 [3] Hazama F. Pell equations for polynomials. Ingad. Mathem. N.S. 8 (3) (1997), p. 387-397 [4] Белов А. Я., Чиликов А. А. Об алгоритмической разрешимости экспоненциально-диофантовых уравнений в кольцах положительной характеристики, Математические методы и приложения. Четвертый математический Симпозиум. М.:1997, с. 78-84. [5] Белов А. Я., Чиликов А. А. Об алгоритмической разрешимости экспоненциально-диофантовых уравнений в кольцах положительной характеристики, Курошевская Алгебраическая Конференция98. Тезисы докладов. Под редакцией Ю. А. Бахтурина, А. И. Кострикина, А. Ю. Ольшанского - М.:1998, с. 144-145. [6] Белов А. Я., Чиликов А. А. Экспоненциальные диофантовы уравнения в кольцах положительной характеристики, Фундаментальная и Прикладная Математика, с.649-668, Т.6, Выпуск 3, 2000. [7] Чиликов A. A. Ряды Тейлора алгебраических функций над полями положительной характеристики., Фундаментальная и Прикладная Математика, с.235-256, Т.7, Выпуск 1, 2001. [8] Чиликов А. А. Ряды Тейлора алгебраических функций над полями положительной характеристики., Международный алгебраический семинар, посвященный 70-летию научно-исследовательского семинара МГУ по алгебре, основанного О. Ю. Шмидтом. 2000. Тезисы докладов. М.:2000, с. 56-58. [9] A. Ya. Kanel-Belov., A. A. Chilikov Normal basises of algebras and Exponential Diophantine equations in rings of positive characteristic, https://arxiv.org/pdf/1712.00508.pdf [10] Шестая Летняя Конференция Турнира Городов, https://olympiads.mccme.ru/lktg/1994/all.htm # Метод перераспределения зарядов* Егор Бакаев, Вера Буланкина,
Александр Полянский, Андрей Рябичев, Григорий Челноков ## Вместо введения Данный проект посвящён одному методу, применяющемуся в теории графов. В частности, он используется для доказательства многих «структурных» результатов о графах, изображённых на плоскости, но не обязательно планарных. Своей популярностью он обязан тому факту, что применялся в доказательстве теоремы о четырёх красках ${ }^{1}$. Речь идёт о так называемом методе перераспределения зарлдов. По сути это обобщение подсчёта двумя способами для элементов графов, изображенных на плоскости, основанная на формуле Эйлера. Наш проект достаточно непростой: вам предстоит самостоятельно понять, как работает данный подход. Ниже для ознакомления приведены три интересные задачи, которые позже встретятся в проекте - они довольно сложные и их решение мы рекомендуем отложить на потом. 1. Конечная конфигурация точек на плоскости называется магической, если выполняется следующее условие: точкам можно так приписать положительные числа, что сумма чисел на точках любой прямой, проходящей через хотя бы две из точек конфигурации, равна 1. Опишите все магические конфигурации. 2. Граф, изображенный на плоскости, называется квазипланарным² ${ }^{2}$, если у него не найдется трёх ребер, попарно пересекающихся во внутренних точках. Докажите, что число рёбер в квазипланарном графе на $n$ вершинах не превосходит $8 n-20$. 3. Граф называется спичечным, если он может быть изображён на плоскости так, чтобы все его ребра являлись непересекающимися друг с другом отрезками длины 1. Докажите, что не существует спичечного графа, степень каждой вершины которого равна 5 . Наконец, стоит отметить, что в заключительном разделе проекта, который будет выдан после промежуточного финиша, можно найти открытые исследовательские задачи, которые, вероятно, решаются с помощью данного метода. Решение любой из них является результатом, достойным публикации.[^34] ## Часто используемые обозначения Через Г обозначим граф ${ }^{3}$. Через $V(\Gamma)$ и $E(\Gamma)$ обозначим, соответственно, множества вериин и ребер графа $Г$. Кроме того, если задано вложение Г в плоскость, то $F(\Gamma)$ обозначает множество граней этого вложения $\Gamma$. Также напомним, что для связного плоского графа Г выполняется формула Эйлера $$ |V(\Gamma)|-|E(\Gamma)|+|F(\Gamma)|=2 $$ Обозначим степень вершины $v \in V(\Gamma)$ через $\operatorname{deg}(v)$, а через $\delta(\Gamma)$ и $\Delta(\Gamma)$ минимальную и максимальную степень вершины в $V(\Gamma)$ соответственно. Граф называется $n$-регулярным, если степень каждой вершины равна $n$. Весом ребра назовем сумму степеней его концов. Назовем степенъю грани $\operatorname{deg}(f)$ длину цикла, образующего грань (таким образом, ребро считается дважды, если грань проходит по нему два раза). Будем говорить, что граф правильно раскрашен в $k$ цветов, если каждой вершине поставлено в соответствие одно из чисел от 1 до $k$ (называемых цветами) так, что соседние вершины имеют разные цвета. Минимальное такое $k$, что граф $\Gamma$ можно правильно покрасить в $k$ цветов, называется хроматическим числом и обозначается через $\chi(\Gamma)$. Подграф Г' графа Г называется индуцированным, если он содержит все ребра графа $\Gamma$, соединяющие вершины $V\left(\Gamma^{\prime}\right)$. ## 1 Подсчёт двумя способами 1.1. В некоторых клетках прямоугольной таблицы нарисованы звездочки. Известно, что для любой отмеченной клетки количество звездочек в её столбце совпадает с количеством звездочек в её строке. Докажите, что число строк в таблице, в которых есть хоть одна звездочка, равно числу столбцов таблицы, в которых есть хоть одна звездочка. 1.2. На фестиваль Зиланткон приехало $E$ эльфов и $D$ гномов. После фестиваля каждый гном подрался по крайней мере с одним эльфом, а каждый эльф - не более чем с десятью гномами. Также известно, что у каждого гнома соперников-эльфов было больше, чем у любого из них - соперников-гномов. Докажите, что $D \leqslant \frac{10}{11} E$. 1.3. В прямоугольной таблице $m$ строк и $n$ столбцов, где $mАлександр Полянский, Андрей Рябичев, Григорий Челноков ## 1 Подсчёт двумя способами 1.1. В некоторых клетках прямоугольной таблицы нарисованы звездочки. Известно, что для любой отмеченной клетки количество звездочек в её столбце совпадает с количеством звездочек в её строке. Докажите, что число строк в таблице, в которых есть хоть одна звездочка, равно числу столбцов таблицы, в которых есть хоть одна звездочка. Решение. Заменим каждую звёздочку числом $1 / k$, где $k$ - количество звёздочек в ее строке (столбце). Тогда сумма чисел в каждой непустой строке равна 1 , следовательно, сумма всех чисел в таблице равна числу непустых строк. Но, аналогично, она равна и числу непустых столбцов. 1.2. На фестиваль Зиланткон приехало $E$ эльфов и $D$ гномов. После фестиваля каждый гном подрался по крайней мере с одним эльфом, а каждый эльф - не более чем с десятью гномами. Также известно, что у каждого гнома соперников-эльфов было больше, чем у любого из них - соперников-гномов. Докажите, что $D \leqslant \frac{10}{11} E$. Решение. Дадим каждому эльфу заряд 1 , тогда суммарный заряд равен $E$. Если эльф подрался с $k$ гномами, то пусть отдаст каждому из них заряд $\frac{1}{k}$. Если гном подрался с $m$ эльфами, то каждый из них подрался не больше чем с $m-1$ гномами $(m \leqslant 11)$, а значит отдаст ему заряд не меньше $\frac{1}{m-1}$. Значит каждый гном получит заряд не меньше $\frac{m}{m-1}$, что не меньше $\frac{11}{10}$. Тогда суммарный заряд гномов не меньше $\frac{11}{10} D$, при этом он равен $E$. Отсюда получаем неравенство $\frac{11}{10} D \leqslant E$, которое легко преобразуется к требуемому виду. 1.3. В прямоугольной таблице $m$ строк и $n$ столбцов, где $m0$. 2.3. Дан планарный граф $\Gamma$ с $\delta(\Gamma) \geqslant 5$, все грани которого - треугольники, а также нет двух соседних вершин степени 5. Докажите, что найдётся грань, степени вершин которой равны 5,6 и 6 соответственно. Решение. Подставим $\alpha=\frac{1}{3}$ в равенство (1) и умножим на 3 , получим $$ \sum_{v \in V}(\operatorname{deg} v-6)+\sum_{f \in F}(2 \operatorname{deg} f-6)=-12 $$ Т.к. все грани треугольные, то второе слагаемое равно нулю: $$ \sum_{v \in V}(\operatorname{deg} v-6)=-12 $$ Дадим каждой вершине $v$ заряд $\operatorname{deg} v-6$. Пусть каждая вершина степени по крайней мере 7 отдаст заряд равный $\frac{1}{3}$ в каждую соседнюю с ней вершину степени 5 . Предположим, что нет грани $(5,6,6)$. Тогда у вершины степени 5 не более двух соседей имеют степень 6 ; т.е. у нее по крайней мере три соседа степени не меньше 7 . При $n \geqslant 7$ вершина степени $n$ инцидентна не более $\frac{n}{2}$ вершинам степени 5 , т.к. все грани треугольные и нет ребёр вида $5-5$. Тогда заряд каждой вершины степени 5 станет не менее $-1+3 \cdot \frac{1}{3}=0$. Заряд вершины степени 6 останется 0 . Заряд вершины степени 7 останется не менее $1-3 \cdot \frac{1}{3}=0$. При $n \geqslant 8$ заряд вершины степени $n$ останется не менее $n-6-\frac{n}{2} \cdot \frac{1}{3}=\frac{5}{6} n-6 \geqslant \frac{5}{6} \cdot 8-6>0$. 2.4. Дан планарный граф $\Gamma$ с $\delta(\Gamma) \geqslant 5$. Докажите, что найдется ребро веса не больше 11 . Решение. Умножая на 2 равенство (1) при $\alpha=\frac{1}{2}$, получаем $$ \sum_{v \in V(\Gamma)}(\operatorname{deg} v-4)+\sum_{f \in F(\Gamma)}(\operatorname{deg} f-4)=-8 $$ Дадим каждой вершине $v$ заряд ( $\operatorname{deg} v-4$ ), а каждой грани $f$ заряд ( $\operatorname{deg} f-4$ ). Пусть каждая вершина степени 5 отдает всем граням, которым принадлежит, по $\frac{1}{5}$, степени 6 - по $\frac{1}{3}$, степени не меньше 7 - по $\frac{2}{5}$. У вершин степени 5 и 6 заряд станет равным 0 . При $n \geqslant 7$ у вершины степени $n$ заряд станет равным $n-4-\frac{2}{5} n=\frac{3}{5} n-4 \geqslant \frac{3}{5} \cdot 7-4>0$. у всех граней степени больше 3 заряд изначально неотрицательный и при перераспределении не уменьшается. Теперь рассмотрим треугольную грань. Предположим, что у каждого ребра вес не меньше 12. Тогда набор степеней вершин треугольной грани имеет вид либо $(5,7+, 7+)$, либо $(6+, 6+, 6+)^{2}$. В первом случае грань получает от вершин заряд $\frac{1}{5}+\frac{2}{5}+\frac{2}{5}=1$, во втором $\frac{1}{3}+\frac{1}{3}+\frac{1}{3}=1$, что делает ее заряд равным 0 . 2.5. Дан планарный граф $\Gamma$ с $\delta(\Gamma) \geqslant 3$. Докажите, что найдется такая пара из грани $f$ и вершины $v$ в ней, что $\operatorname{deg}(v)=3$ и $\operatorname{deg}(f) \leqslant 5$ или $\operatorname{deg}(v) \leqslant 5$ и $\operatorname{deg}(f)=3$. Решение. Умножая на 3 равенство (1) при $\alpha=\frac{1}{2}$, получаем $$ \sum_{v \in V(\Gamma)}(\operatorname{deg} v-4)+\sum_{f \in F(\Gamma)}(\operatorname{deg} f-4)=-8 $$ Дадим каждой вершине $v$ заряд ( $\operatorname{deg} v-4$ ), а каждой грани $f$ заряд $(\operatorname{deg} f-4$ ). Предположим, что у каждой грани степени 3 все вершины имеют степень не меньше 6 , и вершины степени 3 есть только у граней степени не меньше 6 . Пусть каждая вершина степени не меньше 6 отдает, граням в которых она лежит, по $\frac{1}{3}$. Заряд таких вершин останется неотрицательным, а у треугольных граней станет равным 0 . Также пусть каждая грань, у которой не менее 6 вершин, отдает своим вершинам по $\frac{1}{3}$. У таких граней заряд останется неотрицательным, а у вершин степени 3 он станет равным 0 . У остальных вершин и граней заряд неотрицательный изначально и он не уменьшится.[^44] 2.6. Дан планарный граф $\Gamma$ с $\delta(\Gamma) \geqslant 3$. Докажите, что у $\Gamma$ есть не более чем 5 -угольная грань, в которой степени всех вершин, кроме возможно одной, не превосходят 11. Решение. Умножая на 3 равенство (1) при $\alpha=\frac{2}{3}$, получаем $$ \sum_{v \in V(\Gamma)}(2 \operatorname{deg} v-6)+\sum_{f \in F(\Gamma)}(\operatorname{deg} f-6)=-12 $$ Дадим вершине $v$ заряд, равный $2 \operatorname{deg} v-6$, а грене $f$ заряд, равный $\operatorname{deg} f-6$. Предположим противное, то есть у каждой грани степени по крайней мере 5 найдётся хотя бы 2 вершины степени не менее 12 . Пусть каждая вершина степени $n \geqslant 12$ отдаст заряд $\frac{3}{2}$ каждой гране, в которую она входит. Тогда её заряд становится равным $2 n-6-\frac{3}{2} n=\frac{1}{2} n-6 \geq 0$. Каждая грань степени не более 5 тогда получит заряд от по крайней мере двух вершин. Поэтому её заряд будет не менее $-3+\frac{3}{2} \cdot 2=0$. Все остальные вершины и грани изначально имели неотрицательный вес, а в процессе перераспределния зарядов их заряд не уменьшился. Противоречие. ## Легкие раскраски 2.7. а) Докажите, что любой планарный граф можно раскрасить в 6 цветов правильным образом. б) Пусть для любого подграфа некоторого графа найдётся вершина графа степени не превосходящей $d-1$ или индуцированный чётный цикл, степень каждой вершины которого не превосходит $d$. Докажите, что граф можно правильно раскрасить в $d$ цветов. Решение. а) В каждой грани не менее трех ребер, значит, $3 F \leqslant 2 E$. Подставив это неравенство в формулу Эйлера, получим: $6=3 V-3 E+3 F \leqslant 3 V-3 E+2 E=3 V-E$. Отсюда $0<3 V-E$ и $2 E<6 V$. Таким образом, сумма степеней вершин меньше чем $6 V$. Значит, в любом планарном графе найдется вершина со степенью меньше 6 . Докажем по индукции, что любой планарный граф на $n$ вершинах можно правильным образом раскрасить в 6 цветов. База очевидна. Докажем переход: пусть это верно для всех графов на $k$ вершинах. Рассмотрим произвольный граф $\Gamma$ на $k+1$ вершине, пусть $v-$ его вершина степени меньше 6. Раскрасим граф $\Gamma \backslash v$ в 6 цветов, после чего покрасим вершину $v$ - для нее найдется доступный цвет. б) Докажем это утверждение для всех подграфов $\Gamma$ по индукции по количеству вершин $n$. База очевидна. Докажем переход: пусть это верно для всех графов на не более чем $k$ вершинах. Рассмотрим произвольный подграф $\Gamma$ на $k+1$ вершине графа $\Gamma$. Если в $\Gamma$ найдётся вершина $v$ степени не превосходящей $d-1$, то раскрасим граф $\Gamma \backslash v$, затем раскрасим вершину $v$. Иначе в $\Gamma$ найдётся чётный цикл $v_{1} v_{2} \ldots v_{k}$, степень каждой вершины которого не превосходит $d$. Обозначим как $C$ множество его вершин. Правильно раскрасим граф $\Gamma \backslash C$. Теперь покажем, как покрасить $C$. Из каждой вершины множества $C$ выходит не более $d-2$ ребер в множество $\Gamma \backslash C$, значит для этих вершин остается хотя бы по 2 варианта, в какие цвета их можно покрасить. Докажем следующую лемму. Лемма. Если каждой вершине четного цикла назначен список из 2 цветов, в которые ее возможно покрасить, то этот цикл можно раскрасить правилъным образом согласно спискам. Предположим, это сделать нельзя. Если у всех вершин списки одинаковые, то цикл можно раскрасить, чередуя цвета. В противном случае найдутся две соседние вершины с разными списками, пусть это $v_{1}$ и $v_{k}$. Покрасим $v_{1}$ в цвет, которого нет в списке $v_{k}$. Для вершины $v_{2}$ останется хотя бы один доступный цвет, покрасим ее. Потом покрасим $v_{3}, v_{4}$ и так далее до $v_{k}$. Таким образом, любые две соседние вершиныт покрашены в разные цвета, включая пару $v_{1}$ и $v_{k}$. 2.8. a) Докажите, что для любого заданного $k$ найдется такое $n$, что полный двудольный граф $K_{n, n}$ не будет $k$-выбираемым. Хотя и является 2 -раскрашиваемым, как любой двудольный граф. б) Докажите, что цикл четной длины не только 2-раскрашиваем, но и 2-выбираем. Решение. а) Пусть в первой доле будут вершины с номерами от 1 до $k$, у вершины номер $i$ список $\left(c_{i 1}, c_{i 2}, \ldots, c_{i k}\right)$ для всех $i$, и пусть все эти $k^{2}$ цветов различны. А во второй доле - $k^{k}$ вершин со всеми возможными списками вида $\left(c_{1 x_{1}}, c_{2 x_{2}}, \ldots, c_{k x_{k}}\right)$, где $x_{i}$ это число от 1 до $k$ для всех $i$. Тогда для любой раскраски вершин первой доли найдется вершина второй доли, список которой состоит как раз из этих цветов, и ее не получится раскрасить. По условию требовалось, чтобы в долях было поровну вершин, но этого легко достичь, добавляя вершины с произвольными списками в меньшую долю. б) См. лемму из решения задачи 2.76 . 2.9. а) Дан планарный граф, внешняя грань которого это цикл $v_{1} \ldots v_{k} v_{1}$, остальные грани треугольные. Вершинам $v_{1}$ и $v_{2}$ соответствуют списки из двух цветов, каждой другой вершине внешней грани приписан список из трех цветов, а каждой внутренней вершине приписан список из 5 цветов. Докажите, что существует правильная списочная раскраска, соответствующая этим спискам. б) Докажите, что для планарного графа $\Gamma$ выполняется $\operatorname{ch}(\Gamma) \leqslant 5$. Решение. См. теорему 5.4.2 книгу Дистеля [2] 2.10. Пусть $\Gamma$ - планарный граф с $\Delta(\Gamma) \geqslant 11$ Каждой вершине назначен список из $\Delta(\Gamma)$, а каждому ребру - из $(\Delta(\Gamma)+2)$. Тогда можно так выбрать цвета из списков, чтобы получить тотальную списочную покраску. Решение. Сперва докажем следующую лемму: Лемма. В планарном графе $\Gamma$, у которого степени всех вершин не менее 3, есть ребро с весом не больше 11 или цикл из 4 вершин, проходящий через 2 вершины степени 3 и соседнююо с ними вершину степени не больше 10. Доказательство лемми. Назовем ребро лёгким, если его вес не больше 11. Пусть в $\Gamma$ нет лёгких рёбер и специальных 4-циклов, описанных в условии. Если у какой-то грани больше 3 ребер, в ней можно провести нелёгкую диагональ. Действительно, средний вес диагоналей в грани равен среднему весу рёбер её периметра, следовательно найдётся диагональ с весом хотя бы 12 . Заметим, что после добавления ребра специальные 4-циклы из условия леммы появиться не могут. Добавляя нелегкие рёбра, получим граф $\Gamma^{\prime}$ (возможно, имеющий кратные рёбра), у которого все грани треугольные. Из формулы Эйлера получаем $$ \sum_{v \in V\left(\Gamma^{\prime}\right)}(\operatorname{deg} v-6)+\sum_{f \in F\left(\Gamma^{\prime}\right)}(2 \operatorname{deg} f-6)=-12 $$ Назначим каждой вершине $v$ вес $(\operatorname{deg} v-6)$, а грани $f$ вес $(2 \operatorname{deg} f-6)$. Пусть каждая вершина $v$ степени не больше 5 забирает вес $\frac{6-\operatorname{deg} v}{\operatorname{deg} v}$ от каждой соседней вершины степени не меньше 7 , после чего вес $v$ становится равным нулю. Покажем, что после перераспределения остальные веса останутся неотрицательными, и, поскольку суммарный вес отрицателен, получим противоречие. $\Gamma^{\prime}$ - это триангуляция без легких ребер, то каждая вершина степени $k$ имеет не более $\left\lfloor\frac{k}{2}\right\rfloor$ соседей, которым она отдаёт вес. Вершина степени 7 отдает вес только вершинам степени 5 , поэтому она отдает веса не больше $3 \cdot \frac{1}{5}$ и ее вес остается положительным. Вершина степени 8 отдает вес только вершинам степени не меньше 4 , поэтому всего отдает веса не больше $4 \cdot \frac{1}{2}$ и ее вес остается неотрицательным. Если у вершины $w$ степень не меньше 11, то не более чем $\left\lfloor\frac{\operatorname{deg} w}{2}\right\rfloor$ её соседей забирают не больше чем по 1 , и вес $w$ остается не менее $\left\lceil\frac{\operatorname{deg} w}{2}\right\rceil-6$, то есть остается неотрицательным. Осталось рассмотреть случай, когда степень вершины $w$ равна 9 или 10. Так как все грани треугольные, то соседи $w$ образуют замкнутый маршрут длины $\operatorname{deg} w$. Ни одна вершина степени 3 не встречается в таком маршруте дважды, поскольку кратные рёбра не могут выходить из вершины степени 3 . Так как нет легких ребер и запрещены некоторые специальные 4 -циклы, то в этом маршруте между двумя вершинами степени 3 должно быть хотя бы три ребра. Итак, у вершины $w$ степени 9 не более трёх соседей степени 3 , и когда их ровно 3 , то у $w$ остается вес 0 , т.к. больше никакие вершины не отдают ей вес. Если у неё 2 соседа степени 3 , то не больше двух соседей степени 4 или 5 , и снова вес останется неотрицательным. Если у неё 1 сосед степени 3 , то не больше 3 соседей степени 4 или 5 , и снова вес останется положительным. Случаи для вершины степени 10 рассматриваются аналогично. Лемма доказана. Вернемся к решению задачи. Предположим, что есть графы, не удовлетворяющие условию задачи, рассмотрим минимальный (по количеству ребер) такой граф $\Gamma$ и соответствующие списки вершин. Либо у Г есть вершина степени не больше 2 , либо к нему применима лемма. Рассматривая степени вершин у лёгких рёбер, имеем три возможных случая: 1. В $\Gamma$ есть ребро $u v$ с $\operatorname{deg} u \leqslant 2$. 2. В $\Gamma$ есть ребро $u v$ с $\operatorname{deg} u=3$ и $\operatorname{deg} v \leqslant 10$. 3. В $\Gamma$ есть ребро $u v$ с $\operatorname{deg} u \leqslant 5$ и $\operatorname{deg}(u) \leqslant 7$. Так как $\Gamma$ минимальный граф, для которого нет раскраски, то для графа $\Gamma-u v$ раскраска есть. Временно удалим цвет вершины $u$, мы вновь раскрасим ее после того, как выберем цвет ребра $u v$. Рассмотрим, в каждом из случаев во сколько цветов могут быть покрашены элементы, смежные с ребром uv (считая, что $u$ не имеет цвета). В первом случае оно не больше $\Delta(\Gamma)+1$, во втором - не больше $2+1+9$, в третьем $-4+1+6$. Длина списка цветов у ребра $\Delta(\Gamma)+2 \geqslant 13$, так что для ребра $u v$ останется свободный цвет. Осталось покрасить вершину $u$. Так как $\operatorname{deg} u \leqslant 5$, то недоступно не более 10 цветов, а количество цветов в списке $\Delta(\Gamma) \geqslant 11$, так что и для $u$ найдется свободный цвет. ## 3 Серьезные задачи ## 3.1 Спичечные графы Ключевая задача 1. Докажите, что не существует 5-регулярного спичечного графа. 3.1.1. Дан граф минимальных расстояний на $n$ вершинах, все вершины которого находятся в общем положении ${ }^{3}$. Докажите, что а) число рёбер меньше $5 n / 2$; б) найдётся постоянная c меньшая $5 / 2$ такая, что число рёбер не больше $c n$. в) Граф называется интересным, если он является графом минимальных расстояний и для любой вершины графа она и все её соседи находятся в общем положении ${ }^{4}$. Докажите, что для любого $c<5 / 2$ найдется интересный граф с не менее чем $c|V(\Gamma)|$ ребрами. Решение. б) На протяжении этого доказательства Г всегда обозначает какой-либо граф минимальных расстояний. Поместим по единичному заряду в каждую вершину и каждую грань. Теперь заряд каждой грани поровну раздадим ее вершинам. Т.к. Г это граф минимальных расстояний, то степень каждой вершины не больше 5 . Кроме того, в таком графе треугольная грань не может граничить с двумя другими треугольными гранями (в этой конструкции три вершины лежат на одной прямой).[^45] Таким образом, никакая вершина не может иметь заряд больше $1+\frac{1}{2} \cdot 3=\frac{5}{2}$, и заряд $\frac{5}{2}$ может быть только у вершины степени 5 , смежной с гранями, степени которых это 3,3 , $4+, 3,4+$ с точностью до перестановки по циклу. Назовем такую вершину регулярной. Максимальный из зарядов нерегулярных вершин обозначим $\frac{5}{2}-\varepsilon$. Как уже доказано, $\varepsilon>0$. Дальнейшее решение будет заключаться в том, что мы применим следующую лемму (докажите ее). Лемма 1. Если для некоторого $\delta>0$ в $\Gamma$ есть хотя бы $\delta V(\Gamma)$ нерегулярных вершин, то $E(\Gamma) \leqslant\left(\frac{5}{2}-\delta \varepsilon\right) V(\Gamma)$. Теперь докажем, что $\Gamma$ удовлетворяет условиям леммы 1 для некоторого $\delta>0$. Для этого понадобятся еще некоторые определения. Назовем $k$-окрестностью вершины $O$ такое множество вершин графа $\Gamma$, что до каждой из них есть путь из $O$ не более чем из $k$ ребер. Количество вершин в этом множестве обозначим $s(\Gamma, O, k)$. Назовем такую окрестность достаточно большой, если в ней есть нерегулярная вершина. Лемма 2. Докажите, что существует оценка сверху количества вершин в $k$-окрестности, не зависящая от выбора вершины $O$ графа $\Gamma$. Иными словами, есть функция $S(k)$ такая, что $s(\Gamma, O, k) \leqslant S(k)$ для любых $\Gamma$ и $O$. Докажите эту лемму самостоятельно. Подсказка: степени вершин ограничены сверху. Теперь возьмем такое $k$, что для каждой регулярной вершины ее $k$-окрестность является достаточно большой. В каждой из них есть нерегулярная вершина. Таким образом, каждой регулярной вершине $X$ сопоставлена нерегулярная вершина $Y$ такая, что путь $X Y$ содержит не более $k$ ребер. Тогда каждая нерегулярная вершина $Y$ сопоставлена не более чем $S(k)$ регулярным вершинам. Таким образом, отношение количества регулярных вершин к количеству нерегулярных не больше $S(k)$, и можно применить лемму 1 для $\delta=\frac{1}{S(k)+1}$. Осталось доказать, что такое $k$, не зависящее от графа, действительно можно рассмотреть: Лемма 3. Докажите, что существует такое $k$, что любая $k$-окрестность любого графа минимальных расстояний является достаточно большой. в) Указание: Соберите граф из конструкций из 3 . 3.1.2. Решите ключевую задачу 1 . Решение. Данное решение взято из статьи [3]. Предположим противное, пусть такой граф Г существует. Умножая на 3 равенство (1) при $\alpha=\frac{1}{3}$, получаем $$ -12=\sum_{v \in V(\Gamma)}(\operatorname{deg} v-6)+\sum_{f \in F(\Gamma)}(2 \operatorname{deg} f-6)=\sum_{v \in V(\Gamma)}(-1)+\sum_{f \in F(\Gamma)}(2 \operatorname{deg} f-6) $$ Раздадим каждой вершине заряд ( -1$)$, а каждой грани $f$ - заряд, равный $(2 \operatorname{deg} f-6)$. Из (2) следует, что суммарный заряд по всем вершинам и граням отрицательный. Перераспределение зарядов будет следующим Рассмотрим грань $f \in F$ и вершину $x \in$ $V$, принадлежащую этой грани. Пусть $\alpha$ - мера внутреннего угла грани $f$ при вершине $x$. Если $\alpha>\frac{\pi}{3}$ мы забираем у грани $f$ заряд, равный $\min \left\{1, \frac{3}{\pi} \alpha-1\right\}$ и отдаем его вершине $x$. Покажем, что после такого перераспределения заряды каждой грани и каждой вершины будут неотрицательны. Рассмотрим вершину $x \in V$. Обозначим через $\ell$ количество внутренних углов граней при этой вершине, которые больше, чем $\frac{\pi}{3}$, обозначим их величины через $\alpha_{1}, \alpha_{2}, \ldots, \alpha_{\ell}$. Так как $\operatorname{deg} x=5$, то $\ell>0$. Если через какой-то из этих углов от грани его содержащей вершине $x$ передастся заряд 1 , то заряд в $x$ уже будет неотрицателен. С другой стороны, если через каждый из этих углов вершине $x$ передался заряд $\frac{3}{\pi} \alpha_{i}-1$, где $i=1, \ldots, \ell$, то общий заряд вершины увеличился на $\frac{3}{\pi} \sum_{i=1}^{\ell} \alpha_{i}-\ell$. При этом сумма углов при вершине $x$ равна $2 \pi \leq \sum_{i=1}^{\ell} \alpha_{i}+(5-\ell) \frac{\pi}{3}$, откуда следует, что $$ \sum_{i=1}^{\ell} \alpha_{i} \geq 2 \pi-(5-\ell) \frac{\pi}{3}=\frac{\pi}{3}(\ell+1) $$ Но тогда заряд, полученный вершиной $x$ по крайней мере $\frac{3}{\pi} \frac{\pi}{3}(\ell+1)-\ell=1$. Таким образом заряд в вершине $x$ после перераспределения неотрицательный. Теперь докажем неотрицательность зарядов граней. Сначала рассмотрим случаи ограниченных граней. Начальный заряд грани $f \in F$ был равен $2 \operatorname{deg} f-3 \geq 0$. Тогда, если после перераспределения, заряд грани стал отрицательным, то значит по крайней мере один из ее внутренних углов больше, чем $\frac{\pi}{3}$. Если $\operatorname{deg} f=3$, то грань $f$ - равносторонний треугольником (в силу спичечности графа), все углы в котором равны $\frac{\pi}{3}$. Но тогда ее заряд остается неотрицательным. Если $\operatorname{deg} f=4$, то грань $f$ - ромб. Если только два его внутренних угла больше, чем $\frac{\pi}{3}$, то тогда отданный этой гранью заряд не больше 2 , что оставляет общий заряд грани неотрицательным. Если все четыре угла больше $\frac{\pi}{3}$,, то заряд, отданный гранью, опять же не больше, чем $\frac{3}{\pi} 2 \pi-4=2$. Тогда заряд снова остается неотрицательным. Случай $\operatorname{deg} f^{\pi}=5$ разбивается на два. Если не более четырех углов больше, чем $\frac{\pi}{3}$, то отдаваемый заряд не больше 4 , то есть суммарный остается неотрицательным. Если же все углы больше, чем $\frac{\pi}{3}$, то, так как сумма углов в пятиугольнике равна $3 \pi$, отданный заряд не более $\frac{3}{\pi} 3 \pi-5=4$. Опять же заряд грани остается неотрицательным. Наконец, если $\operatorname{deg} f \geq 6$, то отдаваемый заряд не более, чем $\operatorname{deg} f$ и оставшийся заряд неотрицателен: $2 \operatorname{deg} f-3-\operatorname{deg} f \geq 0$. Осталось рассмотреть случай внешней грани $f \in \Gamma$. Если $\operatorname{deg} f \geq 6$, то он аналогичен случаю ограниченной грани. Случаи, когда $\operatorname{deg} f \leq 5$, также легко исключаются. С другой стороны, даже не рассматривая их подробно, можно заметить, что для таких случаев общий отданный заряд не превосходит 5 , то есть, оставшийся заряд не менее -5 (а на самом деле не менее -3 ), что со всеми остальными неотрицательными зарядами граней и вершин, никак не даст в сумме изначально посчитанный общий заряд равный -12 . Получаем противоречие. Следовательно, такого спичечного графа не существует. 3.1.3*. Докажите, что в 4-регулярном спичечном графе не менее 20 вершин. Решение. См. раздел 3 и теорему 3.6 в статье [3]. ## 3.2 Квазипланарные графы Ключевая задача 2. а) Квазипланарный граф на $n$ вершинах содержит не более $8 n-20$ ребер. б) Попробуйте усилить утверждение предыдущей задачи. Интересуют как улучшения оценки, так и обобщения для других классов графов. Решение. С полным решение задачи можно познакомиться в статье [4]. Основная идея в том, чтобы свести задачу к рассмотрению плоского графа, и применить к нему метод перераспределения зарядов. Рассмотрим следующий плоский граф $\Gamma^{\prime}$. Его множество вершин состоит из вершин квазипланарного графа $\Gamma$, а также точек пересечения рёбер $\Gamma$. Рёбрами $\Gamma$ будут пары вершин, являющиеся концами участков кривых, определявших рёбра Г. Обозначим через $V, E, F, V^{\prime}, E^{\prime}, F^{\prime}$ количества вершин, ребер и граней в графах $\Gamma$ и $\Gamma^{\prime}$ соответственно. Вершины $V(\Gamma)$ будем называть старыми, а остальные - новыми. Под степенями вершин и граней будем понимать их степени в новом графе $\Gamma^{\prime}$. Нетрудно убедиться (докажите это самостоятельно), что $E^{\prime}-E=2\left(V^{\prime}-V\right)$, т.е. что количество ребер увеличилось на удвоенное количество новых вершин. Итак, $$ 2 E^{\prime}=2 E+4\left(V^{\prime}-V\right) $$ Именно последнее равенство является ключевым новым моментом решения. Теперь запишем формулу Эйлера для графа $\Gamma^{\prime}$, домноженную на $(-2)$ : $$ -2 V^{\prime}+2 E^{\prime}-2 F^{\prime}=-4 $$ Тогда, используя равенства $2 E^{\prime}=\sum_{v \in V\left(\Gamma^{\prime}\right)} \operatorname{deg} v, 2 E^{\prime}=\sum_{f \in F\left(\Gamma^{\prime}\right)} \operatorname{deg} f$, а также (3) и (4), получим (поймите самостоятельно, как это делается): $$ 2 \gamma E+\sum_{v \in V\left(\Gamma^{\prime}\right) \backslash V(\Gamma)}(\alpha \operatorname{deg} v-2+4 \gamma)+\sum_{v \in V(\Gamma)}(\alpha \operatorname{deg} v-2)+\sum_{f \in F\left(\Gamma^{\prime}\right)}(\beta \operatorname{deg} f-2)=-4 $$ где $\alpha+\beta+\gamma=1$ Далее в каждом конкретном случае будем по-своему подбирать коэффициенты $\alpha, \beta$ и $\gamma$. Во всех трёх задачах новым вершинам мы будем давать нулевой заряд (и они вообще не будут участвовать в перераспределении зарядов). Для этого нам потребуется выполнение равенства $4 \alpha-2+4 \gamma=0$, т.к. степень каждой новой вершины равна 4 . Тогда $\beta=0.5$ и (5) преобразуется в $$ 2 \gamma E+\sum_{v \in V(\Gamma)}((0.5-\gamma) \operatorname{deg} v-2)+\sum_{f \in F\left(\Gamma^{\prime}\right)}(0.5 \operatorname{deg} f-2)=-4 $$ 3.2.1. Пусть граф удовлетворяет дополнительному условию: не существует тройки ребер $e_{1}, e_{2}, e_{3}$ такой, что $e_{1}$ и $e_{2}$ выходят из одной вершины $A$, а ребро $e_{3}$ пересекает их во внутренних точках $B$ и $C$, причем нет других точек пересечения на участках $A B$ и $A C$. Докажите, что число ребер не больше $4 n-8$. Решение. В этом случае возьмём ${ }^{5} \gamma=0.25$. Учитывая, что нам нужно доказать, что $E \leqslant 4 V-8$, преобразуем (6) в $$ -E+4 V-8=\sum_{v \in V(\Gamma)} 0.5 \operatorname{deg} v+\sum_{f \in F\left(\Gamma^{\prime}\right)}(\operatorname{deg} f-4) $$ Покажем, что правая часть неотрицательна. Дадим каждой старой вершине $v$ заряд, равный $0.5 \operatorname{deg} v$, а каждой грани $\Gamma^{\prime}-$ заряд, равный $(\operatorname{deg} f-4)$. Покажем, как перераспределить заряды, чтобы у каждого элемента он стал неотрицательным. От каждой вершины дадим заряд, равный 0.5 , в каждую из граней, в которых она лежит. Отрицательный заряд мог остаться только у треугольников с двумя или тремя новыми вершинами. Но, как несложно убедиться, вспомнив условие задачи, таких граней в графе $\Gamma^{\prime}$ нет. 3.2.2. Без дополнительного условия из прошлой задачи докажите оценку на число ребер $10 n-20$. Решение. В этом случае возьмём ${ }^{6} \gamma=0.1$. Учитывая, что нам нужно доказать, что $E \leqslant 10 V-20$, преобразуем (6) в $$ -E+10 V-20=\sum_{v \in V(\Gamma)} 2 \operatorname{deg} v+\sum_{f \in F\left(\Gamma^{\prime}\right)}(2.5 \operatorname{deg} f-10) $$ Покажем, что правая часть неотрицательна. Дадим каждой старой вершине $v$ заряд, равный $2 \operatorname{deg} v$, а каждой грани $\Gamma^{\prime}$ - заряд, равный $(2.5 \operatorname{deg} f-10)$. Перераспределение зарядов будет следующим. От каждой вершины дадим заряд, равный 2 , в каждую из граней, в которых она лежит. Таким образом, заряд всех вершин станет равным 0. От каждой грани $f \in F\left(\Gamma^{\prime}\right)$ с $\operatorname{deg} f \geqslant 5$ отдадим заряд, равный 0.5 , в каждую из смежных по ребру граней. Если соседняя по ребру грань является четырехугольником без старых вершин (такие четырехугольные грани будем называть новыми, а другие - старыми), то передадим от неё заряд через противоположное ребро в следующую грань (рис. 2), если следующая вновь новый четырехугольник, то будем передавать заряд дальше до тех пор, пока не встретим грань, отличную от нового четырехугольника (убедитесь в том, что этот процесс закончится).[^46] ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-225.jpg?height=302&width=740&top_left_y=183&top_left_x=612) Рис. 2: Передача заряда по цепочке новых четырехугольников. После этого мы совершим вторую фазу перераспределения зарядов. Заметим, что старые четырехугольники будут иметь заряды по крайней мере 2 , которые они получили от инцидентных им вершин. Тогда раздадим их заряды поровну (по 0.5) во соседние по стороне грани. Если соседняя по ребру грань является новым четырехугольником, то передадим от неё заряд через противоположное ребро в следующую грань, а если следующая грань вновь новый четырехугольник, то передадим заряд дальше аналогичным образом до тех пор, пока вновь не встретится грань, отличная от нового четырехугольника. У граней степени не меньше 4 заряд окажется неотрицательным. Рассмотрим треугольную грань. Если у нее две старые вершины, то в итоге у них положительный заряд. Треугольников без старых вершин возникнуть не могло по определению квазипланарных графов. Убедитесь самостоятельно, что у треугольников с одной старой вершиной будет в итоге тоже неотрицательный заряд. 3.2.3. Решите ключевую задачу $2 \mathrm{a}$. Скорее всего, при решении предыдущего пункта вы использовали такую систему зарядов, которая сработает и здесь. Решение. Попробуйте самостоятельно решить эту задачу. Для этого воспользуйтесь решением предыдущей задачи. Ключевой момент, за счёт которого удаётся решить задачу: среди граней, примыкающих к старым вершинам, имеется по крайней мере две грани, не являющиеся треугольниками с в точности одной старой вершиной. ## 3.3 Списочные покраски ребер Ключевая задача 3. Докажите, что $\Gamma$ является реберно $(\Delta(\Gamma)+1)$-выбираемым, если $\Delta(\Gamma) \geqslant 6$. Решение. С полным доказательством можно познакомиться в статье [5]. ### 3.3.1. Докажите, что при $\Delta(\Gamma) \geqslant 7$ граф Г является реберно $(\Delta(\Gamma)+1)$-выбираемым. Решение. Достаточно доказать что найдется ребро с весом не больше $\Delta(\Gamma)+2$ (такие ребра будем называть легкими), это позволит сделать легкую раскраску ребер. Если в графе есть вершины степени 1 или 2, то ребра из них автоматически легкие, значит достаточно рассмотреть случай $\delta(\Gamma) \geq 3$. Присвоим каждой вершине $v$ и каждой грани $f$ заряды $(\operatorname{deg}(v)-4)$ и $(\operatorname{deg}(f)-4)$ соответственно. Тогда по формуле Эйлера сумма всех зарядов равна -8. Наш план - придумать такое правило перераспределения зарядов, чтобы из отсутствия легкого ребра следовала неотрицательность зарядов у всех элементов, что привело бы к противоречию. Правило следующее. Пусть каждая вершина степени не менее 5 отдаст $\frac{1}{2}$ каждой треугольной грани, в которой находится. Кроме того, каждая вершина степени $\Delta(\Gamma)$ отдаст по $\frac{1}{3}$ заряда всем вершинам степени 3 , с которыми соединена. Докажем, что все заряды неотрицательны. Для граней это очевидно: заряд был отрицателен только у треугольных, но в каждой треугольной грани находятся хотя бы две вершины степени не меньшей, чем 5 , иначе рядом стоят две вершины степени не больше 4 - ребро между ними легкое. Теперь разберемся с вершинами. - вершина степени 3 должна быть соединена только с вершинами степени $\Delta(\Gamma)$, иначе соответсвующее ребро легкое. Значит, эта вершина получила трижды по $\frac{1}{3}$ - заряд стал нулевым. - Вершина степени 4 изначально имела заряд 0 и никому ничего не отдавала. - Вершина степени $t$ от 5 до $\Delta(\Gamma)$-1 отдает что-то только треугольным граням, которых у нее не больше чем $\left[\frac{t}{2}\right]$, легко проверить что при $t \geq 5$ выполняется $t-4-\frac{1}{2}\left[\frac{t}{2}\right] \geq 0$. - Вершина степени $\Delta(\Gamma)$ отдает заряд соседним вершинам степени 3 и треугольникам, но в каждом треугольнике вершин степени 3 не больше одной. Итого, вершины этого типа отдают максимум $\frac{5}{6}$ заряда за каждую пару своих ребер, входящих в треугольник, и максимум $\frac{1}{3}$ за каждое не входящее в треугольник ребро. Также легко проверяется, что оставшийся заряд положителен. Кстати, это единственное место в доказательстве, где нужна строгость неравенства $\Delta(\Gamma)>6$. Таким образом, получили противоречие. 3.3.2. В случае $\Delta(\Gamma)=6$ нас интересует еще одна конфигурация: вершина степени 6 , входящая в три треугольника, два из которых с набором степеней $(6,6,3)$, а третий $(6,6,3),(6,5,4)$ или $(6,6,4)$. Докажите, что при наличии такой конфигурации, можно свести вопрос раскраски $\Gamma$ к вопросу раскраски его подграфа. Решение. Пусть описанная в условии конфигурация нашлась (можно считать, что нашлась вершина, принадлежащая треугольникам с набором степеней $(6,6,3),(6,6,3)$ и $(6,6,4)$, остальные являются ее подмножествами). Выбросим ребра этой концфигурации, покрасим остальные по предположению индукции и попробуем докрасить выброшенные. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-226.jpg?height=457&width=597&top_left_y=1325&top_left_x=684) Рис. 3: Конфигурация. На рисунке ребра конфигурации пронумерованы латинскими буквами, и к каждому ребру приписано минимальное число доступных для него цветов. Если у какого-то ребра доступных цветов на самом деле больше - запретим лишний (любой). Будем красить ребра, при этом каждый раз на 1 уменьшая число доступных цветов у соседей, кроме специально оговоренных случаев. Если ни разу не опустились до нуля - все сработало. Ребрам $g$ и $j$ доступно 3 и 2 цвета соответсвенно, а ребру $h$ - только 3. Значит или есть цвет, доступный обоим ребрам $g$ и $j$, или есть цвет, доступный одному из $g$ и $j$ но недоступный $h$. В первом случае покрасим в этот общий цвет ребра $g$ и $j$, и дальше красим ребра в порядке ebdafcih. Во втором случае в тот цвет, который не доступен для $h$, красим то из ребер $g$ и $j$ для которого он доступен, потом оставшееся из двух - в любой (после этого для $h$ все еще доступно минимум 2 цвета), дальше то же самое - красим ребра в порядке ebdafcih. 3.3.3. Наконец, докажите, что при $\Delta(\Gamma)=6$ граф $\Gamma$ содержит хотя бы один из трех типов подграфов ${ }^{7}$, позволяющих свести вопрос раскраски графа к раскраске меньшего графа.[^47] Решение. Докажем, что планарный граф $\Gamma$ с $\Delta(\Gamma)=6$ и без треугольников с общим ребром имеет одну из трех конфигураций: (i) ребро веса не больше 8 ; (ii) индуцированный 4 -цикл $u v w x$, такой что $\operatorname{deg}(u)=\operatorname{deg}(w)=3$; (iii) 6-вершину $A$ входящую в три треугольника с наборами степеней вершин $(6,6,3)$, $(6,6,3)$ и $\left(6,6^{-}, 4^{-}\right)^{8}$. Доказательство. Если в графе есть вершины степени 1 или 2 , то ребра из них автоматически не тяжелее 8 , далее рассматриваем случай $\delta(\Gamma) \geq 3$. Наш дальнейший план действий - раздать заряды с отрицательной суммой вершинам и граням и показать, что если запрещенных конфигураций нет, то есть способ пререспределить заряды так, что каждый заряд станет неотрицательным. Это противоречие докажет, что запрещенные конфигурации есть. В дальнейшем следует помнить, что вершина степени 3 смежна только с вершинами степени 6 , иначе, опять же, появляется легкое ребро. Раздадим в вершины $v$ и грани $f$ заряды $\operatorname{deg}(v)-4$ и $\operatorname{deg}(f)-4$ соответственно. Теперь перераспределим их по следующим правилам: - 5- и 6-грани отдают по $\frac{1}{2}$ каждой 3-вершине, которая в них лежит; - 5-вершина отдает $\frac{1}{2}$ каждому треугольнику, в котором лежит; - 6-вершина, во-первых, отдает смежной с ней 3 -вершине $\frac{1}{6}$, если эта 3 -вершина лежит в 5 - или 6 -грани, иначе отдает $\frac{1}{3}$, во-вторых, отдает каждому треугольнику, в котором лежит, $\frac{1}{3}$ если этот треугольник не содержит 3 - или 4 -вершин, иначе отдает $\frac{1}{2}$. Докажем что все получилось. - В треугольнике не может быть двух $4^{-}$-вершин, ребро между ними было бы легким. Треугольник в котором есть одна $4^{-}$-вершина с каждой из двух других получает по $\frac{1}{2}$. Треугольник, в котором только $5^{+}$-вершины с каждой из них получает минимум пО $\frac{1}{3}$. - Четырехугольник имеет изначально неотрицательный заряд и никому ничего не отдает. - $t$-грань при $t \geq 5$ содержит не больше $\left[\frac{t}{2}\right] 3$-вершин, легко проверить что $$ t-4-\frac{1}{2} \cdot\left[\frac{t}{2}\right] \geq 0 $$ - 3 -вершина или лежит в $5^{+}$-грани, и тогда получает $\frac{1}{2}$ от нее и по $\frac{1}{6}$ от трех соседних 6 -вершин, или не лежит, и тогда получает по $\frac{1}{3}$ от трех соседних 6 -вершин. - 4-вершина имеет изначально ненулевой заряд и никому ничего не отдает. - 5-вершина отдает по $\frac{1}{2}$ максимум двум треугольникам, в которых лежит. - Чтобы убедиться, что некоторая вершина $B$ степени 6 имеет положительный заряд, придется отдельно рассмотреть четыре случая чему может быть равно число $k$ треугольников, которым принадлежит $B$. Во всех случаях полезно помнить, что $B$ может быть смежна не более чем с $6-k$ вершинами степени 3 (потому что в треугольнике максимум одна). $-k=0-B$ отдает максимум шести 3 -вершинам по не более чем $\frac{1}{3}$ - остается неотрицательный заряд 9 .[^48]$-k=1$. Если $B$ соединена не более чем с четырьмя 3 -вершинами, то она отдает $\frac{1}{2}$ треугольнику и максимум $4 \cdot \frac{1}{3}$ вершинам степени 3 - остается положительный заряд. Если $B$ соединена ровно с пятью, то какие-то два ребра в 3 -вершины идут подряд по часовой стрелке (на самом деле, минимум 4 пары ребер идущих подряд). Пусть это ребра $A B$ и $B C$. Тогда они в одной грани, и, чтобы это не был случай (ii), надо, чтобы грань, содержащая $A, B$ и $C$, (вида ... $A B C \ldots$ ) содержала еще как минимум две вершины. Но тогда вершинам $A$ и $B$ положено отдавать только $\frac{1}{6}$ - снова остался положительный заряд. $-k=2$ разбирается аналогично предыдущему случаю. $-k=3$ Если из трех треугольников не более чем два имеют $4^{-}$вершину - то $B$ отдает не более чем $2 \times \frac{1}{2}+\frac{1}{3}$ треугольникам, и максимум двум 3 -вершинам (третьей нет, потому что тогда был бы третитй треугольник с $4^{-}$-вершиной) по $\frac{1}{3}$ - сумма неотрицательна. Итого, если заряд отрицательный, то в каждом из трех треугольников есть $4^{-}$-вершина - треугольники утянули $\frac{3}{2}$ заряда. Значит, чтобы оставшийся стал отрицательным, минимум две из $4^{-}$-вершин должны быть ровно 3-вершинами. Это в точности конфигурация (iii). ## 3.4 Магические конфигурации Ключевая задача 4. Описать все магические конфигурации. Решение. Очень непросто записать кратко решение данной задачи (даже разбитой на большое количество вспомогательных шагов), поэтому ниже предлагаются, скорее, подсказки к решению, чем само решение. Доказательство опиратается на ряд важных (но в тоже время несложных) свойств набора прямых на плоскости (теорема Сильвестра, её усиление, теорема Леви). Некоторые из них мы оставим читателю для самостоятельного изучения. Полное доказательство можно найти в статье [6]. 3.4.1. Докажите теорему Сильвестра, используя формулу Эйлера и двойственный язык. Решение. Предположим, что для любых двух экваторов найдётся третий, проходящий через их точку пересечения. Рассмотрим плоский граф, с вершинами в точках пересечения экваторов, а рёбрами будут участки дуг между соседними точками на какой-то экваторе. Тогда у этого плоского графа степень каждой вершины будет по крайней мере 6 , что возможно по формуле Эйлера только, если в графе есть кратные рёбра. Это означает, что все экваторы проходят через две диаметрально противоположные точки на сфере. 3.4.2. Пусть в магической конфигурации некоторой прямой (некоторому экватору) приписано число большее $\frac{1}{2}$. Докажите, что все остальные прямые (экваторы) пересекаются в одной точке (двух антиподальных точках). Решение. Через е обозначим экватор с приписанным числом большим $\frac{1}{2}$. Заметим, что другим экваторам приписаны числа меньшие $\frac{1}{2}$, так как иначе в точке пересечения другого экватора с е сумма чисел была бы больше 1. Чтобы завершить доказательство, нужно показать, что найдётся два экватора, отличных от $e$, таких, что через одну из их точек пересечения не проходят другие экваторы (включая $e$ ). Это можно сделать, например, следующим образом. Рассматриваем рисунок на плоскости, соответствующий рисунку из экваторов на сфере, у которого образ е был бы бесконечно удалённой прямой. Далее применяется доказательство теоремы Сильвестра через минимальные расстояния между точкой пересечения и прямой, не проходящей через эту точку. Попробуйте самостоятельно восставновить доказательство. Например, можно посмотреть, как это делается в доказательстве леммы 5.6 в книге [7]. 3.4.3. Докажите, что есть точка, в которой пересекаются ровно две синие и одна красная прямая (экватор). Напомним, что конфигурация состоит из прямых (экваторов), а не из точек. Решение. Как мы знаем, по теореме Сильвестра для синих экваторов найдётся две синих экватора, через точку пересечения которых не проходит больше синих экваторов. Следовательно через их точку пересечения должна проходить красный экватор, так как сумма весов для каждой точки пересечения должна равняться 1. 3.4.4. Докажите, что в конфигурации есть синий четырехугольник разбиения с красными диагоналями. Решение. Рассмотрим плоский граф, образованный точками пересечений синих экваторы, рёбрами которого будут пары вершин, лежащих на одном синем экваторе, между которыми на экваторе нет других вершин. Воспользуемся формулой (1) при $\alpha=\frac{1}{3}$. Дадим каждой вершине заряд равный ( $\operatorname{deg} v-6)$, а каждой грани заряд равный $(2 \operatorname{deg} f-6)$. Заметим, что для каждой вершины степени 4 (через которую проходят две синие прямые), найдётся красная прямая, проходящая через неё, так как суммарный заряд должен равняться 1. Начнём перераспределять заряды (это процесс мы продолжим в последующих задачах). От каждой грани, для которой какой-то красный экватор является диагональю, передадим по заряду, равному 1 , в концы этой диагонали. Убедитесь самостоятельно, что если не найдётся синего четырёхугольника с красными диагоналями, то заряд каждой грани и каждой вершины будет неотрицательным. Таким образом, мы получаем противоречие. 3.4.5. Докажите, что найдется синий треугольник разбиения, имеющий общую сторону с плохим четырехугольником. Решение. На текущий момент единственный граней с отрицательным зарядом - плохой четырехугольник (напомним, что у вершин заряды получаются неотрицательными). Точнее у плохого четырехугольника будет заряд равный ( -2 ). Далее передадим от каждой грани смежной с плохим четырехугольником по заряду равному $\frac{1}{2}$. Так как суммарный заряд отрицательный, то обязательно найдётся плохой треугольник. 3.4.6. Докажите, что если плохой треугольник имеет общие стороны сразу с двумя плохими четырехугольниками, то наша конфигурация является двойственной конфигурацией недоФано. Решение. Убедитесь в этом самостоятельно. Это нетрудно следует из того, что один из красных экваторов будет пересекать синий экватор в двух точках (вершины двух плохих четырехугольников). 3.4.7. Докажите, что найдутся плохой треугольник $t$, его плохой четырехугольник $d$ и их общая вершина $A$ такие, что кусок разбиения, вертикальный ${ }^{10} t$ относительно $A$, является четырехугольником. Решение. На данный момент единственный тип граней с отрицательным зарядом плохие треугольники. Заряд в каждом из них будет равен $-\frac{1}{2}$, так как мы предположили, что плохой треугольник соседствует всего с одни плохим четыреугольником. Теперь от каждой грани степени по крайней мере 5 , которая вертикальна плохому треугольнику и смежна его плохому четыреугольнику передадим по заряд равный $\frac{1}{2}$ в плохой треугольник. Убедитесь самостоятельно, что в этом случае каждая из граней степени по крайней мере 5 будет с положительным зарядом. Из этого будет следовать наличие требуемая конструкция. На самом деле всегда можно найти плохой треугольник $t$ смежный плохому четырехугольнику $d$, который либо будет смежен обычным четырехугольникам, вертикальным $t$, либо будет смежен плохому треугольнику и обычному четырехугольнику, которые вертикальны $t$. Только у таких треугольников $t$ будут отрицательные заряды, равные $-\frac{1}{2}$. Плохие треугольники в последней конструкции будем называть оченъ плохими, смежные им плохие четырехугольник будем[^49]называть оченъ плохими, а хорошим четырехугольником будет называться четырехугольник вертикальный очень плохому треугольнику. 3.4.8. Рассмотрим разбиение проективной плоскости (сферы) синими прямыми (экваторами). Докажите, что найдётся такая прямая (экватор), что среди частей разбиения, примыкающих к этой прямой (экватору), ровно две (четыре) являются треугольными. Решение. Итак, на текущий момент гранями с отрицательным зарядом являются очень плохие треугольники. Заряд каждого такого треугольника всё ещё $-\frac{1}{2}$. Далее мы хотим показать, что заряда в хороших четырехугольниках достаточно, чтобы компенсировать заряд во всех очень плохих треугольниках. Теперь рассмотрим двудольный граф. Вершинами первой доли этого двудольного графа будут пары $(b, e)$, где $b$ - это очень плохой четырехугольник, а $e$ - экватор, образующий смежную сторону с его очень плохим треугольником $t$. Вершины другой доли будут пары $(g, e)$, где $g$ - хороший четырехугольник, а $e$ - экватор, который образует сторону $g$ и сторону очень плохого четырехугольника (соотвествующего $g$ ), при этом $g$ и его очень плохой четырехугольник лежат по одну сторону относительно $e$. Пары вершин (из разных долей) вида $(b, e)$ и $(g, e)$ соединим ребром, если $b$ и $g$ смежные и лежат по одну сторону относительно $e$. Степень каждой вершины в получившемся графе равна 1 или 2 . Если в получившемся графе рёбер больше, количества вершин в первой доле, то несложно убедиться (это предлагается сделать самостоятельно), что суммарный заряд неотрицательный Если же суммарный заряд отрицательный, то обязательно найдётся следующая конструкция: индуцированный путь $\left(b_{1}, e\right),\left(g_{1}, e\right), \ldots,\left(g_{k}, e\right),\left(b_{k+1}, e\right)$, где $b_{1}, g_{1}, \ldots, g_{k}, b_{k+1}-$ смежные четырехугольники, лежащие от ограничевающего их экватора $e$ с одной стороны, а также четырехугольники $b_{1}$ и $b_{k+1}$ смежны с треугольниками. Теперь докажите самостоятельно, что найдется искомая прямаяю 3.4.9. Покажите, что случай, описанный в предыдущей задаче, тоже невозможен. Решение. Докажите самостоятельно теорему Леви, утверждающую, что если есть семейство экваторов на сфере, то к каждому из них примыкает по крайней мере шесть треугольников (исключение состоит в случае, когда все прямые кроме возможно одной проходят через одну точку). См. предложение 5.13 в [7]. ## Список литературы [1] А.Канель and А.Ковальджи. Треугольники и катастрофы. Квант, (11), 1992. [2] Р Дистель. Теория графов. Изд-во Ин-та математики Новосибирск, 2002. [3] Sascha Kurz and Rom Pinchasi. Regular matchstick graphs. arXiv preprint arXiv:1401.4372, 2014 . [4] Eyal Ackerman and Gábor Tardos. On the maximum number of edges in quasi-planar graphs. Journal of Combinatorial Theory, Series A, 114(3):563-571, 2007. [5] Daniel W Cranston. Edge-choosability and total-choosability of planar graphs with no adjacent 3-cycles. arXiv preprint math/0512518, 2005. [6] Eyal Ackerman, Kevin Buchin, Christian Knauer, Rom Pinchasi, and Günter Rote. There are not too many magic configurations. Twentieth Anniversary Volume: Discrete \& Computational Geometry, page 1, 2009. [7] Stefan Felsner. Geometric graphs and arrangements: some chapters from combinatorial geometry. Springer Science \& Business Media, 2012. # Discharging method* Egor Bakaev, Vera Bulankina,
Grigory Chelnokov, Alexandr Polyanskii, Andrey Ryabichev ## Introduction This project is devoted to a certain classical idea in graph theory. It plays a crucial role in the proofs of many "structural" results on graphs drawn in the plane that are not necessarily planar. This idea became well-known because of its application in the proof of the celebrated four color theorem. This idea is the so-called discharging method: it is a certain form of double counting based on Euler's Formula for a planar graph. Our project is aimed at an experienced problem solver. We do not plan to show stepby-step instructions and explain how considered method works. Instead, we are going to provide sufficient amount of problems of gradually increasing difficulty, which will form a ladder both to understanding the method and to the valued results achieved throw it. Below you can find three interesting problems of the project, which are very difficult, and so we recommend you to postpone solving them for now and to return to them only after the semifinal. 1. Let $P$ be a finite set of points in the plane. $P$ is called a magical configuration if there is an assignment of positive weights to the points of $P$ such that, for every line $\ell$ containing at least two points of $P$, the sum of the weights of all points of $P$ on $\ell$ equals 1. Describe all magical configurations. 2. We say that a graph $\Gamma$ is drawn in the plane if the vertices $V(\Gamma)$ are represented by distinct points and the edges $E(\Gamma)$ are represented by (Jordan) arcs, each connecting two vertices and containing no other vertex. We say a graph is quasi-planar if it can be drawn in the plane in such a way that no 3 of its edges are pairwise crossing in their inner points ${ }^{1}$. Prove that, for a quasi-planar graph $\Gamma$, the inequality $|E(\Gamma)| \leq 8|V(\Gamma)|-20$ holds. 3. A graph $\Gamma$ is called a matchstick graph if $\Gamma$ can be drawn in the plane in such a way that edges are represented by segments of length 1 , and no two edges cross. Prove that a matchstick graph must have a vertex of degree different from 5 . Also, it is worth mentioning that in the last section of our project (that will be given to you after the semifinal) you can find some open research problems that probably could be solved using this method. A solution of any of them is a publicable result.[^50] ## Useful notations Denote by $V(\Gamma)$ and $E(\Gamma)$, respectively, the sets of the vertices and the edges of a graph ${ }^{2}$ $\Gamma$. Moreover, if an embedding of $\Gamma$ in the plane is given, then $F(\Gamma)$ denotes the set of the faces of the drawing of $\Gamma$. Famous Euler's Formula states that, for a connected planar graph $\Gamma$, $$ |V(\Gamma)|-|E(\Gamma)|+|F(\Gamma)|=2 $$ Denote the degree of the vertex $v \in V(\Gamma)$ by $\operatorname{deg}(v)$. Also, denote by $\delta(\Gamma)$ and $\Delta(\Gamma)$ the minimum and maximum degree of a vertex in $V(\Gamma)$, respectively. A graph is called $n$-regular if the degree of any vertex is $n$. The weight of an edge $u v$ is $\operatorname{deg}(u)+\operatorname{deg}(v)$. Call the degree $\operatorname{deg}(f)$ of a face $f$ the number of edges along the boundary of $f$ (that is a cut-edge is counted twice). Definition. We say that a graph has a proper coloring in $k$ colors (or it is $k$-colorable) if for each vertex one can assign one of the numbers $1, \ldots, k$ (called colors) in such a way that adjacent vertices are colored differently. The chromatic number $\chi(\Gamma)$ is the least $k$ such that a graph $G$ has a proper coloring in $k$ colors. A subgraph $\Gamma^{\prime}$ of a graph $\Gamma$ is induced if $E\left(\Gamma^{\prime}\right)$ contains all edges of $E(\Gamma)$, that connect vertices of $V\left(\Gamma^{\prime}\right)$. ## 1 Double counting 1.1. Some cells of a given table are marked. For every marked cell, the number of marked cells in its column equals to the number of the marked cells in its row. Prove that the number of rows with at least one marked cell is equal to the number of columns with at least one marked cell. 1.2. $E$ elves and $D$ dwarfs arrived to the Zilantkon convention. It turns out that during the convention every dwarf got into a fight with at least one elf and every elf got into a fight with at most ten dwarfs. Also, it is known that every dwarf has more opponents-elves than any of his opponents-elves has opponents-dwarfs. Prove that $D \leq \frac{10}{11} E$. 1.3. A table has $m$ rows and $n$ columns, where $mGrigory Chelnokov, Alexandr Polyanskii, Andrey Ryabichev ## 1 Double counting 1.1. Some cells of a given table are marked. For every marked cell, the number of marked cells in its column equals to the number of the marked cells in its row. Prove that the number of rows with at least one marked cell is equal to the number of columns with at least one marked cell. Solution. Let us put in every marked cell $c$ the number $\frac{1}{k}$, where $k$ is the number of marked cells in the column (the row) of $c$. Clearly, the sum of numbers in each nonempty row is 1 , and so the total sum of all numbers in the table equals the number of non-empty rows. Analogously the total sum of all numbers in the table equals the number of non-empty columns. 1.2. $E$ elves and $D$ dwarfs arrived to the Zilantkon convention. It turns out that during the convention every dwarf got into a fight with at least one elf and every elf got into a fight with at most ten dwarfs. Also, it is known that every dwarf has more opponents-elves than any of his opponents-elves has opponents-dwarfs. Prove that $D \leq \frac{10}{11} E$. Solution. Let us asing charge 1 to every elf, then the total charge is equal to $E$. If an elf has $k$ opponents-dwarfs, then he gives the charge $\frac{1}{k}$ to each of those. If an dwarf Duda has $m$ opponents-elfs, then each of those has at most $m-1$ opponentsdwarf, where $m \leq 11$, and thus give Duda a charge at least $\frac{1}{m-1}$. Therefore, each dwarf obtains charge at least $\frac{m}{m-1} \geq \frac{11}{10}$. Thus, the final charge of dwarfs is at least $\frac{11}{10} D$. Hence, $\frac{11}{10} D \leq E$. 1.3. A table has $m$ rows and $n$ columns, where $m0$. Initially every vertex of degree at least 4 was non-negative and after rearrangement of charges it is still non-negative. Now consider a triangular face. Suppose the contrary the charge of every edge is at least 12 . Then the set of degrees of every face must be $(5,7+, 7+)$ or $(6+, 6+, 6+)^{2}$. It is easy to see in both cases the charge of every face is non-negative. 2.4. Let $\Gamma$ be a planar graph with $\delta(\Gamma) \geq 5$. Prove that there is an edge of weight at most 11 . Solution. See the subsection 'An easy example' on the webpage https://en.wikipedia. org/wiki/Discharging_method_(discrete_mathematics) 2.5. Let $\Gamma$ be a planar graph such that $\delta(\Gamma) \geq 3$. Prove that there are a face $f$ and a vertex $v \in f$ such that either $\operatorname{deg}(v)=3$ and $\operatorname{deg}(f) \leq 5$ or $\operatorname{deg}(v) \leq 5$ and $\operatorname{deg}(f)=3$. Solution. Multiplying the equality (1) for $\alpha=\frac{1}{2}$ and multiply, we have $$ \sum_{v \in V}(\operatorname{deg} v-4)+\sum_{f \in F}(\operatorname{deg} f-4)=-8 $$ Assign to a vertex $v$ a charge $\operatorname{deg} v-4$ and to a face $f$ a charge $\operatorname{deg} f-4$. Suppose the contrary that every face of degree 3 does not have vertices of degree less than 6 and every vertex of degree 3 is not incident to a face of degree at less than 6 . Then every vertex of degree at least 6 gives to every incident face a charge $\frac{1}{3}$. Therefore, the charge of such a vertex is non-negative and every triangular face has charge equal to 0 . Also, every face of degree at least 6 gives to its vertices the charge $\frac{1}{3}$. Therefore, the charges of such faces are non-negative and every triangular face has a charge equal to 0 . A contradiction. 2.6. Let $\Gamma$ be a connected planar graph such that $\delta(\Gamma) \geq 3$. Prove that there is a face of length at most 5 with the degrees of all its vertices but one do not exceed 11 . Solution. Multiplying the equality (1) for $\alpha=\frac{2}{3}$ by 3 , we have $$ \sum_{v \in V(\Gamma)}(2 \operatorname{deg} v-6)+\sum_{f \in F(\Gamma)}(\operatorname{deg} f-6)=-12 $$ Assign to a vertex $v$ a charge $2 \operatorname{deg} v-6$ and to a face $f$ charge $\operatorname{deg} f-6$. Suppose the contrary, that is, every face of degree at most 5 has at least two vertices of degree at least 12 . Let every vertex of degree $n$, where $n \geq 12$, give to every incident face a charge $\frac{3}{2}$. Then its charge becomes equal to $2 n-6-\frac{3}{2} n=\frac{1}{2} n-6 \geq 0$. Every face of degree at most 5 will obtain charge from at least two vertices, and so its charge is at least $-3+\frac{3}{2} \cdot 2=0$. All the rest vertices and faces initially had non-negative charges and these charges do not decrease. ## Light colorings 2.7. a)Prove that any planar graph is 6 -colorable. b) Suppose that, for any subgraph of a given graph $\Gamma$, one of the following conditions holds: 1) there is a vertex of degree at most $d-1 ; 2$ ) there is an even induced cycle such[^59]that the degree of every its vertex does not exceed $d$. Prove that the induced graph $\Gamma$ is $d$-colorable. Solution. a) Every face of a planar graph $\Gamma$ has at least 3 edges, thus, $3|F(\Gamma)| \leq$ $2|E(\Gamma)|$. Substituting this inequality in Euler's Formula, we obtain $$ 6=3|V(\Gamma)|-3|E(\Gamma)|+3|F(\Gamma)| \leq 3|V(\Gamma)|-|E(\Gamma)| $$ It means $3 V>E$, and so there is a vertex of degree at less than 6 . Now, it is easy to finish the proof using induction. We leave it as an exercise for the reader. b) Prove this statement for all subgraphs of $\Gamma$ by induction on the number $n$ of vertices. If in $\Gamma$ there is a vertex $v$ of degree at most $d-1$, then color $\Gamma \backslash\{v\}$ in $d$ colors (by induction) and then choose the proper color for $v$. Otherwise, in $\Gamma$ there is an even induced cycle $v_{1} \ldots v_{k}$ such that the degree of each vertex is at most $d$. Denote by $C$ the set of its vertices. By the induction hypothesis, we color $\Gamma \backslash C$ in a proper way. Now, let us show how we can color vertices of $C$. For every vertex in $C$, there is at most $d-2$ edges connecting it with vertices in $\Gamma \backslash C$, and thus for each vertex in $C$ there are at least 2 possible colors. Let us show the following Lemma. Lemma. If, to every vertex of an even cycle, a list of 2 colors is assigned, then it is possible to color it in proper way choosing colors from the corresponding list. If lists of the vertices of the cycle are the same, then, clearly, it is possible to color it alternatively. Otherwise, there are two neighbouring vertices with distinct lists, say, the vertices $v_{1}$ и $v_{k}$. Color $v_{1}$ in a color that is not in the list of $v_{k}$. Then there is a choice for coloring $v_{2}$, and so we can color it. Analogously, we can color $v_{3}, \ldots, v_{k}$. Thus, all vertices are colored in the proper way, including $v_{1}$ and $v_{k}$. 2.8. a)Prove that for a given $k$ there exists an integer $n$ such that the complete bipartite graph $K_{n, n}$ is not $k$-choosable. Although it is 2 -colorable as any bipartite graph. b) Prove that any even cycle is not only 2-colorable but also 2-choosable. Solution. a) Let the first part of the graph has vertices, indexed by numbers $1, \ldots, k$. Also, assume that the vertex $i$ has list $\left(c_{i 1}, c_{i 2}, \ldots, c_{i k}\right)$, here we assume that all these $k^{2}$ colors are distinct. In the second part there are $k^{k}$ vertices with all possible lists of the form $\left(c_{1 x_{1}}, c_{2 x_{2}}, \ldots, c_{k x_{k}}\right)$, where $x_{i} \in\{1, \ldots, k\}$ for all $i$. Then for any coloring of the first part there is a vertex of the second part, whose list coinside with the coloring of the first part, and therefore it is impossible to color it properly. Since we are going to construct an example with the same number of vertices in both parts, we add extra vertices to the first part (with arbitrary lists). b) See the lemma in the proof of Problem 2.7b. 2.9. a)Suppose that every inner face of a planar graph $\Gamma$ is a triangle and its outer face is bounded by a cycle $v_{1} \ldots v_{k} v_{1}$. Suppose further that with $v_{1}$ and $v_{2}$ lists of 2 colors are associated, with every other vertex of the outer face a list of 3 colours, and with every inner vertex a list of 5 colours. Then there exists a proper list coloring of $\Gamma$ for the given lists. b) Every planar graph is 5 -choosable. Solution. See the proof of Theorem 5.4.2 in [Die12]. 2.10. Consider a planar graph $\Gamma$ with $\Delta(\Gamma) \geq 11$. Assign lists of size $\Delta(\Gamma)$ to the vertices of $\Gamma$ and lists of size $\Delta(\Gamma)+2$ to the edges. Prove that $\Gamma$ has a proper total list coloring for this set of lists. Solution. See the proof of Theorem 6.3 (and the proof of Lemma 6.1) in [CW13]. ## 3 Main problems ### 3.1 Matchstick graphs Key Problem 1. Prove that a 5-regular matchstick graph does not exist. Note. Actually, in any matchstick graph, there is a vertex with degree less than 5 . 3.1.1. Let $\Gamma$ be a graph of minimal distances on $n$ vertices that are in general position ${ }^{3}$. Prove that a) $E(\Gamma)<5 n / 2, \mathbf{b})$ there exists a constant $c<5 / 2$ such that $E(\Gamma) \leq c n$. c) An interesting graph is a graph of minimal distances such that, for every vertex, it and its neighbors are in general position ${ }^{4}$. Prove that, for every $c<5 / 2$, there is an interesting graph $\Gamma$ with at least $c|V(\Gamma)|$ edges. Solution. See [Tót97]. 3.1.2. Solve Key Problem 1. Solution. See $[K P 11]$. 3.1.3*. Prove that a 4-regular matchstick graph contains at least 20 vertices. Solution. See Section 3 and Theorem 3.6 in [KP14]. ### 3.2 Quasi-planar graphs Key Problem 2. a) Prove that $|E(\Gamma)| \leq 8 n-20$ for a quasi-planar graph $\Gamma$ on $n$ vertices. b) Try to improve the statement of the previous problem. Both sharpening of the inequality and its generalizations for other classes of graphs are interesting. Solution. See $[$ AT07]. ### 3.3 List edge coloring ## Key Problem 3. Solution. See the proof of Theorem 12 in [Cra05]. ### 3.4 Magical configurations Key Problem 4. Describe all magical configurations. Solution. See $\left[\mathrm{ABK}^{+}\right.$09].[^60] ## Список литературы $\left[\mathrm{ABK}^{+}\right.$09] Eyal Ackerman, Kevin Buchin, Christian Knauer, Rom Pinchasi, and Günter Rote. There are not too many magic configurations. Twentieth Anniversary Volume: Discrete \& Computational Geometry, page 1, 2009. [AT07] Eyal Ackerman and Gábor Tardos. On the maximum number of edges in quasi-planar graphs. Journal of Combinatorial Theory, Series A, 114(3):563$571,2007$. [Cra05] Daniel W Cranston. Edge-choosability and total-choosability of planar graphs with no adjacent 3-cycles. arXiv preprint math/0512518, 2005. [CW13] Daniel W. Cranston and Douglas B. West. An introduction to the discharging method via graph coloring. 2013. [Die12] Reinhard Diestel. Graph theory, volume 173 of. Graduate texts in mathematics, page 7, 2012. [Fel12] Stefan Felsner. Geometric graphs and arrangements: some chapters from combinatorial geometry. Springer Science \& Business Media, 2012. [KP11] Sascha Kurz and Rom Pinchasi. Regular matchstick graphs. The American Mathematical Monthly, 118(3):264-267, 2011. [KP14] Sascha Kurz and Rom Pinchasi. Regular matchstick graphs. arXiv preprint arXiv:1401.4372, 2014. [KPT17] Andrey Kupavskii, János Pach, and Gábor Tardos. Tilings with noncongruent triangles, 2017. [LP07] Hagit Last and Rom Pinchasi. At least n-1 intersection points in a connected family of $\mathrm{n}$ unit circles in the plane. Discrete 8 Computational Geometry, $38(2): 321-354,2007$. [Mat09] Jiří Matoušek. Blocking visibility for points in general position. Discrete \& Computational Geometry, 42(2):219-223, 2009. [Tót97] Géza Tóth. The shortest distance among points in general position. Computational Geometry, 8(1):33-38, 1997. # Об инверсных образах точки Фейербаха, полюсах треугольника и теореме Куланина П. В. Бибиков, М. И. Бидва, В. Д. Попов, А. А. Шевцов
Проект представляют А. А. Заславский, О. А. Заславский, П. А. Кожевников,
И. И. Фролов ## Введение Точка Фейербаха является одной из наиболее известных замечательных точек треугольника. Она определяется как точка касания вписанной окружности и окружности Эйлера. Соответствующую теорему доказал Карл Вильгельм Фейербах в 1822 году. С тех пор было придумано свыше 300 других доказательств этой теоремы, причем многие из них используют инверсию. Многочисленность таких доказательств свидетельствует о том, что при изучении геометрических свойств точки Фейербаха инверсия является довольно полезным инструментом. В первой части проекта мы предлагаем новый способ изучения точки Фейербаха, основанный на рассмотрении ее инверсных образов относительно некоторых окружностей. Этот подход позволит нам не только получить ряд новых красивых результатов, но и значительно упростить доказательства некоторых уже известных фактов. Во второй части проекта будет предложен способ обобщить многие утверждения, полученные в первой части, на более общие конфигурации. Грубо говоря, мы утверждаем, что многие свойства точки Фейербаха являются лишь частными случаями значительно более общих теорем, связанных с совершенно иными конструкциями, ранее особо не привлекавшимися для ее исследования. Стартуя с одного красивого утверждения (так называемой теоремы Куланина), мы последовательно построим теорию, обобщающую различные свойства точки Фейербаха. Техника, требующаяся для решения задач данного проекта, весьма разнообразна, поэтому от решателей предполагается знакомство со следующими методами и конструкциями: - инверсия и ее свойства; - полюсы и поляры, гармонические четверки; - линейные движения точек; - сведения о кривых второго порядка (необязательно); - комплексные числа (необязательно). Также зафиксируем обозначения, которые будут использоваться на протяжении всего проекта. - $A, B, C$ - вершины треугольника; - $M_{a}, M_{b}, M_{c}$ - середины сторон $B C, A C, A B$ соответственно; - $H_{a}, H_{b}, H_{c}$ - основания высот треугольника $A B C$, опущенных из вершин $A, B, C$ соответственно; - $L_{a}, L_{b}, L_{c}$ - основания биссектрис треугольника $A B C$, проведенных из вершин $A, B, C$ соответственно; - $\omega$ - вписанная окружность треугольника $A B C$ с центром $I$; - $G_{a}, G_{b}, G_{c}$ - точки касания вписанной окружности треугольника $A B C$ со сторонами $B C$, $A C, A B$ - $G_{a}^{\prime}, G_{b}^{\prime}, G_{c}^{\prime}$ - точки касания вневписанных окружности треугольника $A B C$ со сторонами $B C, A C, A B$ - $K_{a}, K_{b}, K_{c}$ - точки, симметричные точкам $G_{a}, G_{b}, G_{c}$ относительно прямых $A I, B I, C I$ соответственно; - $\lambda_{a}$ - окружность, построенная на стороне $B C$ как на диаметре; - $S_{a b}, S_{a c}$ - точки Шарыгина, т.е. точки пересечения окружности $\lambda_{a}$ со средними линиями $M_{a} M_{b}$ и $M_{a} M_{c}$ соответственно; - $\varepsilon, \varepsilon_{a}$ - окружности девяти точек треугольника $A B C$ и треугольника $I B C$ соответственно; - $F$ - точка касания вписанной окружности и окружности девяти точек треугольника $A B C$ (точка Фейербаха). ## 0. Вспомогательные факты Задачи этого раздела содержат достаточно известные факты, и сдавать их необязательно. Тем не менее, если какие-то из задач вам неизвестны, рекомендуем их решить. Это поможет вам при решении последующих задач проекта. 0.1 (Окружность Эйлера) Докажите, что середины сторон треугольника, основания его высот и середины отрезков, соединяющих вершины с ортоцентром, лежат на одной окружности. 0.2 (Прямая Эйлера) Докажите, что центр $E$ этой окружности, точки $O, H$ и $M$ лежат на одной прямой, причем $H E: E M: M O=3: 1: 2$. 0.3 (Точка Микеля) Рассмотрим четыре прямые общего положения, которые образуют четыре треугольника. Докажите, что описанные окружности этих треугольников проходят через одну точку. 0.4 Докажите, что прямая Эйлера треугольника $G_{a} G_{b} G_{c}$ проходит через точку $O$. 0.5 (Прямая Симсона) Докажите, что основания перпендикуляров, опущенных из точки, лежащей на описанной окружности $\Omega$ треугольника $A B C$, на его стороны (или их продолжения), лежат на одной прямой. 0.6 Докажите, что прямая Симсона точки $P$ делит отрезок $P H$ пополам. 0.7 (Изогональное сопряжение) Рассмотрим треугольник $A B C$ и точку $P$, отличную от его вершин. Докажите, что прямые, симметричные прямым $A P, B P$ и $C P$ относительно биссектрис $A I, B I, C I$ соответственно, пересекаются в одной точке $P^{\prime}$. 0.8 Докажите, следующие свойства изогонального сопряжения: - точки $O$ и $H$ изогонально сопряжены; - педальные окружности (т.е. окружности, проходящие через основания перпендикуляров, опущенных из данных точек на стороны треугольника) изогонально сопряженных точек $P$ и $P^{\prime}$ совпадают, причем центр этой окружности лежит в середине отрезка $P P^{\prime}$; - точки, изогонально сопряженные точкам описанной окружности, являются бесконечно удаленными, причем они лежат в направлениях, перпендиклярных прямой Симсона данной точки; - точка $A$ изогонально сопряжена точкам прямой $B C$, отличным от $B$ и $C$. 0.9 Докажите, что если точки $P$ и $Q$ изогонально сопряжены, а точки $P_{b}, P_{c}, Q_{b}, Q_{c}-$ их проекции на стороны $A C$ и $A B$ соответственно, то прямые $P_{b} Q_{c}, P_{c} Q_{b}$ и $P Q$ пересекаются в одной точке. 0.10 Докажите, что коника, проходящая через точки $A, B, C$ и $H$, является равнобокой гиперболой. 0.11 Докажите, что изогональный образ прямой является коникой. ## 1. Теорема Фейербаха 1.1 (Задача 255) Докажите, что тройки прямых $\left(G_{b} G_{c}, B I, M_{a} M_{b}\right)$ и $\left(G_{b} G_{c}, C I, M_{a} M_{c}\right)$ пересекаются в точках $S_{a b}$ и $S_{a c}$. Эти точки лежат на окружности $\lambda_{a}$, построенной на стороне $B C$ как на диаметре 1.2 Докажите, что угол между прямой $B C$ и окружностью Эйлера $\varepsilon$ равен $|\angle B-\angle C|$. 1.3 Докажите, что четверка точек ( $H_{a}, L_{a}, G_{a}, G_{a}^{\prime}$ ) образует гармоническую четверку. 1.4 Докажите теорему Фейербаха, т.е. факт касания вписанной окружности $\omega$ и окружности Эйлера $\varepsilon$ 1.5 Обобщите доказательство из предыдущей задачи на случай вневписанных окружностей. Докажите, что треугольник с вершинами в точках касания окружности Эйлера с вневписанными окружностями перспективен треугольнику $L_{a} L_{b} L_{c}$ с центром в точке $F$. $1.6\left(*_{*}^{\text {* }}\right.$ Найдите другие замечательные окружности, касающиеся окружности Эйлера. Какими интересными свойствами обладают точки касания? 1.7 Докажите, что прямая $F G_{a}$ является биссектрисой угла $\angle H_{a} F M_{a}$. 1.8 Докажите, что для произвольного четырехвершинника (т.е. четверки точек общего положения, соединенными прямыми) окружности Эйлера возникающих треугольников с вершинами в вершинах четырехсторонника пересекаются в одной точке (точке Понселе). 1.9 Пусть $P$ - произвольная точка, отличная от $O$ и $H$. Докажите, что педальная окружность точки $P$ проходит через точку Понселе четырехсторонника $A B C P$. 1.10 Докажите, что треугольник $K_{a} K_{b} K_{c}$ гомотетичен треугольнику $M_{a} M_{b} M_{c}$ с центром в точке F 1.11 Докажите, что точка $F$ лежит на окружности Эйлера $\varepsilon_{a}$ треугольника $B I C$ 1.12 Пусть $\widetilde{G}_{a}$ - точка, диаметрально противоположная точке $G_{a}$ относительно вписанной окружности $\omega$. Докажите, что прямая $F \widetilde{G}_{a}$ пересекает отрезок $A I$ в его середине ## 2. Инверсные образы точки Фейербаха Теперь перейдем к содержательной части программы, которая, наконец, будет использовать инверсию. Рассмотрим уже знакомую нам окружность $\lambda_{a}$, построенную на отрезке $B C$ как на диаметре. Рассмотрим точку $F_{a}^{\prime}$ - образ точки $F$ при инверсии относительно окружности $\lambda_{a}$. Точка $F_{a}^{\prime}$ будет основным объектом наших дальнейших рассуждений в этом разделе. Также в этой серии задач будут активно использоваться точки $K_{a}, K_{b}$ и $K_{c}$, симметричные точкам $G_{a}, G_{b}$ и $G_{c}$ относительно биссектрис соответствующих углов. 2.1 Докажите, что точка $F_{a}^{\prime}$ является радикальным центром окружностей $\lambda_{a}, \varepsilon$ и $\varepsilon_{a}$ 2.2 Пусть $K_{b}^{\prime}$ и $K_{c}^{\prime}$ - точки пересечения лучей $K_{a} K_{b}$ и $K_{a} K_{c}$ с прямой $H_{b} H_{c}$. Докажите, что пятерки точек $\left(F, F_{a}^{\prime}, K_{b}, K_{b}^{\prime}, S_{a b}\right)$ и ( $F, F_{a}^{\prime}, K_{c}, K_{c}^{\prime}, S_{a c}$ ) лежат на окружностях $\psi_{a b}$ и $\psi_{a c}$ 2.3 Докажите, что прямая $M_{a} S_{a b}$ касается окружности $\psi_{a b}$. 2.4 Докажите, что прямая $S_{a b} S_{a c}$ является биссектрисой угла $\angle K_{b} F_{a}^{\prime} H_{b}$. 2.5 Точка $F_{a}^{\prime}$ лежит на прямых $K_{b} K_{c}, L_{b} L_{c}$ и $G_{b}^{\prime} G_{c}^{\prime}$. 2.6 Пусть $T_{a}$ - точка пересечения прямых $F G_{a}$ и $S_{a b} S_{a c}$. Тогда четверки точек $\left(F, T_{a}, M_{b}, S_{a b}\right)$ и $\left(F, T_{a}, M_{c}, S_{a c}\right)$ лежат на окружностях $\psi_{a b}^{\prime}$ и $\psi_{a c}^{\prime}$ 2.7 Докажите, что $F$ является точкой Микеля для треугольника $M_{a} M_{b} M_{c}$ и прямой $S_{a b} S_{a c}$. 2.8 Докажите, что точки $T_{a}, T_{b}$ и $T_{c}$ лежат на средних линиях $M_{b} M_{c}, M_{c} M_{a}$ и $M_{a} M_{b}$ соответственно 2.9 Рассмотрим произвольный треугольник $\triangle=P_{a} P_{b} P_{c}$, гомотетичный серединному треугольнику $M_{a} M_{b} M_{c}$ с центром в точке Фейербаха $F$. Рассмотрим окружности $\psi_{a b}^{\triangle}$ и $\psi_{a c}^{\triangle}$, проходящие через тройки точек $\left(F, P_{b}, S_{a b}\right.$ ) и ( $F, P_{c}, S_{a c}$ ) соответственно. Докажите, что точка их пересечения, отличная от $F$, совпадает с точкой пересечения $A^{\triangle}$ прямых $P_{b} P_{c}$ и $S_{a b} S_{a c}$ 2.10 Докажите, что в случае, когда точка $A^{\triangle}$ совпадает с точкой $G_{b}$, окружность $\psi_{a c}^{\triangle}$ проходит через точки $F, H_{b}, G_{b}, S_{a c}, S_{c a}$ и точку, симметричную $I$ относительно $S_{a c} S_{c a}$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-250.jpg?height=54&width=1784&top_left_y=2126&top_left_x=136) их пересечения с прямыми $F A^{\triangle}, F B^{\triangle}, F C^{\triangle}$, соответственно, пересекаются в одной точке, причем эта точка лежит на прямой $O I$. Более того, окружность, проходящая через эти точки пересечения, проходит через точку Фейербаха $F$. При каком положении треугольника $\triangle$ конфигурация из задачи 2.9 переходит в конфигурации из задач 2.2 и 2.6 ? Последняя задача является очень сложной. Следующий раздел будет посвящен конструкции, позволяющей не только решить ее, но и обобщить на другие аналогичные конфигурации. ## 3. Обобщенные полюсы треугольника и теорема Куланина Целью этого раздела является обобщение понятия точки Фейербаха и доказательство представленных выше утверждений (в частности, задачи 2.11 ) в более общем случае. Мы будем дополнительно использовать следующие обозначения. - $O$ - центр описанной окружности $\Omega$ треугольника $A B C$; - $\ell$ - произвольная прямая, проходящая через точку $O$; - $A_{\ell}, B_{\ell}, C_{\ell}$ - точки пересечения прямой $\ell$ с прямыми $B C, C A$ и $A B$ соответственно; - $P_{a} P_{b} P_{c}$ - педальный треугольник точки $P$ относительно треугольника $A B C$, а $\Omega_{P}$ - его педальная окружность; - $F_{\ell}$ - обобщенная точка Фейербаха (см. теорему Куланина). Для того, чтобы сформулировать необходимые понятия, нам потребуется следующая теорема. В дальнейшем ее можно использовать без доказательства. Теорема 1 (Куланин). Пусть $P$ - произвольная точка прямой $\ell$, проходящей через точку $O$. Тогда все педальные окружности $\Omega_{P}$ проходят через одну точку. 3.1 (*) Докажите эту теорему. Общую точку всех педальных окружностей $\Omega_{P}$ точек прямой $\ell$ мы будем обозначать через $F_{\ell}$ и называть обобщенной точкой Фейербаха. Эта точка зависит от прямой $\ell$, проходящей через центр описанной окружности $O$ треугольника $A B C$. В случае $\ell=O I$ точка $F_{\ell}$ совпадает с обычной точкой Фейербаха $F$. Отметим также, что все точки $F_{\ell}$ лежат на окружности Эйлера $\varepsilon$ треугольника $A B C$ (поскольку окружность Эйлера является педальной окружностью точки $O)$. 3.2 Рассмотрим произвольный треугольник $A B C$, прямую $\ell$, проходящую через $O$, и точку $P$ на ней. Обозначим через $A_{p p}$ точку пересечения прямых $P_{b} P_{c}$ и $M_{b} M_{c}$. Тогда точки $P_{a}, A_{p p}$ и $F_{\ell}$ лежат на одной прямой. 3.3 ($\left.^{*}\right)$ (Основная теорема) Рассмотрим произвольную прямую $\ell$, проходящую через центр $O$ описанной окружности треугольника $A B C$. Пусть $P$ и $Q$ - произвольные точки на этой прямой. Обозначим через $A_{p q}$ точку пересечения прямых $P_{b} P_{c}$ и $F_{\ell} Q_{a}$. Аналогично определим точки $B_{p q}$ и $C_{p q}$. Возьмем произвольную точку $R$ на прямой $\ell$ и проведем через точки $A_{p q}, B_{p q}$, $C_{p q}$ прямые, параллельные $R_{b} R_{c}, R_{c} R_{a}, R_{a} R_{b}$ соответственно. Тогда получившийся треугольник будет гомотетичен треугольнику $R_{a} R_{b} R_{c}$ с центром в точке $F_{\ell}$. 3.4 Выведите из основной теоремы результаты задач $2.5,2.8,2.11$. 3.5 Докажите, что описанная окружность треугольника, образованного прямыми $A_{p q} A_{q p}, B_{p q} B_{q p}$, $C_{p q} C_{q p}$, проходит через точку $F_{\ell}$. 3.6 Докажите, что прямые, симметричные прямой $\ell$ относительно средних линий треугольника $A B C$, пересекаются в точке $F_{\ell}$. 3.7 Пусть $A_{\ell}, B_{\ell}$ и $C_{\ell}$ точки пересечения прямой $\ell$ с прямыми $B C, C A$ и $A B$ соответственно. Докажите, что окружности, построенные на отрезках $A A_{\ell}, B B_{\ell}, C C_{\ell}$ как на диаметрах, пересекаются в двух точках: $F_{\ell}$ и $F_{\ell}^{\prime}$, причем точка $F_{\ell}^{\prime}$ лежит на описанной окружности $\Omega$ треугольника $A B C$, а ортоцентр $H$ этого треугольника принадлежит прямой $F_{\ell} F_{\ell}^{\prime}$. 3.8 Пусть $O H$ - прямая Эйлера треугольника $A B C$. Докажите, что полюсы $A_{h h}, B_{h h}, C_{h h}$ образуют автополярный треугольник относительно окружности Эйлера $\varepsilon$, причем вершины треугольника $A B C$ лежат на сторонах треугольника $A_{h h} B_{h h} C_{h h}$. 3.9 Пусть треугольники $P_{a} P_{b} P_{c}$ и $Q_{a} Q_{b} Q_{c}$ перспективны треугольнику $A B C$, а их вершины лежат на соответствующих сторонах треугольника $A B C$. Пусть $A_{p q}, B_{p q}, C_{p q}$ - точки пересечения соответствующих сторон треугольников $P_{a} P_{b} P_{c}$ и $Q_{a} Q_{b} Q_{c}$. Тогда: a) Вершины треугольника $A B C$ лежат на сторонах треугольника $A_{p q} B_{p q} C_{p q}$; b) Треугольник $A_{p q} B_{p q} C_{p q}$ автополярен относительно коники $\Omega$, проходящей через точки $P_{a}$, $P_{b}, P_{c}, Q_{a}, Q_{b}, Q_{c}$ c) Треугольник $A_{p q} B_{p q} C_{p q}$ перспективен треугольникам $P_{a} P_{b} P_{c}$ и $Q_{a} Q_{b} Q_{c}$, причем центры перспективы лежат на конике $\Omega$. 3.10 Пусть $X_{a}$ - произвольная точка на прямой $B C$. Обозначим через $X_{b}$ точку пересечения прямых $X_{a} B_{h h}$ и $A C$, а через $X_{c}$ - точку пересечения прямых $X_{a} C_{h h}$ и $A B$. Докажите, что - прямая $X_{b} X_{c}$ проходит через точку $A_{h h}$; - прямые $A X_{a}, B X_{b}$ и $C X_{c}$ пересекаются в одной точке $X$; - описанная окружность треугольника $X_{a} X_{b} X_{c}$ проходит через точку $F_{O L}$, где $L$ - точка Лемуана треугольника $A B C$. В заключение сформулируем ряд вопросов, ответы на которые получить пока не удалось. 3.11 Определение полюсов $A_{p q}, B_{p q}, C_{p q}$ зависит от порядка выбора точек $P$ и $Q$ на прямой $\ell$. Возникает вопрос: как связаны полюсы $A_{p q}$ и $A_{q p}$, т.е. полюсы, соответствующие одним и тем же точкам, взятым в разном порядке? Имеет место следующая гипотеза. Точки пересечения $Z_{a}, Z_{b}$, $Z_{c}$ соответствуюших сторон треуголвников $A_{p q} B_{p q} C_{p q}$ и $A_{q p} B_{q p} C_{q p}$ образуют треугольник, гомотетичный треугольнику $A B C$ с центром в точке $F_{\ell}$. 3.12 Существенный интерес представляет такой выбор точек $P$ и $Q$ на прямой $\ell$, при котором вершины треугольника $A B C$ лежат на сторонах треугольника $A_{p q} B_{p q} C_{p q}$. Всегда ли на прямой $\ell$ можно выбрать такую пару точек? Эксперименты показывают, что такие точки существуют всегда, однако они не могут располагаться слишком далеко от точки $O$. Как можно точно описать положения точек $P$ и $Q$, при которых вершины треугольника $A B C$ лежат на сторонах треугольника $A_{p q} B_{p q} C_{p q}$ ? ## Об инверсных образах точки Фейербаха, полюсах треугольника и теореме Куланина Решения ## 0. Вспомогательные факты 0.1 Поскольку $\angle B H C=180^{\circ}-\angle B A C$, то точки, симметричные $H$ относительно $B C$ и середины $B C$, лежат на описанной окружности треугольника. При гомотетии с центром $H$ и коэффициентом $1 / 2$ описанная окружность переходит в окружность Эйлера. 0.2 Серединные перпендикуляры исходного треугольника являются высотами серединного, а значит $O$ - ортоцентр серединного треугольника. Гомотетия с центром $M$ и коэффициентом $-1 / 2$ переводит исходный треугольник в серединный, и $H$ переходит в $O$. Отсюда $M$ лежит на отрезке $H O$ и $H M: M O=2$. Из решения предыдущей задачи $E$ лежит на отрезке $H O$ и $H E: E O=1: 1$. Равенство $H E: E M: M O=3: 1: 2$ следует из двух доказанных. 0.3 Пусть три из прямых образуют треугольник $A B C$ и пусть прямые $A B, A C, B C$ пересекают четвертую прямую в точках $D, E, F$ соответственно. Обозначим через $P$ точку пересечения описанных окружностей треугольников $A B C$ и $C E F$, отличную от точки $C$. Докажем, что $P$ принадлежит описанной окружности треугольника $B D F$. Для этого достаточно проверить, что $\angle(B P, P F)=\angle(B D, D F)$. Ясно, что $\angle(B P, P F)=\angle(B P, P C)+\angle(P C, P F)=$ $\angle(B A, A C)+\angle(E C, E F)=\angle(B D, A C)+\angle(A C, D F)=\angle(B D, D F)$. Аналогично доказывается, что $P$ принадлежит описанной окружности треугольника $A D E$. 0.4 Пусть $I_{a}, I_{b}, I_{c}$ - центры вневписанных окружностей треугольника $A B C$. Заметим, что $I_{a} A$, $I_{b} B, I_{c} C$ - высоты треугольника $I_{a} I_{b} I_{c}$. Стороны треугольников $I_{a} I_{b} I_{c}$ и $G_{a} G_{b} G_{c}$ параллельны, т.е. эти треугольники гомотетичны. Поэтому их прямые Эйлера параллельны. Точка $I$ является центром описанной окружности треугольника $G_{a} G_{b} G_{c}$ и ортоцентром треугольника $I_{a} I_{b} I_{c}$. Поэтому прямые Эйлера этих треугольников проходят через $I$ и параллельны, т.е. совпадают. Точка $O$ является центром окружности Эйлера треугольника $I_{a} I_{b} I_{c}$, а значит лежит на его прямой Эйлера. ## 0.5 Следует из решения следующей задачи (или из легкого счета углов). 0.6 Сделаем гомотетию с центром $P$ и коэффициентом 2. Пусть $P_{a}^{\prime}, P_{b}^{\prime}, P_{c}^{\prime}$ и $H_{a}^{\prime}, H_{b}^{\prime}, H_{c}^{\prime}$ - точки, симметричные $P$ и $H$ относительно соответствующих сторон треугольника. Достаточно доказать, что $P_{b}^{\prime}, H, P_{c}^{\prime}$ лежат на одной прямой (доказательство того, что $P_{a}^{\prime}, H, P_{c}^{\prime}$ лежат на одной прямой, аналогично). Из симметрии относительно $A B$ имеем $\angle\left(P_{c}^{\prime} H, H B\right)=\angle\left(B H_{c}^{\prime}, H_{c}^{\prime} P\right)$ и $\angle\left(C H, H P_{b}^{\prime}\right)=\angle\left(P H_{b}^{\prime}, H_{b}^{\prime} C\right)$. Используя то, что $P, H_{b}^{\prime}, H_{c}^{\prime}$ лежат на описанной окружности, получаем $\angle\left(P_{c}^{\prime} H, H P_{b}^{\prime}\right)=\angle\left(P_{c}^{\prime} H, H B\right)+\angle(H B, C H)+\angle\left(C H, H P_{b}^{\prime}\right)=\angle\left(B H_{c}^{\prime}, H_{c}^{\prime} P\right)+\angle(H B, C H)+$ $\angle\left(P H_{b}^{\prime}, H_{b}^{\prime} C\right)=\angle(B A, A P)+\angle(H B, C H)+\angle(P A, A C)=\angle(B A, A C)+\angle(H B, C H)=0$, т.е. $P_{b}^{\prime}, H, P_{c}^{\prime}$ лежат на одной прямой. 0.7 Пусть $P_{a}^{\prime}, P_{b}^{\prime}, P_{c}^{\prime}$ - точки, симметричные $P$ относительно соответствующих сторон треугольника. Несложно проверить, что прямая, симметричная $A P$ относительно $A I-$ серединный перпендикуляр к $P_{b}^{\prime} P_{c}^{\prime}$. Поэтому искомая точка $P^{\prime}$ - центр описанной окружности треугольника $P_{a}^{\prime} P_{b}^{\prime} P_{c}^{\prime}$. 0.8 Докажем свойства изогонального сопряжения: - Следует из решения предыдущей задачи и того, что точки, симметричные $H$ относительно сторон треугольника, лежат на описанной окружности. - Применяя к гомотетию с центром $P$ и коэффиентом $1 / 2$ к утверждению из решения предыдущей задачи, получим, что середина $P P^{\prime}$ является центром педальной окружности точки $P$. Поскольку расстояния от середины $P P^{\prime}$ до проекций $P$ и $P^{\prime}$ на любую прямую равны, то педальная окружность точки $P^{\prime}$ совпадает с педальной окружностью точки $P$. - Из предыдущей задачи точка, изогонально сопряженная $P$, должна быть пересечением серединных перпендикуляров к $P_{a}^{\prime} P_{b}^{\prime}, P_{b}^{\prime} P_{c}^{\prime}, P_{c}^{\prime} P_{a}^{\prime}$. Но все эти серединные перпендикуляры перпендикулярны прямой Симсона точки $P$. - Ясно 0.9 По предыдущей задаче точки $P_{b}, P_{c}, Q_{b}, Q_{c}$ лежат на окружности с центром в середине $P Q$. Пусть точки $X$ и $Y$ симметричны $P_{b}$ и $P_{c}$ соответственно относительно центра окружности. Тогда $X$ лежит на $Q Q_{b}$, а $Y$ лежит на $Q Q_{c}$. По теореме Паскаля для шестиугольника $P_{b} Q_{c} Y P_{c} Q_{b} X$ середина $P Q$, точка $Q$ и точка пересечения $P_{b} Q_{c}$ с $P_{c} Q_{b}$ лежат на одной прямой. Значит, прямые $P_{b} Q_{c}, P_{c} Q_{b}, P Q$ пересекаются в одной точке. 0.10 Поскольку $A B C H$ - невыпуклый четырехугольник, то коника, проходящая через его вершины является гиперболой. Пусть $P, Q$ - ее бесконечно удаленные точки. Нужно показать, что напрвления на $P$ и $Q$ перпендикулярны. Запишем теорему Паскаля для шестиугольника $A B C H P Q$. Пусть $X$ - точка пересечения $A B$ и $H P$, а $Y$ - точка пересечения $C H$ и $Q A$. Тогда $X Y$ проходит через точку пересечения $B C$ и $P Q$, т.е. $X Y \| B C$, т.е. $X Y \perp A H$. Значит, $Y-$ ортоцентр треугольника $A H X$, откуда $A Y \perp H X$. Т.е. $A Q \perp H P$, что и требовалось. 0.11 Пусть дана прямая $\ell$. Выберем на ней три точки $P, Q, R$. Проведем через их изогональные образы $P^{\prime}, Q^{\prime}, R^{\prime}$ и вершины $B, C$ конику. Рассмотрим два преобразования прямой $\ell$ на конику, сохраняющих двойные отношения. Первое сопоставляет точке $X$ прямой $\ell$ прямую $B X$, затем отражает эту прямую относительно биссектрисы угла $B$, и пересекает полученную прямую с коникой. Второе преобразование аналогично, только вершина $B$ заменена на $C$. Эти преобразования совпадают в трех точках $P, Q, R$, а значит, совпадают всюду. Поэтому изогональный образ $\ell$ - коника, проходящая через вершины $B$ и $C$. Ясно, что она также проходит через $A$. ## 1. Теорема Фейербаха 1.1 Т.к. $S_{a b}$ - точка пересечения $\lambda_{a}$ со средней линией $M_{a} M_{b}$, то треугольник $B M_{a} S_{a b}$ равнобедренный, и $\angle S_{a b} B M_{a}=\frac{1}{2} \angle S_{a b} M_{a} C=\frac{1}{2} \angle A B C=\angle I B C$. Значит, $S_{a b}$ лежит на $B I$. Поскольку $B S_{a b} C$ прямой, то точки $I, G_{b}, S_{a b}, C$ лежат на одной окружности (с диаметром $C I)$. Значит, $\angle S_{a b} G_{b} C=\angle S_{a b} I C=90^{\circ}-\frac{1}{2} \angle B A C=\angle A G_{b} C$. Поэтому $S_{a b}$ лежит на прямой $G_{b} G_{c}$ (см. рис. 1). 1.2 Искомый угол между прямой $H_{a} M_{a}$ и окружностью $\varepsilon$ равен половине дуги $H_{a} M_{a}$ окружности $\varepsilon$, т.е. равен $\angle H_{a} M_{b} M_{a}=\left|\angle C M_{a} M_{b}-\angle C H_{a} M_{b}\right|=\left|\angle C B A-\angle H_{a} C M_{b}\right|=|\angle B-\angle C|$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-255.jpg?height=1092&width=1355&top_left_y=796&top_left_x=358) Рис. 1. 1.3 Пусть $I_{a}$ - центр вневписанной окружности, касающейся отрезка $B C$. Тогда $B I$ и $B I_{a}-$ биссектрисы внутреннего и внешнего угла $A B L_{a}$. Поэтому четверка точек $A, I, L_{a}, I_{a}$ - гармоническая. Ортогонально проектируя на $B C$, получаем утверждение задачи. 1.4 Из предыдущей задачи следует, что при инверсии относительно окружности с диаметром $G_{a} G_{a}^{\prime}$ окружность девяти точек $\varepsilon$ переходит в прямую $\ell$, проходящую через $L_{a}$. С другой стороны, эта инверсия оставляет вписанную и вневписанную окружности $\omega$ и $\omega_{a}$ на месте, т.к. обе они ортогональны окружности инверсии. Так как инверсия - конформное преобразование, угол между прямыми $\ell$ и $B C$ равен углу между окружностью $\omega$ и $B C$, т.е. $|\angle B-\angle C|$. Прямая, симметрична прямой $B C$ относительно биссектрисы $A I$ образует такой же угол с $B C$. Поэтому $\ell$ симметрична прямой $B C$ относительно биссектрисы $A I$, а значит, $\ell$ касается окружностей $\omega$ и $\omega_{a}$ (см. рис. 2). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-256.jpg?height=1017&width=1309&top_left_y=806&top_left_x=381) Рис. 2. 1.5 Касание окружности Эйлера и вневписанной окружности $\omega_{a}$ в некоторой точке $F_{a}$ следует из решения предыдущей задачи. Точка $L_{a}$ - центр отрицательной гомотетии окружностей $\omega$ и $\omega_{a}$. Поэтому перспективность следует из теоремы о трех центрах гомотетии для вписанной, вневписанной окружностей и окружности Эйлера. 1.7 Сделаем гомотетию с центром $F$, переводящую вписанную окружность в окружность Эйлера. Прямая $H_{a} M_{a}$ перейдет в прямую, параллельную $H_{a} M_{a}$ и касающуюся окружности Эйлера в некоторой точке $X$. Из симметрии относительно серединного перпендикуляра к $H_{a} M_{a}$ точка $X$ - середина дуги $H_{a} M_{a}$ окружности Эйлера. Значит, $F X$ является биссектрисой угла $\angle H_{a} F M_{a}$ Утверждение задачи следует из того, что при гомотетии $G_{a}$ переходит в $X$. 1.8 Будем обозначать середину отрезка $X Y$ через $M_{X Y}$. Пусть даны точки $A, B, C, P$. Пусть $X$ - точка пересечения окружностей Эйлера треугольников $A B P$ и $B C P$, отличная от $M_{B P}$. Тогда $\angle\left(M_{A P} X, X M_{C P}\right)=\angle\left(M_{A P} X, X M_{B P}\right)+\angle\left(M_{B P} X, X M_{C P}\right)=\angle\left(M_{A P} M_{A B}, M_{A B} M_{B P}\right)+$ $\angle\left(M_{B P} M_{B C}, M_{B C} M_{C P}\right)=\angle(B P, A P)+\angle(C P, B P)=\angle(C P, A P)=\angle\left(M_{A P} M_{A C}, M_{A C} M_{C P}\right)$. Т.е. $X$ лежит на окружности Эйлера треугольника $A C P$. Аналогично, $X$ лежит на окружности Эйлера треугольника $A B C$ см. рис. 3 ). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-257.jpg?height=764&width=1244&top_left_y=434&top_left_x=164) Рис. 3. 1.9 Пусть $P_{a}, P_{b}, P_{c}$ - проекции $P$ на соответствующие стороны треугольника $A B C$. В обозначениях из решения предыдущей задачи $\angle\left(P_{c} X, X P_{a}\right)=\angle\left(P_{c} X, X M_{B P}\right)+\angle\left(M_{B P} X, X P_{a}\right)=\angle\left(P_{c} M_{A B}, M_{A B} M_{B P}\right)+\angle\left(M_{B P} M_{B C}, M_{B C} P_{a}\right)=$ $\angle\left(P_{c} A, A P\right)+\angle\left(P C, C P_{a}\right)=\angle\left(P_{c} P_{b}, P_{b} P\right)+\angle\left(P P_{b}, P_{b} P_{a}\right)=\angle\left(P_{c} P_{b}, P_{b} P_{a}\right)$. 1.10 Сначала докажем, что соответствующие стороны треугольников $K_{a} K_{b} K_{c}$ и $M_{a} M_{b} M_{c}$ параллельны. Хорды $G_{a} K_{b}$ и $G_{b} G_{c}$ симметричны относительно $B I$, поэтому они равны. Аналогично, хорды $G_{b} G_{c}$ и $G_{a} K_{c}$ симметричны относительно $C I$ и потому также равны. Тогда треугольник $K_{c} G_{a} K_{b}$ равнобедренный, откуда следует, что $\angle G_{a} K_{c} K_{b}=\angle G_{a} K_{b} K_{c}=\angle C G_{a} K_{b}$. Отсюда очевидна параллельность прямых $B C$ и $K_{b} K_{c}$. Для двух оставшихся пар сторон рассуждения аналогичны. По теореме Дезарга треугольники $K_{a} K_{b} K_{c}$ и $M_{a} M_{b} M_{c}$ гомотетичны. Точка пересечения прямых $M_{a} K_{a}, M_{b} K_{b}$ и $M_{c} K_{c}$ является центром этой гомотетии, а значит и центром гомотетии вписанной окружности и окружности Эйлера, то есть точкой Фейербаха $F$ (см. рис. 4). 1.11 Следует из теоремы Фейербаха и задач $1.8,1.9$ для четырехугольника $A B C I$ (см. рис. 5). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-258.jpg?height=813&width=1550&top_left_y=320&top_left_x=141) Рис. 4. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-258.jpg?height=632&width=786&top_left_y=1668&top_left_x=713) Рис. 5. 1.12 Из предыдущей задачи точка Фейербаха $F$ лежит на окружности Эйлера треугольника $A I C$. Середина $X$ отрезка $A I$ там тоже лежит. Достаточно доказать, что $\angle\left(G_{b} F, F X\right)=$ $\angle\left(G_{b} F, F \widetilde{G}_{a}\right)$. Имеем $\angle\left(G_{b} F, F X\right)=\angle\left(G_{b} M_{b}, M_{b} X\right)=\angle\left(G_{b} C, C I\right)=\angle\left(G_{b} G_{a}, G_{a} I\right)=\angle\left(G_{b} G_{a}, G_{a} \widetilde{G}_{a}\right)=\angle\left(G_{b} F, F \widetilde{G}_{a}\right)$ (см. рис. 6). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-259.jpg?height=618&width=792&top_left_y=558&top_left_x=710) Рис. 6. ## 2. Инверсные образы точки Фейербаха 2.1 При инверсии относительно $\lambda_{a}$ окружность Эйлера $\varepsilon$ перейдет в прямую $H_{b} H_{c}$, окружность $\varepsilon_{a}$ перейдет в прямую $S_{a b} S_{a c}$. Поскольку точка $F$ является точкой пересечения окружностей $\varepsilon$ и $\varepsilon_{a}$, то точка $F_{a}^{\prime}$ является точкой пересечения прямых $H_{b} H_{c}$ и $S_{a b} S_{a c}$, которые являются радикальными осями $\lambda_{a}$ и $\varepsilon, \lambda_{a}$ и $\varepsilon_{a}$ (см. рис. 7). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-260.jpg?height=683&width=897&top_left_y=707&top_left_x=796) Рис. 7. 2.2 Сначала докажем, что четыре точки $F, F_{a}^{\prime}, K_{b}, K_{b}^{\prime}$ лежат на одной окружности (рассуждения для четверки точек $F, F_{a}^{\prime}, K_{c}, K_{c}^{\prime}$ аналогичны). Рассмотрим инверсию $\operatorname{Inv}_{K_{a}}$ с центром в точке $K_{a}$, переводящую точку $F_{a}^{\prime}$ в точку $F$. Тогда композиция инверсий $\operatorname{Inv}_{K_{a}} \circ \operatorname{Inv}_{\lambda_{a}}$ переводит окружность Эйлера $\varepsilon$ в некоторую окружность $\omega^{\prime}$, проходящую через точки $F$ и $K_{a}$. В силу того, что инверсия - конформное преобразование, эта композиция сохраняет углы между прямыми и окружностями. Раз центры обеих инверсий лежат на прямой $F F_{a}^{\prime}$, то окружности $\varepsilon$ и $\omega^{\prime}$ образуют с прямой $F F_{a}^{\prime}$ равные углы. Но это означает, что окружности $\varepsilon$ и $\omega^{\prime}$ касаются в точке $F$. Отсюда следует, что окружности $\omega$ и $\omega^{\prime}$ совпадают, поскольку существует единственная окружность, проходящая через точку $K_{a}$ и касающаяся окружности $\varepsilon$ в точке $F$. Далее, заметим, что точка $K_{b}^{\prime}$ пересечения луча $K_{a} K_{b}$ с прямой $H_{b} H_{c}$ - это образ точки $K_{b}$ при инверсии $\operatorname{Inv}_{K_{a}}$. Значит, точки $F, F_{a}^{\prime}, K_{b}$ и $K_{b}^{\prime}$ лежат на одной окружности. Осталось загнать на окружности точки $S_{a b}$ и $S_{a c}$. Для этого используем вписанные углы. Boпервых, $\angle K_{a} F K_{b}=\angle K_{a} K_{c} K_{b}$. Во-вторых, пусть прямая $F K_{b}$ вторично пересекает окружность $\varepsilon_{a}$ в точке $R$. Тогда $\angle M_{a} S_{a c} R=\angle M_{a} F R$. Прямые $M_{a} S_{a c}$ и $K_{a} K_{c}$ параллельны, поэтому вторые лучи равных углов $\angle M_{a} S_{a c} R$ и $\angle M_{a} F R$ также параллельны, т.е. $S_{a c} R\left\|K_{b} K_{c}\right\| B C$. Отсюда сразу следует равенство углов $\angle M_{a} F R$ и $\angle S_{a c} S_{a b} G_{a}$. Значит, точки $F, F_{a}^{\prime}, K_{b}, K_{b}^{\prime}$ и $S_{a b}$ лежат на одной окружности. Для точки $S_{a c}$ рассуждения аналогичны (см. рис. 8). 2.3 Из решения предыдущей задачи окружности $\psi_{a b}$ и $\lambda_{a}$ ортогональны. Поэтому прямая $M_{a} S_{a b}$ касается окружности $\psi_{a b}$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-261.jpg?height=1390&width=1640&top_left_y=644&top_left_x=210) Рис. 8. 2.4 Поскольку прямые $M_{a} S_{a b}$ и $M_{a} S_{a c}$ касаются окружностей $\psi_{a b}$ и $\psi_{a c}$ соответственно, и пары прямых $\left(M_{a} S_{a b}, K_{a} K_{b}\right)$ и ( $\left.M_{a} S_{a c}, K_{a} K_{c}\right)$ параллельны, то $K_{b} S_{a b}=K_{b}^{\prime} S_{a b}$ и $K_{c} S_{a c}=K_{c}^{\prime} S_{a c}$ (параллельные прямые высекают на окружности равные хорды). Тогда $\angle K_{b} F_{a}^{\prime} S_{a b}=\angle K_{b}^{\prime} F_{a}^{\prime} S_{a b}$, что и требовалось. 2.5 Точка $F_{a}^{\prime}$ лежит на $K_{b} K_{c}$ следует из предыдущей задачи. Заметим, что прямые $L_{b} K_{b}$ и $L_{c} K_{c}$ касаются вписанной окружности в точках $K_{b}$ и $K_{c}$ соответственно. Следующее утверждение является вырожденным случаем теоремы Брианшона: Пусть четырехугольник $K_{c} G_{b} K_{b} G_{c}$ вписан в окружность, касательные к которой в вершинах четырехугольника образуют другой четырехугольник. Тогда у этих двух четырехугольников точки пересечения диагоналей совпадают. Применяя его, получаем, что прямые $K_{b} K_{c}, G_{b} G_{c}, L_{b} L_{c}$ пересекаются в одной точке. Прямая $G_{b}^{\prime} G_{c}^{\prime}$ проходит через эту точку из задачи 1.3. 2.6 Приведем доказательство для окружности $\psi_{a b}^{\prime}$, для окружности $\psi_{a c}^{\prime}$ рассуждения аналогичны. $\angle M_{a} F M_{b}=\angle M_{a} M_{c} M_{b}=\angle A C B=\angle G_{a} S_{a b} T_{a}$ (последнее равенство следует из того, что точки $C, G_{a}, G_{b}, S_{a b}$ лежат на окружности с диаметром $C I$ ). По задаче $1.11 \angle G_{a} F M_{a}=\angle G_{a} S_{a b} M_{a}$. Тогда $\angle T_{a} F M_{b}=\angle M_{a} F M_{b}+\angle G_{a} F M_{a}=\angle G_{a} S_{a b} T_{a}+\angle G_{a} S_{a b} M_{a}=\angle T_{a} S_{a b} M_{b}$ (см. рис. 9). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-262.jpg?height=629&width=789&top_left_y=1163&top_left_x=709) Рис. 9. 2.7 Точка $F$ лежит на описанной окружности треугольников $M_{a} M_{b} M_{c}$ и $S_{a b} M_{a} S_{a c}$. 2.8 Следует из предыдущих двух задач (см. рис. 10). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-263.jpg?height=632&width=794&top_left_y=313&top_left_x=709) Рис. 10. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-263.jpg?height=1052&width=1827&top_left_y=1339&top_left_x=295) Рис. 11. 2.9 Используя окружность $\psi_{a b}^{\prime}$ из задачи 2.6 имеем: $\angle A^{\triangle} P_{b} F=\angle T_{a} M_{b} F=\angle T_{a} S_{a b} F=\angle A^{\triangle} S_{a b} F$. Отсюда следует, что точка $A^{\triangle}$ лежит на окружности $\psi_{a b}^{\triangle}$. Аналогично доказывается, что точка $A^{\triangle}$ лежит на окружности $\psi_{a c}^{\triangle}$ (см. рис. 11). 2.10 Вначале докажем, что четверка точек ( $\left.H_{b}, G_{b}, S_{a c}, S_{c a}\right)$ лежит на одной окружности. Заметим, что прямые $B S_{c a}$ и $H_{b} S_{a c}$ параллельны, т.к. обе они перпендикулярны биссектрисе угла $\angle B A C$. Аналогично, прямые $B S_{a c}$ и $G_{b} S_{c a}$. Поэтому четырехугольник $B S_{a c} H_{b} S_{c a}$ является параллелограммом и $H_{b} S_{a c}=B S_{c a}=G_{b} S_{c a}$. Значит, $S_{a c} H_{b} G_{b} S_{c a}$ - равнобокая трапеция, и потому вокруг нее можно описать окружность (см. рис. 8). Теперь докажем, что точка Фейербаха также лежит на этой окружности. По лемме Архимеда следует, что $\angle H_{b} F G_{b}=\frac{1}{2} \angle H_{b} F M_{b}=\frac{1}{2}|\angle A-\angle C|$. Далее, $\angle H_{b} S_{a c} S_{c a}=\angle S_{c a} G_{b} C=\frac{\pi}{2}-\frac{\angle C}{2}$ и $\angle G_{b} S_{a c} S_{c a}=\angle A G_{b} S_{a c}=\frac{\pi}{2}-\frac{\angle A}{2}$, поэтому $$ \angle H_{b} S_{a c} G_{b}=\left|\angle H_{b} S_{a c} S_{c a}-\angle G_{b} S_{a c} S_{c a}\right|=\frac{1}{2}|\angle A-\angle C|=\angle H_{b} F G_{b} $$ откуда следует, что точка $F$ лежит на описанной окружности трапеции $H_{b} G_{b} S_{c a} S_{a c}$. Несложно видеть, что точка $I$ является ортоцентром треугольника $S_{a c} S_{c a} G_{b}$. Значит, точка, симметричная $I$ относительно $S_{a c} S_{c a}$ лежит на описанной окружности треугольника $S_{a c} S_{c a} G_{b}$. 2.11 Будем двигать треугольник $P_{a} P_{b} P_{c}$ линейно. Точки $A^{\triangle}, B^{\triangle}, C^{\triangle}$ двигаются дробно-линейно. Точка $X_{a}$ пересечения $F A^{\triangle}$ с $B C$ (и аналогично определяемые точки $X_{b}, X_{c}$ ) двигаются дробнолинейно. Точки пересечения перпендикуляров к сторонам треугольника в точках $A^{\triangle}, B^{\triangle}, C^{\triangle}$ с OI будет двигаться дробно-линейно. Чтобы доказать, что они все совпадают, достаточно 3 положений. Два положения $P_{a} P_{b} P_{c}=K_{a} K_{b} K_{c}$ и $P_{a} P_{b} P_{c}=M_{a} M_{b} M_{c}$ следуют из предыдущих задач. Третье положение - когда треугольник $P_{a} P_{b} P_{c}$ вырождается в 3 прямые, проходящие через $F$ параллельно сторонам исходного треугольника. В этом положении все перпендикуляры уходят на бесконечность. Утверждение про окружность будет следовать из задачи 3.1. Или же можно вывести все сразу из основной теоремы (см. задачу 3.4). ## 3. Обобщенные полюсы треугольника и теорема Куланина 3.1 Из задач 0.11 и 0.10 (и первого пункта 0.8 ) точка $P^{\prime}$, изогонально сопряженная точке $P$, движется по равнобокой гиперболе, проходящей через вершины треугольника $A B C$. По второму пункту задачи 0.8 достаточно доказать, что педальные окружности точки $P^{\prime}$ проходят через центр этой гиперболы. Используя задачи 1.9 и 1.8 , сводим к следующему хорошо известному утверждению: если треугольник вписан в равнобокую гиперболу, то его окружность Эйлера проходит через центр гиперболы (см. рис. 12). 3.2 Пусть точка $P$ двигается линейно по прямой $\ell$. Тогда точки $P_{a}, P_{b}, P_{c}$ движутся линейно по сторонам треугольника. Значит, точка $A_{o p}$ пересечения $F_{\ell} P_{a}$ со средней линией $M_{b} M_{c}$, двигается линейно по $M_{b} M_{c}$. Нужно доказать, что точки $P_{b}, P_{c}, A_{o p}$ лежат на одной прямой. Достаточно проверить три положения точки $P$. Случай $P=O$ очевиден. Два случая, когда $P$ лежит на описанной окружности, следуют из теоремы Куланина (педальная окружность точки $P$ вырождается в прямую Симсона) (см. рис. 13). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-265.jpg?height=680&width=1634&top_left_y=381&top_left_x=251) Рис. 12. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-265.jpg?height=708&width=1641&top_left_y=1608&top_left_x=250) Рис. 13. 3.3 Будем линейно двигать точку $P$ по прямой $\ell$. Обозначим через $t$ ориентированную длину отрезка $O P$ (т.е. выберем положительное направление на прямой $\ell$ и будем считать длину $t$ отрезка $O P$ положительной, если точка $P$ находится на положительном луче, и отрицательной в противном случае). Введем декартову систему координат с началом в точке $F_{\ell}$ и обозначим через $\left(X_{a}(t), Y_{a}(t)\right),\left(X_{b}(t), Y_{b}(t)\right),\left(X_{c}(t), Y_{c}(t)\right)$ координаты точек $A_{p q}, B_{p q}, C_{p q}$ соответственно. Докажем, что каждая из этих координатных функций является рациональной по переменной $t$, причем степень числителей равна 2 , а степень знаменателей равна 1 . Для определенности рассмотрим точку $A_{p q}$ (рассуждения для точек $B_{p q}$ и $C_{p q}$ аналогичны). Ясно, что точки $P_{b}, P_{c}$ будут двигаться линейно, поэтому их координаты являются линейными функциями от переменной $t$. Далее, запишем явный вид уравнения прямой, проходящей через точки $\left(x_{1}, y_{1}\right)$ и $\left(x_{2}, y_{2}\right)$ : $$ \left(y-y_{1}\right)\left(x_{2}-x_{1}\right)=\left(x-x_{1}\right)\left(y_{2}-y_{1}\right) $$ Отсюда следует, что коэффициенты при $x$ и $y$ уравнения прямой $P_{b} P_{c}$ также линейно зависят от $t$, а свободный член этого уравнения является квадратичным по $t$. Наконец, пересекая прямую $P_{b} P_{c}$ с неподвижной прямой $F_{\ell} Q_{a}$, получаем искомые координаты точки $A_{p q}$ пересечения этих прямых. Аналогично доказывается, что координаты точек $B_{p q}$ и $C_{p q}$ имеют аналогичный вид. Теперь докажем, что при линейном движении точки $P$ прямые $A_{p q} B_{p q}, B_{p q} C_{p q}, C_{p q} A_{p q}$ будут двигаться параллельно (но не линейно). Это условие равносильно равенствам $$ F_{\ell} A_{p q}=\alpha \cdot F_{\ell} C_{p q} \quad \text { и } \quad F_{\ell} B_{p q}=\beta \cdot F_{\ell} C_{p q} $$ где $\alpha$ и $\beta$ - некоторые константы, не зависящие от $t$ (например, можно вычислить их для некоторого конкретного момента времени $t_{0}$, а затем доказать, что они остаются неизменными в любой другой момент времени). Будем доказывать только первое равенство, поскольку второе доказывается абсолютно аналогично. Чтобы доказать истинность первого равенства, перепишем его в координатном виде: $$ X_{a}(t)=\alpha_{1} \cdot X_{c}(t) \quad \text { и } \quad Y_{a}(t)=\alpha_{2} \cdot Y_{c}(t) $$ (здесь $\alpha_{1}$ и $\alpha_{2}$ - фиксированные константы). Вспомним, что каждая из функций, фигурирующих в этих соотношениях, является отношением многочленов степеней 2 и 1. Поэтому, домножая каждое из равенств на знаменатели соответствующих функций, получаем, что данные равенства равносильны обращению в 0 двух кубических многочленов, зависящих от $t$. Ясно, что для проверки этого условия необходимо найти четыре различных значения $t$, или, что то же самое, четыре различных положения точки $P$, при которые утверждение теоремы (равносильное занулению этих кубических многочленов) верно (поскольку если у многочлена степени не выше 3 есть 4 различных корня, то этот многочлен тождественно равен 0). Укажем эти положения. Во-первых, рассмотрим два момента времени, в которые точка $P$ совпадает с $O$ (этот случай соответствует $t=0$ ) и с $Q$. Из предыдущей задачи следует, что точки $\left\{A_{p q}, C_{p q}\right\}$ совпадают в эти два момента времени, причем они отличны от начала координат $F_{\ell}$. Поэтому, если определить константы $\alpha_{1}$ и $\alpha_{2}$ по формулам $\alpha_{1}=X_{a}(0) / X_{c}(0)$ и $\alpha_{2}=Y_{a}(0) / Y_{c}(0)$, то равенства $$ X_{a}(t)=\alpha_{1} \cdot X_{c}(t) \quad \text { и } \quad Y_{a}(t)=\alpha_{2} \cdot Y_{c}(t) $$ верны при двух моментах времени $t$. Во-вторых, рассмотрим два положения точки $P$, являющихся концами диаметра описанной окружности $\Omega$ треугольника $A B C$, содержащегося на прямой $\ell$. В это случае точки $A_{p q}$ и $C_{p q}$ совпадают с $F_{\ell}$, а потому их координаты равны 0 и требуемые равенства также верны. Таким образом, мы нашли четыре момента времени, при которых соотношения между координатами точек $A_{p q}$ и $C_{p q}$ верны. Значит, эти соотношения верны в любой момент времени $t$, откуда следует параллельность прямых вида $A_{p q} C_{p q}$. Аналогично доказывается параллельность прямых $A_{p q} B_{p q}$ и $B_{p q} C_{p q}$. Таким образом, все треугольники вида $A_{p q} B_{p q} C_{p q}$ гомотетичны с центром в точке $F_{\ell}$ при всевозможных положениях точки $P$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-267.jpg?height=735&width=1748&top_left_y=397&top_left_x=259) Рис. 14. Осталось понять, почему при проведении через точки $A_{p q}, B_{p q}, C_{p q}$ прямых, параллельных $R_{b} R_{c}, R_{c} R_{a}, R_{a} R_{b}$ соответственно, треугольник, образованный этими прямыми, будет гомотетичен треугольнику $R_{a} R_{b} R_{c}$ с центром в точке $F_{\ell}$. Для этого рассмотрим момент времени, в который точка $P$ совпадает с $R$. Тогда точки $A_{p q}=A_{r q}, B_{p q}=B_{r q}$ и $C_{p q}=C_{r q}$ будут лежать на прямых $R_{b} R_{c}, R_{c} R_{a}$ и $R_{a} R_{b}$ соответственно. Для произвольного положения точки $P$ рассмотрим гомотетию с центром в $F_{\ell}$, которая переведет $A_{r q}$ в $A_{p q}$. Тогда по доказанному выше эта же гомотетия переведет $B_{r q}$ в $B_{p q}$ и $C_{r q}$ в $C_{p q}$. Поэтому прямая $R_{b} R_{c}$ перейдет в прямую, проходящую через $A_{p q}$ параллельно ей. Аналогичное произойдет с двумя оставшимися прямыми $R_{c} R_{a}$ и $R_{a} R_{b}$ (см. рис. 14 ). 3.4 Задачи 2.5 и 2.8 несложно доказываются и без использования основной теоремы, поэтому мы только упомянем, что конструкция задачи 2.5 - случай $P=I, Q=O$ в основной теореме, а конструкция задачи 2.8 - случай $P=Q=I, R=O$. Выведем задачу 2.11. Пусть перпендикуляр к прямой $B C$ в точке ее пересечения с $F A^{\triangle}$ пересекает $O I$ в точке $Q$. Прямая $P_{b} P_{c}$ проходит через точку $A^{\triangle}=A_{i q}$. Применим основную теорему к $I, Q, R=O$. Получим, что прямые $P_{a} P_{b}$ и $P_{a} P_{c}$ проходят через точки $C_{i q}$ и $B_{i q}$ соответственно. Следовательно, $B^{\triangle}=B_{i q}$ и $C^{\triangle}=C_{i q}$. Поэтому основание перпендикуляров из $Q$ на $A B$ и $A C$ - точки пересечения $A B$ и $A C$ с $F C^{\triangle}$ и $F B^{\triangle}$ соответственно. Утверждение про окружность следует из теоремы Куланина. (см. рис. $15,16,17$ ). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-268.jpg?height=691&width=1651&top_left_y=698&top_left_x=234) Рис. 15. Случай $P=Q=I, R=O$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-268.jpg?height=694&width=1640&top_left_y=1678&top_left_x=248) Рис. 16. Случай $P=I, Q=R=O$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-269.jpg?height=681&width=1634&top_left_y=207&top_left_x=251) Рис. 17. Случай $P=I, R=O$ 3.5 Основная теорема равносильно равенству $F A_{r q} / F A_{p q}=F B_{r q} / F B_{p q}=F C_{r q} / F C_{p q}$. Аналогично, выполнено равенство $F A_{r p} / F A_{q p}=F B_{r p} / F B_{q p}=F C_{r p} / F C_{q p}$. Пусть точка $R$ такова, что $R_{b} R_{c} \| A_{p q} A_{q p}$. Тогда существует гомотетия с центром $F$, переводящая $R_{b} R_{c}$ в $A_{p q} A_{q p}$. Эта гомотетия переводит $A_{r q}$ в $A_{p q}$ и $A_{r p}$ в $A_{q p}$. Из приведенных выше равенств следует, что эта гомотетия переводит треугольник $R_{a} R_{b} R_{c}$ в треугольник, образованный прямыми $A_{p q} A_{q p}, B_{p q} B_{q p}$, $C_{p q} C_{q p}$, откуда следует утверждение задачи. 3.6 Хорошо известно, что прямые, симметричные $\ell$ относительно средних линий треугольника пересекаются в некоторой точке $X$. При этом точка $P$, симметричная $X$ относительно $M_{b} M_{c}$, лежит на прямой $\ell$ и на окружности $\left(A M_{b} M_{c}\right)$. (Если прямая проходит через ортоцентр треугольника, то прямые, симметричные ей относительно сторон, пересекаются на описанной окружности). Используя задачу 3.2 и прямую Симсона точки $P$ относительно треугольника $A M_{b} M_{c}$, получаем $P F_{\ell} \perp B C$. Ясно, что $P X \perp B C$. Аналогичные утверждения верны для точек, симметричных $X$ относительно других средних линий. Значит, $F_{\ell}$ совпадает с $X$ (см.рис. 18). 3.7 Точка $F_{\ell}$ лежит на этих окружностях по теореме Куланина для случая, когда $P$ совпадает с $A_{\ell}, B_{\ell}, C_{\ell}$. Пусть $A^{\prime}, B^{\prime}, C^{\prime}$ - точки, симметричные $A, B, C$ относительно $O$. По теореме Паскаля точка пересечения $A^{\prime} A_{\ell}$ и $B^{\prime} B_{\ell}$ лежит на описанной окружности треугольника $A B C$. Значит, прямые $A^{\prime} A_{\ell}, B^{\prime} B_{\ell}, C^{\prime} C_{\ell}$ пересекаются в некоторой точке $F_{\ell}^{\prime}$ на описанной окружности. Т.к. $\angle A_{\ell} F_{\ell}^{\prime} A=\angle A^{\prime} F_{\ell}^{\prime} A=90^{\circ}$, то точка $F_{\ell}^{\prime}$ лежит на нужных окружностях. Точка $H$ лежит на $F_{\ell} F_{\ell}^{\prime}$, поскольку степени $H$ относительно окружностей равны $H H_{a} \cdot H A=H H_{b} \cdot H B=H H_{c} \cdot H C$ (см. рис. 19). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-270.jpg?height=629&width=1470&top_left_y=398&top_left_x=295) Рис. 18. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-270.jpg?height=623&width=1480&top_left_y=1648&top_left_x=298) Рис. 19. 3.8 Рассмотрим четырехугольник $H_{a} H_{b} H_{c} F_{\ell}$, вписанный в окружность Эйлера $\varepsilon$. Его противоположные стороны $H_{a} H_{c}$ и $F_{\ell} H_{b}$ пересекаются в точке $B_{h h}$, а стороны $H_{a} H_{b}$ и $F_{\ell} H_{c}$ - в точке $C_{h h}$, в то время как диагонали $H_{a} F_{\ell}$ и $H_{b} H_{c}$ пересекаются в точке $A_{h h}$. Отсюда следует, что треугольник $A_{h h} B_{h h} C_{h h}$ является автополярным относительно окружности Эйлера $\varepsilon$. Теперь докажем, что стороны этого треугольника содержат вершины треугольника $A B C$. В самом деле, рассмотрим четырехугольник $H_{c} M_{c} H_{b} M_{b}$, также вписанный в окружность $\varepsilon$. Его диагонали пересекаются в точке $A_{h h}$, а боковые стороны - в точке $A$, поэтому точка $A$ лежит на поляре точки $A_{h h}$. Но полярой точки $A_{h h}$ является прямая $B_{h h} C_{h h}$, поэтому точка $A$ лежит на $B_{h h} C_{h h}$. Рассуждения для остальных вершин абсолютно аналогичны (см. рис. 20). Кроме того, эта задача является частным случаем следующей. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-271.jpg?height=702&width=1659&top_left_y=749&top_left_x=252) Рис. 20. 3.9 а) Двойное отношение прямых $P_{a} B, P_{a} P_{c}, P_{a} A, P_{a} P_{b}$ равно -1. Пересекая с прямой $A B_{p q}$ получаем, что точки $A B_{p q} \cap B C, B_{p q}, A, A B_{p q} \cap P_{a} P_{b}$ образуют гармоническую четверку. Аналогично, точки $A B_{p q} \cap B C, B_{p q}, A, A B_{p q} \cap Q_{a} Q_{b}$ образуют гармоническую четверку. Поэтому $A B_{p q}, P_{a} P_{b}, Q_{a} Q_{b}$ пересекаются в одной точке, т.е. $A$ лежит на прямой $B_{p q} C_{p q}$. $\mathrm{b}, \mathrm{c})$ По теореме Паппа точка пересечения $P_{b} Q_{c}$ с $P_{c} Q_{b}$ лежит на $P Q$. Аналогично для точек пересечения $P_{a} Q_{c}$ с $P_{c} Q_{a}$ и $P_{a} Q_{b}$ с $P_{b} Q_{a}$. По теореме Паскаля шестиугольник $P_{a} Q_{b} P_{c} Q_{a} P_{b} Q_{c}$ вписан в конику $\Omega$. Пусть $F$ - точка пересечения $B_{p q} P_{b}$ с $C_{p q} P_{c}$. По теореме Паскаля и пункту а) шестиугольник $F P_{b} Q_{b} Q_{a} Q_{c} P_{c}$ вписан в конику $\Omega$. Аналогичное утверждение верно для точки пересечения $A_{p q} P_{a}$ с $C_{p q} P_{c}$, откуда прямые $A_{p q} P_{a}, B_{p q} P_{b}, C_{p q} P_{c}$ пересекаются на конике $\Omega$. Это доказывает пункт c). Теперь пункт b) доказывается аналогично предыдущей задаче. 3.10 Пусть точка $X_{a}$ движется дробно-линейно по $B C$. Тогда $X_{b}$ и $X_{c}$ тоже двигаются дробнолинейно по $A C$ и $A B$ соответственно. - достаточно доказать для трех положений точки $X_{a}$. Случаи $X_{a}=B, C, M_{a}$ следуют из предыдущей задачи; - Точка $X$ пересечения $B X_{b}$ и $C X_{c}$ двигается дробно-линейно по некоторой конике, проходящей через $B$ и $C$ (доказательство аналогично решению задачи 0.11). Эта коника проходит через $A$, т.к. $X$ совпадает с $A$, когда $X_{a}$ - точка пересечения $B C$ с $B_{h h} C_{h h}$. Поэтому достаточно доказывать, что $A X$ проходит через $X_{a}$ для трех положений точки $X_{a}$. Одно положение $-X_{a}=X=B$, второе $-X_{a}=X=C$, третье $-X_{a}=M_{a}, X=M$ (и четвертое $-X_{a}=H_{a}, X=H$ ). - Точка $X$ движется по равнобокой гиперболе $A B C H M$. Известно, что описанная окружность треугольника $X_{a} X_{b} X_{c}$ проходит через центр этой гиперболы (простого доказательства этого факта нам не известно; см. книгу А.Акопян, А.Заславский "Геометрические свойства кривых второго порядка", Теорема 4.3). Центр этой гиперболы совпадает с обобщенной точкой Фейербаха для прямой, изогонально сопряженной данной гиперболе, т.е. с точкой $F_{O L}$ (см. рис. 21). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-272.jpg?height=724&width=1678&top_left_y=760&top_left_x=234) Рис. 21. 3.11 Приведем план доказательства того, что треугольники $Z_{a} Z_{b} Z_{c}$ гомотетичны (информации про центр гомотетии не будет). Нужно доказать, что $B C, Z_{b} Z_{c}, A_{p p} A_{q q}$ параллельны (используем задачу 3.2: $A_{p p} A_{q q}$ - средняя линия). Сводим к проективному утверждению: Даны треугольники $P_{a} P_{b} P_{c}$ и $Q_{a} Q_{b} Q_{c}$ и точка $F$. Пусть $A_{p p}$ - точка пересечения $P_{b} P_{c}$ с $F P_{a}$, точки $A_{p q}, B_{q p}, \ldots$ определяются аналогично (как в задаче 3.3). Пусть $Z_{b}$ и $Z_{c}$ - точки пересечения $A_{p q} C_{p q}$ с $A_{q p} C_{q p}$ и $A_{p q} B_{p q}$ с $A_{q p} B_{q p}$. Тогда прямые $P_{a} P_{b}, A_{p p} A_{q q}$ и $Z_{b} Z_{c}$ пересекаются в одной точке. Двойные отношения прямых $A_{q p} Q_{c}, A_{q p} B_{q p}, A_{q p} F, A_{q p} Q_{a}$ и прямых $A_{p q} P_{b}, A_{p q} C_{p q}, A_{p q} F$, $A_{p q} P_{a}$ равны - проецируем из $A_{q p}$ на прямую $Q_{c} Q_{a}$, потом из $F$ на прямую $P_{b} P_{a}$, потом из $A_{p q}$ (и меняем порядок, не меняя двойное отношение). Аналогично, двойные отношения прямых $A_{q p} Q_{c}$, $A_{q p} C_{q p}, A_{q p} F, A_{q p} Q_{a}$ и прямых $A_{p q} P_{b}, A_{p q} B_{p q}, A_{p q} F, A_{p q} P_{a}$ равны. Поэтому можно применить следующий факт: Пусть $f$ - сохраняющее двойные отношения преобразование, сопоставляющее каждой прямой $l$ через точку $A_{q p}$ прямую $f(l)$ через точку $A_{p q}$. Пусть $m$ и $n$ - любые две прямые через $A_{q p}$. Рассмотрим прямую, соединяющую точку пересечения $m$ и $f(n)$ с точкой пересечения $n$ и $f(m)$. Тогда эта прямая проходит через фиксированную точку, не зависящую от выбора прямых $m$ и $n$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-273.jpg?height=715&width=1718&top_left_y=984&top_left_x=239) Рис. 22. # Inversions of the Feuerbach point, the poles of triangle and the Kulanin theorem P. V. Bibikov, M. I. Bidva, V. D. Popov, A. A. Shevtsov The project is presented by P. A. Kozhevnikov, I.I. Frolov, A. A. Zaslavsky, O. A. Zaslavsky ## Introduction The Feuerbach point is one of well known triangle centers. It is defined as a tangency point of the incircle of triangle and its nine-points circle. The corresponding theorem was proved by Karl Wilgelm Feuerbach in 1822. After this more than 300 another proofs of this theorem were found, many of them use the inversion. The plurality of such proofs shows that the inversion is very useful instrument for analysis of properties of the Feuerbach point. In the first part of the project we propose new method of analysis of the Feuerbach point based on the considering of its reflections about several circles. This approach allows us to obtain several new nice results and to simplify the proofs of some known facts. In the second part of the project we propose a method extending the obtaining results to more general configurations. We can say that several properties of the Feuerbach point are partial cases of general theorems, concerning absolutely distinct constructions which were not be used for its study. Starting from one nice assertion (the Kulanin theorem) we construct a theory generalizing different properties of the Feuerbach point. The technic for solving the problems of the project is very manifold, thus we suppose that the following methods and constructions are known to the participants: - inversion and its properties; - poles and polars, harmonic quadruples; - linear moving of points; - properties of conics (not necessary); - complex numbers (not necessary). Fix the notations which will be used in all problems of the project. - $A, B, C$ - the vertices of triangle; - $M_{a}, M_{b}, M_{c}$ - the midpoints of segments $B C, A C, A B$ respectively; - $H_{a}, H_{b}, H_{c}$ - the feet of altitudes of triangle $A B C$ from $A, B, C$ respectively; - $L_{a}, L_{b}, L_{c}$ - the feet of bisectors of triangle $A B C$ from $A, B, C$ respectively; - $\omega$ - the incircle of triangle $A B C$ centered at $I$; - $G_{a}, G_{b}, G_{c}$ - the tangency points of the incircle of triangle $A B C$ with sides $B C, A C, A B$ respectively; - $G_{a}^{\prime}, G_{b}^{\prime}, G_{c}^{\prime}$ - the tangency points of the excircles of triangle $A B C$ with sides $B C, A C, A B$ respectively; - $K_{a}, K_{b}, K_{c}$ - the reflections of $G_{a}, G_{b}, G_{c}$ about lines $A I, B I, C I$ respectively; - $\lambda_{a}$ - the circle with diameter $B C$; - $S_{a b}, S_{a c}$ - the Sharygin points, i.e. the common points of $\lambda_{a}$ with medial lines $M_{a} M_{b}$ and $M_{a} M_{c}$ respectively; - $\varepsilon, \varepsilon_{a}$ - the nine-points circles of triangles $A B C$ and $I B C$ respectively; - $F$ - the tangency point of the incircle and the nine-points circle of triangle $A B C$ (the Feuerbach point). ## 0. Auxiliary facts The problems of this section contain known facts and it is not necessary to present their solutions. But we recommend to solve the problems unknown to you. This will help to solve the remaining problems of the project. $\mathbf{0 . 1}$ (The nine-points circle) Prove that the midpoints of sides of triangle, the feet of its altitudes, and tyhe midpoints of segments joining its vertices with the orthocenter are concyclic. $\mathbf{0 . 2}$ (The Euler line) Prove that the center $E$ of the nine-points circle, the circumcenter $O$, the orthocenter $H$, and the centroid $M$ are collinear, and $H E: E M: M O=3: 1: 2$. $\mathbf{0 . 3}$ (The Mickel point) Consider four lines in general position forming four triangles. Prove that the circumcircles of these triangles have a common point. 0.4 Prove that the Euler line of triangle $G_{a} G_{b} G_{c}$ passes through $O$. 0.5 (The Simson line) Prove that the projections of a point lying on the circumcircle $\Omega$ of triangle $A B C$ to its sidelines are collinear. 0.6 Prove that the Simson line of point $P$ bisects segment $P H$. 0.7 (Isogonal conjugation) Consider a triangle $A B C$ and a point $P$ distinct from its vertices. Prove that the reflections of lines $A P, B P$, and $C P$ about the bisectors $A I, B I, C I$ respectively concur at a point $P^{\prime}$. 0.8 Prove the following properties of isogonal conjugation: - the points $O$ and $H$ are isogonally conjugated; - the pedal circles (i.e. the circles passing passing through the projections of the given points to the sidelines of triangle) of isogonally conjugated points $P$ and $P^{\prime}$ coincide, and this circle is centered at the midpoint of segment $P P^{\prime}$; - the point isiogonally conjugated to a point $P$ of the circumcircle is the infinite point of the perpendicular to the Simson line of $P$; - the point $A$ is isogonally conjugate to points of line $B C$ distinct from $B$ and $C$. 0.9 Let $P$ and $Q$ be isogonally conjugated, and $P_{b}, P_{c}, Q_{b}, Q_{c}$ be their projections to $A C$ and $A B$ respectively. Prove that $P_{b} Q_{c}, P_{c} Q_{b}$, and $P Q$ concur. 0.10 Prove that any conic passing through $A, B, C$, and $H$ is an equilateral hyperbola. 0.11 Prove that the isogonal image of a line is a conic. ## 1. The Feuerbach theorem 1.1 Prove that the triplets of lines $\left(G_{b} G_{c}, B I, M_{a} M_{b}\right)$ and $\left(G_{b} G_{c}, C I, M_{a} M_{c}\right)$ concur at points $S_{a b}$ and $S_{a c}$ repectively. These points lie on the circle $\lambda_{a}$ with diameter $B C$. 1.2 Prove that the angle between the line $B C$ and the nine-points circle $\varepsilon$ equals to $|\angle B-\angle C|$. 1.3 Prove that the quadruple $\left(H_{a}, L_{a}, G_{a}, G_{a}^{\prime}\right)$ is harmonic. 1.4 Prove the Feuerbach theorem, i.e. the tangency of the incircle $\omega$ and the nine-points circle $\varepsilon$ 1.5 Generalize the proof from the previous problem to the case of the excircles. Prove that the triangle formed by the tangency points of the nine-points circle with the excircles is perspective to the triangle $L_{a} L_{b} L_{c}$ with center at $F$. $\left.1.6{ }^{*}\right)$ Find another remarkable circles tangent to the nine-points circle. Which interesting properties of tangency points can be noted? 1.7 Prove that the line $F G_{a}$ bisects angle $H_{a} F M_{a}$. 1.8 Prove that the nine-points circles of four triangles formed by arbitrary four points in general position have a common point (the Poncelet point). 1.9 Let $P$ be an arbitrary point distinct from $O$ and $H$. Prove that the pedal circle of $P$ passes through the Poncelet point of $A B C P$. 1.10 Prove that the triangles $K_{a} K_{b} K_{c}$ and $M_{a} M_{b} M_{c}$ are homothetic with center $F$. 1.11 Prove that $F$ lies on the nine-points circle $\varepsilon_{a}$ of triangle $B I C$. 1.12 Let $\widetilde{G}_{a}$ be the point of $\omega$ opposite to $G_{a}$. Prove that the line $F \widetilde{G}_{a}$ bisects segment $A I$. ## 2. Inversion images of the Feuerbach point Now pass to the substantial part using the inversion. Consider again the circle $\lambda_{a}$ with diameter $B C$. Let $F_{a}^{\prime}$ be the reflection of $F$ about $\lambda_{a}$. The point $F_{a}^{\prime}$ will be the main object of studying in this section. Also we will actively use in the problems of this part the points $K_{a}, K_{b}$, and $K_{c}$ symmetric to $G_{a}, G_{b}$, and $G_{c}$ with respect to the bisectors of the corresponding angles. 2.1 Prove that $F_{a}^{\prime}$ is the radical center of circles $\lambda_{a}, \varepsilon$, and $\varepsilon_{a}$. 2.2 Let $K_{b}^{\prime}$ and $K_{c}^{\prime}$ be the common points of rays $K_{a} K_{b}$ and $K_{a} K_{c}$ respectively with the line $H_{b} H_{c}$. Prove that the fives of points $\left(F, F_{a}^{\prime}, K_{b}, K_{b}^{\prime}, S_{a b}\right)$ and $\left(F, F_{a}^{\prime}, K_{c}, K_{c}^{\prime}, S_{a c}\right)$ are concyclic on circles $\psi_{a b}$ and $\psi_{a c}$ respectively. 2.3 Prove that the line $M_{a} S_{a b}$ touches the circle $\psi_{a b}$. 2.4 Prove that the line $S_{a b} S_{a c}$ bisects the angle $K_{b} F_{a}^{\prime} H_{b}$. 2.5 Prove that $F_{a}^{\prime}$ is the common point of lines $K_{b} K_{c}, L_{b} L_{c}$, and $G_{b}^{\prime} G_{c}^{\prime}$. 2.6 Let $T_{a}$ be the common point of lines $F G_{a}$ and $S_{a b} S_{a c}$. Then the quadruples $\left(F, T_{a}, M_{b}, S_{a b}\right.$ ) and $\left(F, T_{a}, M_{c}, S_{a c}\right)$ are concyclic on circles $\psi_{a b}^{\prime}$ and $\psi_{a c}^{\prime}$ respectively. 2.7 Prove that $F$ is the Mickel point for the triangle $M_{a} M_{b} M_{c}$ and the line $S_{a b} S_{a c}$. 2.8 Prove that $T_{a}, T_{b}$, and $T_{c}$ lie on medial lines $M_{b} M_{c}, M_{c} M_{a}$, and $M_{a} M_{b}$ respectively. 2.9 Consider an arbitrary triangle $\triangle=P_{a} P_{b} P_{c}$ homothetic to the triangle $M_{a} M_{b} M_{c}$ with center at $F$. Consider circles $\psi_{a b}^{\triangle}$ and $\psi_{a c}^{\triangle}$ passing through $\left(F, P_{b}, S_{a b}\right)$ and $\left(F, P_{c}, S_{a c}\right)$ respectively. Prove that their common point distinct from $F$ coincides with the common point $A^{\triangle}$ of lines $P_{b} P_{c}$ and $S_{a b} S_{a c}$. 2.10 Prove that when $A^{\triangle}$ coincides with $G_{b}$ the circle $\psi_{a c}^{\triangle}$ passes through $F, H_{b}, G_{b}, S_{a c}, S_{c a}$, and the reflection of $I$ about $S_{a c} S_{c a}$. 2.11 (*) Prove that the perpendiculars to lines $B C, C A$, and $A B$ from their common points with $F A^{\triangle}, F B^{\triangle}, F C^{\triangle}$ respectively concur at point lying on $O I$. Also the circle passing through these common points passes through $F$. For which triangle $\triangle$ the configuration of problem 2.9 passes to the configurations of problems $\mathbf{2 . 2}$ and 2.6 ? The last problem is very difficult. The following section is dedicated to the construction allowing not only to solve this problem but to generalize it to similar configurations. ## 3. General poles of triangle and the Kulanin theorem In this section we generalize the notion of Feuerbach point and prove the assertions presented above (partially in problem $\mathbf{2 . 1 1}$ ) for general cases. We will use next additional notations. - $O$ - the center of circumcircle $\Omega$ of triangle $A B C$; - $\ell$ - an arbitrary line passing through $O$; - $A_{\ell}, B_{\ell}, C_{\ell}$ - the common points of $\ell$ with $B C, C A$, and $A B$ respectively; - $P_{a} P_{b} P_{c}$ - the pedal triangle of $P$ with respect to triangle $A B C$, and $\Omega_{P}$ - its pedal circle; - $F_{\ell}$ - general Feuerbach point (see. the Kulanin theorem). To formulate the necessary notions we use the following theorem. Now it can be used without proof. Theorem 1 (Kulanin). Let $P$ be an arbitrary point of line $\ell$ passing through $O$. Then all pedal circles $\Omega_{P}$ have a common point. $3.1\left({ }^{*}\right)$ Prove this theorem. Denote by $F_{\ell}$ the common point of pedal circles $\Omega_{P}$ of points lying on $\ell$. We will call $F_{\ell}$ the general Feuerbach point. This point depends on line $\ell$ passing through the circumcenter $O$ of triangle $A B C$. In the case $\ell=O I$ the point $F_{\ell}$ coincides with the Feuerbach point $F$. Note also that all points $F_{\ell}$ lie on the nine-points circle $\varepsilon$ of triangle $A B C$ (because the nine-points circle is the pedal circle of $O)$. 3.2 Consider a triangle $A B C$, a line $\ell$ passing through $O$, and a point $P$ on it. Denote by $A_{p p}$ the common point of lines $P_{b} P_{c}$ and $M_{b} M_{c}$. Prove that $P_{a}, A_{p p}$, and $F_{\ell}$ are collinear. $3.3\left(^{*}\right)$ (Main theorem) Consider an arbitrary line $\ell$ passing through the circumcenter $O$ of triangle $A B C$. Let $P$ and $Q$ be arbitrary points of this line. Note as $A_{p q}$ the common point of lines $P_{b} P_{c}$ and $F_{\ell} Q_{a}$. Define points $B_{p q}$ and $C_{p q}$ similarly. Take an arbitrary point $R$ on $\ell$ and draw the lines through $A_{p q}, B_{p q}, C_{p q}$ parallel to $R_{b} R_{c}, R_{c} R_{a}, R_{a} R_{b}$ respectively. Then the obtained triangle is homothetic to triangle $R_{a} R_{b} R_{c}$ with homothety center $F_{\ell}$. 3.4 Using the main theorem prove the results of problems $2.5,2.8,2.11$. 3.5 Prove that the circumcircle of triangle formed by the lines $A_{p q} A_{q p}, B_{p q} B_{q p}, C_{p q} C_{q p}$ passes through $F_{\ell}$. 3.6 Prove that the reflections of $\ell$ about the medial lines of triangle $A B C$ concur at point $F_{\ell}$. 3.7 Let $A_{\ell}, B_{\ell}$, and $C_{\ell}$ be the common points of $\ell$ with $B C, C A$, and $A B$ respectively. Prove that the circles with diameters $A A_{\ell}, B B_{\ell}, C C_{\ell}$ have two common points $F_{\ell}$ and $F_{\ell}^{\prime}, F_{\ell}^{\prime}$ lies on the circumcircle $\Omega$ of triangle $A B C$, and the orthocenter $H$ of this triangle lies on the line $F_{\ell} F_{\ell}^{\prime}$. 3.8 Let $O H$ be the Euler line of triangle $A B C$. Prove that the triangle formed by the poles $A_{h h}, B_{h h}$, $C_{h h}$ is autopolar with respect to the nine-points circle $\varepsilon$, in such a way that the vertices of $A B C$ lie on the sidelines of $A_{h h} B_{h h} C_{h h}$. 3.9 Let triangles $P_{a} P_{b} P_{c}$ and $Q_{a} Q_{b} Q_{c}$ be perspective to $A B C$ and their vertices lie on the corresponding sidelines of $A B C$. Let $A_{p q}, B_{p q}, C_{p q}$ be the common points of the corresponding sidelines of triangles $P_{a} P_{b} P_{c}$ and $Q_{a} Q_{b} Q_{c}$. Prove that: a) The vertices of $A B C$ lie on the sidelines of triangle $A_{p q} B_{p q} C_{p q}$; b) Triangle $A_{p q} B_{p q} C_{p q}$ is autopolar ith respect to conic $\Omega$ passing through $P_{a}, P_{b}, P_{c}, Q_{a}, Q_{b}, Q_{c}$; c) Triangle $A_{p q} B_{p q} C_{p q}$ is perspective to triangles $P_{a} P_{b} P_{c}$ and $Q_{a} Q_{b} Q_{c}$ and the corresponding perspective centers lie on the conic $\Omega$. 3.10 Let $X_{a}$ be an arbitrary point of line $B C$. Denote by $X_{b}$ the common point of lines $X_{a} B_{h h}$ and $A C$, note as $X_{c}$ the common point of lines $X_{a} C_{h h}$ and $A B$. Prove that - the line $X_{b} X_{c}$ passes through $A_{h h}$; - the lines $A X_{a}, B X_{b}$, and $C X_{c}$ concur at point $X$; - the circumcircle $X_{a} X_{b} X_{c}$ passes through $F_{O L}$, where $L$ is the Lemoine point of $A B C$. Finally formulate several open problems. 3.11 The definition of poles $A_{p q}, B_{p q}, C_{p q}$ depends on the order of choosing of points $P$ and $Q$ lying on $\ell$. The question: how the poles $A_{p q}$ and $A_{q p}$, i.e. the points of the same points taken in different order are correlated? We have the following hypothesis. The common points $Z_{a}, Z_{b}, Z_{c}$ of the corresponding sidelines of triangles $A_{p q} B_{p q} C_{p q}$ and $A_{q p} B_{q p} C_{q p}$ form a triangle homothetic to ABC with center $F_{\ell}$. 3.12 It is interesting to find such pair of points $P$ and $Q$ of line $\ell$ that the vertices of $A B C$ lie on the sidelines of triangle $A_{p q} B_{p q} C_{p q}$. Is always possible to choose such pair of points of $\ell$ ? Experiments show that such points always exist, but they can not lie too far from $O$. How can we describe such positions of $P$ and $Q$ that the vertices of $A B C$ lie on the sidelines of triangle $A_{p q} B_{p q} C_{p q}$ ? ## On inverse images of the Feuerbach point, poles of a triangle, and Kulanin's theorem Solutions ## 0. Auxiliary facts 0.1 Since $\angle B H C=180^{\circ}-\angle B A C$, points symmetric to $H$ in $B C$ and in the midpoint of $B C$, lie on the circumcircle. Homothety with center $H$ and ratio $1 / 2$ maps the circumcircle to Euler circle. 0.2 Perpendicular bisectors of the original triangle are the altitudes for the medial triangle, hence $O$ is its orthocenter. Homothety with center $M$ and ratio $-1 / 2$ maps the original triangle to the medial one, hence it maps $H$ to $O$. Hence $M$ belongs to the segment $H O$ and $H M: M O=2$. By the previous problem, $E$ belons to the segment $H O$ and $H E: E O=1: 1$. The equality $H E: E M: M O=3: 1: 2$ follows. 0.3 Let theee lines form a traingle $A B C$, while the fourth line intersects $A B, A C, B C$ at $D, E, F$, respectively. Let the circles $(A B C)$ and $(C E F)$ intersect at $\neq C$. Further, we prove that $P$ lies on the circle $(B D F)$. It is sufficient to prove that $\angle(B P, P F)=\angle(B D, D F)$. It is clear that $\angle(B P, P F)=$ $\angle(B P, P C)+\angle(P C, P F)=\angle(B A, A C)+\angle(E C, E F)=\angle(B D, A C)+\angle(A C, D F)=\angle(B D, D F)$. Similarly, $P$ lies on the circle $(A D E)$. 0.4 Let $I_{a}, I_{b}, I_{c}$ be the centers of the excircles of $A B C$. Note that $I_{a} A, I_{b} B, I_{c} C$ are the altitudes in $I_{a} I_{b} I_{c}$. The corresponding sides of $I_{a} I_{b} I_{c}$ and $G_{a} G_{b} G_{c}$ are parallel, i.e. these triangles are homothetic. Hence, their Euler lines are parallel. Note that $I$ is the circumcenter of $G_{a} G_{b} G_{c}$ and the orthocenter of $I_{a} I_{b} I_{c}$. Hence both Euler lines pass through $I$, and therefore, coincide. $O$ is the center of Euler circle in triangle $I_{a} I_{b} I_{c}$, hence $O$ belongs to its Euler line. 0.5 By angle chasing, follows from the solution of the next problem. 0.6 Consider the homothety with center $P$ and ratio 2 . Let $P_{a}^{\prime}, P_{b}^{\prime}, P_{c}^{\prime}, H_{a}^{\prime}, H_{b}^{\prime}, H_{c}^{\prime}$ be points symmetric to $P$ and $H$ in the corresponding sidelines. It suffices to prove that $P_{b}^{\prime}, H, P_{c}^{\prime}$ are collinear (the proof of collinearity of $P_{a}^{\prime}, H, P_{c}^{\prime}$ is analogous). By symmetry in $A B, \angle\left(P_{c}^{\prime} H, H B\right)=\angle\left(B H_{c}^{\prime}, H_{c}^{\prime} P\right)$ $\angle\left(C H, H P_{b}^{\prime}\right)=\angle\left(P H_{b}^{\prime}, H_{b}^{\prime} C\right)$. Since $P, H_{b}^{\prime}, H_{c}^{\prime}$ are concyclic, we have $\angle\left(P_{c}^{\prime} H, H P_{b}^{\prime}\right)=\angle\left(P_{c}^{\prime} H, H B\right)+$ $\angle(H B, C H)+\angle\left(C H, H P_{b}^{\prime}\right)=\angle\left(B H_{c}^{\prime}, H_{c}^{\prime} P\right)+\angle(H B, C H)+\angle\left(P H_{b}^{\prime}, H_{b}^{\prime} C\right)=\angle(B A, A P)+\angle(H B, C H)+$ $\angle(P A, A C)=\angle(B A, A C)+\angle(H B, C H)=0$, therefore, $P_{b}^{\prime}, H, P_{c}^{\prime}$ are collinear. 0.7 Let $P_{a}^{\prime}, P_{b}^{\prime}, P_{c}^{\prime}$ be points symmetric to $P$ in the corresponding sidelines. It is easy to check that the line symmetric to $A P$ in $A I$ is the perpendicular bisector of $P_{b}^{\prime} P_{c}^{\prime}$. Therefre, $P^{\prime}$ is the circumcenter of $P_{a}^{\prime} P_{b}^{\prime} P_{c}^{\prime}$. 0.8 We prove the following properties of isogonal conjugacy: - Follows from the previous problem and the fact that the reflections of $H$ in the sidelines lie on the circumcircle. - Using the previous and applying homothety with center $P$ and ratio $1 / 2$, we obtain that the midpoint of $P P^{\prime}$ is the center if the pedal circle of $P$. The midpoint of $P P^{\prime}$ is equidistant from the projections of $P$ and $P^{\prime}$ onto any sideline. It follows that the pedal circles of $P^{\prime}$ and $P$ coincide. - By the previous problem, the isogonal conjugate of $P$ is the intersection of the perpendicular bisectors of $P_{a}^{\prime} P_{b}^{\prime}, P_{b}^{\prime} P_{c}^{\prime}, P_{c}^{\prime} P_{a}^{\prime}$. Note that each perpendicular bisector is perpendicular to Simson line of $P$. - Clear. 0.9 By the previous problem, $P_{b}, P_{c}, Q_{b}, Q_{c}$ lie on a circle centered at the midpoint of $P Q$. Let $X$ and $Y$ be reflections of $P_{b}$ and $P_{c}$ in the center, respectively. Hence $X$ lies on $Q Q_{b}$, while $Y$ lies on $Q_{c}$. By Pascal theorem for $P_{b} Q_{c} Y P_{c} Q_{b} X$, the midpoint of $P Q$, the point $Q$, and the point $P_{b} Q_{c} \cap P_{c} Q_{b}$ are collinear. This means that $P_{b} Q_{c}, P_{c} Q_{b}, P Q$ are concurrent. 0.10 Since $A B C H$ is non-convex, each its circumconic is a hyperbola. Let $P, Q$ be its points at infinity. It suffices to prove that the directions defined by $P$ and $Q$ are perpendicular. Use pascal theorem for $A B C H P Q$. Let $X=A B \cap H P, Y=C H \cap Q A$. $X Y$ passes through $B C \cap P Q$, i.e. $X Y \| B C$, or, equivalently, $X Y \perp A H$. Hence $Y$ is the orthocenter of $A H X$, and it follows $A Y \perp H X$. Thus $A Q \perp H P$, QED. 0.11 Choose points $P, Q, R$ on a line $\ell$. Take a conic $c$ through its isogonal conjugates $P^{\prime}, Q^{\prime}, R^{\prime}$ and through $B, C$. Consider two maps $\ell \rightarrow c$ keeping cross-ratios. The first map to each $X \in \ell$ assings the intersection of $c$ with the reflection of $B X$ in the bisector of the angle $B$. The second map is the same with $B$ replaced by $C$. These two maps coincide at $P, Q, R$, hence in fact they coincide. Therefore, the image of $\ell$ under isogonal conjugacy is a conicn through $B$ and $C$. It is clear that this conic also passes through $A$. ## 1. Feuerbach theorem 1.1 Since $S_{a b}$ is the intersection point of $\lambda_{a}$ and the midline $M_{a} M_{b}$, triangle $B M_{a} S_{a b}$ is isosceles. We have $\angle S_{a b} B M_{a}=\frac{1}{2} \angle S_{a b} M_{a} C=\frac{1}{2} \angle A B C=\angle I B C$. Hence $S_{a b}$ lies on $B I$. Since $\angle B S_{a b} C=90^{\circ}, I, G_{b}, S_{a b}, C$ are concyclic (lie on the circle with diameter $C I$ ). Hence $\angle S_{a b} G_{b} C=\angle S_{a b} I C=90^{\circ}-\frac{1}{2} \angle B A C=\angle A G_{b} C$. Therefore, $S_{a b}$ lies on the line $G_{b} G_{c}$ (see fig. 1 ). 1.2 The angle between $H_{a} M_{a}$ and $\varepsilon$ is equal to the half of arc $H_{a} M_{a}$ in $\varepsilon$, hence it is equal to $\angle H_{a} M_{b} M_{a}=\left|\angle C M_{a} M_{b}-\angle C H_{a} M_{b}\right|=\left|\angle C B A-\angle H_{a} C M_{b}\right|=|\angle B-\angle C|$. 1.3 Note that $B I$ and $B I_{a}$ are internal and external bisectors of the angle $A B L_{a}$. Hence $A, I, L_{a}, I_{a}$ is a harmonic quadruple. The required statement follows from projection onto $B C$. 1.4 By the previous problem, the inversion in the circle with diameter $G_{a} G_{a}^{\prime}$ takes the nine-point circle $\varepsilon$ to the line $\ell$ passing through $L_{a}$. On the other hand, $\omega$ and $\omega_{a}$ are invariant under this inversion, since each of them is orthogonal to circle of inversion. Therefore, the angle between $\ell$ and $B C$ equals $|\angle B-\angle C|$, that is the angle between $\omega B C$. The line symmetric to $B C$ in $A I$ forms the same angle with $B C$. Therefore, $\ell$ is symmetric to $B C$ in $A I$, hecne $\ell$ is tangent to $\omega$ and $\omega_{a}$ (see fig. 2). 1.5 The fact of tangency of Eulaer circle with the excircle $\omega_{a}$ at some point $F_{a}$ follows from the solution of the previous problem. $L_{a}$ is the center of homothety with a negative ratio. The required perspectivity follows from Monge theorem for the incircle, the excircle, and Euler circle. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-282.jpg?height=1073&width=1339&top_left_y=190&top_left_x=358) Figure 1. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-282.jpg?height=1017&width=1312&top_left_y=1470&top_left_x=379) Figure 2. 1.7 Consider the homothety with center $F$ taking the incircle to the Euler circle. The line $H_{a} M_{a}$ maps to the line parallel to $H_{a} M_{a}$ and touching the Euler circle at $X$. By symmetry in the perpendicular bisector of $H_{a} M_{a}, X$ is the midpoint of the arc $H_{a} M_{a}$ of Euler circle. Hence $F X$ is the bisector of $\angle H_{a} F M_{a}$. Now the statement follows, since the homothety maps $G_{a}$ to $X$. 1.8 By $M_{X Y}$ denote the midpoint of $X Y$ Let $A, B, C, P$ be points, and let $X \neq M_{B P}$ be the common point of the Euler circles of $A B P$ and $B C P$. We have $\angle\left(M_{A P} X, X M_{C P}\right)=\angle\left(M_{A P} X, X M_{B P}\right)+$ $\angle\left(M_{B P} X, X M_{C P}\right)=\angle\left(M_{A P} M_{A B}, M_{A B} M_{B P}\right)+\angle\left(M_{B P} M_{B C}, M_{B C} M_{C P}\right)=\angle(B P, A P)+\angle(C P, B P)=$ $\angle(C P, A P)=\angle\left(M_{A P} M_{A C}, M_{A C} M_{C P}\right)$. This means that $X$ lies on the Euler circle of $A C P$. Similarly, $X$ lies on the Euler circle of $A B C$ (see fig. 3). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-283.jpg?height=767&width=1241&top_left_y=771&top_left_x=168) Figure 3. 1.9 Let $P_{a}, P_{b}, P_{c}$ be $P$ projections of $P$ to the sides of the triangle $A B C$. We use notation from the solution to the previous problem. $\angle\left(P_{c} X, X P_{a}\right)=\angle\left(P_{c} X, X M_{B P}\right)+\angle\left(M_{B P} X, X P_{a}\right)=\angle\left(P_{c} M_{A B}, M_{A B} M_{B P}\right)+\angle\left(M_{B P} M_{B C}, M_{B C} P_{a}\right)=$ $\angle\left(P_{c} A, A P\right)+\angle\left(P C, C P_{a}\right)=\angle\left(P_{c} P_{b}, P_{b} P\right)+\angle\left(P P_{b}, P_{b} P_{a}\right)=\angle\left(P_{c} P_{b}, P_{b} P_{a}\right)$. 1.10 First, let us prive that the corresponding sides of $K_{a} K_{b} K_{c}$ and $M_{a} M_{b} M_{c}$ are parallel. Chords $G_{a} K_{b}$ and $G_{b} G_{c}$ are symmetric in $B I$, hence they are equal. The similar is true for chords $G_{b} G_{c}$ and $G_{a} K_{c}$. Therefore, $K_{c} G_{a} K_{b}$ is isosceles, which implies $\angle G_{a} K_{c} K_{b}=\angle G_{a} K_{b} K_{c}=\angle C G_{a} K_{b}$. Hence $B C \| K_{b} K_{c}$. For the other pairs the proof is similar. Now by Desargues, $K_{a} K_{b} K_{c}$ are $M_{a} M_{b} M_{c}$ homothetic. The common point of lines $M_{a} K_{a}, M_{b} K_{b}$, and $M_{c} K_{c}$ is the center of this homothety, and moreover, the center of homothety taking the incircle to the Euler circle, that is the Feuerbach point $F$ (see fig. 4 ). 1.11 Follows from the Feuerbach theorem and problems 1.8, 1.9 for quadrilateral $A B C I$ (see fig. 5). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-284.jpg?height=811&width=1548&top_left_y=321&top_left_x=142) Figure 4. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-284.jpg?height=628&width=786&top_left_y=1670&top_left_x=713) Figure 5. 1.12 By the previous problem, the point $F$ lies on the Euler circle of $A I C$. The midpoint $X$ of $A I$ also lies on this circle. It suffices to prove that $\angle\left(G_{b} F, F X\right)=\angle\left(G_{b} F, F \widetilde{G}_{a}\right)$. We have $\angle\left(G_{b} F, F X\right)=\angle\left(G_{b} M_{b}, M_{b} X\right)=\angle\left(G_{b} C, C I\right)=\angle\left(G_{b} G_{a}, G_{a} I\right)=\angle\left(G_{b} G_{a}, G_{a} \widetilde{G}_{a}\right)=\angle\left(G_{b} F, F \widetilde{G}_{a}\right)$ (see fig. 6). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-285.jpg?height=605&width=786&top_left_y=516&top_left_x=713) Figure 6 . ## 2. The inverse images of the Feuerbach points 2.1 The inversion in $\lambda_{a}$ takes the Euler circle $\varepsilon$ to $H_{b} H_{c}$, while takes $\varepsilon_{a}$ to $S_{a b} S_{a c}$. Since $F$ is a common point of $\varepsilon$ and $\varepsilon_{a}, F_{a}^{\prime}$ is the common point of lines $H_{b} H_{c}$ and $S_{a b} S_{a c}$, which are radical axis of $\lambda_{a}$ and $\varepsilon, \lambda_{a}$ and $\varepsilon_{a}$ (see fig. 7). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-286.jpg?height=684&width=900&top_left_y=669&top_left_x=797) Figure 7. 2.2 First, we prove that $F, F_{a}^{\prime}, K_{b}, K_{b}^{\prime}$ are concyclic (similarly, for $F, F_{a}^{\prime}, K_{c}, K_{c}^{\prime}$ ). Consider the inversion $\operatorname{Inv}_{K_{a}}$ with center $K_{a}$ taking $F_{a}^{\prime}$ to $F$. Hence $\operatorname{Inv}_{K_{a}} \circ \operatorname{Inv}_{\lambda_{a}}$ takes $\varepsilon$ to a circle $\omega^{\prime}$ passingh through $F$ and $K_{a}$. This transformation preserves angles between lines and circles. Note that the centers of both inversions lie on $F F_{a}^{\prime}$. Therefore, circles $\varepsilon$ and $\omega^{\prime}$ form equal angles with $F F_{a}^{\prime}$. This means that $\varepsilon$ and $\omega^{\prime}$ touch at $F$. It follows that $\omega$ and $\omega^{\prime}$ coincide, sine there exists a unique circle passing through $K_{a}$ and touching $\varepsilon$ at $F$. Further, note that $K_{b}^{\prime}$ that is the point of intersection of ray $K_{a} K_{b}$ with $H_{b} H_{c}$ is the image of $K_{b}$ under $\operatorname{Inv}_{K_{a}}$. Hence $F, F_{a}^{\prime}, K_{b}$, and $K_{b}^{\prime}$ are concyclic. To complete with $S_{a b}$ and $S_{a c}$ use inscribed angles. First, $\angle K_{a} F K_{b}=\angle K_{a} K_{c} K_{b}$. Further, let $F K_{b}$ meet $\varepsilon_{a}$ again at $R$. We have $\angle M_{a} S_{a c} R=\angle M_{a} F R$. Lines $M_{a} S_{a c}$ and $K_{a} K_{c}$ are parallel. Since $\angle M_{a} S_{a c} R=\angle M_{a} F R$, we have $S_{a c} R\left\|K_{b} K_{c}\right\| B C$. it follows $\angle M_{a} F R=\angle S_{a c} S_{a b} G_{a}$, which means that $F, F_{a}^{\prime}, K_{b}, K_{b}^{\prime}, S_{a b}$ are concyclic. For $S_{a c}$ the proof is analogous (see fig. 8). 2.3 From the solution of the previous problem $\psi_{a b}$ and $\lambda_{a}$ are orthogonal. Therefore, $M_{a} S_{a b}$ touches $\psi_{a b}$. 2.4 Since $M_{a} S_{a b}$ and $M_{a} S_{a c}$ touch $\psi_{a b}$ and $\psi_{a c}$, respectively, and $M_{a} S_{a b}\left\|K_{a} K_{b} \quad M_{a} S_{a c}\right\| K_{a} K_{c}$, we have $K_{b} S_{a b}=K_{b}^{\prime} S_{a b}$ and $K_{c} S_{a c}=K_{c}^{\prime} S_{a c}$ (equal chords between parallel lines). Hence $\angle K_{b} F_{a}^{\prime} S_{a b}=$ $\angle K_{b}^{\prime} F_{a}^{\prime} S_{a b}$, QED. 2.5 $F_{a}^{\prime}$ lies on $K_{b} K_{c}$, by the previous problem. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-287.jpg?height=1390&width=1640&top_left_y=641&top_left_x=207) Figure 8. Note that $L_{b} K_{b}$ and $L_{c} K_{c}$ are tangent to the incircle at points $K_{b}$ and $K_{c}$ respectively. We use the following claim, which is a degenerate case of Brianchon's theorem: Let $K_{c} G_{b} K_{b} G_{c}$ be a cyclic quadrilateral. Tangents to its circumcircle through its vertices form another quadrilateral. Then the diagonals of these quadrilaterals have a common point. So the lines $K_{b} K_{c}, G_{b} G_{c}, L_{b} L_{c}$ are concurrent. By problem 1.3 the line $G_{b}^{\prime} G_{c}^{\prime}$ passes through their common point. 2.6 Let us prove for $\psi_{a b}^{\prime}$; for $\psi_{a c}^{\prime}$ the proof is similar. $\angle M_{a} F M_{b}=\angle M_{a} M_{c} M_{b}=\angle A C B=\angle G_{a} S_{a b} T_{a}$ (the last equality follows since $C, G_{a}, G_{b}, S_{a b}$ lie on the circle with diameter $C I$ ). By the problem 1.11, $\angle G_{a} F M_{a}=\angle G_{a} S_{a b} M_{a}$. Hence $\angle T_{a} F M_{b}=\angle M_{a} F M_{b}+\angle G_{a} F M_{a}=\angle G_{a} S_{a b} T_{a}+\angle G_{a} S_{a b} M_{a}=$ $\angle T_{a} S_{a b} M_{b}$ (see fig. 9). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-288.jpg?height=637&width=789&top_left_y=820&top_left_x=709) Figure 9 . 2.7 $F$ lies on circles $\left(M_{a} M_{b} M_{c}\right)$ and $\left(S_{a b} M_{a} S_{a c}\right)$. 2.8 Follows from two previous problems (see fig. 10). 2.9 Using the circle $\psi_{a b}^{\prime}$ from the problem 2.6 we have: $\angle A^{\triangle} P_{b} F=\angle T_{a} M_{b} F=\angle T_{a} S_{a b} F=$ $\angle A^{\triangle} S_{a b} F$. It follows that $A^{\triangle}$ lies on the circle $\psi_{a b}^{\triangle}$. Similarly, prove that $A^{\triangle}$ lies on $\psi_{a c}^{\triangle}$ (see fig. 11). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-289.jpg?height=635&width=792&top_left_y=306&top_left_x=707) Figure 10. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-289.jpg?height=1041&width=1827&top_left_y=1339&top_left_x=295) Figure 11. 2.10 First, we prove that $H_{b}, G_{b}, S_{a c}, S_{c a}$ are concyclic. Note that $B S_{c a}$ and $H_{b} S_{a c}$ are parallel, since both these lines are perpendicular to the bisector of $\angle B A C$. Similarly, $B S_{a c} \| G_{b} S_{c a}$. Hence $B S_{a c} H_{b} S_{c a}$ is a parallelogram, and $H_{b} S_{a c}=B S_{c a}=G_{b} S_{c a}$. Hence $S_{a c} H_{b} G_{b} S_{c a}$ is an isosceles trapezoid, thus it is inscribed (see fig. 8). Now let us prove that the Feuerbach point also lies on this circle. By Archimedes lemma, $\angle H_{b} F G_{b}=\frac{1}{2} \angle H_{b} F M_{b}=\frac{1}{2}|\angle A-\angle C|$. Further, $\angle H_{b} S_{a c} S_{c a}=\angle S_{c a} G_{b} C=\frac{\pi}{2}-\frac{\angle C}{2}$, and $\angle G_{b} S_{a c} S_{c a}=$ $\angle A G_{b} S_{a c}=\frac{\pi}{2}-\frac{\angle A}{2}$, hence $$ \angle H_{b} S_{a c} G_{b}=\left|\angle H_{b} S_{a c} S_{c a}-\angle G_{b} S_{a c} S_{c a}\right|=\frac{1}{2}|\angle A-\angle C|=\angle H_{b} F G_{b} $$ which imples that $F$ lies on the circumcircle of the trapezoid $H_{b} G_{b} S_{c a} S_{a c}$ It is easy to see that the point $I$ is the orthocenter of the triangle $S_{a c} S_{c a} G_{b}$. It follows that the reflection of $I$ in the line $S_{a c} S_{c a}$ lies on the circle $\left(S_{a c} S_{c a} G_{b}\right)$, QED. 2.11 Let us move $P_{a} P_{b} P_{c}$ linearly. Points $A^{\triangle}, B^{\triangle}, C^{\triangle}$ move keeping cross-ratio. $X_{a}=F A^{\triangle} \cap B C$ (and similarly, $X_{b}, X_{c}$ ) move keeping cross-ratio. The intersection points of perpendiculars to the sidelines through $A^{\triangle}, B^{\triangle}, C^{\triangle}$ with $O I$ also move keeping cross-ratio. To prove that they coincide, it suffices to find 3 particular positions. Two positions: $P_{a} P_{b} P_{c}=K_{a} K_{b} K_{c}$ and $P_{a} P_{b} P_{c}=M_{a} M_{b} M_{c}$ follow from the previous problems. The third position is a case, where $P_{a} P_{b} P_{c}$ degenerates to 3 lines passing through $F$ parallel to the sidelines. In this position all perpendiculars are lines at infinity. The statement on a circle follows from the problem 3.1. Alternatively, one could derive all the statements from the main theorem (see problem 3.4). ## 3. Generalized poles of a triangle and Kulanin's theorem 3.1 From problems $\mathbf{0 . 1 1}, \mathbf{0 . 1 0}$, and the first statement of the problem $\mathbf{0 . 8}$, it follows that the isogonal conjugate $P^{\prime}$ of $P$ traces along a rectangular hyperbola passing through the vertices of $A B C$. By the second statement of the problem $\mathbf{0 . 8}$, it suffices to prove that the pedal circles of $P^{\prime}$ pass through the center of this hyperbola. Using problems 1.9 and 1.8 , reduce to the known statement: the Euler circle of a triangle inscribed into a rectangular hyperbola passes through the center of this hyperbola (see fig. 12). 3.2 Let $P$ move linearly along the line $\ell$. Then $P_{a}, P_{b}, P_{c}$ move linearly along the sidelines. Hence $A_{o p}=F_{\ell} P_{a} \cap M_{b} M_{c}$ moves linearly along $M_{b} M_{c}$. It suffices to prove that $P_{b}, P_{c}, A_{o p}$ are collinear. For this purpose we check 3 positions of $P$. The case $P=O$ is trivial. Two cases, where $P$ lie on the circumcircle, follow from Kulanon's theorem (the pedal circle of $P$ degenerates to the Simson line) (see fig. 13). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-291.jpg?height=688&width=1634&top_left_y=369&top_left_x=251) Figure 12. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-291.jpg?height=700&width=1640&top_left_y=1607&top_left_x=248) Figure 13. 3.3 Let $P$ move linearly along the line $\ell$. Let $t$ be the length of the oriented segment $O P$. Introduce the Cartesian coordinates with origin $F_{\ell}$, and by $\left(X_{a}(t), Y_{a}(t)\right),\left(X_{b}(t), Y_{b}(t)\right),\left(X_{c}(t), Y_{c}(t)\right)$ denote coordinates of $A_{p q}, B_{p q}, C_{p q}$, respectively. We show that each of coordinate functions is a rational function of the form $f(t) / g(t)$ with $\operatorname{deg} f=2$, and $\operatorname{deg} g=1$. For instance, consider $A_{p q}$ (similar reasoning for points $B_{p q}$ and $C_{p q}$ ). It is clear that $P_{b}$ and $P_{c}$ move linearly, i.e. its coordinates are linear functions of $t$. The line through the points $P_{b}\left(x_{1}, y_{1}\right)$ and $P_{c}\left(x_{2}, y_{2}\right)$ is $\left(y-y_{1}\right)\left(x_{2}-x_{1}\right)=\left(x-x_{1}\right)\left(y_{2}-y_{1}\right)$. It is of the form $\alpha x+\beta y+\gamma=0$, where $\alpha$ and $\beta$ are linear in $t$, while $\gamma$ is quadratic in $t$. Finally, intersect $P_{b} P_{c}$ with the constant line $F_{\ell} Q_{a}$, find the coordinates of $A_{p q}$. Now let us prove that the directions of $A_{p q} B_{p q}, B_{p q} C_{p q}, C_{p q} A_{p q}$ are constant (though these lines do not move linearly). This condition is equivalent to equalities $$ F_{\ell} A_{p q}=\alpha \cdot F_{\ell} C_{p q} \quad \text { and } \quad F_{\ell} B_{p q}=\beta \cdot F_{\ell} C_{p q} $$ where $\alpha$ and $\beta$ are some constants. Let us prove the first equalilty (similarly, for the second one) in coordinates: $$ X_{a}(t)=\alpha_{1} \cdot X_{c}(t) \quad Y_{a}(t)=\alpha_{2} \cdot Y_{c}(t) $$ (where $\alpha_{1}$ and $\alpha_{2}$ are constants). Recall that each function in these relation has the form $f(t) / g(t)$ with $\operatorname{deg} f=2$, and $\operatorname{deg} g=1$. Multiplying by denominators, we obtain cubic equations in $t$. Thus it suffices to find 4 different values of $t$, or, equivalently, 4 different positions of $P$, for which the condition holds First, consider cases $P=O$ and $P=Q$. the previous problem implies that $\left\{A_{p q}, C_{p q}\right\}$ coincide, and do not coincide with $F_{\ell}$. Defining constants $\alpha_{1}$ and $\alpha_{2}$ by $\alpha_{1}=X_{a}(0) / X_{c}(0)$ and $\alpha_{2}=Y_{a}(0) / Y_{c}(0)$, we have equalities $$ X_{a}(t)=\alpha_{1} \cdot X_{c}(t) \quad \text { and } \quad Y_{a}(t)=\alpha_{2} \cdot Y_{c}(t) $$ which are true for two distinct values of $t$. Secondly, consider two positions of $P$ that are antipodal points of the the circumcircle $\Omega$ of $A B C$, lying on $\ell$. In this case $A_{p q}=C_{p q}=F_{\ell}$, i.e. their coordinates are 0 , and required equalities hold. Thus we have proved that all lines $A_{p q} C_{p q}$ are parallel, and the same is true for lines $A_{p q} B_{p q}$ and $B_{p q} C_{p q}$. Hence, while $P$ is moving, $A_{p q} B_{p q} C_{p q}$ traces a family of triangles homothetic with center $F_{\ell}$. Now it remains to show that the triangles formed by limes through $A_{p q}, B_{p q}, C_{p q}$ parallel to $R_{b} R_{c}$, $R_{c} R_{a}, R_{a} R_{b}$, respectively, is homothetic to $R_{a} R_{b} R_{c}$ with homothety center $F_{\ell}$. For this purpose, consider the case $P=R$. We have $A_{p q}=A_{r q}, B_{p q}=B_{r q}$, and $C_{p q}=C_{r q}$ lying on the lines $R_{b} R_{c}$, $R_{c} R_{a}$, and $R_{a} R_{b}$, respectively. For an arbitrary position of $P$ consider the homothety with center $F_{\ell}$ taking $A_{r q}$ to $A_{p q}$. By the arguments above, this homothety takes $B_{r q}$ to $B_{p q}$, and it takes $C_{r q}$ to $C_{p q}$. Hence $R_{b} R_{c}$ maps to the line through $A_{p q}$ parallel to $R_{b} R_{c}$. Similarly, for the lines $R_{c} R_{a}$ and $R_{a} R_{b}$ (see fig. 14). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-293.jpg?height=732&width=1759&top_left_y=970&top_left_x=259) Figure 14. 3.4 The problems 2.5 and 2.8 could be proved without using the main theorem, thus we just mention that the construction in the problem 2.5 is the case $P=I, Q=O$ of the main theorem, while the construction in the problem 2.8 is the case $P=Q=I, R=O$. Now derive the problem 2.11 . Let the perpendicular to $B C$ through its intersection point with $F A^{\triangle}$ meets $O I$ at $Q$. The line $P_{b} P_{c}$ passes through $A^{\triangle}=A_{i q}$. Applying the mein theorem to $I, Q, R=O$. obtain that $P_{a} P_{b}$ and $P_{a} P_{c}$ pass through $C_{i q}$ and $B_{i q}$, respectively. Therefore, $B^{\triangle}=B_{i q}$ and $C^{\triangle}=C_{i q}$. Hence the feet of perpendiculars from $Q$ onto $A B$ and $A C$ are the intersection points of $A B$ and $A C$ with $F C^{\triangle}$ and $F B^{\triangle}$, respectively. The statement about the circle follows from Kulanin's theorem. (see fig. 15, 16, 17). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-294.jpg?height=686&width=1634&top_left_y=703&top_left_x=251) Figure 15. $P=Q=I, R=O$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-294.jpg?height=696&width=1637&top_left_y=1674&top_left_x=252) Figure 16. $P=I, Q=R=O$ ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-295.jpg?height=677&width=1634&top_left_y=209&top_left_x=251) Figure 17. $P=I, R=O$ 3.5 The main theorem is equivalent to the equality $F A_{r q} / F A_{p q}=F B_{r q} / F B_{p q}=F C_{r q} / F C_{p q}$. Similarly, we have $F A_{r p} / F A_{q p}=F B_{r p} / F B_{q p}=F C_{r p} / F C_{q p}$. Let $R$ be a point such that $R_{b} R_{c} \| A_{p q} A_{q p}$. Then there exists a homothety with center $F$ taking $R_{b} R_{c}$ to $A_{p q} A_{q p}$. This homothety takes $A_{r q}$ to $A_{p q}$, and takes $A_{r p}$ to $A_{q p}$. From the equalities above it follows that this homothety takes $R_{a} R_{b} R_{c}$ to the triangle formed by $A_{p q} A_{q p}, B_{p q} B_{q p}, C_{p q} C_{q p}$, and the required statement follows. 3.6 It is known that the lines symmetric to $\ell$ in the midlines of the triangle intersect at some point $X$. The point $P$ symmetric to $X$ in $M_{b} M_{c}$ lies on $\ell$ and on the circle $\left(A M_{b} M_{c}\right)$. (Since the reflections in the sidelines of a line through the orthocenter meet on the circumcircle). Using the problem 3.2 and the Simson line of $P$ with respect to $A M_{b} M_{c}$, we have $P F_{\ell} \perp B C$. It is clear that $P X \perp B C$. Similarly, for the other reflections of $X$ in the midlines. Hence $F_{\ell}=X$ (see fig. 18). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-295.jpg?height=607&width=1464&top_left_y=1746&top_left_x=298) Figure 18. 3.7 If $P$ coincides to one of the points $A_{\ell}, B_{\ell}, C_{\ell}$, then $F_{\ell}$ lies on these circles, by Kulanin's theorem. Let $A^{\prime}, B^{\prime}, C^{\prime}$ be reflections of $A, B, C$ in $O$. By Pascal theorem, $A^{\prime} A_{\ell} \cap B^{\prime} B_{\ell}$ lies on the circle $(A B C)$. Hence, lines $A^{\prime} A_{\ell}, B^{\prime} B_{\ell}, C^{\prime} C_{\ell}$ meet at some point $F_{\ell}^{\prime}$ of the circumcircle. As $\angle A_{\ell} F_{\ell}^{\prime} A=\angle A^{\prime} F_{\ell}^{\prime} A=$ $90^{\circ}$, the point $F_{\ell}^{\prime}$ lies on the required circles. $H$ lies on $F_{\ell} F_{\ell}^{\prime}$, since the powers of $H$ with respect to these circles are equal to $H H_{a} \cdot H A=H H_{b} \cdot H B=H H_{c} \cdot H C$ (see fig. 19). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-296.jpg?height=610&width=1460&top_left_y=451&top_left_x=300) Figure 19. 3.8 Consider the quadrilateral $H_{a} H_{b} H_{c} F_{\ell}$ inscribed in the Euler circle $\varepsilon$. We set $H_{a} H_{c} \cap F_{\ell} H_{b}=B_{h h}$, $H_{a} H_{b} \cap F_{\ell} H_{c}=C_{h h}, H_{a} F_{\ell} \cap H_{b} H_{c}=A_{h h}$. It follws that $A_{h h} B_{h h} C_{h h}$ is an autopolar traingle for $\varepsilon$. Now let us show that the sidelines of $A_{h h} B_{h h} C_{h h}$ contain $A, B, C$. Indeed, consider $H_{c} M_{c} H_{b} M_{b}$ that is a quadrilateral inscribed in $\varepsilon$. By the construction, $A$ lies on the polar line of $A_{h h}$. But the polar line of $A_{h h}$ is $B_{h h} C_{h h}$, hence $A$ lies on $B_{h h} C_{h h}$. Similarly, for the other vertices (see fig. 20). Moreover, this problem is a particular case of the next one. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-297.jpg?height=705&width=1656&top_left_y=553&top_left_x=251) Figure 20. 3.9 a) the cross-ratio of the lines $P_{a} B, P_{a} P_{c}, P_{a} A, P_{a} P_{b}$ equals -1 . Projecting otno the line $A B_{p q}$, we get the harmonic quadruple $A B_{p q} \cap B C, B_{p q}, A, A B_{p q} \cap P_{a} P_{b}$. Similarly, $A B_{p q} \cap B C, B_{p q}, A$, $A B_{p q} \cap Q_{a} Q_{b}$ is a harmonic quadruple. Hence $A B_{p q}, P_{a} P_{b}, Q_{a} Q_{b}$ are collinear, i.e. $A$ lies on $B_{p q} C_{p q}$. b,c) By Pappus theorem, $P_{b} Q_{c} \cap P_{c} Q_{b}$ lies on $P Q$. Similarly, for $P_{a} Q_{c} \cap P_{c} Q_{a}$ and $P_{a} Q_{b} \cap P_{b} Q_{a}$. By Pascal theorem, $P_{a} Q_{b} P_{c} Q_{a} P_{b} Q_{c}$ is inscribed into a conic $\Omega$. Let $F=B_{p q} P_{b} \cap C_{p q} P_{c}$. by Pascal theorem and the item a), $F P_{b} Q_{b} Q_{a} Q_{c} P_{c}$ is inscribed into $\Omega$. Analogous statement is true for $A_{p q} P_{a} \cap C_{p q} P_{c}$. it follows that $A_{p q} P_{a}, B_{p q} P_{b}, C_{p q} P_{c}$ meet on $\Omega$, and c) follows. Now be could be proved similarly to the previous problem. 3.10 Let $X_{a}$ move along $B C$ linearly. Then $X_{b}$ and $X_{c}$ move along $A C$ and $A B$, respectively, preserving cross-ratio. - It suffices to prove for 3 positions of $X_{a}$. Cases $X_{a}=B, C, M_{a}$ follow from the previous problem; - $X=B X_{b} \cap C X_{c}$ moves along some conic passing through $B$ and $C$ preserving cross-ratio (the proof is analogous to one for the problem 0.11). This conic also passes through $A$, since $X=A$, for $X_{a}=B C \cap B_{h h} C_{h h}$. Thus it suffices to prove that $A X$ passes through $X_{a}$, for some 3 positions of $X_{a}$. The first position is $X_{a}=X=B$, the second is $X_{a}=X=C$, the third is $X_{a}=M_{a}, X=M$ (and the fourth is $X_{a}=H_{a}, X=H$ ). - $X$ moves along a rectangular hyperbola $A B C H M$. It is known that the circle $\left(X_{a} X_{b} X_{c}\right)$ passes through the center of this hyperbola (we do not know an easy proof of this fact; see A. Akopyan, A.Zaslavsky "Geometrical properties of conics", Theorem 4.3). The center of this hyperbola is $F_{O L}$, that is the generalized Feuerbach point for the isogonal conjugate on this hyperbola (see fig. 21). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-298.jpg?height=707&width=1675&top_left_y=384&top_left_x=239) Figure 21. 3.11 We present a sketch of the proof that all traingles $Z_{a} Z_{b} Z_{c}$ are homothetic (though we do not specify the center of homothety). We need to prove that $B C\left\|Z_{b} Z_{c}\right\| A_{p p} A_{q q}$ (here we use the problem 3.2: $A_{p p} A_{q q}$ is the midline). Reduce it to a statement of projective geometry: Assume that traingles $P_{a} P_{b} P_{c}$ and $Q_{a} Q_{b} Q_{c}$ are given, and let $F$ be an arbitrary point. Let $A_{p p}=$ $P_{b} P_{c} \cap F P_{a}$, while $A_{p q}, B_{q p}, \ldots$ are defined similarly, like in the problem 3.3. Let $Z_{b}=A_{p q} C_{p q} \cap A_{q p} C_{q p}$ and $Z_{c}=A_{p q} B_{p q} \cap A_{q p} B_{q p}$. Hence $P_{a} P_{b}, A_{p p} A_{q q}$, and $Z_{b} Z_{c}$ are concurrent. Note that cross-ratios of quadruples $A_{q p} Q_{c}, A_{q p} B_{q p}, A_{q p} F, A_{q p} Q_{a}$ and $A_{p q} P_{b}, A_{p q} C_{p q}, A_{p q} F, A_{p q} P_{a}$ are equal (project with center $A_{q p}$ onto $Q_{c} Q_{a}$, then project with center $F$ onto $P_{b} P_{a}$, then with center $A_{p q}$. Similarly, cross-ratios of quadruples $A_{q p} Q_{c}, A_{q p} C_{q p}, A_{q p} F, A_{q p} Q_{a}$ and $A_{p q} P_{b}, A_{p q} B_{p q}, A_{p q} F, A_{p q} P_{a}$ are equal. Now one can apply the following fact: Let $f$ be a map $l \mapsto f(l)$ from the band of lines through $A_{q p}$ to the band of lines through $A_{p q}$ that preserves cross-ratio. Let $m$ and $n$ be arbitrary lines through $A_{q p}$. Consider the line through $m \cap f(n)$ and $n \cap f(m)$. Then this line passes through a fixed point, regardless of a choice of $m$ and $n$. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-299.jpg?height=710&width=1718&top_left_y=984&top_left_x=239) Figure 22. [^0]: ${ }^{1}$ Удобно считать, что значение $y$ фиксировано, а клетки $(-y, y),(-y+2, y), \ldots,(y, y)$ - это все возможные исходы эксперимента. Например, $y$-тая горизонталь может быть фотопластиной, детектирующей электрон. Знакомство с теорией вероятностей не требуется для решения предложенных задач. Осторожно! Такое правило вычисления вероятности годится только для базовой модели; позже мы его изменим, когда будем уточнять модель. Мы будем делать подобные замечания каждый раз, когда будем нарушать какой-то фундаментальный принцип ради простоты. ${ }^{2}$ Здесь появляется дополнительный исход эксперимента, когда электрон поглощается и не достигает фотопластины. [^1]: ${ }^{3}$ Обратите внимание, что в размерности 3 и больше правый и левый означают не направление движения, а нечто совсем другое. Хотя это часто представляют, как направление вращения электрона, эти состояния не имеют неквантовых аналогов. ${ }^{4}$ И киралъностъю; будьте внимательны: термин спин обычно обозначает свойство, не имеющее никакого отношения к направлению движения. ${ }^{5}$ Таким образом, результатом эксперимента является пара (конечная координата по $x$, направление последнего хода), а конечная координата по $y$ фиксирована. Именно эти вероятности являются фундаментальными, а $P(x, y)$ в общем случае по определению равна выражению из решения задачи 11 а не приведённому выше выражению $P(x, y)=|\vec{a}(x, y)|^{2}$ (которое случайно совпало с нужным в простом случае) [^2]: ${ }^{6}$ Объяснить, откуда в пределах берётся нормирующий множитель $n$, несколько сложнее; мы не будем это делать. ${ }^{7}$ Называемые функииями Бесселя, они почти так же хорошо исследованы, как синус и косинус; предварительное знакомство с ними не требуется. [^3]: ${ }^{8}$ Обозначение $\vec{a}(x, y, \lambda / \varepsilon, \pm)$ надо не путать с $\vec{a}(x, y, m \varepsilon, \pm)$ выше. ${ }^{9}$ Осторожно: как правило, шашки Фейнмана не подходят для описания света; частичное отражение - это замечательное исключение. ${ }^{10}$ Масса пропорциональна $(n-1) /(2 \sqrt{n})$, где $n$ показатель преломления; у стекла $n \approx 1.5$ и $(n-1) /(2 \sqrt{n}) \approx 0.2$. 11 Это упрощающее предположение требует отрицательной массы на левой поверхности; причина этого станет ясна после решения задачи 25 [^4]: ${ }^{12}$ Концептуальнее было бы определить $P(\lambda, L,-)=\lim _{\varepsilon \rightarrow 0} \lim _{\Delta \rightarrow+\infty} \lim _{y \rightarrow+\infty} \sum_{x \in \mathbb{Z}} P\left(x, y, m_{3}(x) \varepsilon, \frac{\lambda}{\varepsilon}, \Delta,-\right)$, но мы не будем так делать. 13Ятот дополнительный поворот можно объяснить следующим образом. Свет может рассеяться в каждой клетке внутри стекла несколько раз. Каждое рассеивание добавляет множитель - imв к нашему вектору (рассматриваемому, как комплексное число) и может поменять или не поменять направление движения шашки. Предположим, $m \varepsilon<1$. Таким образом, ход без изменения направления движения добавляет множитель [^5]: ${ }^{14}$ Обратите внимание, что хотя этот метод рассмотрения магнитного поля хорошо известен, он сильно отличается от метода, изложенного в Feynman. ${ }^{15}$ Или, точнее, электромагнитный потенииал. Поле следует интерпретировать как магнитное или электрическое в зависимости от того, интерпретируется ли координата по $y$ как положение или как время. [^6]: 16 Здесь важно, что $s$ и $s^{\prime}$ - это пути однотипных частиц, например двух электронов. Иначе вторая сумма опускается. Знак перед второй суммой меняется на плюс для некоторых частиц, например фотонов (частиц света). ${ }^{17}$ Удобно считать, что значение $y$ фиксировано, а четвёрки $\left(F, F^{\prime}, E, E^{\prime}\right)$ - это возможные исходы эксперимента. 18 эт определение осмысленно только при $x_{0} \geq 2 y ;$ иначе аннигиляцией частиц нельзя пренебречь, и модель становится неадекватной. [^7]: ${ }^{19}$ Если шаг решётки $\epsilon$ не фиксирован, то $g$ может зависеть от него. Это называется перенормировкой; мы не будем обсуждать её. ${ }^{20}$ Он определён корректно, так как путь $s$ целиком содержится в $R$, не считая первого и последнего хода. Будем считать, что $g^{n}=1$ при $g=n=0$. ${ }^{21}$ Здесь мы суммируем не вектора, а вероятности. Не путайте обозначение $P(x, y, g,+)$ с $P(x, y, m \varepsilon,+)$. [^8]: ${ }^{22}$ Эта расстановка не имеет никакого отношения к магнитному полю из $\S 6$ ${ }^{23}$ Мы рассматриваем именно знаки, а не возможные пути шашки через конкретную клетку, как на рисунке 6 справа. [^9]: ${ }^{24}$ Не надо путать это с приниипом неопределённости, который устанавливает предел точности одновременного измерения координат и импульса. [^10]: ${ }^{1}$ One should think of the value $y$ as fixed, and the squares $(-y, y),(-y+2, y), \ldots,(y, y)$ as all the possible outcomes of an experiment. For instance, the $y$-th horizontal might be a photoplate detecting the electron. Familiarity with probability theory is not required for solving the presented problems. Beware that our rule for the probability computation is valid only for the basic model in question; we are going to change the rule slightly in the upgrades. We make similar remarks each time we break some fundamental principles for simplicity. ${ }^{2}$ Thus an additional outcome of the experiment is that the electron has been absorbed and has not reached the photoplate. [^11]: ${ }^{3}$ Beware that in 3 or more dimensions 'right' and 'left' mean something very different from the movement direction. Although often visualized as the direction of the electron rotation, these states cannot be explained in nonquantum terms. ${ }^{4}$ And chirality; beware that the term spin usually refers to a property, not related to the movement direction at all. ${ }^{5}$ Thus an experiment outcome is a pair (final $x$-coordinate, last-move direction), whereas the final $y$-coordinate is fixed. These are the fundamental probabilities, whereas $P(x, y)$ should in general be defined by the formula from the solution of Problem 11 rather than the above formula $P(x, y)=|\vec{a}(x, y)|^{2}$ (being a coincidence). [^12]: ${ }^{6}$ In the limits, the normalization factor of $n$ is a bit harder to explain; we not discuss it. ${ }^{7}$ Called Bessel functions, which are almost as well-studied as sine and cosine; but familiarity with them is not required. ${ }^{8}$ The notation $\vec{a}(x, y, \lambda / \varepsilon, \pm)$ should not be confused with $\vec{a}(x, y, m \varepsilon, \pm)$. [^13]: ${ }^{9}$ Beware: in general Feynman checkerboard is inappropriate to describe light; partial reflection is a remarkable exception. ${ }^{10}$ The mass is proportional to $(n-1) / 2 \sqrt{n}$, where $n$ is the refractive index; for glass $n \approx 1.5$ and $(n-1) / 2 \sqrt{n} \approx 0.2$. ${ }^{11}$ This simplifying assumption requires negative mass for the left surface; the origin of that becomes clear after solving 25 . ${ }^{12}$ It is more conceptual to define $P(\lambda, L,-)=\lim _{\varepsilon \rightarrow 0} \lim _{\Delta \rightarrow+\infty} \lim _{y \rightarrow+\infty} \sum_{x \in \mathbb{Z}} P\left(x, y, m_{3}(x) \varepsilon, \frac{\lambda}{\varepsilon}, \Delta,-\right)$ but we do not. ${ }^{13}$ This additional rotation is explained as follows. The light can be scattered in each square inside the glass several times. Each individual scattering gives a factor of -ime to our vector (viewed as a complex number) and may or may not change the movement direction. Assume that $m \varepsilon<1$. Thus a move without changing the direction contributes a factor of $$ 1 \text { (no scattering })-i m \varepsilon(1 \text { scattering })+(-i m \varepsilon)^{2}(2 \text { scatterings })+\cdots=\frac{1}{1+i m \varepsilon} \text {. } $$ Without a scattering, the checker moves straightly. Thus a turn in a particular square inside the glass contributes a factor $$ -i m \varepsilon(1 \text { scattering })+(-i m \varepsilon)^{2}(2 \text { scatterings })+(-i m \varepsilon)^{3}(3 \text { scatterings })+\cdots=\frac{-i m \varepsilon}{1+i m \varepsilon} $$ [^14]: ${ }^{14}$ Beware that this method of adding the magnetic field, although well-known, is very different from the one from [Feynman]. ${ }^{15} \mathrm{Or}$ electromagnetic vector-potential, to be precise. The field is interpreted as magnetic or electric depending on if the $y$-coordinate is interpreted as position or time. ${ }^{16}$ Here it is essential that $s$ and $s^{\prime}$ are paths of particles of the same sort, e.g., two electrons. Otherwise the 2nd sum is omitted. The sign before the 2nd sum is changed to plus for some other sorts of particles, e.g., photons (particles of light). [^15]: ${ }^{17}$ One should think of the value $y$ as fixed, and the quadruples $\left(F, F^{\prime}, E, E^{\prime}\right)$ as the possible outcomes of an experiment. ${ }^{18}$ This definition makes sense only for $x_{0} \geq 2 y$; otherwise annihilation of particles cannot be ignored, and the model becomes inappropruate. ${ }^{19}$ If the lattice step $\epsilon$ is not fixed, then $g$ might depend on it. This is called renormalization; we do not discuss it. [^16]: ${ }^{20}$ It is well-defined because the path $s$ is contained in $R$ except the first and the last move. We set $g^{n}=1$ for $g=n=0$. ${ }^{21}$ Here we sum probabilities rather than vectors. The notation $P(x, y, g,+)$ should not be confused with $P(x, y, m \varepsilon,+)$. [^17]: ${ }^{22}$ The assignment has nothing to do with the magnetic field from $\S 6$. ${ }^{23}$ We consider positions of signs, but not possible checker paths in a particular black square like in Figure 6 to the right. [^18]: ${ }^{24}$ This should not be confused with uncertainty principle, which does not allow simultaneous measurement of the coordinates and momentum. [^19]: ${ }^{25}$ Fractions in those equalities are respectively $$ \binom{k}{q-2-k, \frac{p-q}{2}+1+k, k-\frac{p+q}{2}+1} \text { and }\binom{k}{q-2-k, \frac{p-q}{2}+1+k, k-\frac{p+q}{2}+1} \cdot \frac{k+p+1}{q-1-k} $$ and upwards-left moves respectively. These numbers are uniquely determined by the conditions $L+R=y$ and $R-L=x$. Denote by $2 r+1$ the number of turns in the path. Let $x_{1}, x_{2}, \ldots, x_{r+1}$ be the number of consecutive upwards-right moves before the first, the third, ..., the last turn respectively. Let $y_{1}, y_{2}, \ldots, y_{r+1}$ be the number of consecutive upwards-left moves after the first, the third, ..., the last turn respectively. Then $x_{k}, y_{k} \geq 1$ and $$ \begin{aligned} R & =x_{1}+\cdots+x_{r+1} \\ L & =y_{1}+\cdots+y_{r+1} \end{aligned} $$ Conversely, each collection $\left(x_{1}, x_{2}, \ldots, x_{r+1}, y_{1}, y_{2}, \ldots, y_{r+1}\right)$ satisfying the resulting equations determines a path from $(0,0)$ to $(x, y)$ with an odd number of turns. The problem now reduces to a combinatorial one: find the number of positive integer solutions of the resulting equations. For the first equation, this number equals to the number of ways to put $r$ sticks between $R$ coins in a row, that is, $\binom{R-1}{r}$. Thus $$ a_{1}(x, y)=2^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{R-1}{r}\binom{L-1}{r} $$ Since $L+R=y$ and $R-L=x$, the required formula from Answer 2 follows. The formula for $a_{1}(x, y)$ from Answer 1 can be derived using the polynomial from the remark after the answer. Depending on the sign of $x$ we get $$ 2^{(1-y) / 2}(1+t)^{y-k-1}(1-t)^{k-1}=\left\{\begin{array}{l} 2^{(1-y) / 2}\left(1-t^{2}\right)^{\frac{y-x-2}{2}}(1-t)^{x}, \text { for } x \geq 0 \\ 2^{(1-y) / 2}\left(1-t^{2}\right)^{\frac{y+x-2}{2}}(1+t)^{-x}, \text { for } x<0 \end{array}\right. $$ For each $|x||x|$ is analogous. The case $y \leq|x|$ is obvious. Solution by Alexander Kudryavtsev (participant of Summer conference of Tournament of Towns). Denote $c_{p, q}=a_{1}(2 p-q+1, q+1) \cdot 2^{q / 2}, d_{p, q}=a_{2}(2 p-q+1, q+1) \cdot 2^{q / 2}$ (they are well-defined for $\left.0 \leq p \leq q-1\right)$. Let $C(x, y), D(x, y)$ be generating functions of $c_{p, q}$ and $d_{p, q}$ respectively. From the answer to Problem 4 we obtain the following equations: $$ \left\{\begin{array}{l} \frac{C(x, y)-C(x, 0)}{y}=D(x, y)+C(x, y) \\ \frac{D(x, y)-D(x, 0)}{y}=x \cdot(D(x, y)-C(x, y)) \end{array}\right. $$ Since $C(x, 0)=0$ and $D(x, 0)=1$, we get $$ \left\{\begin{array}{l} C(x, y)=y \cdot(D(x, y)+C(x, y)) \\ D(x, y)-1=x y \cdot(D(x, y)-C(x, y)) \end{array}\right. $$ The solution of this system is $C(x, y)=\frac{y}{1-y-x y+2 x y^{2}}$ and $D(x, y)=\frac{1-y}{1-y-x y+2 x y^{2}}$. Denote $$ E(x, y):=\frac{C(x, y)}{y}=\frac{D(x, y)}{1-y}=1+y(1+x-2 x y)+y^{2}(1+x-2 x y)^{2}+\ldots $$ The coefficient of $x^{n} \cdot y^{m}$ in the $E(x, y)$ is equal to $$ \sum_{k=\max (n, m-n)}^{m}(-2)^{m-k} \cdot \frac{k!}{(m-k)!(n-m+k)!(k-n)!} $$ since from every summand of the form $y^{k}(1+x-2 x y)^{k}$ we must take exactly one combination of factors: - for the power of $y$ to be equal to $m$, the number of factors $-2 x y$ must be exactly $m-k$; - for the power of $x$ to be equal to $n$, the number of factors $x$ must be exactly $n-(m-k)$; - for the total number of factors to be $k$, the number of factors 1 must be $k-(n-(m-k))-(m-k)=k-n$. And the number of ways to choose $m-k, k-n$ and $n-(m-k)$ factors $(1+x-2 x y)$ equals to the fraction in the formula for $E(x, y)$. Now Answer 3 can be easily obtained from the equations $C(x, y)=y \cdot E(x, y)$ and $D(x, y)=E(x, y)-C(x, y)$. 10. Answer: $$ \vec{a}(x, y)=\frac{\sqrt{2}\left(\sin \left(\frac{\pi y}{4}-\frac{x^{2}}{2 y}\right), e^{x / y} \cos \left(\frac{\pi y}{4}+\frac{x(2-x)}{2 y}\right)\right)}{\sqrt{\pi y}}+\mathrm{O}\left(\frac{\log ^{2} y}{y}\right) \text { for }|x|=\mathrm{O}(\sqrt{y}) $$ Remark. Compare the graphs of $P(x, 400), a_{1}(x, 400)$, and $a_{2}(x, 400)$ (blue) with the graphs of their approximations in the right-hand side for $y=400$ (orange): ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-103.jpg?height=414&width=1878&top_left_y=541&top_left_x=150) Hint. Note that by Problem 9 $$ \begin{aligned} & a_{2}(-n+2 k+1, n+1)=2^{-n / 2}\left[x^{k-1}\right](-1)^{k}(1+x)^{n-k-1}(1-x)^{k} \\ & a_{1}(-n+2 k+1, n+1)=2^{-n / 2}\left[x^{k}\right](-1)^{k}(1+x)^{n-k-1}(1-x)^{k} \end{aligned} $$ Thus $$ a_{2}(-n+2 k+1, n+1)=2^{-n / 2}(-1)^{k} \int_{0}^{1}\left(1+e^{2 \pi i t}\right)^{n-k-1}\left(1-e^{2 \pi i t}\right)^{k} e^{-2 \pi i(k-1) t} d t $$ For an odd $n$ denote $l=k-\frac{n-1}{2}$. We have $$ a_{2}(-n+2 k+1, n+1)=\frac{i^{k}}{\sqrt{2}} \int_{0}^{1}(\sin (2 \pi t))^{\frac{n-1}{2}}(\tan (\pi t))^{l} e^{-2 \pi i(l-1) t} d t $$ Since $l<3$, thus $\max _{y} P(y,-)=$ $P(2,-)=P(3,-)=1 / 2$. By the Newton binomial theorem we get $\sum_{k=0}^{\infty}\binom{2 k}{k} x^{k}=\frac{1}{\sqrt{1-4 x}}$ for each $x \in\left[-\frac{1}{4}, \frac{1}{4}\right)$. Setting $x=-\frac{1}{4}$ we obtain $\lim _{y \rightarrow \infty} P(y,-)=\lim _{y \rightarrow \infty} \frac{1}{2} \sum_{k=0}^{\infty}\binom{2 k}{k}\left(-\frac{1}{4}\right)^{k}=\frac{1}{2 \sqrt{2}}$. Remark. Using the Stirling formula (see $\S 2$ ) one can estimate the convergence rate: $$ \left|P(k,-)-\lim _{y \rightarrow \infty} P(y,-)\right|<\frac{1}{2 \cdot 4\lfloor k / 2\rfloor}\binom{2\lfloor k / 2\rfloor}{\lfloor k / 2\rfloor}<\frac{e}{2 \pi \sqrt{2\lfloor k / 2\rfloor}}<\frac{1}{2 \sqrt{k}} $$ 13. Answer: $$ \begin{gathered} P(x, y, 0,+)= \begin{cases}1, & \text { for } x=y \\ 0, & \text { for } x \neq y\end{cases} \\ P(x, y, \infty,+)= \begin{cases}1, & \text { for } x=1, y \equiv 1 \quad \bmod 2 \\ 0, & \text { otherwise }\end{cases} \end{gathered} $$ Solution. By the definition, if the number of turns in a path $s$ is nonzero then $\vec{a}(s, 0)=0$. Therefore, $\vec{a}(x, y, 0,+) \neq 0$ if and only if $x=y$. Hence, we get the answer for $P(x, y, 0,+)$. The case $x=y$ is obvious. One cannot prove the answer for $P(x, y, \infty,+)$ because it is a definition; but let us provide a motivation for the definition. If $m \varepsilon$ is "very large", then $\vec{a}(s, m \varepsilon)$ is "small" unless the checker turns at each move; this gives the formula for $P(x, y, \infty,+)$. Remark. $P(x, y, \infty,+)=\lim _{m \varepsilon \rightarrow \infty} P(x, y, m \varepsilon,+)$. Indeed, $$ \lim _{m \varepsilon \rightarrow \infty}|\vec{a}(s, m \varepsilon)|=\lim _{m \varepsilon \rightarrow \infty} \frac{(m \varepsilon)^{t}(s)}{\left(\sqrt{1+(m \varepsilon)^{2}}\right)^{y-1}}=\lim _{m \varepsilon \rightarrow \infty} \frac{1}{\left(\sqrt{1+(m \varepsilon)^{2}}\right)^{y-1-t(s)}}= \begin{cases}1, & \text { for } t(s)=y-1 \\ 0, & \text { otherwise }\end{cases} $$ where $t(s)$ is the number of turns in the path $s$. To compute $P(x, y, m \varepsilon,+)$, we take only paths with $t \equiv 0 \bmod 2$, hence we get the answer. 14. Answer. Recall the notation introduced at the beginning of section "Hints...". We have $$ \begin{aligned} & a_{1}(x, y, m \varepsilon,-)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{1}(x+1, y-1, m \varepsilon,-)+m \varepsilon a_{2}(x+1, y-1, m \varepsilon,+)\right) \\ & a_{2}(x, y, m \varepsilon,+)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{2}(x-1, y-1, m \varepsilon,+)-m \varepsilon a_{1}(x-1, y-1, m \varepsilon,-)\right) \\ & a_{2}(x, y, m \varepsilon,-)=a_{1}(x, y, m \varepsilon,+)=0 \end{aligned} $$ For each $y>|x|$ we have $$ \begin{aligned} & a_{1}(x, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r}(m \varepsilon)^{2 r+1} \\ & a_{2}(x, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \sum_{r=1}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{(x+y-2) / 2}{r}\binom{(y-x-2) / 2}{r-1}(m \varepsilon)^{2 r} \end{aligned} $$ For each $y=x>0$ we have $$ \begin{aligned} & a_{1}(y, y, m \varepsilon)=0 \\ & a_{2}(y, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \end{aligned} $$ For each $00$ we have $$ \begin{aligned} & a_{1}(x, y, m \varepsilon)=0 \\ & a_{2}(x, y, m \varepsilon)=0 \end{aligned} $$ Hint. First let us solve the analogue of Problem 4 for $m \varepsilon \neq 1$. Let us derive the formula for $a_{2}(x, y, m \varepsilon,+)$. Consider any path $s$ from $(0,0)$ to $(x, y)$. The last move in the path $s$ is made either from $(x-1, y)$ or from $(x+1, y)$. Denote by $s^{\prime}$ the path $s$ without the last move. If the directions of the last moves in $s$ and $s^{\prime}$ coincide, then $\vec{a}(s, m \varepsilon)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}} \vec{a}\left(s^{\prime}, m \varepsilon\right)$, otherwise $\vec{a}(s, m \varepsilon)=\frac{m \varepsilon}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{2}\left(s^{\prime}, m \varepsilon\right),-a_{1}\left(s^{\prime}, m \varepsilon\right)\right)$. Therefore, $a_{2}(x, y, m \varepsilon,+)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}\left(a_{2}(x-1, y-1, m \varepsilon,+)-m \varepsilon a_{1}(x-1, y-1, m \varepsilon,-)\right)$. The formula for $a_{1}(x, y, m \varepsilon,-)$ is proved analogously. Now let us solve the analogue of Problem 5: we prove that $\sum_{x \in \mathbb{Z}} P(x, y, m \varepsilon)=1$ for all $y \geq 1$ by induction over $y$ using these results. Obviously, $\sum_{x \in \mathbb{Z}} P(x, 1, m \varepsilon)=1$. The step of induction follows immediately from the following computation: $$ \begin{aligned} & \sum_{x \in \mathbb{Z}} P(x, y+1, m \varepsilon)=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y+1, m \varepsilon,-)^{2}+a_{2}(x, y+1, m \varepsilon,+)^{2}\right]=\sum_{x \in \mathbb{Z}} a_{1}(x, y+1, m \varepsilon,-)^{2}+\sum_{x \in \mathbb{Z}} a_{2}(x, y+1, m \varepsilon,+)^{2} \\ &= \frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{1}(x+1, y, m \varepsilon,-)+m \varepsilon a_{2}(x+1, y, m \varepsilon,+)\right)^{2}+\frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{2}(x-1, y, m \varepsilon,+)-m \varepsilon a_{1}(x-1, y, m \varepsilon,-)\right)^{2}= \\ &=\frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{1}(x, y, m \varepsilon,-)+m \varepsilon a_{2}(x, y, m \varepsilon,+)\right)^{2}+\frac{1}{1+m^{2} \varepsilon^{2}} \sum_{x \in \mathbb{Z}}\left(a_{2}(x, y, m \varepsilon,+)-m \varepsilon a_{1}(x, y, m \varepsilon,+)\right)^{2}= \\ &=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y, m \varepsilon,-)^{2}+a_{2}(x, y, m \varepsilon,+)^{2}\right]=\sum_{x \in \mathbb{Z}} P(x, y, m \varepsilon) . \end{aligned} $$ Let us now solve the analogue of Problem 9 for $m \varepsilon \neq 1$. Let us find $a_{1}(x, y)$ for $y>|x|$. We need to consider paths with an odd number of turns only. Denote by $2 r+1$ the number of turns in a path. Denote by $R$ and $L$ the number of upwards-right and upwards-left moves respectively. Let $x_{1}, x_{2}, \ldots, x_{r+1}$ be the number of upwards-right moves before the first, the third, $\ldots$, the last turn respectively. Let $y_{1}, y_{2}, \ldots, y_{r+1}$ be the number of upwards-left moves after the first, the third, $\ldots$, the last turn respectively. Then $x_{k}, y_{k} \geq 1$ for $1 \leq k \leq r+1$ and $$ \begin{aligned} & R=x_{1}+\cdots+x_{r+1} \\ & L=y_{1}+\cdots+y_{r+1} \end{aligned} $$ The problem now reduces to a combinatorial one: find the number of positive integer solutions of the resulting equations. For the first equation, this number equals to the number of ways to put $r$ sticks between $R$ coins in a row, that is, $\binom{R-1}{r}$. Thus $$ a_{1}(x, y, m \varepsilon)=\left(1+m^{2} \varepsilon^{2}\right)^{(1-y) / 2} \sum_{r=0}^{\lfloor y / 2\rfloor}(-1)^{r}\binom{R-1}{r}\binom{L-1}{r}(m \varepsilon)^{2 r+1} $$ Since $L+R=y$ and $R-L=x$, the required formula follows. The proof of the formula for $a_{2}(x, y)$ for $y>|x|$ is analogous. The case $y \leq|x|$ is obvious. 15. Answer: for $|x|0$. Take such $N$ that $\sum_{i=0}^{N} b_{i}>b-\varepsilon$. For each $0 \leq i \leq N$ take $M_{i}$ such that $\forall t \geq M_{i}: a_{i}(t)>b_{i}-\frac{\varepsilon}{2^{i+1}}$. Then for all $n>\max \left(M_{0}, M_{1}, \ldots, M_{N}\right)$ we have $\sum_{i=0}^{\infty} a_{i}(n)>b-\varepsilon$. Therefore, $\lim _{n \rightarrow \infty} \sum_{i=0}^{\infty} a_{i}(n)=b$. From Lemmas 5 and 3 it follows that $$ \lim _{n \rightarrow \infty} n \cdot \sum_{l=0 ; 2 \mid l}^{\infty} C_{A-1}^{l} C_{B-1}^{l}\left(\frac{m}{n}\right)^{2 l+1}=\sum_{l=0 ; 2 \mid l}^{\infty} \frac{\left(\frac{x+y}{2}\right)^{l}\left(\frac{y-x}{2}\right)^{l} \cdot m^{2 l+1}}{(l!)^{2}} $$ Therefore, by Problem 14 and Lemma 2 we have $$ \begin{aligned} \lim _{n \rightarrow \infty} n \vec{a}\left(x_{n}, y_{n}, \frac{m}{n},-\right) & =\lim _{n \rightarrow \infty} n \cdot \sum_{l=0}^{\infty} n(-1)^{l} C_{A-1}^{l} C_{B-1}^{l}\left(\frac{m}{n}\right)^{2 l+1}= \\ & =\left(\sum_{l=0 ; 2 \mid l}^{\infty} \cdot \frac{\left(\frac{x+y}{2}\right)^{l}\left(\frac{y-x}{2}\right)^{l} \cdot m^{2 l+1}}{(l!)^{2}}\right)-\left(\sum_{l=0 ; 2 \nmid l}^{\infty} \cdot \frac{\left(\frac{x+y}{2}\right)^{l}\left(\frac{y-x}{2}\right)^{l} \cdot m^{2 l+1}}{(l!)^{2}}\right)=m \cdot J_{0}\left(m \sqrt{y^{2}-x^{2}}\right) \end{aligned} $$ It can be proven analogously that $\lim _{n \rightarrow \infty} n \vec{a}\left(x_{n}, y_{n}, \frac{m}{n},+\right)=-i m \frac{x+y}{\sqrt{y^{2}-x^{2}}} \cdot J_{1}\left(m \sqrt{y^{2}-x^{2}}\right)$. Solution. Denote modified Bessel functions $$ I_{0}(z):=\sum_{k=0}^{\infty} \frac{(z / 2)^{2 k}}{(k!)^{2}} \quad \text { and } \quad I_{1}(z):=\sum_{k=0}^{\infty} \frac{(z / 2)^{2 k+1}}{k!(k+1)!} $$ The answer follows from the following stronger theorem. Theorem 2. Let $0<\varepsilon_{0} \leq 1 / 2, \min \{y+x, y-x\} \geq \varepsilon_{0}, z:=m \sqrt{y^{2}-x^{2}}, n \geq n_{0}=n_{0}\left(z, \varepsilon_{0}\right)=4 e^{3 z} / \varepsilon_{0}^{2}$. Then $$ \begin{aligned} & a_{1}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},-\right)=\frac{m}{n}\left(J_{0}(z)+O\left(\frac{\log ^{2} n}{\varepsilon_{0} n} I_{0}(z)\right)\right) \\ & a_{2}\left(2\left\lfloor\frac{n x}{2}\right\rfloor, 2\left\lfloor\frac{n y}{2}\right\rfloor, \frac{m}{n},+\right)=-\frac{m}{n} \frac{y+x}{\sqrt{y^{2}-x^{2}}}\left(J_{1}(z)+O\left(\frac{\log ^{2} n}{\varepsilon_{0} n} I_{1}(z)\right)\right) \end{aligned} $$ Proof. Denote $x_{n}=2\left\lfloor\frac{n x}{2}\right\rfloor, y_{n}=2\left\lfloor\frac{n y}{2}\right\rfloor$. By Problem 9 we have $$ \begin{aligned} & a_{1}\left(x_{n}, y_{n}, \frac{m}{n},-\right)=\left(1+\frac{m^{2}}{n^{2}}\right)^{\frac{1-y_{n}}{2}} \frac{m}{n} \sum_{r=0}^{\left[y_{n} / 2\right]}(-1)^{r}\binom{\left(y_{n}+x_{n}-2\right) / 2}{r}\binom{\left(y_{n}-x_{n}-2\right) / 2}{r}\left(\frac{m}{n}\right)^{2 r} \\ & a_{2}\left(x_{n}, y_{n}, \frac{m}{n},+\right)=\left(1+\frac{m^{2}}{n^{2}}\right)^{\frac{1-y_{n}}{2}} \frac{m}{n} \sum_{r=1}^{\left[y_{n} / 2\right]}(-1)^{r}\binom{\left(y_{n}+x_{n}-2\right) / 2}{r}\binom{\left(y_{n}-x_{n}-2\right) / 2}{r-1}\left(\frac{m}{n}\right)^{2 r-1} \end{aligned} $$ Since $1+\frac{m^{2}}{n^{2}} \leq e^{m^{2} / n^{2}}$ it follows that $\left(1+\frac{m^{2}}{n^{2}}\right)^{\frac{1-y_{n}}{2}} \geq e^{\frac{m^{2}}{n^{2}}\left(\frac{1-y_{n}}{2}\right)}$ because $y_{n} \geq 1$. Also $\frac{1-y_{n}}{2} \geq \frac{-y n}{2}$. Hence, $$ 1 \geq\left(1+\frac{m^{2}}{n^{2}}\right)^{\frac{1-y_{n}}{2}} \geq e^{\frac{m^{2}}{n^{2}}\left(\frac{-y n}{2}\right)} \geq 1-\frac{m^{2} y}{2 n} $$ Thus $\left(1+\frac{m^{2}}{n^{2}}\right)^{\frac{1-y_{n}}{2}}=1+O\left(\frac{m^{2} y}{n}\right)$. To find the asymptotic form for $a_{1}$ and $a_{2}$ consider the $T$-th partial sums with $T=\lfloor\log n\rfloor+1$ summands. The inequalities $y \geq \varepsilon_{0}$ and $n \geq \frac{4}{\varepsilon_{0}^{2}}$ guarantee that the total number of summands $\left\lfloor\frac{y_{n}}{2}\right\rfloor \geq T$ because $\frac{n y}{2} \geq \sqrt{n} \geq \log n+1$ for $n>0$. For $r \geq T$ the ratios of consecutive summands in the expressions for $a_{1}$ and $a_{2}$ equal respectively $$ \begin{aligned} &\left(\frac{m}{n}\right)^{2} \frac{\left(\left(y_{n}+x_{n}-2\right) / 2-r\right)\left(\left(y_{n}-x_{n}-2\right) / 2-r\right)}{(r+1)^{2}}<\left(\frac{m}{n}\right)^{2} \cdot \frac{n(y+x)}{2 T} \cdot \frac{n(y-x)}{2 T}=\frac{z^{2}}{4 T^{2}} \\ &\left(\frac{m}{n}\right)^{2} \frac{\left(\left(y_{n}+x_{n}-2\right) / 2-r\right)\left(\left(y_{n}-x_{n}-2\right) / 2-r+1\right)}{(r-1) r}<\left(\frac{m}{n}\right)^{2} \cdot \frac{n(y+x)}{2 T} \cdot \frac{n(y-x)}{2 T-2}=\frac{z^{2}}{4 T(T-1)} \end{aligned} $$ From the condition $n>e^{3 z+1}$ it follows that $T=\lfloor\log n\rfloor+1 \geq 3 z+1$ and $\frac{z^{2}}{4 T^{2}}<\frac{z^{2}}{4 T(T-1)}<\frac{1}{2}$. Therefore, in both cases the error term (i.e., the sum over $r \geq T$ ) is less then the sum of geometric series with ratio $\frac{1}{2}$. Hence, $$ \begin{aligned} a_{1}\left(n_{x}, n_{y}, \frac{m}{n},-\right) & =\frac{m}{n}\left(1+O\left(\frac{m^{2} y}{n}\right)\right) \\ & \cdot\left[\sum_{r=0}^{T-1}(-1)^{r}\binom{\left(y_{n}+x_{n}-2\right) / 2}{r}\binom{\left(y_{n}-x_{n}-2\right) / 2}{r}\left(\frac{m}{n}\right)^{2 r}+O\left(\frac{\left(y^{2}-x^{2}\right)^{T}}{(T!)^{2}} \cdot\left(\frac{m}{2}\right)^{2 T}\right)\right] \\ a_{2}\left(n_{x}, n_{y}, \frac{m}{n},+\right) & =\frac{m}{n}\left(1+O\left(\frac{m^{2} y}{n}\right)\right) \\ & \cdot\left[\sum_{r=1}^{T-1}(-1)^{r}\binom{\left(y_{n}+x_{n}-2\right) / 2}{r}\binom{\left(y_{n}-x_{n}-2\right) / 2}{r-1}\left(\frac{m}{n}\right)^{2 r-1}+O\left(\frac{(y+x)^{T}(y-x)^{T-1}}{(T-1)!T!} \cdot\left(\frac{m}{2}\right)^{2 T-1}\right)\right] \end{aligned} $$ The error term in the latter formulae are estimated as follows. Since $T!\geq(T / 3)^{T}$ and $T \geq \sqrt{y^{2}-x^{2}} \frac{3 m \sqrt{e}}{2}=$ $\frac{3 \sqrt{e}}{2} z$, it follows that $$ \frac{\left(y^{2}-x^{2}\right)^{T}}{(T!)^{2}} \cdot\left(\frac{m}{2}\right)^{2 T} \leq \frac{\left(y^{2}-x^{2}\right)^{T}}{(T)^{2 T}} \cdot\left(\frac{3 m}{2}\right)^{2 T} \leq e^{-T} \leq \frac{1}{n} $$ Now transform binomial coefficients in the following way: $$ \binom{\alpha n-1}{r}=\frac{(\alpha n-1) \cdots(\alpha n-r)}{r!}=\frac{(\alpha n)^{r}}{r!}\left(1-\frac{1}{\alpha n}\right) \cdots\left(1-\frac{r}{\alpha n}\right) $$ If $\alpha \geq \varepsilon_{0}$ and $r $(-1,0)_{-}$ | | $(0,1)_{+}$
$(0,0)_{-}$ | | $(0,-2)_{+}$
$(0,0)_{-}$ | | $(0,2)_{+}$
$(-1,0)_{-}$ | | $(0,-1)_{+}$
$(0,0)_{-}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 2 | | $(0,0)_{+}$
$(-1,0)_{-}$ | | $(0,-1)_{+}$
$(1,0)_{-}$ | | $(0,1)_{+}$
$(-1,0)_{-}$ | | $(0,-1)_{+}$
$(0,0)_{-}$ | | | 1 | | | $(0,-1)_{+}$
$(0,0)_{-}$ | | $(0,1)_{+}$
$(0,0)_{-}$ | | $(0,-1)_{+}$
$(0,0)_{-}$ | | | | $\mathrm{y}$ x | -3 | -2 | -1 | 0 | 1 | 2 | 3 | 4 | 5 | Hint. Note that we are interested only in even $x_{0}$, thus $2 \pi x_{0} \varepsilon / \lambda=\frac{\pi x_{0}}{2}$ is a multiple of $\pi$. Analogously to Problem 1, that means that $\vec{a}(x, y, 4,3,+)$ is parallel to $O y$ and $\vec{a}(x, y, 4,3,-)$ is parallel to $O x$ (be careful: this is not true for an arbitrary $\varepsilon / \lambda$ ). 17. Let us give a more general formula; the solution of the problem is the particular case $m \varepsilon=1$. Denote $$ \begin{aligned} & a_{-}(x, y):=\vec{a}(x, y, m \varepsilon, \lambda / \varepsilon, \Delta,-) \\ & a_{+}(x, y):=\vec{a}(x, y, m \varepsilon, \lambda / \varepsilon, \Delta,+) \end{aligned} $$ where the vectors in the right-hand side are defined analogously to $\vec{a}(x, y, \lambda / \varepsilon, \Delta, \pm)$ and $\vec{a}(x, y, m \varepsilon, \pm)$. (Beware that $a_{-}(x, y)$ and $a_{+}(x, y)$ are vector-valued functions; unlike $a_{1}(x, y)$ and $a_{2}(x, y)$ they are not coordinates of any vector.) Then we have the Dirac equation (where $i:=R^{\pi / 2}$ is the counterclockwise rotation through $90^{\circ}$ ) $$ \left\{\begin{array}{l} a_{-}(x, y)=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{-}(x+1, y-1)+\frac{-i m \varepsilon}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{+}(x+1, y-1) \\ a_{+}(x, y)=\frac{-i m}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{-}(x-1, y-1)+\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}} a_{+}(x-1, y-1) \end{array}\right. $$ From Problem 14 we know that this equation holds for $\Delta=1$. It can be generalised for any greater $\Delta$ because one can change the order of rotations $R^{2 \pi x_{0} \varepsilon / \lambda}$ and $i=R^{\pi / 2}$. Using this equation, analogously to Problem 14 we get $\sum_{x \in \mathbb{Z}} P(x, y, \lambda / \varepsilon, \Delta)=1$ for $y \geq 1$ by induction over $y$. 18. Since $\Delta P(x, y, \lambda / \varepsilon, \Delta,+)=|\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)|^{2}$, it suffices to prove that $\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)$ does not depend on $\Delta$ for $\Delta>y+|x|$. By definition $\sqrt{\Delta} \vec{a}(x, y, \lambda / \varepsilon, \Delta,+)=\sum_{\substack{x_{0}=1-\Delta \\ x_{0} \text { even }}}^{x_{0}=\Delta-1} \sum_{s} R^{2 \pi x_{0} \varepsilon / \lambda} \vec{a}(s)$. But there are no paths $s$ starting at $\left(x_{0}, 0\right)$ and ending at $(x, y)$ for $\left|x-x_{0}\right|>y$, thus increasing $\Delta$ over $y+|x|$ does not change the sum. 19. Answer: if $x_{1}$ and $x_{2}$ are both even, then $$ \begin{aligned} & \vec{a}\left(x_{1}, y, \lambda / \varepsilon,+\right)=R^{\left(x_{1}-x_{2}\right) 2 \pi \varepsilon / \lambda} \vec{a}\left(x_{2}, y, \lambda / \varepsilon,+\right) \\ & \vec{a}\left(x_{1}, y, \lambda / \varepsilon,-\right)=R^{\left(x_{1}-x_{2}\right) 2 \pi \varepsilon / \lambda} \vec{a}\left(x_{2}, y, \lambda / \varepsilon,-\right) \end{aligned} $$ Solution. From the solution of Problem 18 it follows that $\vec{a}(x, y, \lambda / \varepsilon,+)=\sum_{x_{0}} \sum_{\text {even }} R^{2 \pi x_{0} \varepsilon / \lambda} \vec{a}(s)$, where the second sum is over those paths from $\left(x_{0}, 0\right)$ to $(x, y)$, which both start and finish with an upwards-right move. To each path $s_{1}$ from $\left(x_{0}, 0\right)$ to $\left(x_{1}, y\right)$ assign the unique path $s_{2}$ from $\left(x_{0}+\left(x_{2}-x_{1}\right), 0\right)$ to $\left(x_{2}, y\right)$, obtained from $s_{1}$ by translation by $\left(x_{1}-x_{2}, 0\right)$ and vice versa. Thus we sum over the same vectors, but for $x=x_{1}$ they are rotated through $2 \pi x_{0} \varepsilon / \lambda$ while for $x=x_{2}$ they are rotated through $2 \pi\left(x_{0}+\left(x_{2}-x_{1}\right)\right) \varepsilon / \lambda$, and we are done. 20. Let us give a more general formula; the solution of the problem is the particular case $m \varepsilon=1$. Answer. Denote $E=\arccos \left(\frac{\cos (2 \pi \varepsilon / \lambda)}{\sqrt{1+m^{2} \varepsilon^{2}}}\right)$. Then for even $x+y$ we have: $$ \begin{aligned} \vec{a}(x, y, \lambda / \varepsilon,-)= & \left(\frac{m \varepsilon \cos (2 \pi \varepsilon x / \lambda) \sin ((y-1) E)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E}, \frac{m \varepsilon \sin (2 \pi \varepsilon x / \lambda) \sin ((y-1) E)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E}\right) \\ \vec{a}(x, y, \lambda / \varepsilon,+)= & \left(\frac{\cos (2 \pi \varepsilon(x-2) / \lambda) \sin (y E)-\cos (2 \pi \varepsilon x / \lambda) \sin ((y-2) E)}{2 \cos (2 \pi \varepsilon / \lambda) \sin E}\right. \\ & \left.\frac{-\sin (2 \pi \varepsilon(x-2) / \lambda) \sin (y E)+\sin (2 \pi \varepsilon x / \lambda) \sin ((y-2) E)}{2 \cos (2 \pi \varepsilon / \lambda) \sin E}\right) \\ P(x, y, \lambda / \varepsilon,-)= & \frac{m^{2} \varepsilon^{2} \sin ^{2}((y-1) E)}{m^{2} \varepsilon^{2}+\sin ^{2}(2 \pi \varepsilon / \lambda)} \\ P(x, y, \lambda / \varepsilon,+)= & \frac{m^{2} \varepsilon^{2} \cos ^{2}((y-1) E)+\sin ^{2}(2 \pi \varepsilon / \lambda)}{m^{2} \varepsilon^{2}+\sin ^{2}(2 \pi \varepsilon / \lambda)} \\ P(x, y, \lambda / \varepsilon,-)+P(x, y, \lambda / \varepsilon,+)= & \end{aligned} $$ Remark. In complex form, $$ \begin{aligned} & \vec{a}(x, y, \lambda / \varepsilon,-)=r^{x} \frac{m \varepsilon \sin ((y-1) E)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E} \\ & \vec{a}(x, y, \lambda / \varepsilon,+)=r^{x-1}\left(\frac{\sin ((y-1) E) \sin (2 \pi \varepsilon / \lambda)}{\sqrt{1+m^{2} \varepsilon^{2}} \sin E}+i \cos ((y-1) E)\right) \end{aligned} $$ where $r=\cos (2 \pi \varepsilon / \lambda)+i \sin (2 \pi \varepsilon / \lambda$ ). This is the solution of the Dirac equation (see the solution of Problem 17) with the (quasi)periodic initial condition $a_{-}(x, 0)=0$ and $a_{+}(x, 0)=\left(-\sin \frac{2 \pi x \varepsilon}{\lambda}, \cos \frac{2 \pi x \varepsilon}{\lambda}\right)$. The number $E$ has physical meaning of energy. Hint. Formulas for $\vec{a}(x, y, \lambda / \varepsilon,+)$ and $\vec{a}(x, y, \lambda / \varepsilon,-)$ can be proven by induction over $y$. The base is obvious; the inductive step follows from Problem 17. Then $P(x, y, \lambda / \varepsilon,+)$ and $P(x, y, \lambda / \varepsilon,-)$ are computed directly. Path to solution. Let us view vectors as complex numbers. Then $R^{\alpha} \vec{a}=(\cos \alpha+i \sin \alpha) \vec{a}$. Denote $r=$ $\cos (2 \pi \varepsilon / \lambda)+i \sin (2 \pi \varepsilon / \lambda)$; then $r^{-1}=\cos (2 \pi \varepsilon / \lambda)-i \sin (2 \pi \varepsilon / \lambda)$. Denote $f(y)=r^{-y} \vec{a}(y, y, \lambda / \varepsilon,+)$ and $g(y)=$ $r^{-y} \vec{a}(y, y, \lambda / \varepsilon,-)$. Denote $M=\frac{m \varepsilon}{\sqrt{1+m^{2} \varepsilon^{2}}}$ and $U=\frac{1}{\sqrt{1+m^{2} \varepsilon^{2}}}$. By Problem 19 for even $x+y$ we have $\vec{a}(x, y, \lambda / \varepsilon,+)=$ $r^{x} f(y)$ and $\vec{a}(x, y, \lambda / \varepsilon,-)=r^{x} g(y)$. By Problem 17 we have $r^{x} f(y)=U r^{x-1} f(y-1)-i M r^{x-1} g(y-1)$ and $r^{x} g(y)=U r^{x+1} g(y-1)-i M r^{x+1} f(y-1)$. Since $r \neq 0$, we can divide each side by $r^{x}$ and obtain the following system: $$ \left\{\begin{array}{l} f(y)=U r^{-1} f(y-1)-i M r^{-1} g(y-1) \\ g(y)=U r g(y-1)-i M r f(y-1) \\ f(1)=i r^{-1} \\ g(1)=0 \end{array}\right. $$ There are multiple ways of solving this system (we have thought about generating functions and about diagonalising the 'transfer matrix') but the one which is probably most accessible (to math high-school students in Russia) is transforming this system into a linear recurrence relation of degree 2. From (1) we have $g(y-1)=\frac{1}{M} \operatorname{irf}(y)-$ $\frac{U}{M} i f(y-1)$. From this and (2) we have $g(y)=\frac{U}{M} i r^{2} f(y)-\frac{1}{M} i r f(y-1)$. From this and (1) we have $f(y)=$ $U\left(r+r^{-1}\right) f(y-1)-f(y-2)$. The solution of this recurrence equation is $f(n)=c_{1} d_{+}^{n}+c_{2} d_{-}^{n}$, where $$ \begin{aligned} & d_{+}=\frac{U\left(r+r^{-1}\right)+\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}}{2}=U \cos (2 \pi \varepsilon / \lambda)+i \sqrt{U^{2} \sin ^{2}(2 \pi \varepsilon / \lambda)+M^{2}} \\ & d_{-}=\frac{U\left(r+r^{-1}\right)-\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}}{2}=U \cos (2 \pi \varepsilon / \lambda)-i \sqrt{U^{2} \sin ^{2}(2 \pi \varepsilon / \lambda)+M^{2}} \end{aligned} $$ are the roots of the equation $t^{2}-U\left(r+r^{-1}\right) t+1=0$, and $c_{1}$ and $c_{2}$ are constants that are found using (3) and (4). Since we are dealing with complex numbers, the notation $\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}$ is ambiguous. But we can choose any of the two roots; $\sqrt{U^{2}\left(r+r^{-1}\right)^{2}-4}$ denotes $2 i \sqrt{U^{2} \sin ^{2}(2 \pi \varepsilon / \lambda)+M^{2}}$. It follows that $$ c_{1}=i \frac{r^{-1}-U d_{-}}{d_{+}-d_{-}} \text {and } c_{2}=i \frac{r^{-1}-U d_{+}}{d_{-}-d_{+}} $$ Note that $E=\arccos (U \cos (2 \pi \varepsilon / \lambda))$. We have $d_{+}=\cos E+i \sin E$ and $d_{-}=\cos E-i \sin E$. Thus we have $\vec{a}(x, y, \lambda / \varepsilon,+)=r^{x} f(y)=i r^{x} \frac{\left(r^{-1}-U d_{-}\right) d_{+}^{y}-\left(r^{-1}-U d_{+}\right) d_{-}^{y}}{d_{+}-d_{-}}=i r^{x-1} \frac{\sin (y E)}{\sin E}-i U r^{x} \frac{\sin ((y-1) E)}{\sin E}$. Now we compute $P(x, y, \lambda / \varepsilon,+)=\vec{a}(x, y, \lambda / \varepsilon,+) \cdot \vec{a}(x, y, \lambda / \varepsilon,+)$. It follows from formulas above that $\bar{r}=r^{-1}$, $\overline{d_{+}}=d_{-}, \overline{d_{-}}=d_{+}, \overline{c_{1}}=i \frac{r-U d_{+}}{d_{+}-d_{-}}$and $\overline{c_{2}}=i \frac{r-U d_{-}}{d_{-}-d_{+}}$. From this (with some simplification) we finally have $P(x, y, \lambda / \varepsilon,+)=\frac{\left(d_{+}-d_{-}\right)^{2}-M^{2}\left(d_{+}^{y-1}-d_{-}^{y-1}\right)^{2}}{\left(d_{+}-d_{-}\right)^{2}}=1-\frac{M^{2} \sin ^{2}((y-1) E)}{\sin ^{2} E}$. Analogously we get $\vec{a}(x, y, \lambda / \varepsilon,-)=r^{x} M \frac{\sin ((y-1) E)}{\sin E}$ and $P(x, y, \lambda / \varepsilon,-)=\frac{M^{2} \sin ^{2}((y-1) E)}{\sin ^{2} E}$. 21. Solution. Let $(0,0),(1,1),\left(x_{2}, 2\right), \cdots,\left(x_{y}, y\right)$ be the squares passed by the path $s$. To define the vector $\vec{a}(s, m(x) \varepsilon)$, start with the vector $(0,1)$. While the checker moves straightly, the vector is not changed, but if the checker changes the direction after the $n$-th move, the vector is rotated through $90^{\circ}$ clockwise and multiplied by $m\left(x_{n}\right) \varepsilon$. In addition, at the very end the vector is divided by $\prod_{n=1}^{y-1} \sqrt{1+m^{2}\left(x_{n}\right) \varepsilon^{2}}$, where $y$ is the total number of moves. The final position of the vector is what we denote by $\vec{a}(s, m(x) \varepsilon)$. The numbers $a_{1}(x, y, m(x) \varepsilon)$ and $a_{2}(x, y, m(x) \varepsilon)$ are defined analogously to $a_{1}(x, y)$ and $a_{2}(x, y)$, only $\vec{a}(s)$ is replaced by $\vec{a}(s, m(x) \varepsilon)$. The formulas analogous to the ones from Problem 4 are the following: $$ \begin{aligned} & a_{1}\left(x_{0}, y, m(x) \varepsilon\right)=\frac{1}{\sqrt{1+m^{2}\left(x_{0}+1\right) \varepsilon^{2}}}\left(a_{1}\left(x_{0}+1, y, m(x) \varepsilon\right)+m\left(x_{0}+1\right) \varepsilon \cdot a_{2}\left(x_{0}+1, y, m(x) \varepsilon\right)\right) \\ & a_{2}\left(x_{0}, y, m(x) \varepsilon\right)=\frac{1}{\sqrt{1+m^{2}\left(x_{0}-1\right) \varepsilon^{2}}}\left(a_{2}\left(x_{0}-1, y, m(x) \varepsilon\right)-m\left(x_{0}-1\right) \varepsilon \cdot a_{1}\left(x_{0}-1, y, m(x) \varepsilon\right)\right) \end{aligned} $$ 22. Answer: For each $x, y$ we have $P\left(x, y, m_{0}(x), \lambda / \varepsilon,+\right)=1$ $P\left(x, y, m_{0}(x), \lambda / \varepsilon,-\right)=0$. $P\left(x, y, m_{1}(x), \lambda / \varepsilon,+\right)= \begin{cases}\frac{1}{1.04}=\frac{25}{26}, & \text { for } 01$ we have $y=2 L<3(L-1)$ (for $L=1$ the function $f(L)$ is undefined), it follows that the number of turns must be exactly 1 . Therefore, either the path starts in the square $(0,0)$ and changes the direction in the square $(L, L)$, or the path starts in the square $(2-2 L, 0)$ and changes the direction in $(1,2 L-1)$. The vectors for these two paths are related by the rotation through the angle $\pi-\frac{\pi(L-1)}{4}$ and division by 1.04. Since the vector for the latter path has square $\frac{1}{26}$ and the one for the former one has square $\frac{1}{26 \cdot 1.04^{2}}$, by the law of cosines we get $$ P\left(0,2 L, m_{2}(x), 16,-\right)=\frac{1}{26}\left(1+\frac{1}{1.04^{2}}-\frac{2}{1.04} \cos \frac{\pi(L-1)}{4}\right) \approx 0.074\left(1-\cos \frac{\pi(L-1)}{4}\right) $$ The graph of this function is shown in the following figure (kindly provided by Punnawith Thuwajit, a participant of Summer Conference of Tournament of Towns). ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-112.jpg?height=377&width=577&top_left_y=1194&top_left_x=745) 24. Answer: $$ \begin{array}{ll} P(\lambda, L,-)=\frac{0.1664}{1.1664+\cot ^{2} \frac{2 \pi(L-1)}{\lambda}} ; & \max _{L} P(\lambda, L,-)=\frac{0.1664}{1.1664} \approx 0.14 \\ P(\lambda, L,+)=\frac{1}{1+0.1664 \sin ^{2} \frac{2 \pi(L-1)}{\lambda}} ; & P(\lambda, L,-)+P(\lambda, L,+)=1 \end{array} $$ Hint. Denote $m=0.2$ and $\phi=\frac{2 \pi}{\lambda}(L-1)$. For $y>2 L-2$ we have $$ \begin{aligned} & P\left(0, y, m_{2}(x), \lambda,-\right)= \\ & \quad=\left|e^{\frac{2 \pi i}{\lambda}(2-y)} \frac{m}{\sqrt{1+m^{2}}}\left(-1+\frac{1}{1+m^{2}} e^{2 i \phi}+\frac{m^{2}}{\left(1+m^{2}\right)^{2}}\left(e^{2 i \phi}\right)^{2}+\cdots+\frac{1}{m^{2}}\left(\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right)^{\left[\frac{y-2}{2(L-1)}\right] 2(L-1)}\right)\right|^{2} \end{aligned} $$ Then, since $\cos 2 \phi=1-2 \sin ^{2} \phi$, we have $$ \begin{gathered} P(\lambda, L,-)=\frac{m^{2}}{1+m^{2}} \cdot\left|1-\frac{1}{1+m^{2}} e^{2 i \phi} \cdot \sum_{k=0}^{\infty}\left(\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right)^{k}\right|^{2}= \\ =\frac{m^{2}}{1+m^{2}} \cdot\left|1-\frac{1}{1+m^{2}} e^{2 i \phi} \frac{1}{1-\frac{m^{2}}{1+m^{2}} e^{2 i \phi}}\right|^{2}=\frac{m^{2}}{1+m^{2}} \cdot \frac{\left|1-e^{2 i \phi}\right|^{2}}{\left|1-\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right|^{2}}= \\ =\frac{m^{2}\left(1+m^{2}\right)(2-2 \cos (2 \phi))}{\left(1+m^{2}\right)^{2}+m^{4}-2 m^{2}\left(1+m^{2}\right) \cos (2 \phi)}=\frac{4 m^{2}\left(1+m^{2}\right)}{\left(1+2 m^{2}\right)^{2}+\cot ^{2} \phi}=\frac{0.1664}{1.1664+\cot ^{2} \frac{2 \pi}{\lambda}(L-1)} \end{gathered} $$ Therefore, $\max _{L} P(\lambda, L,-)=\frac{0.1664}{1.1664} \approx 0.14$. Analogously, $$ \begin{aligned} P(\lambda, L,+) & =\left|\frac{1}{1+m^{2}} \cdot \sum_{k=0}^{\infty}\left(\frac{m^{2}}{1+m^{2}} e^{2 i \phi}\right)^{k}\right|^{2}=\left|\frac{1}{1+m^{2}} \cdot \frac{1}{1-\frac{m^{2}}{1+m^{2}} e^{2 i \phi}}\right|^{2}= \\ & =\frac{1}{\left(1+m^{2}\right)^{2}+m^{4}-2\left(1+m^{2}\right) m^{2} \cos (2 \phi)}=\frac{1}{1+4 m^{2}\left(1+m^{2}\right) \sin ^{2} \phi}=\frac{1}{1+0.1664 \sin ^{2} \frac{2 \pi}{\lambda}(L-1)} \end{aligned} $$ Note that $$ 1-\frac{1}{1+0.1664 \sin ^{2} \frac{2 \pi}{\lambda}(L-1)}=\frac{0.1664}{\frac{1}{\sin ^{2} \frac{2 \pi}{\lambda}(L-1)}+0.1664}=\frac{0.1664}{1.1664+\cot ^{2} \frac{2 \pi}{\lambda}(L-1)} $$ Hence, $P(\lambda, L,-)+P(\lambda, L,+)=1$. 25. Answer: $$ \frac{\left(n^{2}-1\right)^{2}}{\left(n^{2}+1\right)^{2}+4 n^{2} \cot ^{2} \frac{2 \pi L n}{\lambda}} \quad \text { and } \quad m=\frac{(0.3328+0.864 \sqrt{1.04}) \cdot \pi}{\lambda} $$ where $n=\sqrt{1+m \lambda / \pi}$ (the refractive index; notice that the relation between $m$ and $n$ has changed when we passed to reflection inside the glass). For the latter $m$ the limit equals $$ \frac{0.1664}{1.1664+\cot ^{2} \frac{2 \pi L(1.08+0.08 \sqrt{1.04})}{\lambda}} $$ Remark. The first formula in the answer is well-known and confirmed by experiment. A remarkable detailed popular science discussion of Problems 23-25 can be found in [Feynman], and a derivation of the equations on a physical level of rigor - in [Landafshitz, Chapter X, §86, Problem 4]. Path to solution. Consider Dirac's equation from the solution of Problem 17. Search for a solution of the form $$ \left(a_{+}(x, y), a_{-}(x, y)\right)=\left(a_{+}(x), a_{-}(x)\right) e^{-2 \pi i \varepsilon y / \lambda} $$ Use Problem 20 to guess $a_{ \pm}(x)$ separately for $x \leq 0,0 \leq x \leq L / \varepsilon$, and $x \geq L / \varepsilon$. Assume for simplicity that $1 / \varepsilon$ is an odd integer. Abbreviate $m_{3}(x)=: m(x)$. First consider the half-plane $x \leq 0$ (to the left from glass). We need to specify what is meant by a solution of Dirac's equation in the half-plane. Analogously to Problem 17, by Dirac's equation in a black square ( $x, y$ ) we mean the system $$ \left\{\begin{array}{l} a_{-}(x-1, y+1)=\frac{1}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{-}(x, y)+\frac{-i m(x) \varepsilon}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{+}(x, y) \\ a_{+}(x+1, y+1)=\frac{-i m(x) \varepsilon}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{-}(x, y)+\frac{1}{\sqrt{1+m(x)^{2} \varepsilon^{2}}} a_{+}(x, y) \end{array}\right. $$ A solution of Dirac's equation in the half-plane $x \leq 0$ is a pair $\left(a_{+}(x, y), a_{-}(x, y)\right)$ satisfying the equation in each black square in the half-plane. In particular, $a_{-}(x, y)$ is defined for $x \leq 0$, whereas $a_{+}(x, y)$ is defined for $x \leq 1$. Search for a solution of the form $\left(a_{+}(x), a_{-}(x)\right)=\left(i e^{2 \pi i x \varepsilon / \lambda}, r(\varepsilon) e^{-2 \pi i x \varepsilon / \lambda}\right)$ for some $r(\varepsilon) \in \mathbb{C}$ to be determined later, where the formulae for $a_{+}(x)$ and $a_{-}(x)$ hold for $x \leq 1$ and $x \leq 0$ respectively. (Physically, $a_{+}(x, y)$ is the incident wave, $a_{-}(x, y)$ is the reflected wave, and $\left|\lim _{\varepsilon \rightarrow 0} r(\varepsilon)\right|^{2}$ is the reflection probability.) For the half-plane $x \geq L / \varepsilon$ take $\left(a_{+}(x), a_{-}(x)\right)=\left(t(\varepsilon) e^{2 \pi i x \varepsilon / \lambda}, 0\right)$ for some $t(\varepsilon) \in \mathbb{C}$ to be determined later, where the formulae for $a_{+}(x)$ and $a_{-}(x)$ hold for $x \geq L / \varepsilon+1$ and $x \geq L / \varepsilon$ respectively. (Physically, $a_{+}(x)$ is the transmitted wave, and $\left|\lim _{\varepsilon \rightarrow 0} t(\varepsilon)\right|^{2}$ is the transmission probability.) For $0 \leq x \leq L / \varepsilon$ search for a solution of the form $$ \begin{aligned} & a_{+}(x)=a(\varepsilon) e^{2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)}+b(\varepsilon) e^{-2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)} \\ & a_{-}(x)=c(\varepsilon) e^{2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)}+d(\varepsilon) e^{-2 \pi i x \varepsilon / \lambda^{\prime}(\varepsilon)} \end{aligned} $$ for some coefficients $a, b, c, d \in \mathbb{C}$ and $\lambda^{\prime}>0$ depending on $\varepsilon$ (we have a combination of refracted and reflected waves). In what follows $a, b, c, d, \lambda^{\prime}, r, t$ denote the limits of the above numbers as $\varepsilon \rightarrow 0$. The limit $\lambda^{\prime}$ (wavelength of refracted wave) is determined by Dirac's equation as follows. Substituting the expressions for $a_{ \pm}(x)$ into Dirac's equation, canceling common factors, and taking the limit $\varepsilon \rightarrow 0$ we get $$ \begin{aligned} m a & =\left(m-\frac{2 \pi}{\lambda^{\prime}}+\frac{2 \pi}{\lambda}\right) c \\ m c & =\left(m+\frac{2 \pi}{\lambda^{\prime}}+\frac{2 \pi}{\lambda}\right) a \end{aligned} $$ We also get an analogous equation with $b$ and $d$ instead of $c$ and $a$. This implies that $\lambda^{\prime}=\lambda / n$, where $n=$ $\sqrt{1+m \lambda / \pi}$ (refractive index). Sewing together the solutions (namely, $a_{+}(x)$ at $x=1$ and $x=L / \varepsilon+1$, whereas $a_{-}(x)$ at $x=0$ and $x=L / \varepsilon$ ) gives a system of linear equations in $a, b, c, d, r, t$ : $a_{-}(0)=r(\varepsilon)=c(\varepsilon)+d(\varepsilon), \quad a_{-}(L / \varepsilon)=0=c(\varepsilon) e^{2 \pi i L / \lambda^{\prime}}+d(\varepsilon) e^{-2 \pi i L / \lambda^{\prime}}$, $a_{+}(1)=i e^{2 \pi i L / \lambda}=a(\varepsilon) e^{2 \pi i L / \lambda^{\prime}}+b(\varepsilon) e^{-2 \pi i L / \lambda^{\prime}}, \quad a_{+}(L / \varepsilon+1)=t(\varepsilon) e^{2 \pi i(L+\varepsilon) / \lambda}=a(\varepsilon) e^{2 \pi i(L+\varepsilon) / \lambda^{\prime}}+b(\varepsilon) e^{-2 \pi i(L+\varepsilon) / \lambda^{\prime}}$. Passing to the limit $\varepsilon \rightarrow 0$ and solving the resulting system of linear equations we find $a, b, c, d, r, t$ Surely, one still needs to check that even before passing to the limit, the expressions for $a_{ \pm}(x)$ determined by our equations indeed satisfy Dirac's equation in each black square. We omit that. It remains to prove that $\left(a_{+}(x), a_{-}(x)\right)=\lim _{\substack{y \rightarrow+\infty \\ y+x \text { even }}} R^{2 \pi y \varepsilon / \lambda} \vec{a}\left(x, y, m_{3}(x) \varepsilon, \frac{\lambda}{\varepsilon},-\right)$; then the number $|r|^{2}$ is the required limit. We do not know any elementary proof of that assertion and thus omit the details. Informally, the idea is to show that the limit does not depend on the initial values $a_{+}(x, 0)$ for $x>0$ and $a_{-}(x, 0)$ for $xL / \varepsilon$ and $a_{-}(x, 0)$ for $x<0$ is obvious. To prove independence on $a_{ \pm}(x, 0)$ for $0 \leq x \leq L / \varepsilon$, assume that $a_{ \pm}(x, 0)$ vanishes outside the segment $0 \leq x \leq L / \varepsilon$ and consider the linear transformation $$ \begin{aligned} U:\left(a_{-}(0,0), a_{-}(2,0), \ldots, a_{-}(L / \varepsilon-1,0)\right. & \left., a_{+}(0,0), \ldots, a_{+}(L / \varepsilon-1,0)\right) \\ & \mapsto\left(a_{-}(0,2), a_{-}(2,2), \ldots, a_{-}(L / \varepsilon-1,2), a_{+}(0,2), \ldots, a_{+}(L / \varepsilon-1,2)\right) \end{aligned} $$ Let us show that the absolute values of all the eigenvalues of $U$ are less than 1. Consider an eigenvector $v$ with the coordinates $v_{1}=a_{-}(0,0), \ldots, v_{L / \varepsilon+1}=a_{+}(L / \varepsilon-1,0)$. Let $v_{k}$ be the leftmost nonzero coordinate. If $k>1$ then $U v$ has nonzero ( $k-1$ )-th coordinate; but $U v$ is proportional to $v$, a contradiction. Thus $k=1$. Hence $a_{-}(-2,2) \neq 0$ by Dirac's equation. Then by the probability conservation $$ \begin{aligned} \|v\| & =\left|a_{-}(0,0)\right|^{2}+\left|a_{-}(2,0)\right|^{2}+\cdots+\left|a_{+}(L / \varepsilon-1,0)\right|^{2} \\ & =\left|a_{-}(-2,2)\right|^{2}+\left|a_{-}(0,2)\right|^{2}+\cdots+\left|a_{+}(L / \varepsilon-1,2)\right|^{2} \\ & >\left|a_{-}(0,2)\right|^{2}+\cdots+\left|a_{+}(L / \varepsilon-1,2)\right|^{2} \\ & =\|U v\| \end{aligned} $$ Thus the absolute value of the eigenvalue $v$ is strictly less than 1 . Thus $\left\|U^{y}\right\| \rightarrow 0$ as $y \rightarrow \infty$, which proves that the values $a_{ \pm}(x, 0)$ for $0 \leq x \leq L$ do not affect the limit. 26. Answer: In the table, the vector in the cell $(x, y)$ is $\vec{a}(x, y, u,+)$, and an empty cell $(x, y)$ means that $\vec{a}(x, y, u,+)=(0,0)$. | 4 | | | $\left(0, \frac{1}{2 \sqrt{2}}\right)$ | | $(0,0)$ | | $\left(0, \frac{1}{2 \sqrt{2}}\right)$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | | | $\left(0,-\frac{1}{2}\right)$ | | $\left(0, \frac{1}{2}\right)$ | | | 2 | | | | | $\left(0,-\frac{1}{\sqrt{2}}\right)$ | | | | 1 | | | | $(0,-1)$ | | | | | $\mathrm{y}$ x | -2 | -1 | 0 | 1 | 2 | 3 | 4 | In the following table the number in the cell $(x, y)$ is $P(x, y, u,+)$, and an empty cell $(x, y)$ means that $P(x, y, u,+)=0$. | 4 | | | $1 / 8$ | | 0 | | $1 / 8$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 3 | | | | $1 / 4$ | | $1 / 4$ | | | 2 | | | | | $1 / 2$ | | | | 1 | | | | 1 | | | | | $\mathrm{y}$ x | -2 | -1 | 0 | 1 | 2 | 3 | 4 | (The vectors $\vec{a}(x, y, u,+)$ can be easily computed using Problem 31. ) For any positive $y$ we have $\sum_{x \in \mathbb{Z}}(P(x, y, u,+)+P(x, y, u,-))=1$. (This is one of the statements of Problem 31.) 27. Answer (kindly provided by Gleb Minaev and Ivan Russkikh, participants of Summer Conference of Tournament of Towns): ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-115.jpg?height=520&width=674&top_left_y=114&top_left_x=177) 28. Solution. Suppose that $u^{\prime}$ is the result of changing the values of $u$ in all the vertices of a black square $(a, b)$. Notice that if a path $s$ neither starts nor ends in the square $(a, b)$, then it either passes through two vertices with a changed value of $u$ or does not pass through such vertices at all. Therefore, $\vec{a}(s, u)=\vec{a}\left(s, u^{\prime}\right)$. The equation $P\left(x, y, u^{\prime},+\right)=P(x, y, u,+)$ for any $(x, y) \neq(a, b)$ follows immediately. The only case left to prove is $(a, b)=(x, y)$ or $(a, b)=(0,0)$. For each path $s$ starting or ending in $(a, b)$ we have $\vec{a}\left(s, u^{\prime}\right)=-\vec{a}(s, u)$ because $s$ passes through exactly one vertex with a changed value of $u$. Therefore, $P\left(a, b, u^{\prime},+\right)=P(a, b, u,+)$. 29. Solution. The top-right vertex of the square $(x, y)$ is called vertex $(x, y)$. Denote by $u(x, y)$ the value of $u$ at the vertex $(x, y)$. Reversing a black square $(x, y)$ means changing the values of $u$ at all the four vertices of the square. Notice that reversing a black square does not change the sign of any white squares, because the number of changed vertices in each white square is even. If all values of a field in a rectangle are positive, then all the white squares inside the rectangle are positive. Therefore, it is impossible to make $u$ identically +1 by reversing black squares, if initially there is a negative white square. Suppose that there are no negative white squares in a rectangle $M \times N$, where $N$ is the height of the rectangle. Let the vertex $(0,0)$ be at the bottom-left of the rectangle. Let us prove that $u$ can be made identically +1 through reversal of black squares by induction over $N$. Base: $N=0$. Let us describe the algorithm for making $u$ identically +1 in the rectangle $M \times 0$ (i.e in vertices $(0,0),(1,0), \ldots,(M, 0))$. Note that for all $k$ we can reverse either square $(k+1,0)$ or square $(k+1,1)$, thereby we can change the signs of the values of $u$ in vertices $(k, 0)$ and $(k+1,0)$. Denote this operation reversing $k$. The algorithm is as follows: Find the minimum $0 \leq k \leq M$ such that $u(k, 0)=-1$, and reverse $k$. If after that there still exists $0 \leq k \leq M$ such that $u(k, 0)=-1$, than the minimum such $k$ is at least 1 more than before using the operation. Therefore, after at most $M+1$ operations we make $u$ to be identically +1 in the rectangle $M \times 0$. Step: Suppose that the statement is true for $N$ and prove it for $N+1$ by giving an explicit algorithm. First make $u$ identically +1 in the lower rectangle $M \times N$ by applying the algorithm for $N$. Then for each white square $(k, N+1)$ in the rectangle reverse the square $(k, N+2)$ if $u(k-1, N+1)=u(k, N+1)=-1$, otherwise do nothing. Finally, if the square $(0, N+1)$ is black and $u(-1, N+1)=-1$ then reverse the square $(-1, N+2)$, and if the square $(M, N+1)$ is black and $u(M, N+1)=-1$ then reverse the square $(M+1, N+2)$. The result is a field, positive in the whole rectangle $(M+1) \times N$. 30. Answer: Zero, if one of the sides of the rectangle is 1 , and any even number no greater than the number of white squares in the rectangle, if both sides are greater than 1. Hint. Denote by $n(x, y, u)$ the number of vertices of the square $(x, y)$ where the value of $u$ is negative. Notice that any vertex that is strictly inside the rectangle (not on the boundary) is a vertex of exactly four squares in that rectangle, two of which are white. Therefore, the sum of $n(x, y, u)$ over all white squares in the rectangle is even. Hence the number of negative white squares is even. Let us show that any even number no greater than the number of white squares in the rectangle is indeed possible. If both the length and the width of the rectangle are even, then for any field that is positive everywhere in the rectangle except for $n$ vertices with both coordinates even there are exactly $2 n$ negative white squares. For odd length or width (greater than 1) see the figure below. The identically +1 field has no negative white squares. The field $u_{2 n+2}$ with $2 n+2$ negative white squares is obtained from the field $u_{2 n}$ by changing all the values of $u_{2 n}$ to the opposite in any single encircled area where they have not been changed before. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-116.jpg?height=746&width=945&top_left_y=61&top_left_x=153) 31. Answer. Recall that $u(x, y)$ is the value of $u$ in the top-right vertex of the square $(x, y)$. Then $$ \begin{aligned} & a_{1}(x, y, u)=\frac{1}{\sqrt{2}} u(x, y-1)\left(a_{1}(x+1, y-1, u)+a_{2}(x+1, y-1, u)\right) \\ & a_{2}(x, y, u)=\frac{1}{\sqrt{2}} u(x-1, y-1)\left(-a_{1}(x-1, y-1, u)+a_{2}(x-1, y-1, u)\right) \end{aligned} $$ Hint. The solution of the analogue of Problem 4 is very similar to the original one, only the factors $u(x, y-1)$ and $u(x-1, y-1)$ are added due to the last step of the path $s$ passing through the top-right vertex of either $(x, y-1)$ or $(x-1, y-1)$. Let us prove the analogue of Problem 5 by induction over $y$. The step of induction follows immediately from the following computation: $$ \begin{gathered} \sum_{x \in \mathbb{Z}} P(x, y+1, u)=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y+1, u)^{2}+a_{2}(x, y+1, u)^{2}\right]=\sum_{x \in \mathbb{Z}} a_{1}(x, y+1, u)^{2}+\sum_{x \in \mathbb{Z}} a_{2}(x, y+1, u)^{2}= \\ =\frac{1}{2} \sum_{x \in \mathbb{Z}} u(x, y)^{2}\left(a_{1}(x+1, y, u)+a_{2}(x+1, y, u)\right)^{2}+\frac{1}{2} \sum_{x \in \mathbb{Z}} u(x-1, y)^{2}\left(a_{2}(x-1, y, u)-a_{1}(x-1, y, u)\right)^{2}= \\ =\sum_{x \in \mathbb{Z}} \frac{\left(a_{1}(x, y, u)+a_{2}(x, y, u)\right)^{2}}{2}+\sum_{x \in \mathbb{Z}} \frac{\left(a_{2}(x, y, u)-a_{1}(x, y, u)\right)^{2}}{2}=\sum_{x \in \mathbb{Z}}\left[a_{1}(x, y, u)^{2}+a_{2}(x, y, u)^{2}\right]=\sum_{x \in \mathbb{Z}} P(x, y, u) . \\ \text { 32. Answer: } P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)= \begin{cases}P(x, y,-) P\left(x^{\prime}-x_{0}, y,-\right), & \text { if } E=(x+1, y), E^{\prime}=\left(x^{\prime}+1, y\right), \\ P(x, y,-) P\left(x^{\prime}-x_{0}, y,+\right), & \text { if } E=(x+1, y), E^{\prime}=\left(x^{\prime}-1, y\right), \\ P(x, y,+) P\left(x^{\prime}-x_{0}, y,-\right), & \text { if } E=(x-1, y), E^{\prime}=\left(x^{\prime}+1, y\right), \\ P(x, y,+) P\left(x^{\prime}-x_{0}, y,+\right), & \text { if } E=(x-1, y), E^{\prime}=\left(x^{\prime}-1, y\right) .\end{cases} \end{gathered} $$ Solution. Notice that $\vec{a}\left(s, s^{\prime}\right)=-i \vec{a}(s) \vec{a}\left(s^{\prime}\right)$, where $\vec{a} \vec{b}$ denotes the product of complex numbers $\vec{a}$ and $\vec{b}$. Without loss of generality suppose that $x^{\prime}>x$. Due to the condition $x_{0} \geq 2 y$ there are no paths starting at $A$ and ending in $F^{\prime}$ and no paths starting at $A^{\prime}$ and ending in $F$. Therefore $\sum_{\substack{s: A B \rightarrow E^{\prime} F^{\prime} \\ s^{\prime}: A^{\prime} B^{\prime} \rightarrow E F}} \vec{a}\left(s, s^{\prime}\right)=0$ and $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=\sum_{\substack{s: A B \rightarrow E F \\ s^{\prime}: A^{\prime} \rightarrow B^{\prime} \rightarrow E^{\prime} F^{\prime}}} \vec{a}\left(s, s^{\prime}\right)$. If $E=(x-1, y)$ and $E^{\prime}=\left(x^{\prime}-1, y\right)$ then $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=-i \sum_{s, s^{\prime}} \vec{a}(s) \vec{a}\left(s^{\prime}\right)$, where the sum is over $s$ and $s^{\prime}$ ending with an upwards-right move. Therefore, $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=-i \vec{a}(x, y,+) \vec{a}\left(x^{\prime}-x_{0}, y,+\right)$. Hence $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=P(x, y,+) P\left(x^{\prime}-x_{0}, y,+\right)$. If $E=(x+1, y)$ and $E^{\prime}=\left(x^{\prime}-1, y\right)$ then $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=-i \sum_{s, s^{\prime}} \vec{a}(s) \vec{a}\left(s^{\prime}\right)$, where the sum is over $s$ ending with an upwards-left move and $s^{\prime}$ ending with an upwards right move. Hence $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=$ $P(x, y,-) P\left(x^{\prime}-x_{0}, y,+\right)$. If $E=(x-1, y)$ and $E^{\prime}=\left(x^{\prime}+1, y\right)$ then $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=-i \sum_{s, s^{\prime}} \vec{a}(s) \vec{a}\left(s^{\prime}\right)$, where the sum is over $s$ ending with an upwards-right move and $s^{\prime}$ ending with an upwards left move. Hence $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=$ $P(x, y,+) P\left(x^{\prime}-x_{0}, y,-\right)$. If $E=(x+1, y)$ and $E^{\prime}=\left(x^{\prime}+1, y\right)$ then $\vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=-i \sum_{s, s^{\prime}} \vec{a}(s) \vec{a}\left(s^{\prime}\right)$, where the sum is over $s$ and $s^{\prime}$ ending with an upwards-left move. Hence $P\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)=P(x, y,-) P\left(x^{\prime}-x_{0}, y,-\right)$. 33. Hint. Suppose that paths $s: A B \rightarrow E F$ and $s^{\prime}: A^{\prime} B^{\prime} \rightarrow E^{\prime} F^{\prime}$ have a common move. Suppose that $C D$ is the first common move of $s$ and $s^{\prime}$. Denote by $\tilde{s}$ the path that coincides with $s$ up to the square $C$ and with $s^{\prime}$ after $C$. and $\tilde{s^{\prime}}$ the path that coincides with $s^{\prime}$ before $C$ and with $s$ after. The paths $\tilde{s}, \tilde{s}^{\prime}$ start with the moves $A B, A^{\prime} B^{\prime}$ and end with moves $E^{\prime} F^{\prime}$ and $E F$ respectively. The total number of turns in the paths $s$ and $s^{\prime}$ equals the total number of turns in $\tilde{s}$ and $\tilde{s}^{\prime}$, because $C D$ is a common move for all of them. Therefore, $\vec{a}\left(s, s^{\prime}\right)=\vec{a}\left(\tilde{s}, \tilde{s}^{\prime}\right)$. Notice that $f:\left(s, s^{\prime}\right) \mapsto\left(\tilde{s}, \tilde{s}^{\prime}\right)$ is a bijection, because $C D$ is exactly the first common move. One of $\left(s, s^{\prime}\right)$ and $\left(\tilde{s}, \tilde{s}^{\prime}\right)$ contributes to $\vec{a}\left(E F, E^{\prime} F^{\prime}\right)$ with the coefficient $(+1)$ and the other one with the the coefficient $(-1)$. Therefore, these pairs cancel each other. 34. Hint. Let us prove this equality by induction over $y$. The base is obvious. Note that by analogy with Problem $11, \vec{a}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ is parallel to either the line $x=0$ or the line $y=0$. Denote by $a_{1}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ and $a_{2}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$ its coordinates. Fix the starting moves $A B$ and $A^{\prime} B^{\prime}$ of the paths. Denote by $a\left(E, E^{\prime}, \uparrow,+,-\right)=a_{1}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right)$, where $E F$ is an upwards-right move, $E^{\prime} F^{\prime}$ is an upwards-left move, and $a_{1}\left(A B, A^{\prime} B^{\prime} \rightarrow E F, E^{\prime} F^{\prime}\right.$ ) is the (possibly) non-zero coordinate of the corresponding vector. $a\left(E, E^{\prime}, \uparrow,+,+\right), a\left(E, E^{\prime}, \uparrow,-,+\right)$ and $a\left(E, E^{\prime}, \uparrow,-,-\right)$ are defined analogously. Denote $a\left(E, E^{\prime}, \downarrow,+,-\right)=a_{1}\left(A B, A^{\prime} B^{\prime} \rightarrow D E, D^{\prime} E^{\prime}\right)$, where $D E$ is an upwards-right move and $D^{\prime} E^{\prime}$ is an upwards-left move. $a\left(E, E^{\prime}, \downarrow,+,+\right), a\left(E, E^{\prime}, \downarrow,-,+\right)$ and $a\left(E, E^{\prime}, \downarrow,-,-\right)$ are defined analogously. Note that the induction step is a proof of the equality $$ \begin{aligned} & \sum_{E, E^{\prime}}\left(\left(a\left(E, E^{\prime}, \uparrow,+,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \uparrow,+,-\right)\right)^{2}+\left(a\left(E, E^{\prime}, \uparrow,-,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \uparrow,-,-\right)\right)^{2}\right)= \\ = & \sum_{E, E^{\prime}}\left(\left(a\left(E, E^{\prime}, \downarrow,+,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \downarrow,+,-\right)\right)^{2}+\left(a\left(E, E^{\prime}, \downarrow,-,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \downarrow,-,-\right)\right)^{2}\right) \end{aligned} $$ We will prove this equation by proving that for every pair $\left(E, E^{\prime}\right)$, where $E$ and $E^{\prime}$ have the same $y$-coordinate, $$ \begin{aligned} & \left(a\left(E, E^{\prime}, \uparrow,+,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \uparrow,+,-\right)\right)^{2}+\left(a\left(E, E^{\prime}, \uparrow,-,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \uparrow,-,-\right)\right)^{2}= \\ = & \left(a\left(E, E^{\prime}, \downarrow,+,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \downarrow,+,-\right)\right)^{2}+\left(a\left(E, E^{\prime}, \downarrow,-,+\right)\right)^{2}+\left(a\left(E, E^{\prime}, \downarrow,-,-\right)\right)^{2} \end{aligned} $$ Analogously to Problem 4 we obtain $$ \begin{aligned} & a\left(E, E^{\prime}, \uparrow,+,+\right)=\frac{1}{2}\left(a\left(E, E^{\prime}, \downarrow,+,+\right)-a\left(E, E^{\prime}, \downarrow,+,-\right)-a\left(E, E^{\prime}, \downarrow,-,+\right)-a\left(E, E^{\prime}, \downarrow,-,-\right)\right) \\ & a\left(E, E^{\prime}, \uparrow,+,-\right)=\frac{1}{2}\left(a\left(E, E^{\prime}, \downarrow,+,+\right)-a\left(E, E^{\prime}, \downarrow,+,-\right)+a\left(E, E^{\prime}, \downarrow,-,+\right)+a\left(E, E^{\prime}, \downarrow,-,-\right)\right) \\ & a\left(E, E^{\prime}, \uparrow,-,+\right)=\frac{1}{2}\left(a\left(E, E^{\prime}, \downarrow,+,+\right)+a\left(E, E^{\prime}, \downarrow,+,-\right)-a\left(E, E^{\prime}, \downarrow,-,+\right)+a\left(E, E^{\prime}, \downarrow,-,-\right)\right) \\ & a\left(E, E^{\prime}, \uparrow,-,-\right)=\frac{1}{2}\left(-a\left(E, E^{\prime}, \downarrow,+,+\right)-a\left(E, E^{\prime}, \downarrow,+,-\right)-a\left(E, E^{\prime}, \downarrow,-,+\right)+a\left(E, E^{\prime}, \downarrow,-,-\right)\right) \end{aligned} $$ With this, we can prove our equation by substitution and expansion. 35. Answer: $P\left(A B, B^{\prime} A^{\prime} \rightarrow E F, F^{\prime} E^{\prime}\right)= \begin{cases}P(x, y,-) P\left(x_{0}-x^{\prime}, y,-\right), & \text { if } E=(x+1, y), E^{\prime}=\left(x^{\prime}+1, y\right), \\ P(x, y,-) P\left(x_{0}-x^{\prime}, y,+\right), & \text { if } E=(x+1, y), E^{\prime}=\left(x^{\prime}-1, y\right), \\ P(x, y,+) P\left(x_{0}-x^{\prime}, y,-\right), & \text { if } E=(x-1, y), E^{\prime}=\left(x^{\prime}+1, y\right), \\ P(x, y,+) P\left(x_{0}-x^{\prime}, y,+\right), & \text { if } E=(x-1, y), E^{\prime}=\left(x^{\prime}-1, y\right) .\end{cases}$ Hint. Notice that $\vec{a}\left(s, s^{\prime}\right)=i \vec{a}(s) \vec{a}^{*}\left(s^{\prime}\right)$, where $\vec{a} \vec{b}$ denotes the product of complex numbers $\vec{a}$ and $\vec{b}$, and $\vec{a}^{*}$ denotes conjugation. For the remaining problems, we only give preliminary draft solutions. 36. Answer: for each $u_{\mathrm{fin}}$ we have $$ \begin{array}{lr} P\left(2,2, u_{\mathrm{fin}}, 1,+\right)=1 / 4, & P\left(0,2, u_{\mathrm{fin}}, 1,-\right)=1 / 4 \\ P\left(3,3, u_{\mathrm{fin}}, 1,+\right)=1 / 32, & P\left(1,3, u_{\mathrm{fin}}, 1,-\right)=1 / 32 \\ P\left(1,3, u_{\mathrm{fin}}, 1,+\right)=1 / 32, & P\left(-1,3, u_{\mathrm{fin}}, 1,-\right)=1 / 32 \end{array} $$ all the other probabilities vanish for $y=2,3$. 37. Answer: $\sum_{u} \vec{a}(g, s, u)=2^{3 y-4}\left(1+g^{2}\right)^{\frac{y-1}{2}} \cos ((r-l) \arctan (g)) \vec{a}(s)$. Hint. For brevity, denote $Z=\frac{1}{2^{(y-2)^{2}}\left(1+g^{2}\right)^{\frac{(y-1)^{2}}{2}}}$. Suppose that $r$ and $l$ are the numbers of white squares in $R$ to the right and to the left from $s$ respectively. Notice that $r$ and $l$ are odd if and only if the number of "-" signs on the path $s$ is odd (that is, $\vec{a}(s, u)=-\vec{a}(s))$. Easy-to-prove lemma: if $b$ is the number of black squares in $R$ not touching the boundary, then the number of fields $u$ with exactly $2 k$ negative white squares equals $2^{b}$ times the number of ways to assign "-" to $2 k$ white squares in $R$. The factor $2^{b}$ is the number of gauge transformations from the Problem 28 . By the lemma we have $$ \begin{array}{r} \sum_{u} \vec{a}(g, s, u)=Z \sum_{u}(-i g)^{n} \vec{a}(s, u)=2^{b} Z\left[\sum_{k, m \text { even }}(-i g)^{k+m}\binom{r}{k}\binom{l}{m}-\sum_{k, m \text { odd }}(-i g)^{k+m}\binom{r}{k}\binom{l}{m}\right] \vec{a}(s)= \\ =2^{b-2} Z\left[\left((1+i g)^{r}+(1-i g)^{r}\right)\left((1+i g)^{l}+(1-i g)^{l}\right)-\left((1+i g)^{r}-(1-i g)^{r}\right)\left((1+i g)^{l}-(1-i g)^{l}\right)\right] \vec{a}(s)= \\ =2^{b} Z R e\left((1+i g)^{r}(1-i g)^{l}\right) \vec{a}(s)=2^{b}\left(1+g^{2}\right)^{\frac{r+l}{2}} Z \cos ((r-l) \arctan (g)) \vec{a}(s) \end{array} $$ Since $b=r+l=y^{2}-y$, it follows that $\sum_{u} \vec{a}(g, s, u)=2^{3 y-4}\left(1+g^{2}\right)^{\frac{y-1}{2}} \cos ((r-l) \arctan (g)) \vec{a}(s)$. 38. Answer. Denote $$ \begin{array}{ll} a_{--}(x, y)=a_{2}\left(x, y, u_{-}, g,-\right), & a_{-+}(x, y)=a_{1}\left(x, y, u_{-}, g,+\right) \\ a_{+-}(x, y)=a_{1}\left(x, y, u_{+}, g,-\right), & a_{++}(x, y)=a_{2}\left(x, y, u_{+}, g,+\right) \end{array} $$ Denote by $R(y)=R$ the rectangle formed by all the squares $\left(x^{\prime}, y^{\prime}\right)$ such that $1-yА. Скопенков, М. Фёдоров, Н. Хорошавкина ${ }^{\dagger}$ ## Аннотация. Мы определим простые инварианты узлов и зацеплений (коэффициент зацепления, инварианты Арфа и Кэссона, полиномы Александера-Конвея), мотивированные интересными результатами, формулировки которых доступны неспециалистам или школьникам (например, теоремы 1.1.3 и 1.2.2). Мы покажем как простейшие инварианты появляются естественным образом в результате попыток развязать узел или расцепить зацепление. Затем мы представим некоторые рекурсивные скейн-соотношения для простейших инвариантов, которые позволяют ввести более сильные инварианты. Мы сформулируем теорему Васильева-Концевича способом, удобным для подсчета самих инвариантов, а не только размерности пространства инвариантов. Мы также представим инварианты раскрасок, для определения которых мы не знаем элементарной мотивации. Мы дадим строгие определения основных понятий, не препятствующие их интуитивному пониманию. Для изучения этого текста не требуется никаких предварительных знаний. ## Рекомендации участникам. Если условие задачи является формулировкой утверждения, то в задаче требуется это утверждение доказать. Если задача выделена словом «теорема» («лемма», «следствие» и т. д.), то её утверждение более важное. Как правило, мы приводим (в виде задачи) формулировку красивого или важного утверждения перед его доказательством. В таких случаях для доказательства утверждения могут потребоваться последующие задачи. Мы не лишаем Вас удовольствия самостоятельно найти момент, когда Вы наконец-то сможете доказать такое утверждение. Вообще, если Вы застряли на какой-то задаче, попробуйте перейти к следующим, они могут оказаться полезными. Замечания формально не используются в дальнейшем. В тексте определения важных понятий помечены жирным шрифтом, чтобы затем было проще их найти. Мы рекомендуем всем участником, работающим над проектом, консультироваться с жюри по любым вопросам, касающимся проекта. Те, кто успешно работают над проектом, получат интересные дополнительные задачи (пункты, помеченные звездочками). Участник (или группа участников) конференции, решающий задачи проекта, получает «боб» за каждое записанное решение, оцененное в «+» или «+.». Дополнительные бобы могут выдаваться за красивые решения, решения сложных задач или оформление некоторых решений в *Мы благодарны А. Рябичеву, А. Сосинскому и С. Чмутову за полезные обсуждения, а также В. Прасолову и издательству МЦНМО за возможность использовать некоторые задачи и рисунки из [Pr95]. Этот текст основан на курсе лекций А. Скопенкова, прочитанном в Независимом Московском Университете (включая программу Math in Moscow) и Московском Физико-Техническом Институте. ${ }^{\dagger}$ А. Глебов: Новосибирский Государственный университет. Д. Елисеев: Самарский университет. А. Еннэ: Петрозаводский Государственный университет. P. Живалевич: http://www.rade-zivaljevic.appspot.com/, Институт математики в САНУ, Белград. Е. Морозов, М. Фёдоров, Н. Хорошавкина: Высшая Школа Экономики, Москва. A. Скопенков: http://www.mccme.ru/ skopenko, Московский Физико-Технический Институт, Независимый Московский Университет. системе $\mathrm{TE}_{\mathrm{E}} \mathrm{X}$. У жюри бесконечно много бобов. Решения можно сдавать и устно, отдавая один боб за каждые пять попыток (не важно, удачных или нет). Пожалуйста, сообщите нам, если Вы знаете решения каких-то из предложенных задач. Если Вы подтвердите свои знания, сообщив нам решения некоторых из них, Вам будет разрешено пользоваться ими при решении остальных, не получая по ним плюсы. ## 1 Задачи до промежуточного финиша ## 1.1 Основные определения и результаты об узлах Начнем с неформального описания основных понятий (строгие определения даны после задачи 1.1.1). Узел можно представить себе как тонкую эластичную веревку, концы которой склеены - см. рис. 1. Как и на этом рисунке, узлы обычно изображаются с помощью их «удобной» проекции на плоскость, которая называется диаграммой узла. Представьте, что вы раскладываете веревку на столе, тщательно записывая как она пересекает себя (т.е. какая часть веревки лежит сверху). Необходимо помнить, что проекции одного и того же узла на разные плоскости могут иметь разный вид. Тривиальный узел - это контур (граница) треугольника. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-122.jpg?height=438&width=1590&top_left_y=1081&top_left_x=316) Рис. 1: Узлы, изотопные трилистнику (верхний ряд) и восьмерке (нижний ряд) Под изотопией узла понимается его непрерывная деформация в пространстве как тонкой эластичной нити. В процессе деформации самопересечения не допускаются. Два узла называются изотопными, если один из них можно преобразовать в другой с помощью изотопии. В качестве доказательства изотопности конкретных узлов жюри принимает большую и понятную картинку или эксперимент с веревкой, воспроизводимый членом жюри. Задача 1.1.1. (а) Некоторые два узла из верхнего ряда рис. 1 изотопны самому левому узлу. Для одного из этих двух узлов представьте изотопию с помощью движений Рейдемайстера (см. рис. 5). (b)* Все узлы, представленные в верхнем ряду рис. 1 изотопны друг другу. $\left(c, \mathrm{~d}^{*}\right)$ То же верно для узлов в нижнем ряду рис. 1. (e) Все узлы с одинаковыми диаграммами изотопны. Замечание. Здесь мы покажем необходимость строгого определения изотопии. На рис. 2 показана изотопия между трилистником и тривиальным узлом. Действительно ли это изотопия? Это так называемая «кусочно-линейная необъемлемая изотопия», которая omличается от «кусочно-линейной объемлемой изотопии», определенной и используемой далее. (Первое понятие лучше отражает идею непрерывной деформации без самопересечений, но ма- ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-123.jpg?height=228&width=1718&top_left_y=81&top_left_x=168) Рис. 2: Необъемлемая изотопия между трилистником и тривиальным узлом лодоступно для старшеклассников, cp. [Sk16i].) На самом деле, любые два узла кусочно-линейно необъемлемо изотопны! Обычная проблема с интуитивно понятными определениями не в том, что их трудно формализовать, а в том, что это можно сделать несколькими способами. Узел - это пространственная замкнутая несамопересекающаяся ломаная. Диаграмма узла - это его проекция общего положения ${ }^{1}$ на плоскость ${ }^{2}$, вместе с информацией, какая часть узла «проходит под» (проход), а какая - «проходит над» (переход) в каждом конкретном перекрестке. Задача 1.1.2. Для любой диаграммы узла существует узел, проецирующийся на эту диаграмму. (Такой узел не обязательно единственный; см. впрочем задачу 1.1.1.е.) ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-123.jpg?height=232&width=896&top_left_y=1022&top_left_x=577) Рис. 3: Элементарное движение Предположим, что стороны $A C$ и $C B$ треугольника $A B C$ - ребра некоторого узла. Кроме того, предположим, что узел и часть плоскости, ограниченная треугольником $A B C$, ни в каких других точках не пересекаются. Элементарное движение $A C B \rightarrow A B-$ это замена двух ребер $A C$ и $C B$ ребром $A B$, либо обратная операция $A B \rightarrow A C B$ (рис. 3). ${ }^{3}$ Два узла $K, L$ называются (кусочно-линейно объемлемо) изотопными, если существует последовательность узлов $K_{1}, \ldots, K_{n}$ такая, что $K_{1}=K, K_{n}=L$, и каждый узел последовательности $K_{j+1}$ получен из предыдущего узла $K_{j}$ элементарным движением. Теорема 1.1.3. (а) Тривиальный узел не изотопен трилистнику. (b) Тривиалъный узел не изотопен восъмерке. (c) Трилистник не изотопен восъмерке. (d) Существует бесконечное число попарно не изотопных узлов. Эта теорема доказывается с использованием инвариантов Арфа и Кэссона, см. §1.4 и §2.1, или с использованием правилъных раскрасок, см. §2.2 (так что нет необходимости тратить много времени, доказывая эти результаты прямо сейчас). Замечание. Это замечание может быть полезным в качестве подсказки и предостережения к задачам 1.1.4 и 1.1.5. ![](https://cdn.mathpix.com/cropped/2024_06_05_666601c4e2d753147acag-123.jpg?height=43&width=1738&top_left_y=2255&top_left_x=172) единением ребер такая, что никакие три вершины ломаной $L$ не лежат на одной прямой и никакие три отрезка, соединяющие некоторые вершины ломаной $L$, не имеют общей внутренней точки. ${ }^{2}$ Терминология университетской математики - «образ общего положения при проекции на плоскость». ${ }^{3}$ Если треугольник $A B C$ вырожденный, то элементарное движение является либо разбиением ребра, либо обратной этому операцией. [^20]: ${ }^{4}$ Эта задача будет мотивировкой для введения Арф-инварианта (§1.4). Доказательство иллюстрирует в малых размерностях основные идеи выдающегося доказательства Зимана многомерной теоремы о незаузленности сфер, см. обзор [Sk16c, Theorem 2.3]. 5Эта простая лемма будет использована для индуктивного построения инвариантов, использующих скейнсоотношения, см. далее. ${ }^{6}$ Строгое определение первого движения Рейдемейстера легко дается с помощью рис. 6 (слева). У других движений Рейдемейстера есть аналогичные строгие определения. От участников не требуется использование этих строгих определений в решении задач. Вы можете использовать неформальное описание движений Рейдемейстера на рис. 5 и пропустить движения изотопии плоскости. [^21]: ${ }^{7}$ Более точно, рассмотрим непересекающиеся ориентированные диаграммы двух ориентированных узлов. Найдем в плоскости прямоугольник, одна пара сторон которого совпадает с некоторыми ребрами этих узлов, а остальная часть прямоугольника с узлами не пересекается, и эти ребра имеют одинаковую ориентацию вдоль контура прямоугольника. Теперь соединим две диаграммы, удалив эти ребра из узлов и добавив в качестве ребер отрезки, образующие другую пару сторон прямоугольника. Получившаяся диаграмма связной суммы имеют ориентацию, согласующуюся с ориентацией исходных диаграмм. [^22]: ${ }^{8}$ Теорема 1.4.3 является аналогом задачи 2.3 .1 для однокомпонентных зацеплений (узлов). Определение arf, данное в 1.4 применимо только к узлам, и наша цель здесь - обобщить его на трехкомпонентные зацепления. [^23]: ${ }^{9}$ Не вполне ясно, которая из двух версий сформулирована в [CDM12, §2.3.1], так что мы приводим обе версии и выводим более строгую версию из более слабой. [^24]: ${ }^{10}$ Это также показывает, что наличие движений изотопии плоскости в утверждении из $\$ 1.7$ [PS96] не делает само утверждение строгим, поэтому этого следует избегать. На интуитивном уровне движения изотопии плоскости следует пропускать. С альтернативным строгим определением ниже движения изотопии плоскости могут быть выражены с помощью движений Рейдемейстера и поэтому их следует опустить в утвреждении. Мы предложим альтернативное строгое определение первого движения Рейдемейстера. У других движений Рейдемейстера есть аналогичные строгие определения. Возьмем на плоскости замкнутую несамопересекающуюся ломаную $L$, внутренность которой (см. теорему Жордана в замечании после теоремы 1.1.3) пересекает диаграмму $D$ по несамопересекающейся ломаной $M$, соединяющей две точки на $L$. Пусть $N$ - замкнутая несамопересекающаяся ломаная во внутренности $L$ такая, что $N \cap L=\emptyset, N \cap M-$ одна точка, и $M \cup N$ можно сделать (несамопересекающейся) ломаной общего положения. Первое движение Рейдемейстера - это замена $M$ на $M \cup N$ в $D$ с сохранением «информации» в соответствующем перекрестке. [^25]: ${ }^{1}$ A polygonal line in the plane is generic if there is a polygonal line $L$ with the same union of edges such that no three vertices of $L$ belong to any line and no three segments joining some vertices of $L$ have a common interior point. ${ }^{2}$ A university-mathematics terminology is 'a generic image under projection onto a plane'. ${ }^{3}$ If the triangle $A B C$ is degenerate, then elementary move is either subdivision of an edge or inverse operation. [^26]: ${ }^{4}$ This problem would be a motivation for introduction of the Arf invariant (§1.4). The proof illustrates in low dimensions one of the main ideas of the celebrated Zeeman's proof of the higher-dimensional Unknotting Spheres Theorem, see survey [Sk16c, Theorem 2.3]. ${ }^{5}$ This simple lemma will be used for inductive construction of invariants using skein relations, see below. ${ }^{6}$ A rigorous definition of the first Reidemeister move is easily given using fig. 6 (left). The other Reidemeister moves have analogous rigorous definitions. The participants are not required to use these rigorous definitions in solutions. You can use informal description of Reidemeister moves in fig. 5 and so ignore plane isotopy moves. [^27]: ${ }^{7}$ More precisely, consider disjoint oriented plane diagrams of the two oriented knots. Find a rectangle in the plane where one pair of sides are edges of each knot, but the rectangle is otherwise disjoint from the knots, and the edges are oriented around the outline of the rectangle in the same direction. Now join the two diagrams together by deleting these edges from the knots and adding the edges that form the other pair of sides of the rectangle. The resulting connected sum diagram inherits an orientation consistent with the orientations of the two original diagrams. [^28]: ${ }^{8}$ Theorem 1.4.3 is the analogue of problem 2.3 .1 for 1-component links (knots). The definition of arf given in $\S 1.4$ applies to knots only and here the point is to extend it to 3 -component links. [^29]: ${ }^{9}$ It is not clear which of the two versions is stated in [CDM12, §2.3.1], so we present both versions and deduce the stronger version from the weaker version. [^30]: ${ }^{10}$ This also shows that having plane isotopy in the statement [PS96, §1.7] does not make the statement rigorous, and thus should be avoided. On an intuitive level, plane isotopies should better be ignored. With the alternative rigorous definition below, plane isotopies can be expressed via Reidemeister moves and so should better be ignored in the statement. Let us present an alternative rigorous definition of the first Reidemeister move. The other Reidemeister moves have analogous rigorous definitions. On the plane take a closed non-self-intersecting polygonal line $L$ whose interior (see the Jordan Theorem in remark after theorem 1.1.3) intersects a knot diagram $D$ by a non-self-intersecting polygonal line $M$ joining two points on $L$. Let $N$ be a closed non-self-intersecting polygonal line in the interior of $L$ such that $N \cap L=\emptyset, N \cap M$ is one point and $M \cup N$ cam be made a generic (self-intersecting) polygonal line. The first Reidemeister move is replacement of $M$ to $M \cup N$ in $D$, with any 'information' at the appearing crossing. [^31]: ${ }^{1}$ На самом деле можно положить $m=11$ [^32]: ${ }^{1}$ In fact it suffices to take $m=11$. [^33]: *College of Mathematics and Statistics, Shenzhen University, Shenzhen, 518061, China ${ }^{\dagger} \mathrm{BIU}$ ${ }^{\ddagger}$ Московский Физико-Технический Институт, факультет инноваций и высоких технологий, лаборатория продвинутой комбинаторики и сетевых приложений ${ }^{\S}$ Московский Государственный Технический Университет им. Н.Э. Баумана, факультет Информатика и системы управления, кафедра ИУ-8 Информационная безопасность, ${ }^{\text {® Passware, Research Department, chilikov@passware.com }}$ [^34]: *Здесь вы можете найти текущие результаты: https://clck.ru/HMnm8 ${ }^{1}$ Одна из сложных задач проекта будет связана с раскраской плоских графов, но саму теорему о четырёх красках мы не будем доказывать. ${ }^{2}$ Отметим, что если в этом определении заменить “трех ребер" на "двух ребер" - получится обычное определение планарного графа, что и объясняет выбор термина "квазипланарный". [^35]: ${ }^{3}$ Здесь и далее, если не оговаривается обратное, мы будем рассматривать графы без петель и кратных рёбер. [^36]: ${ }^{4}$ Крестом клетки называется объединение её столбца и её строки [^37]: ${ }^{5}$ Две вершины называются соседними, если у них есть общее ребро; два ребра - если у них общая вершина; вершина и ребро - если это ребро выходит из этой вершины. [^38]: ${ }^{6}$ То есть никакие три не лежат на одной прямой. ${ }^{7}$ при этом все вершины графа не обязаны быть в общем положении ${ }^{8}$ Например, требование, чтобы каждое ребро имело не более чем $t$ пересечений, или чтобы среди ребер было не более $d$ попарно пересекающихся и так далее. [^39]: ${ }^{9}$ Один их них вы нашли при решении задачи 3.3.1, другой подсказан в задаче 3.3.2, до третьего типа догадайтесь самостоятельно. [^40]: 10 Экватором на сфере называется окружность, получающаяся как пересечение сферы с плоскостью, проходящей через центр сферы [^41]: ${ }^{11}$ Два куска разбиения с общей вершиной $A$ называются вертикалъными относительно $A$, если их углы при $A$ вертикальны [^42]: *Вопросы и замечания пишите на alexander.polyanskii@yandex.ru [^43]: ${ }^{1}$ Крестом клетки называется объединение её столбца и её строки. [^44]: ${ }^{2}$ Под $k+$ подразумеваем число, не меньшее $k$. [^45]: ${ }^{3}$ То есть никакие три не лежат на одной прямой. ${ }^{4}$ при этом все вершины графа не обязаны быть в общем положении [^46]: ${ }^{5}$ Обратите внимание, что в искомой оценке свободный член равен -8 . ${ }^{6}$ Обратите внимание, что в искомой оценке свободный член равен -20 . [^47]: ${ }^{7}$ Один их них вы нашли при решении задачи 3.3.1, другой подсказан в задаче 3.3.2, до третьего типа догадайтесь самостоятельно. [^48]: ${ }^{8}$ Запись $k^{-}$обозначает «число, меньшее либо равное $k$ » ${ }^{9}$ На самом деле можно показать, что в этом случае заряд строго положителен, но это не нужно. [^49]: ${ }^{10}$ Два куска разбиения с общей вершиной $A$ называются вертикалъными относителъно $A$, если их углы при $A$ вертикальны [^50]: *Here you can find the current results: https://clck.ru/HMnm8 ${ }^{1}$ Note that if one replaces 3 with 2 in the above definition, one gets the definition of a planar graph. This explains the choice of the term "quasi-planar". [^51]: ${ }^{2}$ From now on, we consider graphs without loops and multiple edges unless otherwise stated. ${ }^{3} \mathrm{~A}$ cross of a cell is the union of cells lying with it in the same row or the same column [^52]: ${ }^{4}$ Two vertices are called neighboring if they are connected by an edge, two edges are neighboring if they share a common vertex, a vertex and an edge are neighboring if the vertex belongs to the edge. [^53]: ${ }^{5}$ that is, no three of them are on the same line ${ }^{6}$ all vertices are not necessary in general position ${ }^{7}$ For example, the condition that each edge has at most $t$ inner crossings, or that there are at most $d$ edges pairwise crossing in inner points [^54]: ${ }^{8}$ Solving problem 3.3.1, you found the first configuration, another one is defined in problem 3.3.2. Try to figure out what is the third one. [^55]: ${ }^{9}$ An equator is the intersection of a sphere with a plane passing through its center. [^56]: ${ }^{10}$ Two cells containing a vertex $A$ are vertical with respect to $A$ if their angles $A$ are vertical. [^57]: *Here you can find the current results: https://clck.ru/HMnm8 [^58]: ${ }^{1} \mathrm{~A}$ cross of a cell is the union of cells lying with it in the same row or the same column [^59]: ${ }^{2}$ Here, we denote by $k+$ a number that is at least $k$. [^60]: ${ }^{3}$ that is, no three of them are on the same line ${ }^{4}$ all vertices are not necessary in general position