# Произведения последовательных натуральных чисел Вадим Бугаенко, Константин Кохась, Ярослав Абрамов, Мария Илюхина ## Версия 0.9 ## Условия задач Мы будем решать следующую задачу: Может ли произведение нескольких последовательных натуральных чисел быть некоторой степенъю натурального числа Иными словами, ставится вопрос разрешимости в целых положительных числах уравнения $$ x(x+1)(x+2) \ldots(x+k-1)=y^{m} $$ при всевозможных $k \geqslant 2$ и $m \geqslant 2$. Решение задачи для любого частного случая будет засчитываться, как продвижение. Рекомендуем начать со следующих случаев. ## 1 Некоторые частные случаи 1.1. $k=2, m=2$. 1.2. $k=2, m-$ произвольное. 1.3. $k=3, m=2$. 1.4. $k=3, m-$ произвольное. 1.5. $k=4, m=2$. 1.6. $k=m$. 1.7. $k=8, m=4$. 1.8. $k=8, m=2$. 1.9. $k=4, m-$ произвольное. 1.10. $k=5, m=2$. ## 2 Близкие вопросы 2.1. Докажите, что при $m=2$ и четном $k$ уравнение не может иметь бесконечно много решений $(x, y)$. 2.2. Из пяти последовательных натуральных чисел выбирают четыре и перемножают. Может ли результат оказаться точным квадратом? 2.3. Докажите, что уравнение $x(x+d)(x+2 d)=y^{2}$ имеет бесконечно много решений $(x, y, d)$ в натуральных числах. 2.4. Докажите, что при фиксированном $k \neq 2,4$ многочлен вида $x(x+1)(x+2) \ldots(x+k-1)+c$, где $c-$ рациональное число, не является квадратом многочлена. ## 3 Удобные числа Назовем число $k$ удобным, если среди любых $k$ последовательных натуральных чисел найдется хотя бы одно, взаимно простое с остальными При ссылке на наше основное уравнение (1) будем называть его уравнением $(k, m)$, если речь идет о произвольных $k$ и $m$, или указывать конкретные значения $k$ и $m$. Например, в самой первой задаче обсуждалось уравнение $(2,2)$. 3.1. Докажите, что для удобных $k$ при $m>k$ уравнение $(k, m)$ не может иметь бесконечно много решений. 3.2. Докажите, что для каждого удобного числа $k$ существует такое число $m_{0}(k)$, что при $m>m_{0}(k)$ уравнение $(k, m)$ не имеет решений. 3.3. Докажите, что для удобных $k$ уравнение $(k, m)$ не имеет решений при $m \geqslant 2 k$. 3.4. Докажите, что для удобных $k$ уравнение $(k, m)$ не имеет решений при $m \geqslant k+2 \log _{2} k$. 3.5. Докажите, что уравнение $(5,7)$ не имеет решений. 3.6. Докажите, что все натуральные числа, не превосходящие 16 , удобны. 3.7. Докажите, что 17 - неудобное число. 3.8. Докажите, что все натуральные числа, большие 17 , неудобные. ## 4 Общие свойства решений уравнения ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-002.jpg?height=228&width=1590&top_left_y=1121&top_left_x=220) $$ x+i=a_{i} z_{i}^{m}, \quad 0 \leqslant i \leqslant k-1 $$ где числа $a_{i}$ не делятся ни на какую точную $m$-ю степень, т. е. все простые множители входят в разложение чисел $a_{i}$ в степени меньше $m$. 4.1. Докажите, что для каждого решения уравнения $(k, m)$ верно неравенство $x>k$. 4.2. Докажите, что для каждого решения уравнения $(k, m)$ верно неравенство $x>k^{m}$. 4.3. Докажите, все простые множители чисел $a_{i}$ меньше $k$. 4.4. Решите уравнение $(7,2)$. 4.5. Решите уравнение $(6,2)$. 4.6. Докажите, что если $x$ - решение уравнения $(k, m)$, то равенство $\left(x+i_{1}\right) \ldots\left(x+i_{m-1}\right)=\left(x+j_{1}\right) \ldots\left(x+j_{m-1}\right), \quad$ где $\quad 0 \leqslant i_{1} \leqslant \cdots \leqslant i_{m-1} \leqslant k-1 ; 0 \leqslant j_{1} \leqslant \cdots \leqslant j_{m-1} \leqslant k-1$ возможно лишь в случае, когда наборы индексов в левой и правой частях совпадают. 4.7. Докажите, все числа $a_{i}$ различны. 4.8. Пусть $m=3$. Докажите, все попарные произведения $a_{i} a_{j}, 0 \leqslant i \leqslant j \leqslant k-1$, различны. 4.9. Пусть $m=3$. Докажите, что частное вида $\frac{a_{i} a_{j}}{a_{r} a_{s}}$, где $0 \leqslant i \leqslant j \leqslant k-1,0 \leqslant r \leqslant s \leqslant k-1$, не может быть кубом рационального числа (не равного 1 ). 4.10. Пусть $m=3, k=75$. Докажите, что не менее двадцати чисел $a_{i}$ содержат в своем разложении только простые множители, не превосходящие 10. 4.11. Докажите, что уравнение $(75,3)$ не имеет решения. 4.12. Обозначим через $\pi(k)$ количество простых чисел, не превосходящих $k$. Докажите, что почти все числа $a_{i}$ невелики в следующем смысле: можно выбрать $k-\pi(k)$ чисел $a_{i}$ так, что их произведение будет делителем числа $k!$. (Вы можете ограничиться случаями $m=2,3$.) 4.13. Пусть $m=2, B_{n}(k)=a_{0} a_{1} \ldots a_{k-1}$. Докажите, что $B_{x}(k)>\left(\frac{4}{3}\right)^{k} k!$ при достаточно больших $k$. ## Решения задач Изучаемое уравнение вообще не имеет решений. Поэтому при разборе отдельных пунктов мы не приводим этот повторяющийся ответ "решений нет", а сразу приступаем к доказательству. ## 1 Некоторые частные случаи 1.1. Следует из утверждения следующего пункта. 1.2. Числа $x$ и $x+1$ взаимно просты, и поэтому оба должны быть точными $m$-ми степенями, что, как нетрудно понять, невозможно. 1.3. Следует из утверждения следующего пункта. 1.4. Заметим, что числа $x+1$ и $x(x+2)=(x+1)^{2}-1$ взаимно просты. Следовательно, каждое из них точная $m$-я степень. Пусть $x+1=u^{m}$, тогда $x(x+2)=\left(u^{2}\right)^{m}-1=v^{m}$. Последнее невозможно, так как разность между $m$-ми степенями натуральных чисел всегда больше 1 . 1.5. Воспользуемся тождеством $x(x+1)(x+2)(x+3)+1=\left(x^{2}+3 x+1\right)^{2}$. 1.6. Заметим, что $x10$ $$ (a+6)^{4}2 u^{3 m-3}+u^{m}>2 u^{3 m-3}+u^{m}-2 \end{aligned} $$ 1.10. Заметим, что наибольший общий делитель любых двух из чисел $x, x+1, \ldots, x+4$ не превосходит 4 . Поэтому все крупные простые множители числа $y$ должны входить в разложение чисел $x, \ldots, x+4$ в четных степенях. Следовательно, каждый множитель в левой части уравнения имеет вид $n^{2}, 2 n^{2}, 3 n^{2}$ или $6 n^{2}$. Так как в левой части уравнения 5 множителей, какие-то два множителя имеют один и тот же вид. Но разности между соседними квадратами, удвоенными квадратами и т. д. обычно больше 4 , поэтому уравнение не имеет решений. ## 2 Близкие вопросъ 2.1. Следующее рассуждение взято из [4]. Многочлен в правой части уравнения (1) обозначим для краткости через $f(x)$. Допустим, что при $m=2, k=2 n$ уравнение (1) имеет бесконечно много решений $\left(x_{i}, y_{i}\right)$, где $x_{i} \rightarrow+\infty$ и $f\left(x_{i}\right)=y_{i}^{2}$. Заметим, что $f(x)$ не является квадратом многочлена, поскольку все его корни имеют кратность 1. Подберем тогда такой многочлен $a(x)$ степени $n$, что $\operatorname{deg}\left(f-a^{2}\right) \leqslant n-1$. Пусть $r=f-a^{2}$, $a\left(x_{i}\right)=z_{i}$. Тогда $z_{i} \sim x_{i}^{n}$ при $i \rightarrow+\infty$ (читатель, не знакомый с теорией пределов, может читать эту фразу так: $z_{i}>0.99 x_{i}^{n}$ при больших $i$ ). Кроме того, при больших $i y_{i}^{2}-z_{i}^{2}=r\left(x_{i}\right) \neq 0$ и при этом $\left|r\left(x_{i}\right)\right| \leqslant$ const $\cdot x_{i}^{n-1}$. Но с другой стороны, $\left|r\left(x_{i}\right)\right|=\left|y_{i}^{2}-z_{i}^{2}\right|=\left(y_{i}+z_{i}\right)\left|y_{i}-z_{i}\right| \geqslant z_{i} \sim x_{i}^{n}$, что противоречит только что полученной оценке. 2.2. Ответ: да, может. $2 \cdot 3 \cdot 4 \cdot 6=12^{2}$. 2.3. Пусть $x=k d$. Тогда $k(k+1)(k+2) d^{3}=y^{2}$. Положим теперь $d=k(k+1)(k+2)$. Вот другое решение с использованием Пифагоровых троек. Пусть $\bar{x}=x+d$. Тогда уравнение запишется в виде $\bar{x}^{2}\left(\bar{x}^{2}-d^{2}\right)=y^{2}$. В качестве решения подойдут $d=2 a b\left(a^{2}-b^{2}\right), \bar{x}=\left(a^{2}+b^{2}\right)^{2}$. А можно просто воспользоваться однородностью: заметим, например, что $(1,35,24)$ - решение. Кроме того, если $(x, y, d)$ - решение, то при каждом $k$ тройка $\left(k^{2} x, k^{3} y, k^{2} d\right)$ - тоже решение. 2.4. Мы приводим рассуждение из [4]. Обозначим для краткости $P_{k, c}(x)=x(x+1)(x+2) \ldots(x+k-1)+c$. Предположим, что $P_{k, c}(x)=a(x)^{2}, k=2 n$. Тогда $$ P_{k, c}(x+1)-P_{k, c}(x)=k(x+1)(x+2) \ldots(x+k-1)=a(x+1)^{2}-a(x)^{2} . $$ Следовательно, $$ (a(x+1)-a(x))(a(x+1)+a(x))=k(x+1)(x+2) \ldots(x+k-1) . $$ Поскольку график многочлена $y=a(x+1)$ получается из графика $y=a(x)$ сдвигом на 1 влево, каждое из $n-1$ решения уравнения $a(x+1)=a(x)$ лежит между парой корней многочлена $a(x)+a(x+1)$ (который всего имеет $n$ корней). Значит, $$ \begin{aligned} & a(x+1)-a(x)=n(x+2)(x+4) \ldots(x+2 n-2) \\ & a(x+1)+a(x)=2(x+1)(x+3) \ldots(x+2 n-1) \end{aligned} $$ Складывая эти выражения, получаем $$ 2 a(x+1)=2(x+1)(x+3) \ldots(x+2 n-1)+n(x+2)(x+4) \ldots(x+2 n-2) $$ А вычитая и подставив в результат $x+1$ вместо $x,-$ $$ 2 a(x+1)=2(x+2)(x+4) \ldots(x+2 n)-n(x+3)(x+5) \ldots(x+2 n-1) $$ Два полученных выражения для $2 a(x+1)$ несовместны при $n>2$. В этом можно убедиться, например, так. Подставим в эти равенства $x=0$ и вычтем одно из другого. Мы получим равенство $$ (n+2)(1 \cdot 3 \cdots(2 n-1))=3 n(2 \cdot 4 \cdots(2 n-2)) $$ в котором при $n>2$ правая часть делится на бо́льшую степень двойки, чем левая. ## 3 Удобные числа 3.1. Пусть число $x+i$ взаимно просто с остальными множителями левой части. Тогда $x+i=u^{m}$ и $$ \left(u^{m}-k+1\right)\left(u^{m}-k+2\right) \ldots u^{m} \leqslant x(x+1) \ldots(x+k-1) \leqslant u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right) $$ Проверим, что при больших $u$ $$ \left(u^{k}-1\right)^{m}<\left(u^{m}-k+1\right)\left(u^{m}-k+2\right) \ldots u^{m} \leqslant u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right)<\left(u^{k}+1\right)^{m} $$ Если это так, то при больших $u$ выполнено неравенство $u^{k}-1u^{k m}+m u^{k m-k} $$ С другой стороны, $$ u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right)k$ и больших $u$ мы имеем неравенство $$ u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right)\left(u^{m}-k+1\right)^{k} $$ то дстаточно при $m \geqslant 2 k$ доказать неравенство $$ \left(u^{m}-k+1\right)^{k}>\left(u^{k}-1\right)^{m} $$ Мы докажем неравенство (5) индукцией по $m$. База, $m=2 k$ $$ \left(u^{2 k}-k+1\right)^{k}>\left(u^{k}-1\right)^{2 k} $$ Извлечем корень $k$-й степени и раскроем скобки, получим очевидное неравенство $2 u^{k}>k$. Переход. Достаточно проверить неравенство $$ \left(u^{m}-k+1\right)^{k}\left(u^{k}-1\right)<\left(u^{m+1}-k+1\right)^{k} $$ Запишем его в виде $$ u^{k}-1<\left(\frac{u^{m+1}-k+1}{u^{m}-k+1}\right)^{k} $$ Это неравенство очевидно, поскольку дробь, написанная в скобках в правой части, не меньше $u$. 3.4. Пусть число $z=x+i$ взаимно просто с остальными множителями правой части уравнения. Тогда $z=x+i=u^{m}$ и $$ \left(u^{m}-k+1\right)^{k}<(z-k+1) \ldots(z-1) z \leqslant x(x+1) \ldots(x+k-1) \leqslant z(z+1) \ldots(z+k-1)<\left(u^{m}+k-1\right)^{k} $$ Докажем, что при $m \geqslant k+2 \log _{2} k$ и $u \geqslant 2$ выполнены неравенства $$ \begin{aligned} & \left(u^{m}+k-1\right)^{k}<\left(u^{k}+1\right)^{m} \\ & \left(u^{m}-k+1\right)^{k}>\left(u^{k}-1\right)^{m} \end{aligned} $$ Доказательство неравенства (6). Запишем частное правой и левой частей и воспользуемся неравенством Бернулли $$ \frac{\left(u^{k}+1\right)^{m}}{u^{k m}} \cdot \frac{u^{k m}}{\left(u^{m}+k-1\right)^{k}}=\left(1+\frac{1}{u^{k}}\right)^{m}\left(1-\frac{k-1}{u^{m}+k-1}\right)^{k} \geqslant 1+\frac{m}{u^{k}}-\frac{k(k-1)}{u^{m}+k-1}-\frac{m k(k-1)}{u^{k}\left(u^{m}+k-1\right)} $$ Нам нужно проверить, что последнее выражение больше 1. Для этого достаточно установить, что сумма трех последних слагаемых положительна, т. е. $$ \frac{m}{u^{k}}>\frac{k(k-1)}{u^{m}+k-1}+\frac{m k(k-1)}{u^{k}\left(u^{m}+k-1\right)} $$ Домножим на знаменатели и перенесем $m(k-1)$ в правую часть: $$ m u^{m}>k(k-1) u^{k}+(k-1)^{2} m $$ Так как $m>k+2 \log _{2} k \geqslant k+2 \log _{u} k$, то $u^{m}>k^{2} u^{k}$. Заменим тогда выражение $u^{m}$ в левой части на $k^{2} u^{k}$, а множители $k-1$ в правой части - на $k$, от этих замен неравенство усилится: $$ m k^{2} u^{k}>k^{2} u^{k}+k^{2} m $$ Мы пришли к верному неравенству $a b>a+b$. Доказательство неравенства (7). Рассмотрим частное $\frac{\left(u^{m}-k+1\right)^{k}}{\left(u^{k}-1\right)^{m}}=\frac{\left(u^{m}-k+1\right)^{k}}{u^{k m}} \frac{u^{k m}}{\left(u^{k}-1\right)^{m}}$ и докажем, что оно больше 1. В самом деле, в силу неравенства Бернулли $$ \left(1-\frac{k-1}{u^{m}}\right)^{k}\left(\frac{u^{k}}{u^{k}-1}\right)^{m}>\left(1+\frac{m}{u^{k}-1}\right)\left(1-\frac{k(k-1)}{u^{m}}\right) $$ Кроме того, $1-\frac{k(k-1)}{u^{m}}>1-\frac{k^{2}}{u^{m}}$. Значит, достаточно доказать, что $$ 1<\left(1+\frac{m}{u^{k}-1}\right)\left(1-\frac{k^{2}}{u^{m}}\right)=1+\frac{m}{u^{k}-1}-\frac{k^{2}}{u^{m}}-\frac{m k^{2}}{u^{m}\left(u^{k}-1\right)}=1+\frac{m u^{m}-k^{2} u^{k}+k^{2}-m k^{2}}{u^{m}\left(u^{k}-1\right)} $$ Но ведь $u^{m} \geqslant k^{2} u^{k}$, значит, $$ m u^{m}-k^{2} u^{k}+k^{2}-m k^{2}>m k^{2} u^{k}-k^{2} u^{k}+k^{2}-m k^{2}=k^{2}(m-1)\left(u^{k}-1\right)>0 $$ что и требовалось. 3.5. Неравенство $7 \geqslant 5+2 \log _{2} 5$ неверно, так что мы не можем сослаться на результат предыдущей задачи. Правда, в решении предыдущей задачи мы пользовались лишь неравенством $m \geqslant k+2 \log _{u} k$, так что такая ссылка все же возможна, если дополнить ее перебором малых значений $x$. Но мы дадим прямое решение. Как и в предыдущих задачах, достаточно доказать два неравенства $$ \begin{aligned} & u^{7}\left(u^{7}+1\right)\left(u^{7}+2\right)\left(u^{7}+3\right)\left(u^{7}+4\right)<\left(u^{5}+1\right)^{7} \\ & u^{7}\left(u^{7}-1\right)\left(u^{7}-2\right)\left(u^{7}-3\right)\left(u^{7}-4\right)>\left(u^{5}-1\right)^{7} \end{aligned} $$ Для доказательства неравенства (8) раскроем скобки $$ \begin{aligned} & u^{35}+10 u^{28}+35 u^{21}+50 u^{14}+24 u^{7}< \\ & u^{35}-7 u^{30}+21 u^{25}-35 u^{20}+35 u^{15}-21 u^{10}+7 u^{5}-1 $$ Соберем в каждой части слагаемые одного знака. $$ \begin{aligned} 6 u^{30}+u^{30}+35 u^{21}+35 u^{20}+24 u^{7}+21 u^{10}+1> \\ >10 u^{28}+21 u^{25}+35 u^{15}+50 u^{14}+7 u^{5} \end{aligned} $$ Нетрудно видеть, что при $u \geqslant 2$ каждое слагаемое нижней строки меньше расположенного над ним слагаемого верхней строки. 3.6. Утверждение этой и двух следующих задач мы почерпнули в [1]. 3.7. В качестве примера набора из 17 чисел, каждое из которых имеет общий делитель с каждым из остальных подходит набор, в котором наименьшее число равно $2184+30030 n$. 3.8. ## 4 Общие свойства решений уравнения 4.1. При $x \leqslant k$ все натуральные числа, лежащие между $x+k$ и $\frac{1}{2}(x+k)$, входят как сомножители в левую часть нашего уравнения. Согласно постулату Бертрана, среди этих чисел встречается простое число $p$, и очевидно, что $p$ входит в произведение в левой части в первой степени. Поэтому при $x \leqslant k$ левая часть не может быть никакой точной степенью. 4.2. По теореме Сильвестра некоторый множитель $x+i$ из левой части уравнения имеет простой делитель $p>k$. Поскольку в левой части лишь $x+i$ может делиться на $p$, то чтобы все произведение оказалось точной степенью, необходимо, чтобы $x+i$ делилось хотя бы на $p^{m}$. Тогда $x+i \geqslant p^{m} \geqslant(k+1)^{m}$. Если при этом $x \leqslant k^{p}$, то $k^{p}+i \geqslant x+i \geqslant(k+1)^{p}$. Откуда $i>p k$, что неверно. 4.3. Если бы число $a_{i}$ имело бы простой делитель $p>k$, то среди чисел $x, x+1, \ldots, x+k-1$ только $x+i$ делилось бы на $p$. Но тогда $p$ входит в разложение $x+i$ на простые множители в степени, кратной $m$, и, поскольку $x+i=a_{i} z_{i}^{m}$, множитель $p$ входит только в разложение $z_{i}$ и не входит в разложение $a_{i}$. Противоречие. 4.4. Заметим, что если мы найдем среди чисел $a_{i}$ не менее 5 чисел, в разложение которых входят только простые множители 2 и 3 , то, рассуждая дальше как в задаче 1.10 , мы сразу получим, что уравнение не имеет решений. Но такие 5 чисел совсем нетрудно найти: всего чисел у нас 7 , они могут содержать только простые делители $2,3,5$ и при этом количество чисел $a_{i}$, делящихся на 5, - не более двух! 4.5. Как и в предыдущей задаче, попробуем найти среди чисел $a_{i}$ не менее 5 чисел, в разложение которых входят только простые множители 2 и 3 . Числа $a_{i}$ могут содержать только множители $2,3,5$ - причем каждый множитель может присутствовать лишь в степени 0 или 1 . Произведение чисел $a_{i}$ должно быть точным квадратом. Среди исходных последовательных натуральных чисел не более двух делятся на 5 , поэтому среди чисел $a_{i}$ тоже не более двух чисел, делящихся на 5. Нам достаточно рассмотреть только случай, когда этих чисел ровно 2. Очевидно, это возможно, лишь если $a_{0}$ и $a_{5}$ делятся на 5 . Сделаем на всякий случай очевидное замечание. Среди трех последовательных натуральных чисел ровно одно делится на 3. Но если эта тройка входит в четной степени, то среди соответствующих чисел $a_{i}$ не будет ни одного, делящегося на 3. Рассмотрим 4 числа $x+1, x+2, x+3, x+4$. Как мы знаем из задачи 1.5 , их произведение не может быть точным квадратом. Следовательно, суммарная степень вхождения в числа $a_{1}, a_{2}, a_{3}, a_{4}$ множителей 2 или суммарная степень вхождения в эти же числа $a_{i}$ множителей 3 должна быть нечетна. Это может быть лишь в двух (невзаимоисключающих) случаях: 1) среди чисел $a_{1}, a_{2}, a_{3}, a_{4}$ ровно одно делится на 2 ; 2) среди чисел $a_{1}, a_{2}, a_{3}, a_{4}$ ровно одно делится на 3 . Заметим еще, что количество чисел $a_{i}$, делящихся на 2 (и, аналогично, на 3), не превосходит 2. Тогда, как нетрудно видеть, в каждом из случаев среди чисел $x+1, x+2, x+3, x+4$ обязательно найдутся два числа вида $t^{2}$, или два числа вида $2 t^{2}$, два числа вида $3 t^{2}$, чего не может быть. 4.6. Это утверждение - часть леммы 1 из [3]. Сократим равные множители. Так как $\operatorname{HOД(~} n+i, n+j)k^{m}$, получается, что ни одна из скобок в левой части не делит произведения в правой части, поэтому равенство невозможно. 4.7. Пусть $a_{i}=a_{j}$, где $0 \leqslant jn+j=a_{j} z_{j}^{m}$, то $z_{i} \geqslant z_{j}+1$. Следовательно, $k>a_{j} z_{i}^{m}-a_{j} z_{j}^{m}=a_{j}\left(\left(z_{j}+1\right)^{m}-z_{j}^{m}\right)>a_{j} m z_{j}^{m-1} \geqslant a_{j}^{(m-1) / m} z_{j}^{m-1}=\left(a_{j} z_{j}^{m}\right)^{(m-1) / m}=(x+j)^{(m-1) / m}>x^{1 / m}$, что противоречит утверждению задачи 4.2. 4.8. Похоже на 4.7. Следует из утверждения следующей задачи. Достаточно в ее решении всюду полагать $u=v=1$. 4.9. Это лемма 1 из [3]. Допустим, что $$ a_{i_{1}} a_{i_{2}}=a_{j_{1}} a_{j_{2}} t^{3} $$ Проверим, что тогда $t=1$ и наборы индексов в левой и правой частях совпадают. Не умаляя общности, можно считать, что $\left(x+i_{1}\right)\left(x+i_{2}\right)>\left(x+j_{1}\right)\left(x+j_{2}\right)$ (равенство этих выражений невозможно в силу утверждения задачи 4.6). Пусть $t=u / v$ (несократимая запись). Тогда $a_{i_{1}} a_{i_{2}} / u^{3}=a_{j_{1}} a_{j_{2}} / v^{3}$ и в силу того, что $u$ и $v$ взаимно просты, обе части равенства представляют собой целое число. Обозначим $A=a_{i_{1}} a_{i_{2}} / u^{3}=a_{j_{1}} a_{j_{2}} / v^{3}$. По определению чисел $a_{i}$ имеем $x+i=a_{i} z_{i}^{3}$, и тогда $$ \begin{aligned} & \left(x+i_{1}\right)\left(x+i_{2}\right)=a_{i_{1}} a_{i_{2}} \cdot \frac{s^{3}}{u^{3}} \\ & \left(x+j_{1}\right)\left(x+j_{2}\right)=a_{j_{1}} a_{j_{2}} \cdot \frac{r^{3}}{v^{3}} \end{aligned} $$ где $s=u z_{i_{1}} z_{i_{2}}, r=v z_{j_{1}} z_{j_{2}}$. Тогда $A s^{3}>A r^{3}$ и, значит, $s \geqslant r+1$. Таким образом, $$ \left(x+i_{1}\right)\left(x+i_{2}\right)-\left(x+j_{1}\right)\left(x+j_{2}\right) \geqslant A\left((r+1)^{3}-r^{3}\right)>3 A r^{2} $$ Заметим, что $A r^{3}=\left(x+j_{1}\right)\left(x+j_{2}\right)>x^{2}$, и тогда последнее неравенство может быть переписано так: $$ \left(x+i_{1}\right)\left(x+i_{2}\right)-\left(x+j_{1}\right)\left(x+j_{2}\right) \geqslant 3 A \cdot\left(\frac{x^{2}}{A}\right)^{2 / 3} \geqslant 3 x^{4 / 3} $$ С другой стороны, $$ \left(x+i_{1}\right)\left(x+i_{2}\right)-\left(x+j_{1}\right)\left(x+j_{2}\right)<(x+k)^{2}-x^{2}<3 k x $$ Полученные оценки противоречат друг другу, так как из задачи 4.2 мы знаем, что $x>k^{3}$, и следовательно, $3 k x<3 x^{4 / 3}$. 4.10. Нам нужно найти двадцать чисел $a_{i}$, которые содержат только простые множители $2,3,5,7$. Мы знаем, что числа $a_{i}$ могут делиться только на простые множители, не превосходящие 75 , т. е. на 2,3 , $5,7,11,13,17,19,23,29,31,37,41,43,47,53,59,61,67,71,73$. Поскольку числа $a_{i}$ являются делителями 75 последовательных натуральных чисел, то для каждого простого числа из диапазона [41;73] (в этом отрезке 9 простых чисел) лишь не более двух из чисел $a_{i}$ делятся на это простое число; а для каждого простого числа из диапазона $[29 ; 37]$ (3 простых числа) лишь не более трех из чисел $a_{i}$ делятся на это простое число. Далее, не более четырех из чисел $a_{i}$ делятся на 23 ; не более четырех - на 19 ; не более пяти - на 17 ; не более шести - на 13 и не более семи - на 11. Итого мы насчитали не более $$ 18+9+8+5+6+7=53 $$ чисел. Значит, осталось не менее 22 чисел. Эти числа могут иметь только множители $2,3,5,7$. 4.11. Рассмотрим только те числа $a_{i}$, которые содержат только простые множители $2,3,5,7$. По утверждению предыдущей задачи, таких чисел не меньше 20 , но нам, впрочем, достаточно будет десяти из них. Пользуясь этими десятью числами, мы можем составить $90=10 \cdot 9$ различных частных вида $a_{i} / a_{j}$. С другой стороны, по утверждению задачи 4.9 частное двух произведений вида $a_{i} a_{j}$ (и, в том числе, частное вида $a_{i} / a_{j}$ ) не может быть кубом рационального числа. Но тогда, с точностью до отличия на множитель-куб существует не более $3^{4}=81$ классов значений для частных выбранных нами десяти чисел (каждый множитель $2,3,5,7$ может входить в такое частное в степени $0+3 s, 1+3 s, 2+3 s$ ). Таким образом, в какой-то класс попадут два частных, т. е. будет выполнено соотношение $$ \frac{a_{i}}{a_{j}}=\frac{a_{u}}{a_{v}} \cdot t^{3} $$ запрещенное утверждением задачи 4.9. 4.12. Это изящное рассуждение мы целиком взяли из статьи [3]. Для каждого простого числа $p_{0}B^{\prime}(k)$, где $B^{\prime}(k)$ - произведение первых $k$ чисел, свободных от квадратов. Достаточно доказать, что $B^{\prime}(k)>(4 / 3)^{k} k$ ! при $k \geqslant 24$. Докажем это по индукции. База, $k=24$. Проверяется непосредственно: $$ \frac{26 \cdot 29 \cdot 30 \cdot 31 \cdot 33 \cdot 34 \cdot 35 \cdot 37}{4 \cdot 8 \cdot 9 \cdot 12 \cdot 16 \cdot 18 \cdot 20 \cdot 24}>\left(\frac{4}{3}\right)^{24} $$ Индукционный переход. При $r \geqslant 9$ количество свободных от квадратов чисел, не превосходящих $r$, не превосходит $r-\left[\frac{r}{4}\right]-1<\frac{3}{4} r$. Значит, при $n \geqslant 7 n$-е свободное от квадратов число больше, чем $\frac{4}{3} n$. ## ЛитератУра [1] Флейшман Д. Китайская теорема об остатках и гипотеза Ченцова // Квант. 1997. № 3. [2] Сендеров В. Об одной замечательной теореме Эрдёша // Квант. 2008. В печати. [3] Erdốs P., Selfridge J.L. The product of consecutive integers is never a power // Illinois Math. J. 1975. Vol. 19. P. 292-301. [4] Poorten A. van der, Woeginger G. Squares from products of consecutive integers // Amer. Math. Monthly. 2002. Vol. 109. № 5. P. 459-461. ## Products of consecutive Integers Vadim Bugaenko, Konstantin Kokhas, Yaroslav Abramov, Maria Ilyukhina 8 августа 2008 г. ## Problems We shall consider the following problem: Could a product of some consecutive integers be a power of an integer? In other words we ask whether the equation $$ x(x+1)(x+2) \ldots(x+k-1)=y^{m} $$ has a solution in positive integers (for each $k \geqslant 2$ and $m \geqslant 2$ ). A solution of the problem for any partial case will be accepted as a progression. We recommend to consider the following cases first. ## 1 Some partial cases 1.1. $k=2, m=2$. 1.2. $k=2, m$ is arbitrary. 1.3. $k=3, m=2$. 1.4. $k=3, m$ is arbitrary. 1.5. $k=4, m=2$. 1.6. $k=m$. 1.7. $k=8, m=4$. 1.8. $k=8, m=2$. 1.9. $k=4, m$ is arbitrary. 1.10. $k=5, m=2$. ## 2 Variations of the question 2.1. Prove that for $m=2$ and even $k$ the equation does not have infinitely many solutions $(x, y)$. 2.2. We take 5 consecutive integers, choose 4 of them and multiply. Is it possible the result to be an exact square? 2.3. Prove that the equation $x(x+d)(x+2 d)=y^{2}$ has infinitely many solutions $(x, y, d)$ in nonnegative integers. 2.4. Prove that for every $k \neq 2,4$ a polynom of a form $x(x+1)(x+2) \ldots(x+k-1)+c$, where $c$ is a rational number, is not a square of a polynom. ## 3 Convenient numbers We call a number $k$ convenient if among each $k$ consecutive positive integers there is at least one which is relatively prime to the others. We shall refer to our main equation (1) by the notation $(k, m)$. For example, in the very first problem we spoke about the equation $(2,2)$. 3.1. Prove that the equation $(k, m)$ could not have infinitely many solutions for convinient $k$ and $m>k$. 3.2. Prove that for each convinient $k$ there is a number $m_{0}(k)$, such that for $m>m_{0}(k)$ the equation $(k, m)$ has no solutions. 3.3. Prove that the equation $(k, m)$ has no solutions for convinient $k$ and $m \geqslant 2 k$. 3.4. Prove that the equation $(k, m)$ has no solutions for convinient $k$ and $m \geqslant k+2 \log _{2} k$. 3.5. Prove that the equation $(5,7)$ has no solutions. 3.6. Prove that all positive integers less than or equal to 16 are convenient. 3.7. Prove that 17 is not convenient. 3.8. Prove that all positive integers greater than 17 are not convenient. ## 4 Common properties of solutions ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-011.jpg?height=229&width=1419&top_left_y=1120&top_left_x=224) $$ x+i=a_{i} z_{i}^{m}, \quad 0 \leqslant i \leqslant k-1 $$ where integers $a_{i}$ are free of $m$-th powers, i.e. the power of each prime factor of $a_{i}$ is less than $m$. 4.1. Prove that $x>k$ for any solution of the equation $(k, m)$. 4.2. Prove that $x>k^{m}$ for any solution of the equation $(k, m)$. 4.3. Prove that all prime factors of integers $a_{i}$ are less than $k$. 4.4. Solve the equation $(7,2)$. 4.5. Solve the equation $(6,2)$. 4.6. Let $x$ be a solution of the equation $(k, m)$. Prove that the equality $\left(x+i_{1}\right) \ldots\left(x+i_{m-1}\right)=\left(x+j_{1}\right) \ldots\left(x+j_{m-1}\right), \quad$ where $\quad 0 \leqslant i_{1} \leqslant \cdots \leqslant i_{m-1} \leqslant k-1 ; 0 \leqslant j_{1} \leqslant \cdots \leqslant j_{m-1} \leqslant k-1$ is possible only if the sets of indices coincide. 4.7. Prove that all the integers $a_{i}$ are different. 4.8. Let $m=3$. Prove that all the pairwise products $a_{i} a_{j}, 0 \leqslant i \leqslant j \leqslant k-1$, are different. 4.9. Let $m=3$. Prove that the fractions of the form $\frac{a_{i} a_{j}}{a_{r} a_{s}}$, where $0 \leqslant i \leqslant j \leqslant k-1,0 \leqslant r \leqslant s \leqslant k-1$, are not equal to the cube of rational number (excluding 1 ). 4.10. Let $m=3, k=75$. Prove that at least 20 integers $a_{i}$ have no prime factors greater than 10 . 4.11. Prove that the equation $(75,3)$ has no solutions. 4.12. Denote by $\pi(k)$ the number of primes which are less or equal to $k$. Prove that almost all integers $a_{i}$ are "small" in the following sense: we can choose $k-\pi(k)$ of them, such that the product of chosen numbers divides $k$ !. (You may consider the cases $m=2,3$ only.) 4.13. Let $m=2, B_{x}(k)=a_{0} a_{1} \ldots a_{k-1}$. Prove that $B_{x}(k)>\left(\frac{4}{3}\right)^{k} k$ ! for an arbitrary large $k$. ## Solutions The equation under consideration has no solutions at all. So the answer "There are no solutions" will not be repeated in each solution. ## 1 Some partial cases 1.1. Follows from the next problem. 1.2. Numbers $x$ and $x+1$ are relatively prime, thus both must be perfect $m$-th power. However this is impossible. 1.3. Follows from the next problem. 1.4. Let us denote that numbers $x+1$ and $x(x+2)=(x+1)^{2}-1$ are relatively prime. So each of them is a perfect $m$-th power. Let $x+1=u^{m}$, then $x(x+2)=\left(u^{2}\right)^{m}-1=v^{m}$. This is impossible because two perfect $m$-th powers differ by more than 1 . 1.5. Let us apply an identity $x(x+1)(x+2)(x+3)+1=\left(x^{2}+3 x+1\right)^{2}$. 1.6. Note that $x10$ $$ (a+6)^{4}0.99 x_{i}^{n}$ for large $i$ ). Moreover $y_{i}^{2}-z_{i}^{2}=r\left(x_{i}\right) \neq 0$ for large $i$ and at the same time $\left|r\left(x_{i}\right)\right| \leqslant$ const $\cdot x_{i}^{n-1}$. But on the other hand $\left|r\left(x_{i}\right)\right|=\left|y_{i}^{2}-z_{i}^{2}\right|=\left(y_{i}+z_{i}\right)\left|y_{i}-z_{i}\right| \geqslant z_{i} \sim x_{i}^{n}$, which contradicts to the estimation just obtained. 2.2. Answer: Yes, it is possible. $2 \cdot 3 \cdot 4 \cdot 6=12^{2}$. 2.3. First solution. Let $x=k d$. Then $k(k+1)(k+2) d^{3}=y^{2}$. Put $d=k(k+1)(k+2)$. Second solution. It uses Pythagorean triples. Let $\bar{x}=x+d$. Then the equation will be written in a form $\bar{x}^{2}\left(\bar{x}^{2}-d^{2}\right)=y^{2}$. We could take $d=2 a b\left(a^{2}-b^{2}\right), \bar{x}=\left(a^{2}+b^{2}\right)^{2}$ as a solution. Third solution. Note that if $(x, y, d)$ - a solution then for each $k$ the triple $\left(k^{2} x, k^{3} y, k^{2} d\right)$ is also a solution. So to solve the problem it is enough to find one partial solution, for example $(1,35,24)$. 2.4. The proof is taken from [4]. Let us denote $P_{k, c}(x)=x(x+1)(x+2) \ldots(x+k-1)+c$. Suppose that $P_{k, c}(x)=a(x)^{2}, k=2 n$. Then $$ P_{k, c}(x+1)-P_{k, c}(x)=k(x+1)(x+2) \ldots(x+k-1)=a(x+1)^{2}-a(x)^{2} . $$ Hence $$ (a(x+1)-a(x))(a(x+1)+a(x))=k(x+1)(x+2) \ldots(x+k-1) . $$ Since the graph of the polynomial $y=a(x+1)$ could be obtained by translation to the left by 1 from the graph $y=a(x)$, each of $n-1$ solutions of the equation $a(x+1)=a(x)$ lies between a pair of roots of the polynomial $a(x)+a(x+1)$ (which have $n$ roots). Hence $$ \begin{aligned} & a(x+1)-a(x)=n(x+2)(x+4) \ldots(x+2 n-2) \\ & a(x+1)+a(x)=2(x+1)(x+3) \ldots(x+2 n-1) \end{aligned} $$ Adding these expressions we get $$ 2 a(x+1)=2(x+1)(x+3) \ldots(x+2 n-1)+n(x+2)(x+4) \ldots(x+2 n-2) $$ And substituting the same changing $x$ by $x+1$ we obtain $$ 2 a(x+1)=2(x+2)(x+4) \ldots(x+2 n)-n(x+3)(x+5) \ldots(x+2 n-1) . $$ Two obtained expressions contradict to each other. To be ensure this put $x=0$ to both and subtract one from another. We get $$ (n+2)(1 \cdot 3 \cdots(2 n-1))=3 n(2 \cdot 4 \cdots(2 n-2)) $$ Here the right hand contains two as a factor with more power than left hand side. ## 3 Convenient numbers 3.1. Let an integer $x+i$ be relatevely prime with all other factors from the left hand side. Then $x+i=u^{m}$ and $$ \left(u^{m}-k+1\right)\left(u^{m}-k+2\right) \ldots u^{m} \leqslant x(x+1) \ldots(x+k-1) \leqslant u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right) $$ Let us check that if $u$ is large $$ \left(u^{k}-1\right)^{m}<\left(u^{m}-k+1\right)\left(u^{m}-k+2\right) \ldots u^{m} \leqslant u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right)<\left(u^{k}+1\right)^{m} $$ If this is true then for large $u$ the inequality $u^{k}-1u^{k m}+m u^{k m-k} $$ On the other hand $$ u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right)k$ and $u$ is large we have inequality $$ u^{m}\left(u^{m}+1\right) \ldots\left(u^{m}+k-1\right)\left(u^{m}-k+1\right)^{k} $$ then it is enought to prove $$ \left(u^{m}-k+1\right)^{k}>\left(u^{k}-1\right)^{m} $$ for $m \geqslant 2 k$. We would prove the inequality (5) by induction by $m$. Base, $m=2 k$ $$ \left(u^{2 k}-k+1\right)^{k}>\left(u^{k}-1\right)^{2 k} $$ Taking the root of degree $k$-th and expanding we get the obvious inequality $2 u^{k}>k$. Step of induction. It is enought to check $$ \left(u^{m}-k+1\right)^{k}\left(u^{k}-1\right)<\left(u^{m+1}-k+1\right)^{k} $$ Let us write this inequality in the form $$ u^{k}-1<\left(\frac{u^{m+1}-k+1}{u^{m}-k+1}\right)^{k} $$ This inequality is obvious because the fraction in parenthesis in the right hand side is not less than $u$ This inequality is obvious since the fraction in parenthesis in the right hand side is not less than $u$. 3.4. Let an integer $z=x+i$ be relatively prime with all other factors of the right hand side of the equations. Then $z=x+i=u^{m}$ and $$ \left(u^{m}-k+1\right)^{k}<(z-k+1) \ldots(z-1) z \leqslant x(x+1) \ldots(x+k-1) \leqslant z(z+1) \ldots(z+k-1)<\left(u^{m}+k-1\right)^{k} $$ Let us prove that for $m \geqslant k+2 \log _{2} k$ and $u \geqslant 2$ we have inequalities $$ \begin{aligned} & \left(u^{m}+k-1\right)^{k}<\left(u^{k}+1\right)^{m} \\ & \left(u^{m}-k+1\right)^{k}>\left(u^{k}-1\right)^{m} \end{aligned} $$ Proof of inequality (6). Let us apply Bernoully inequality for the quotient of right and left hand sides $$ \frac{\left(u^{k}+1\right)^{m}}{u^{k m}} \cdot \frac{u^{k m}}{\left(u^{m}+k-1\right)^{k}}=\left(1+\frac{1}{u^{k}}\right)^{m}\left(1-\frac{k-1}{u^{m}+k-1}\right)^{k} \geqslant 1+\frac{m}{u^{k}}-\frac{k(k-1)}{u^{m}+k-1}-\frac{m k(k-1)}{u^{k}\left(u^{m}+k-1\right)} $$ We want prove that the last expression is greater than 1. It is sufficient to establish that the sum of the three last expressions is positive, i.e. $$ \frac{m}{u^{k}}>\frac{k(k-1)}{u^{m}+k-1}+\frac{m k(k-1)}{u^{k}\left(u^{m}+k-1\right)} $$ Multiply by the denominators $$ m u^{m}>k(k-1) u^{k}+(k-1)^{2} m $$ Since $m>k+2 \log _{2} k \geqslant k+2 \log _{u} k$, then $u^{m}>k^{2} u^{k}$. Replacement of the expression $u^{m}$ at the left hand side by $k^{2} u^{k}$, and factors $k-1$ at the right hand side by $k$ makes the inequality stronger: $$ m k^{2} u^{k}>k^{2} u^{k}+k^{2} m $$ We obtained to the correct inequality $a b>a+b$. Proof of inequality (7). Consider the quotient $\frac{\left(u^{m}-k+1\right)^{k}}{\left(u^{k}-1\right)^{m}}=\frac{\left(u^{m}-k+1\right)^{k}}{u^{k m}} \frac{u^{k m}}{\left(u^{k}-1\right)^{m}}$ and prove that it is greater then 1. Indeed due to Bernoully inequality we have $$ \left(1-\frac{k-1}{u^{m}}\right)^{k}\left(\frac{u^{k}}{u^{k}-1}\right)^{m}>\left(1+\frac{m}{u^{k}-1}\right)\left(1-\frac{k(k-1)}{u^{m}}\right) $$ Besides $1-\frac{k(k-1)}{u^{m}}>1-\frac{k^{2}}{u^{m}}$. Hence it is sufficient to prove that $$ 1<\left(1+\frac{m}{u^{k}-1}\right)\left(1-\frac{k^{2}}{u^{m}}\right)=1+\frac{m}{u^{k}-1}-\frac{k^{2}}{u^{m}}-\frac{m k^{2}}{u^{m}\left(u^{k}-1\right)}=1+\frac{m u^{m}-k^{2} u^{k}+k^{2}-m k^{2}}{u^{m}\left(u^{k}-1\right)} $$ But $u^{m} \geqslant k^{2} u^{k}$, therefore, $$ m u^{m}-k^{2} u^{k}+k^{2}-m k^{2}>m k^{2} u^{k}-k^{2} u^{k}+k^{2}-m k^{2}=k^{2}(m-1)\left(u^{k}-1\right)>0 $$ QED. 3.5. Like in previous problems it is sufficient to prove two inequalities $$ \begin{aligned} & u^{7}\left(u^{7}+1\right)\left(u^{7}+2\right)\left(u^{7}+3\right)\left(u^{7}+4\right)<\left(u^{5}+1\right)^{7} \\ & u^{7}\left(u^{7}-1\right)\left(u^{7}-2\right)\left(u^{7}-3\right)\left(u^{7}-4\right)>\left(u^{5}-1\right)^{7} \end{aligned} $$ To prove (8) expand parenthesis $$ \begin{aligned} & u^{35}+10 u^{28}+35 u^{21}+50 u^{14}+24 u^{7}< \\ & u^{35}-7 u^{30}+21 u^{25}-35 u^{20}+35 u^{15}-21 u^{10}+7 u^{5}-1 $$ Collect together summands of the same sign. $$ \begin{aligned} 6 u^{30}+u^{30}+35 u^{21}+35 u^{20}+24 u^{7}+21 u^{10}+1> \\ >10 u^{28}+21 u^{25}+35 u^{15}+50 u^{14}+7 u^{5} \end{aligned} $$ We see that if $u \geqslant 2$ then each summand of lower row less than corresponding one of the upper row. 3.6. This problem and next two ones are taken from [1]. 3.7. As an example one can take a set of 17 integers starting from 2184 . 3.8. ## 4 Общие свойства решений уравнения 4.1. If $x \leqslant k$ all integers between $x+k$ and $\frac{1}{2}(x+k)$ are factors of the left hand side of the equation. According to Bertrand postulate one of them is a prime number $p$. Obviously that the left hand side is not divisible by $p^{2}$. Therefore it could not be an $m$-th power. 4.2. According to Sylvester theorem there is a factor $x+i$ that is divisible by a prime $p>k$. Since other factors at the left hand side are not divisible by $p$ the product could be an $m$-th power only if $x+i$ is divisible by $p^{m}$. Then $x+i \geqslant p^{m} \geqslant(k+1)^{m}$. If at the same time $x \leqslant k^{p}$, then $k^{p}+i \geqslant x+i \geqslant(k+1)^{p}$. Hence $i>p k$, which is wrong. 4.3. If a number $a_{i}$ has a prime divisor $p>k$, then other integers among $x, x+1, \ldots, x+k-1$ are not divisible by $p$. Then $x+i$ is divisible by $p^{d m}$. Since $x+i=a_{i} z_{i}^{m}$, then $p$ is relatively prime with $a_{i}$. We got a contradiction. 4.4. Note, if we find at least 5 integers $a_{i}$ which are divisible by primes 2 and 3 only, then we deduce that the equation has no solutions similarly to the problem 1.10. Such 5 integers really exist, because we have 7 integers $a_{i}$ with divisors 2,3 and 5 only, and at most two of them are divisible by 5 . 4.5. Similarly to the previous problem we would find 5 integers divisible by 2 and 3 only. The integers $a_{i}$ have prime factors 2,3 , and 5 only (each factor in the power 0 or 1 ). The product of all $a_{i}$ is a perfect square. At most 2 of initial consecutive numbers are divisible by 5 , hence at most 2 of $a_{i}$ are divisible by 5 . It is sufficient to consider a case when we have exactly 2 integers $a_{i}$ which are divisible by 5 . This case is possible only if $a_{0}$ and $a_{5}$ are divisible by 5 . Consider 4 integers $x+1, x+2, x+3, x+4$. We know from the problem 1.5 that their product is not a perfect square. Hence the total power of divisor 2 or the total power of divisor 3 in the product $a_{1} a_{2} a_{3} a_{4}$ is odd. It is possible only in two cases: 1) There is only one integer among $a_{1}, a_{2}, a_{3}, a_{4}$ divisible by 2 ; 2) There is only one integer among $a_{1}, a_{2}, a_{3}, a_{4}$ divisible by 3 ; Besides that we have at most 2 integers $a_{i}$ divisible by 2 ; the same is true for 3 . Then it is easy to see that in each case we have among integers $x+1, x+2, x+3, x+4$ two integers either of the form $t^{2}$ or of the form $2 t^{2}$, or of the form $3 t^{2}$. But this is impossible. 4.6. This statement is a part of lemma 1 of [3]. Cancel equal factors. Since $\operatorname{GCD}(n+i, n+j)k^{m}$, we see that any factor at the left hand side does not divide the product at the right hand side. 4.7. Let $a_{i}=a_{j}$, where $0 \leqslant jn+j=a_{j} z_{j}^{m}$, then $z_{i} \geqslant z_{j}+1$. Therefore $k>a_{j} z_{i}^{m}-a_{j} z_{j}^{m}=a_{j}\left(\left(z_{j}+1\right)^{m}-z_{j}^{m}\right)>a_{j} m z_{j}^{m-1} \geqslant a_{j}^{(m-1) / m} z_{j}^{m-1}=\left(a_{j} z_{j}^{m}\right)^{(m-1) / m}=(x+j)^{(m-1) / m}>x^{1 / m}$, that contradicts to the statement of the problem 4.2. 4.8. It follows from the next solution. One can put $u=v=1$. 4.9. This statement is lemma 1 from [3]. Assume that $$ a_{i_{1}} a_{i_{2}}=a_{j_{1}} a_{j_{2}} t^{3} $$ Let us check that $t=1$ and sets of indices coincide. WLOG $\left(x+i_{1}\right)\left(x+i_{2}\right)>\left(x+j_{1}\right)\left(x+j_{2}\right)$ (the equality is impossible due to problem 4.6). Let $t=u / v(\operatorname{GCD}(u, v)=1)$. Then $a_{i_{1}} a_{i_{2}} / u^{3}=a_{j_{1}} a_{j_{2}} / v^{3}$ and both sides are integers. Let $A=a_{i_{1}} a_{i_{2}} / u^{3}=$ $a_{j_{1}} a_{j_{2}} / v^{3}$. By the definition of $a_{i}$ we have $x+i=a_{i} z_{i}^{3}$, then $$ \begin{aligned} & \left(x+i_{1}\right)\left(x+i_{2}\right)=a_{i_{1}} a_{i_{2}} \cdot \frac{s^{3}}{u^{3}} \\ & \left(x+j_{1}\right)\left(x+j_{2}\right)=a_{j_{1}} a_{j_{2}} \cdot \frac{r^{3}}{v^{3}} \end{aligned} $$ where $s=u z_{i_{1}} z_{i_{2}}, r=v z_{j_{1}} z_{j_{2}}$. Then $A s^{3}>A r^{3}$, so $s \geqslant r+1$. Thus $$ \left(x+i_{1}\right)\left(x+i_{2}\right)-\left(x+j_{1}\right)\left(x+j_{2}\right) \geqslant A\left((r+1)^{3}-r^{3}\right)>3 A r^{2} $$ Note that $A r^{3}=\left(x+j_{1}\right)\left(x+j_{2}\right)>x^{2}$, then the last inequality could be rewritten as $$ \left(x+i_{1}\right)\left(x+i_{2}\right)-\left(x+j_{1}\right)\left(x+j_{2}\right) \geqslant 3 A \cdot\left(\frac{x^{2}}{A}\right)^{2 / 3} \geqslant 3 x^{4 / 3} $$ On the other hand $$ \left(x+i_{1}\right)\left(x+i_{2}\right)-\left(x+j_{1}\right)\left(x+j_{2}\right)<(x+k)^{2}-x^{2}<3 k x $$ Obtained estimations contradict to each other, because due to problem 4.2 we know that $x>k^{3}$, and therefore $3 k x<3 x^{4 / 3}$. 4.10. We need to find 20 integers $a_{i}$, which are divisible by primes $2,3,5,7$ only. We know that integers $a_{i}$ are divisible by the primes are not greater than 75 only, i.e. 2, 3, 5, 7, 11,13,17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73. The integers $a_{i}$ are divisors of 75 consecutive integers therefore: 1) for every prime $p \in[41 ; 73]$ (there are 9 prime numbers in this interval) at most 2 of $a_{i}$ are divisible by $p$; 2) for every prime $p \in[29 ; 37]$ (there are 3 prime numbers in this interval) at most 3 of $a_{i}$ are divisible by $p$; 3) at most 4 of $a_{i}$ are divisible by 23; at most 4 of $a_{i}$ are divisible by 19 ; 4) at most 5 of $a_{i}$ are divisible by 17 ; 5 ) at most 6 of $a_{i}$ are divisible by 13 ; 6) at most 7 of $a_{i}$ are divisible by 11 ; So we have at most $$ 18+9+8+5+6+7=53 $$ integers. Hence at least $75-23=22$ integers have prime divisors $2,3,5,7$ only. 4.11. Consider integers $a_{i}$ which have prime divisors $2,3,5,7$ only. By the previous problem we know that at least 20 integers have that property, but for our purposes it is sufficient to take 10 of them. If we have 10 numbers $a_{i}$, we can construct $90=10 \cdot 9$ different formal quotients of the form $a_{i} / a_{j}$. On the other hand, due to the statement of problem 4.9 the quotient of two products of the form $a_{i} a_{j}$ (including quotients of the form $a_{i} / a_{j}$ ) does not equal to qube of rational number. Therefore there exist at most $3^{4}=81$ classes for the values of quotients $a_{i} / a_{j}$ for our 10 numbers (each prime 2, 3, 5, 7 divides this quotient in power $0+3 s, 1+3 s, 2+3 s)$. Hence two quotients belong to the same class and we have a relation $$ \frac{a_{i}}{a_{j}}=\frac{a_{u}}{a_{v}} \cdot t^{3} $$ which is restricted by the statement of problem 4.9. 4.12. This beautiful solution we take from [3]. For each prime $p_{0}B^{\prime}(k)$, where $B^{\prime}(k)$ is a product of first $k$ square free numbers. It is sufficient to prove that $B^{\prime}(k)>(4 / 3)^{k} k$ ! when $k \geqslant 24$. Induction by $k$. Base, $k=24$. $$ \frac{26 \cdot 29 \cdot 30 \cdot 31 \cdot 33 \cdot 34 \cdot 35 \cdot 37}{4 \cdot 8 \cdot 9 \cdot 12 \cdot 16 \cdot 18 \cdot 20 \cdot 24}>\left(\frac{4}{3}\right)^{24} $$ Step of induction. For $r \geqslant 9$ the number of square free integers that does not exceed $r$ is at most $r-\left[\frac{r}{4}\right]-1<\frac{3}{4} r$. Therefore for $n$-th square free integer for $n \geqslant 7$ is at most $\frac{4}{3} n$. ## REFERENCES [1] Флейшман Д. Китайская теорема об остатках и гипотеза Ченцова // Квант. 1997. № 3. [2] Сендеров В. Об одной замечательной теореме Эрдёша // Квант. 2008. В печати. [3] Erdós P., Selfridge J.L. The product of consecutive integers is never a power // Illinois Math. J. 1975. Vol. 19. P. 292-301. [4] Poorten A. van der, Woeginger G. Squares from products of consecutive integers // Amer. Math. Monthly. 2002. Vol. 109. № 5. P. 459-461. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-019.jpg?height=365&width=320&top_left_y=89&top_left_x=248) ## Кубик Рубика и проблема Хигмана Объектом нашего изучения будет кубик Рубика и похожие на него головоломки. Прежде, чем приступить к изучению непосредственно кубика Рубика, мы решим несколько подготовительных задач. ## Подготовительные задачи. P1. В ряд стоят 12 кубиков, пронумерованных по порядку числами от 1 до 12. (На самом левом - 1, на самом правом - 12). При каждом ударе волшебного барабана какие-то два соседних кубика меняются местами. После 333 ударов каждый кубик начинает прыгать от счастья, если видит справа от себя нечётное число кубиков с меньшим номером. Могут ли запрыгать от счастья ровно 6 кубиков? P2. В ряд стоят 42 кубика, пронумерованных в порядке возрастания. За один ход можно переставить местами любые два. Может ли ровно через 2008 ходов оказаться так, что первые два кубика поменялись местами, а остальные оказались на тех же местах? P3. По кругу стоят 20 разноцветных кубиков. За один ход можно поднять любые три и поставить первый на место второго, второй на место третьего, третий - на место первого. Могло ли так получиться, что после какого-то хода все кубики оказались сдвинуты по циклу на один по сравнению с первоначальным положением? P4. Грани кубика раскрашены в различные цвета. Из нескольких таких кубиков выложен прямоугольник $m \times n$. Можно выбрать любой ряд кубиков (по вертикали или горизонтали) и повернуть все кубики одновременно относительно горизонтали (или вертикали). Докажите, что все кубики можно повернуть вверх одинаковыми гранями. ## А. Кубик Рубика. Для определённости кубом будем называть весь большой куб, а кубиками - маленькие кубики, из которых он состоит. Любую грань куба, состоящую из 9 кубиков, можно повернуть по часовой или против часовой стрелки. Можно провести несколько таких поворотов подряд, это будет называться комбинация поворотов или просто комбинация. Грани куба будем обозначать заглавными буквами, например, $A, B, C$. Поворот по часовой стрелки соответствующей грани будем обозначать той же буквой, например $A$. Поворот против часовой стрелки $A^{-1}$. Комбинацию поворотов будем записывать как последовательность букв: например, запись $A B A^{-1} C$ означает, что сначала поворачивают грань $A$ по часовой стрелке, потом грань $B$ по часовой стрелке, потом грань $A$ против часовой стрелки, потом грань $C$ по часовой стрелке. Для поворотов (или комбинаций) $X$ и $Y$ назовём их коммутатором комбинацию $X Y X^{-1} Y^{-1}$. Кубики бывают трёх видов: центральные - в серединах граней, угловые в углах куба, и средние - в серединах рёбер куба. Ясно, что центральные кубики не меняют своего положения относительно друг друга, угловые всегда остаются угловыми, а средние - средними. Представим, что средние и угловые кубики ни с чем не скреплены, то есть их можно свободно вытаскивать и переставлять. При этом средние кубики меняются местами со средними, а угловые - с угловыми так, что грани маленьких кубиков, смотрящие наружу до перестановки, остаются наружными и после неё. (Центральные же кубики никогда переставлять не будем.) Любое положение, достигаемое такой перестановкой, назовём состоянием. Будем говорить, что кубик, все грани которого одноцветны, находится в правильном состоянии. Будем говорить, что маленький кубик находится в правильном положении, если цвета его граней такие же, как и при правильном состоянии всего куба. Если из одного состояния можно с помощью некоторой комбинации получить другое, то эти состояния назовём связанными. Разрешённое состояние - состояние, связанное с правильным. A1. Некоторая комбинация поворотов вывела куб из начального состояния. Докажите, что если повторить ее еще несколько раз, можно опять получить начальное состояние. A2. Существует ли такая универсальная комбинация поворотов, применяя которую разное количество раз, можно собирать кубик из любого разрешенного положения? А3. Придумайте комбинацию поворотов, позволяющую циклически переставить кубики 1, 2, 3 и оставляющую остальные средние кубики на своих местах (рис. 1). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-021.jpg?height=584&width=1104&top_left_y=650&top_left_x=478) Рисунок 1. A4. Покажите, что комбинация $A^{-1} C^{-1} B^{-1} A^{-1} B A C$ меняет местами кубики 1 и 2 и оставляет остальные средние кубики на своих местах (рис. 2). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-021.jpg?height=571&width=491&top_left_y=1579&top_left_x=480) Рисунок 2. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-021.jpg?height=571&width=488&top_left_y=1582&top_left_x=1098) Рисунок 3. - A5. Найдите комбинацию, позволяющую одновременно повернуть в своих гнёздах кубики 1 и 2 и оставляющую остальные средние кубики на тех же местах и в тех же положениях (рис. 3)? - А6. Докажите, что не существует комбинации, позволяющей повернуть кубик 1 в своём гнезде и оставляющей остальные средние кубики на тех же местах и в тех же положениях (рис. 3). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-022.jpg?height=197&width=174&top_left_y=144&top_left_x=193) A7. Допустим, состояние куба - разрешённое. Опишите, как расставить все средние на свои места. Пусть состояние куба не обязательно разрешенное. Рассмотрите состояния средних кубиков и ппишите все связанные возможные положения. А8. Найдите нетривиальную комбинацию поворотов, такую, что будучи повторенной трижды, она не вызывает никаких изменений. А9. Придумайте комбинацию поворотов, позволяющую циклически переставить кубики 1, 2, 3 (рис. 4), не меняющую положение остальных угловых кубиков и оставляющую все средние кубики на тех же местах и в тех же положениях. A10. Допустим, состояние куба - разрешённое. Опишите, как расставить все угловые кубики на свои места, не меняя положений средних кубиков. Пусть состояние куба не обязательно разрешенное. Рассмотрите положения угловых кубиков и опишите возможные связанные состояния. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-022.jpg?height=577&width=500&top_left_y=1442&top_left_x=481) Рисунок 4. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-022.jpg?height=574&width=500&top_left_y=1438&top_left_x=1098) Рисунок 5. A11. Докажите, что не существует комбинации поворотов, позволяющих повернуть только один угловой кубик и оставляющих все остальные кубики в тех же положениях. A12. Придумайте комбинацию поворотов, позволяющую одновременно повернуть кубики 1, 2, 3 (рис. 5) на 120 градусов по часовой стрелке и не меняющую положение остальных кубиков. A13. Опишите, как собрать кубик Рубика, если известно, что состояние разрешённое. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-023.jpg?height=195&width=172&top_left_y=148&top_left_x=194) A14. Как по положению угловых кубиков определить, можно ли собрать кубик Рубика, если они стоят на своих местах, а средние находятся в правильных положениях? A15. Найдите максимальное количество попарно несвязанных состояний кубика. A16. Посчитайте число разрешённых состояний кубика Рубика. ## Цикл В. В задачах этого цикла мы разберём несколько похожих ловоломок. Будем считать две комбинации (поворотов) различными, если, применяя их к одному состоянию, мы получаем разные результаты. B1. На шахматной доске расставили все натуральные числа от 1 до 64 . Разрешается выбрать любой квадратик $2 \times 2$ и переставить числа в нем по часовой стрелке. Докажите, что используя эту операцию, можно добиться любой расстановки. - В2. Рассмотрим кубик $2 \times 2 \times 2$. Опишите все разрешенные состояния. Сколько их существует? ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-023.jpg?height=480&width=757&top_left_y=1756&top_left_x=658) Рисунок 6 . В3. Рассмотрим игру «Венгерские кольца» (рис. 6). Это плоская головоломка, состоящая из двух или более связанных овалов, на каждом из которых отмечено несколько пронумерованных кружков. Некоторые из этих кружков могут принадлежать более чем одному овалу. Ход состоит в движении одного из овалов на одно или несколько делений, вместе со всеми его кружками. Расстояния между кружками-делениями одинаковые. Кружок, лежащий более ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-024.jpg?height=195&width=183&top_left_y=148&top_left_x=183) чем на одном овале, может двигаться вместе с любым из них. Для простоты, рассмотрим головоломку с двумя овалами, каждый из которых содержит 6 кружков. Кружки 1 и 3 могут перемещаться с любым из овалов. Обратите внимание, что каждый ход отвечает своей перестановке на множестве чисел $\{1,2, \ldots$, $10\}$. Опишите все возможные допустимые состояния. - B4. Рассмотрим игру «Экватор» (рис. 7). Головоломка состоит из сферы, опоясанной 3 лентами, каждая из которых разделена на 12 частей, имеющих форму части сферы. Любые две ленты пересекаются под прямым углом и имеют два общих куска, назовем их узлами. Всего есть 6 узлов. Разрешено сдвигать любую ленту так, чтобы ее части переходили друг в друга. Общее число перемещающихся частей $3 \times 12-6=30$. Опишите все возможные допустимые состояния. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-024.jpg?height=648&width=659&top_left_y=1304&top_left_x=710) B5. Рассмотрим кубик $4 \times 4 \times 4$. Опишите все возможные допустимые состояния. B6. Рассмотрим игру «15». В квадрате $4 \times 4$ расположены 15 пронумерованных плиток. Одно поле остаётся свободным. За один ход разрешается выбрать соседнюю по стороне со свободным полем плитку и переместить её на свободное место. Опишите все возможные допустимые состояния игры. - В7. Каково максимальное количество несвязанных состояний у куба $4 \times 4 \times 4$ ? - В8. Найдите систему инвариантов для куба $4 \times 4 \times 4$. - B9. Найдите систему инвариантов для куба $n \times n \times n$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-025.jpg?height=200&width=186&top_left_y=146&top_left_x=181) ## Цикл С. - C1. Рассмотрим правильный тетраэдр. Его можно повернуть так, что он перейдёт в себя, но при этом какие-то вершины и ребра могут поменяться местами. Сколько различных движений существует? - $\mathrm{C} 2$. Тот же вопрос для куба. Как ведут себя большие диагонали куба при движениях? Пусть у нас есть правильный многогранник. Аналогично поворотам в кубике Рубика и других головоломках, мы можем провести комбинацию двух движений. Ясно, что эта комбинация сама по себе является движением, переводящим многогранник в себя. Будем считать такую комбинацию двух движений произведением. Будем называть единичным такое движение при котором многогранник вообще не сдвигается, и все вершины остаются на своих местах. Если какоенибудь движение «умножить» на единичное движение, то ничего не изменится. - C3. Пусть есть $n$ элементов. Рассмотрим все преобразования, переставляющие эти $n$ элементов в другом порядке. Например, если $n=3$, то таких перестановок 6 : 1) поменять местами 1 и 2,3 оставить на месте; 2) поменять местами 2 и 3,1 оставить на месте; 3) поменять местами 1 и 3,2 оставить на месте; 4) поставить 1 на место 2,2 на место 3,3 на место 1 (цикл длины 3); 5) поставить 1 на место 3,3 на место 2,2 на место 1 (второй цикл длины 3 ); 6) оставить все элементы на месте. Аналогично можно выписать все перестановки в случае любого $n$. Произведением двух перестановок назовём перестановку, полученную применением сначала первой, потом второй. Какая перестановка будет единичной? Проверьте, что для каждого движения $A$ найдётся движение $A^{-1}$ такое, что $A A^{-1}=A^{-1} A=E$, где $E$ - единичное движение. Проверьте выполнение закона $(A B) C=A(B C)$. C4. Рассмотрим множество $A$ - все движения куба, переводящие его в себя, и множество $B$ - все возможные перестановки четырёх элементов. Проведите соответствие между этими множествами так, чтобы произведение в одном множестве соответствовало произведению в другом. Определение. Множество $G$, для которого выполнены следующие требования: 1) для любых двух элементов определено произведение, подчиняющееся ассоциативному закону $(A B) C=A(B C)(A, B, C \in G)$; 2) присутствует единичный элемент $E \in G: A E=E A=A$ для любого $A \in G$; 3) для каждого элемента $A \in G$ есть обратный ему $A^{-1} \in G: A A^{-1}=$ $A^{-1} A=E ;$ называется группой. Группа перестановок из задачи С3 обозначается $S_{n}$. - C5. Докажите, что следующие множества с операциями являются группами: 1) множество целых чисел по сложению; 2) множество положительных рациональных чисел по умножению; 3) множество последовательностей поворотов кубика Рубика относительно композиции. - C6. Являются ли группами следующие множества: 1) множество рациональных чисел по умножению; 2) множество слов в конечном алфавите (включая пустое слово) относительно приписывания (конкатенации) одного слова к другому; 3) множество слов из букв $\{a, b, c\}$ (включая пустое слово), при условии, что можно заменять любое из слов $X a b c Y, X b c a Y, X c a b Y$ на $X Y$ для любых слов $X$ и $Y$ (то есть вычёркивать $a b c, b c a, c a b$ из любых слов), а также делать обратную операцию (добавлять соответсвующие слова); 4) множество двойных транспозиций четырёх элементов (12)(34), (13)(24), (14)(23) и тождественное преобразование? Примечание: запись (123)(4567) означает, что в каждой скобке элементы меняются по циклу: $1 \rightarrow 2 \rightarrow 3 \rightarrow 1$ и $4 \rightarrow 5 \rightarrow 6 \rightarrow 7 \rightarrow 4$. Определение. Пусть $G$ - группа, $H$ - подмножество $G$. Если $H$ содержит единицу (так мы будем называть единичный элемент) группы $G$, а также все произведения элементов из $H$ и их обратные, то $H$ называется подгрупnoй. C7. Пусть $H$ - подгруппа в $G$. Докажите, что $H$ - группа. - C8. Найдите все подгруппы $S_{3}$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-027.jpg?height=197&width=174&top_left_y=147&top_left_x=193) - C9. Теорема Лагранжа. Докажите, что количество элементов в любой группе делится на число элементов в любой её подгруппе, - C10. Найдите подгруппу из $\frac{n!}{2}$ элементов в $S_{n}$, где $n \geqslant 2$. Группа из предыдущей задачи обозначается $A_{n}$. - C11. Докажите, что любой элемент $A_{n}$ является произведением тройных циклов - перестановок, представляющих собой цикл длины 3). Определение. Элемент $a b a^{-1} b^{-1}$ называется коммутатором элементов $a$ и $b$. Определение. Коммутантом групы $G$ называется подгруппа из всевозможных произведений коммутаторов. - C12. Найдите коммутанты групп $S_{3}, A_{3}, A_{4}, S_{n}, A_{n}$. Выберем в некоторой группе $G$ элемент $a$. Каждому элементу группы $g$ поставим в соответствие элемент $a^{-1} g a$. Этот элемент называется сопряженным к $g$ относительно $a$, или просто сопряженным. - C13. Пусть $H$ - подгруппа $G$. Докажите, что $a^{-1} H a$ - множество элементов, сопряженных к $H$ - тоже подгруппа. Подгруппы $H$ и $a^{-1} H a$ называются сопряженными. Определение. Подгрупа, которая при всех сопряжениях переходит в себя, называется нормальной. - C14. Докажите, что коммутант группы, единичный элемент и вся группа являются нормальными подгруппами. В любой группе всегда есть две тривиальные нормальные подгруппы единичный элемент и вся группа целиком. Остальные нормальные подгруппы, если они есть, называются нетривиальными. Групы, в которых нет нетривиальных нормальных подгрупп, называются простыми. - C15. При каких $n S_{n}$ является простой группой? C16. Докажите, что $A_{n}$ - простая при $n \geqslant 5$. - C17. Каким из групп $S_{n}$ или $A_{n}$ соответствуют группы движений куба и тетраэдра. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-028.jpg?height=197&width=186&top_left_y=141&top_left_x=181) - C18. Докажите, что группа движений икосаэдра соответствует $A_{5}$. - C19. Придумайте группу из 8 элементов такую, что в ней найдутся два элемента $a$ и $b$, для которых $a b \neq b a$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-029.jpg?height=195&width=172&top_left_y=148&top_left_x=194) ## Цикл D. Группы. Определение. Группы $G$ и $H$ называются изоморфными, если между ними можно провести взаимно однозначное соответствие (изоморфизм) $\varphi$, такое, что единица переходит в единицу и произведение двух элементов в одной группе переходит в произведение соответствующих двух элементов в другой группе: $\varphi\left(g_{1} g_{2}\right)=\varphi\left(g_{1}\right) \times \varphi\left(g_{2}\right)$. - D1. Докажите, что следующие пары групп изоморфны: 1) група движений куба и $S_{4}$; 2) группа действительных чисел по сложению и группа параллельных переносов вдоль оси $O X$; 3) группа целых чисел по сложению и группа чисел вида $2^{k}$ (для целых $k$ ) по умножению; 4) группа движений додекаэдра (икосаэдра) и $A_{5}$. Изоморфные группы, это, в некотором смысле, одинаковые группы. Пусть $G$ - группа, $M$ - множество. Говорят, что $G$ действует на $M$, если каждому $m \in M$ и $g \in G$ соответствует элемент $m^{\prime}=g(m)$, при этом $\left(g_{1} g_{2}\right) m=g_{1}\left(g_{2} m\right)$ для любого $m \in M$. ## Примеры. 1. Группа всех движений действует на трехмерном пространстве. 2. Группы $S_{n}$ и $A_{n}$ действуют на множестве $\{1, \ldots, n\}$. 3. Группа действует сама на себе левыми умножениями: каждый элемент $h \in G$ определяет отображение $\varphi_{h}(g)=h g$. 4. Группа $G$ действует сама на себе сопряжениями: каждому $h \in G$ соответствует отображение $\varphi_{h}(g)=h^{-1} g h$. - D2. Проверьте, что это действительно действия. - D3. Пусть $n$ - количество элементов в группе $G$ (обозначение $n=|G|$ ). Докажите, что в $S_{n}$ можно найти подгруппу, которая изоморфна $G$. - D4. Группа $G$ раскрашена в несколько цветов так, что цвет произведения зависит только от цветов сомножителей. Единица красная. Докажите, что множество красных элементов образует нормальную подгруппу. Пусть $H$ подгруппа в $G$. Левой раскраской называется раскраска элементов $G$ в несколько цветов, такая, что ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-030.jpg?height=200&width=183&top_left_y=146&top_left_x=183) 1) для любого $g \in G$ и $h \in H$ элементы $g$ и $h g$ раскрашены в один цвет; 2) если $g_{1}$ и $g_{2}$ раскрашены в один цвет, то $g_{1}=h g_{2}$ для некоторого $h \in H$. Аналогично определяется правая раскраска. ## $\checkmark$ D5. a) Докажите, что левая раскраска совпадает с правой тогда и только тогда, когда $H$ - нормальна. b) Докажите, что если $H$ - нормальна, то цвет произведения однозначно определяется цветами сомножителей, а цвет обратного элемента однозначно определяется цветом элемента. Пусть $H$ - подгруппа в $G$. Для любого $g \in G$ можно рассмотреть множество $g H$ - совокупность элементов $g h_{i}$ для различных $h_{i} \in H$. Это множество называется левым классом смежности по подгруппе $H$ (обратите внимание, что это то же самое, что одноцветные элементы при левой раскраске). Каждый элемент $G$ лежит в каком-то смежном классе, причем, только в одном. Два элемента лежат в одном и том же классе, если они оба представляются в виде $g h_{i}$ для различных элементов $h_{i} \in H$. Каждый элемент $g_{1}=g_{1} e$ содержится в классе $g_{1} H$ (так как $e \in H$ ). Элемент $g_{2}$ попадает в этот же класс, если существует такой $h$, что $g_{2} h=g_{1}$, или, что тоже самое, $g_{1} g_{2}^{-1}=h$. Произведением двух смежных классов $g_{1} H$ и $g_{2} H$ назначим смежный класс $g_{1} g_{2} H$. Конечно, мы можем выбрать других представителей этих двух смежных классов, например, $g_{1}^{\prime} \in g_{1} H$ и $g_{2}^{\prime} \in g_{2} H$. вместо $g_{1}$ и $g_{2}$. Тогда произведением будет класс $g_{1}^{\prime} g_{2}^{\prime} H$. D6. Докажите, что произведение не зависит от выбора представителей, то есть $g_{1} g_{2} H=g_{1}^{\prime} g_{2}^{\prime} H$. Аналогично можно назначить обратный элемент на множестве смежных классов: обратным к классу $g H$ будет класс $g^{-1} H$. Множество смежных классов (или множество цветов из задачи D5) можно рассматривать как группу. Эта группа называется факторгруппой по нормальной подгруппе. Ее элементами являются смежные классы. - D7. Найдите факторгруппы $G$ по $H$ : 1) $G=S_{n}$ и $H=A_{n}$ 2) $G=A_{4}$ и $H$ - группа двойных транспозиций четырёх элементов (12)(34), (13)(24), (14)(23) плюс тождественное преобразование; 3) $G$ - группа действительных чисел по сложению и $H$ - подгруппа целых чисел. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-031.jpg?height=197&width=174&top_left_y=147&top_left_x=193) ## D8. a) Докажите, что группа комбинаций поворотов кубика $2 \times 2 \times 2$ есть факторгруппа группы комбинаций поворотов кубика Рубика $3 \times 3 \times 3$. b) Докажите, что группа комбинаций поворотов кубика Рубика $3 \times 3 \times 3$, а также кубика $4 \times 4 \times 4$ есть факторгруппа группы поворотов кубика $5 \times 5 \times 5$. Определение. Орбитой элемента $m \in M$ называется множество $\left\{g_{i} m\right\}$ для всех различных $g_{i} \in G$. D9. Опишите орбиты элементов при следующих действиях: 1) угловой кубик при действии группы комбинаций поворотов кубика Рубика; 2) произвольная точка плоскости при действии группы параллельных переносов вдоль оси $O X$; 3) цикл из трех элементов при действии группы перестановок на себе сопряжениями. D10. Докажите, что любые две орбиты либо не пересекаются, либо совпадают. Пусть задано действие $\varphi$ группы $G$ на множестве $M$. Некоторые элементы группы не сдвигают элементов множества: Определение. Стабилизатором элемента $m \in M$ называется множество элементов $g \in G$ таких, что $g(m)=m$. Определение. A set of elements $g \in G$ is called a stabilizer of en element $m \in M$ if $g(m)=m$. D11. Докажите, что стабилизатор Stab $m$ есть подгруппа. Докажите, что $\left|O_{m}\right| \cdot|\operatorname{Stab} m|=|G|$. Определение. Действие называется хорошим, если стабилизатор любого элемента состоит из одной единицы. Определение. An act is called good if stabilizer of any element is unity only. Рассмотрим два действия группы $G$ на множествах $M_{1}$ и $M_{2}$. Пусть множества взаимно однозначно соответствуют друг другу: каждому элементу $M_{1}$ соответствует свой элемент $M_{2}$, и наоборот: $\Psi\left(M_{1}\right)=M_{2}$. Тогда эти два действия $\varphi$ и $\psi$ называются одинаково устроенными или сопряженными, если группа действует на этих множествах одинаковым образом: элементы $g_{\varphi}\left(M_{1}\right)$ и $g_{\psi}\left(M_{2}\right)$ соответствуют друг другу при отображении $\Psi$. В качестве множеств $M_{1}$ и $M_{2}$ может выступать одно и тоже множество M. D12. а) Пусть даны два хороших действия $G \varphi_{1}$ и $\varphi_{2}$ на множестве $M$. Обязательно ли они сопряжены? b) Пусть даны два хороших действия $G \varphi_{1}$ и $\varphi_{2}$ на множестве $M$ и количество орбит в этих действиях одинаково (либо счетно в обоих случаях). Докажите, что они сопряжены. Пусть $G$ - группа, $g$ - ее элемент. Вместе с каждым элементом $h$ можно рассмотреть элемент $g^{-1} h g$. При этом, произведению $h_{1} h_{2}$ соответствует элемент $g^{-1} h_{1} g g^{-1} h_{2} g=g^{-1} h_{1} h_{2} g$. То есть, $h \rightarrow g^{-1} h g-$ взаимно однозначное соответствие, которое произведение переводит в произведение. Таким образом, получается изоморфизм группы на себя. - D13. Пусть $G$ - группа, $H_{1}$ и $H_{2}$ - изоморфные подгруппы, $\varphi: H_{1} \rightarrow$ $H_{2}$ - изоморфизм. Всегда ли $\varphi$ можно продолжить до изоморфизма всей группы на себя? ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-033.jpg?height=195&width=174&top_left_y=148&top_left_x=193) D14. Пусть $G$ - группа, $H_{1}$ и $H_{2}$ - изоморфные подгруппы $G$ такие, что количества элементов в смежных классах по $H_{1}$ и по $H_{2}$ одинаковы (либо счётны в обоих случаях). Докажите, что тогда существует такая группа $G^{\prime}$, что $G$ - её подгруппа в $G^{\prime} t \in G^{\prime}$ и в этой группе для любого $h \in H_{1}$ выполнено: $$ t h t^{-1}=\varphi(h) \in H_{2} $$ D15. То же, но если в $H_{1}$-классах и $H_{2}$-классах не обязательно равное количество элементов. Определение. Группа $G$ порождена элементами $x_{i}$ (мы будем писать $G=$ $\left\langle x_{i}\right\rangle$ ), если любой элемент из $G$ является произведением каких-то элементов из $\left\{x_{i}\right\}$. Определение. Группа называется $n$-порождённой, если $\left\{x_{i}\right\}$ состоит из $n$ элементов. Определение. Група называется свободно $n$-порождённой, если она изоморфна группе слов в алфавите $\left\{g_{1}, g_{1}^{-1}, g_{2}, g_{2}^{-1}, \ldots, g_{n}, g_{n}^{-1}\right\}$. D16. Пусть группа $G$ содержит пары изоморфных подгрупп $\varphi_{i}: H_{i} \rightarrow H_{i}^{\prime}$ при $i=1 \ldots n$. Докажите, что тогда существует такая группа $G^{\prime}$, что $G-$ её подгруппа и в этой группе выполнено: 1) $t_{i} h_{i} t_{i}^{-1}=\varphi_{i}\left(h_{i}\right)$ 2) $\left\langle t_{i}\right\rangle$ порождена ими свободно. D17. Докажите, что если группа $G$ свободно порождена элементами $t_{i}$ $i \in \mathbb{N}$, то существует изоморфизм между $G$ и $H=\left\langle t_{2}, t_{3}, \ldots\right\rangle$. D18. Докажите, что любая счетная группа может быть представлена в качестве подгрупы 3-порождённой группы. ## Цикл Е. Группы и сборка кубика Рубика высших порядков Мы говорим, что куб почти собран, если все кубики стоят на своих местах, но возможен непорядок с ориентацией. Следующие несколько пунктов относятся к почти сборке кубика. E1. а) Даны два зацепляющихся цикла длины 4 с одной общей вершиной. Докажите, что они порождают всю группу $S_{7}$. b) Даны два зацепляющихся цикла длины 4 с парой общих соседних вершин. Докажите, что они порождают всю группу $S_{6}$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-034.jpg?height=197&width=186&top_left_y=147&top_left_x=181) E2. а) Покажите, что группа $A_{12}$ порождается 11 -членными циклами. b) Пусть $x \in S_{8}$. Докажите, что тогда $x^{8}$ ! есть единичная перестановка. Пусть $s$ - 11-членный цикл. Докажите, что тогда $s=t^{8}$ для некоторого 11 -членного цикла $t$. c) Докажите, что если переставить в кубе $3 \times 3 \times 3$ средние и угловые кубики перестановками одинаковой чётности, то кубик почти собирается. d) Докажите, что любая расстановка в кубике $4 \times 4 \times 4$ почти собирается. Определение. Говорят, что группа $G$ является прямой суммой групп $G_{1}$ и $G_{2}$, если она состоит из пар $\left(g_{1}, g_{2}\right)$, таких что $g_{1} \in G_{1}, g_{2} \in G_{2}$. При этом произведение пар подчиняется закону $\left(g_{1}, g_{2}\right) \times\left(h_{1}, h_{2}\right)=\left(g_{1} h_{1}, g_{2} h_{2}\right)$. Прямая сумма обозначается как $G=G_{1} \oplus G_{2}$. - E3. Пусть $G_{1}$ есть простая конечная группа с образующими $a_{1}, \ldots, a_{k}, G_{2}$ есть простая конечная группа с образующими $b_{1}, \ldots, b_{k}, G=G_{1} \oplus G_{2}, H$ есть подгруппа в $G$ порожденная элементами $z_{i}=\left(a_{i}, b_{i}\right), i=1, \ldots, k$. Докажите, что либо $H=G$, либо для некоторого изоморфизма $\varphi: G_{1} \rightarrow G_{2}$ выполняется равенство $b_{i}=\varphi\left(a_{i}\right)$. Докажите, что группа комбинаций поворотов кубика $3 \times 3 \times 3$ содержит $A_{8} \oplus A_{12}$ в качестве подгруппы. - E4. Докажите, что любая расстановка в кубике $2 n \times 2 n \times 2 n$ почти собирается. - E5. Докажите, что куб $2 \times \ldots \times 2$ почти собирается в любой размерности. - E6. Докажите, что куб $(2 n)^{m}$ почти собирается в любой размерности. Из того факта, что куб $2 \times 2 \times 2 \times 2$ почти собирается, следует полезное наблюдение. Пусть можно развернуть $k$ угловых кубиков. Тогда можно развернуть любые другие $k$ угловых кубиков аналогичным образом. Это помогает исследовать ситуацию с полной сборкой кубика. E7. Докажите, что в четырёхмерном пространстве куб $2 \times 2 \times 2 \times 2$ имеет 3 класса связанных состояний. Указание: Используйте то, что факторгрупnа $A_{4}$ по группе пар транспозиций (пример 2 из задачи $\mathrm{D}$ 7) это группа из трёх элементов. - E8. Докажите, что в пространстве размерности 5 и выше у куба $2 \times 2 \times \ldots \times 2$ все состояния разрешённые (инвариантов нет). E9. Найдите число классов связанных состояний у куба $3 \times 3 \times \ldots \times 3$ для пространств размерности 4 и выше. E10. Найдите число классов связанных состояний у куба $n \times n \times \ldots \times n$ для пространств размерности 4 и выше. ## Кубик Рубика с высшей точки зрения. 9 августа 2008 г. Мы поместили Квантовскую статью и книгу, посвященную Кубику в раздаточные материалы. В статье содержится решение п.а), а в книге серии а, б) и необходимую теоретико групповую технику. Мы не исключаем, что часть пунктов перейдет в заочный конкурс, поэтому здесь мы постараемся прежде всего показать логику исследования, каким образом общий подход позволяет разобраться в ситуации, как математик, не имея дела с кубиком может создать общую теорию. Допустим, мы ничего не знаем про кубик и нам надо разобраться в общей ситуации. Что значит общая ситуация? Это - кубик $n \times n \times n$ - многомерный куб. - прочие головоломки. Нелепо сразу пытаться разобраться в общей ситуации. Нужно - сперва разобрать классический кубик - кубик $4 \times 4 \times 4$ - кубик $5 \times 5 \times 5$ - кубик $n \times n \times n$ Рассмотрим $k$-мерный куб $n \times \cdots \times n$. Он разбит на $n^{k}$ кубиков. Разрешается взять один слой из $n^{k-1}$ кубика, представляющий собой ( $k-1$ )мерный куб и повернуть его как единое целое внутри слоя так чтобы он совместился. (Так, в случае обычного трехмерного куба действие заключается в повороте слоя.) Каждая грань такого куба раскрашена в свой цвет. Вопрос заключается в возможности сборки. Если $n=2$, то такой куб называется дамским. Ключевое значение имеет - многомерный дамский куб. Поскольку группа движений $n$-мерного кубика, переводящих фиксированную вершину в себя есть $A_{n}$, группа самосовмещений углового кубика есть $A_{n}$. Поэтому дамский куб ведет себя по-разному в размерностях 3,4 ,(когда эта группа разрешима) и в размерности 5 (когда она проста и некоммутативна). Высшие размерности должны себя вести как пятимерье. - общий случай. С чего начать при изучении классического кубика? С очевидных, но полезных наблюдений. 1) Центральные и угловые кубики не смешиваются. 2) Можно временно забыть об ориентации - от ориентаций не зависят перестановки кубиков. 3) Центральные кубики неподвижны. 4) Интересно взаимодействие соседних граней. Отсюда - программа исследований. ## 1. Изучить почти сборку кубика. (Мы говорим, что куб почти собран, если все кубики стоят на своих местах, но возможен непорядок с ориентацией.) ## 2. Изучить ориентации. Чтобы осуществить (1) нужно a) Изучить почти сборку углов. б) Изучить почти сборку ребер. в) Изучить взаимодействие а) и б), лучше всего - научиться расщеплять а) и б). Но прежде всего надо рассмотреть ключевую ситуацию - взаимодействие поворотов двух соседних граней. Это составляет содержание задачи E1 а) Даны два зацепляющихся цикла длины 4 с одной общей вершиной. Докажите, что они порождают всю группу $S_{7}$. b) Даны два зацепляющихся цикла длины 4 с парой общих соседних вершин. Докажите, что они порождают всю группу $S_{6}$. Теперь интуитивно ясно, как можно почти собрать дамский куб почти собирается в любой размерности. Ясно также, что что множества реберных и угловых кубиков кубика $3 \times 3 \times 3$ почти собирается, правда по-отдельности. Более того, видно, как установить что множество реберных кубиков каждого сорта кубика $n \times n \times n$ почти собирается, правда - по отдельности. Ясно также, что аналогичное утверждение для многомерного кубика доказывается. Теперь нам надо понять, как почти собирать кубики разных сортов вместе. Отсюда - такие задачи. E3 a) Покажите, что группа $A_{12}$ порождается 11-членными циклами. б) Пусть $s$ - 11-членный цикл. Тогда $s=s^{88}$ для некоторого 11членного цикла $s^{\prime}$. Пусть $x \in S_{8}$. Тогда $x^{8!}$ есть единичная перестановка. в) Если переставить реберные и угловые кубики перестановками одинаковой четности, то такая перестановка почти собирается. г) Докажите, что любая расстановка в кубике $4 \times 4 \times 4$ почти собирается. Приведенное решение все же использует специфику чисел 8 и 12 - а именно то, что между ними есть простое число. Основное наблюдение надо обобщить. Отсюда возникает следующая задача Пусть $G_{1}$ есть простая конечная группа с образующими $a_{1}, \ldots, a_{k}$, $G_{2}$ есть простая конечная группа с образующими $b_{1}, \ldots, b_{k}, G=G_{1} \oplus G_{2}$, $H$ есть подгрупи в $G$ порожденная элементами $z_{i}=\left(a_{i}, b_{i}\right), i=1, \ldots, k$. Доказать, что либо $H=G$, либо для некоторого изоморфизма $\varphi$ : $G_{1} \rightarrow G_{2}$ выполняется равенство $b_{i}=\varphi\left(a_{i}\right)$. Хочется уметь расщеплять несколько объектов одновременно. Тогда мы приходим к следующему утверждению: Пусть $G_{1}$ есть простая конечная группа с образующими $a_{1}, \ldots, a_{k}$, $G_{2}$ есть простал конечная группа с образующими $b_{1}, \ldots, b_{k}, G_{3}$ есть простал конечная группа с образующими $c_{1}, \ldots, c_{k}, G=G_{1} \oplus G_{2} \oplus G_{3}$, $H$ есть подгруппа в $G$ порожденная элементами $z_{i}=\left(a_{i}, b_{i}, c_{i}\right), i=$ $1, \ldots, k$. Тогда либо $H=G$, либо для некоторого изоморфизма $\varphi: G_{1} \rightarrow G_{2}$ выполняется равенство $b_{i}=\varphi\left(a_{i}\right)$ и при этом $H=H_{12} \oplus G_{3}$, где $H_{12}=$ $\{(x, \varphi(x))\}$ есть подгруппа в $G_{1} \oplus G_{2}$, либо аналогично $H=H_{23} \oplus G_{1}$ либо аналогично $H=H_{31} \oplus G_{2}$, либо для некоторых изоморфизмов $\varphi$ : $G_{1} \rightarrow G_{2}$ и $\psi: G_{1} \rightarrow G_{3}$ выполняются равенства $b_{i}=\varphi\left(a_{i}\right) c_{i}=\psi\left(a_{i}\right)$. Можно сформулировать аналогичное утверждение для прямых сумм от произвольного числа элементов. Теперь можно разобраться с почти сборкой. При этом лучше начать с изучения полной системы инвариантов для почти-сборки куба $5 \times 5 \times 5$, перейти к кубу $n \times n \times n$, а затем изучить многомерье. Теперь займемся ориентацией. Все реберные инварианты нам известны. Однако из решения не совсем ясно, как их можно придумать. Поэтому полезно получить общее объяснение, позволяющее понять и общую ситуацию (пусть даже неоправданно длинное если иметь в виду только кубик Рубика). Хулиган Вася может взять реберный кубик (стоящий на некотором месте) и, перевернув его, поставить на исходное место. Из данной расстановки Васиными действиями можно получить $2^{12}$ степени различных расстановок. $e_{i}$ означает Васино действие с $i$-м ребром. Тогда результат Васиной деятельности определяется множеством $I \subset\{1, \ldots, 12\}$ таким, что произошло переключение на ребрах с номерами $i \in I$. Иногда результат Васиных действий можно обозначить в виде $f=\sum_{i \in I} e_{i}$. Легко видеть, что результаты складываются по модулю 2. Действие называется честным, если его можно осуществить только вращая кубик. В следующие утверждения и рассуждения совершенно стандартны с точки зрения линейной алгебры и теории инвариантов. Пусть $и, v$ - честные действия, тогда $u+v$ тоже честное действие. Если $\sum_{i \in I} e_{i}$ - честное действие, $\sigma \in S_{12}$, то $\sum_{i \in I} e_{\sigma(i)}$ также является честным действием. Пространством называется группа, порожденная множеством действий. Множество честных действий называется честным пространством. Пусть $L$ - пространство. Назовем $L$ инвариантным, если для любого $\sigma \in S_{n}$ и $\sum_{i \in I} e_{i} \in L$ величина $\sum_{i \in I} e_{\sigma(i)}$ снова попадает в $L$. Мы установили, что пространство честных действий есть инвариантное пространство. Пусть $L$ есть инвариантное пространство. Тогда имеет место одна из следующих возможностей. 1) $L=0$ 2) $L$ совпадает со всем пространством. 3) $L$ порождено элементами $e_{i}+e_{j}$. 4) $L=\left\{0, \sum i=1^{12} e_{i}\right.$. Это утверждение проясняет природу реберного инварианта в кубике Рубика, и показывает, как к ним можно прийти. Вторая возможность для честного пространства исключается проверкой справедливости инварианта, а первая и четвертая - указанием способа осуществить разворот. Несложно сформулировать и доказать аналог утверждения F8 для вращения угловых кубиков. Кроме того, несложно показать, что реберные и угловые действия разделяются, даже если учитывать ориентацию. Этих идей достаточно чтобы найти полную систему инвариантов для кубика $n \times n \times n$ а также полную систему инвариантов для додекаэдра и икосаэдра Рубика произвольного размера. Чтобы провести исследование в многомерье, надо рассмотреть ключевой случай дамского куба. (Ladies Cube). Прежде всего, необходимо изучить группы вращений. Отсюда - вспомогательные утверждения. а) Пусть $K$ - куб в п-мерном пространстве, $G$ - группа, порожденная его поворотами. Тогда $G \simeq S_{n}$. b) Пусть $K$ - куб в п-мерном пространстве, $B$ - его вершина, $H$ группа, порожденная его поворотами, переводящими вершину $B$ в себя. Тогда $H \simeq A_{n}$. Изучим четырехмерный дамский куб. Известно, что $A_{4} / K_{4}=A_{3}$. Выберем две пары ребер, выходящих из вершины $B$ (это можно сделать 3 разными способами). На этих способах действует группа $A_{4} / K_{4}=A_{3}$. Таким образом находится нетривиальный инвариант для четырехмерного дамского куба. Неожиданным образом он оказывается по модулю 3 ! Чтобы получить этот инвариант достаточно покрасить пары двумерных граней, выходящих из вершин кубика плоскости которых пересекаются по точке ( $X O Y, Z O T)$, и рассуждать так же как и для трехмерного куба. Можно показать (это будет следовать из дальнейших рассмотрений) что других инвариантов нет. Как разворот части (некоторых, но не всех!!) угловых кубиков дамского $n>3$ мерного куба, при котором каждый кубик будет оставаться на своем месте? Это очень просто - надо действовать так, чтобы кубики, у которых первые три координаты совпадают, вращались как единое целое. Это рассуждение работает для любой размерности. Изучим возможные развороты кубика. Пусть $G$ - простая некоммутативная конечная группа. $R=G \oplus$ $G \oplus \cdots \oplus G$. Назовем подгруппу $H$ группы $G$ удобной, если ее проекиия покрывает каждое слагаемое. Назовем группу элементарной, если она имеет вид $\left(h, \varphi_{1}(h), \ldots, \varphi_{l}(h)\right), h \in G, \varphi_{i}$ - изоморфизми $G$ на себя. Тогда удобная группа есть прямая сумма элементарных. Из этого утверждения выводится, что у пятимерного дамского кубика нет инвариантов, т.е. количество его допустимых состояний равно $2^{5}$ !. $60^{3} 2$. Аналогично обстоит дело в пространстве бо́льшей размерности. Чтобы изучить ситуацию в четырехмерном пространстве, надо изучить поведение групп $S_{3}$ и $A_{4}$. Пусть $G=S_{3} . R=G \oplus G \oplus \cdots \oplus G$. Назовем подгруппу $Н$ группи $G$ удобной, если ее проекиия покрывает каждое слагаемое. Назовем группу элементарной, если она имеет вид $\left(h, \varphi_{1}(h), \ldots, \varphi_{l}(h)\right), h \in G, \varphi_{i}-$ изоморфизмы $G$ на себя. Тогда удобная группа есть прямая сумма элементарных, либо она содержит $A_{3} \oplus A_{3} \oplus \cdots \oplus A_{3}$. Пусть $G=A_{4} . R=G \oplus G \oplus \cdots \oplus G$. Назовем подгруппу $H$ группы $G$ удобной, если ее проекция покрывает каждое слагаемое. Назовем груп$n у$ элементарной, если она имеет вид $\left(h, \varphi_{1}(h), \ldots, \varphi_{l}(h)\right), h \in G, \varphi_{i}-$ изоморфизмы $G$ на себя. Тогда удобная группа есть прямал сумма элементарных, либо она содержит $K_{4} \oplus K_{4} \oplus \cdots \oplus K_{4}$. Для завершения исследования в кубе произвольного размера и произвольной размерности надо выбрать такую систему координат, чтобы координаты центра куба были нулями, а координаты центров кубиков по модулю не превосходили $n$. Каждый сорт определяется набором $M_{k}$ координат, по модулю равных $k$ для всех $k$. Группа разворотов для каждого сорта суть $\bigoplus_{k} S_{\left|M_{k}\right|}$ если $\left|M_{0}\right| \neq 0$ и $\bigoplus_{k} A_{\left|M_{k}\right|}$ если $\left|M_{0}\right|=0$. Приведенных выше соображений достаточно для нахождения полной системы инвариантов. С группами $S_{i}$ связаны только инварианты по модулю 2 , с группами $A_{3}$ и $A_{4}$ - только инварианты по модулю 3 , с группами $A_{n}$ при $n \geq 5$ - вообще никакие инварианты не связаны. Следующая задача завершает исследование. ## F1 a) Пусть координаты центра четырехмерного куба $3 \times \cdots \times 3$ суть $(0,0,0,0)$, а координаты центров кубиков $-0, \pm 1$. Покажите, что инварианты разворота, связанные с кубиками с координатами центров $(0, \pm 1, \pm 1, \pm 1),( \pm 1,0, \pm 1, \pm$ связаны только с четностью. б) Пусть координаты центра пятимерного куба $4 \times \cdots \times 4$ суть $(0,0,0,0,0)$, а координаты центров кубиков $-0, \pm 1, \pm 2$. Покажите, что инварианты разворота, связанные с кубиками с координаты центров которых суть: $(0, \pm 1, \pm 1, \pm 1, \pm 1),( \pm 1,0, \pm 1, \pm 1, \pm 1),( \pm 1, \pm 1,0, \pm 1, \pm 1),( \pm 1, \pm 1, \pm 1,0, \pm 1)$, $( \pm 1, \pm 1, \pm 1, \pm 1,0)$ связаны только с модулем 3 . F2 Какие несмешиваемые типы кубиков есть в $k$-мерном кубе размера $n$ ? F3 Приведите разные примеры инвариантов для $k$-мерного куба размера $n$. F4* Постройте полную систему инвариантов для $k$-мерного куба размера $n$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-041.jpg?height=368&width=323&top_left_y=87&top_left_x=244) ## Rubic's cube and Higman problem In this project we shall explore Rubic's cube and similar puzzles. Before this we should solve some preliminary problems. ## Preliminary problems. P1. There are 12 labeled cubes in a row. The left cube is labeled with 1 and the right one with 12 . If the crazy drummer beats his magic drumhead then two neighboring cubes transpose. After 333 beats every cube goes crazy if there are odd number of cubes with less number to the right of him. Can the number of the crazy cubes be equal to 6 ? P2. There are 42 cubes in the row labeled in the increased order. One can transpose every two neighboring cubes. Is it possible to transpose the first two cubes and leave the remaining cubes at their places by 2008 moves? P3. There are 20 color cubes in the circle. One can choose every three cubes and put the first one on the second one's place, second one to the third one's place and the third one to the first one's place. Is it possible to obtain the situation that all the cubes would be cyclically shifted in comparison with the initial position? P4. The facets of a cube are painted into different colors. Some of these cubes are used to form a rectangle $m \times n$. One can choose a row or a column and rotate all the cubes with regard to the common axis. Prove that it is possible to obtain the situation that all the cubes be rotated to the up with the same color. ## A. Rubic's cube. Let us call the whole large cube by cube and small cubes by bricks. Every 9 -bricks facet of the cube can be rotated clockwise or counter-clockwise. One can do several such rotates in series. This sequence of rotates is simply called combination. Denote the facets of the cube by $A, B, C$. We shall say that clockwise rotation of the facet is denoted by the same letter, for example, $A$. We also denote the counterclockwise rotation by $A^{-1}$. Further, we shall denote the sequence of rotations by the sequence of the letters. For example, $A B A^{-1} C$ means "clockwise rotation of the $A$ facet then clockwise rotation of the $B$ facet then counter-clockwise rotation of the $A$ facet then clockwise rotation of the $C$ facet". ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-042.jpg?height=203&width=186&top_left_y=144&top_left_x=181) The sequence $X Y X^{-1} Y^{-1}$ is called the commutator of the rotations $X$ and $Y$. There are three types of bricks inside the cube: central bricks are located in the centers of facets, corner bricks form the corners of the cube and middle ones are located in the centers of the cube's edges. In is clear that the central bricks doesn't move (with regard to each other). Also corner bricks will stay corner and middle ones will stay middle. Suppose that middle and corner bricks don't fasten with anything. Hence we can easily take they out, transpose and put they in. Let us transpose middle bricks with other middle ones. Similarly we do with the corner ones. (We shall not transpose the central bricks.) Moreover, we check that the external facets of the bricks stay external after any transformation. Any position after such transformations is called a state. A state of the cube is called right if every facet of this cube is one-color. We shall say that a brick has the right position if it has the same colors of the facets as for right state of the whole cube. States are called connected if there exists a sequence of rotations such that the first state converts to the second one. A state is said to be admissible if it is connected with the right state. - A1. Suppose that we apply a sequence of rotations to the initial cube. Prove that it is possible to apply this combination several times again and return to the initial state. A2. Is there exists a sequence of rotations that would arrange the cube starting from any state (by applying it several times)? A3. Find a combination of rotations that would cause to cyclically shift the bricks 1, 2, 3 and leave the remaining middle-bricks at their places (fig. 1). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-042.jpg?height=582&width=1118&top_left_y=2059&top_left_x=478) Figure 1. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-043.jpg?height=203&width=186&top_left_y=141&top_left_x=181) A4. Show that the combination $A^{-1} C^{-1} B^{-1} A^{-1} B A C$ cause to transpose the 1 and 2 bricks and leave the remaining middle-bricks at their places (fig. 2). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-043.jpg?height=585&width=506&top_left_y=570&top_left_x=478) Figure 2. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-043.jpg?height=583&width=505&top_left_y=571&top_left_x=1095) Figure 3. - A5. Find a combination that would cause to rotate the bricks 1 and 2 within their sockets and leave the remaining middle-bricks at their places and states (fig. 3)? A6. Prove that there is no combination that would cause to rotate the brick 1 within its socket and leave the remaining middle-bricks at their places and states (fig. 3). A7. Suppose that the state of the cube is admissible. Show how to place all the middle bricks to the right states. Suppose that the state is not necessarily admissible. Consider the states of the middle bricks and describe all possible connected states. A8. Find a non-trivial combination of rotations that would cause no effect to the cube while applying it exactly three times. A9. Find a combination of rotations that would cause to cyclically shift the bricks 1, 2, 3 (fig. 4), and leave the remaining corner-bricks at their places and middle-bricks at their places and states. A10. Consider an admissible state of the cube. Find a combination that would arrange all the corner bricks to their places and leave the middle-bricks at their places and states. Suppose that the state is not necessarily admissible. Consider the states of the corner bricks and describe all possible connected states. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-044.jpg?height=200&width=177&top_left_y=146&top_left_x=191) ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-044.jpg?height=572&width=506&top_left_y=385&top_left_x=475) Figure 4. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-044.jpg?height=572&width=505&top_left_y=385&top_left_x=1095) Figure 5. A11. Prove that there is no combination that would rotate exactly one corner brick and leave the remaining bricks at their places and states. - A12. Find a combination that would cause to rotate clockwise the bricks 1, 2, 3 (fig. 5) by 120 degrees, and leave the remaining bricks at their places and states. A13. Suppose that the state of the cube is admissible. How to arrange the cube? A14. Suppose that the all corner bricks are in the right places and middle ones are in the right states. Consider the states of the corner bricks. How to define if it possible to arrange the cube? A15. How many pairwise unconnected states of the cube would be? A16. Calculate the number of admissible states of the cube. ## Section B. In this section we shall consider some similar puzzles. Let us apply two combinations (of rotations) at the same state. We shall say that these two combinations are different if they make different results. B1. The chessboard is labeled with all integer numbers from 1 to 64 . One can choose a square $2 \times 2$ and rotate clockwise the numbers inside it. Prove that it is possible to achieve any possible numbers arrangement. B2. Consider the $2 \times 2 \times 2$ cube. Describe all the admissible states of this cube. How many such states are there? ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-045.jpg?height=480&width=759&top_left_y=371&top_left_x=657) Figure 6. B3. Consider the game "Hungarian rings" (fig. 6). There is a planar puzzle consisting of two or more interwoven ovals each of which has several labeled pieces, some of which may belong to more than one oval. A puzzle move consists of shifting an oval by one or more "increments", and hence all the pieces on it, along the oval's grooved track. The pieces are equally spaced apart (in spite of the typed depiction below) and whose pieces which lie on more than one oval can be moved along either oval. For simplicity, consider the puzzle consisting of only two ovals, each having 6 pieces. The pieces 1 and 3 can be moved along either oval. Note that each move corresponds to an unique permutation of the numbers in $\{1,2, \ldots, 10\}$. Describe all the admissible states. - B4. Consider the game 'Equator' (fig. 7). This puzzle is in the shape of the sphere but has 3 circular bands encircling a sphere, each having 12 square-shaped pieces and each band intersecting each other at a 90 degree angle. Each pair of circles intersects at two points, or "nodes" and at each such node there is a puzzle piece shared by the two circular bands. There are 6 nodes total. One can rotate any band such that their pieces would transfer to each other. The total number of movable pieces is therefore $3 \times 12-6=30$. Describe all the admissible states. B5. Consider the $4 \times 4 \times 4$ cube. Describe all the admissible states. B6. Consider the game "15". There are 15 numerated tiles placed in the $4 \times 4$ square. One square is empty. One can choose a neighboring (by side) tile of the empty square and move it on the empty place. Describe all the admissible states of the game. - B7. How many pairwise unconnected states of the cube $4 \times 4 \times 4$ would be? B8. Find the invariants system for $4 \times 4 \times 4$ cube. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-046.jpg?height=814&width=1168&top_left_y=140&top_left_x=183) Figure 7. B9. Find the invariants system for $n \times n \times n$ cube. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-047.jpg?height=197&width=186&top_left_y=144&top_left_x=181) ## Section C. - C1. Consider a regular tetrahedra. We can rotate it such that it maps to itself and some edges and vertexes may transpose. Find the number of the different rotations. - C2. Similar task for cube. Describe behavior of the main diagonals of cube. Consider a regular polyhedra. Similarly to Rubic's cube rotations we can apply a sequence of two motions. It is easy to see that this composition is a motion itself. This composition is called a product. The motion of polyhedra is called an identity if it preserves all its vertexes. Composition of every motion $F$ with identical motion $I$ does not change $F$, i.e. $F=I F=F I$. Identical motion of a polyhedra is an identity of all space. - C3. Let $M$ be an $n$-element set. Consider the permutations of $M$ elements. For example, if $M$ consists of 3 elements, then there are 6 such permutations: 1) exchange 1 and 2, 3 does not move; 2) exchange 2 and 3,1 does not move; 3) exchange 1 и 3,2 does not move; 4) sent 1 into 2, 2 into 3, 3 into 1 (cycle of length 3); 5) $1 \rightarrow 3,3 \rightarrow 2,2 \rightarrow 1$ (cycle of length 3 ); 6) don't move anything. Similarly for each $n$ all permutations can be listed. Suppose that we apply some permutation then apply another one. This combination of two permutations is called a product of them. Which permutation is identical? Prove that for any motion $A$ there exists a motion $A^{-1}$ such that $A A^{-1}=A^{-1} A=E$, there $E$ is the identical motion. Check the associative law: $(A B) C=A(B C)$. - C4. Let $A$ be a set of all motions of a cube. Find a correspondence between $A$ and the set $B$ of permutations of 4 elements such that product in $A$ corresponds to the product in $B$. Definition. A group is a set $G$ with operation of multiplication such that the following properties are satisfied: 1) $(A B) C=A(B C)$ (associative law); 2) there is an unit element $E$ such that $A E=E A=A$ for any $A$; 3) for any element $A$ there exists an inverse $A^{-1}$ such that $A A^{-1}=A^{-1} A=E$. Permutation group from problem C3 is denoted by $S_{n}$. - C5. Prove that the following sets with operations are groups: 1) the set of integers with respect to addition; 2) the set of all positive rational numbers with respect to multiplication; 3) Rubic's cube transformation with respect to composition. - C6. Which of the following objects are groups? 1) the rational numbers with respect to multiplication; 2) set of all words in the Latin alphabet (including empty word) with respect to concatenation (concatenation of $u$ and $v$ is $u v$ ). $3)$ set of all words in the alphabet $\{a, b, c\}$ (including empty word), if for any words $X, Y$ we can replace any of the words $X a b c Y, X b c a Y, X c a b Y$ with the word $X Y$ (in other words, we can remove $a b c, b c a, c a b$ from any word) and do inverse operation (add these words); 4) Three double transpositions (12)(34), (13)(24), (14)(23) and identity? Note: (123)(4567) means that the elements are cyclically shifted in every bracket: $1 \rightarrow 2 \rightarrow 3 \rightarrow 1$ and $4 \rightarrow 5 \rightarrow 6 \rightarrow 7 \rightarrow 4$. Definition. Let $G$ be a group, $H$ be a subset of $G$. Suppose that $H$ contains unit element of $G$, (we just call it unit) and also all products of any two elements in $H$ together with their inverses, then $H$ is called a subgroup. - C7. Let $H$ be a subgroup of $G$. Prove that $H$ is a group. - C8. Find all subgroups of $S_{3}$. - C9. Lagrange theorem. Prove that the number of subgroup elements divides the number of group elements. - C10. Find a $\frac{n!}{2}$-elements subgroup of $S_{n}$, for $n \geq 2$. We shall denote the group in the previous problem by $A_{n}$. - C11. Prove that any element of $A_{n}$ is a product of cycles of length 3 (they may intersect with each other). Definition. $a b a^{-1} b^{-1}$ is called a commutator of $a$ and $b$. Definition. Commutant of $G$ is the set of all products of commutators. - C12. Find the commutants of $S_{3}, A_{3}, A_{4}, S_{n}, A_{n}$. Let us fix $a \in G$. For any $g \in G$ consider corresponding element $a^{-1} g a$. This element is called a conjugate $g$ respect to $a$, or just conjugate. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-049.jpg?height=203&width=190&top_left_y=141&top_left_x=182) - C13. Let $H$ be a subgroup of $G$. Suppose that $a^{-1} H a$ is the set of all conjugates of the $H$ elements with respect to $a$. Prove that $a^{-1} H a$ is a subgroup as well. Subgroups $H$ and $a^{-1} H a$ are called conjugated. Definition. A subgroup is called normal if it is equal to all its conjugates. - C14. Prove that commutant, unit element $E$ and whole group $G$ are normal subgroups of $G$. The whole group $G$ and unit element $E$ are the trivial normal subgroups of $G$. All other subgroups are called non-trivial. If $G$ has the trivial normal subgroups only, then $G$ is simple group. - C15. Find all $n$ such that group $S_{n}$ is simple. - C16. Prove that group $A_{n}$ is simple for $n \geq 5$. - C17. Find all $n$ such that $S_{n}$ or $A_{n}$ are equal to the groups of motions of cube or tetrahedra. - C18. Prove that the group of icosahedra motions are equal to $A_{5}$. - C19. Find a 8-element group such that there exist two elements $a, b$ and $a b \neq b a$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-050.jpg?height=212&width=192&top_left_y=134&top_left_x=178) 1010 ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-051.jpg?height=195&width=186&top_left_y=148&top_left_x=181) ## Section D. Groups Definition. Groups $G$ and $H$ are called isomorphic if there exists a one-toone mapping (isomorphism) $\varphi$ such that unit element maps to unit element and composition of any two elements in $G$ maps to composition of the corresponding two elements in $H: \varphi\left(g_{1} g_{2}\right)=\varphi\left(g_{1}\right) \times \varphi\left(g_{2}\right)$. D1. Prove that the following groups are isomorphic: 1) group of cube's motions and $S_{4}$; 2) group of real numbers with respect to addition and group of horizontal motions; 3) group of integer numbers with respect to addition and group of $2^{k}$ for integer $k$ with respect to composition. 4) group of dodecahedra's (icosahedra's) motions and $A_{5}$; Isomorphic groups are really equivalent groups. Suppose that $G$ is a group, and $M$ is a set. We shall say that $G$ acts on $M$, if for any $m \in M$ and $g \in G$ there exists an corresponding element $m^{\prime}=g(m)$ such that $\left(g_{1} g_{2}\right) m=g_{1}\left(g_{2} m\right)$. ## Examples. 1. The group of all motions acts on 3-dimensional space. 2. The groups $S_{n}$ и $A_{n}$ act on the set $\{1, \ldots, n\}$. 3. A group acts on itself by left multiplications: every $h \in G$ defines a mapping $\varphi_{h}(g)=h g$. 4. A group acts on itself by conjugates: every $h \in G$ defines a mapping $\varphi_{h}(g)=$ $h^{-1} g h$. D2. Prove that these are really the acts. D3. Suppose that $n$ is the number of elements in the group $G$ (we shall denote it by $n=|G|)$. Prove that there exists a subgroup in $S_{n}$ such that it is isomorphic to $G$. D4. A group $G$ is colored into several colors such that the color of the product depends on colors of the factors only. The unity is red. Prove that the set of red elements is a normal subgroup. Suppose that $H$ is a subgroup in $G$. The coloring of $G$ is called left if: ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-052.jpg?height=200&width=186&top_left_y=146&top_left_x=181) 1) for any $g \in G$ and $h \in H$ the elements $g$ and $h g$ are colored in the same color; 2) if $g_{1}$ and $g_{2}$ are colored in the same color then $g_{1}=h g_{2}$ for some $h \in H$. ## D5. a) Prove that the left coloring coincides to the right one if and only if $H$ is a normal group. b) Prove that if $H$ is a normal group then the color of the product depends on colors of the factors only and the color of the inverse element depends on color of the original one. Suppose that $H$ is a subgroup in $G$. For any $g \in G$ consider the set of all $g h_{i}$ for $h_{i} \in H$. This set is called left $H$-class with respect to $H$ and denoted by $g H$ (note that this is essentially just the same as one-colored elements). Every element $G$ belongs to one left $H$-class only. Two elements belong to the same class if they can be presented in the form $g h_{i}$ for some $h_{i} \in H$. Every $g_{1}=g_{1} e$ belongs to $g_{1} H$ (because of $e \in H$ ). The element $g_{2}$ belongs to the same class if there exists $h$ such that $g_{2} h=g_{1}\left(\right.$ or $\left.g_{1} g_{2}^{-1}=h\right)$. The product of two classes $g_{1} H$ and $g_{2} H$ is the class $g_{1} g_{2} H$. We can choose other representatives of these two classes, namely, $g_{1}^{\prime} \in g_{1} H$ and $g_{2}^{\prime} \in g_{2} H$. Then the product would be $g_{1}^{\prime} g_{2}^{\prime} H$. - D6. Prove that a product is well-defined (because of $g_{1} g_{2} H=g_{1}^{\prime} g_{2}^{\prime} H$ ). Similarly we can define an inverse element on the left $H$-classes set: the class $g^{-1} H$ is inverse for $g H$. The set of left $H$-classes (or the set of colors from D5) can be considered as a group. This group is called a factorgroup with respect to the normal subgroup $H$.The left $H$-classes are the elements of this group. D7. Find factorgroups of $G$ with respect to the $H$ : 1) $G=S_{n}$ and $H=A_{n}$ 2) $G=A_{4}$ and $H$ is a group of three double transpositions (12)(34), (13)(24), (14)(23) and identity? 3) $G$ is a group of real numbers with respect to addition and $H$ is the subgroup of integers. D8. a) Prove that the group of rotations combinations of the cube $2 \times 2 \times 2$ is the factorgroup of the $3 \times 3 \times 3$-cube group. b) Prove that the groups of rotations combinations of the cubes $3 \times 3 \times 3$ and $4 \times 4 \times 4$ is the factorgroups of the $5 \times 5 \times 5$-cube group. Definition. An orbit of an element $m \in M$ is the set $\left\{g_{i} m\right\}$ for some $g_{i} \in G$. - D9. Describe the orbits of elements for the following acts: 1) the group of rotations combinations acts on the corner brick; 2) the horizontal translation group acts on the point of the plane; 3) group $S_{n}$ of permutations acts by conjugates on itself. The element is 3-cycle. D10. Prove that any two orbits are disjoint or coincide. Definition. Стабилизатором элемента $m \in M$ называется множество элементов $g \in G$ таких, что $g(m)=m$. Suppose that $G$ acts on set $M$. Some elements of the group don't shift the elements of the set: Definition. A set of elements $g \in G$ is called a stabilizer of en element $m \in M$ if $g(m)=m$. - D11. Prove that stabilizer Stab $m$ is a subgroup. Prove that $\left|O_{m}\right| \cdot|\mathrm{Stab} m|=$ $|G|$. Definition. Действие называется хорошим, если стабилизатор любого элемента состоит из одной единицы. Definition. An act is called good if stabilizer of any element is unity only. Consider two actions of $G$ on the sets $M_{1}$ и $M_{2}$. Suppose that these sets one-toone correspond to each other: every element $M_{1}$ corresponds to the unique element $M_{2}$ and vice-versa: $\Psi\left(M_{1}\right)=M_{2}$. Then these two actions are called conjugated if the group acts on these sets by the similar way: element $g_{\varphi}\left(M_{1}\right)$ corresponds to $g_{\psi}\left(M_{2}\right)$. Note that we can choose $M_{1}=M_{2}=M$. ## D12. a) Suppose that $\varphi_{1}$ and $\varphi_{2}$ are two good actions on $M$. Is it always true that they are conjugated? b) The same question if the numbers of orbits are equal. Suppose that $G$ is a group and $g \in G$. Consider an element $g^{-1} h g$ for each $g$. It is easy to see that the element $g^{-1} h_{1} g g^{-1} h_{2} g=g^{-1} h_{1} h_{2} g$ correspond to the product $h_{1} h_{2}$. Hence, $h \rightarrow g^{-1} h g$ is one-to-one mapping such that a product maps to a product. Thus, we have obtain an isomorphism of $G$ to itself. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-054.jpg?height=214&width=192&top_left_y=133&top_left_x=178) - D13. Suppose that $G$ is a group, $H_{1}$ and $H_{2}$ are isomorphic subgroups with isomorphism $\varphi: H_{1} \rightarrow H_{2}$. Is it always possible to continue $\varphi$ to the isomorphism of $G$ to itself. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-055.jpg?height=197&width=186&top_left_y=147&top_left_x=181) D14. Suppose that $G$ is a group and $H_{1}$ and $H_{2}$ are isomorphic subgroups of $G$. Let the numbers of elements in the $H_{1}$-classes and $H_{2}$ be equal (or both are countable). Prove that there exists a group $G^{\prime}$ such that $G$ is a subgroup of $G^{\prime}$, $t \in G^{\prime}$ and the following condition holds for any $h \in H_{1}$ : $$ t h t^{-1}=\varphi(h) \in H_{2} $$ D15. Suppose that the numbers of elements in the $H_{1}$-classes and $H_{2}$ not be equal. Prove the same fact. Definition. If every element of group $G$ can be presented by the product of some elements of $\left\{x_{i}\right\}$ then $G$ is called generated by elements $x_{i}$. We denote this by $G=\left\langle x_{i}\right\rangle$ Definition. A group is called $n$-generated if there are $n$ elements in $\left\{x_{i}\right\}$. Definition. A group is called $n$-free-generated, if it is isomorphic to the group of words in the alphabet $\left\{g_{1}, g_{1}^{-1}, g_{2}, g_{2}^{-1}, \ldots, g_{n}, g_{n}^{-1}\right\}$. D16. Suppose that $G$ contains the following pairs of isomorphic subgroups: $\varphi_{i}: H_{i} \rightarrow H_{i}^{\prime}$ for $i=1 \ldots n$. Prove that there exists a group $G^{\prime}$, such that $G$ is a subgroup of $G^{\prime}$ and the following condition hold in $G^{\prime}$ : 1) $t_{i} h_{i} t_{i}^{-1}=\varphi_{i}\left(h_{i}\right)$ 2) the group generated by $\left\langle t_{i}\right\rangle$ is free. D17. Suppose that group $G$ is free-generated by the elements $t_{i} i \in \mathbb{N}$. Prove that there is a isomorphism between $G$ and $H=\left\langle t_{2}, t_{3}, \ldots\right\rangle$. D18. Prove that any countable group can be presented by the subgroup of a 3 -generated group. ## Section E. Groups and arrangement of a high-dimensional cubes We shall say that a cube is almost solved if all the bricks returned to their places but maybe some of them aren't rotated well. We use some following problems to almost solve a cube. - E1. a) Suppose that two 4-cycle contain one common vertex. Prove that they generate a group $S_{7}$. b) Suppose that two 4-cycle contain two common consecutive vertexes. Prove that they generate a group $S_{6}$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-056.jpg?height=197&width=174&top_left_y=147&top_left_x=193) E2. a) Prove that the group $A_{12}$ is generated by 11-cycles. b) Let $x \in S_{8}$. Prove that $x^{8!}$ is the unity. Suppose that $s$ is a 11-cycle. Prove that there exists a 11-cycle $t$ such that $s=t^{8}$. c) Suppose that middle bricks and corner brick in the $3 \times 3 \times 3$ cube are transposed with permutations of equal parity. Prove that it is possible to almost solve the cube. d) Prove that it is possible to almost solve the cube $4 \times 4 \times 4$ starting from any state. Definition. We shall say that a group $G$ is a sum of groups $G_{1}$ and $G_{2}$ if it is consists of pairs $\left(g_{1}, g_{2}\right)$ such that $g_{1} \in G_{1}, g_{2} \in G_{2}$. Besides, the following rule describes the multiplication in $G$ : $\left(g_{1}, g_{2}\right) \times\left(h_{1}, h_{2}\right)=\left(g_{1} h_{1}, g_{2} h_{2}\right)$. We denote a sum by $G=G_{1} \oplus G_{2}$. E3. Suppose that $G_{1}$ is simple finite group with generators $a_{1}, \ldots, a_{k}$ and $G_{2}$ is simple finite group with generators $b_{1}, \ldots, b_{k}$. Let $G=G_{1} \oplus G_{2}$. Suppose that $H$ is a $G$-subgroup generated by the elements $z_{i}=\left(a_{i}, b_{i}\right), i=1, \ldots, k$. Prove that $H=G$ or there is an isomorphism $\varphi: G_{1} \rightarrow G_{2}$ such that $b_{i}=\varphi\left(a_{i}\right)$. Prove that the group of rotations combinations of the $3 \times 3 \times 3$ cube contains a subgroup $A_{8} \oplus A_{12}$. - E4. Prove that any state in the cube $2 n \times 2 n \times 2 n$ can be almost solved. - E5. Prove that $2 \times 2 \times \ldots \times 2$ cube can be almost solved for any dimension. - E6. Prove that $2 n \times 2 n \times \ldots \times 2 n$ cube can be almost solved for any dimension. Using the fact that a $2 \times 2 \times 2 \times 2$ cube can be almost solved we can do the useful note. Suppose that we can rotate $k$ corner bricks. Then we can similarly rotate any other $k$ corner bricks. This note helps us to describe a full cube solvability. - E7. Prove that $2 \times 2 \times 2 \times 2$ cube has 3 classes of connected states in 4 -dimensional space. Note: Use the fact that factorgroup of $A_{4}$ with respect to the group of double transpositions (example 2 in $\mathrm{D} 7$ ) is a 3-element group. E8. Prove that all states of the $2 \times 2 \times \ldots \times 2$ cube are admissible in the dimension 5 or higher. E9. Find the number of classes of connected states for $3 \times 3 \times \ldots \times 3$ cube in the dimension 4 or higher. E10. Find the number of classes of connected states for $n \times n \times \ldots \times n$ cube in the dimension 4 or higher. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-057.jpg?height=534&width=1667&top_left_y=183&top_left_x=176) НАУЧНО-ПОПУЛЯРНЫЙ ФИЗИКО-МАТЕМАТИЧЕСКИЙ ЖУРНАЛ АКАДЕМИИ НАУК сссР И АКАДЕМИИ ПЕДАГОГИपЕСКИХ НАУК сссР ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-057.jpg?height=1972&width=1715&top_left_y=919&top_left_x=166) при которой каждая грань куба снова будет одного цвета. Задача эта совсем не проста. Не вооруженному теорией человеку, даже способному, редко удается сразу собрать более одной грани. Число расстановок кубиков, которые можно получить (подсчитано, что их $N=43252003274489856000)$, деB. Залгаллер, C. Залгаллер лает ее недоступной для перебора даже на ЭВМ. Заметим, впрочем, что не любая расстановка может ## Венгерский шарнирный кубик быть получена вращениями плиток куба: если разрешить разборку куба на составляющие его 26 кубиков, то можно составить $12 \cdot N=$ $=529024039393878272000$ разных расстановок (см. задачу 8 из Добавления).В қастоящей заметке мы предла- В Венгерской Народной Республике распространена занятная математическая головоломка, созданная в 1975 г. венгерским архитектором, профессором Эрнё Рубиком. Ее внешний вид показан на четвертой странице обложки: это пластмассовый куб, разбитый на 27 конгруэнтных кубиков. Внутренний кубик удален, a 26 наружных кубиков с помощью специальных выступов сцеплены так, что любая плитка из 9 кубиков, прилегающих к одной грани куба, может быть повернута в любую сторону на $90^{\circ}$. (Начало двух таких поворотов изображено на рисунке 1.) После поворота на $90^{\circ}$ вся система сохраняет прежнюю свободу вращений: снова любую плитку в любую сторону можно повернуть в ее плоскости на $90^{\circ}$. Об устройстве шарнирного скрепления этих кубиков можно написать отдельную статью - сейчас же мы будем обсуждать другой вопрос. Первоначально каждая из граней большого куба была окрашена в свой цвет (красный, оранжевый, желтый, зеленый, синий, белый). После ряда случайно выбранных вращений окраска граней куба становится пестрой: на грани присутствуют клетки разных цветов. Головоломка состоит в том, чтобы, получив в руки такой пестрый куб, добиться с помощью вращений правильной расстановки кубиков, то есть такой расстановки, ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-059.jpg?height=437&width=717&top_left_y=213&top_left_x=290) Рнс. 2. чает, что положение всех центральных кубиков в пространстве остается неизменным. Это означает также, что из девяти плоскостей куба мы будем поворачивать только шесть. Если на рисунке некоторая центральная клетка отмечена буквой $A$, то поворот (не «средней») плитки, содержащей эту клетку, на $90^{\circ}$ по часовой стрелке мы будем обозначать через $A$, а поворот этой же плитки на $90^{\circ}$ против часовой стрелки будем обозначать через $A^{-1}$. Очевидно, $A^{-1} A^{-1} A^{-1}=A$ и $A^{2}=A^{-2}$, где $A^{2}=A \quad A$ и $A^{-2}=A^{-1} \quad A^{-1}$. Кубики, содержащие среднюю часть ребра большого куба, будем называть средними. Средний кубик всегда остается средним и смотрит наружу двумя клетками определенного цвета. Для каждой пары цветов (кроме тех пар цветов, которыми первоначально были раскрашены противоположные грани куба) имеется единственный средний кубик с клетками этих цветов. Угловыми назовем кубики, занимающие в составе куба угловые места. Каждый угловой кубик всегда остается угловым и смотрит наружу тремя клетками, окрашенными в разные цвета. Сочетание этих трех цветов у каждого из угловых кубиков свое. ## Основные этапы Заманчивый, на первый взгляд, путь: постепенно увеличивая пятно одноцветных клеток, получить одноцветную грань, а потом взяться за другую - видимо, приводит лишь к непреодолимым трудностям. Не справившись с головоломкой, ее нередко портят - от злости или из любопытства к устройству шарнира. Мы на- ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-059.jpg?height=394&width=345&top_left_y=223&top_left_x=1027) Рис. 3. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-059.jpg?height=386&width=326&top_left_y=224&top_left_x=1416) Рис. 4. деемся, что предлагаемый ниже способ решения сделает головоломку доступной, но и он требует определенных усилий при реализации. Предлагаемые действия разобьем на четыре больших этапа, Мы уже договорились, что в ходе решения головоломки центральные кубики не меняют своего положения в пространстве. Что касается средних кубиков, то каждый из них должен в процессе решения занять вполне определенное место: оказаться на ребре между двумя «своими» гранями куба, то есть теми гранями, чьи центральные клетки такого же цвета, как две клетки данного среднего кубика. Кроме того, он должен быть правильно повернут: его цветные клетки должны прилегать к центральным клеткам того же цвета. Совершенно аналогично обстоит дело с угловыми кубиками: у каждого из них есть свое место (на стыке трех граней с центральными клетками его цветов) и единственный правильный разворот. В соответствии с этим порядок наших действий будет следующим: Этап 1: поставить на нужные места все средние кубики. Этап 2: правильно повернуть на своих местах все средние кубики. Этап 3: поставить на нужные места все угловые кубики. Этап 4: правильно повернуть все угловые кубики. Қаждый из этапов мы будем выполнять только после того, как предыдущий этап полностью закончен. При этом очередной этап будем выполнять так, чтобы после его завершения оказались не нарушенными достижения предшествующих этапов. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-060.jpg?height=391&width=346&top_left_y=227&top_left_x=318) Рис. 5. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-060.jpg?height=391&width=325&top_left_y=227&top_left_x=817) Рис. 6. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-060.jpg?height=388&width=326&top_left_y=226&top_left_x=1296) Рис. 7. После выполнения второго этапа на каждой из граней куба образуется крест из пяти клеток одного цвета. После выполнения четвертого этапа задача окажется решенной - каждая грань куба станет одноцветной. ## Невозможные положения Достаточность предлагаемых ниже «комбинаций» для выполнения перечисленных этапов опирается на три свойства рассматриваемого шарнирного куба. В рамках этой статьи мы будем считать эти свойства экспериментальными фактами. Их можно, однако, доказать, решив задачи, придуманные В. Дубровским (см. Приложение). Если кубики выведены из правильного положения только допустимыми вращениями (а не разборкой и новой сборкой всего устройства или перекраской граней), то не может возникнуть положение, при котором: I. все средние кубики стоят на своих местах и только один из них повернут неправильно; II. все средние кубики и стоят. и повернуты правильно, а все угловые кубики, кроме двух, стоят (в любых положениях) на своих местах; III. все средние кубики и стоят. и повернуты правильно, а все угловые кубики стоят на своих местах и только один из них повернут неправильно. ## Предварительные комоинации Чтобы привыкнуть к тому, как записываются повороты, мы сначала рассмотрим несколько важных для дальнейшего стандартных комбинаций. Первая комбинация $A^{-1} B A$ (в отличие от школьного учебника, это будет означать, что сначала совершается поворот $A^{-1}$, затем $B$, затем $A$ ) называется сопряжением элемента $B$ с помощью элемента $A$. На рисунке 2 показано, как эта комбинация позволила собрать белую грань целиком, поставив белую клетку 1 на место желтой 2. Вторая комбинация $A^{-1} B^{-1} A B$ называется коммутатором элементов A, B. Проверьте, что в положени॥ куба, изображенном на рисунке 2 , серые клетки остаются на месте, а средние кубики 1, 2, 3 циклически переставляются, то есть переходят по схеме $1 \rightarrow 3 \rightarrow 2 \rightarrow 1$. Проследите еще, как переставляют средние кубики комбинации $A B A^{-1} B^{-1}$, $A B^{-1} A^{-1} B, \quad A^{2} B^{2} A^{2} B^{2}, \quad A^{2} B A^{2} B^{-1}$, $A^{2} B^{-1} A^{2} B, \quad A B^{2} A^{-1} B^{2}, \quad A^{-1} B^{2} A B^{2}-$ коммутаторы других поворотов (каких?). Этап 1: средние кубики - на место Қомбинация из семи поворотов $$ K_{1}=A^{-1} C^{-1} B \quad ' A^{-1} B A C $$ взаимно переставляет местами средние кубики 1,2 (рис. 3) и сохраняет местоположение остальных средних кубиков (проверьте!"). Комтіинашия $K_{1}$ фактически позволяет менять местами любые два средних кубика. Если они не прилегали к одной грани куба или не были на ней соседними, то всегда можно вспомогательными поворотами (запомнив эти повороты по цвету центра поворачиваемой плитки) привести желаемые кубики в положение I, 2 (рис. 3), переставить их комбинацией $K_{1}$, а затем, в обратной очередности и обратных направлениях повторить сделанные вспомогательные повороты.[^0] Очевидно, попарные перестановки средних кубиков позволяют осуществить этап 1 : поставить на свои места все средние кубики. Этап 2: повернем средние кубнкн Қомбинация из двенадцати поворотов $$ K_{2}=\left(A B^{-1} A^{-1} B\right)\left(B C^{-1} B^{-1} C\right)\left(C A^{-1} C^{-1} A\right) $$ одновременно поворачивает в своих гнездах кубики 1, 2 (рис. 4 ) и не меняет ни местоположений, ни поворотов остальных средних кубиков (проверьте!). Комбинация $\kappa_{2}$ позволяет повернуть на своих местах любые два средних кубика. Действительно, если они не прилегали к одной грани куба или не были на ней соседними, то всегда можно вспомогательными поворотами (запомнив эти повороты по цвету центра поворачиваемой плитки) привести желаемые два кубика в положение 1, 2 (рис. 4), повернуть их комбинацией $K_{2}$, а затем, в обратной очередности и в обратных направления повторить вспомогательные повороты. Попарные совместные повороты средних кубиков позволяют осуществить этап 2, поскольку (ввиду свойства 1) не может случиться, чтобы требовал поворота только один средний кубик. ## Этап 3: угловые кубики - на место Комбинация из двенадщати поворотов $$ K_{3}=\left(A B A^{\mathrm{t}} B^{\mathrm{t}}\right)^{3} $$ осуществляет одновременно перестановки кубиков 1, 2 и кубиков 3, 4 (рис. 5), сохраняет все достижения этапов 1,2 " не изменяет местоположения угловых кубиков, отличных от 1, 2, 3, 4 (проверьте!) ротов Комбинация из двадцати четырех пово- $$ K_{4}=\left(A C A^{-1} C^{-1}\right)^{3}\left(B^{-1} A^{-1} B A\right)^{3} $$ осуществляет перестановку кубиков I, 2, 3 (рис. 6) по схеме $1 \rightarrow 2 \rightarrow 3 \rightarrow 1$, сохраняет все достижения этапов 1,2 и не изменяет местоположения угловых кубиков, отличных от $1,2,3$. Поясним, как,- пользуясь комбинациями $К_{3}$ и $K_{4}$, осуществлять этап 3. Допустим, что после завершения этапа 2 часть угловых кубиков не стоит на свонх местах. Ввиду свойства II таких кубиков будет не менее, чем три. Выберем один из них и отметим его номером 1. Кубик, занимающий то место, куда должен встать кубик 1. отметим номером 2. Если кубик 2 сам должен перейти на место кубика I, то номером 3 отметим любой, отличный от первых двух и стоящий не на своем месте угловой кубик, а номером 4 - кубик, стоящий там, куда должен перейти кубик 3. Затем, с помошью вспомогательных поворотов (запоминая их) поставим кубики 1, 2 и 3,4 в положение пар 1,2 , и 3,4 , изображенных на рисунке 5. С помощью комбинации $K_{3}$ осушествим перестановку $\xrightarrow[\rightleftarrows]{\rightleftarrows} 2,3 \underset{\rightleftarrows}{\rightleftarrows}$, после чего в обратной очередности и обратных на- правлениях повторим вспомогательные повороты. В итоге, кроме ранее стоявших на своих местах угловых кубиков, заведомо попадут на свои места кубики I, 2, 3. Если же кубик 2 не должен был перейти на место кубика 1 , то отметим номером 3 кубик, стоящий там, куда должен перейти кубик 2. Ёли кубик 3 не должен перейти на место кубика 1 , то отметим номером 4 кубик, который стоит там, куда должен перей. ти кубик 3. После этого, как и выше, то есть с помощью вспомогательных поворотов и комбинации $K_{3}$, осуществляем перестановку $\rightrightarrows \rightleftarrows 2$. $3 \rightleftarrows 4$, которая ставит на свон места кубики 1 и 3. Остается случай, когда требуется перестановка кубиков $1 \rightarrow 2 \rightarrow 3 \rightarrow 1$. В этом случае. с помощью вспомогательных поворотов (запоминая их) ставим кубики 1, 2, 3 в каком-то порядке на места 1, 2. 3 (рис. 6 ). Одна из двух комбннаций $K_{4}, K_{4}$ ', которыхе осуществляют перестановки мест $1 \rightarrow 2 \rightarrow 3 \rightarrow 1$. $1 \rightarrow 3 \rightarrow 2 \rightarrow 1$, будет оуушествлять перестановку ку 6 иков $1 \rightarrow 2 \rightarrow 3 \rightarrow 1$ (находящихся сейчас на местах I, 2, 3). Выполним именно эту комбинацию. Ilocле этого в обратной очередности и обратных направлениях повторим вспомогательные повороты. В итоге кубики $1,2,3$ станут на место. Повторяя указанный процесс, мы выполним этап 3 - поставим на свои места все угловые кубики. ## Этал 4: повернем угловые кубики Нам потребуются следующие две комбинации: $K_{5}=\left\{\left.\left(A^{1} C A C^{\prime}\right)\left(C^{\prime} B C B^{1}\right)\left(B^{-1} A B A{ }^{1}\right)\right|^{2}\right.$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-061.jpg?height=52&width=711&top_left_y=1473&top_left_x=1049) Комбинация $K_{\text {и }}$ поворачивает одновременно каждый из кубиков 1, 2, 3 (рис. 7) вокру! юси, идущей от его центра к центру куба. на $120^{\circ}$ по часовой стрелке. Комбинация $K_{5}$ ' поворачивает те же кубики против часовой стрелки. Местоположение и поворот любого из остальных кубиков при этом не меняются. Знанне комбинаций $K_{5}$ и $K_{5}{ }^{\prime}$ позволяет совместно повернуть в желаемом направлении (но в одну и ту же сторону) лю 6 ы е три угловых кубика. Достаточно собрать их с помощью вспомогательных поворотов в одну грань, совместно повернуть с помощью комбинации $K_{5}$ или $K_{3}{ }^{\prime}$ (в зависимости от нужного направления поворота), а затем вернуть на своим места поворотами, обратными вспомогательным. Поясним теперь, как, пользуясь комбинациями $K_{5}$ и $K_{5}{ }^{1}$, осушествить этап 4 . Если после завершения этапа 3 имеются неправильно повернутые угловые кубики, то, ввиду свойства III. таких кубиков не менее двух. Допустим, что их больше двух. Возьмем любые три из них. Хотя бы два из этих трех кубиков требуют поворота в одну и ту же сторону. Повернув в эту сторону все три кубика, мы уменьшим число неверно повернутых кубиков. Так мы придем к положению, когда неверно повернутых кубиков либо нет, либо ровно два. Если неверно повернутых кубиков только два, они не могут оба требовать поворота в одну и ту же сторону. Иначе мы бы совместно повернули их и еще один кубик появилось бы положение с одним неверно повернутым кубиком, что невозможно в силу III. Итак, два неверно повернутых кубика требуют поворота в разные стороны. Повернем в нужную сторону однн их этих кубиков и два кубнка, ранее стоявших правильно. А затем - в другую сторону - эти два новых кубнка и второй из ранее повернутых неправильно. В результате все кубики займут правильное положение! ## Приложение*) Обозначнм через $S_{0}$ правильное состояние куба. Занумеруем числами $i=1,2, \ldots, 8$ его вершины (угловые кубики) и числами $j=1$. 2,..., 12 его ребра (средние кубики). На ребрах выберем (и запомним) любую ориентацию (и нарисуем соответствующие стрелки на средних кубиках) так, чтобы параллельные ребра были сонаправлены. Предположим, что в некотором законком состоянии**) $S$ куба $j$-й средний кубик попал на ребро $i^{\prime}$; сравним ориентацию ребра $j^{\prime}$ (которую мы запомнили) со стрелкой, нарисованной на $j$-м кубике. Обозначим через $n_{j}(S)$ величину, равную 0 , если указанные ориентации совпадают, и равную I в противном случае. Если сумма $$ n_{1}(S)+n_{2}(S)+\ldots+n_{12}(S) $$ четна, мы полагаем $n(S)=0$, иначе $n(S)=1$; можно сказать, что $n(S)$ - это षчетность суммарного поворота средних кубиков». Докажите, что 1. Величияа $n$ является инвариантом, то есть одинакова при всех законмых состояниях куба: $n(S)=n\left(S_{0}\right)=0 . \quad$ (У ка зание. Достаточно показать, что $n(S)$ не меняется при любом повороте од ной плитки). 2. В законном состоянии куба не может быть повернут ровно один средний кубик (свойство I, с. 19). (Указан и е. В таком состоянии $n(S)=1$ ). Вспомним, что положение централь ных клеток куба в пространстве не меняется при поворотах: для определенности предположим, что нижний центральный кубик - зеленый, верхний - синий: эти два цвета будем считать выделенкыми. Возьмем $i$-й угловой кубик нашего куба, находящегося в состоянии $S$. Ровно одна клетка углового кубика выделена (почему?). Если эта клетка горизонтальна, положим (по определению) $N_{i}(S)=0$; если эта клетка становится горнзонтальной при повороте кубика на $120^{\circ}$ (по часовой стрелке) вокруг диагонали большого куба, положим $N_{i}(S)=1$; если же горизонтальность получается поворотом на[^1] $240^{\circ}$, то $N_{i}(S)=-1$. Если сумма $$ N_{1}(S)+N_{2}(S)+\ldots+N_{8}(S) $$ при делении на 3 дает в остатке $0,1,2$, мы полагаем величину $N(S)$ равной $0,1,-1$ соответственно. Можно сказать, что величина $N$ - это єнаправление суммарного поворота угловых кубиков». Докажите, что 3. Величина $N$ является инвариантом. то есть одинакова при всех законных состояниях куба: $N(S)=N\left(S_{0}\right)=0$. 4. В законмом состоянии у куба не может быть повернут неправильно ровно одик угловой кубик (свойство III, с. 19). Предположим теперь, что в состоянии $S$ средние кубики с номерами $1,2, \ldots, 12$ занимают места с номерами $j_{1}, j_{2}, \ldots, j_{12}$ соответственно, а угловые кубики $1,2, \ldots, 8$ - места $i_{1}, i_{2}, \ldots, i_{8}$. Наборы $I(S)=\left(i_{1}, i_{2}, \ldots, i_{8}\right)$ и $J(S)=\left(j_{1}, j_{2}, \ldots, j_{12}\right)$ - это просто номера кубиков, записанные в другом порядке: математики такне наборы называют перестамовками. Перестановку называют четной, если в ней имеется четное число беспорядков. то есть четное число пар цифр, стоящих не в порядке возрастания, и нечетной в противном случае. Например, перестановка (12453687) нечетна, так как она содержит 3 беспорядка: $(4,3),(5,3),(8,7)$. Обозначим через $\varepsilon(S)$ число, равное 0 , если $I(S)$ и $J(S)$ имеют одинаковую четность, и равное 1 в противном случае. Можно сказать, что $\varepsilon$ - это «четность расстано́вки всех кубиков». Докажите, что 5. Величина \& является инвариантом, притом $\varepsilon(S)=\varepsilon\left(S_{0}\right)=0$ для любого законного состояния S. (Указание. Покажите, что при любом повороте плитки меняется четность I и четность $J$ ). 6. Ровно два угловых кубика не могут поменяться местами (свойство II, с. I9), 7*. Состояние S куба законно тогда и только тогда, когда $$ n(S)=N(S)=\varepsilon(S)=0 $$ 8*. Все состояния (в том числе полученкые разборкой и сборкой куба разбиваются на 12 классов; при этом два состояния $S, S^{\prime}$ переводятся друг в друга поворотами пмит тогда и только тогда, когда $n(S)=n\left(S^{\prime}\right)$, $N(S)=N\left(S^{\prime}\right), \varepsilon(S)=\varepsilon\left(S^{\prime}\right)$. Три параболы
Ф.Нилов, А.Заславский Мы будем рассматривать следующую конфигурацию: треугольник и три параболы, каждая из которых проходит через две его вершины и касается в этих вершинах соответствующих сторон. Основная задача - изучение свойств этой конфигурации. # 1 Необходимые теоретические сведения Определение 1. Параболой называется геометрическое место точек, равноудаленных от данной точки и данной прямой. Эти точка и прямая назыаются соответственно фокусом и директрисой параболы. Определение 2. Пусть дан треугольник $A B C$ и произвольная точка $P$. Прямые, симметричные прямым $A P, B P, C P$ относительно биссектрис соответствующих углов, пересекаются в одной точке, которая называется изогонально сопряженной $P$. Точка $L$, изогонально сопряженная центру тяжести $M$, называется точкой Лемуана. Определение 3. Треугольник, образованный проекциями точки $P$ на стороны треугольника $A B C$, называется педальным треугольником $P$ относительно $A B C$. Окружность, описанная около педального треугольника, называется педальной окружностью $P$. Определение 4. Два треугольника называются перспективными, если прямые, соединяющие их соответствующие вершины, пересекаются в одной точке, которая называется центром перспективы. Теорема Дезарга. Два треугольника перспективны тогда и только тогда, когда точки пересечения их соответствующих сторон лежат на одной прямой. Эта прямая называется осъю перспективы. Определение 5. Два треугольника называются ортологичными, если перпендикуляры, опущенные из вершин одного на соответствующие стороны другого, пересекаются в одной точке, которая называется центром ортологичности. Свойство ортологичности является симметричным, так что ортологичные треугольники имеют два, в общем случае различных, центра ортологичности. В любом треугольнике точки $M, O$ и ортоцентр $H$ лежат на одной прямой, которая называется прямой Эйлера. Середины сторон и основания высот лежат на одной окружности, которая называется окружностью Эйлера. Определение 6. Точки $Q_{1}, Q_{2}$, такие что $\angle Q_{1} A B=\angle Q_{1} B C=\angle Q_{1} C A$ и $\angle Q_{2} B A=$ $\angle Q_{2} C B=\angle Q_{2} A C$, называются точками Брокара треугольника $A B C$. Угол $\phi=\angle Q_{1} A B=$ $\angle Q_{2} B A$ называется углом Брокара. Для любого неравностороннего треугольника существуют две точки, педальные треугольники которых правильные. Эти точки инверсны друг другу относительно описанной окружности треугольника. Та, которая лежит внутри окружности, называется первой точкой Аполлония, а другая - второй точкой Аполлония. Точки, изогонально сопряженные точкам Аполлония называются первой и второй точками Торричелли. ## 2 Вводные задачи 1. Докажите оптическое свойство параболы: прямая, касающаяся параболы с фокусом $F$ в точке $X$, образует равные углы с прямой $F X$ и осью симметрии параболы. 2. Докажите, что точка, симметричная фокусу параболы относительно касательной к ней, лежит на директрисе параболы, причем совпадает с проекцией на директрису точки касания. 3. Пусть касательные к параболе в точках $X$ и $Y$ пересекаются в точке $P$. Докажите, что $P$ - центр окружности, описанной около треугольника $F X^{\prime} Y^{\prime}$, где $F$ - фокус параболы, а $X^{\prime}, Y^{\prime}$ - проекции $X$ и $Y$ на директрису. 4. Докажите, что в обозначениях предыдущей задачи медиана треугольника $P X Y$ параллельна оси параболы. 5. Докажите, что в обозначениях задачи 3 ось параболы и прямая $P F$ образуют равные углы с биссектрисой угла $P$. 6. Докажите, что педальные окружности изогонально сопряженных точек совпадают. Пусть дан треугольник $A B C$. Обозначим параболу, проходящую через $A$ и $B$ и касающуюся в этих точках прямых $A C$ и $B C$, через $\Pi_{c}$, а ее фокус - через $F_{c}$. Аналогично определим параболы $\Pi_{a}, \Pi_{b}$ и их фокусы $F_{a}, F_{b}$. 7. Докажите, что $\Pi_{a}, \Pi_{b}$ имеют, помимо $C$, ровно одну общую точку $C^{\prime}$. 8. Определим точки $A^{\prime}, B^{\prime}$ аналогично точке $C^{\prime}$. Докажите, что треугольники $A B C$ и $A^{\prime} B^{\prime} C^{\prime}$ перспективны. 9. Докажите, что треугольники $A B C$ и $F_{a} F_{b} F_{c}$ перспективны. 10. Докажите, что треугольник, образованный директрисами парабол, перспективен треугольнику $A B C$. ## Три параболы ## 3 Основные задачи 11. Докажите, что $F_{c} C$ - биссектриса угла $A F_{c} B$. 12. Докажите, что точки $A, B, F_{c}$ и центр $O$ описанной около $A B C$ окружности лежат на одной окружности. 13. Докажите, что парабола $\Pi_{c}$ касается средней линии, параллельной $A B$. 14. Докажите, что точки $F_{a}, F_{b}, F_{c}, O, L$ лежат на одной окружности. 15. Докажите, что директриса параболы $\Pi_{c}$ проходит через точку пересечения медианы, проведенной из вершины $C$, с окружностью Эйлера. 16. Докажите, что центры тяжести $A B C$ и треугольника, образованного директрисами парабол, совпадают. 17. Докажите, что $A B C$ и треугольник, образованный директрисами парабол, ортологичны и их центры ортологичности совпадают. 18. Докажите, что прямая Эйлера треугольника $A B C$ проходит через точку Лемуана треугольника, образованного директрисами. 19. Докажите, что прямая Эйлера треугольника, образованного директрисами, проходит через точку $L$. 20. Пусть прямые $A F_{a}, B F_{b}, C F_{c}$ пересекают противоположные стороны треугольника $A B C$ в точках $P_{a}, P_{b}, P_{c}$. Докажите, что треугольник $P_{a} P_{b} P_{c}$ перспективен треугольнику, образованному директрисами парабол. 21. Докажите, что попарные центры перспективы $A B C, P_{a} P_{b} P_{c}$ и треугольника, образованного директрисами, лежат на одной прямой. 22. Докажите, что треугольник, образованный директрисами перспективен ортотреугольнику $A B C$. 23. Докажите, что углы Брокара треугольника $A B C$ и треугольника, образованного директрисами, равны. ## Три параболы ## 4 Дополнительные задачи 24. Пусть $T$ - любая точка Торричелли треугольника $A B C, T_{d}$ - соответствующая точка Торричелли треугольника, образованного директрисами. Докажите, что прямые, соединяющие $T$ и $T_{d}$ с соответствующими вершинами треугольников, параллельны. 25. Докажите, что прямая $T T_{d}$ и прямая, соединяющая соответствующие точки Аполлония, проходит через центр тяжести $A B C$. 26. Вычислить $\frac{S^{\prime}}{S}$, где $S^{\prime}$ - площадь криволинейного "параболического"треугольника $A^{\prime} B^{\prime} C^{\prime}$ а $S-$ площадь $A B C$. 27. Докажите, что $N_{1} N_{2}$ делит пополам отрезок, соединяющий центры тяжести $A B C$ и $P_{a} P_{b} P_{c}$, где $N_{1}$ и $N_{2}$ - центры перспективы треугольника, образованного директрисами, и треугольников $A B C$ и $P_{a} P_{b} P_{c}$ соответственно. 28. Пусть $F_{a}^{\prime}, F_{b}^{\prime}, F_{c}^{\prime}$ - точки, изогонально сопряжённые точкам $F_{a}, F_{b}, F_{c}, H$ - ортоцентр, $M$ - точка пересечения медиан. Докажите, что a) $F_{a}^{\prime}, F_{b}^{\prime}, F_{c}^{\prime}$ лежат на окружности, построенной на отрезке $H M$ как на диаметре. b) точки $F_{a}$ и $F_{a}^{\prime}$ равноудалены от центра окружности Эйлера треугольника $A B C$. Назовём проекиией точки $C$ на параболу $\Pi_{c}$ точку параболы, ближайшую к $C$. Обозначим через $C^{*}$ проекцию точки $C$ на параболу $\Pi_{c}$. 29. Докажите, что прямая $C C^{*}$ перпендикулярна параболе $\Pi_{c}$. 30. Докажите, что $\angle A C^{*} C=\angle B C^{*} C$. 31. Обозначим через $C_{1}$ и $C_{2}$ точки пересечения касательной, проведённой к параболе $\Pi_{c}$ в точке $C^{*}$, с $C A$ и $C B$. Обозначим через $C^{* *}$ точку пересечения медианы треугольника $A B C$, проведённой из вершины $C$ с $C_{1} C_{2}$. Докажите, что $C_{1} C^{*}=C^{* *} C_{2}$. Определим точки $A^{*}$ и $B^{*}$ аналогично $C^{*}$. 32. (Гипотеза) Прямые $A A^{*}, B B^{*}$ и $C C^{*}$ пересекаются в одной точке. ## Три параболы ## 5 Похожий сюжет Рассмотрим множество прямых, делящих периметр треугольника $A B C$ пополам. 33. Докажите, что эти прямые огибают три параболы, каждая из которых касается двух прямых, содержащих стороны треугольника $A B C$, в тех же точках, что и соответствующая вневписанная окружность. Обозначим через $\Pi_{A}$ параболу, которая касается прямых $A B$ и $A C$. Аналогично определим $\Pi_{B}$ и $\Pi_{C}$. 34. Докажите, что центр описанной окружности треугольника $A B C$ является центром ортологичности треугольника $A B C$ и треугольника, образованного директрисами парабол $\Pi_{A}, \Pi_{B}$ и $\Pi_{C}$. 35. Докажите, что ортоцентр треугольника $A B C$ лежит на прямой Эйлера треугольника, образованного директрисами парабол. 36. Докажите, что центр вписанной окружности треугольника $A B C$ и центр описанной окружности треугольника, образованного директрисами парабол, симметричны относительно центра окружности Эйлера треугольника $A B C$. 37. Пусть $T$ - любая точка Торричелли треугольника $A B C, T_{d}$ - соответствующая точка Торричелли треугольника, образованного директрисами. Докажите, что прямые, соединяющие $T$ и $T_{d}$ с соответствующими вершинами треугольников, параллельны. ## 6 Похожий сюжет №2 Рассмотрим точки $A_{c}$ и $A_{b}$ на сторонах $A B$ и $A C$ такие, что $C A=C A_{c}$ и $B A=B A_{b}$. Рассмотрим параболу $\Pi_{a}^{\prime}$, касающуюся сторон $A C$ и $A B$ в точках $A_{b}$ и $A_{c}$. Аналогично определим параболы $\Pi_{b}^{\prime}$ и $\Pi_{c}^{\prime}$. 38. Докажите, что фокусы парабол $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ и $\Pi_{c}^{\prime}$ совпадают с точками $F_{a}^{\prime}, F_{b}^{\prime}$ и $F_{c}^{\prime}$. 39. Докажите, что треугольник, образованный директрисами парабол $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ и $\Pi_{c}^{\prime}$, перспективен треугольнику $A B C$. 40. Докажите, что треугольник, образованный директрисами парабол $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ и $\Pi_{c}^{\prime}$, ортологичен треугольнику $A B C$, причём их центры ортологичности совпадают. 41. Докажите, что треугольник, образованный директрисами парабол $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ и $\Pi_{c}^{\prime}$, перспективен ортотреугольнику треугольника $A B C$. 42. Обозначим через $O^{\prime}$ и $M^{\prime}$ центр описанной окружности и центр тяжести треугольника, образованного директрисами. Докажите, что прямые $O^{\prime} M$ и $M^{\prime} O$ пересекаются в точке Лемуана треугольника $A B C$. # Три параболы
Решения Ф.Нилов, А.Заславский ## 2 Вводные задачи 1. Обозначим проекцию точки $X$ на директрису параболы через $X^{\prime}$. Тогда $F X=X X^{\prime}$. Утверждение задачи равносильно тому, что прямая $l$, содержащая биссектрису угла $X^{\prime} X F$, является касательной к параболе. Предположим противное. Тогда $l$ не является касательной и пересекает параболу в некоторой точке $Y$, отличной от $X$. Заметим, что треугольники $F X Y$ и $X^{\prime} X Y$ равны. Значит, $F Y=Y X^{\prime}$. Обозначим через $Y^{\prime}$ проекцию точки $Y$ на директрису параболы. Тогда $Y Y^{\prime}=F Y=Y X^{\prime}$. Значит, $Y Y^{\prime}=Y X^{\prime}$. Но этого быть не может, поскольку отрезки $Y Y^{\prime}$ и $Y X^{\prime}$ являются гипотенузой и катетом прямоугольного треугольника $Y X^{\prime} Y^{\prime}$. 2. Обозначим через $l$ касательную, через $X$ - точку касания, через $F$ - фокус, через $X^{\prime}$ - проекцию $\mathrm{X}$ на директрису. По определению параболы $F X=X X^{\prime}$. Из задачи 1 следует, что $l$ содержит биссектрису равнобедренного треугольника $F X X^{\prime}$. Поэтому $l$ является серединным перпендикуляром к отрезку $F X^{\prime}$. Значит, точки $X^{\prime}$ и $F$ симметричны относительно $l$. 3. По задаче 2 прямые $P X$ и $P Y$ являются серединными перпендикулярами к отрезкам $F X^{\prime}$ и $F Y^{\prime}$. Поэтому их точка пересечения $P$ является центром описанной окружности треугольника $F X^{\prime} Y^{\prime}$. 4. Рассмотрим среднюю линию трапеции $X X^{\prime} Y^{\prime} Y$. Она перпендикулярна директрисе и поэтому параллельна оси параболы. Кроме того, она является медианой треугольника $P X Y$, поскольку она проходит через середину $M$ отрезка $X Y$ и через точку $P$ (по задаче 3). Поэтому медиана $P M$ параллельна оси параболы. 5. Рассмотрим прямую $l$, параллельную оси параболы и проходящую через $P$. Утверждение задачи равносильно тому что угол $\phi$ между прямыми $l$ и $P X$ равен углу $F P Y$. Обозначим через $X^{\prime \prime}$ и $Y^{\prime \prime}$ проекции точки $F$ на прямые $P X$ и $P Y$. По задаче 2 точки $X^{\prime \prime}$ и $Y^{\prime \prime}$ являются серединами отрезков $F X^{\prime}$ и $F Y^{\prime}$. Поэтому прямые $X^{\prime} Y^{\prime}$ и $X^{\prime \prime} Y^{\prime \prime}$ параллельны. Значит, прямые $l$ и $X^{\prime \prime} Y^{\prime \prime}$ перпендикулярны. Поэтому $\angle \phi=90^{\circ}-\angle P X^{\prime \prime} Y^{\prime \prime}=90^{\circ}-\angle P F Y^{\prime \prime}=\angle F P Y$, что и требовалось. 6. Обозначим через $P$ и $Q$ изогонально сопряженные точки. Обозначим через $P_{c}$ и $Q_{c}$, $P_{a}$ и $Q_{a}, P_{b}$ и $Q_{b}$ точки, симметричные $P$ и $Q$ относительно сторон $A B, B C$ и $A C$. Ясно, что $B P_{c}=B P=B P_{a}$ и что $\angle P_{c} B P_{a}=2 \angle B$. Заметим, что $\angle Q B C=\angle P B A=$ $\angle A B P_{c}$. Поэтому $\angle P_{c} B Q=\angle B=1 / 2 \angle P_{c} B P_{a}=\angle P_{a} B Q$. Поэтому $Q P_{c}=Q P_{a}$. Аналогично $Q P_{a}=Q P_{b}$ и $P Q_{c}=P Q_{a}=P Q_{b}$. Понятно, что окружность $\omega$ центром в середине отрезка $P Q$ и радиусом, равным $1 / 2 Q P_{c}$ проходит через проекцию точки $P$ на сторону $A B$. Аналогично $\omega$ проходит и через остальные вершины педальных треугольников точек $P$ и $Q$. 7. Поскольку параболы $\Pi_{a}$ и П в вписаны в углы $A$ и $B$, то их точки пересечения должны принадлежать пересечению углов $A$ и $B$. Значит, точки пересечения парабол лежат внутри или на границе треугольника $A B C$. Понятно, что $C$ - не единственная точка пересечения парабол: иначе параболы касались бы в точке $C$ и касательные к параболам в точке $C$ совпадали, но это не так. Значит, существует вторая точка пересечения парабол $C^{\prime}$. Ясно, что точек пересечения, отличных от $C$ и лежащих на границе треугольника $A B C$, нет. Поэтому все точки пересечения, отличные от $C$, лежат внутри треугольника $A B C$. Предположим, что у $\Pi_{a}$ и $\Pi_{b}$ есть третья точка пересечения $C^{\prime \prime}$. Легко убедиться в том, что любые четыре точки параболы являются вершинами выпуклого четырёхугольника. Поскольку точки $A, C, C^{\prime}$ и $C^{\prime \prime}$ лежат на $\Pi_{b}$, то $A C C^{\prime} C^{\prime \prime}$ - выпуклый четырёхугольник. По аналогичным соображениям $B C C^{\prime} C^{\prime \prime}$ - выпуклый, чего быть, очевидно, не может. Поэтому $\Pi_{a}$ и П Пм $_{b}$ имеют ровно две общие точки. 8. Известно, что образ параболы при произвольном аффинном преобразовании является параболой. Рассмотрим аффинное преобразование, которое переводит треугольник $A B C$ в правильный. Понятно, что в правильном треугольнике рассматриваемые параболы пересекаются на его медианах. Поэтому прямые $A A^{\prime}, B B^{\prime}$ и $C C^{\prime}$ являются медианами треугольника $A B C$ и пересекаются в одной точке. 9. Обозначим через $M$ точку пересечения медиан треугольника $A B C$. Из задач 4 и 5 следует, что прямые $A M$ и $A F_{a}$ симметричны относительно биссектрисы угла $A$. Поэтому прямые $A F_{a}, B F_{b}$ и $C F_{c}$ пересекаются в точке Лемуана $L$ треугольника $A B C$. 10. Обозначим через $d_{a}$ директрису параболы $\Pi_{a}$. Обозначим точку пересечения $d_{a}$ и $B C$ через $A^{\prime}$. Обозначим проекции точек $B$ и $C$ на $d_{a}$ через $B_{1}$ и $C_{1}$. Ясно, что $B B_{1}=B F_{a}$ и $C C_{1}=C F_{a}$. Из подобия треугольников $A^{\prime} B B_{1}$ и $A^{\prime} C C_{1}$ получаем, что $\frac{A^{\prime} B}{A^{\prime} C}=\frac{B B_{1}}{C C_{1}}=\frac{B F_{a}}{C F_{a}}$. Из задач 1 и 5 следует, что $\angle F_{a} B A=\angle F_{a} A C$ и $\angle F_{a} C A=\angle F_{a} A B$. По теореме синусов для треугольника $F_{a} B A$ получаем, что $B F_{a}=\frac{\sin \left(\angle F_{a} A B\right) \cdot A F_{a}}{\sin \left(\angle A B F_{a}\right)}$. Аналогично $C F_{a}=\frac{\sin \left(\angle F_{a} A C\right) \cdot A F_{a}}{\sin \left(\angle A C F_{a}\right)}$. Поэтому $\frac{B F_{a}}{C F_{a}}=\frac{\sin ^{2}\left(\angle F_{a} A B\right)}{\sin ^{2}\left(\angle F_{a} A C\right)}$. Обозначим через $d_{b}$ и $d_{c}$ директрисы парабол $\Pi_{b}$ и $\Pi_{c}$. Обозначим через $B^{\prime}$ точку пересечения $d_{b}$ с $A C$, а через $C^{\prime}$ - точку пересечения $d_{c}$ с $A B$. Тогда $$ \frac{A^{\prime} B \cdot B^{\prime} C \cdot C^{\prime} A}{A^{\prime} C \cdot B^{\prime} A \cdot C^{\prime} B}=\frac{B F_{a} \cdot C F_{b} \cdot A F_{c}}{C F_{a} \cdot A F_{b} \cdot B F_{c}}=\frac{\sin ^{2}\left(\angle F_{a} A B\right) \cdot \sin ^{2}\left(\angle F_{b} B C\right) \cdot \sin ^{2}\left(\angle F_{c} C B\right)}{\sin ^{2}\left(\angle F_{a} A C\right) \cdot \sin ^{2}\left(\angle F_{b} B A\right) \cdot \sin ^{2}\left(\angle F_{c} C A\right)}=1 $$ Последнее равенство в этой цепочке следует из тригонометрической теоремы Чевы, записанной для чевиан $A F_{a}, B F_{b}$ и $C F_{c}$. Следовательно, по теореме Менелая точки $A^{\prime}, B^{\prime}$ и $C^{\prime}$ лежат на одной прямой. Поэтому из теоремы Дезарга следует перспективность рассматриваемых в задаче треугольников. # Три параболы Решения Ф.Нилов, А.Заславский ## 3 Основные задачи 11. Обозначим середину отрезка $A B$ через $C_{0}$. Тогда $\angle C_{0} C B=\angle F_{c} C A$. Из задач 1,4 и 5 следует, что $\angle C_{0} C B=\angle F_{c} B C$ и $\angle C_{0} C A=\angle F_{c} A C$. Поэтому $\angle A F_{c} C=180^{\circ}-$ $\angle F_{c} A C-\angle F_{c} C A=180^{\circ}-\angle C=180^{\circ}-\angle F_{c} C B-\angle F_{c} B C=\angle B F_{c} C$. 12. Из решения задачи 11 следует, что $\angle A F_{c} C=\angle B F_{c} C=180^{\circ}-\angle C$. Поэтому $\angle A F_{c} B=$ $360^{\circ}-2\left(180^{\circ}-\angle C\right)=2 \angle C=\angle A O B$. Значит, точки $A, B, F_{c}$ и $O$ лежат на одной окружности $\omega$. 13. Обозначим через $A_{0}$ и $B_{0}$ середины сторон $B C$ и $A C$. Из задачи 5 легко выводится следующая полезная лемма. Лемма. Описанная окружность треугольника, описанного около некоторой параболь проходит через фокус этой параболь. Поэтому достаточно доказать, что описанная окружность треугольника $A_{0} B_{0} C$ проходит через $F_{c}$. Обозначим через $C^{\prime}$ точку пересечения $C F_{c}$ с описанной окружностью треугольника $A B C$. Из задач 1 и 5 следует, что $\angle F_{c} B C=\angle F_{c} C A$ и $\angle F_{c} A C=\angle F_{c} C B$. Поэтому $\angle C^{\prime} B A=\angle F_{c} C A=\angle F_{c} B C$ и $\angle B A C^{\prime}=\angle F_{c} C B$. Значит, треугольники $F_{c} C B$ и $C^{\prime} A B$ подобны и $\frac{C F_{c}}{C^{\prime} A}=\frac{B F}{B C^{\prime}}$. По задаче $11 \angle B F_{c} C^{\prime}=\angle A F_{c} C^{\prime}$. Кроме того, $\angle C^{\prime} B F_{c}=\angle A B C=\angle A C^{\prime} F_{c}$. Поэтому треугольники $F_{c} B C^{\prime}$ и $F_{c} C^{\prime} A$ подобны. Значит, $\frac{F_{c} C^{\prime}}{C^{\prime} A}=\frac{B F}{B C^{\prime}}=\frac{C F_{c}}{C^{\prime} A}$. Следовательно, $F_{c} C=F_{c} C^{\prime}$. При гомотетии с центром в точке $C$ и коэффициентом $\frac{1}{2}$ точки $B$ и $A$ переходят в $A_{0}$ и $B_{0}$, и $C^{\prime}$ переходит в $F_{c}$. Откуда и следует утверждение. 14. По задаче 11 точки $A, B, F_{c}$ и $O$ лежат на одной окружности $\omega$. Рассмотрим точку $O^{\prime}$ пересечения прямой $C F_{c}$ с окружностью $\omega$. По задаче $10 \angle A F_{c} O^{\prime}=\angle B F_{c} O^{\prime}$. Поэтому $O^{\prime}$ - середина дуги $A B$ окружности $\omega$, т.е. отрезок $O O^{\prime}$ является диаметром $\omega$. Значит, $\angle L F_{c} O=\angle O F_{c} O^{\prime}=90^{\circ}$. Аналогично $\angle L F_{a} O=\angle L F_{b} O=90^{\circ}$. Следовательно, точки $F_{a}, F_{b}$ и $F_{c}$ лежат на окружности с диаметром $O L$. 15. Обозначим через $A_{0}$ и $B_{0}$ середины сторон $B C$ и $A C$. Рассмотрим точку $F_{c}^{\prime}$, симметричную точке $F_{c}$ относительно $A_{0} B_{0}$. По задаче $13 A_{0} B_{0}$ касается параболы $\Pi_{c}$. Поэтому по задаче 2 точка $F^{\prime}$ лежит на директрисе $\Pi_{c}$. Из решения задачи 13 следует, что четырёхугольник $C A_{0} F_{c} B_{0}$ вписанный. Поэтому $\angle A_{0} F_{c}^{\prime} B_{0}=\angle A_{0} F_{c} B_{0}=$ $180^{\circ}-\angle C$. Значит, точка $F_{c}^{\prime}$ лежит на окружности Эйлера треугольника $A B C$. Обозначим через $M^{\prime}$ середину отрезка $A_{0} B_{0}$. Из решения задачи 13 следует, что прямая $A_{1} B_{1}$ делит угол $C M^{\prime} F_{c}$ пополам. Значит, точка $F_{c}^{\prime}$ лежит на медиане $C M^{\prime}$. Поэтому точка пересечения медианы и окружности Эйлера лежит на директрисе параболы $\Pi_{c}$. 16. Пусть $A^{\prime \prime}, B^{\prime \prime}, C^{\prime \prime}$ - точки пересечения $d_{b}$ и $d_{c}, d_{a}$ и $d_{c}, d_{a}$ и $d_{b}$. Обозначим через $A^{\prime}$, $B^{\prime}$ и $C^{\prime}$ точки пересечения медиан треугольника $A B C$ с его окружностью Эйлера. По задаче 15 эти точки лежат на сторонах $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$. Обозначим через $C_{0}$ середину $A B$. Пусть $G$ - центр тяжести треугольника $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$. Заметим, что $$ \frac{\sin \left(\angle G C^{\prime \prime} B^{\prime \prime}\right)}{\sin \left(\angle G C^{\prime \prime} A^{\prime \prime}\right)}=\frac{\sin \left(\angle B^{\prime \prime}\right)}{\sin \left(\angle A^{\prime \prime}\right)}=\frac{\sin \left(\angle C_{0} M A\right)}{\sin \left(\angle C_{0} M B\right)}=\frac{M B}{M A}=\frac{M A^{\prime}}{M B^{\prime}}=\frac{\sin \left(\angle M C^{\prime \prime} B^{\prime \prime}\right)}{\sin \left(\angle M C^{\prime \prime} A^{\prime \prime}\right)} $$ Здесь первое и третье равенства следуют из того, что $C^{\prime \prime} G$ и $M C_{1}$ - медианы треугольников $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$ и $A M B$. Поэтому $\angle G C^{\prime \prime} B^{\prime \prime}=\angle M C^{\prime \prime} B^{\prime \prime}$. Значит, точка $G$ лежит на $C^{\prime \prime} M$. Аналогично $G$ лежит на $A^{\prime \prime} M$ и $B^{\prime \prime}$. Следовательно, точки $G$ и $M$ совпадают. 17. Из задачи 4 следует, что прямая $A M$ перпендикулярна $d_{a}$. Аналогично $B M$ перпендикулярна $d_{b}$ и $C M$ перпендикулярна $d_{c}$. Поэтому рассматриваемые треугольники ортологичны и точка $M$ является центром их ортологичности. Обозначим через $A^{\prime}$, $B^{\prime}$ и $C^{\prime}$ точки пересечения директрис $d_{b}$ и $d_{c}, d_{a}$ и $d_{c}, d_{a}$ и $d_{b}$. То, что медианы треугольника $A B C$ перпендикулярны соответствующим сторонам треугольника $A^{\prime} B^{\prime} C^{\prime}$, можно записать так: (1) $\left(\overrightarrow{C^{\prime} A^{\prime}}+\overrightarrow{A^{\prime} B^{\prime}}\right) \cdot(\overrightarrow{A B}+\overrightarrow{A C})=0$. (2) $\overrightarrow{A^{\prime} B^{\prime}} \cdot(2 \overrightarrow{C A}+\overrightarrow{A B})=0$. (3) $\overrightarrow{C^{\prime} A^{\prime}} \cdot(2 \overrightarrow{A B}+\overrightarrow{C A})=0$. Складывая (1) и (3) и вычитая из результата (2), получаем, что $\overrightarrow{C A} \cdot \overrightarrow{A^{\prime} B^{\prime}}=\overrightarrow{C^{\prime} A^{\prime}} \cdot \overrightarrow{A B}$. Из этого и равенства (1) получаем, что $\left(\overrightarrow{C^{\prime} A^{\prime}}-\overrightarrow{A^{\prime} B^{\prime}}\right) \cdot(\overrightarrow{A B}-\overrightarrow{A C})=0$, что равносильно тому, что прямые $A^{\prime} G$ и $B C$ перпендикулярны, где $G$ - центр тяжести $A^{\prime} B^{\prime} C^{\prime}$. По задаче 16 точки $G$ и $M$ совпадают. Значит, прямые $A^{\prime} M$ и $B C$ перпендикулярны. Аналогично перпендикулярны прямые $B^{\prime} M$ и $A C, C^{\prime} M$ и $A B$. Поэтому центры ортологичности треугольников $A B C$ и $A^{\prime} B^{\prime} C^{\prime}$ совпадают с точкой $M$. 18. Поскольку директриса $d_{a}$ перпендикулярна медиане $A A_{0}$, то из задачи 15 следует, что она проходит через точку $A^{\prime \prime}$ пересечения высоты $A A^{\prime}$ с окружностью Эйлера. Аналогично точки пересечения $B^{\prime \prime}$ и $C^{\prime \prime}$ окружности Эйлера с высотами $B B^{\prime}$ и $C C^{\prime}$ лежат на $d_{b}$ и $d_{c}$. Заметим, что окружность Эйлера треугольника $A B C$ является педальной окружностью точки $M$ относительно треугольника, образованного директрисами. По задаче 16 точка $M$ является точкой пересечения медиан треугольника, образованного директрисами. Поэтому прямые $A^{\prime \prime} L^{\prime}, B^{\prime \prime} L^{\prime}$ и $C^{\prime \prime} L^{\prime}$ параллельны соответствующим медианам треугольника $A B C$, где $L^{\prime}$ - точка Лемуана треугольника, образованного директрисами. Как известно, треугольники $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$ и $A B C$ гомотетичны с центром гомотетии $H$ и коэффициентом $\frac{1}{2}$. Следовательно, $L^{\prime}$ является серединой отрезка $H M$, поскольку $L^{\prime}$ является точкой пересечения медиан треугольника $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$. Откуда и следует утверждение. 19. Обозначим через $T$ треугольник, образованный директрисами. Рассмотрим параболы $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ и $\Pi_{c}^{\prime}$, касающиеся сторон треугольника $T$ в его вершинах. Обозначим через $T^{\prime}$ треугольник, образованный директрисами этих парабол. Заметим, что соответствующие стороны треугольников $T^{\prime}$ и $A B C$ перпендикулярны соответствующим медианам треугольника $T$. Поэтому они параллельны. Поэтому треугольники $T^{\prime}$ и $A B C$ гомотетичны, причём их центр гомотетии совпадает с $M$, поскольку по задаче 16 центры тяжести $T^{\prime}$ и $A B C$ совпадают. Обозначим точку Лемуана треугольника $T^{\prime}$ через $L^{\prime}$. Из гомотетичности $T^{\prime}$ и $A B C$ следует, что точки $M, L$ и $L^{\prime}$ лежат на одной прямой, где $L$ - точка Лемуана треугольника $A B C$. По задаче 18 точка $L^{\prime}$ лежит на прямой Эйлера треугольника $T$. Значит, прямая, проходящая через точки $L, M$ и $L^{\prime}$ совпадает с прямой Эйлера треугольника $T$. Откуда и следует утверждение задачи. 20. Обозначим через $C^{\prime}$ точку пересечения прямых $A B$ и $P_{a} P_{b}$. Заметим, что четвёрка точек $C^{\prime}, P_{c}, A$ и $B$ является гармонической, т.е. $\frac{C^{\prime} A}{C^{\prime} B}=\frac{A P_{c}}{P_{c} B}$. Из задачи 11 следует, что $\frac{A P_{c}}{P_{c} B}=\frac{A F_{c}}{F_{c} B}$. Обозначим через $C^{\prime \prime}$ точку пересечения $d_{c}$ и $A B$. Из решения задачи 10 Следует, что $\frac{C^{\prime \prime} A}{C^{\prime \prime} B}=\frac{A F_{c}}{F_{c} B}$. Поэтому $\frac{C^{\prime} A}{C^{\prime} B}=\frac{A F_{c}}{F_{c} B}=\frac{C^{\prime \prime} A}{C^{\prime \prime} B}$. Значит, точки $C^{\prime}$ и $C^{\prime \prime}$ совпадают. Из решения задачи 10 следует, что точки $A^{\prime \prime}, B^{\prime \prime}$ и $C^{\prime \prime}$ лежат на одной прямой. Поэтому и точки $A^{\prime}, B^{\prime}$ и $C^{\prime}$ лежат на одной прямой и из теоремы Дезарга следует перспективность треугольников $A B C$ и $P_{a} P_{b} P_{c}$. 21. Из решения задачи 20 следует, что прямые перспективы треугольников $A B C, P_{a} P_{b} P_{c}$ и треугольника, образованного директрисами, совпадают. Осталось лишь воспользоваться следующей леммой. Лемма. Центры перспективы трёх взаимно перспективных треугольников, оси перспективы которых совпадают, лежат на одной прямой. Доказательство леммы. Рассмотрим проективное преобразование, которое переводит общую ось перспективы треугольников в бесконечно удалённую прямую. При таком преобразовании рассматриваемые треугольники перейдут в гомотетичные треугольники. По теореме о трёх центрах гомотетии центры перспективы этих треугольников лежат на одной прямой. Откуда и следует лемма. 22. Обозначим через $A_{1}, B_{1}$ и $C_{1}$ середины сторон треугольника $A B C$, а через $A_{2}, B_{2}$ и $C_{2}$ - основания соответствующих высот. Пусть $C^{\prime \prime \prime}$ - точка пересечения медианы $C C_{1}$ и окружности Эйлера. По задаче $15 C^{\prime \prime \prime}$ принадлежит $d_{c}$. Обозначим через $C^{\prime}$ точку пересечения $d_{c}$ и $A^{\prime} B^{\prime}$, а через $C^{\prime \prime}$ - точку пересечения $C C_{1}$ и $A_{2} B_{2}$. Заметим, что $B_{2} C_{1}=A C_{1}=B C_{1}=A_{2} C_{1}$. Поэтому $\angle B_{2} C^{\prime \prime \prime} C_{1}=\angle A_{2} C^{\prime \prime \prime} C_{1}$, поскольку $A_{2} C^{\prime \prime \prime} B_{2} C_{1}$ - вписанный. По задаче $5 d_{c}$ перпендикулярна $C^{\prime \prime \prime} C^{\prime \prime}$. Поэтому $C^{\prime \prime \prime} C^{\prime}$ является биссектрисой внешнего угла $C^{\prime \prime \prime}$. Значит, $\frac{C^{\prime} A_{2}}{C^{\prime} B_{2}}=\frac{C^{\prime \prime} A_{2}}{C^{\prime \prime} B_{2}}$. Аналогично $\frac{B^{\prime} C_{2}}{B^{\prime} A_{2}}=\frac{B^{\prime \prime} C_{2}}{B^{\prime \prime} A_{2}}$ и $\frac{A^{\prime} B_{2}}{A^{\prime} C_{2}}=\frac{A^{\prime \prime} B_{2}}{A^{\prime \prime} C_{2}}$, где $A^{\prime}$ и $B^{\prime}$ - точки пересечения $d_{a}$ и $B_{2} C_{2}, d_{b}$ и $A_{2} C_{2}$, а $A^{\prime \prime}$ и $B^{\prime \prime}-$ точки пересечения медиан $A A_{1}$ и $B B_{1}$ с окружностью Эйлера. Заметим, что $$ \frac{C^{\prime} A_{2} \cdot B^{\prime} C_{2} \cdot A^{\prime} B_{2}}{C^{\prime} B_{2} \cdot B^{\prime} A_{2} \cdot A^{\prime} C_{2}}=\frac{C^{\prime \prime} A_{2} \cdot B^{\prime \prime} C_{2} \cdot A^{\prime \prime} B_{2}}{C^{\prime \prime} B_{2} \cdot B^{\prime \prime} A_{2} \cdot A^{\prime \prime} C_{2}}=1 $$ Поэтому по теореме Менелая точки $A^{\prime}, B^{\prime}, C^{\prime}$ лежат на одной прямой и по теореме Дезарга треугольник, образованный директрисами, и ортотреугольник перспективны. 23. Как известно, из медиан треугольника можно составить треугольник. Рассмотрим треугольник, составленный из медиан треугольника $A B C$. Из задачи 5 следует, что углы треугольника, образованного директрисами, равны углам треугольника, образованного медианами. Поэтому достаточно доказать, что равны углы Брокара треугольника $A B C$ и треугольника, составленного из его медиан. Воспользуемся следующими формулами, справедливыми для произвольного треугольника: (1) $\operatorname{ctg} \phi=\operatorname{ctg} \alpha+\operatorname{ctg} \beta+\operatorname{ctg} \gamma$. (2) $\operatorname{ctg} \alpha+\operatorname{ctg} \beta+\operatorname{ctg} \gamma=\frac{a^{2}+b^{2}+c^{2}}{4 S}$. (3) $S=\frac{3}{4} S_{m}$. (4) $a^{2}+b^{2}+c^{2}=\frac{3}{4}\left(m_{a}^{2}+m_{b}^{2}+m_{c}^{2}\right)$. где $\alpha, \beta$ и $\gamma$ - углы треугольника, а $\phi$ - его угол Брокара; $S$ и $S_{m}$ - площади данного треугольника и треугольника, составленного из его медиан; $a, b$ и $c$ - длины стороны треугольника; $m_{a}, m_{b}$ и $m_{c}$ - длины его медиан. Применяя эти формулы к треугольнику $A B C$, получаем, что $$ \operatorname{ctg} \phi=\operatorname{ctg} \alpha+\operatorname{ctg} \beta+\operatorname{ctg} \gamma=\frac{a^{2}+b^{2}+c^{2}}{4 S}=\frac{m_{a}^{2}+m_{b}^{2}+m_{c}^{2}}{4 S_{m}}=\operatorname{ctg} \phi_{m} $$ где $\phi_{m}$ - угол Брокара треугольника, составленного из медиан. Значит, $\phi=\phi_{m}$, что и требовалось. ## Three parabolas
F.Nilov, A.Zaslavsky We will examine the properties of the next configuration: a triangle and three parabolas, each passing through two vertices and touching in these vertices the respective sidelines. ## 1 Necessary definitions and theorems Definition 1. A parabola is the locus of points with equal distances from a fixed point $F$ and a fixed line $l$. $F$ is the focus and $l$ is the directrix of the parabola. Definiton 2. Given triangle $A B C$ and point $P$. Then the reflections of lines $A P, B P, C P$ in respective bisectors concur in point $P^{\prime}$. This point is called isogonally conjugate to $P$. Point $L$, isogonally conjugate to the centroid $M$, is called the Lemoine point. Definition 3. The projections of point $P$ to the sidelines of $A B C$ form the triangle called the pedal triangle of $P$ wrt $A B C$. The circumcircle of the pedal triangle is called the pedal circle of $P$. Definition 4. The triangles $A B C$ and $A^{\prime} B^{\prime} C^{\prime}$ are called perspective if the lines $A A^{\prime}, B B^{\prime}$ and $C C^{\prime}$ concur. The common point of these lines is called the perspective center. The Desargues theorem. Two triangles are perspective iff the common points of their respective sidelines are collinear. The line passing through these points is called the perspective axis. Definition 5. Triangles $A B C$ and $A^{\prime} B^{\prime} C^{\prime}$ are called orthological if the perpendiculars from $A$ to $B^{\prime} C^{\prime}$, from $B$ to $A^{\prime} C^{\prime}$ and from $C$ to $A^{\prime} B^{\prime}$ concur. The common point of these perpendiculars is called the orthology center. This property is reciprocal, so the orthological triangles have two orthology centers, distinct in general. The centroid $M$, the circumcenter $O$ and the orthocenter $H$ of an arbitrary triangle lie on the line called the Euler line. The midpoints of the sides and the feet of the altitudes lie on the circle called the Euler circle. Definition 6. For an arbitrary triangle $A B C$, there exist two points $Q_{1}, Q_{2}$, such that $\angle Q_{1} A B=$ $\angle Q_{1} B C=\angle Q_{1} C A$ and $\angle Q_{2} B A=\angle Q_{2} C B=\angle Q_{2} A C$. These points are called the Brocard points of $A B C$. The angle $\phi=\angle Q_{1} A B=\angle Q_{2} B A$ is called the Brocard angle. For any non-regular triangle there exist two points with regular pedal triangle. These points are inverse wrt the circumcircle of the triangle. The point lying inside the circumcircle is called the first Apollonius point, another point is called the second Apollonius point. Two points isogonally conjugate to the Apollonius points are called the first and the second Torricelli points. ## 2 Introductory problems 1. Prove the optic property of the parabola: the tangent to a parabola in a point $X$ forms equal angles with the line $F X$ and the axis of parabola. 2. Prove that the reflection of the focus in the tangent to the parabola in point $X$ coincides with the projection of $X$ to the directrix. 3. Let $P$ be the common point of tangents to parabola in points $X$ and $Y$. Prove that $P$ is the circumcenter of triangle $F X^{\prime} Y^{\prime}$, where $X^{\prime}, Y^{\prime}$ are the projections of $X$ and $Y$ to the directrix. 4. Prove that in notation of the previous problem, the median of $P X Y$ is parallel to the axis of parabola. 5. Prove that in notation of problem 3, the bisector of angle $P$ forms equal angles with the line $P F$ and the axis of the parabola. 6. Prove that two isogonally conjugate points have common pedal circle. Given triangle $A B C$. Denote by $\Pi_{c}$ the parabola touching the lines $A C$ и $B C$ in points $A$ and $B$. Furthermore denote by $F_{c}$ the focus of $\Pi_{c}$. The parabolas $\Pi_{a}, \Pi_{b}$ and their focuses $F_{a}, F_{b}$ are defined similarly. 7. Prove that $\Pi_{a}$ and $\Pi_{b}$ have only one common point $C^{\prime}$, different from $C$. 8. Define points $A^{\prime}, B^{\prime}$ similarly to $C^{\prime}$. Prove that triangles $A B C$ and $A^{\prime} B^{\prime} C^{\prime}$ are perspective. 9. Prove that the triangles $A B C$ and $F_{a} F_{b} F_{c}$ are perspective. 10. Prove that the triangle formed by the directrices of $\Pi_{a}, \Pi_{b}, \Pi_{c}$ is perspective to $A B C$. ## 3 Basic problems 11. Prove that $F_{c} C$ is the bisector of angle $A F_{c} B$. 12. Prove that $A, B, F_{c}$ and the circumcenter $O$ of $A B C$ are cocyclic. 13. Prove that the medial line parallel to $A B$ touches $\Pi_{c}$. 14. Prove that $F_{a}, F_{b}, F_{c}, O, L$ are cocyclic. 15. Prove that the directrix of $\Pi_{c}$, the median from $C$ and the Euler circle concur. 16. Prove that the centroids of $A B C$ and of the triangle formed by three directrices coincide. 17. Prove that $A B C$ and the directrix triangle are orthological and their orthology centers coincide. 18. Prove that the Euler line of $A B C$ passes through the Lemoine point of the triangle formed by three directrices. 19. Prove that the Euler line of the directrix triangle passes through $L$. 20. Let $A F_{a}, B F_{b}, C F_{c}$ intersect the respective sidelines of $A B C$ in the points $P_{a}, P_{b}, P_{c}$. Prove that triangle $P_{a} P_{b} P_{c}$ is perspective to the directrix triangle. 21. Prove that the pairwise perspective centers of $A B C, P_{a} P_{b} P_{c}$ and of the directrix triangle are collinear. The respective line passes through the midpoint of the segment between the centroids of $A B C$ and $P_{a} P_{b} P_{c}$. 22. Prove that the directrix triangle is perspective to the orthotriangle of $A B C$. 23. Prove that the Brocard angles of $A B C$ and the directrix triangle are equal. ## 4 Additional problems 24. Let $T$ be some Torricelli point of $A B C$ and $T_{d}$ the respective Torricelli point of the directrix triangle. Prove that the lines $T A, T B, T C$ are parallel to the lines joining $T_{d}$ with respective vertices of the directrix triangle. 25. Prove that $T T_{d}$ and the line joining the respective Apollonius points pass through the centroid of $A B C$. 26. Find $\frac{S}{S^{\prime}}$, where $S^{\prime}$ is the area of the curvilinear "parabolic"triangle $A^{\prime} B^{\prime} C^{\prime}$, and $S$ is the area of $A B C$. 27. Let $N_{1}, N_{2}$ be the perspective centers of the directrix triangle and the triangles $A B C, P_{a} P_{b} P_{c}$ respectively. Prove that the line $N_{1} N_{2}$ bisects the segment between the centroids of $A B C$ and $P_{a} P_{b} P_{c}$. 28. Let $F_{a}^{\prime}, F_{b}^{\prime}, F_{c}^{\prime}$ be isogonally conjugate to $F_{a}, F_{b}, F_{c} ; H$ is the orthocenter, $M$ the centroid of $A B C$. Prove that a) $F_{a}^{\prime}, F_{b}^{\prime}, F_{c}^{\prime}$ lie on the circle with diameter $H M$. b) The distances from $F_{a}$ and $F_{a}^{\prime}$ to the center of the Euler circle of $A B C$ are equal. The point of parabola $\Pi_{c}$, nearest to point $C$, will be called the projection of point $C$ to the parabola and will be denoted by $C^{*}$. 29. Prove that line $C C^{*}$ is perpendicular to parabola $\Pi_{c}$. 30. Prove that $\angle A C^{*} C=\angle B C^{*} C$. 31. Denote by $C_{1}$ and $C_{2}$ the common points of the tangent to the parabola at point $C^{*}$ with $C A$ and $C B$ respectively. Denote by $C^{* *}$ the common point of $C_{1} C_{2}$ and the median of triangle $A B C$ drawn from vertex $C$. Prove that $C_{1} C^{*}=C^{* *} C_{2}$. Define points $A^{*}$ and $B^{*}$ similarly to $C^{*}$. r liner 32. (Conjecture.) Lines $A A^{*}, B B^{*}$ and $C C^{*}$ concur. ## Three parabolas ## 5 Related problems Consider the set of lines bissecting the perimeter of triangle $A B C$. 33. Prove that these lines envelop three parabolas, each touching two sidelines of $A B C$ in the same points that the respective excircle. Note by $\Pi_{A}$ the parabola touching $A B$ and $A C$. Define $\Pi_{B}$ and $\Pi_{C}$ similarly. 34. Prove that the circumcircle of $A B C$ is the orthology center of $A B C$ and the triangle formed by the directrix of $\Pi_{A}, \Pi_{B}$ and $\Pi_{C}$. 35. Prove that the orthocenter of $A B C$ lies on the Euler line of triangle formed by the directrix. 36. Prove that the incenter of $A B C$ and the circumcenter of directrix triangle are symmetric wrt the center of Euler circle of $A B C$. 37. Let $T$ be some Torricelli point of triangle $A B C, T_{d}$ be the respective point of directrix triangle. Prove that the lines, joining $T$ and $T_{d}$ with respective vertex are parallel. ## 6 Related problems-2 Consider points $A_{c}$ and $A_{b}$ on rays $A B$ and $A C$ such that $C A=C A_{c}$ and $B A=B A_{b}$. Let parabola $\Pi_{a}^{\prime}$ touche $A C$ and $A B$ in points $A_{b}$ and $A_{c}$. Parabolas $\Pi_{b}^{\prime}$ and $\Pi_{c}^{\prime}$ are defined similarly. 38. Prove that the focus of $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ and $\Pi_{c}^{\prime}$ coincide with $F_{a}^{\prime}, F_{b}^{\prime}$ and $F_{c}^{\prime}$. 39. Prove that the triangle formed by directrix of $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ and $\Pi_{c}^{\prime}$ is perspective to $A B C$. 40. Prove that the directrix triangle of $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ and $\Pi_{c}^{\prime}$ is ortologic to $A B C$ and their othology centers coincide. 41. Prove that the directrix triangle of $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ and $\Pi_{c}^{\prime}$ is perspective to the orthotriangle of $A B C$. 42. Let $O^{\prime}$ and $M^{\prime}$ be the circumcenter and the centroid of directrix triangle. Prove that the common point of lines $O^{\prime} M$ and $M^{\prime} O$ coincide with the Lemoine point of $A B C$. ## Three parabolas
Solutions
F.Nilov, A.Zaslavsky ## 2 Introductory problems 1. Let $X^{\prime}$ be the projection of $X$ to the directrix. Then $F X=X X^{\prime}$. Suppose that the bisectrix $l$ of the angle $X^{\prime} X F$ isn't the tangent to parabola. Then $l$ intersect the parabola in some point $Y$, distinct from $X$. Note that the triangles $F X Y$ and $X^{\prime} X Y$ are equal and so $F Y=Y X^{\prime}$. Let $Y^{\prime}$ be the projection of $Y$ to the directrix. Then $Y Y^{\prime}=F Y=Y X^{\prime}$. But it is impossible because $Y Y^{\prime}$ and $Y X^{\prime}$ are the hypothenuse and the cathetus of rectangle triangle $Y X^{\prime} Y^{\prime}$. 2. Let $l$ touche the parabola in the point $X$. Then $F X=X X^{\prime}$. By problem 1 is the bisectrix of isosceles triangle $F X X^{\prime}$. So $l$ is the medial perpendicular to the segment $F X^{\prime}$, and the points $X^{\prime}$ and $F$ are symmetric wrt $l$. 3. By problem 2 the lines $P X$ and $P Y$ are medial perpendiculars of segments $F X^{\prime}$ и $F Y^{\prime}$. So their common point $P$ is the circumcenter of $F X^{\prime} Y^{\prime}$. 4. Consider the medial line of the trapezoid $X X^{\prime} Y^{\prime} Y$. It is perpendicular to the directrix and so parallel to the axis of the parabola. Also it is the median of $P X Y$, because it pass through the midpoint of segment $X Y$ and (by problem 3) through $P$. 5. Consider the line $l$ passing through $P$ and parallel to the axis of the parabola. We must prove that the angle $\phi$ between $l$ and $P X$ is equal to the angle $F P Y$. Let $X^{\prime \prime}$ and $Y^{\prime \prime}$ be the projections of $F$ to the lines $P X$ and $P Y$. By problem $2 X^{\prime \prime}$ and $Y^{\prime \prime}$ are the midpoints of $F X^{\prime}$ and $F Y^{\prime}$. So the lines $X^{\prime} Y^{\prime}$ and $X^{\prime \prime} Y^{\prime \prime}$ are parallel. This yields that $l$ and $X^{\prime \prime} Y^{\prime \prime}$ are perpendicular. Now we have $\angle \phi=90^{\circ}-\angle P X^{\prime \prime} Y^{\prime \prime}=90^{\circ}-\angle P F Y^{\prime \prime}=\angle F P Y$, q.e.d. 6. Let $P$ and $Q$ be isogonal conjugated. Note as $P_{c}$ and $Q_{c}, P_{a}$ and $Q_{a}, P_{b}$ and $Q_{b}$ the reflections of $P$ and $Q$ in $A B, B C$ and $A C$. It is evident that $B P_{c}=B P=B P_{a}$ and $\angle P_{c} B P_{a}=2 \angle B$. Note that $\angle Q B C=\angle P B A=\angle A B P_{c}$. So $\angle P_{c} B Q=\angle B=$ $1 / 2 \angle P_{c} B P_{a}=\angle P_{a} B Q$, and $Q P_{c}=Q P_{a}$. Similarly $Q P_{a}=Q P_{b}$ и $P Q_{c}=P Q_{a}=P Q_{b}$. Using the homothety with center $P$ and coefficient $1 / 2$ we receive that the circle $\omega$ with center in the midpoint of $P Q$ and the radius $Q P_{c} / 2$ is the pedal circle of $P$. Similarly $\omega$ is the pedal circle of $Q$. 7. As $\Pi_{a}$ and $\Pi_{b}$ are inscribed in the angles $A$ and $B$ their common points lie inside or on the sidelines of the triangle $A B C . C$ isn't the unique common point of parabolas because their tangents in $C$ doesn't coincide. So there exists the common point $C^{\prime}$ distonct from $C$. It is clear that the sidelines of $A B C$ doesn't contain the common points distinct from $C$. Suppose that $\Pi_{a}$ and $\Pi_{b}$ have inside the triangle the common point $C^{\prime \prime}$ distinct from $C^{\prime}$. As $A, C, C^{\prime}$ и $C^{\prime \prime}$ lie on $\Pi_{b}$, they are the vertex of convex quadrilateral. Similarly $B$, $C, C^{\prime}$, and $C^{\prime \prime}$ are the vertex of convex quadrilateral. But it is impossible. So $\Pi_{a}$ and $\Pi_{b}$ have exactly two common points 8. It is known that an arbitrary affine map transforms any parabola to the parabola. Consider the map transforming the triangle $A B C$ to regular triangle. It is evident that in regular triangle the common points of parabolas lie on the medians. So $A A^{\prime}, B B^{\prime}$ and $C C^{\prime}$ are the medians of $A B C$ and concur in its centroid $M$. 9. By problems 4 and 5 the lines $A M$ and $A F_{a}$ are symmetric wrt the bisectrix of angle $A$. So $A F_{a}, B F_{b}$ and $C F_{c}$ concur in the Lemoine point $L$ of $A B C$. 10. Let the directrix $d_{a}$ of $\Pi_{a}$ intersect $B C$ in the point $A^{\prime}$. Note the projections of $B$ and $C$ to $d_{a}$ as $B_{1}$ and $C_{1}$. It is clear that $B B_{1}=B F_{a}$ and $C C_{1}=C F_{a}$. As the triangles $A^{\prime} B B_{1}$ and $A^{\prime} C C_{1}$ are similar $\frac{A^{\prime} B}{A^{\prime} C}=\frac{B B_{1}}{C C_{1}}=\frac{B F_{a}}{C F_{a}}$. By problems 1 and $5 \angle F_{a} B A=\angle F_{a} A C$ and $\angle F_{a} C A=\angle F_{a} A B$. Usung the sinus theorem for the triangle $F_{a} B A$ we receive that $B F_{a}=\frac{\sin \left(\angle F_{a} A B\right) \cdot A F_{a}}{\sin \left(\angle A B F_{a}\right)}$. Similarly $C F_{a}=\frac{\sin \left(\angle F_{a} A C\right) \cdot A F_{a}}{\sin \left(\angle A C F_{a}\right)}$. So $\frac{B F_{a}}{C F_{a}}=\frac{\sin ^{2}\left(\angle F_{a} A B\right)}{\sin ^{2}\left(\angle F_{a} A C\right)}$. Let $d_{b}$ and $d_{c}$ be the directrix of $\Pi_{b}$ and $\Pi_{c}$. Note as $B^{\prime}, C^{\prime}$ the common points of $d_{b}$ and $A C, d_{c}$ and $A B$. Then $$ \frac{A^{\prime} B \cdot B^{\prime} C \cdot C^{\prime} A}{A^{\prime} C \cdot B^{\prime} A \cdot C^{\prime} B}=\frac{B F_{a} \cdot C F_{b} \cdot A F_{c}}{C F_{a} \cdot A F_{b} \cdot B F_{c}}=\frac{\sin ^{2}\left(\angle F_{a} A B\right) \cdot \sin ^{2}\left(\angle F_{b} B C\right) \cdot \sin ^{2}\left(\angle F_{c} C B\right)}{\sin ^{2}\left(\angle F_{a} A C\right) \cdot \sin ^{2}\left(\angle F_{b} B A\right) \cdot \sin ^{2}\left(\angle F_{c} C A\right)}=1 $$ The last equality follows from the Ceva theorem for the cevians $A F_{a}, B F_{b}$ and $C F_{c}$. By Menelaus theorem $A^{\prime}, B^{\prime}$ and $C^{\prime}$ are collinear. So by Desargues theorem the triangles are perspective. # Three parabolas Solutions F.Nilov, A.Zaslavsky ## 3 Basic problems 11. Note the midpoint of $A B$ as $C_{0}$. We have $\angle C_{0} C B=\angle F_{c} C A$. By problems 1,4 and $5 \angle M C B=\angle F_{c} B C$ и $\angle M C A=\angle F_{c} A C$. So $\angle A F_{c} C=180^{\circ}-\angle F_{c} A C-\angle F_{c} C A=$ $180^{\circ}-\angle C=180^{\circ}-\angle F_{c} C B-\angle F_{c} B C=\angle B F_{c} C$. 12. By problem $11 \angle A F_{c} C=\angle B F_{c} C=180^{\circ}-\angle C$. So $\angle A F_{c} B=360^{\circ}-2\left(180^{\circ}-\angle C\right)=$ $2 \angle C=\angle A O B$. This follows that $A, B, F_{c}$ and $O$ lie on the circle $\omega$. 13. Let $A_{0}$ и $B_{0}$ be the midpoints of $B C$ and $A C$. The problem 5 yields next Lemma. The focus of the parabola inscribed in the triangle lies on the circumcircle of this triangle Prove that the circumcircle of $A_{0} B_{0} C$ pass through $F_{c}$. Note as $C^{\prime}$ the second common point of $C F_{c}$ with the cirumcircle of $A B C$. By problems 1 and $5 \angle F_{c} B C=\angle F_{c} C A$ and $\angle F_{c} A C=\angle F_{c} C B$. So $\angle C^{\prime} B A=\angle F_{c} C A=\angle F_{c} B C$ и $\angle B A C^{\prime}=\angle F_{c} C B$. From this the triangles $F_{c} C B$ and $C^{\prime} A B$ are similar and $\frac{C F_{c}}{C^{\prime} A}=\frac{B F}{B C^{\prime}}$. By problem $11 \angle B F_{c} C^{\prime}=\angle A F_{c} C^{\prime}$. Also $\angle C^{\prime} B F_{c}=\angle A B C=\angle A C^{\prime} F_{c}$. So the triangles $F_{c} B C^{\prime}$ and $F_{c} C^{\prime} A$ are similar and $\frac{F_{c} C^{\prime}}{C^{\prime} A}=\frac{B F}{B C^{\prime}}=\frac{C F_{c}}{C^{\prime} A}$. It follows that $F_{c} C=F_{c} C^{\prime}$. The homothety with center $C$ and coefficient $\frac{1}{2}$ transforms $B, A$ and $C^{\prime}$ to $A_{0}, B_{0}$ and $F_{c}$, q.e.d. 14. By problem 11 the points $A, B, F_{c}$ and $O$ lie on the circle $\omega$. Let $O^{\prime}$ be the second common point of $C F_{c}$ and $\omega$. By problem $10 \angle A F_{c} O^{\prime}=\angle B F_{c} O^{\prime}$. So $O^{\prime}$ is is the midpoint of an arc $A B$ of $\omega$ and the segment $O O^{\prime}$ is the diameter of $\omega$. From this $\angle L F_{c} O=\angle O F_{c} O^{\prime}=90^{\circ}$. Similarly $\angle L F_{a} O=\angle L F_{b} O=90^{\circ}$. So $F_{a}, F_{b}$ lie $F_{c}$ lie on the circle with diameter $O L$. 15. Let $A_{0}$ and $B_{0}$ be the midpoints of $B C$ and $A C$. Consider the point $F_{c}^{\prime}$ symmetric to $F_{c}$ wrt $A_{0} B_{0}$. By problem $13 A_{1} B_{1}$ touches $\Pi_{c}$. So by problem $2 F^{\prime}$ lies on the directrix of $\Pi_{c}$. By problem 13 the quadrilateral $C A_{1} F_{c} B_{1}$ is cyclic. So $\angle A_{1} F_{c}^{\prime} B_{1}=\angle A_{1} F_{c} B_{1}=180^{\circ}-\angle C$ and $F_{c}^{\prime}$ lies on the Euler circle of $A B C$. Note as $M^{\prime}$ the midpoint of $A_{0} B_{0}$. By problem $13 A_{1} B_{1}$ is the bisectrix of angle $C M F_{c}$. So $F_{c}^{\prime}$ lies on $C M^{\prime}$. Thus the common point of the median and Euler circle lies on the directrix of $\Pi_{c}$. 16. Let $A^{\prime \prime}, B^{\prime \prime}, C^{\prime \prime}$ be the vertex of directrix triangle. Note as $A^{\prime}, B^{\prime}$ and $C^{\prime}$ the common points of medians of $A B C$ and its Euler circle. By problem 15 these points lie on the sidelines of $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$. Note as $C_{0}$ the midpoint of $A B$. Let $G$ be the centroid of $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$. Note that $$ \frac{\sin \left(\angle G C^{\prime \prime} B^{\prime \prime}\right)}{\sin \left(\angle G C^{\prime \prime} A^{\prime \prime}\right)}=\frac{\sin \left(\angle B^{\prime \prime}\right)}{\sin \left(\angle A^{\prime \prime}\right)}=\frac{\sin \left(\angle C_{0} M A\right)}{\sin \left(\angle C_{0} M B\right)}=\frac{M B}{M A}=\frac{M A^{\prime}}{M B^{\prime}}=\frac{\sin \left(\angle M C^{\prime \prime} B^{\prime \prime}\right)}{\sin \left(\angle M C^{\prime \prime} A^{\prime \prime}\right)} $$ The first and the third equalities are correct because $C^{\prime \prime} G$ and $M C_{0}$ are the medians of $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$ and $A M B$. So $\angle G C^{\prime \prime} B^{\prime \prime}=\angle M C^{\prime \prime} B^{\prime \prime}$ and $G$ lies on $C^{\prime \prime} M$. Similarly $G$ lies on $A^{\prime \prime} M$ and $B^{\prime \prime} M$. this follows that $G$ and $M$ coincide. 17. By problem $4 A M$ is perpendicular to $d_{a}$. Similarly $B M$ is perpendicular to $d_{b}$ and $C M$ is perpendicular to $d_{c}$. So $M$ is the orthology center of considering triangles. Let $A^{\prime}, B^{\prime}$ и $C^{\prime}$ be the common points of respective directrix. As the medians of $A B C$ are perpendicular to the sidelines of $A^{\prime} B^{\prime} C^{\prime}$ we have : (1) $\left(\overrightarrow{C^{\prime} A^{\prime}}+\overrightarrow{A^{\prime} B^{\prime}}\right)$. $\cdot(\overrightarrow{A B}+\overrightarrow{A C})=0$ (2) $\overrightarrow{A^{\prime} B^{\prime}} \cdot(2 \overrightarrow{C A}+\overrightarrow{A B})=0$. (3) $\overrightarrow{C^{\prime} A^{\prime}} \cdot(2 \overrightarrow{A B}+\overrightarrow{C A})=0$. Summing (1) and (3) and substracting from the result (2), we receive that $\overrightarrow{C A} \cdot \overrightarrow{A^{\prime} B^{\prime}}=$ $\overrightarrow{C^{\prime} A^{\prime}} \cdot \overrightarrow{A B}$. From this and (1) $\left(\overrightarrow{C^{\prime} A^{\prime}}-\overrightarrow{A^{\prime} B^{\prime}}\right) \cdot(\overrightarrow{A B}-\overrightarrow{A C})=0$. This follows that $B C$ is perpendicular to the median $A^{\prime} G$ of $A^{\prime} B^{\prime} C^{\prime}$. By problem 16 the points $G$ and $M$ coincide. So $A^{\prime} M$ and $B C$ are perpendicular. Similarly $B^{\prime} M$ and $A C, C^{\prime} M$ and $A B$ are perpendicular. So $M$ is the common orthology center of $A B C$ and $A^{\prime} B^{\prime} C^{\prime}$. Note that two orthological triangles with coinciding orthology centers are perspective. So we receive another solution of problem 10. 18. As $d_{a}$ is perpendicular to the median $A A_{0}$, then by problem $15 d_{a}$ pass through the common point $A^{\prime \prime}$ of an altitude $A A^{\prime}$ and the Euler circle. Similarly the common points of Euler circle $B^{\prime \prime}, C^{\prime \prime}$ of Euler circle and the altitudes $B B^{\prime}, C C^{\prime}$ lie on $d_{b}$ and $d_{c}$. Note that Euler circle of $A B C$ is the pedal circle of $M$ wrt the directrix triangle. By problem $16 M$ is the centroid of directrix triangle. So the medians of $A B C$ are parallel to the lines $A^{\prime \prime} L^{\prime}, B^{\prime \prime} L^{\prime}$ and $C^{\prime \prime} L^{\prime}$, where $L^{\prime}$ is the Lemoine point of the directrix triangle. It is known that the triangles $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$ and $A B C$ are homothetic with center $H$ and coefficient $1 / 2$. So, as $L^{\prime}$ is the centroid of $A^{\prime \prime} B^{\prime \prime} C^{\prime \prime}$ it is the midpoint of $H M$ and lies on the Euler line of $A B C$. 19. Let $T$ be the directrix triangle. Consider the parabolas $\Pi_{a}^{\prime}, \Pi_{b}^{\prime}$ и $\Pi_{c}^{\prime}$, touching the sidelines of $T$ in its vertex. Note as $T^{\prime}$ the triangle formed by the directrix of these parabolas. The respective sidelines of $T^{\prime}$ and $A B C$ are perpendicular to the medians of $T$. So $T^{\prime}$ and $A B C$ are homothetic and $M$ is the homothety center because by problem 16 the centroids of $T^{\prime}$ and $A B C$ coinside. Note the Lemoine point of $T^{\prime}$ as $L^{\prime}$. By th homothety of $T^{\prime}$ and $A B C$ the points $M, L$ and $L^{\prime}$ are collinear. By problem $18 L^{\prime}$ Lies on the Euler line of $T$. So the line passing through $L, M$ and $L^{\prime}$ coincide with the Euler line of $T$. 20. Let $C^{\prime}$ be the common point of $A B$ and $P_{a} P_{b}$. The points $C^{\prime}, P_{c}, A B$ are harmonic because $\frac{C^{\prime} A}{C^{\prime} B}=\frac{A P_{c}}{P_{c} B}$. By problem $11 \frac{A P_{c}}{P_{c} B}=\frac{A F_{c}}{F_{c} B}$. Let $C^{\prime \prime}$ be the common point of $d_{c}$ and $A B$. By problem $10 \frac{C^{\prime \prime} A}{C^{\prime \prime} B}=\frac{A F_{c}}{F_{C} B}$. So $\frac{C^{\prime} A}{C^{\prime} B}=\frac{A F_{c}}{F_{c} B}=\frac{C^{\prime \prime} A}{C^{\prime \prime} B}$ and the points $C^{\prime}$ and $C^{\prime \prime}$ coincide. By problem $10 A^{\prime \prime}, B^{\prime \prime}$ and $C^{\prime \prime}$ are collinear. So $A^{\prime}, B^{\prime}$ and $C^{\prime}$ are also collinear and by the Desargues theorem the triangles $A B C$ and $P_{a} P_{b} P_{c}$ are perspective. 21. By problem 20 the perspective axis of $A B C, P_{a} P_{b} P_{c}$ and the directrix triangle coincide. Now use next Lemma. If the perspective axis of three mutually perspective triangles coincide then their pair perspective centers are collinear. Proof. Consider the projective map transforming the commo perspective axis to the infinite line. It transforms given triangles to the homothetic triangles. Their homothtety centers are collinear. 22. Let $A_{1}, B_{1}$ и $C_{1}$ be the midpoints of the sides of $A B C$ and $A_{2}, B_{2}$ и $C_{2}$ be the feet of its altitudes. Note as $C^{\prime \prime \prime}$ the common point of $C C_{1}$ and the Euler circle. By problem 15 $C^{\prime \prime \prime}$ lies on $d_{c}$. Let $C^{\prime}$ be the common point of $d_{c}$ and $A^{\prime} B^{\prime}$, and $C^{\prime \prime}$ be the common point of $C C_{1}$ and $A_{2} B_{2}$. Note that $B_{2} C_{1}=A C_{1}=B C_{1}=A_{2} C_{1}$. So $\angle B_{2} C^{\prime \prime \prime} C_{1}=\angle A_{2} C^{\prime \prime \prime} C_{1}$, because the quadrilateral $A_{2} C^{\prime \prime \prime} B_{2} C_{1}$ is cyclic. By problem $5 d_{c}$ is perpendicular to $C^{\prime \prime \prime} C^{\prime \prime}$. So $C^{\prime \prime \prime} C^{\prime}$ is the external bisectrix of angle $C^{\prime \prime \prime}$, and $\frac{C^{\prime} A_{2}}{C^{\prime} B_{2}}=\frac{C^{\prime \prime} A_{2}}{C^{\prime \prime} B_{2}}$. Similarly $\frac{B^{\prime} C_{2}}{B^{\prime} A_{2}}=\frac{B^{\prime \prime} C_{2}}{B^{\prime \prime} A_{2}}$ and $\frac{A^{\prime} B_{2}}{A^{\prime} C_{2}}=\frac{A^{\prime \prime} B_{2}}{A^{\prime \prime} C_{2}}$, where $A^{\prime}, B^{\prime}$ are the common points of $d_{a}$ and $B_{2} C_{2}, d_{b}$ and $A_{2} C_{2} ; A^{\prime \prime}$, $B^{\prime \prime}$ are the common points of $A A_{1}, B B_{1}$ and the Euler circle. Note that $$ \frac{C^{\prime} A_{2} \cdot B^{\prime} C_{2} \cdot A^{\prime} B_{2}}{C^{\prime} B_{2} \cdot B^{\prime} A_{2} \cdot A^{\prime} C_{2}}=\frac{C^{\prime \prime} A_{2} \cdot B^{\prime \prime} C_{2} \cdot A^{\prime \prime} B_{2}}{C^{\prime \prime} B_{2} \cdot B^{\prime \prime} A_{2} \cdot A^{\prime \prime} C_{2}}=1 $$ So by Menelaus theorem $A^{\prime}, B^{\prime}, C^{\prime}$ are collinear and by Desargues theorem the directrix triangle and the orthotriangle are perspective . 23. Consider the triangle formed by the medians of $A B C$. By problem 5 its angles are equal to the angles of directrix triangle. Prove that the Brocard angles of $A B C$ and the triangles formed by its medians are equal. Next formulaes are correct for any triangle: (1) $\operatorname{ctg} \phi=\operatorname{ctg} \alpha+\operatorname{ctg} \beta+\operatorname{ctg} \gamma$. (2) $\operatorname{ctg} \alpha+\operatorname{ctg} \beta+\operatorname{ctg} \gamma=\frac{a^{2}+b^{2}+c^{2}}{4 S}$. (3) $S=\frac{3}{4} S_{m}$. (4) $a^{2}+b^{2}+c^{2}=\frac{3}{4}\left(m_{a}^{2}+m_{b}^{2}+m_{c}^{2}\right)$. there $\alpha, \beta$ and $\gamma$ are the angles of triangle, $\phi$ is its Brocard angle; $S$ and $S_{m}$ are the areas of the triangle and the triangle formed by its medians; $a, b$ and $c$ are the lengths of the sides; $m_{a}, m_{b}$ and $m_{c}$ are the lengths of the medians. Using these formulaes to $A B C$ we receive $$ \operatorname{ctg} \phi=\operatorname{ctg} \alpha+\operatorname{ctg} \beta+\operatorname{ctg} \gamma=\frac{a^{2}+b^{2}+c^{2}}{4 S}=\frac{m_{a}^{2}+m_{b}^{2}+m_{c}^{2}}{4 S_{m}}=\operatorname{ctg} \phi_{m} $$ where $\phi_{m}$ is the Brocard angle of the triangle formed by the medians. So, $\phi=\phi_{m}$, q.e.d. # Взвешивания со сломанными весами К.А. Кноп, Г.Р. Челноков, И.И. Богданов ## 1 Постановка проблемы ## Базовые понятия Как обычно, у нас имеется множество из нескольких одинаковых на вид монет, одна из которых фальшивая - она немного легче, чем настоящая; требуется определить, какая именно. При этом в разных задачах мы можем пользоваться разными тестирующими устройствами. Детектор - это устройство, которое за одно действие (тестирование) про любое выбранное подмножество монет сообщает, содержится в нем фальшивая или нет. Таким образом, детектор при любом разбиении полного множества на 2 подмножества определяет, в каком из подмножеств оказалась фальшивая монета). Рычажные весы (или просто весы) - это устройство, позволяющее сравнивать между собой веса двух подмножеств. Таким образом, для двух подмножеств из равного количества монет весы сравнивают количества фальшивых в них. Поэтому в нашей задаче весы говорят нам, в каком из трех подмножеств разбиения находится фальшивая монета (при этом два из подмножеств должны иметь поровну элементов). Главная изюминка данной задачи: мы разрешаем некоторым устройствам сообщать неверную информацию. При этом устройства не уподобляются лжецам, которые «всегда врут». Они просто сломаны - то есть информация от такого устройства может быть как истинной, так и ложной. Грубо говоря, вместо правильного работающего устройства у нас работает генератор случайных ответов. Итого, у нас есть несколько устройств, и мы знаем лишь, сколько из них сломаны, но не знаем, какие. (При этом за одну операцию тестирование проводится только на одном из устройств!) Введем обозначения. Пусть $Д_{x, y}(z)$ обозначает минимальное число тестирований детекторами, необходимое для гарантированного нахождения одной фальшивой монеты среди $z$ при помощи $x$ детекторов, из которых $y$ сломаны (эту тестирующую систему мы будем обозначать через $x$ д $[y]$ ). Аналогично, $\mathrm{B}_{x, y}(z)$ будет обозначать то же для весов (а соответствующая система обозначается $x \mathrm{~B}[y]$ ). На протяжении первых разделов мы будем приводить для каждой задачи две формулировки: с использованием введенных обозначений, и (для тех, кому этот язык еще непривычен) без их использования. ## 2 Вводные задачи: конкретные случаи ## Что точно можно? 2.1. Тремя весами, из которых одни сломаны, можно из 3 монет найти фальшивую за 3 взвешивания (На удобном языке: $\mathrm{B}_{3,1}(3) \leqslant 3$ ). 2.2. a) Тремя детекторами, из которых один сломан, можно из 8 монет найти фальшивую за 6 тестирований. (На удобном языке: Д ${ }_{3,1}(8) \leqslant 6$.) b) Тремя весами, из которых одни сломаны, можно из 9 монет найти фальшивую за 4 взвешивания. (На удобном языке: $\mathrm{B}_{3,1}(9) \leqslant 4$ ). 2.3. a) Тремя детекторами, из которых один сломан, можно из 32 монет найти фальшивую за 9 тестирований. (На удобном языке: Д $_{3,1}(32) \leqslant 9$.) b) Тремя весами, из которых одни сломаны, можно из 81 монеты найти фальшивую за 7 взвешиваний. (На удобном языке: $\mathrm{B}_{3,1}(81) \leqslant 7$ ). ## Что точно нельзя? 2.4. Из двух монет нельзя найти фальшивую за 2 тестирования/взвешивания (любым числом любых устройств, среди которых есть сломанные). (На удобном языке: $Д_{x, 1}(2) \geqslant 3$, $\left.\mathrm{B}_{x, 1}(2) \geqslant 3\right)$. 2.5. a) Любым числом детекторов, среди которых есть сломанный, нельзя найти фальшивую монету из $2^{k}$ монет за $k$ тестирований. (На удобном языке: $Д_{x, 1}\left(2^{k}\right)>k$ ). b) Любым числом весов, среди которых есть сломанные, нельзя найти фальшивую монету из $3^{k}$ монет за $k$ взвешиваний. (На удобном языке: $\mathrm{B}_{x, 1}\left(3^{k}\right)>k$.) 2.6. a) Любым числом детекторов, среди которых есть сломанные, нельзя найти фальшивую монету из $2^{k}$ монет за $k+1$ тестирование. (На удобном языке: $Д_{x, 1}\left(2^{k}\right)>k+1$ ). b) Любым числом весов, среди которых есть сломанные, нельзя найти фальшивую монету из $3^{k}$ монет за $k+1$ взвешивание. (На удобном языке: $\mathrm{B}_{x, 1}\left(3^{k}\right)>k+1$.) c) Любым числом весов, среди которых есть двое сломанных, нельзя найти фальшивую монету из $n>3^{6}$ монет за 11 взвешиваний. (На удобном языке: $\mathrm{B}_{x, 2}(n)>11$, если $n>3^{6}$.) ## 3 Грубые, но серийные результаты ## Серийные оценки сверху В этом разделе $k$ - натуральное число. 3.1. a) Для каждого $k$ найдите минимальное такое $K$, что $Д_{K, k}(n)<\infty$ при любом $n$ (то есть, найдите минимальное число детекторов, при котором вообще возможно найти фальшивую монету). b) Для каждого $k$ найдите минимальное такое $K$, что $\mathrm{B}_{K, k}(n)<\infty$ при любом $n$. 3.2. a) Докажите, что Д $_{3,1}\left(2^{k}\right) \leqslant 2 k+1$. b) Докажите, что $\mathrm{B}_{3,1}\left(3^{k}\right) \leqslant 2 k+1$. Как видно из предыдущей задачи, при наличии одного сломанного устройства из $n$ монет фальшивая ищется примерно за $2 \log _{2} n$ тестирований детекторами, и примерно за $2 \log _{3} n$ взвешиваний весами. Но эта оценка очень неточна. Задачи этого раздела будут посвящена уменьшению константы при логарифме, то есть числа $c$ в оценках вида $Д_{x, 1}(n) \lesssim c \log _{2} n$ и $\mathrm{B}_{x, 1}(n) \lesssim c \log _{3} n$. 3.3. a) Докажите, что $\mathrm{B}_{3,1}\left(3^{2 k}\right) \leqslant 3 k+1$. b) Докажите, что Д $_{3,1}\left(2^{2 k}\right) \leqslant 3 k+2$. Предыдущая задача показывает, что коэффициент $c$ можно уже сделать строго меньшим 2. Следующая цель - доказать, что он на самом деле равен 1. Сначала предлагается это сделать для бо́льшего числа устройств. Обозначение $o(k)$ означает функцию, растущую медленнее, чем $k$, то есть $f(k)=o(k)$, если $f(k) / k \rightarrow 0$ $(k \rightarrow \infty)$. Например, $\log _{2} k=o(k)$ и $\sqrt{k}=o(k)$. 3.4. a) Докажите, что при бесконечном числе детекторов, из которых один сломан, можно выявить одну фальшивую монету из $2^{k}$ за $k+o(k)$ взвешиваний. Иными словами, Д $_{\infty, 1}\left(2^{k}\right)=k+o(k)$. b) Докажите, что при бесконечном числе весов, из которых одни сломаны, можно выявить одну фальшивую монету из $3^{k}$ за $k+o(k)$ взвешиваний. Иными словами, $\mathrm{B}_{\infty, 1}\left(3^{k}\right)=$ $k+o(k)$. 3.5. a) Докажите, что существует $x$ такое, что Д ${ }_{x, 1}\left(2^{k}\right)=k+o(k)$. b) Докажите, что существует $x$ такое, что $\mathrm{B}_{x, 1}\left(3^{k}\right)=k+o(k)$. 3.6. a) Докажите, что $Д_{3,1}\left(2^{k(k+1)}\right) \leqslant(k+1)^{2}$ при $k \geqslant 5$. b) Докажите, что $\mathrm{B}_{3,1}\left(3^{k(k+1)}\right) \leqslant(k+1)^{2}$ при $k \geqslant 2$. ## Серийные оценки снизу 3.7. a) Докажите, что Д Д $_{x, 1}(n) \leqslant Д_{x, 1}\left(2^{k}\right)$ при $n<2^{k}$. b) Докажите, что $\mathrm{B}_{x, 1}(n) \leqslant \mathrm{B}_{x, 1}\left(3^{k}\right)$ при $n<3^{k}$. 3.8. a) Докажите, что $Д_{x, 1}(n) \leqslant Д_{x, 1}(N)$ при $n4$. b) Докажите, что $\mathrm{B}_{x, 1}(n)=\mathrm{B}_{4,1}(n)$ для любого $n$ и $x>4$. Если (при фиксированном $s$ ) для некоторого $t$ выполняются равенства $Д_{t, s}(n)=Д_{x, s}(n)$ $\left(\mathrm{B}_{t, s}(n)=\mathrm{B}_{x, s}(n)\right)$ при всех $x>t$, то такое число устройств мы будем называть идеальным (для данного числа сломанных устройств $s$ ). Иными словами, увеличивать количество устройств бессмысленно. Предыдущая задача показывает, что 4 устройства - идеальное число для одного сломанного. 4.6. a) Верна ли оценка того же вида, что и в задаче 4.4 , для $Д_{3,1}(n)$ ? b) Тот же вопрос про $\mathrm{B}_{3,1}(n)$. # Взвешивания со сломанными весами. Решения К. Кноп, Г. Челноков, И. Богданов ## 2 Вводные задачи: конкретные случаи ## Что точно можно? Мы будем для краткости называть фальшивую монету ФМ. Заметим, что если есть всего один сломанный тестер, и показания двух тестеров на каком-то наборе совпали, то эти показания истинны 2.1. Условие. Тремя весами, из которых одни сломаны, можно из 3 монет найти фальшивую за 3 взвешивания (На удобном языке: $B_{3,1}(3) \leqslant 3$ ). Решение. Взвесим первые две монеты на всех трех весах. Как минимум два результата совпадут - это и будет правильный результат этого взвешивания. По нему, очевидно, определяется ФМ. 2.2. a) Условие. Тремя детекторами, из которых один сломан, можно из 8 монет найти фальшивую за 6 тестирований. (На удобном языке: Д $_{3,1}(8) \leqslant 6$.) Решение. Разделим все монеты на 4 группы по две монеты. Проверим первым детектором группы 1 и 2 , а вторым - группы 1 и 3. Без ограничения общности можно считать, что оба ответа положительны (почему?). Тогда третьим детектором проверим группы 2 и 3. Возможны два случая. 1. Пусть ответ отрицателен; тогда ФМ в группе 1 (иначе хотя бы два детектора соврали!). Проверив одну монету из этой группы на всех трех детекторах, мы найдем ФМ. 2. Пусть ответ на третье испытание положителен. Тогда уже точно один из детекторов соврал (и если бы у нас имелся в наличии четвертый, то он точно бы был рабочим - но и без него все получается!). Проверим группы 1 и 2 - теперь на втором детекторе. Если ответ положителен, то ФМ точно в этих группах, так как это утверждают два детектора; значит, первый детектор на первом испытании говорил правду, а врал тогда либо второй, либо третий. Если же ответ отрицателен, то либо первый, либо второй детектор лжет, а значит, третий - рабочий, поэтому ФМ в группе 2 или 3. Итого, после 4 тестирования мы определили рабочий детектор и две группы, в которых находится ФМ. Тогда ее легко найти за 2 оставшихся испытания. b) Условие. Тремя весами, из которых одни сломаны, можно из 9 монет найти фальшивую за 4 взвешивания. (На удобном языке: $B_{3,1}(9) \leqslant 4$ ). Решение. Для удобства занумеруем монеты числами от 0 до 8 и запишем их в троичной записи (таким образом, каждой монете сопоставлена пара цифр от 0 до 2 ). Первое взвешивание делаем первыми весами в соответствии с первой цифрой номера: на левую чашу кладем монеты, у номера которых первая цифра 0 , на правую - у которых она 1 . Второе взвешивание делаем аналогично вторыми весами в соответствии со второй цифрой номера. Без ограничения общности можем считать, что оба раза весы говорили, что фальшивая монета на левой чаше. Тогда 4 монеты без нулей в номере точно не фальшивые (иначе соврали и первые, и вторые весы). Теперь разобьем монеты на три группы следующим образом: в одну поместим 00 , в другую 01 и 02 , в третью - 10 и 20 ; дополним все группы до трех монет точно не фальшивыми и взвесим третьими весами. Если весы сказали, что фальшивая в группе с 00 , то это и есть 00 (иначе двое весов соврали), и четвертое взвешивание не понадобилось. Если взвешивание сказало, что фальшивая в группе с 01 и 02 , то третьи весы противоречат вторым. Поэтому хотя бы одни из них соврали, а значит, первые точно исправны. Но тогда у фальшивой первая цифра действительно 0; таким образом, остались лишь три кандидата на ФМ и одни точно исправные весы, которыми мы находим ФМ за 1 ход. Аналогично поступаем, если третье взвешивание сказало, что фальшивая в группе с 10 и 20. 2.3. a) Условие. Тремя детекторами, из которых один сломан, можно из 32 монет найти фальшивую за 9 тестирований. (На удобном языке: ДЗ, $(32) \leqslant 9$.) Решение. Разобьем все монеты на 4 группы по 8 монет в каждой и применим тот же алгоритм, что и в задаче 2.2a). Тогда в случае 1 мы за 3 тестирования найдем группу из 8 монет, которая содержит фальшивую; применяя ту же задачу, мы сделаем требуемое. Во втором же случае мы за 4 тестирования найдем две группы, одна из которых содержит фальшивую, а также найдем один рабочий детектор. С помощью него мы еще за 4 испытания найдем ФМ - в этом случае мы даже обошлись 8 тестированиями. b) Условие. Тремя весами, из которых одни сломаны, можно из 81 монеты найти фальшивую за 7 взвешиваний. (На удобном языке: $B_{3,1}(81) \leqslant 7$ ). Решение. Аналогично $2.2 \mathrm{~b}$ ), занумеруем монеты последовательностями из 4 цифр 0,1 или 2 , и проведем первое взвешивание по первой цифре на первых весах, а второе по второй цифре - на вторых. Без ограничения общности можем считать, что оба раза весы говорили, что фальшивая монета имеет цифру 0 в соответствующем разряде. Отсюда опять же следует, что монеты, у которых первая и вторая цифра точно не нули настоящие. Теперь разобьем монеты на три группы следующим образом: в одну поместим 9 монет, у которых первые две цифры - 00 , в другую - 18 монет с первыми двумя цифрами 01 или 02 , а в третью - с цифрами 10 и 20 ; дополним все группы до 27 монет точно не фальшивыми и взвесим третьими весами. Если весы сказали, что фальшивая в группе с 00, то это действительно так, и у нас осталось 9 монет, среди которых есть фальшивая, и 4 взвешивания. Это сводит задачу к $2.2 \mathrm{~b}$ ). Если взвешивание сказало, что фальшивая в группе с 01 и 02 , то третьи весы противоречат вторым. Значит, хотя бы одни из них соврали, поэтому первые точно исправны. Но тогда у ФМ первая цифра действительно 0 . Теперь мы имеем 27 монет, среди которых есть фальшивая, и одни точно исправные весы; с помощью них мы находим фальшивую за 3 хода (а седьмой даже не понадобился). Аналогично мы поступаем, если третье взвешивание сказало, что фальшивая в группе с 10 и 20. ## Что точно нельзя? 2.4. Условие. Из двух монет нельзя найти фальшивую за 2 тестирования/взвешивания (любым числом любых устройств, среди которых есть сломанные). (На удобном языке: Д $x_{x, 1}(2) \geqslant 3, B_{x, 1}(2) \geqslant 3$ ). Решение. Если оба тестирования мы делаем одним устройством, то оно может оказаться сломанным, и мы не получим никакой информации. Если мы использовали разные устройства, то в случае противоречащих ответов мы не можем определить фальшивую. 2.5. a) Условие. Любым числом детекторов, среди которых есть сломанный, нельзя найти фальшивую монету из $2^{k}$ монет за $k$ тестирований. (На удобном языке: Д Решение. Предположим, что это возможно. Пусть некоторые тестирования уже сделаны. Рассмотрим все монеты, которые сейчас могут оказаться фальшивыми. Заметим, что при некотором исходе следующего взвешивания их количество уменьшится не более, чем вдвое (ибо каждый кандидат на фальшивость хотя бы при одном из исходов следующего взвешивания продолжит оставаться таким кандидатом). На первом ходу количество кандидатов в фальшивые не уменьшалось (ибо первый детектор мог соврать). Поэтому за $k$ шагов при некотором наборе исходов оно уменьшилось не более, чем в $2^{k-1}$ раз, то есть осталось больше 1. Таким образом, при этом наборе исходов мы явно не смогли однозначно определить фальшивую монету. b) Условие. Любым числом весов, среди которых есть сломанные, нельзя найти фальшивую монету из $3^{k}$ монет за $k$ взвешиваний. (На удобном языке: $B_{x, 1}\left(3^{k}\right)>k$.) Решение. Аналогично, только количество кандидатов делится на 3 , а не на 2. 2.6. a) Условие. Любым числом детекторов, среди которых есть сломанные, нельзя найти фальшивую монету из $2^{k}$ монет за $k+1$ тестирование. (На удобном языке: $Д_{x, 1}\left(2^{k}\right)>k+1$ ). Решение. Будем говорить, что тестирование отказало данной монете, если оно показало, что монета находится в другой группе, нежели фальшивая. Иными словами, если тестирование отказало монете, то или монета не фальшивая, или детектор сломан. Аналогично доказанному в предыдущей задаче, если бы за $k+1$ тестирование можно было определить фальшивую, то после второго тестирования монет, которые могут оказаться фальшивыми, должно быть не более $2^{k-1}$. Но любая монета, которой не отказало хотя бы одно тестирование из первых двух, может оказаться фальшивой (если другое тестирование делалось на сломанном детекторе). Первое тестирование при одном из исходов отказывает не более, чем $\frac{1}{2}$ всех монет; таким образом, в этом случае первое тестирование не отказывает хотя бы половине всех монет. Второе же при некотором исходе не откажет еще какой-нибудь монете. Значит, строго больше половины монет могут оказаться фальшивыми противоречие. b) Условие. Любым числом весов, среди которых есть сломанные, нельзя найти фальшивую монету из $3^{k}$ монет за $k+1$ взвешивание. (На удобном языке: $B_{x, 1}\left(3^{k}\right)>k+1$. Решение. Аналогично: получаем, что первое взвешивание не откажет хотя бы трети всех монет, а второе не откажет еще какой-нибудь монете. Поэтому после первых двух взвешиваний кандидатов будет больше, чем треть всех монет. c) Условие. Любым числом весов, среди которых есть двое сломанных, нельзя найти фальшивую монету из $n>3^{6}$ монет за 11 взвешиваний. (На удобном языке: $B_{x, 2}(n)>11$, если $n>3^{6}$.) Решение. Решение этого пункта находится в конце раздела 3. ## 3 Грубые, но серийные результаты ## Серийные оценки сверху 3.1. Условие. а) Для каждого $k$ найдите минимальное такое $K$, что Д $_{K, k}(n)<\infty$ при любом $n$ (то есть, найдите минимальное число детекторов, при котором вообще возможно найти фальшивую монету). b) Тот же вопрос для весов. Решение. Решение не зависит от типа тестеров. Ответ. $K=2 k+1$. Пусть число тестеров $\leqslant 2 k$, а фальшивая монета $-A$. Заставим все сломанные тестеры действовать так, как будто фальшивая монета - $B$. Тогда никогда не удастся выяснить, то ли все сломанные сломаны (и фальшивая - $A$ ), то ли все наоборот (и фальшивая - $B$ ). Если же тестеров $2 k+1$, то можно каждую монету (или каждую пару для весов) проверить на всех тестерах; $k+1$ из результатов совпадут - это и будут истинные результаты. 3.2. a) Условие. Докажите, что Д $_{3,1}\left(2^{k}\right) \leqslant 2 k+1$. Решение. Индукция по $k$. При $k=1$ проверим первую монету всеми тремя детекторами; два совпавших ответа будут истинными. Пусть при $k=t$ утверждение доказано. Рассмотрим $2^{t+1}$ монет и проверим первые $2^{t}$ из них на первых двух тестерах. Если результаты совпадают, то они истинны, и мы сократили число монет до $2^{t}$; применяя предположение индукции, получаем требуемое. Если же результаты различны, то один из первых двух детекторов сломан; тогда третьим (заведомо рабочим) детектором мы за $t+1$ испытание найдем $\Phi М ;$ в этом случае мы использовали не более $t+3 \leqslant 2(t+1)+1$ испытаний. b) Условие. Докажите, что $B_{3,1}\left(3^{k}\right) \leqslant 2 k+1$. Решение. Совершенно аналогично. 3.3. а) Условие. Докажите, что $B_{3,1}\left(3^{2 k}\right) \leqslant 3 k+1$. Решение. Опять индукция; база при $k=1$ - это задача $2.2 \mathrm{~b}$ ). Переход абсолютно аналогичен решению 2.3b). b) Условие. Докажите, что Д $_{3,1}\left(2^{2 k}\right) \leqslant 3 k+2$. Решение. Аналогично, индукция; переход аналогичен решению 2.3a). База доказывается из тех же соображений; именно, мы либо за 3 испытания находим $Ф М$, либо за 4 детекции находим пару монет, среди которых есть фальшивая, и рабочий детектор. Тогда за оставшееся испытание мы найдем ФМ. 3.4. а) Условие. Докажите, что при бесконечном числе детекторов, из которых один сломан, можно выявить одну фальшивую монету из $2^{k}$ за $k+o(k)$ взвешиваний. Иными словами, $Д_{\infty, 1}\left(2^{k}\right)=k+o(k)$. Решение. Занумеруем все монеты $k$-значными двоичными числами (от $0 \ldots 0$ до $1 \ldots 1$ ). Первыми $k$ взвешиваниями применим первые $k$ детекторов: каждый - к своему двоичному разряду. Без ограничения общности, пусть каждый из них ответил, что у фальшивой монеты в соответствующем разряде стоит 0 . Значит, в любом случае единиц в номере $\Phi М$ не больше одной. Применим ( $k+1$ )-й детектор к монете $0 \ldots 0$. Если он говорит, что она фальшивая, то так оно и есть (иначе два детектора солгали!). Иначе мы понимаем, что какой-то детектор из первых $k+1$ солгал, и при этом остался $k+1$ кандидат на фальшивость. Из этих кандидатов можно найти ФМ ( $k+2$ )-м (рабочим!) детектором за $\leqslant \log _{2}(k+1)+1$ испытаний. Итого мы справились не более, чем за $k+1+\log _{2}(k+1)+1=k+o(k)$ испытаний. b) Условие. Докажите, что при бесконечном числе весов, из которых одни сломаны, можно выявить одну фальшивую монету из $3^{k}$ за $k+o(k)$ взвешиваний. Иными словами, $B_{\infty, 1}\left(2^{k}\right)=k+o(k)$. Решение. Абсолютно аналогично. 3.5. Условие. a) Докажите, что существует $x$ такое, что $Д_{x, 1}\left(2^{k}\right)=k+o(k)$. b) Докажите, что существует $x$ такое, что $B_{x, 1}\left(3^{k}\right)=k+o(k)$. Решение. Следует из следующей задачи. 3.6. b) Условие. Докажите, что $B_{3,1}\left(3^{k(k+1)}\right) \leqslant(k+1)^{2}$ при $k \geqslant 2$. Решение. Решим задачу индукцией по $k$, база для $k=1$ доказана в 2.2.b. Как и раньше, нумеруем монеты в троичной системе счисления. Первые $k$ взвешиваний делаем по первым $k$ разрядам на первых весах, вторые $k$ взвешиваний - по следующим $k$ разрядам на вторых весах. Без ограничения общности можно считать, что все взвешивания показывали, что у фальшивой монеты в соответствующем разряде стоит 0 . Теперь, если у номера ФМ есть не ноль как в первых $k$ разрядах, так и во вторых, то и первые, и вторые весы соврали. Значит, все такие монеты точно настоящие (будем их называть эталонами). Для $(2 k+1)$-го взвешивания разделим все неэталоны так: в одну группу поместим монеты, у которых в первых $2 k$ разрядах только нули, во вторую - монеты, у которых нули в первых $k$ разрядах, но есть не нули в разрядах с $(k+1)$-го по $2 k$-й, а в третью - монеты, у которых есть не нули в первых $k$ разрядах, но в разрядах с $(k+1)$-го по $2 k$-й все цифры - нули. Ясно, что во второй и третьей группах монет поровну; взвесим их третьими весами. Если взвешивание указало на первую группу, то все результаты взвешиваний верны (иначе хотя бы двое весов давали неверный результат); значит, нам осталось за $k^{2}$ взвешиваний узнать $k(k-1)$ разрядов, что возможно по предположению индукции. Если $(2 k+1)$-е взвешивание указало на вторую группу (для третьей аналогично), то показания третьих весов противоречат показаниям вторых, значит, первые весы точно исправны. Тогда мы уже выяснили первые $k$ разрядов, и за оставшиеся $k^{2}$ взвешиваний исправными весами выясним оставшиеся. a) Условие. Докажите, что Д $_{3,1}\left(2^{k(k+1)}\right) \leqslant(k+1)^{2}$ при $k \geqslant 5$. Решение. Аналогично пункту b) можно получить чуть более слабую оценку $Д_{3,1}\left(2^{k(k-1) / 2-1}\right) \leqslant$ $\leqslant k(k+1) / 2-1$. Похоже, что более сильная, к сожалению, следует лишь из 4.4 . ## Серийные оценки снизу 3.7. a) Условие. Докажите, что $Д_{x, 1}(n) \leqslant Д_{x, 1}\left(2^{k}\right)$ при $n<2^{k}$. Решение. См. решение $3.8 \mathrm{a})$. b) Условие. Докажите, что $B_{x, 1}(n) \leqslant B_{x, 1}\left(3^{k}\right)$ при $n<3^{k}$. Решение. Жюри не известно простого доказательства этого факта. 3.8. а) Условие. Докажите, что $Д_{x, 1}(n) \leqslant Д_{x, 1}(N)$ при $n3^{6}$ монет за 11 взвешиваний. (На удобном языке: $B_{x, 2}(n)>11$, если $n>3^{6}$.) Решение. Рассмотрим произвольный алгоритм, позволяющий найти ФМ из $n$ монет за 11 ходов. Пусть мы каким-то образом его провели; запишем результаты испытаний в строчку. У нас получилась строка из 11 знаков «<», «=» и «>». Как и в решении $3.9 \mathrm{a}$ ), каждой такой строчке однозначно соответствует ФМ. Теперь выясним, сколько таких строчек соответствует одному варианту ФМ. Их хотя бы $1+2 \cdot 11+4 \cdot 55=243$; действительно, есть как минимум такие строчки: 0) в которой все ответы правильные; 1) в которых ровно один ответ неверен (таких строчек 22: можно выбрать место, в котором стоит неверный ответ, 11 способами, и в каждом из них может стоять один из двух возможных неверных ответов); 2) в которых ровно два ответа неверны (таких $4 \cdot 55$, ибо есть 55 пар мест и 4 варианта пар неверных ответов). Нетрудно понять, что все описанные строчки различны. Таким образом, строк не меньше, чем $243 n$; с другой стороны, всех возможных строк не больше, чем $3^{11}$. Таким образом, $n \cdot 243 \leqslant 3^{11}$, откуда $n \leqslant 3^{6}$. ## 4 Точные результаты 4.1. Условие. а) Докажите, что Д $_{4,1}\left(2^{4}\right)=7$. b) Докажите, что $B_{4,1}\left(3^{6}\right)=9$ Решение. См. решение 4.4. Хотя это можно сделать и без применения столь общих приемов. 4.2. Условие. a) Среди какого наибольшего числа монет можно выявить фальшивую на 4д[1] за 15 тестирований? (Иначе говоря, найдите наибольшее $n$ такое, что $Д_{4,1}(n) \leqslant 15$.) b) Среди какого наибольшего числа монет можно выявить фальшивую на 4в[1] за 13 тестирований? (Иначе говоря, найдите наибольшее $n$ такое, что $B_{4,1}(n) \leqslant 13$.) c) А за 40 тестирований? Решение. Ответы. а) $2^{11}$; b) $3^{10}$; c) $3^{36}$. Верхние оценки следуют из задачи 3.9. Доказательство того, что эти оценки достигаются, вытекает из решения 4.4. Тем не менее, при желании эти алгоритмы можно найти ручными методами. 4.3. Условие. Приведите пример такого $n$, что $B_{4,1}(n)b_{1}$, то уменьшим $b_{2}$, иначе можно уменьшить $b_{1}$ ). Итого, можно считать, что $a_{1}+a_{2}+a_{3}=b_{1}+b_{2}$ Выстроим всех школьников в ряд: сначала все из первого класса, затем - из третьего, и в конце - из второго. Затем рассадим первых $b_{1}$ из них в первый автобус, затем следующих $b_{2}$ во второй. Тогда все первоклассники окажутся в первом автобусе (непосредственно из условия), а все второклассники окажутся во втором автобусе (в противном случае $b_{2}a_{1}+a_{3}>a_{3} \geqslant b_{2}$, что неверно). Обратно, если $b_{1}13$. Предположим противное; это может случиться, только если значимость на каждом шагу уменьшается ровно втрое (при любом исходе взвешивания!). А это, в свою очередь, значит, что в любой момент существуют отказники не более, чем двух типов - а именно не того типа, на котором производится очередное взвешивание. Рассмотрим, как это может происходить. Будем обозначать ситуацию тройкой чисел $(a, b, c)$, где $a$ - количество подозреваемых, а $b$ и $c$ - количества отказников присутствующих двух типов. Тогда исходная ситуация есть $\left(3^{10}, 0,0\right)$, после первого взвешивания - $\left(3^{9}, 2 \cdot 3^{9}, 0\right)$. Пусть во втором взвешивании количества подозреваемых в группах равны $a, b, c$. Пусть это взвешивание показало на первую группу, тогда мы пришли к ситуации $\left(a, 3^{9}-a, t\right)$, где число $t$ вычисляется из суммарной значимости: $23 a+\left(3^{9}-a\right)+t=3^{11}$, откуда $t=8 \cdot 3^{9}-22 a$. Заметим, что $t \geqslant 0$, то есть $a \leqslant \frac{4}{11} 3^{9}$. Аналогично $b, c \leqslant \frac{4}{11} 3^{9}$, поэтому $a=3^{9}-(b+c) \geqslant \frac{3}{11} 3^{9}$. Посмотрим на третье взвешивание. Если все $a$ подозреваемых лежат в одной группе, то значимость этой группы будет не меньше $21 \cdot \frac{3}{11} 3^{9}>3^{10}$, что невозможно. Значит, подозреваемые попадут хотя бы в две группы. Тогда, чтобы после каждого возможного исхода оставались отказники лишь двух типов, необходимо, чтобы в каждой группе лежали бы отказники только одного типа. Значит, в какой-то группе будут лежать все отказники одного типа - пусть их количество равно $s$. Если на ней лежат еще $x$ подозреваемых, то ее значимость после взвешивания равна $3^{10}=21 x+(a-x)+s=s+a+20 x$; таким образом, $3^{10}-s-a$ должно делиться на 20. При этом $s=3^{9}-a$ либо $s=t=8 \cdot 3^{9}-22 a$. В первом случае получаем, что $3^{10}-s-a=2 \cdot 3^{9}$ не делится на 20 , что невозможно. Значит, выполнен второй случай, а тогда $3^{10}-s-a=21 a-5 \cdot 3^{9}$. Так как это число делится на 20 , то $a$ должно делиться на 5 . Аналогично, $b$ и $c$ также должны делиться на 5 ; это противоречит тому, что $a+b+c=3^{9}$, которое на 5 не делится. 4.5. а) Условие. Докажите, что $Д_{x, 1}(n)=Д_{4,1}(n)$ для любого $n$ и $x>4$. Решение. Пусть мы должны найти фальшивую из $n$ монет. Заметим, что если $t$ таково, что $2^{t}<$ $4$. Решение. Аналогично; нужно вдобавок добиться, чтобы в двух из получающихся на очередном шаге групп будет поровну элементов. Это можно сделать добавлением эталонов. Замечание. Теми же методами (хотя и с бо́льшим числом технических подробностей) можно показать, что идеальное число тестеров существует для любого числа сломанных тестеров. ## 4.6. а) Условие. Верна ли оценка того же вида, что и в задаче 4.4, для $Д_{3,1}(n)$ ? Решение. Пусть мы хотим (при достаточно большом $n$ ) из $n$ монет найти фальшивую тремя детекторами, из которых один сломан. Рассмотрим минимальное $d$, такое что $2^{d}+d+1 \geqslant n(2 d+1)$. Докажем, что нам хватит $d$ ходов для того, чтобы осталось не более двух монет, которые могут оказаться фальшивыми; это и будет момент, относительно которого мы определяем значимость. Тогда исходная значимость равна $n(2 d+1)$. Обозначим $x_{i}=2^{i}+i+1$. Построим процесс так, чтобы за $i$ взвешиваний до конца значимость не превосходила $x_{i}$, а отказники всегда были не более, чем двух разных типов. Заметим, что если мы этого добились, то после $d$-го хода значимость не превосходит 2 , значит, не более чем две монеты являются подозреваемыми или отказниками. Построим процесс по индукции. До первого взвешивания оба условия выполнены. Пусть осталось построить $i$ последних взвешиваний, покажем, как сделать очередное. Здесь вместо леммы об автобусах используется следующая (очевидная) Лемма о разрезании. Пусть в ряд выстроены несколько предметов, стоимость каждого не превосходит $x$, а общая их стоимость равна $S$. Тогда ряд можно разбить на $t$ частей так, чтобы стоимость каждой части не превосходила $\frac{S+(t-1) x}{t}$. Пусть, без ограничения общности, присутствуют отказники первого и второго типа. Расставим монеты в ряд: сначала отказников первого типа, затем подозреваемых, и в конце отказников второго типа. Обозначим количество подозреваемых через $a$. Посчитаем отказников с коэффициентом 1 , а подозреваемых $-i-1$. Тогда сумма коэффициентов не превосходит $x_{i}-2 a$; значит, по лемме о разрезании этот ряд можно разрезать на две части так, чтобы сумма коэффициентов в каждой части была $\leqslant \frac{x_{i}-2 a+i-1}{2}$. Теперь, если протестировать детектором монеты любой из частей разрезания, то значимость полученной части будет равна сумме ее коэффициентов, увеличенной на количество подозреваемых (ибо подозреваемых в этой части мы должны считать с коэффициентом $i$ вместо $i-1$, а в другой - с коэффициентом 1 вместо 0 ). Таким образом, при любом исходе значимость не превысит $\frac{x_{i}+i-1}{2}=x_{i-1}$. При этом, что если в какую-то часть разбиения попали отказники обоих типов, то туда же попали все подозреваемые; поэтому при любом исходе проверки остались отказники не больше, чем двух типов. Итак, через $d$ взвешиваний у нас осталось не более двух кандидатов на ФМ. Осталось заметить, что из двух монет одну фальшивую можно найти за ограниченное число ходов (например, за 3). Таким образом, доказано, что $d+3$ тестирований хватит. Осталось заметить, что $d=\log _{2}(n(2 d+1)-d-1)$, откуда $d=$ $=\log _{2} n+\log _{2} \log _{2} n+O(1)$, как и требовалось. b) Условие. Тот же вопрос про $B_{3,1}(n)$. Решение. Аналогично, мы собираемся (при достаточно больших $n$ ) из $n$ монет оставить только 3 кандидатов в фальшивые за $d$ взвешиваний. Мы утверждаем, что достаточно выбрать $d$ таким, что $$ (2 d+1)+2 \cdot 3^{d} \geqslant n(2 d+1) $$ Тогда (по задаче 2.1) мы найдем ФМ за $d+3$ взвешивания; при этом $d=\log _{3} n+\log _{3} \log _{3} n+O(1)$. Опять же, моментом, относительно которого отсчитывается значимость, мы называем момент после $d$-го взвешивания; тогда исходная значимость равна $n(2 d+1)$. Обозначим $y_{i}=(2 i+1)+2 \cdot 3^{i}$; мы будем строим процесс так, чтобы за $i$ взвешиваний до конца значимость не превосходила $y_{i}$, а отказники всегда были не более, чем двух разных типов; тогда после $d$-го взвешивания значимость не превосходит 3 , что и требовалось. Шаг алгоритма строится опять же аналогично. Присвоим всем подозреваемым коэффициент $D=2 i-2$, а отказникам коэффициент 1 ; тогда сумма всех коэффициентов не превосходит $y_{i}-3 a$, где $a-$ количество подозреваемых. Разложим все монеты в ряд: сначала отказников первого типа, затем подозреваемых, и в конце отказников второго типа. Тогда по лемме о разрезании, из можно разрезать на три части так, что сумма коэффициентов в каждой части не больше $\frac{y_{i}-3 a+2 D}{3}$. Это значит, что значимость любой ситуации после взвешивания с такими тремя частями не превосходит $\frac{y_{i}-3 a+2 D}{3}+a=y_{i-1}$, что и требовалось. Здесь может возникнуть еще одна проблема: количества монет в трех полученных группах могут различаться. Покажем, что нам хватит эталонов для того, чтобы сделать две группы равными; тогда мы можем соорудить требуемое взвешивание. Легко видеть, что на первых двух шагах таких проблем не возникает: в первом взвешивании две из трех групп имеют равное количество монет, а во втором взвешивании легко распределить отказников и подозреваемых по трем группам так, что в двух из них монет будет поровну (немного изменив алгоритм разбиения, но не ухудшив оценку!). После этих взвешиваний появится хотя бы $\frac{4}{9} n-2$ эталонов. Тогда мы раскладывали по трем группам не более, чем $\frac{5}{9} n+2$ монет, и значит, разность количеств монет в двух из этих групп будет не больше, чем $\frac{5}{27} n+1 \leqslant \frac{4}{9} n-2$ при $n \geqslant 12$. Тогда эталонов хватит. Решение окончено. # Weighings using broken balances K. Knop, G. Chelnokov, I. Bogdanov ## 1 Introduction ## Basic notions As in the usual weighing problems, we have a set of coins indistinguishable by sight, though one of them is false. The weight of a false coin is a bit smaller than the weight of a genuine one, ans we want to find this false coin. In different problems we will use the different types of testers. A detector is a tester, which at one turn (or testing) tells whether a given subset of coins contains a false one or not. Thus, for any subdivision of the total set into two subsets, a detector shows which one of them contains a false coin. A bi-scaled balance (or simply balance) is a tester comparing the weights of two subsets. Thus, for two subsets with equal numbers of coins, the balance compares the numbers of false coins in these subsets. In our problem, the balance shows which of three subsets in a subdivision contains a false coin (two of three subsets should have the same cardinality). The zest of this problem set is that some testers can report a wrong information. Actually, such a tester does not behave as a "liar": it does not necessarily tell the wrong messages each time. It is simply broken, so its responses may appear wrong and correct from time to time; they have no relation to the reality, and we can consider this balance as a generator of random answers. So, we have several testers, and we know only the number of broken testers among them; but we do not know, which ones are broken. (Note that at each testing, only one tester is used!) Let us introduce the notation. We denote by $\mathrm{D}_{x, y}(z)$ the minimal number of testings by detectors which are necessary to find one false coin from $n$ using $x$ detectors with $y$ broken ones among them. (We should be able to specify the false coin for each possible sequence of detector responses.) Analogously, by $\mathrm{B}_{x, y}(z)$ we denote the same number for the testing system consisting of $x$ balances with $y$ broken ones among them. The systems of $x$ detectors (balances) with $y$ broken ones will be referred to as $x \mathrm{~d}[y](x \mathrm{~b}[y])$. Throughout the first sections, we will write down two formulations of the problem: one in an usual language, and another in our notation. This is to get all the readers acquainted with the notation. ## 2 Introductory problems: some particular cases ## What are we able to achieve? 2.1. Using three balances with one broken among them, we can find a false coin of 3 ones in 3 weighings. (Using the notation: $\mathrm{B}_{3,1}(3) \leq 3$.) 2.2. a) Using three detectors with one broken among them, we can find a false coin of 8 ones in 6 detections. (Using the notation: $\mathrm{D}_{3,1}(8) \leq 6$.) b) Using three balances with one broken among them, we can find a false coin of 9 ones in 4 weighings. (Using the notation: $\mathrm{B}_{3,1}(9) \leq 4$.) 2.3. a) Using three detectors with one broken among them, we can find a false coin of 32 ones in 9 detections. (Using the notation: $\mathrm{D}_{3,1}(32) \leq 9$.) b) Using three balances with one broken among them, we can find a false coin of 81 ones in 7 weighings. (Using the notation: $\mathrm{B}_{3,1}(81) \leq 7$.) ## What are we unable? 2.4. If we have two coins with one of them being false, it is not possible to find the false one in 2 testings using any number of testers with at least one broken among them. (Using the notation: $\mathrm{D}_{x, 1}(2) \geq 3, \mathrm{~B}_{x, 1}(2) \geq 3$.) 2.5. a) It is not possible to find a false coin from $2^{k}$ ones in $k$ detections by any number of detectors with one broken among them. (Using the notation: $\mathrm{D}_{x, 1}\left(2^{k}\right)>k$ ). b) It is not possible to find a false coin from $3^{k}$ ones in $k$ weighings by any number of balances with one broken among them. (Using the notation: $\mathrm{B}_{x, 1}\left(3^{k}\right)>k$ ). 2.6. a) It is not possible to find a false coin from $2^{k}$ ones in $k+1$ detections by any number of detectors with one broken among them. (Using the notation: $\mathrm{D}_{x, 1}\left(2^{k}\right)>k+1$ ). b) It is not possible to find a false coin from $3^{k}$ ones in $k+1$ weighings by any number of balances with one broken among them. (Using the notation: $\mathrm{B}_{x, 1}\left(3^{k}\right)>k+1$ ). c) It is not possible to find a false coin from $n>3^{6}$ ones in 11 weighings by any number of balances with two broken ones among them. (Using the notation: $\mathrm{B}_{x, 2}(n)>11$, if $n>3^{6}$ ). ## 3 Rough but serial results ## Upper bounds In this section, $k$ is always a positive integer. 3.1. a) For each $k$ find the minimal value of $K$ such that $\mathrm{D}_{K, k}(n)<\infty$ for every $n$ (in other words, find the least number of detectors with $k$ broken among them such that it is possible to find the false coin using them). b) For each $k$ find the minimal value of $K$ such that $\mathrm{B}_{K, k}(n)<\infty$ for every $n$. ## 3.2. a) Prove that $\mathrm{D}_{3,1}\left(2^{k}\right) \leq 2 k+1$. b) Prove that $\mathrm{B}_{3,1}\left(3^{k}\right) \leq 2 k+1$. It follows from the previous problem that, using some balances with one broken, one can find a false coin from $n$ in approximately $2 \log _{2} n$ detections, or in approximately $2 \log _{3} n$ weighings. But this estimate is not sharp. The aim of this section is to find the correct constant in the estimates of the form $\mathrm{D}_{x, 1}(n) \lesssim c \log _{2} n$ and $\mathrm{B}_{x, 1}(n) \lesssim c \log _{3} n$. ## 3.3. a) Prove that $\mathrm{B}_{3,1}\left(3^{2 k}\right) \leq 3 k+1$. b) Prove that $\mathrm{D}_{3,1}\left(2^{2 k}\right) \leq 3 k+2$. The previous problem shows that $c$ can be made less than 2 . Our next aim is to prove that $c=1$. It is easier to make this using more than 3 testers. We write $f(k)=o(k)$ for the function growing slower than $k$, that is, $f(k) / k \rightarrow 0(k \rightarrow \infty)$. For instance, $\log _{2} k=o(k)$ and $\sqrt{k}=o(k)$. 3.4. a) Prove that having an infinite number of detectors with one broken, one can find a false coin from $2^{k}$ in $k+o(k)$ weighings; that is, $\mathrm{D}_{\infty, 1}\left(2^{k}\right)=k+o(k)$. b) Prove that having an infinite number of balances with one broken, one can find a false coin from $3^{k}$ in $k+o(k)$ weighings; that is, $\mathrm{B}_{\infty, 1}\left(2^{k}\right)=k+o(k)$. 3.5. $\quad$ a) Prove that there exists $x$ such that $\mathrm{D}_{x, 1}\left(2^{k}\right)=k+o(k)$. b) Prove that there exists $x$ such that $\mathrm{B}_{x, 1}\left(3^{k}\right)=k+o(k)$. 3.6. a) Prove that $\mathrm{D}_{3,1}\left(2^{k(k+1)}\right) \leq(k+1)^{2}$ for every $k \geq 5$. b) Prove that $\mathrm{B}_{3,1}\left(3^{k(k+1)}\right) \leq(k+1)^{2}$ for every $k \geq 2$. ## Lower bounds 3.7. a) Prove that $\mathrm{D}_{x, 1}(n) \leq \mathrm{D}_{x, 1}\left(2^{k}\right)$ if $n<2^{k}$. b) Prove that $\mathrm{B}_{x, 1}(n) \leq \mathrm{B}_{x, 1}\left(3^{k}\right)$ if $n<3^{k}$. 3.8. a) Prove that $\mathrm{D}_{x, 1}(n) \leq \mathrm{D}_{x, 1}(N)$ if $n4$. b) Prove that $\mathrm{B}_{x, 1}(n)=\mathrm{B}_{4,1}(n)$ for every $n$ and $x>4$. Fix some value of $s$. If there exists $t$ such that $\mathrm{D}_{t, s}(n)=\mathrm{D}_{x, s}(n)\left(\mathrm{B}_{t, s}(n)=\mathrm{B}_{x, s}(n)\right)$ for all $x>t$, then we say that $t$ is an ideal number of testers (for he given number $s$ of broken ones). In other words, it is senseless to increase the number of testers. The previous problem states that 4 testers is an ideal number for 1 broken tester. 4.6. a) Find whether the estimate of the same form as in problem 4.4 is valid for $\mathrm{D}_{3,1}(n)$. b) The same question for $\mathrm{B}_{3,1}(n)$. # Weighings using broken balances. Solutions K. Knop, G. Chelnokov, I. Bogdanov ## 2 Introductory problems: some particular cases ## What are we able to achieve? For convenience, we will denote a false coin as FC. Note that in the situation with only one broken tester, if the readings of two testers on identical questions coincide, then this reading is true since one of the testers is operable. 2.1. Statement. Using three balances with one broken among them, we can find a false coin of 3 ones in 3 weighings. (Using the notation: $B_{3,1}(3) \leqslant 3$.) Solution. Compare two coins consecutively on all three balances. At least two results will coincide - these ones will be correct. From this result, one can easily find a FC. 2.2. a) Statement. Using three detectors with one broken among them, we can find a false coin of 8 ones in 6 detections. (Using the notation: $D_{3,1}(8) \leqslant 6$.) Solution. Split the coins into 4 groups with 2 coins in each. Apply the first detector to groups 1 and 2, and the second one to groups 1 and 3. Without loss of generality, we may assume that both answers are affirmative (why?). Then apply the third detector to groups 2 and 3. Two cases may occur. 1. If the answer is negative, then FC is in group 1 (otherwise two balances made wrong statements). Then we check one coin from this group by all three detectors; these results show which coin from group 1 is false. 2. Suppose that the third answer is affirmative. Then one of detectors lied (note that if we had the fourth detector, then it would be operable, and the remaining would be easy; nevertheless, we can also proceed without it!). Check groups 1 and 2 - now by the second detector. If the result is affirmative (so it coincides with the result of the first detector), then it is true, and FC is in group 1 or 2 . Moreover, the first balance told the truth on first use, hence either second or third balance lied. Otherwise, if the result of the third detection is negative, then either first or second balance is broken; hence the third one is operable, and FC is in group 2 or 3. So, in each case we get one operable balance and 4 candidates to be FC. It is now easy to find FC in 2 detections. b) Statement. Using three balances with one broken among them, we can find a false coin of 9 ones in 4 weighings. (Using the notation: $B_{3,1}(9) \leqslant 4$.) Solution. For the convenience, we enumerate the coins by numbers from 0 to 8 , and write these numbers in a ternary notation (so, each coin corresponds to an ordered pair of digits from 0 to 2 ). In the first weighing, we apply the first balance to the groups constructed by the leftmost digit: the left hand contains the coins with 0 at this digit, and the right hand contains those with 1. Analogously, the second weighing (on the second balance) compares the coins with 0 in the last digit with those with 1 . Without loss of generality, we may assume that both weighings told that a false coin has 0 in a corresponding digit. Then 4 coins having no zeroes in a ternary notation are surely genuine. Now we split all the coins into three parts in a following way: put coin 00 into one group, coins 01 and 02 into the second one, and 10 and 20 into the third. Complete each group by genuine coins so that each group would contain 3 coins. Compare two of these groups by the third balance. This weighing claims that FC is in one of these groups. If this is a group with 00 , then this is true (otherwise two balances lie), and we do not need the fourth weighing. If this is a group with 01 and 02 , then this claim contradicts the readings of the second balance. Hence, the first balance is operable, so the false coin has 0 as its first digit, and there are three candidates for FC. Thus, we can easily find FC in one weighing. If the false coin in a group with 10 and 20 , we proceed analogously. 2.3. a) Statement. Using three detectors with one broken among them, we can find a false coin of 32 ones in 9 detections. (Using the notation: $D_{3,1}(32) \leqslant 9$.) Solution. Split all the coins into 4 groups with 8 coins in each, and apply the same algorithm as in 2.2a). In case 1 , after 3 detections we get 8 candidates for $\mathrm{FC}$; applying problem 2.2a) again, we make the desired. In case 2, in 4 detections we find two groups with FC in one of them, and we find one operable detector. So in 4 other detections we find FC; in this case we do not need a 9th detection. b) Statement. Using three balances with one broken among them, we can find a false coin of 81 ones in 7 weighings. (Using the notation: $B_{3,1}(81) \leqslant 7$.) Solution. Analogously to problem 2.2b), we identify the coins with the sequences of 4 digits, each being from 0 to 2. Again, we take the first weighing on first balance according to the first digit of a coin, and the second weighing on the second balance by the second digit. Again, without loss of generality we may assume that zero digits are suspicious by opinions of the balances, and then the coins with nonzero first two digits are genuine. Now we again split coins into three groups: first one containing 9 coins beginning with 00 , the second containing 18 coins beginning with 01 and 02 , and the third one containing the coins beginning with 10 or 20 . we can distribute the genuine coins so that each group would contain 27 coins; then we compare two groups on the third balance. If the result claims that FM is in the group beginning with 00, then this claim is true, and we are left to work out 9 coins in 4 weighings; this is possible by problem 2.2b). Otherwise, suppose that the result claims that the false coin begins either with 01 or with 02 (the case with 10 and 20 is analogous). Then the readings of the third balance contradict those of the second balance; hence the first one is surely operable, and (by the first weighing) the false coin should begin with 0 . Thus we get 27 coins and one operable balance; then we are able to find FC in 3 weighings (in this case we do not need the 7th weighing). ## What are we unable? 2.4. Statement. If we have two coins with one of them being false, it is not possible to find the false one in 2 testings using any number of testers with at least one broken among them. (Using the notation: $D_{x, 1}(2) \geqslant 3$, $B_{x, 1}(2) \geqslant 3$.) Solution. If both weighings are made on one tester, then it can appear to be broken, and we get no information at all. Otherwise, we cannot find FC if the readings of two balances contradict each other. 2.5. a) Statement. It is not possible to find a false coin from $2^{k}$ ones in $k$ detections by any number of detectors with one broken among them. (Using the notation: $D_{x, 1}\left(2^{k}\right)>k$ ). Solution. Suppose this is possible. Suppose that some tests are already taken. Consider all the coins which can appear to be false. For each of them, there exists a result of the next test which allows this coin to remain false. Hence the number of such coins divides by not more than 2 for some result of the next test. Note that this number did not change on the first test, since this test might use the broken tester. Hence after $d$ steps (with an appropriate choice of test results) we get not less than $2^{k} / 2^{d}$ possible FCs. Hence, if we find FC, then $2^{k} / 2^{d} \leqslant 1 \Longleftrightarrow d \geqslant k$, QED. b) Statement. It is not possible to find a false coin from $3^{k}$ ones in $k$ weighings by any number of balances with one broken among them. (Using the notation: $B_{x, 1}\left(3^{k}\right)>k$ ). Solution. Analogously, but the number of candidates divides by 3 instead of 2 . 2.6. a) Statement. It is not possible to find a false coin from $2^{k}$ ones in $k+1$ detections by any number of detectors with one broken among them. (Using the notation: $\left.D_{x, 1}\left(2^{k}\right)>k+1\right)$. Solution. We say that a testing refuses to a coin, if it claimed that this coin is not in a group with a false one. In other words, if a testing refused to a coin, then either the coin is genuine or the tester is broken. By the same reasons as in a previous problem, if one can find FC in $k+1$ testing, then after the second testing we get not more than $2^{k-1}$ candidates to be false. On the other hand, each coin, to which at least one of first two testings did not refuse, is such a candidate (if another testing was made on a broken detector). Let us estimate the number of such coins. The first testing (for one of the results) refuses to not more than $\frac{1}{2}$ of coins; hence it does not refuse to a half at least. The second testing - for one of the results - does not refuse to some coin such that the first one refused to it. Then, the number of candidates is more than a half on a total number of coins; hence the desired is not possible. b) Statement. It is not possible to find a false coin from $3^{k}$ ones in $k+1$ weighings by any number of balances with one broken among them. (Using the notation: $\left.B_{x, 1}\left(3^{k}\right)>k+1\right)$. Solution. Analogously: we get that the first weighing does not refuse to at least a third of the total number of coins, and the second one does not refuse to at least one more coin. Hence, after these weighings the number of candidates is greater than a third of the total number of coins. c) Statement. It is not possible to find a false coin from $n>3^{6}$ ones in 11 weighings by any number of balances with two broken ones among them. (Using the notation: $B_{x, 2}(n)>11$, if $n>3^{6}$ ). Solution. The solution of this problem is in the end of Section 3. ## 3 Rough but serial results ## Upper bounds 3.1. Statement. a) For each $k$ find the minimal value of $K$ such that $D_{K, k}(n)<\infty$ for every $n$ (in other words, find the least number of detectors with $k$ broken among them such that it is possible to find the false coin using them). b) For each $k$ find the minimal value of $K$ such that $B_{K, k}(n)<\infty$ for every $n$. ## Solution. The solution does not depend on the type of testers. Answer. $K=2 k+1$. Suppose the number of testers is $\leqslant 2 k$, and the coin $A$ is false. Let all broken testers work so as the false coin is $B$. In this case we cannot determine whether all broken testers are broken (and the false coin is $A$ ) or we have the opposite case (and the false coin is $B$ ). If the number of testers is $2 k+1$ then each coin (or, in the case of balances, each pair of coins) can be tested by all testers; $k+1$ results will coincide, and this will be the valid result. 3.2. a) Statement. Prove that $D_{3,1}\left(2^{k}\right) \leqslant 2 k+1$. Solution. We use induction on $k$. For $k=1$ we test the first coin by all three detectors; two coincident answers are valid. Suppose the assertion is proved for $k=t$. Consider $2^{t+1}$ coins and test first $2^{t}$ of them by first two detectors. If the results coincide then they are valid, so we reduce the number of coins to $2^{t}$; by the induction hypothesis, we obtain our assertion. In the case of distinct results, one of two first detectors is broken; then the third (apparently operable) detector can find FC in $t+1$ detections; in this case we have used not more than $t+3 \leqslant 2(t+1)+1$ detections. b) Statement. Prove that $B_{3,1}\left(3^{k}\right) \leqslant 2 k+1$. Solution. Quite similarly. 3.3. a) Statement. Prove that $B_{3,1}\left(3^{2 k}\right) \leqslant 3 k+1$. Solution. Again by induction; the base case $k=1$ is Problem $2.2 \mathrm{~b}$ ). The induction step is quite similar to the solution of $2.3 \mathrm{~b}$ ). b) Statement. Prove that $D_{3,1}\left(2^{2 k}\right) \leqslant 3 k+2$. Solution. Similarly, by induction; the induction step is similar to the solution of $2.3 \mathrm{a})$. The base case is proved by a similar argument; namely, we either find FC in 3 detections, or in 4 detections we find a pair of coins including the false one, and an operable detector. Then the remaining testing enables to find FC. 3.4. a) Statement. Prove that having an infinite number of detectors with one broken, one can find a false coin from $2^{k}$ in $k+o(k)$ weighings; that is, $D_{\infty, 1}\left(2^{k}\right)=k+o(k)$. Solution. Enumerate all coins by $k$-digit binary numbers (from $0 \ldots 0$ to $1 \ldots 1$ ). In the first $k$ testings we apply first $k$ detectors, each to the corresponding position. Without loss of generality, we may assume that each detector responses that the false coin has 0 in the corresponding position. Hence the number corresponding to FC contains not more than one 1. Apply $(k+1)$ th detector to the coin $0 \ldots 0$. If the response is that it is false then this is valid (otherwise two detectors lie!). In the opposite case, some detector among first $k+1$ ones has lied, and we have $k+1$ coins possibly false. Among them, we can find FC by $(k+2)$ th (operable!) detector in $\leqslant \log _{2}(k+1)+1$ testings. In all, we have used not more than $k+1+\log _{2}(k+1)+1=k+o(k)$ testings. b) Statement. Prove that having an infinite number of balances with one broken, one can find a false coin from $3^{k}$ in $k+o(k)$ weighings; that is, $B_{\infty, 1}\left(2^{k}\right)=k+o(k)$. Solution. Quite similarly. 3.5. a) Statement. Prove that there exists $x$ such that $D_{x, 1}\left(2^{k}\right)=k+o(k)$. b) Prove that there exists $x$ such that $B_{x, 1}\left(3^{k}\right)=k+o(k)$. Solution. Follows from the next problem. 3.6. b) Statement. Prove that $B_{3,1}\left(3^{k(k+1)}\right) \leqslant(k+1)^{2}$ for every $k \geqslant 2$. Solution. We will use induction on $k$. The base case $k=1$ was proved in $2.2 . \mathrm{b}$. As above, enumerate coins in the ternary notation. In the first $k$ weighing, we examine first $k$ position by the first balance, and in the next $k$ weighings we examine the next $k$ positions by the second balance. Without loss of generality, we may assume that according to all weighings, the false coin has 0 in the corresponding positions. Now if FC has not 0 both in the first and in the second group of positions, then both balances have lied. Hence all such coins are genuine (we call them standards). For the $(2 k+1)$ th weighing, we split all non-standards as follows. The first group includes coins having only zeros in the first $2 k$ positions; the second group includes coins having only zeros in the first $k$ positions but not only zeros in the next $k$ positions; and the third group consists of coins which have not only zeros in the first $k$ positions but only zeros in the next $k$ positions. Clearly the second and the third group include equal number of coins; apply the third balance to them. If the weighing indicates the first group then the results of all weighings are valid (otherwise at least two balances lied); hence it remains to determine $k(k-1)$ positions in $k^{2}$ weighings which is possible by the induction hypothesis. If $k^{2}$ th weighing indicates the second group (similarly for the third group) then the indicating of the second and of the third balance contradict each other. Hence the first balance is definitely operable. Then we already know the first $k$ positions, and in the remaining $k^{2}$ testing, we can determine the remaining positions by the operable balance. a) Statement. Prove that $D_{3,1}\left(2^{k(k+1)}\right) \leqslant(k+1)^{2}$ for every $k \geqslant 5$. Solution. An estimate analogous to part b) would be weaker, namely, $\mathrm{D}_{3,1}\left(2^{k(k-1) / 2-1}\right) \leqslant k(k+1) / 2-1$. Unfortunately, it seems that this problem does not have a solution much easier than 4.4. ## Lower bounds 3.7. a) Statement. Prove that $D_{x, 1}(n) \leqslant D_{x, 1}\left(2^{k}\right)$ if $n<2^{k}$. Solution. See the solution of $3.8 \mathrm{a}$ ). b) Statement. Prove that $B_{x, 1}(n) \leqslant B_{x, 1}\left(3^{k}\right)$ if $n<3^{k}$. Solution. The authors know no simple proof of this fact. 3.8. a) Statement. Prove that $D_{x, 1}(n) \leqslant D_{x, 1}(N)$ if $n3^{6}$ ones in 11 weighings by any number of balances with two broken ones among them. (Using the notation: $B_{x, 2}(n)>11$, if $n>3^{6}$ ). Solution. Consider an arbitrary algorithm which enables to find FC among $n$ coins in 11 steps. Suppose we have carried it out; write down the results of detections in a line. We obtain a sequence of 11 symbols " $<$ ", "=" and ">". By the same reasons as in 3.9a), each of these sequences uniquely determines an FC. Now we shall find the number of sequences corresponding to the same FC. This number is not less than $1+2$. $11+4 \cdot 55=243$; indeed, we have at least the sequences of the following types: 0 ) all answers are valid; 1) just one answer is wrong (the number of such sequences is 22: there are 11 ways to choose the position for the wrong answer, and each of them may contain one of two possible wrong answers); 2) just 2 answers are wrong (the number of such sequences is $4 \cdot 55$, since there are 55 pairs of positions and 4 pairs of wrong answers). Clearly all the above sequences are distinct. Thus the number of sequences is not less than $243 n$; on the other hand, the number of all possible sequences does not exceed $3^{11}$. Hence $n \cdot 243 \leqslant 3^{11}$, and so $n \leqslant 3^{6}$. ## 4 Sharp results 4.1. Statement. a) Prove that $D_{4,1}\left(2^{4}\right)=7$. b) Prove that $B_{4,1}\left(3^{6}\right)=9$. Solution. See the solution 4.4. Nevertheless, these problems can be solved without use of such general methods. 4.2. Statement. a) Find the maximal number of coins $n$ such that $D_{4,1}(n) \leqslant 15$. b) Find the maximal number of coins $n$ such that $B_{4,1}(n) \leqslant 13$. c) Find the maximal number of coins $n$ such that $B_{4,1}(n) \leqslant 40$. Solution. Answers. a) $2^{11}$; b) $3^{10}$; c) $3^{36}$. The upper bounds follow from 3.9. Their attainability follows from the solution of 4.4. Nevertheless it is possible to obtain the corresponding algorithms by direct methods. 4.3. Statement. Find some value of $n$ such that $B_{4,1}(n)b_{1}$ then we decrease $b_{2}$, otherwise we decrease $b_{1}$ ). Thus we may assume $a_{1}+a_{2}+a_{3}=b_{1}+b_{2}$. Put all the students in a line: first, all the students from the first class, then from the third, and after that from the second one. Then place first $b_{1}$ of them in the first bus, the next $b_{2}$ ones in the second bus. Then all students from the first class will be in the first bus (by the condition), and the students from the second class will be in the second bus (otherwise $b_{2}a_{1}+a_{3}>a_{3} \geqslant b_{2}$, which is wrong). Conversely, if $b_{1}13$. Suppose the contrary; this is possible only if the significance decreases exactly thrice at each step (for any result of weighing!). This in turn means that at any moment there exist refusers of not more than two types, namely not of the type involved in the weighing in question. Let us examine how this occurs. Denote a situation by a triple of numbers $(a, b, c)$ where $a$ is the number of suspected coins, and $b$ and $c$ are the numbers of refusers of two existing types. Then the initial situation is $\left(3^{10}, 0,0\right)$, and after the first weighing we have $\left(3^{9}, 2 \cdot 3^{9}, 0\right)$. In the second weighing, let the numbers of suspected coins in the groups be $a, b, c$. Suppose the weighing indicates the first group. Then we have the situation $\left(a, 3^{9}-a, t\right)$ where the number $t$ is determined by the total significance: $23 a+\left(3^{9}-a\right)+t=3^{11}$ and so $t=8 \cdot 3^{9}-22 a$. Note that $t \geqslant 0$, hence $a \leqslant \frac{4}{11} 3^{9}$. Similarly $b, c \leqslant \frac{4}{11} 3^{9}$, hence $a=3^{9}-(b+c) \geqslant \frac{3}{11} 3^{9}$. Consider the third weighing. If all $a$ suspected coins are in the same group then its significance is not less than $21 \cdot \frac{3}{11} 3^{9}>3^{10}$ which is impossible. Hence the suspected coins belong to at least two groups. To have refusers of not more than two types for each possible result, it is necessary to have refusers of a single type in each group. Thus one of the groups will contains all refusers of some type; let their number be $s$. If this group contains $x$ suspected coins besides that then its significance after weighing equals $3^{10}=21 x+(a-x)+s=s+a+20 x$; thus $3^{10}-s-a$ is divisible by 20. Furthermore, we have either $s=3^{9}-a$ or $s=t=8 \cdot 3^{9}-22 a$. In the first case $3^{10}-s-a=2 \cdot 3^{9}$ is not divisible by 20 which is impossible. Hence we have the second case, and then $3^{10}-s-a=21 a-5 \cdot 3^{9}$. Since 20 divides this number, 5 divides $a$. Similarly, 5 divides $b$ and $c$; this contradicts $a+b+c=3^{9}$ since the last number is not divisible by 5 . 4.5. a) Statement. Prove that $D_{x, 1}(n)=D_{4,1}(n)$ for every $n$ and $x>4$. Solution. Suppose we must find the false coin among $n$ coins. Note that if $t$ satisfies the injequality $2^{t}4$. Solution. Similarly; in addition, we have to provide that two of groups obtained at any step contain equal number of elements. It is possible to do by adding standards. Remark. The same methods (although with more technical details) enable us to show that the ideal number of testers does exist for any number of broken testers. 4.6. a) Statement. Find whether the estimate of the same form as in problem 4.4 is valid for $D_{3,1}(n)$. Solution. Suppose we wish to find the false coin among $2^{k}$ coins using 3 testers one of which is broken. Consider the minimal $d$ such that $2^{d}+d+1 \geqslant 2^{k}(2 d+1)$. We will prove that $d$ steps are sufficient to retain not more than two coins which can occur to be false (this will be the moment relative to which we determine the significance). Put $x_{i}=2^{i}+i+1$. We construct a process such that for $i$ weighings before the end the significance does not exceed $x_{i}$, and refusers always belong to not more than two distinct types. Note that if have provided this then the significance after the last step does not exceed 2 , hence not more than 2 coins will be suspected or refusers at the end. We construct the process by induction. Before the first weighing, both conditions are satisfied. Suppose we have to carry out $i$ last weighings. We will define the first of them. Instead of the bus lemma, we use the following (obvious) Lemma on dissection. Suppose we have some objects arranged in a line such that each of them costs not more than $x$, and the total price of them is $S$. Then the line can be cut into $t$ parts so that each part costs not more than $\frac{S+(t-1) x}{t}$. Without loss of generality, we may assume that there are refusers of the first and the second types. Put coins in a line: first the refusers of the first type, then the suspected coins, and then the refusers of the second type. Denote the number of suspected coins by $a$. Sum the refusers with coefficient 1 , and the suspected coins with coefficient $i-1$. Then the sum of the coefficients does not exceed $x_{i}-2 a$; hence by lemma on dissection this line can be dissected into two parts such that the sum of coefficients in each part is $\leqslant \frac{x_{i}-2 a+i-1}{2}$. If now we test coins of any part of the dissection, the significance of the part obtained equals the sum of its coefficients plus the number of suspected coins (since the suspected coins in this part are summed with coefficient $i$ instead of $i-1$, and in the other part with coefficient 1 instead of 0 ). Thus for any result the significance does not exceed $\frac{x_{i}+i-1}{2}=x_{i-1}$. Moreover if some part of the dissection includes refusers of both types then it includes all suspected coins; hence for each result of testing refusers of not more than two types can remain. Thus after $d$ weighings we have not more than two possible FC. It remains to observe that a single false coin among two coins can be determined in a bounded number of steps (for instance, in 3 steps). Hence we have proved that $d+3$ testings are sufficient. Furthermore $d=\log _{2}\left(2^{k}(2 d+1)-d-1\right)$ hence $d=\log _{2} k+\log _{2} \log _{2} k+O(1)$ as required. b) Statement. Тот же вопрос про $B_{3,1}(n)$. Solution. Similarly, we intend (for $n$ sufficiently great) to keep only 3 possibly false coins among $n$ coins in $d$ weighings. We assert that it suffices to take $d$ such that $$ (2 d+1)+2 \cdot 3^{d} \geqslant n(2 d+1) $$ Then (by problem ) we will find FC in $d+3$ weighings; moreover $d=\log _{3} n+\log _{3} \log _{3} n+O(1)$. Again, the moment for determining the significance is the moment after the $d$ th weighing; then the initial significance equals $n(2 d+1)$. Denote $y_{i}=(2 i+1)+2 \cdot 3^{i}$; we arrange the process so that in $i$ weighings before the end of the process the significance does not exceed $y_{i}$ and the number of types of refusers does not exceed 2 ; then the significance after the $d$ th weighing does not exceed 3 as required. The step of the algorithm again is constructed similarly. Attach the factor $D=2 i-2$ to all suspected coins, and the factor 1 to all refusers; then the total sum of factors does not exceed $y_{i}-3 a$ where $a$ is the number of suspected coins. Put all coins in a line: first the refusers of the first type, then the suspected coins, and at last the refusers of the second type. Then the lemma on dissection implies that they can be split into three parts so that the sum of factors in each part does not exceed $\frac{y_{i}-3 a+2 D}{3}$. This means that the significance of any situation after weighing with these three parts does not exceed $\frac{y_{i}-3 a+2 D}{3}+a=y_{i-1}$ as required. The last obstacle which might appear is that the numbers of coins in three groups obtained may differ. We will show that the number of standards is sufficient to equalize two groups: then we can arrange the required weighing. As is easily seen, this trouble does not exist at the first two steps: in the first weighing, two groups contain equal number of coins, and in the second weighing we can easily redistribute refusers and suspected coins in three groups so that two groups will contain equal number of coins (by a slight modification of the algorithm for dissection but with no weakening of the estimate). After these weighings, we have not less than $\frac{4}{9} n-2$ standards. We have distributed not more than $\frac{5}{9} n+2$ coins in 3 groups, hence the difference of numbers of coins in two of these groups does not exceed $\frac{5}{27} n+1 \leqslant \frac{4}{9} n-2$ for $n \geqslant 12$. Thus the number of standards is sufficient. The solution is complete. # УСТОЙЧИВОСТЬ ПЕРЕСЕЧЕНИЙ ПУТЕЙ НА ПЛОСКОСТИ П. Дергач, И. Нетай, А. Скопенков, М. Скопенков Основными результатами являются критерии устойчивости самопересечений путей и циклов на плоскости (задачи D2d и D3d; доказательство намечено в задачах D2abc и D3abc; определения см. далее). Критерии даются в терминах дифференцирования графов и путей ${ }^{1}$. Если условие задачи является формулировкой утверждения, то подразумевается, что это утверждение и надо доказать. Части А, В, С предлагаются до промежуточного финиша, остальные - после. 0. Город $N$ на плоскости состоит из нескольких площадей (кругов), соединенных непересекающимися дорогами (прямолинейными отрезками). Известно, что существует маршрут, проходящий по каждой дороге ровно один раз (этот маршрут может проходить по площадям несколько раз). Докажите, что существует несамопересекающийся маршрут, проходящий по каждой дороге ровно один раз. ## А. Устойчивость пересечений пары путей. Проблема устойчивости пересечений пары путей ${ }^{2}$. Два охотника охотятся в лесу. Каждый из них ведет на коротком поводке собаку. Собаки слушаются охотников и движутся так, как те им говорят. Если одна из них пересечет следы другой, то она станет лаять и спугнет дичь. Как по данным путям охотников определить, смогут ли они избежать срыва охоты (если смогут, то пересечение путей охотников называется неустойчивым ${ }^{3}$ )? Будем считать, что пути охотников и собак составлены из конечного числа прямолинейных отрезков (т.е. кусочно-линейны). Для таких путей охотников известен медленный, "переборныи" алгоритм распознавания устойчивости пересечений. Нахождение быстрого алгоритма - нерешенная проблема. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-107.jpg?height=616&width=622&top_left_y=1528&top_left_x=200) a ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-107.jpg?height=460&width=708&top_left_y=1686&top_left_x=868) b Рис. 1: Трансверсальное пересечение путей и пара путей в букве " $H$ " Везде в дальнейшем мы предполагаем, что охотники движутся по некоторой системе дорожек на плоскости. И охотника, и собаку, мы считаем точками (при этом мы разрешаем[^2]охотнику и собаке находиться в один момент времени в одной и той же точке плоскости). Длины поводков считаются равными 1 м (то есть расстояние между охотником и его собакой в каждый момент времени не превосходит 1 м). A-1. Два охотника движутся по дорожке в форме отрезка длины 1 км. При этом они могут менять направление своего движения. Докажите, что независимо от движения охотников собаки смогут двигаться так, чтобы не пересекать следы друг друга. Пример. Два охотника прошли (равномерно не меняя направления) по прямолинейным дорожкам, пересекающимся под прямым углом в точке, отстоящей от каждого из их концов на 1 км (рис. 1.а, на котором $\varphi\left(I_{1}\right)$ и $\varphi\left(I_{2}\right)$ - пути охотников, а $f_{1}\left(I_{1}\right)$ и $f_{2}\left(I_{2}\right)$ возможные пути собак). Тогда одна собака пересекала следы другой. A-2. Система лесных дорожек имеет форму буквы " $H$ " (см. рис. 1.b), причем длина каждого из отрезков $A B, B C, B E, D E, E F$ равна 1 км. Один из охотников прошел по пути $A B E F$, а второй - по пути $C B E D$. Тогда одна из собак пересекала следы другой. Для доказательства того, что некоторые пути собак обязаны пересекаться, может оказаться полезной следующая теорема. Ей разрешается пользоваться без доказательства. Циклом называется путь, начало и конец которого совпадают, и при этом забыто, где начало. Мы говорим, что два пути (или цикла) пересекаются трансверсально, если вблизи каждой точки пересечения они выглядят подобно двум путям на рис. 1.a. Теорема о четности. Два (кусочно-линейных) цикла на плоскости, пересекающихся трансверсально, пересекаются в четном числе точек. Заметим, что точки самопересечения (то есть точки, соответствующие пересечению собакой своих собственных следов) не считаются за точки пересечения. A-3. Система лесных дорожек имеет форму буквы "Y", составленной из трех прямолинейных отрезков длины 1 км, образующих в их общей точке углы $2 \pi / 3$ (рис. 4.Y). Придумайте такие пути двух охотников, чтобы срыва охоты невозможно было избежать. ## В. Устойчивость самопересечений пути и цикла. Проблема устойчивости самопересечений пути. Охотник гуляет по лесу, ведя на коротком поводке собаку. Собака слушается охотника и движется так, как тот ей говорит. Если она пересечет свои следы, то залает и спугнет дичь. Как по данному пути охотника определить, сможет ли он избежать срыва охоты (если сможет, то самопересечения пути называются неустойчивыми $)^{4}{ }^{5}$ ? Основной результат данного цикла задач - быстрый алгоритм распознавания устойчивости самопересечений. B-1. (а) Охотник гуляет по лесной дорожке, имеющей форму прямолинейного отрезка длины 1 км. При этом он может менять направление своего движения. Докажите, что независимо от движения охотника собака может двигаться так, чтобы не пересекать свой след. (b) То же для дорожки в форме окружности (радиуса 1 км). B-2. (a) Несамопересекающийся путь имеет неустойчивые самопересечения (наша терминология не должна смущать читателя). (b) Если самопересечения пути неустойчивы, то то же верно для любого его подпути ${ }^{6}$.[^3](c) Если пересечения некоторой пары подпутей данного пути устойчивы, то самопересечения этого пути устойчивы. (d) Существует путь, не содержащий трансверсальных пересечений (рис. 1.a), и все равно имеющий устойчивые самопересечения. Одним из основных результатов данного цикла задач является следующая теорема. Теорема о паре подпутей. Путь на плоскости имеет устойчивые самопересечения, если и только если некоторая пара его подпутей имеет устойчивые пересечения. B-3. Пусть система дорожек образует некоторый граф на плоскости, ребра которого являются отрезками длины 1 км. (Причем расстояние от любой вершины до любого ребpa, ее не содержащего, больше 10 м). Предположим, что охотник прошел по этой системе дорожек, пройдя по каждой из них ровно 1 раз и меняя направление движения только в местах соединения дорожек. Докажите, что путь охотника имеет устойчивые самопересечения, если и только если он содержит трансверсальное самопересечение. B-4. Существует алгоритм проверки устойчивости самопересечений для данного пути на плоскости. B-5. (а) Охотник (равномерно не меняя направления) двигался по лесной дорожке в форме окружности диаметром 1 км, сделав два оборота. Он вел на поводке длиной 1 м собаку, которая в конце движения вернулась в исходную точку. Докажите, что собака обязательно пересекала свой след (в некоторый момент времени, отличный от конечного, рис. 2). (b) Верно ли (a) без предположения о том, что собака в конце движения вернулась в исходную точку? (c) Докажите аналог (а) для случая, когда охотник сделал три оборота. (d) Для какого числа оборотов в (а) собака обязательно пересекала свой след? (е) Предположим, что дорожка имеет форму отрезка длиной 1 км. Докажите, что независимо от движения охотника собака может двигаться так, чтобы не пересекать свои следы и в конце движения вернуться в исходную точку. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-109.jpg?height=317&width=311&top_left_y=1600&top_left_x=884) Рис. 2: Путь собаки ## С. Производные графов и путей. Производной $G^{\prime}$ графа $G$ называется граф, вершины которого находятся во взаимно однозначном соответствии с ребрами графа $G$. Вершины $e^{\prime}$ и $f^{\prime}$, соответствующие ребрам $e$ и $f$, соединены ребром в графе $G^{\prime}$, если ребра $e$ и $f$ имеют общую вершину (рис. 3). C-1. Нарисуйте производные (см. рис. 4) (a) дуги с $n$ ребрами; (b) окружности с $n$ ребрами; (c) звезды с $n$ лучами ( $n$-ода); (d) буквы " $H$ ". C-2. Граф называется планарным, если его можно нарисовать на плоскости без самопересечений. Производная планарного графа не обязательно планарна. Путем в графе $G$ назовем любую последовательность его вершин $v_{0}, v_{1}, \ldots, v_{n}$, такую что $v_{i}$ и $v_{i+1}$ соединены ребром в $G$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-110.jpg?height=494&width=1168&top_left_y=136&top_left_x=455) Рис. 3: Производная пути в графе ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-110.jpg?height=397&width=1280&top_left_y=744&top_left_x=408) Рис. 4: Продифференцируйте нас! Путь (цикл) $v_{0}, \ldots, v_{n}$ называется эйлеровым, если он проходит по каждому ребру ровно один раз, т.е. если среди ребер $v_{0} v_{1}, \ldots, v_{n-1} v_{n}$ встречаются все ребра графа $G$ ровно по одному разу. Пусть путь $\varphi$ в графе $G$ задается последовательностью $v_{0}, v_{1}, \ldots, v_{n}$ вершин. Рассмотрим последовательность $\left(v_{0} v_{1}\right)^{\prime}, \ldots,\left(v_{n-1} v_{n}\right)^{\prime}$ вершин производной графа $G$. В этой последовательности могут стоять подряд одинаковые вершины. Для каждого такого набора одинаковых вершин (стоящих подряд в последовательности вершин производной) заменим этот набор на одну вершину. Полученный путь $\varphi^{\prime}$ в графе $G^{\prime}$ называется производной пути $\varphi$. Производная пары путей - это пара путей, определяемая аналогичным образом. Пример. Пусть $A$ - граф с вершинами $a_{1}, a_{2}, a_{3}, a_{4}, a_{5}$ и ребрами $a_{1} a_{2}, a_{2} a_{3}, a_{2} a_{4}$, $a_{3} a_{4}, a_{4} a_{5}$ (рис. 3). Обозначим через $b_{1}=\left(a_{1} a_{2}\right)^{\prime}, b_{2}=\left(a_{2} a_{3}\right)^{\prime}, b_{3}=\left(a_{2} a_{4}\right)^{\prime}, b_{4}=\left(a_{3} a_{4}\right)^{\prime}$, $b_{5}=\left(a_{4} a_{5}\right)^{\prime}$ вершины производной графа $A$. Пусть $\varphi$ - путь $a_{1} a_{2} a_{4} a_{2} a_{3} a_{4} a_{2} a_{4} a_{5}$ в графе $A$. Тогда производная $\varphi^{\prime}$ - это путь $b_{1} b_{3} b_{2} b_{4} b_{3} b_{5}$. (На рис. 3 изображены не сами пути $\varphi$ и $\varphi^{\prime}$, а некоторые близкие к ним пути на плоскости). C-3. (а) Найдите первые и вторые производные путей и пар путей с рис. 1, рис. 3 (а также путей и пар путей, построенных Вами в решениях задач А3 и В2d). (b) Число вершин в производной пути с $n$ вершинами не превосходит $n-1$. (c) Будем говорить, что путь имеет точку возврата, если по некоторому ребру он проходит два раза подряд. Путь $\varphi$ без точек возврата является эйлеровым тогда и только тогда, когда путь $\varphi^{\prime}$ не имеет самопересечений. ## Формальные определения. Приведем эквивалентную формулировку наших задач на примере задачи В5. (эквивалентность доказана в [Mi97]). Рассмотрим две полянки (т.е. два круга), соединенных двумя тропинками (т.е. полосками) $a$ и $b$, как на рис. 5 и рис. 2. Собака бегала по полянкам и тропинкам и вернулась в исходную точку. Каждый раз, когда собака перебегала с полянки на тропинку, она записывала обозначение этой тропинки. В задаче B5 утверждается, что если получилась запись $a b a b$, то собака обязательно пересекала свой след (в некоторый момент времени, отличный от конечного). Аналогично можно переформулировать другие задачи. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-111.jpg?height=483&width=602&top_left_y=632&top_left_x=744) Рис. 5: Две полянки Приведем формальное определение устойчивости пересечений и самопересечений. Обозначим через $I=[0,1]$ отрезок, через $S^{1}$ окружность (т. е. отрезок со склеенными концами) и через $\mathbb{R}^{2}$ плоскость. Кусочно-линейным путем на плоскости называется отображение $\varphi: I \rightarrow \mathbb{R}^{2}$, для которого существуют такие точки $0=v_{0}0$ существует такой несамопересекающийся путь $f: I \rightarrow \mathbb{R}^{2}$, что расстояние между точками $f(x)$ и $\varphi(x)$ меньше $\varepsilon$ для любой точки $x \in I$ ). Аналогично определяется устойчивость самопересечений цикла $\varphi: S^{1} \rightarrow \mathbb{R}^{2}$. Пара путей $\varphi_{1}, \varphi_{2}: I \rightarrow \mathbb{R}^{2}$ имеет неустойчивые пересечения (или допускает устранение пересечений малым шевелением), если существует пара сколь угодно близких к ним непересекающихся путей (т.е. если для любого $\varepsilon>0$ существуют такие непересекающиеся пути $f_{1}, f_{2}: I \rightarrow \mathbb{R}^{2}$, что расстояние между точками $f_{i}(x)$ и $\varphi_{i}(x)$ меньше $\varepsilon$ для любой точки $x \in I$ и $i=1,2$ ). Например, трансверсальное пересечение (рис. 1.а) двух путей устойчиво. Скажем, задача Blab на этом языке формулируется так: Если образом $\varphi(I)$ пути $\varphi: I \rightarrow \mathbb{R}^{2}$ является отрезок или окружность, то его самопересечения неустойчивы. ## Указания и решения к некоторым вводным задачам 0. Проведём сначала какой-нибудь эйлеров путь. Будем считать, что на площадях путь прямолинеен. Самопересечения случаются только на площадях. Рассмотрим какое-нибудь самопересечение. Оно состоит из четырёх вершин на окружности и двух диагоналей образованного ими выпуклого четырёхугольника. Изменим наш путь. Для этого в нашем четырёхугольнике вместо пары диагоналей проведём пару противоположных сторон. Это можно сделать так, чтобы мы снова получили эйлеров путь (а не путь и цикл, не связанные между собой). Будем действовать так, пока у нашего пути есть самопересечения. Таких операций может произойти только конечное число, потому что длина пути уменьшается, а на фиксированном множестве вершин при прямолинейных рёбрах возможно только конечное множество возможных длин графов. A-1. Проведем прямую через нашу дорогу. Она разделяет плоскость на две полуплоскости. Пусть первый охотник прикажет своей собаке двигаться в одной из данных полуплоскостей, а второй охотник прикажет своей собаке двигаться в другой полуплоскости. Тогда следы собак не пересекутся. A-2. Предположим, что собаки могут двигаться так, чтобы не пересекать следы друг друга. Пусть $A^{\prime} F^{\prime}$ и $C^{\prime} E^{\prime}$ - пути собак. Замкнем эти пути, добавив к ним ломаные $F^{\prime} X A^{\prime}$ и $E^{\prime} X C^{\prime}$, показанные на рисунке. Так как расстояние между охотником и собакой много меньше попарных расстояний между точками $A, C, E, F$, то ломаная $F^{\prime} X A^{\prime}$ не пересекает путь $C^{\prime} E^{\prime}$, а ломаная $E^{\prime} X C^{\prime}$ не пересекает путь $A^{\prime} F^{\prime}$. Значит, два цикла $A^{\prime} F^{\prime} X A^{\prime}$ и $C^{\prime} E^{\prime} X C^{\prime}$ пересекаются трансверсально в единственной точке $X$. А по теореме о четности число их точек пересечения должно быть четно. Полученное противоречие доказывает, что пути собак обязательно пересекаются. Примеры к задачам A3 и B2d приведены на рис. 6, где для наглядности нарисован не сам путь, а близкий к нему путь общего положения. См., впрочем, [Mi97, Sk03]. Указание: можно свести к непланарности графов Куратовского $K_{5}$ и $K_{3,3}$. Пунктирная линия на рис. 6.d поможет сделать это. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-112.jpg?height=294&width=371&top_left_y=1418&top_left_x=571) a) ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-112.jpg?height=331&width=331&top_left_y=1796&top_left_x=614) c) ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-112.jpg?height=257&width=537&top_left_y=1456&top_left_x=974) b) ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-112.jpg?height=254&width=511&top_left_y=1883&top_left_x=955) d) Рис. 6: Пути с устойчивыми самопересечениями B-1. (а) Направим ось $O x$ вдоль дорожки, а ось $O y$ - перпендикулярно к ней. Предположим, что координата охотника (выраженная в метрах) задается функцией $x(t)$, а время охоты $t$ изменяется в пределах от 0 до $T$. Пусть охотник прикажет собаке двигаться по графику функции $x(t)$, сжатому к оси $O x$, то есть положим координаты собаки в момент времени $t$ равными $(x(t) ; t / T)$. Легко видеть, что тогда собака не будет пересекать свои следы, а в каждый момент времени будет находиться не дальше 1 м от охотника. (b) Будем действовать аналогично пункту (a): пусть собака движется так, чтобы в каждый момент времени $t$ она находилась на луче, направленном из центра окружности в точку, где находится охотник, а ее расстояние от охотника равнялось $t / T$ м ( $T$ - общее время охоты). Тогда собака не будет пересекать свои следы, а в каждый момент времени будет находиться не дальше 1 м от охотника. B-2. (а) Пусть собака идет за охотником "след в след", то есть в каждый момент времени собака и охотник находятся в одной и той же точке (это не запрещено условием). Поскольку путь охотника не самопересекается, то и путь собаки не будет самопересекаться. (b) Поскольку самопересечения пути неустойчивы, то собака может двигаться, не пересекая свои следы. Рассмотрим движение собаки только в отрезок времени, соответствующий выбранному подпути. Он также не самопересекается. Значит, выбранный подпуть также имеет неустойчивые самопересечения. (с) Предположим, что самопересечения пути неустойчивы. Тогда собака может двигаться, не пересекая свои следы. Рассмотрим движение собаки только в те два отрезка времени, которые соответствуют выбранным подпутям. Эти два пути не перескаются. Значит, самопересечения соответствующей пары подпутей неустойчивы. B-3. Ясно, что если путь охотника содержит трансверсальное самопересечение, то его самопересечения устойчивы. Докажем, что если в рассматриваемом случае путь не содержит трансверсальных самопересечений, то его самопересечения неустойчивы. Нарисуем вокруг каждой точки соединения дорожек круг радиусом 1 м. Пусть все время, когда охотники находятся вне этих кругов, собака движется след в след за охотником. Как только охотник начинает двигаться внутри некоторого круга, собака срезает маршрут, двигаясь по хорде вместо пары радиусов, как показано на рисунке. Построенный путь собаки не самопересекается. Действительно, так как по каждой дорожке охотник проходит ровно 1 раз, то вне кругов собака не пересекает свои следы. Если же собака пересекает свои следы внутри некоторого круга, то некоторые две из построенных нами хорд пересекаются. А это возможно только в том случае, если путь охотника имеет в этом месте трансверсальное самопересечение. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-113.jpg?height=485&width=491&top_left_y=1385&top_left_x=797) Рис. 7: Собака срезает путь. B-4. Смотри формулировку задачи D7 (простой, но медленный алгоритм) или теоремы Минца D2d (более сложный, но быстрый алгоритм). B-5. (а) Указание. Предположим, что собака может двигаться так, чтобы не пересекать свои следы. Пусть $A$ - некоторая точка окружности. Рассмотрим луч $O A$, направленный из центра окружности в точку $A$. Ясно, что цикл собаки пересекал луч $O A$ как минимум дважды: хотя бы раз, пока охотник делал первый оборот, и хотя бы раз - когда второй. Отметим на этом луче все его точки пересечения с путем собаки. Ясно, что найдутся две "соседние" отмеченные точки $A^{\prime}$ и $A^{\prime \prime}$ (то есть такие отмеченные точки, на отрезке между которыми нет других отмеченных точек), одна из которых относится к моменту, когда охотник делал первый оборот по окружности, а вторая - к моменту, когда охотник делал уже второй оборот. "Разорвем" цикл собаки в точках $A^{\prime}$ и $A^{\prime \prime}$ и добавим к ним пару путей $p^{\prime}$ и $p^{\prime \prime}$, расположенных "вблизи" отрезка $A^{\prime} A^{\prime \prime}$ и пересекающихся трансверсально в точке $X$, как показано на рисунке. В итоге мы из цикла собаки получим пару циклов, которые пересекаются трансверсально в единственной точке $X$. А по теореме о четности количество точек пересечения у двух циклов должно быть четно. Полученное противоречие доказывает, что собака обязана пересекать свои следы. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-114.jpg?height=496&width=994&top_left_y=134&top_left_x=544) Рис. 8: Превращение цикла в пару циклов. (b) Ответ: в этом случае собака может двигаться так, чтобы не пересекать свои следы. (c) Доказательство дословно повторяет наше рассуждение из пункта (a). (d) Ответ: для любого числа оборотов (в любом направлении), не меньшего двух. Доказательство дословно повторяет наше рассуждение из пункта (а). (е) Без ограничения общности можно считать, что охотник побывал во всех точках дорожки, включая его концы (иначе мы просто уменьшим отрезок-дорожку, чтобы добиться этого условия.) Тогда можно считать, что цикл охотника начинается и заканчивается в одном из концов отрезка. Пусть собака движется по графику пути охотника, сжатому в направлении дорожки (аналогично решению задачи B1a). Нам остается замкнуть путь собаки, добавив к нему ломаную, расположенную вблизи конца отрезка, как показано на рисунке. Полученный цикл не самопересекается. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-114.jpg?height=114&width=984&top_left_y=1298&top_left_x=548) Рис. 9: Замыкаем путь собаки. C-1. (a) Дуга с $n-1$ ребром; (b) окружность с $n$ ребрами; (c) полный граф с $n$ вершинами; (d) граф, составленный из двух треугольников ровно с одной общей вершиной. C-2. Пример: звезда с 5 лучами - планарный граф, производная которого - непланарный граф (полный граф на 5 вершинах). С-3. (а) См., например, рис. 10. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-114.jpg?height=540&width=1174&top_left_y=1847&top_left_x=454) Рис. 10: Вторая производная пути (b) Пусть исходный путь состоит из $n$ вершин $v_{1}, v_{2}, \ldots, v_{n}$. Тогда в последовательности $\left(v_{1} v_{2}\right)^{\prime},\left(v_{2} v_{3}\right)^{\prime}, \ldots,\left(v_{n-1} v_{n}\right)^{\prime}$ ровно $(n-1)$ вершин. Чтобы построить производный путь, мы из данной последовательности еще, возможно, выбрасываем несколько вершин. В итоге мы получим путь, содержащий не более $n-1$ вершины. (c) Докажем, что производная эйлерова пути не самопересекается. Пусть $\phi$ - эйлеров путь $v_{0}, v_{1}, \ldots, v_{n-1}, v_{n}=v_{0}$. Так как эйлеров путь проходит по каждому ребру ровно 1 раз, то в последовательности ребер $\left(v_{1} v_{2}\right),\left(v_{2} v_{3}\right), \ldots,\left(v_{n-1} v_{n}\right)$ нет повторяющихся. Значит, в последовательности вершин производной данного пути $\left(v_{1} v_{2}\right)^{\prime},\left(v_{2} v_{3}\right)^{\prime}, \ldots,\left(v_{n-1} v_{n}\right)^{\prime}$ нет повторяющихся. Значит, путь $\varphi^{\prime}$ - несамопересекающийся. Докажем, что если некоторый путь не имеет точек возврата и его производная не самопересекается, то исходный путь - эйлеров. Пусть $\phi$ - данный путь $v_{0}, v_{1}, \ldots, v_{n-1}, v_{n}=v_{0}$. Так как он не имеет точек возврата, то в последовательности ребер $\left(v_{1} v_{2}\right),\left(v_{2} v_{3}\right), \ldots,\left(v_{n-1} v_{n}\right)$ нет двух стоящих подряд одинаковых ребер. Значит, последовательность вершин производной данного пути это в точности $\left(v_{1} v_{2}\right)^{\prime},\left(v_{2} v_{3}\right)^{\prime}, \ldots,\left(v_{n-1} v_{n}\right)^{\prime}$. Так как производная не самопересекается, то в этой последовательности нет повторяющихся вершин. Значит, в последовательности ребер исходного пути нет повторяющихся. То есть, по определению, исходный путь - эйлеров. ## D. Основные задачи D-1. Самопересечения путей на рис. 6 устойчивы. Для кусочно-линейного пути $\varphi: I \rightarrow \mathbb{R}^{2}$ образ $\varphi(I)$ можно рассматривать как граф с вершинами $\varphi\left(v_{1}\right), \ldots, \varphi\left(v_{n}\right)$. По любому кусочно-линейному пути $\varphi$ однозначно строится путь в графе $\varphi(I)$. Обратно, пусть граф $G$ нарисован на плоскости без самопересечений так, что все его ребра являются прямолинейными отрезками. Каждому пути в графе $G$, последовательно проходящему вершины $v_{1}, v_{2}, \ldots, v_{n}$, сопоставим путь $\varphi:[0 ; 1] \rightarrow G$ на плоскости, полагая $\varphi\left(\frac{i}{n}\right)=v_{i}$ для всех $i=0, \ldots, n$ и линейно продолжая $\varphi$ на отрезки $\left[\frac{i}{n} ; \frac{i+1}{n}\right]$. Путь $\varphi:[0 ; 1] \rightarrow \mathbb{R}^{2}$, получающийся таким образом из некоторого пути в графе $G$, называется симплициальным. D-2. (а) Если путь $\varphi$ не содержит трансверсальных самопересечений, то граф $\varphi^{\prime}(I)$ планарен. Зафиксируем "естественный" способ вложения графа $\varphi^{\prime}(I)$ в плоскость (определение придумайте сами). Тогда $\varphi^{\prime}: I \rightarrow \varphi^{\prime}(I)$ будет некоторым путем на плоскости. (b) Если самопересечения пути $\varphi$ неустойчивы, то и самопересечения пути $\varphi^{\prime}$ неустойчивы. В качестве следствия получите доказательство утверждения задачи А2. (c) Если путь $\varphi$ не содержит трансверсальных самопересечений и самопересечения пути $\varphi^{\prime}$ неустойчивы, то и самопересечения пути $\varphi$ неустойчивы. (d) (Теорема Минца) Самопересечения симплициального пути $\varphi: I \rightarrow G$, содержащего $n$ точек, устойчивы тогда и только тогда, когда для некоторого $k=0, \ldots, n$ его $k$-я производная $\varphi^{(k)}$ содержит трансверсальное самопересечение. (е) Докажите теорему о паре подпутей. D-3. (а) При каких $m$ самопересечения цикла "намотки степени $m$ " (рис. 2 для $m=2$ ) устойчивы? (b) Для любого цикла $\varphi$ найдется $k$ такое, что $\varphi^{(k)}$ является намоткой. (c) Утверждения задач D2abc остаются в силе для цикла $\varphi$. (d) Самопересечения симплициального цикла $\varphi: S^{1} \rightarrow G$, содержащего $n$ точек, устойчивы тогда и только тогда, когда для некоторого $k=0, \ldots, n$ его $k$-я производная $\varphi^{(k)}$ либо содержит трансверсальное самопересечение, либо является стандартной намоткой степени $m \neq 0, \pm 1$. D-4.* Как по циклу $\varphi$ в графе определить $m$ такое, что $\varphi^{(\infty)}$ есть $m$-кратная намотка? D-5.* Сформулируйте и докажите критерий того, что данный набор путей в графе на плоскости аппроксимируем набором (a) непересекающихся и несамопересекающихся путей; (b)* непересекающихся (но, возможно, самопересекающихся) путей. D-6. Как по пути построить его "интеграл"? Используйте это для построения новых примеров путей с устойчивыми самопересечениями. D-7. (а) Для симплициального пути $\varphi: I \rightarrow G \subset \mathbb{R}^{2}$ заменим каждое ребро графа $G$ на $k$ близких кратных ребер, если путь $\varphi$ проходит по этому ребру $k$ раз. Обозначим через $\tilde{G} \subset \mathbb{R}^{2}$ полученный граф, а через $\pi: \tilde{G} \rightarrow G$ проекцию, переводящую каждый набор кратных ребер с концами $a$ и $b$ в ребро $a b$. Самопересечения пути $\varphi$ неустойчивы тогда и только тогда, когда существует путь $\psi: I \rightarrow \tilde{G}$ без трансверсальных самопересечений, такой что $\pi \circ \psi=\varphi$. (b)* Придумайте быстрый алгоритм распознавания, имеет ли данный путь в графе на плоскости трансверсальные самопересечения. D-8. Существует бесконечное количество путей $\varphi: I \rightarrow \mathbb{R}^{2}$ с устойчивыми самопересечениями, образами которых (a) являются (содержащие друг друга) деревья, (b) является "буква $Y$ " (см. рис. 4.Y), (с) (А. Халявин) является "буква $P$ " (см. рис. 4.P), пути не имеют точек возврата, и ни один из которых не является подпутем другого. D-9.* Минором графа называется граф, полученный из него несколькими операциями выкидывания (внутренности) ребра или стягивания ребра. Теорема Куратовского имеет следующую равносильную формулировку: граф планарен тогда и только тогда, когда у него нет миноров, изоморфных $K_{5}$ и $K_{33}$. Придумайте понятие минора пути и выясните, существует ли бесконечное количество (кусочно-линейных) путей $\varphi:[0,1] \rightarrow \mathbb{R}^{2}$ с устойчивыми самопересечениями, ни один из которых не является минором другого. Решите ту же задачу для класса путей, образами которых является триод или являются (вложенные друг в друга) деревья. ## Указания и решения к некоторым основным задачам. D-1. Следует из D2b. D-2. (a) Нарисуем граф $\varphi^{\prime}(I)$ на плоскости следующим образом. Поместим вершины графа $\varphi^{\prime}(I)$ в середины соответствующих рёбер исходного графа $G$. Рёбра графа $\varphi^{\prime}(I)$ будем рисовать, пользуясь следующим алгоритмом. Для каждой вершины $v$ графа $G$ сделаем следующее. Нарисуем на плоскости маленькую окружность с центром в этой точке $v$. Каждое ребро графа $G$, выходящее из вершины $v$, пересекает эту окружность в некоторой точке. Будем обходить окружность по часовой стрелке. Занумеруем рёбра графа $G$, выходящие из вершины $v$ в том порядке, в котором эти точки пересечения расположены на окружности: $v_{1}, v_{2}, \ldots, v_{n}$. Вблизи точки пересечения ребра $v_{i}$ с окружностью отметим на окружности $n-1$ точку $v_{i, 1}, v_{i, 2}, \ldots, v_{i, i-1}, v_{i, i+1}, \ldots, v_{i, n}$ в указанном порядке против часовой стрелки. Для каждого ребра $v_{i}^{\prime} v_{j}^{\prime}$ графа $\varphi^{\prime}(I)$ сделаем следующее. Нарисуем его в виде трёхзвенной ломаной, соединяющей середины рёбер $v_{i}$ и $v_{j}$ и проходящей через точки $v_{i, j}$ и $v_{j, i}$. Сделаем это для всех рёбер графа $\varphi^{\prime}(I)$. Покажем, что при этом нарисованный граф не будет иметь самопересечений. Предположим, что он имеет самопересечение. Очевидно, что оно лежит внутри одной из построенных окружностей. Значит, для некоторых рёбер $v_{i}^{\prime} v_{j}^{\prime}$ и $v_{k}^{\prime} v_{l}^{\prime}$ графа $\varphi^{\prime}(I)$ отрезки $v_{i, j} v_{j, i}$ и $v_{k, l} v_{l, k}$ пересекаются. А это возможно, только если исходный путь имеет трансверсальное самопересечение. То же самое решение можно записать на языке полянок и тропинок (см. стр. 5). На рис. 11 показано, как по системе $N$ полянок и тропинок для графа $G$ построить систему $N^{\prime}$ полянок и тропинок для графа $\varphi^{\prime}(I)$. На этом рисунке $a$ и $b$ обозначает пару смежных ребер графа $G, N_{(a)}$ и $N_{(b)}$ - тропинки вокруг ребер $a$ и $b, N_{a \cap b}$ - полянка вокруг их общей вершины. В качестве полянок вокруг вершин $a^{\prime}$ и $b^{\prime}$ возьмем тропинки $N_{(a)}$ и $N_{(b)}$, а в качестве тропинки, соответствующей ребру $a^{\prime} b^{\prime}$ - узкую дорожку $N_{a^{\prime} b^{\prime}}^{\prime}$, проходящую внутри полянки $N_{a \cap b}$. Если исходный путь не имеет трансверсальных самопересечений, то эти дорожки можно выбрать непересекающимися. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-118.jpg?height=503&width=625&top_left_y=1573&top_left_x=727) Рис. 11: Полянки и тропинки для производного графа (b) Указание. Удобнее использовать эквивалентную формулировку рассматриваемой проблемы, приведённую на стр. 5. Рассмотрим систему тропинок и дорожек, соответствующую графу $G$. Проведём внутри неё несамопересекающийся путь собаки, соответствующий пути $\varphi$. Наша цель - построить в системе тропинок и дорожек, соответствующей графу $\varphi^{\prime}(I)$, несамопересекающийся путь собаки, соответствующий пути $\varphi^{\prime}$. Будем считать, что границы полянок - это в точности окружности, построенные нами в решении пункта (а). Нарисуем внутри каждой полянки узкие дорожки вокруг отрезков $v_{i, j} v_{j, i}$, таких что ребро $v_{i}^{\prime} v_{j}^{\prime}$ принадлежит графу $\varphi^{\prime}(I)$. Построенная система тропинок и дорожек является системой полянок и тропинок для производной исходного пути. Посмотрим на исходный путь собаки, как на путь в этой новой системе тропинок и дорожек. В действительности, это почти тот путь, который нам нужен. Нам остаётся продеформировать его так, чтобы он не выходил за пределы дорожек. При этом в точках возврата (см. задачу С3c), а также в начальной и конечной точках, путь нужно дополнительно продеформировать так, чтобы он не заходил на соответствующую полянку. Подробное решение см. в статье [Sko03, Lemma 2.2.A]. (c) Данное утверждение доказывается аналогичным методом. Подробное доказательство приведено в статье [Sko03, Lemma 2.1]. (d) Теорема Минца следует из предыдущих двух пунктов таким образом: во-первых, если самопересечения пути неустойчивы, то все его производные также имеют неустойчивые самопересечения (по пункту (b)). Во-вторых, некоторая производная пути есть путь из единственного ребра, так как количество рёбер при дифференцировании строго убывает. Тогда, если бы все производные не имели трансверсальных пересечений, то отсюда бы следовало бы, что и самопересечения исходного пути неустойчивы по пункту (с). (е) Указание. Пусть самопересечения пути устойчивы. Тогда есть некоторая производная, у которой есть трансверсальное самопересечение. Трансверсальное пересечение образовано двумя путями, каждый из которых состоит из двух ребер. Возьмём два подпути исходного пути, которые переходят при дифференцировании в эти два пути. Пересечение построенных подпутей исходного пути устойчиво. Замечание. Построенные таким образом подпути могут иметь общие участки. Авторам задачи неизвестно, остается ли справедливой теорема о паре подпутей для подпутей без общих участков. D-3. (a) Ответ: для всех $m$, кроме $\{-1,0,1\}$. Указание. Для остальных доказательство можно построить так же, как и для двух оборотов, посчитав всё, что после первого оборота, одним больши́м куском, основываясь на непланарности $K_{3,3}$ : из центра круга выпустим 3 луча под углами $120^{\circ}$. В каждом луче соединим отрезком луча какую-то точку одного витка и какую-то другого. Если изначальный путь не самопересекался, то мы нарисовали на плоскости $K_{3,3}$ без самопересечений. (b) Указание. Обратим внимание на следующий тип путей - пары подпутей исходного пути, где мы проходим по ним дважды, то есть либо отображения подпутей идут по одному и тому же набору рёбер в одном и том же порядке либо в противоположном. Среди таких путей нас интересуют максимальные, то есть те, что нельзя продолжить ребром в какуюто сторону так, чтобы это ребро продолжало оба подпути. Для завершения доказательства достаточно заметить два факта: что если таких нет, то есть любой такой путь хоть в какуюто сторону, но можно продолжить, то это обмотка, и что при дифференцировании длины таких путей уменьшаются, так что за число дифференцирований, равное максимальной длине такого пути, они все исчезнут. Подробное решение см. в статье [Sko03, Lemma 2.3]. $(\mathrm{c}, \mathrm{d})$ Аналогично пунктам (c) и (d) предыдущей задачи. Подробное доказательство приведено в статье [Sko03, Lemma 2.2.A и Lemma 2.1] Интересно обобщить критерии D2d и D3d на кусочно линейные отображения $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$, где $K$ - произвольный граф. В этом случае не известно быстрого алгоритма распознавания устойчивости самопересечений (частный случай разобран в английской версии статьи [Sko03, Theorem 1.5]). Устойчивость самопересечений отображения $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$ определяется аналогично устойчивости самопересечений пути. Оказывается, для таких отображений тоже можно определить производную, причем утверждения задач D2b и D3b остаются в силе. Определение производной. (см. рис. 10, а также часть рис. 12) Пусть дано симплициальное отображение $\varphi: K \rightarrow G$, то есть такое отображение, что каждое ребро графа $K$ отображается в прямолинейный отрезок на плоскости. Сначала построим граф $K_{\varphi}^{\prime}$, который будет областью определения производной $\varphi^{\prime}$. Под $\varphi$-компонентой графа $K$ мы подразумеваем любую связную компоненту $\alpha$ множества $\varphi^{-1} a$, отображаемую на $a$, для некоторого ребра $a \subset G$. Множество вершин графа $K_{\varphi}^{\prime}$ находится в 1 -1 соответствии с множеством всех $\varphi$-компонент. Для $\varphi$-компоненты $\alpha \subset K$ обозначим через $\alpha^{\prime} \in K_{\varphi}^{\prime}$ соответствующую вершину. Две вершины $\alpha^{\prime}$ и $\beta^{\prime}$ соединены ребром в графе $K_{\varphi}^{\prime}$, если и только если $\alpha \cap \beta \neq \emptyset$. Производная $\varphi^{\prime}: K_{\varphi}^{\prime} \rightarrow G^{\prime}$ - это симплициальное отображение, определенное на вершинах графа $K_{\varphi}^{\prime}$ формулой $\varphi^{\prime} \alpha^{\prime}=(\varphi \alpha)^{\prime}$. В дальнейшем заменим $\varphi^{\prime}$ на сюръективное ограничение $\varphi^{\prime}: K_{\varphi}^{\prime} \rightarrow \varphi^{\prime} K_{\varphi}^{\prime}$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-120.jpg?height=692&width=1020&top_left_y=480&top_left_x=536) Рис. 12: Построение производной в два шага Оказывается, что для произвольного графа $K$ остается верным, что если если отображение $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$ имеет неустойчивые самопересечения, то и его производная имеет неустойчивые самопересечения. Рисунки 12 и 13 поясняют основную идею доказательства этого факта. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-120.jpg?height=488&width=799&top_left_y=1481&top_left_x=640) Рис. 13: Аппроксимация производной Оказывается, что и у задач D2c, D3c тоже есть аналоги в случае произвольного графа $K$. А именно, если производная $\varphi^{\prime}$ имеет неустойчивые самопересечения, то и полупроизводная $\bar{\varphi}$ (определение которой ясно из рис. 12) также имеет неустойчивые самопересечения. Идею доказательства этого утверждения поясняет рис. 14. D-5. (а) В этой ситуации теорема Минца (см. задачу D2d) остается справедливой; доказательство аналогично. (b) Покажем, что в этой ситуации теорема Минца перестает быть верной [Sko03, Пример 3.3]. Рассмотрим пару путей, изображенных на рис. 15. Здесь $K, L \cong I-$ графы с вершинами $k_{1}, \ldots, k_{5}$ и $l_{1}, \ldots, l_{7}$, и $G$ - граф с вершинами $a_{1}, \ldots, a_{6}$ и ребрами $a_{1} a_{2}, a_{1} a_{3}$, $a_{1} a_{4}, a_{1} a_{5}, a_{2} a_{3}, a_{2} a_{4}$ и $a_{2} a_{6}$. Симплициальные пути $\varphi, \psi$ задаются формулами $\varphi k_{1}=a_{1}$, $\varphi k_{2}=a_{2}, \varphi k_{3}=a_{3}, \varphi k_{4}=a_{1}, \varphi k_{5}=a_{2}$ и $\psi l_{1}=a_{5}, \psi l_{2}=a_{1}, \psi l_{3}=a_{2}, \psi l_{4}=a_{4}, \psi l_{5}=a_{1}$, ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-121.jpg?height=488&width=1422&top_left_y=133&top_left_x=336) Рис. 14: Аппроксимация "полупроизводной" $\psi l_{6}=a_{2}, \psi l_{7}=a_{6}$. Легко проверить, что пересечение данной пары путей устойчиво, а их производных - неустойчиво, при этом у них нет трансверсальных пересечений. Нахождение удобного критерия представляет собой открытую проблему. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-121.jpg?height=648&width=1179&top_left_y=1018&top_left_x=450) Рис. 15: Пара путей с устойчивым пересечением ## Литература [Mi97] P. Minc. Embedding simplicial arcs into the plane, Topol. Proc. 1997. 22. 305-340. [RS98] D. Repovš and A. Skopenkov. A deleted product criterion for approximability of a map by embeddings, Topol. Appl. 1998. 87, 1-19. [RS02] Д. Реповш и А. Скопенков. Теория препятствий для начинающих, Мат. Просвещение. 2002. 6. 60-77. [KoSk00] П. Кожевников, А. Скопенков. Узкие деревья на плоскости. Материалы 12-й Летней конференции Турнира городов. [Si69] K. Sieklucki. Realization of mappings, Fund. Math. 1969. 65. 325-343. [Sk03] M. Skopenkov. On approximability by embeddings of cycles in the plane, Topol. Appl. 134 (2003), p. 1-22. Русский перевод: М. Скопенков, Об аппроксимируемости вложениями циклов на плоскости, http://arxiv.org/abs/0808.1187. ## Stability of intersections of paths in the plane P. Dergach, I. Netay, A. Skopenkov, M. Skopenkov He wants it rare, but something's there that holds him back from the attack Accept and Deaffy, Dogs on Leads. The main results are criteria for stability of intersections of paths and cycles in the plane (problems D-2.d, D-3.d; the proof is outlined in the problems D-2.abc and D-3.abc; all the definitions are given below). The criteria are formulated in terms of derivation of graphs and cycles. $^{1}$ In the following problems, let us agree that for any assertion formulated as a problem, a proof of this assertion is required. The parts A, B, C are set before a half-way finish, the rest are set after it. $\mathbf{0}$. The city $N$ in the plane is formed by several squares (of side 1 ) joined by disjoint roads (straight line segments). Suppose that there is a path passing through each road exactly once (this path can go through certain squares several times). Prove that there is a path without self-intersections passing through each road exactly once. ## A. Stability of intersections of a pair of paths. ${ }^{2}$ Problem on the stability of intersections of a pair of paths. Two hunters hunt in a forest. Each leads a dog on a short lead. The dogs obey the hunters and move as they are commanded. If one dog intersects the path of another, it barks and scares off a game. Having fixed paths of the hunters, how to determine whether these hunters could maintain successful hunting (in this case the intersection of paths of the hunters is called unstable ${ }^{3}$ )? We shall assume that all the paths of hunters and dogs are formed by finite unions of rectilinear arcs (i.e. are piecewise-linear). For such paths of hunters there is a slow full search algorithm for recognition of the stability of intersections. Finding a fast algorithm is an unsolved problem. In the following problems we assume that the hunters move along a certain system of roads in a plane. We assume that both the hunter and the dog are points (not necessarily distinct). The lengths of leads are $1 \mathrm{~m}$. A-1. Two hunters move along the road that has the form of a segment of $1 \mathrm{~km}$ length. The hunters may change the direction of movement. Prove that independently of hunters' movement one dog could move without intersections with the path of another dog. Example. Two hunters moved (with constant speed and permanent direction) along rectilinear roads of length $2 \mathrm{~km}$; the roads intersect at their middle points and form right angles at the intersection point (fig. 1.a, where $\varphi\left(I_{1}\right)$ and $\varphi\left(I_{2}\right)$ are paths of hunters, and $f_{1}$ and $f_{2}$ are possible paths of dogs). Then one dog did intersect the path of the other. A-2. A system of roads has the form of letter "H" (see fig. 1.b); the lengths of $A B, B C$, $B E, D E, E F$ are $1 \mathrm{~km}$. The first hunter moved along the path $A B E F$ and the second along $C B E D$. Then one dog did intersect the path of the other.[^4] ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-123.jpg?height=614&width=623&top_left_y=138&top_left_x=197) a ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-123.jpg?height=474&width=711&top_left_y=291&top_left_x=867) b Figure 1: The transversal intersection and a path in letter " $H$ " For proving that certain paths of dogs must intersect the following theorem could be useful. You are allowed to use it without proof. A cycle is a path whose starting point and endpoint are the same; it is not specified which point is the starting point (see Figure). Note that self-intersection points (i.e. points corresponding to the dog's intersection of its own path) are not counted as intersection points. Two piecewise-linear paths (or cycles) intersect transversally, if near each intersection point these paths look like two paths $\varphi\left(I_{1}\right)$ and $\varphi\left(I_{2}\right)$ in Fig. 1. Even Number Theorem. Two (piecewise linear) cycles in the plane that intersect transversally have an even number of intersection points. A-3. A system of roads has the form of letter " $Y$ ". This letter is formed by three straightline segments making angles $2 \pi / 3$ at the common point (fig. 4.Y). Give an example of two hunters' paths such that the hunters could not maintain successful hunting. ## B. Stability of self-intersections of path and of cycle. Problem of the stability of self-intersections of path. One hunter walks in a forest and leads a dog on a short lead. The dog obeys the hunter and moves as it is commanded. If the dog intersects its path, it barks and scares off a game. How to determine, for the fixed path of the hunter, whether he could maintain successful hunting? (If he can, the self-intersection of the path of hunter is called unstable. $)^{4}$ The main result of the presented problems is a fast algorithm recognizing the stability of self-intersections ${ }^{5}$. B-1. (a) A hunter moves along the road that has the form of a segment of length $1 \mathrm{~km}$. The hunter may change the direction of movement. Prove that independently of the hunter's movement the dog can move without intersecting its trace. (b) The same for the road that has the form of the circle of radius $1 \mathrm{~km}$.[^5] B-2. (a) Each path without self-intersections has unstable self-intersections (this should not confuse the reader). (b) If self-intersections of a path are unstable, then the same is true for all subpaths of this path ${ }^{6}$. (c) If intersections of a certain pair of subpaths of a path are stable, then self-intersections of this path are stable. (d) There exists a path that does not contain transversal intersections but does have stable self-intersections. One of the main results of this sequence of problems is the following theorem (try to prove it, but not too hard!) Theorem on subpaths. A path in the plane has stable self-intersections if and only if there are two subpaths of this path that have stable intersections. B-3. Assume that a system of roads form certain graph in the plane, so that the edges are straight line segments of length at least $1 \mathrm{~km}$ and the distance from every vertex to every edge that does not contain this vertex is more than $10 \mathrm{~m}$. Suppose that the hunter walked along this system of roads, passing along every road only once and changing the direction of his movement only at vertices. Prove that the path of the hunter has stable self-intersections if and only if the path contains transversal self-intersections. B-4. There exists an algorithm recognizing stability of self-intersections for given path in the plane. B-5. (a) The hunter moved along a circular path of diameter $1 \mathrm{~km}$ (with constant speed and permanent direction along the circle) and winded twice along the circle. He lead a dog on a lead of length $1 \mathrm{~m}$. The dog returned to the starting point at the end of the movement. Prove that the dog necessarily intersect its own path (in a certain moment of time different from the final movent, fig. 2). (b) Is the analogue (a) correct if we do not suppose that the dog returned to the starting point at the final moment? (c) Prove the analogue of (a) for the case when the hunter winded three times along the circle. (d) For which number of windings in the analogue of (a) the dog necessarily intersected its own path? (e) Suppose that the road is the segment of length $1 \mathrm{~km}$. Prove that independently of the hunter's movement the dog can move without self-intersections and so that at the final moment of movement it will return to the starting point. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-124.jpg?height=323&width=328&top_left_y=1946&top_left_x=881) Figure 2: A path of the dog ## C. Derivative of graphs and paths. [^6]The derivative $G^{\prime}$ of a graph $G$ is a graph whose vertices are in one-to-one correspondence with the edges of the graph $G$. Vertices $e^{\prime}$ and $f^{\prime}$ that correspond to edges $e$ and $f$ are connected by an edge in the graph $G^{\prime}$, if the edges $e$ and $f$ have a common vertex (fig. 3). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-125.jpg?height=492&width=1156&top_left_y=343&top_left_x=470) Figure 3: The derivative of a path in a graph C-1. Draw the derivatives graphs of (fig. 4): (a) an arc with $n$ edges; (b) a circle with $n$ edges; (c) a star with $n$ rays; (d) letter " $H$ ". ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-125.jpg?height=388&width=1264&top_left_y=1164&top_left_x=410) Figure 4: Differentiate us! C-2. A graph is called planar if it can be drawn in the plane without self-intersections. The derivative of a planar graph is not necessarily planar. A path in a graph $G$ is a sequence of vertices $v_{0}, v_{1}, \ldots, v_{n}$ such that $v_{i}$ and $v_{i+1}$ are connected by an edge in the graph $G$. A path (a cycle) $v_{0}, \ldots, v_{n}$ is called an Euler path (cycle) if it passes through each edge of the graph $G$ only once, i.e. if each edge of the graph $G$ is present among the edges $v_{0} v_{1}, v_{1} v_{2}, \ldots, v_{n-1} v_{n}$ only once. Suppose that a path $\varphi$ in a graph $G$ is defined by a sequence $v_{0}, v_{1}, \ldots, v_{n}$ of vertices. Consider the sequence $\left(v_{0} v_{1}\right)^{\prime}, \ldots,\left(v_{n-1} v_{n}\right)^{\prime}$ of vertices of the derivative graph. In this sequence there could be the same consecutive vertices. Substitute each set of the same consecutive vertices by one vertex. In this way we obtain a path $\varphi^{\prime}$ that is called the derivative of the path $\varphi$. The derivative of a pair of paths is a pair of paths defined analogously. Example. Let $A$ be the 'letter A' graph, i.e. the graph with vertices $a_{1}, a_{2}, a_{3}, a_{4}, a_{5}$ and edges $a_{1} a_{2}, a_{2} a_{3}, a_{2} a_{4}, a_{3} a_{4}, a_{4} a_{5}$ (fig. 3). Denote by $b_{1}=\left(a_{1} a_{2}\right)^{\prime}, b_{2}=\left(a_{2} a_{3}\right)^{\prime}, b_{3}=\left(a_{2} a_{4}\right)^{\prime}$, $b_{4}=\left(a_{3} a_{4}\right)^{\prime}, b_{5}=\left(a_{4} a_{5}\right)^{\prime}$ the vertices of the derivative of $A$. Let $\varphi$ be a path $a_{1} a_{2} a_{4} a_{2} a_{3} a_{4} a_{2} a_{4} a_{5}$ in the graph $A$. Then the derivative $\varphi^{\prime}$ is the path $b_{1} b_{3} b_{2} b_{4} b_{3} b_{5}$. C-3. (a) Find the first and the second derivatives of the paths and the pairs of paths from fig. 3 and fig. 1 (and also of your paths constructed in problems A3 and B2d). (b) Given a path with $n$ vertices, the number of vertices in the derivative path does not exceed $n-1$. (c) We say that a path $\varphi$ has a return point, if $v_{i} v_{i+1}=v_{i+1} v_{i+2}$ for some integer $i$. A path (cycle) $\varphi$ is an Euler path if and only if it does not have return points and $\varphi^{\prime}$ does not have self-intersections. ## Formal definitions. Let us give an equivalent formulation of problem B-5 (the equivalence is proved in [Mi97]). Consider two clearings (i.e. two disks) in the plane connected by two paths (i.e. strips) $a$ and $b$, as it is shown on Fig. 2. The dog moved on the clearings and paths, and at last returned into the starting point. Each time when the dog moved from the clearing to the path, it wrote down the letter corresponding to this path. It is stated in the problem B-5 that if we get the word $a b a b$ then the dog necessarily intersected its own path (at a certain moment of time different from the final moment). The other problems can be reformulated in a similar way. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-126.jpg?height=485&width=619&top_left_y=908&top_left_x=730) Figure 5: Two clearings We shall give the formal definition of stability of intersections and self-intersections. Denote a segment by $I=[0,1]$, a circle (i.e. a segment with the ends glued) by $S^{1}$ and the plane by $\mathbb{R}^{2}$. A piecewise-linear path in the plane is a mapping $\varphi: I \rightarrow \mathbb{R}^{2}$ for which there exist points $0=v_{0}0$ there exists a path $\varphi: I \rightarrow \mathbb{R}^{2}$ without self-intersections such that the distance between the points $f(x)$ and $\varphi(x)$ is less than $\varepsilon$ for each point $x \in I$ ). The stability of self-intersections of the cycle $\varphi: S^{1} \rightarrow \mathbb{R}^{2}$ is defined analogously. A pair of paths $\varphi_{1}, \varphi_{2}: I \rightarrow \mathbb{R}^{2}$ has unstable intersections (or allows the removal of selfintersections by a small perturbation), if there exist a non-intersecting pair of paths arbitrarily close to our paths (i.e. if for each $\varepsilon>0$ there exist non-intersecting paths $f_{1}, f_{2}: I \rightarrow \mathbb{R}^{2}$ such that distance between the points $f_{i}(x)$ and $\varphi_{i}(x)$ is less than $\varepsilon$ for each point $x \in I$ and for each $i=1,2$. For instance, the transversal intersection (Fig. 1) of two paths is stable. For example, the problem B-1ab can be reformulated in this language as follows: If the image $\varphi(I)$ of a path $\varphi: I \rightarrow \mathbb{R}^{2}$ is a segment or a circle then the self-intersections are unstable. Hints and solutions of some problems from parts A, B, C. A-1. Draw a straight line along our road. It splits the plane into two half-planes. The first hunter orders his dog to move in one of the given half-planes, and the second hunter orders his dog to move in another half-plane. Then the dogs' paths do not intersect. A-2. Assume that the dogs may move so as not to intersect each other's paths. Let $A^{\prime} F^{\prime}$ and $C^{\prime} E^{\prime}$ be the paths of the dogs. We close these paths by adding to them broken lines $F^{\prime} X A^{\prime}$ and $E^{\prime} X C^{\prime}$ shown in the figure 6 . The maximal distance between the hunter and the dog is much less than the distances between each two of the points $A, C, E, F$, hence the broken line $F^{\prime} X A^{\prime}$ does not intersect the path $C^{\prime} E^{\prime}$, and the broken line $E^{\prime} X C^{\prime}$ does not intersect the path $A^{\prime} F^{\prime}$. Thus the two cycles $A^{\prime} F^{\prime} X A^{\prime}$ and $C^{\prime} E^{\prime} X C^{\prime}$ transversally intersect in the unique point $X$. And according to the Even number Theorem the number of their intersection points has to be even. The obtained contradiction proves that the dogs' paths do intersect. Examples to the problems A3 and B2d are presented in Fig. 6, where (for clarity) is shown not the initial path itself, but a general position path close to the initial path is drawn. However, see [Mi97, Sk03]. Hint: it is possible to reduce this problem to non-planarity of the Kuratowski graphs $K_{5}$ and $K_{3,3}$. The dotted line on the Fig. 6 will help to do this. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-127.jpg?height=277&width=374&top_left_y=1092&top_left_x=567) a) ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-127.jpg?height=349&width=325&top_left_y=1456&top_left_x=617) c) ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-127.jpg?height=254&width=540&top_left_y=1118&top_left_x=969) b) ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-127.jpg?height=257&width=511&top_left_y=1551&top_left_x=955) d) Figure 6: Paths with stable self-intersections B-1. (a) Direct the axis $0 x$ along the road, and take an axis $0 y$ perpendicular to the road. Assume that the hunter's coordinate (expressed in meters) is given by the function $x(t)$, and the hunting time changes in the range from 0 to $T$. The hunter orders the dog to move along the graph of the function $x(t)$, compressed towards the axis $0 x$, i.e. let the dog's coordinates at the moment $t$ be $(x(t), t / T)$. It is easy to see that in this case the dog will not intersect its own path, and in every moment of time it will be closer than $1 \mathrm{~m}$ from the hunter. (b) We will act analogously to the item (a): the dog moves so that in each moment of time it is on the ray that is directed from the center of the circle to the point where the hunter is situated, and its distance from the hunter is $t / T \mathrm{~m}$ (where $T$ is the total hunting time). Then the dog will not intersect its own traces, and in every moment of time it will be closer than 1 $\mathrm{m}$ from the hunter. B-2. (a) Suppose that the dog follows in the tracks of the hunter, i.e. in each moment of time the dog and the hunter are in the same point (it is not prohibited by the conditions). The hunter's path will not self-intersect, and thus the dog's path also will not self-intersect. (b) The self-intersections of the path are unstable, and so the dog can move without intersecting its own path. We shall consider the dog's movement only in the segment of time that corresponds to the chosen subpath. It also does not self-intersect. That's why the chosen subpath has unstable self-intersections too. (c) Suppose that self-intersections of the path are unstable. Then the dog may move without intersecting its path. Concern the dog's movement only in two segments of time that correspond to the chosen subpaths. These two paths do not intersect. Thus self-intersections of the corresponding pair of subpaths are unstable. B-3. It is clear that if the hunter's path contains transversal self-intersection then the hunter's self-intersections are stable. Prove that if the path does not contain transversal selfintersections in the considered case then the hunter's self-intersections are unstable. We shall draw a circle with the radius $1 \mathrm{~m}$ around every point of connection of roads. Let the dog follows the tracks of the hunter all the time when the hunter is outside of these circles. At the moment when the hunter begins to move inside a certain circle, the dog takes a short cut, moving along the chord instead of a pair of radii, as it is shown on the figure. The built path does not self-intersect. Indeed, the hunter passes along each road only once, so the dog does not intersect its own path outside the circles. If the dog intersects its path inside a circle then some two of chords drawn by us intersect. And it is possible only if the hunter's path has a transversal self-intersection in this point. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-128.jpg?height=485&width=491&top_left_y=1294&top_left_x=797) Figure 7: The dog takes a short cut B-4. See the formulation of the problem D7 (simple but slow algorithm) and Minc theorem D2d (more complex but fast algorithm). B-5. (a) Assume that the dog may move without intersecting its own traces. Let $A$ be a certain point of the circle. Consider the ray $O A$ that is directed from the center of the circle to the point $A$. It is obvious that the dog's cycle has intersected the ray $O A$ at least twice: at least once when the hunter made the first cycle and at least once when he made the second cycle. We mark on this ray all its intersection points with the dog's path. It is obvious that there exist two "adjacent" marked points $A^{\prime}$ and $A^{\prime \prime}$ (i.e. such points that there are no marked points on the segment between these two ones), one of them is related to the moment when the hunter made his first turn along the circle, and the other - to the moment when the hunter made the second turn. Let us "tear" the dog's cycle in the points $A^{\prime}$ and $A^{\prime \prime}$ and add a pair of paths $p^{\prime}$ and $p^{\prime \prime}$ situated "near" the segment $A^{\prime} A^{\prime \prime}$ which intersect transversally in the point $X$, as it is shown on the figure. As a result, we will obtain from the dog's cycle a pair of cycles that intersect transversally in the unique point $X$. But, according to the Even number Theorem, the quantity of the intersection points of two cycles must be even. The obtained contradiction proves that the dog must intersect its own path. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-129.jpg?height=502&width=994&top_left_y=294&top_left_x=542) Figure 8: Transformation of the cycle (b) Answer: in this case the dog may move without intersecting its own path. (c) The proof repeats word-for-word our reasoning from the item (a). (d) Answer: for every number of turns (in any direction) not less than two. The proof repeats word-for-word our reasoning from the item (a). (e) We may consider (without lack of generality) that the hunter was in all of the points of the road, including its ends (otherwise we shall just diminish the segment of the road to obtain this condition). In this case it is possible to suppose that the hunter's cycle begins and finishes at one of the ends of the segment. Let the dog move along the graph of the hunter's movement, compressed towards the road (analogously to the solution of the problem B1a). Then we have only to close the dog's path, adding a broken line situated near the end of the segment, as it is shown on the figure. The constructed cycle does not self-intersect. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-129.jpg?height=122&width=986&top_left_y=1538&top_left_x=548) Figure 9: We close the dog's path C-1. (a) Arch with $n-1$ edges; (b) circle with $n$ edges; (c) full graph with $n$ vertices; (d) graph made from the two triangles with a unique common vertex. C-2. Example: a star with 5 rays - a planar graph which derivative is a non-planar graph (a full graph with 5 vertices). C-3. (a) E. g., see fig. 10. (b) Let the initial path consist of $n$ vertices $v_{1}, v_{2}, \ldots, v_{n}$. Then there are exactly $n-$ 1 vertices in the sequence $\left(v_{1} v_{2}\right)^{\prime},\left(v_{2} v_{3}\right)^{\prime}, \ldots,\left(v_{n-1} v_{n}\right)^{\prime}$. To construct an arbitrary path we (maybe) exclude some vertices from the given path. As a result we get the path containing not more than $n-1$ vertices. (c) Let $\psi$ be the Euler path $v_{0}, v_{1}, \ldots, v_{n-1}, v_{n}=v_{0}$. Since the Euler path comes along every edge only once, there are not two same edges in the sequence $\left(v_{1} v_{2}\right),\left(v_{2} v_{3}\right), \ldots,\left(v_{n-1} v_{n}\right)$. Thus in the sequence of the vertices of the derivative of the given path $\left(v_{1} v_{2}\right)^{\prime},\left(v_{2} v_{3}\right)^{\prime}, \ldots,\left(v_{n-1} v_{n}\right)^{\prime}$ there are not two same vertices. So the path $\psi^{\prime}$ does not self-intersect. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-130.jpg?height=550&width=1178&top_left_y=132&top_left_x=450) Figure 10: The second derivative of a path ## D. Main problems D-1. Self-intersections of the paths in Figure 6 are stable. For a piecewise linear path $\varphi: I \rightarrow \mathbb{R}^{2}$ the image $\varphi(I)$ can be considered as a graph with vertices $\varphi\left(v_{1}\right), \ldots, \varphi\left(v_{n}\right)$. From every piecewise linear path $\varphi$ a path in the graph $\varphi(I)$ can be constructed in a unique way. Vice versa, suppose that the graph $G$ be drawn in the plane without self-intersections so that all its edges are straight line segments. For every path $v_{1}, \ldots, v_{n}$ in the graph $G$ we shall build a path $\varphi:[0,1] \rightarrow G$ in the plane, setting $\varphi\left(\frac{i}{n}\right)=v_{i}$ for every $i=0, \ldots, n$ and extending $\varphi$ linearly on the segments $\left[\frac{i}{n}, \frac{i+1}{n}\right]$. The path $\varphi:[0,1] \rightarrow \mathbb{R}^{2}$ that is constructed in this way from a certain path in the graph $G$, is called simplicial. D-2. (a) If a path $\varphi$ does not contain transversal self-intersections then the graph $\varphi^{\prime}(I)$ is planar. Fix the "natural" embedding of the graph $\varphi^{\prime}(I)$ into the plane (invent the definition yourself). Then $\varphi^{\prime}: I \rightarrow \varphi^{\prime}(I)$ is a certain path in the plane. (b) If self-intersections of a path $\varphi$ are unstable then self-intersections of the path $\varphi^{\prime}$ are also unstable. As a corollary, obtain the proof of the statement from problem A2. (c) If a path $\varphi$ does not contain transversal self-intersections and self-intersections of the path $\varphi^{\prime}$ are unstable then self-intersections of the path $\varphi$ are also unstable. (d) The Minc Theorem. Self-intersections of a simplicial path $\varphi: I \rightarrow G$ containing $n$ points are stable if and only if for a certain $k=0, \ldots, n$ its $k$-th derivative $\varphi^{(k)}$ contains a transversal self-intersections. (e) Prove the Theorem on two subpaths. D-3. (a) For which $m$ self-intersections of the cycle winding of degree $m$ (Figure 2 for $m=2$ ) are stable? (b) For every cycle $\varphi$ there is $k$ such that $\varphi^{(k)}$ is a winding. (c) The statements of the problems D2abc remain true for a cycle $\varphi$. (d) Theorem. The self-intersections of a simplicial cycle $\varphi: S^{1} \rightarrow G$ that contains $n$ points are stable if and only if for certain $k=0, \ldots, n$ its $k$-th derivative $\varphi^{(k)}$ either contains a transversal self-intersection or is a standard winding of degree $m \neq 0, \pm 1$. D-4.* Consider a cycle $\varphi$. How to determine $m$ such that $\varphi^{(\infty)}$ is a winding of degree $m$ ? D-5.* Consider a set of paths in a given graph in the plane. Formulate and prove a criterion of approximating this set by a set of (a) non-intersecting and non-self-intersecting paths; (b)* non-intersecting (but possibly self-intersecting) paths. D-6. How to construct an "integral" of a given path? Use this to invent new examples of paths with stable self-intersections. D-7. (a) For a simplicial path $\varphi: I \rightarrow G \subset \mathbb{R}^{2}$ substitute every edge of the graph $G$ by $k$ close edges if the path $\varphi$ passes along this edge $k$ times. Denote the constructed graph by $\tilde{G} \subset \mathbb{R}^{2}$. Denote by $\pi: \tilde{G} \rightarrow G$ the projection mapping to an edge $a b$ the union of multiple edges corresponding to the edge $a b$ of $G$. Self-intersections of the path $\varphi$ are stable if and only if there exists a path $\psi: I \rightarrow \tilde{G}$ without transversal self-intersections such that $\pi \circ \psi=\varphi$. (b)* Invent a fast algorithm for recognition whether a given path in the graph in the plane has transversal self-intersections. D-8. There exists an infinite set of paths $\varphi: I \rightarrow \mathbb{R}^{2}$ with stable self-intersections such that their images: (a) are trees (containing each other); (b) are "letter Y" (see the Figure 4.Y); (c) are (A. Chalyavin) "letter P" (see the Figure 4.P), the paths do not have return points and no one of them is a subpath of any other. D-9.* The minor of a graph is a graph that is obtained from the initial graph by several operations of throwing away (the interior of) an edge or gluing ends of an edge. The Kuratovski Theorem has the following equivalent formulation: a graph is planar if and only if it does not have minors isomorphic to $K_{5}$ and $K_{3,3}$. Invent a notion of a minor of a path and find out whether there exists an infinite set of (piecewise linear) paths $\varphi:[0,1] \rightarrow \mathbb{R}^{2}$ with stable self-intersections so that no path of these is a minor of any other graph from this set. Solve the same problem for the set of paths whose images are triods or whose images are trees embedded into each other. ## Hints and solutions of some main problems. ## D-1. Follows by D2b. D-2. (a) Draw the graph $\varphi^{\prime}(I)$ in the plane as follows. Put the vertices of the graph $\varphi^{\prime}(I)$ in the middle points of corresponding edges of the initial graph $G$. Edges of the graph $\varphi^{\prime}(I)$ are drawn as follows. For each vertex $v$ of the graph $G$ make the following construction. Draw a small circle in the plane centered at this point $v$. Every edge of the graph $G$ issuing out of the vertex $v$ intersects this circle at a certain point. Go clockwise along the circle. Enumerate the edges of the graph $G$ issuing out of $v$ in their order along the circle: $v_{1}, v_{2}, \ldots, v_{n}$. Near the intersection point of the edge $v_{i}$ and the circle take $n-1$ points $v_{i, 1}, v_{i, 2}, \ldots, v_{i, i-1}, v_{i, i+1}, \ldots, v_{i, n}$ in the circle in this order counterclockwise. For each edge $v_{i}^{\prime} v_{j}^{\prime}$ of the graph $\varphi^{\prime}(I)$ make the following construction. Draw the edge as a broken line of three segments, so that the edge connects the middle point of the edge $v_{i}$ with the middle point of the edge $v_{j}$ and passes through the points $v_{i, j}$ and $v_{j, i}$. We shall show that the drawn graph $\varphi^{\prime}(I)$ does not have self-intersections. In the opposite case the self-intersection point is inside one of the constructed circles. Thus for certain edges $v_{i}^{\prime} v_{j}^{\prime}$ and $v_{k}^{\prime} v_{l}^{\prime}$ in the graph $\varphi^{\prime}(I)$ the segments $v_{i, j} v_{j, i}$ and $v_{k, l} v_{l, k}$ intersect. But this is possible only if the initial path has a transversal self-intersection. The same solution can be obtained by construction of the system of discs and strips for the graph $\varphi^{\prime}(I)$ (cf. page 5). Figure 11 may help to realize this idea. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-132.jpg?height=505&width=646&top_left_y=1215&top_left_x=725) Figure 11: Discs and strips for the derivative graph (b) A detailed solution can be found in the article [Sko03, Lemma 2.2A]. (c) The given statement is proved using an analogous method. A detailed proof is given in the article [Sko03, Lemma 2.1]. (d) The Minc theorem is deduced from the previous two items in the following way: first, if self-intersections of the path are unstable then by (b) all the derivatives of the path also have unstable self-intersections. Second, certain derivative of the path is a path which consists of a unique edge, because the quantity of edges strictly decreases when we differentiate the path. So if each derivative does not have transversal self-intersections then by (c) self-intersections of the initial path are also unstable. (e) Hint. Suppose that self-intersections of the path are stable. Then there exists certain derivative having a transversal self-intersection. The transversal self-intersection is formed of 2 paths containing 2 edges each. Take two subpaths of the initial path whose derivatives are these paths. Then their self-intersection is stable. D-3. (a) Answer: Each $m$ except $\{-1,0,1\}$. A detailed solution is given in the article [Sko03, Lemma 2.3]. (c, d) Analogously to the items (c) and (d) of the previous problem. A detailed proof is given in the article [Sko03, Lemma 2.2.A and Lemma 2.1] It is interesting to generalize these results to the case of maps $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$, where the graph $K$ is arbitrary. This case is treated in [Sko03]. D-5. (a) A theorem analogous to the Minc theorem D2d is true. The proof is analogous. (b) The analogue of the Minc theorem is not true in this situation. A counterexample is shown in Fig. 12. Finding a fast algorithm for checking the stability of the intersection of two paths is an open problem. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-133.jpg?height=645&width=1176&top_left_y=594&top_left_x=454) Figure 12: A pair of paths with stable intersection ## References [Mi97] P. Minc. Embedding simplicial arcs into the plane, Topol. Proc. 1997. 22. 305-340. [RS98] D. Repovš and A. Skopenkov. A deleted product criterion for approximability of a map by embeddings, Topol. Appl. 1998. 87, 1-19. [RS02] D. Repovš and A. Skopenkov. The theory of obstacles for beginners, Mathematical Education, 2002. 6. 60-77. [Si69] K. Sieklucki. Realization of mappings, Fund. Math. 1969. 65. 325-343. [Sk03] M. Skopenkov. On approximability by embeddings of cycles in the plane, Topol. Appl. 134 (2003), p. 1-22. http://arxiv.org/abs/0808.1187 # ОБ АППРОКСИМИРУЕМОСТИ ВЛОЖЕНИЯМИ ЦИКЛОВ НА ПЛОСКОСТИ Михаил Скопенков #### Abstract Аннотация. Мы получаем критерий аппроксимируемости вложениями кусочно линейных отображений $S^{1} \rightarrow$ $\mathbb{R}^{2}$, аналогичный доказанному Минцем для кусочно линейных отображений $I \rightarrow \mathbb{R}^{2}$. Теорема. Пусть $\varphi: S^{1} \rightarrow \mathbb{R}^{2}$ - кусочно линейное отображение, которое является симплициальным для некоторой триангуляции $S^{1}$ c $k$ вершинами. Отображение $\varphi$ аппроксимируется вложениями, если и только если для каждого $i=0, \ldots, k$ его $i$-я производная $\varphi^{(i)}$ (определенная Минцем) не содержит трансверсальных самопересечений, и не является стандартной намоткой степени $\notin\{-1,0,1\}$. Мы выводим из результата Минца полноту препятствия Ван Кампена к аппроксимируемости вложениями кусочно-линейных отображений $I \rightarrow \mathbb{R}^{2}$. ## 1. ВВЕДЕНИЕ Кусочно линейное отображение $\varphi: K \rightarrow \mathbb{R}^{2}$ графа $K$ в плоскость аппроксимируется вложениями, если для каждого $\varepsilon>0$ существует отображение $f: K \rightarrow \mathbb{R}^{2}$ без самопересечений, $\varepsilon$-близкое к $\varphi$. В большей части статьи мы рассматриваем случай, когда $\varphi$ является путем или циклом, то есть, $K \cong I$ или $K \cong S^{1}$. Пример 1.1. 12 Стандартная $d$-намотка $S^{1} \rightarrow S^{1} \subset \mathbb{R}^{2}$ аппроксимируется вложениями в плоскость, если и только если $d \in\{-1,0,1\}$. Можно доказать также, что симплициальное отображение $S^{1} \rightarrow S^{1}$ аппроксимируется вложениями, если и только если его степень $d \in\{-1,0,1\}$ (см. Теорему 1.3). Трансверсальным самопересечением кусочно линейного отображения $\varphi: K \rightarrow \mathbb{R}^{2}$ называется пара непересекающихся дуг $i, j \subset K$, таких что $\varphi i$ и $\varphi j$ пересекаются на плоскости трансверсально. Пример 1.2. Эйлеров путь или цикл в графе на плоскости аппроксимируется вложениями, если и только если он не имеет трансверсальных самопересечений (следовательно, у любого эйлерова графа на плоскости есть эйлеров цикл, аппроксимируемый вложениями). Понятие аппроксимируемости вложениями появилось в исследованиях вложимости компактов в $\mathbb{R}^{2}$ (см. [12, 15], актуальные обзоры можно найти в статьях [7, §9], [2, §4], [8, §1], мы вернемся к этому вопросу еще раз в конце §11). Существует алгоритм проверки того, является ли данное симплициальное отображение аппроксимируемым вложениями (см. [13). Более удобный для применения критерий аппроксимируемости вложениями симплициального пути на плоскости был доказан в статье 6] (Теорема 1.3 I ниже, обобщающая Пример 1.2). Главный результат этой статьи - аналогичный критерий для аппроксимируемости вложениями цикла на плоскости (Теорема 1.3.S ниже, также обобщающая Пример 1.2). Эти критерии утверждают, что, в некотором смысле, трансверсальное самопересечение - единственное препятствие к аппроксимируемости вложениями. Ясно, что буквально это не верно [12, и нет никакого критерия для рассматриваемой проблемы, аналогичного критерию Куратовского. Мы формулируем наш критерий (Теорему [1.3) в терминах производной пути [5, 6] ("операция d"). Дадим определение этого понятия (см. иллюстрацию 1). Сначала определим производную $G^{\prime}$ графа $G$ (это - синоним для реберного графа и двойственного графа). Множество вершин графа $G^{\prime}$ находится в $1-1$ соответствии с множеством ребер графа $G$. Для ребра $a \subset G$ обозначим через $a^{\prime} \in G^{\prime}$ соответствующую вершину. Вершины $a^{\prime}$ и $b^{\prime}$ в графе $G^{\prime}$ соединены ребром, если и только если ребра $a$ и $b$ являются смежными в графе $G$. Отметим, что производные $G^{\prime}$ и $H^{\prime}$ гомеоморфных, но не изоморфных графов $G$ и $H$ не обязательно гомеоморфны. Теперь пусть $\varphi$ - путь в графе $G$, заданный последовательностью своих вершин $x_{1}, \ldots, x_{k} \in G$, где вершины $x_{i}$ и $x_{i+1}$ соединены ребром. Тогда $\left(x_{1} x_{2}\right)^{\prime}, \ldots,\left(x_{k-1} x_{k}\right)^{\prime}$ является последовательностью вершин графа $G^{\prime}$. В этой последовательности заменим каждый отрезок вида $$ \left(x_{i} x_{i+1}\right)^{\prime},\left(x_{i+1} x_{i+2}\right)^{\prime}, \ldots,\left(x_{j-2} x_{j-1}\right)^{\prime},\left(x_{j-1} x_{j}\right)^{\prime} $$ 1991 Mathematics Subject Classification. 57Q35 (Primary); 54C25, 57M20 (Secondary). Key words and phrases. Аппроксимируемость вложениями, препятствие Ван Кампена, реберный граф, производная графа, производная симплициального отображения, операция $d$, трансверсальное самопересечение, стандартная $d$-намотка, симплициальное отображение, утолщение графа. Автор частично поддержан грантом ИНТАС 06-1000014-6277, грантами Российского Фонда Фундаментальных Исследований 05-01-00993-а, 06-01-72551-НЦНИЛ-а, 07-01-00648-а, Грантом Президента Российской Федерации для государственной поддержки ведущих научных школ Российской Федерации, проект НШ-4578.2006.1, программой Министерства Образования и Науки "Развитие научного потенциала высшей школы", проект РНП 2.1.1.7988, Фондом поддержки молодых ученых "Конкурс Мёбиуса". ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-135.jpg?height=557&width=1194&top_left_y=178&top_left_x=431) Рис. 1. Производные графов и путей где $\left(x_{i} x_{i+1}\right)^{\prime}=\left(x_{i+1} x_{i+2}\right)^{\prime}=\cdots=\left(x_{j-1} x_{j}\right)^{\prime}$, единственной вершиной (то есть заменим несколько идущих подряд одинаковых вершин на одну вершину). Полученная последовательность вершин определяет путь в графе $G^{\prime}$. Этот путь $\varphi^{\prime}$ называют производной пути $\varphi$. Любой 5 -од (то есть конус над 5 точками) является планарным графом, чья производная является графом Куратовкого, то есть непланарным графом. Но если $G \subset \mathbb{R}^{2}$, и путь $\varphi$ не имеет трансверсальных самопересечений, то образ отображения $\varphi^{\prime}$ является планарным подграфом $G_{\varphi}^{\prime} \subset G^{\prime}$ (мы приводим построение естественного вложения $G_{\varphi}^{\prime} \rightarrow \mathbb{R}^{2}$ в 2 см. Определение производной утолщения). Заменим граф $G^{\prime}$ на образ $G_{\varphi}^{\prime}$, и отображение $\varphi^{\prime}$ - на ограничение $\varphi^{\prime}: I \rightarrow G_{\varphi}^{\prime}$. Определим $k$-ю производную $\varphi^{(k)}$ индуктивно. Для цикла $\varphi$ определение производной $\varphi^{\prime}$ аналогично, и это будет снова некоторый цикл в графе на плоскости (который может выродиться в точку). Приведем пример, который будет использоваться в дальнейшем: $\varphi^{\prime}=\varphi$ для стандартной $d$-намотки $\varphi: S^{1} \rightarrow$ $S^{1}$ с $d \neq 0$. Ясно, что $\varphi^{\prime}$ - вложение для любого эйлерового пути или цикла $\varphi$. Таким образом, Пример 1.2действительно частный случай следующей теоремы. Теорема 1.3. I) [6] Пусть $\varphi: I \rightarrow \mathbb{R}^{2}$ - кусочно линейное отображение, являющееся симплицильным для некоторой триангуляции отрезка I с $k$ вершинами. Отображение $\varphi$ аппроксимируется вложениями, если и только если для каждого $i=0, \ldots, k$ его $i$-я производная $\varphi^{(i)}$ не содержит трансверсальных самопересечений. S) Пусть $\varphi: S^{1} \rightarrow \mathbb{R}^{2}$ - кусочно линейное отображение, являющееся симплициальным для некоторой триангуляции окружности $S^{1}$ с $k$ вершинами. Отображение $\varphi$ аппроксимируется вложениями, если и только если для каждого $i=0, \ldots, k$ его $i$-я производная $\varphi^{(i)}$ не содержит трансверсальных самопересечений, и при этом не является стандартной намоткой степени $d \notin\{-1,0,1\}$. Мы доказываем обе теоремы 1.3 I и 1.3, $\mathrm{S}$ в 92 Наше доказательство результата 1.3.I является более простым, чем приведенное в 6 . В $\$ 3$ мы применяем Теорему 1.3 для получения следующего критерия. Следствие 1.4. Кусочно линейное отображение $\varphi: I \rightarrow \mathbb{R}^{2}$ аппроксимируется вложениями, если и только если выполнено любое из следующих эквивалентных условий: D) (свойство взрезанного произведения) Существует отображение $$ \{(x, y) \in I \times I: x \neq y\} \rightarrow S^{1} $$ такое что его ограничение на множество $\{(x, y) \in I \times I: \varphi x \neq \varphi y\}$ гомотопно отображению, заданному формулой $\tilde{\varphi}(x, y)=\frac{\varphi x-\varphi y}{\|\varphi x-\varphi y\|} ;$ V) преплтствие ван Кампена (определенное в \$3) $v(\varphi)=0$. Хотя Критерий 1.4 V и труднее сформулировать, но его легче применять, чем 1.3] и 1.4.D. В Следствии 1.4 отрезок $I$ нельзя заменить на окружность $S^{1}$ : стандартная 3 -намотка является контрпримером [8]. Препятствия, подобные $1.4 \mathrm{D}$ и $1.4 \mathrm{~V}$, существуют и в близкой теории аппроксимируемости сингулярными зацеплениями (то есть, отображениями с непересекающимися образами связных компонент), но критерии, аналогичные [1.4 I и 1.4.DV для них не верны (Пример 3.3 ниже). Гипотеза. Кусочно-линейный путь $\varphi: I \rightarrow \mathbb{R}^{2}$ аппроксимируется вложениями, если и только если для любой пары дуг $I_{1}, I_{2} \subset I$, такой что $I_{1} \cap I_{2}=\emptyset$, пара ограничений $\varphi: I_{1} \rightarrow \mathbb{R}^{2}$ и $\varphi: I_{2} \rightarrow \mathbb{R}^{2}$ аппроксимируется сингулярными зацеплениями. Интересно обобщить критерии 1.3 и 1.4 на кусочно линейные отображения $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$, где $K-$ произвольный граф (см. частный случай в [14). Гипотеза. Пусть $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$ - кусочно линейное отображение, являющееся симплициальным относительно некоторой триангуляции графа $K$ с $k$ вершинами. Тогда отображение $\varphi$ аппроксимируется вложениями, если и только если $v(\varphi)=0$ и для каждого $i=0, \ldots, k$ его $i$-я производная $\varphi^{(i)}$ (определенная в (2) не содержит стандартных намоток степени $d \notin\{-1,0,1\}$. Если данная гипотеза верна, то кусочно-линейное отображение $\varphi: K \rightarrow \mathbb{R}^{2}$ дерева $K$ аппроксимируется вложениями, если и только если $v(\varphi)=0$ [2, Problem 4.5]. Завершим $\S$ несколькими замечаниями по поводу истории возникновения понятия аппроксимируемости вложениями. Дадим определение разложения 1-мерного компакта в обратный предел и покажем, как понятие аппроксимируемости вложениями появляется при исследовании планарности этого компакта. (Мы не будем использовать это определение в нашей статье.) В качестве примера построим 2-адический соленоид Ван Данцига. Возьмем полноторие $T_{1} \subset \mathbb{R}^{3}$. Пусть $T_{2} \subset T_{1}$ - полноторие, обходящее дважды вдоль оси полнотория $T_{1}$. Аналогично, возьмем полноторие $T_{3} \subset T_{2}$, обходящее дважды вдоль оси полнотория $T_{2}$. Продолжая далее подобным образом, мы получаем бесконечную последовательность полноторий $T_{1} \supset T_{2} \supset T_{3} \supset \ldots$ Пересечение всех полноторий $T_{i}$ является 1 -мерным компактом и называется 2-адическим соленоидом Ван Данцига. Обратным пределом бесконечной последовательности графов и симплициальных отображений между ними $K_{1} \stackrel{\varphi_{1}}{\longleftarrow} K_{2} \stackrel{\varphi_{2}}{\longleftarrow} K_{3} \stackrel{\varphi_{3}}{\longleftarrow} \ldots$ мы называем компакт $$ C=\left\{\left(x_{1}, x_{2}, \ldots\right) \in l_{2}: x_{i} \in K_{i} \text { и } \varphi_{i} x_{i+1}=x_{i}\right\} $$ Можно видеть из нашего построения, что для соленоида Ван Данцига все $K_{i} \cong S^{1}$ и все $\varphi_{i}$ суть 2 -намотки. Можно доказать, что любой 1 -мерный компакт может быть представлен в виде обратного предела. Такое представление показывает, что любой 1 -мерный компакт может быть вложен в $\mathbb{R}^{3}$. Оно также предоставляет простое достаточное условие планарности данного компакта: для каждого положительного целого числа $i$ должно существовать вложение $f_{i}: K_{i} \rightarrow \mathbb{R}^{2}$, такое что отображение $f_{i} \circ \varphi_{i}$ аппроксимируется вложениями и $f_{i+1}$ является $2^{-i}$-близким к $f_{i} \circ \varphi_{i}$. ## 2. ДОКАЗАТЕЛЬСТВО КРИТЕРИЯ АППРОКСИМИРУЕМОСТИ ВЛОЖЕНИЯМИ Теорема 1.3 следует из Примера 1.1 и Лемм 2.1, 2.2 (для $K \cong I, S^{1}$ ) и 2.3, которые интересны и сами по себе. Лемма 2.1. (для $K \cong I$ [6]) Предположим, что симплициальное отображение $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$ графа $K \cong S^{1}$ или $K \cong I$ не имеет трансверсальных самопересечений. Тогда если $\varphi^{\prime}$ аппроксимируется вложениями, то и $\varphi$ аппроксимируется вложениями. Лемма 2.2. А) 6] Если симплициальное отображение $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$ аппроксимируется вложениями, то и отображение $\varphi^{\prime}$ аппроксимируется вложениями. V) Если симплициальное отображение $\varphi: K \rightarrow G \subset \mathbb{R}^{2}$ аппроксимируется $\bmod 2$-вложениями, то отображение $\varphi^{\prime}$ аппроксимируется $\bmod 2$-вложениями. Здесь $\bmod 2$-вложение - это отображение общего положения $f: K \rightarrow \mathbb{R}^{2}$, такое что для каждой пары $a, b$ непересекающихся ребер графа $K$ множество $f a \cap f b$ состоит из четного числа точек. Определение производной для симплициального отображения произвольного графа $K$ приводится ниже. Лемма 2.3. Пусть $\varphi: S^{1} \rightarrow G$ - кусочно линейное отображение, которое является симплициальным для некоторой триангуляции окружности $S^{1}$ c $k$ вершинами. Тогда либо область определения отображения $\varphi^{(k)}$ пуста, либо $\varphi^{(k)}$ является стандартной намоткой степени $d \neq 0$. Это число $d$ можно рассматривать как обобщение степени для любого симплициального отображения $S^{1} \rightarrow$ $G$. Таким образом, интересно получить решение следующей задачи (оно может также сделать применение Критериев 1.3 более удобным): найти простой алгоритм для вычисления степени намотки $\varphi^{(\infty)}$ для данного кусочно линейного отображения $\varphi: S^{1} \rightarrow G$. Далее мы используем следующее обобщение определения производной (для пути), данного в $\$ 1$ Определение 2.4 (Производная симплициального отображения). [6] (см. иллюстрацию [1, а также часть иллюстрации 4 ниже) Пусть дано симплициальное отображение $\varphi: K \rightarrow G$. Сначала построим граф $K_{\varphi}^{\prime}$, который будет областью определения производной $\varphi^{\prime}$. Под $\varphi$-компонентой графа $K$ мы подразумеваем любую связную компоненту $\alpha$ множества $\varphi^{-1} a$, отображаемую на $a$, для некоторого ребра $a \subset G$. Множество вершин графа $K_{\varphi}^{\prime}$ находится в 1-1 соответствии с множеством всех $\varphi$-компонент. Для $\varphi$-компоненты $\alpha \subset K$ обозначим через $\alpha^{\prime} \in K_{\varphi}^{\prime}$ соответствующую вершину. Две вершины $\alpha^{\prime}$ и $\beta^{\prime}$ соединены ребром в графе $K_{\varphi}^{\prime}$, если и только если $\alpha \cap \beta \neq \emptyset$. Производная $\varphi^{\prime}: K_{\varphi}^{\prime} \rightarrow G^{\prime}$ - это симплициальное отображение, определенное на вершинах графа $K_{\varphi}^{\prime}$ формулой $\varphi^{\prime} \alpha^{\prime}=(\varphi \alpha)^{\prime}$. В дальнейшем заменим $\varphi^{\prime}$ на сюръективное ограничение $\varphi^{\prime}: K_{\varphi}^{\prime} \rightarrow \varphi^{\prime} K_{\varphi}^{\prime}$. (В оригинальном определении статьи [6] граф $G^{\prime}$ обозначается как $D(G)$, производная $\varphi^{\prime}$ как $d[\varphi]$, и граф $K_{\varphi}^{\prime}$ как $D(\varphi, K)$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-137.jpg?height=523&width=648&top_left_y=167&top_left_x=704) Рис. 2. Производная утолщения графа Доказательство Леммы 2.3. Будем говорить, что симплициальное отображение $\varphi: K \rightarrow G$ является сильно невырожденным, если для каждого ребра $a \subset K$ образ $\varphi a$ является ребром $G$ (а не вершиной) и для каждой пары $a, b \subset K$ смежных ребер мы имеем $\varphi a \neq \varphi b$. Обозначим через $|K|$ число вершин в графе $K$. Ясно, что если $K \cong S^{1}$, то $\left|K_{\varphi}^{\prime}\right| \leq|K|$, причем $\left|K_{\varphi}^{\prime}\right|=|K|$, только если $\varphi$ является сильно невырожденным. Поэтому лемму достаточно доказать только в последнем случае (потому что случаи, когда $K_{\varphi}^{\prime} \cong I$ или граф $K_{\varphi}^{\prime}$ является точкой, тривиальны). В случае сильно невырожденного отображения лемма очевидна, но мы приводим доказательство для полноты. Докажем, что если сильно невырожденное сюръективное симплициальное отображение $\varphi: K \rightarrow G$ графа $K \cong S^{1}$ не является стандартной намоткой степени, отличной от нуля, то $\left|G^{\prime}\right|>|G|$. Заметим, что для сильно невырожденного отображения $\varphi: S^{1} \rightarrow G$ граф $G$ не содержит висящих вершин. Если степень каждой вершины графа $G$ равна двум, то $\varphi$ является сильно невырожденным симплициальным отображением $S^{1} \rightarrow S^{1}$, следовательно, $\varphi$ является стандартной намоткой, вопреки нашему предположению. Значит, граф $G$ содержит вершину степени по крайней мере 3 . Тогда, по доказанному выше, число ребер графа $G$ больше числа вершин, следовательно, $\left|G^{\prime}\right|>|G|$. Поскольку для симплициального на отображения $\varphi: K \rightarrow G$ мы имеем $1 \leq|G| \leq|K|$, то $|G|,\left|G^{\prime}\right|, \ldots,\left|G^{(k)}\right| \leq k$ (напомним, что по определению отображение $\varphi^{\prime}$ сюръективно). Таким образом, есть только две возможности: любой одна (а значит, и $k$-я тоже) из производных $\varphi, \ldots, \varphi^{(k)}-$ стандартная намотка ненулевой степени, либо $\left|G^{(k)}\right|=0$, то есть область определения отображения $\varphi^{(k)}$ пуста. Теперь приведем обещанное в $\S 1$ построение вложения $G_{\varphi}^{\prime} \rightarrow \mathbb{R}^{2}$. Нам будет удобнее рассматривать утолщения графов, а не вложения графов в плоскость. В этом смысле обещанное построение эквивалентно построению производной утолщения (см. определение утолщения $N^{\prime}$ ниже). Далее мы предполагаем, что фиксировано утолщение $N$ графа $G$ на плоскости (то есть, регулярная окрестность графа $G \subset \mathbb{R}^{2}$ ). Мы также предполагаем, что также фиксировано разложение на ручки (обозначаемое через $S$ ) $$ N=\bigcup_{x \in \text { множество вершин графа } G} N_{x} \cup \bigcup_{a \in \text { множество ребер графа } G} N_{(a)} $$ соответствующее графу $G$, где $N_{x}$ обозначают 2-мерные диски, а $N_{(a)}$ - присоединенные к ним ленточки. Обозначим через $N_{a}$ ограничение $N_{x} \cup N_{(a)} \cup N_{y}$ утолщения $N$ на ребро $a=x y$. Фактически, мы не используем планарность $N$ в последующих рассуждениях. Можно считать, что утолщение $N$ является всего лишь ориентируемым (ориентируемость необходима для утверждения Примера 1.1). Дадим определение производной $N^{\prime}$ утолщения $N$. Это утолщение $N^{\prime}$ зависит от симплициального отображения $\varphi: K \rightarrow G \subset N$ и определено корректно, только если $\varphi$ не содержит трансверсальных самопересечений. Кроме того, в случае произвольного графа $K$ мы должны также предположить, что не существует пары дуг $i, j \subset K$ (не обязательно непересекающихся!), таких что пересечение $\varphi i \cap \varphi j$ трансверсально. Определение 2.5 (Производная утолщения графа). (см. иллюстрацию 2) Пусть $\varphi: K \rightarrow G \subset N$ - симплициальное отображение, такое что для любой пары дуг $i, j \subset K$ пересечение $\varphi i \cap \varphi j$ (возможно пустое) не трансверсально. Возьмем по диску $N_{a^{\prime}}^{\prime}$ для каждой вершины $a^{\prime} \in G^{\prime}$ и по ленточке $N_{\left(a^{\prime} b^{\prime}\right)}^{\prime}$ для каждого ребра $a^{\prime} b^{\prime} \subset G^{\prime}$. Тогда $N^{\prime}$ вместе с его разложением ручки $S^{\prime}$ определяется формулой $N^{\prime}=\bigcup N_{a^{\prime}}^{\prime} \cup \bigcup N_{\left(a^{\prime} b^{\prime}\right)}^{\prime}$. Здесь мы полагаем $N_{\left(a^{\prime} b^{\prime}\right)^{\prime}}^{\prime}=N_{(a)}$ для каждого ребра $a \subset G$. Для каждой пары $a, b \subset G$ смежных ребер, для которых $\left(\varphi^{\prime}\right)^{-1}\left(a^{\prime} b^{\prime}\right) \neq \emptyset$, мы соединяем два диска $N_{a^{\prime}}^{\prime}$ и $N_{b^{\prime}}^{\prime}$ узкой ленточкой $N_{\left(a^{\prime} b^{\prime}\right)}^{\prime}$ в $N_{a \cap b}$. Поскольку пересечение дуг $a \cup b$ и $c \cup d$ не трансверсально ни для какой пары смежных ребер $c, d \subset K$, то мы можем выбрать ленточки $N_{\left(a^{\prime} b^{\prime}\right)}^{\prime}$ так, чтобы они не пересекались для различных ребер $a^{\prime} b^{\prime}$. Это определение можно рассматривать как построение вложения $N^{\prime} \rightarrow N$, а также вложения $G_{\varphi}^{\prime} \rightarrow \mathbb{R}^{2}$. Заметим, что разбиение на ручки $S^{\prime}$ и топологический тип утолщения $N^{\prime}$ не зависят от выбора ленточек $N_{\left(a^{\prime} b^{\prime}\right)}^{\prime}$ ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-138.jpg?height=268&width=651&top_left_y=363&top_left_x=266) a ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-138.jpg?height=459&width=871&top_left_y=176&top_left_x=932) b Рис. 3. Перестройки вырожденных отображений в нашем определении. Альтернативное определение производной $D(N)$ утолщения $N$ из статьи [6] не зависит также от выбора отображения $\varphi$. Утолщение $N^{\prime}$ в нашей статье является подутолщением утолщения $D(N)$ (определение которого приводится в статье [6]), соответствующим подграфу $G_{\varphi}^{\prime} \subset G^{\prime}$. Ясно, что для исследования аппроксимируемости вложениями симплициальных отображений $K \rightarrow G \subset \mathbb{R}^{2}$ достаточно рассмотривать только приближения $f: K \rightarrow N$. Теперь мы собираемся свести задачу аппроксимируемости вложениями данного отображения к задаче существования вложения, близкого к нему в некотором смысле ( $S$-близкого). Определение 2.6 ( $S$-аппроксимация). 6] Отображение $f: K \rightarrow N$ называется $S$-аппроксимацией отображения $\varphi$, или, отображение $f S$-близко к $\varphi$, если выполняются следующие условия: (i) $f x \subset N_{\varphi x}$ для каждой вершины или ребра $x$ графа $K$; (ii) $x \cap f^{-1} N_{(\varphi x)}$ связно для каждого ребра $x$ графа $K$ с невырожденным образом $\varphi x$. Согласно Утверждению 2.9 статьи [6], отображение $\varphi: K \rightarrow G$ аппроксимируется вложениями, если и только если существует вложение $f: K \rightarrow N, S$-близкое к $\varphi$. Кусочно линейное отображение $\varphi: K \rightarrow N$ называется вырожденным, если $\varphi с$ является точкой для некоторого ребра $c \subset K$. Докажем следующее несложное Утверждение о стягивании ребра 2.7 которое в некотором смысле позволяет считать, что в Леммах 2.1 и 2.2 отображение $\varphi$ является невырожденным. Утверждение 2.7 (О стягивании ребра). Пусть $\varphi: K \rightarrow G$ - симплициальное отображение, такое что $\varphi с$ лвляется точкой для некоторого ребра $c \subset K$. Пусть $K / с$ - граф, полученный из графа $K$ стягиванием ребра $c$, и пуств $\varphi / c: K / c \rightarrow G-$ соответствующее отображение. Тогда D) $K_{\varphi / c}^{\prime}=K_{\varphi}^{\prime}, G_{\varphi}^{\prime}=G_{\varphi / c}^{\prime} u(\varphi / c)^{\prime}=\varphi^{\prime}$. A) для $K \cong S^{1}$ или $K \cong I$ отображение $\varphi /$ с аппроксимируется вложениями, если и только если $\varphi$ аппроксимируется вложениями. К) для произвольного графа $К$, если $\varphi$ аппроксимируется вложениями, то $\varphi /$ с аппроксимируется вложениями. V) Если $\varphi$ аппроксимируется $\bmod 2$-вложениями, то $\varphi /$ с аппроксимируется $\bmod 2$-вложениями. Доказательство Утверждения 2.7. D) очевидно. А) Докажем прямую импликацию. Пусть $f: K / c \rightarrow N$ - вложение, $S$-близкое к $\varphi / c$. Пусть $a \subset K-$ ребро, смежное с $c$ (если $c$ - связная компонента графа $K$, то требуемое утверждение очевидно). Добавим новую вершину к ребру $a$ графа $K / c$ (иллюстрация 3 , а). Так как $K \cong S^{1}$ или $K \cong I$, то полученный граф изоморфен $K$ и вложение $f: K \rightarrow N$ - искомое. Обратная импликация - частный случай утверждения $\mathrm{K}$ ). K) Пусть $f: K \rightarrow N$ - вложение, $S$-близкое к $\varphi$. Сделаем перестройку, показанную на иллюстрации $3 \mathrm{~b}$. Получим вложение $\bar{f}: K / c \rightarrow N, S$-близкое к $\varphi / c$. V) Пусть $f-\bmod 2$-вложение, $S$-близкое к $\varphi$. Сделаем перестройку, показанную на иллюстрации $3 \mathrm{~b}$. Получим $S$-близкое к $\varphi / c$ отображение $\bar{f}: K / c \rightarrow N$. Достаточно доказать, что $|\bar{f} a \cap \bar{f} b|=0(\bmod 2)$ для каждой пары непересекающихся ребер $a, b \subset(K / c)$. Действительно, $a$ и $b$ являются ребрами также и в графе $K$, причем по крайней мере одно из них не смежно с $c$ (потому что $a$ и $b$ являются непересекающимися в $K / c$ ). Если ни $a$, ни $b$ не смежно с $c$, то $|\bar{f} a \cap \bar{f} b|=|f a \cap f b|=0(\bmod 2)$. Если, например, $b \subset K$ смежно с $c$ и $a$ не смежно с $c$, то $|\bar{f} a \cap \bar{f} b|=|f a \cap f b|+|f a \cap f c|=0(\bmod 2)$, что доказывает утверждение. Вырожденные отображения появляются в нашем доказательстве Лемм 2.1 и 2.2, даже если исходное отображение $\varphi: K \rightarrow G$ является невырожденным. Мы собираемся построить граф $\bar{K}_{\varphi}^{\prime}$ и пару (вырожденных) симплициальных отображений $G \stackrel{\bar{\varphi}}{\stackrel{K}{\varphi}} \bar{\varphi}^{\bar{\varphi}^{\prime}} G^{\prime}$, которые могут быть получены из отображений $\varphi$ и $\varphi^{\prime}$, ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-139.jpg?height=694&width=1037&top_left_y=178&top_left_x=515) Рис. 4. Полупроизводные симплициального отображения соответственно, операцией из Утверждения о стягивании ребра 2.7 (при некоторых дополнительных предположениях относительно $\varphi$, детали представлены ниже). Вместе с построением вложения $N^{\prime} \rightarrow N$ (см. определение утолщения $N^{\prime}$ выше), это немедленно влечет утверждение Леммы 2.1 (см. иллюстрации , 4, 5, 6). Определение 2.8 (Полупроизводная симплициального отображения). (см. иллюстрацию (4) Предположим, что отображение $\varphi$ является невырожденным, и $K$ не имеет вершин степени 0 . Возьмем несвязное объединение всех $\varphi$-компонент графа $K$ (см. Определение производной $\varphi^{\prime}$ выше). Соединим ребром любые две вершины, принадлежащие различным $\varphi$-компонентам и отвечающие одной и той же вершине графа $K$. Обозначим полученную полупроизводную графа $K$ через $\bar{K}_{\varphi}^{\prime}$. Таким образом, $\varphi$-компонента $\alpha \subset K$ является также подграфом графа $\bar{K}_{\varphi}^{\prime}$, обозначаемым через $\bar{\alpha}^{\prime}$. В дальнейшем мы отождествляем точки графов $\alpha$ и $\bar{\alpha}^{\prime}$. Определим симплициальные отображения $\bar{\varphi}$ и $\bar{\varphi}^{\prime}$ ( полупроизводные отображения $\varphi$ ) как очевидные проекции $\bar{K}_{\varphi}^{\prime} \rightarrow G$ и $\bar{K}_{\varphi}^{\prime} \rightarrow G^{\prime}$, соответственно, заданные на вершинах формулами $\bar{\varphi} x=\varphi x$ и $\bar{\varphi}^{\prime} x=(\varphi \alpha)^{\prime}$, где вершина $x \in \bar{K}_{\varphi}^{\prime}$ принадлежит $\varphi$-компоненте $\bar{\alpha}^{\prime}$. Доказательство Леммы 2.1. Согласно Утверждению о стягивании ребра2.7.D, отображение $\varphi$ может считать невырожденным. Мы также можем считать, что граф $K$ не имеет вершин степени 0 . Легко видеть, что $\varphi$ и $\varphi^{\prime}$ могут быть получены из $\bar{\varphi}$ и некоторого сужения $\bar{\varphi}^{\prime}$, соответственно, операцией из Утверждения о стягивании ребра 2.7 Если любые две $\varphi$-компоненты имеют не более одной общей точки, то $\varphi^{\prime}$ может быть получен таким образом непосредственно из $\bar{\varphi}$. Но для $K \cong S^{1}$ последнее условие выполнено всегда, кроме случая, когда граф $K$ состоит из ровно двух $\varphi$-компонент. Очевидно, отображение $\varphi$ аппроксимируется вложениями в указанном случае. Таким образом, достаточно доказать следующее утверждение: $\left.{ }^{*}\right)$ если $\bar{\varphi}^{\prime}$ аппроксимируется вложениями, то $\bar{\varphi}$ аппроксимируется вложениями. Докажем утверждение (*) для произвольного графа $K$. Если $\bar{\varphi}^{\prime}$ аппроксимируется вложениями, то найдется вложение $\bar{K}_{\varphi}^{\prime} \rightarrow N^{\prime}, S^{\prime}$-близкое к $\bar{\varphi}^{\prime}$. Определим вложение $f: \bar{K}_{\varphi}^{\prime} \rightarrow N$ как композицию этого вложения и вложения $N^{\prime} \rightarrow N$, построенного в определении утолщения $N^{\prime}$ (см. иллюстрацию [5, где это построение применяется к отображению $\varphi$ с иллюстрации (4). Ясно, что существует новое разложение на ручки $N=\bigcup \bar{N}_{a} \cup$ $\bigcup \bar{N}_{(a b)}$ утолщения $N$, обозначаемое $\bar{S}$, такое что $f$ будет $\bar{S}$-аппроксимацией отображения $\bar{\varphi}$ (см. иллюстрацию б, сравни с [6], Утверждение 4.9) Тогда $f: \bar{K}_{\varphi}^{\prime} \rightarrow \bar{N}$ (где $\bar{N}$ обозначает утолщение $N$ с новым разложением ручки $\bar{S}$ ) - вложение, $\bar{S}$-близкое к отображению $\bar{\varphi}$. Лемма доказана. Та же самая идея используется в доказательстве Лемм 2.2 A,V. Рассматривается отображение $f: \bar{K}_{\varphi}^{\prime} \rightarrow N$ общего положения, $S$-близкое к $\bar{\varphi}$ и строится полупроизводная $\bar{f}^{\prime}: \bar{K}_{\varphi}^{\prime} \rightarrow N^{\prime}, S$-близкая к $\bar{\varphi}^{\prime}$ (см. иллюстрацию耳). Потом проверяется, что если $f$ - вложение, то $\bar{f}^{\prime}$ - также вложение (см. иллюстрацию (8). Определение 2.9 (Полупроизводная $S$-аппроксимации). (см. иллюстрацию 7 г где приведенное ниже построение построение применяется к отображению $\varphi$, изображенному на иллюстрации (4) Пусть $K$ - граф без вершин степени 0 . Пусть $\varphi: K \rightarrow G \subset N$ - невырожденное симплициальное отображение без трансверсальных самопересечений. Пусть $f: K \rightarrow N-S$-аппроксимация отображения $\varphi$. Тогда полупроизводная отображения $f$ есть $S^{\prime}$-аппроксимация $\bar{f}^{\prime}: \bar{K}_{\varphi}^{\prime} \rightarrow N^{\prime}$ отображения $\varphi^{\prime}$, и строится следующим образом. Для каждого ребра $a \subset G$ выберем гомеоморфизм $h_{a}: N_{a} \rightarrow N_{\prime}^{\prime}$ таким образом, что для каждого любого ребра $b$, смежного с $a$, мы имеем $h_{a}\left(N_{a} \cap N_{(b)}\right) \subset N_{\left(a^{\prime} b^{\prime}\right)}^{\prime}$. Определим $\bar{f}^{\prime}$ на каждой $\varphi$-компоненте $\bar{\alpha}^{\prime} \subset \bar{K}_{\varphi}^{\prime}$ формулой $\left.\bar{f}^{\prime}\right|_{\bar{\alpha}^{\prime}}=\left.h_{\varphi \alpha} f\right|_{\alpha}$ Теперь ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-140.jpg?height=488&width=628&top_left_y=184&top_left_x=720) Рис. 5. Построение $S$-аппроксимации ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-140.jpg?height=308&width=757&top_left_y=800&top_left_x=655) Рис. 6. Построение разбиения на ручки ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-140.jpg?height=489&width=780&top_left_y=1246&top_left_x=638) Рис. 7. Полупроизводная $S$-аппроксимации определим $\bar{f}^{\prime}$ на каждом ребре $x y \subset \bar{K}_{\varphi}^{\prime}$, соединяющем две различные $\varphi$-компоненты $\bar{X}^{\prime}$ и $\bar{Y}^{\prime}$. Возьмем ребро $a \subset \bar{X}^{\prime}$, содержащее вершину $x$. Отождествим $\bar{X}^{\prime}$ с $X$ (см. Определение полупроизводной симплициального отображения $\bar{\varphi}^{\prime}$ ). Тогда $a$ будет отождествлено с некоторым ребром графа $K$, а $x$ - с некоторой вершиной графа $K$. Обозначим через $\bar{x}$ дугу $a \cap f^{-1} N_{\varphi x}$. Определим дугу $\bar{y}$ аналогично. Разрежем ребро $x y$ в три отрезка $x x_{1}$, $x_{1} y_{1}$ и $y_{1} y$. Пусть $\bar{f}^{\prime}$ гомеоморфно отображает отрезок $x x_{1}$ на $h_{\varphi X} f \bar{y}$, отрезок $y_{1} y-$ на $h_{\varphi} f \bar{x}$, а отрезок $x_{1} y_{1}$ - на прямолинейный отрезок в диске $N_{(\varphi X \varphi Y)}^{\prime}$, соединяющий точки $\bar{f}^{\prime} x_{1}$ и $\bar{f}^{\prime} y_{1}$. Таким образом, отображение $\bar{f}^{\prime}: \bar{K}_{\varphi}^{\prime} \rightarrow N^{\prime}$ построено. Заметим, что если $f$ - вложение, то есть более простое альтернативное построение отображения $\bar{f}^{\prime}$, в некотором смысле обратное к построению из доказательства Леммы 2.1. Но это альтернативное построение неприменимо к доказательству Леммы 2.2 V , поэтому мы не пользуемся им в данной статье. Мы собираемся доказать Лемму 2.2 $\mathrm{A}, \mathrm{V}$ только в случае, когда производная $N^{\prime}$ определена корректно, то есть $K$ не содержит пар дуг $i, j$, для которых пересечение $\varphi i \cap \varphi j$ трансверсально. Этого достаточно для доказательства Теоремы [1.3 В общем случае доказательство аналогично, но необходимо всюду вместо $N^{\prime}$ пользоваться производной $D(N)$, определенной в статье [6]. Доказательство Леммы 2.2. $A$. Согласно Утверждению 2.7] Кожно считать, что $\varphi$ невырождено. Возьмем некоторое вложение $f: K \rightarrow N, S$-близкое к $\varphi$. Тогда достаточно показать, что отображение $\bar{f}^{\prime}$ (см. Определение полупроизводной $S$-аппроксимации $\bar{f}^{\prime}$ ) является вложением. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-141.jpg?height=589&width=531&top_left_y=111&top_left_x=771) Рис. 8. Подсчет количества точек пересечения Рассмотрим пару различных ребер $x y, z t$ графа $K_{\varphi}^{\prime}$. Обозначим множество $\bar{f}^{\prime}(x y) \cap \bar{f}^{\prime}(z t)$ через $i$. Достаточно показать, что $i=\bar{f}^{\prime}(x y \cap z t)$. Обозначим $a^{\prime}=\bar{\varphi}^{\prime} x, b^{\prime}=\bar{\varphi}^{\prime} y, c^{\prime}=\bar{\varphi}^{\prime} z$ и $d^{\prime}=\bar{\varphi}^{\prime} t$. Без ограничения общности возможны следующие 3 случая. 1) $a^{\prime}, b^{\prime}, c^{\prime}$ и $d^{\prime}$ попарно различны. Так как $\bar{f}^{\prime}$ является $S^{\prime}$-аппроксимацией,то $\bar{f}^{\prime} x y \subset N_{a^{\prime} b^{\prime}}^{\prime}$ и $\bar{f}^{\prime} z t \subset N_{c^{\prime} d^{\prime}}^{\prime}$, следовательно $i=\emptyset$. 2) ( $a^{\prime}=c^{\prime}$ и $b^{\prime} \neq d^{\prime}$ ) или $\left(a^{\prime}=b^{\prime}=c^{\prime}=d^{\prime}\right)$. Тогда $i \subset N_{a^{\prime}}^{\prime}$, следовательно, $i=h_{a}(f \bar{x} \cap f \bar{z})$ (см. определение $h_{a}$ и $\bar{x}$ в Определении полупроизводной $S$-аппроксимации $\bar{f}^{\prime}$, дуга $\bar{z}$ определяется аналогично $\bar{x}$ ). Если $y \neq t$, то $\bar{x}$ и $\bar{z}$ не пересекаются, так что $f \bar{x} \cap f \bar{z}=\emptyset$ и $i=\emptyset$. Если же $y=t$, то $i=h_{a}(f y)=\bar{f}^{\prime}(x y \cap z t)$. 3) $a^{\prime}=c^{\prime}, b^{\prime}=d^{\prime}$ и $a^{\prime} \neq b^{\prime}$. В этом случае как $x y$, так и $z t$, соединяют вершины из различных $\varphi$-компонент. Докажем, что $x y$ и $z t$ не пересекаются. Например, пусть $y=t$. Тогда все вершины $x, y, z$ и $t$ графа $\bar{K}_{\varphi}^{\prime}$ отвечают одной и той же вершине графа $K$. Обозначим ее через $w$. Обозначим через $X$ и $Z$ те $\varphi$-компоненты множества $\varphi^{-1} a=\varphi^{-1} c$, для которых $x \in \bar{X}^{\prime}$ и $z \in \bar{Z}^{\prime}$. Таким образом, у $\varphi$-компонент $X$ и $Z$ есть общая точка $w$, следовательно $X=Z$. Значит, $x, z \in \bar{X}^{\prime}=\bar{Z}^{\prime}$ отвечают одной и той же вершине $w$, следовательно, $x=z$. Мы получаем, что $y=t$ и $x=z$, тогда по построению графа $\bar{K}_{\varphi}^{\prime}$ мы получаем $x y=z t$, что противоречит выбору этих ребер. Значит, $x y$ и $z t$ не пересекаются. Покажем, что в случае $(3)|i|=0(\bmod 2)$. В дальнейшем будем опускать $\bar{f}^{\prime}$ в обозначениях всех образов при отображении $\bar{f}^{\prime}$. Заметим, что гомеоморфизм $h_{a} \circ h_{b}^{-1}$ отображает $y_{1} y$ и $t_{1} t$ на $\bar{x}$ и $\bar{z}$, соответственно (иллюстрация 8). Из этого следует, что $|i|=|I \cap J|$, где $I=\bar{x} \cup x y_{1}$ и $J=\bar{z} \cup z t_{1}$. Из этого также следует, что две пары точек $\partial I$ и $\partial J$ не зацеплены на окружности $\partial\left(h_{a} N_{a \cap b} \cup N_{\left(a^{\prime} b^{\prime}\right)}^{\prime}\right)$. Так как $I, J \subset h_{a} N_{a \cap b} \cup N_{\left(a^{\prime} b^{\prime}\right)}$, то $|i|=|I \cap J|=0$ $(\bmod 2)$. Таким образом, остается доказать, что $|I \cap J| \leq 1$, тогда $I \cap J=\emptyset$. Последнее утверждение следует из равенства $$ \bar{x} \cap \bar{z}=h_{a}(f \bar{x} \cap f \bar{z})=\emptyset \quad x x_{1} \cap z z_{1}=h_{a}(f \bar{y} \cap f \bar{t})=\emptyset \quad \text { и } \quad\left|x_{1} y_{1} \cap z_{1} t_{1}\right| \leq 1 $$ потому что $x_{1} y_{1}$ и $z_{1} t_{1}$ - прямолинейные отрезки в диске $N_{\left(a^{\prime} b^{\prime}\right)}$. Лемма доказана. Доказательство Леммы[2.2. V. Согласно Утверждению о стягивании ребра2.7.V нам достаточно доказать, что если $f: \bar{K}_{\varphi}^{\prime} \rightarrow N$ является $\bmod 2$-вложением, $S$-близким к $\varphi$, то его полупроизводная $\bar{f}^{\prime}$ также является $\bmod 2$-вложением. Возьмем пару непересекающихся ребер $x y$, zt графа $\bar{K}_{\varphi}^{\prime}$ и рассмотрим те же три случая, что и в доказательстве Леммы 2.2 А. Случай 1) тривиален. В случае 2) мы имеем $f(x y) \cap f(z t) \subset N_{a}$, следовательно, $|i|=\left|h_{a}(f \bar{x} \cap f \bar{z})\right|=$ $\left|h_{a}(f(x y) \cap f(z t))\right|=|f(x y) \cap f(z t)|=0(\bmod 2)$. В доказательстве Леммы 2.2 А мы уже показали, что в случае 3) выполнено равенство $|i|=0(\bmod 2)$. Таким образом, Лемма $2.2 \mathrm{~V}$ доказана. ## 3. ПРеПЯТСТВИе Ван КамПЕНа Препятствие Ван Кампена было придумано Ван Кампеном при исследовании вложимости полиэдров в $\mathbb{R}^{2 n}$ 2, 3, 4, 7, 8. Дадим определение препятствия ван Кампена к аппроксимируемости вложениями симплициальных путей. Наше построение более наглядно, чем построение препятствия Ван Кампена к вложимости. Пусть $\varphi$ : $I \rightarrow \mathbb{R}^{2}$ - симплициальный путь (на иллюстрации 9 приведенная ниже конструкция применяется к пути, показанному на иллюстрации (1). Обозначим через $x_{1}, \ldots, x_{k}$ вершины графа $I$ в порядке их расположения на дуге $I$, и обозначим ребро $x_{i} x_{i+1}$ через $i$. Пусть $I^{*}=\bigcup_{i Устойчивость пересечений путей на плоскости ## Lucky Hunting Stability of intersections of paths in the plane Два охотника... Two hunters... ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-146.jpg?height=1912&width=1262&top_left_y=359&top_left_x=306) Пути охотников Paths of hunters ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-147.jpg?height=200&width=958&top_left_y=420&top_left_x=417) Пути охотников Paths of hunters ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-147.jpg?height=174&width=955&top_left_y=1500&top_left_x=416) Пути собак Paths of dogs Производные путей Derivatives of paths $\cdot 4$ ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-148.jpg?height=130&width=468&top_left_y=1634&top_left_x=800) $\downarrow \varphi$ $\downarrow \varphi^{\prime}$ $\downarrow \varphi^{\prime \prime}$ ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-148.jpg?height=126&width=532&top_left_y=1926&top_left_x=463) $\Delta_{\mathrm{G}}$ ## Спасибо за внимание! ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-149.jpg?height=576&width=766&top_left_y=437&top_left_x=547) ## $C I S$-ГРАФЫ Напомним основные определения. Пусть дан граф $G$. Подграфом графа $G$ на вершинах $v_{1}, v_{2}, \ldots, v_{n}$ (где $v_{1}, v_{2}, \ldots, v_{n}$ - часть вершин графа $G$ ) называется граф, вершинами которого являются вершины $v_{1}, v_{2}, \ldots, v_{n}$, и между ними проведены все рёбра, которые были проведены в графе $G$. Дополнением графа $G$ называется граф, вершинами которого являются вершины графа $G$, но две вершины соединяются ребром в том и только в том случае, если они не были соединены ребром в графе $G$. Кликой в графе $G$ называется любое множество попарно соединённых вершин. Соответственно независимым множеством называется любое множество попарно не соединённых вершин. Клика (независимое множество) называется максимальной (максимальным), если она не содержится ни в какой большей клике (независимом множестве). Главным определением этой серии задач является определение $C I S$-графа. $C I S$-графом называется граф, в котором любая максимальная клика $C$ пересекается с любым максимальным независимым множеством $S$. Например, цикл длины четыре является $C I S$-графом, а П-граф, изображённый на рисунке 9 , не является $C I S$-графом. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-151.jpg?height=117&width=123&top_left_y=952&top_left_x=972) Рис. 1 ## А много ли $C I S$-графов? Задачи на определение. 1. Покажите, что несвязное объединение двух $C I S$-графов $G_{1}$ и $G_{2}$ является $C I S$-графом. Примечание. Чтобы получить несвязное объединение графов $G_{1}$ и $G_{2}$, нужно объединить их (считая все вершины различными!) и не проводить больше никаких ребер. 2. Найдите все $C I S$-графы, в которых нет трёх попарно соединённых вершин. Проверьте, что вершины каждого из них можно покрасить в два цвета так, чтобы любые две соседние вершины были разных цветов. 3. Пусть из $C I S$-графа удалили вершину, из которой выходило одно ребро. Покажите, что получился снова $C I S$-граф. 4. а) Назовём вершину хорошей, если любые два её соседа соединены (то есть она вместе с соседями образует клику). Докажите, что если в графе любая максимальная клика содержит хорошую вершину, то это $C I S$-граф. b) Покажите, что обратное неверно. 5. Рассмотрим граф, множество вершин которого является объединением пересекающихся клики и независимого множества. Покажите, что он является $C I S$-графом. Подстановкой графа $G_{1}$ в граф $G_{2}$ будем называть следующую операцию: некоторая фиксированная вершина $A$ графа $G_{2}$ заменяется на граф $G_{1}$, между вершинами графа $G_{1}$ и графа $G_{2}$ проводятся в точности те же ребра, которые были между $A$ и остальными вершинами $G_{2}$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-151.jpg?height=292&width=874&top_left_y=1872&top_left_x=594) Рис. 2 6. Пусть в результате подстановки графа $G_{1}$ в граф $G_{2}$ получился $C I S$-граф. Покажите, что и $G_{1}$, и $G_{2}$ были $C I S$-графами. 7. Докажите, что любой граф является подграфом некоторого $C I S$-графа. 8. Рассмотрим два множества графов $G_{1}, G_{2}, G_{3}, \ldots$ и $H_{1}, H_{2}, H_{3}, \ldots$, причём графов в множествах поровну, но их число может быть как конечное, так и бесконечное. Теперь рассмотрим множество всех графов, для которых выполнено условие: если среди подграфов данного графа есть граф $G_{i}$, то этот $G_{i}$ дополняется до подграфа $H_{i}$, то есть этот $G_{i}$ содержится в некотором подграфе $H_{i}$ данного графа. a) Подберите графы $G_{i}$ и $H_{i}$ так, чтобы получилось множество всех $C I S$-графов. b) Можно ли выбрать конечное число графов $G_{i}$ и $H_{i}$ так, чтобы получилось множество всех $C I S$-графов? 9*. Пусть в графе существует ровно одна пара из непересекающихся максимальной клики $C$ и максимального независимого множества $S$ a) Докажите, что в $G$, кроме вершин из $C \cup S$, не может быть ровно одной вершины. b) То же для двух вершин. c) Попробуйте доказать, что тогда множество вершин этого графа совпадает с $C \cup S$. ## Расчёски и гребешки. Класс графов называется замкнутым относительно подстановки, если из графов $G_{1}$ и $G_{2}$ этого класса с помощью подстановки получается снова граф этого класса. Класс графов называется точно замкнутым относительно подстановки, если дополнительно потребовать, чтобы графы этого класса могли получаться подстановкой только из графов этого класса. Класс графов называется замкнутым относительно дополнения, если дополнение графа из этого класса также принадлежит этому классу. Класс графов называется наследственным, если из того, что некоторый граф принадлежит этому классу, следует, что и любой его подграф принадлежит этому классу. Очевидно, что класс $C I S$-графов замкнут относительно дополнения. В задаче 6 мы также показали, что он замкнут относительно подстановки. Будем исследовать $C I S$-графы дальше. 10. а) Докажите, что если в графе существует пара непересекающихся максимальной клики и максимального независимого множества, то этот граф содержит в качестве подграфа П-граф. b) Покажите, что обратное неверно. Определение. $k$-расчёской $(k \geqslant 2$ ) называется граф на $2 k$ вершинах, в котором первые $k$ вершин попарно соединены, другие $k$ вершин попарно не соединены, и ещё проведено $k$ рёбер, соединяющие $i$-ю вершину первой группы из $k$ вершин с $i$-ой вершиной второй группы для каждого $i=1,2, \ldots, k$. Далее, $k$-гребешком называется граф, полученный прибавлением одной вершины к $k$-расчёске, причем новая вершина соединена только со всеми $k$ вершинами из первой группы (с вершинами, образующими клику). Соответственно, $k$-антирасческой и $k$ веником называются дополнения к $k$-расческе и $k$-гребешку. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-152.jpg?height=225&width=168&top_left_y=984&top_left_x=453) 3 -расческа ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-152.jpg?height=225&width=172&top_left_y=984&top_left_x=725) 3 -гребешок ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-152.jpg?height=229&width=260&top_left_y=982&top_left_x=1001) 3 -антирасческа ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-152.jpg?height=229&width=271&top_left_y=982&top_left_x=1349) 3 -веник 11. а)В $C I S$-графе любой П-подграф содержится в $A$-подграфе (см. рисунок 3 ). b)В $C I S$-графе для любой $k$-расчёски (являющейся подграфом) существует содержащий её $k$-гребешок (являющийся подграфом исходного графа). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-152.jpg?height=206&width=125&top_left_y=1416&top_left_x=974) Рис. 3 . $A$-граф 12. Приведите пример графа, в котором любая $k$-расчёска содержится в $k$-гребешке, любая $k$-антирасчёска содержится в $k$-венике (для всех $k$ ), а также в этом графе есть пара непересекающихся максимальной клики и независимого множества. Путеводная звезда. Попробуйте доказать следующую важную теорему. Теорема. Пусть среди подграфов данного графа нет 3 -расчёски и её дополнения, и все 2-расчёски дополнены до 2-гребешков. Тогда этот граф является $C I S$-графом. Определение. $d$-графом называется полный граф, каждое ребро которого покрашено в один из $d$ цветов. Обозначим за $E_{i}$ рёбра $i$-го цвета. Граф на вершинах $d$-графа с рёбрами $E_{i}$ будем называть $x$ роматической (или цветной) компонентой $i$-го цвета. Определение $C I S$ - $d$-графа. Пусть для каждого цвета выбрано максимальное множество вершин, среди которых любые две соединены ребром не этого цвета (максимальное означает, что при добавлении любой другой вершины появляется ребро данного цвета). Пусть для любой такой выборки все эти независимые множества имеют общую вершину. Тогда этот $d$-граф называется $C I S$ - $d$-графом. Рассмотрим игру на $d$ человек с полной информацией. Можно представлять ее себе как конечное дерево, вершинами которого являются позиции в игре. Есть корневая вершина (начало игры). Из неё выходят рёбра одного цвета (цвет игрока, который начинает игру). Тем самым игрок выбирает следующую позицию в игре. Из любой новой вершины выходят рёбра также одного цвета (но другого), причем для разных вершин этот цвет может быть разным. Тем самым первый игрок также выбирает, кто будет ходить следующим. Далее аналогично. Так как дерево конечное, то игра когда-нибудь закончится, то есть наступит один из исходов игры. Из вершины графа, соответствующей этой позиции, больше ребер не выходит. Можно построить $d$-граф на вершинах-исходах, покрасив ребро между двумя исходами в цвет исходящих ребер для вершины на пути, их соединяющем, ближайшей к корню. Оказывается, что получится $C I S$ - $d$-граф (можете на досуге обдумать). У $C I S$ - $d$-графа любая цветная компонента является $C I S$-графом (задача), но истинность обратного утверждения неизвестна. Если дополнительно предположить, что в $C I S$ - $d$-графе нет разноцветных треугольников, то это обратное утверждение оказывается верным (задача). Описание таких $C I S$ - $d$-графов сводится к описанию $C I S$-графов (приемлемого описания нет ни у тех, ни у других). Пока неизвестно, существует ли $C I S$ - $d$-граф, в котором есть разноцветный треугольник, но если вершин не больше 12 , то нет разноцветного треугольника (проверено на компьютере). ## Свойство графов Галлаи. Назовём $\Delta$-графом граф на трёх вершинах, в котором три ребра покрашены в разные цвета (разноцветный треугольник). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-153.jpg?height=174&width=215&top_left_y=433&top_left_x=909) Рис. 4 . $\Delta$-граф Определение. $d$-граф, среди подграфов которого нет $\Delta$-графа, называется $d$-графом Галлаи. Переопределим П-граф как $d$-граф на четырёх вершинах, в котором рёбра первого цвета образуют букву П, то есть хроматическая компонента первого цвета является П-графом; остальные рёбра покрашены во второй цвет, и рёбер остальных цветов нет. Будем его писать просто П-граф. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-153.jpg?height=123&width=137&top_left_y=864&top_left_x=959) Рис. 5 . Цветной П-граф Цель этой части - последняя задача, которая использует замечательное свойство $d$-графов Галлаи - декомпозицию Галлаи. Декомпозиция Галлаи описана в задаче 16. 13. Пусть дан $d$-граф $G$, обладающий следующим свойством: если в нём вычеркнуть рёбра любого выбранного цвета, то он останется связным. Пусть $G$ - это не П-граф и не $\Delta$-граф. Докажите, что из него можно выбросить одну вершину так, что он сохранит это свойство. 14. Пусть в $d$-графе $(d \geqslant 3)$ хроматическая компонента каждого цвета связна. Покажите, что $\Delta$-граф является подграфом этого графа. 15. Проверьте, является ли класс $d$-графов Галлаи замкнутым относительно подстановки, точно замкнутым относительно подстановки и является ли он наследственным. 16. Докажите, что $d$-граф Галлаи является результатом подстановки $n d$-графов в некоторый 2 -граф вместо $n$ его вершин. 17. Пусть $F$ - класс графов, точно замкнутый относительно подстановки и замкнутый относительно дополнения. Рассмотрим такой $d$-граф Галлаи, что все, кроме одной, его хроматические компоненты принадлежат классу $F$ (пусть кроме компоненты цвета $d$ ). Предположим еще, что есть хотя бы одно ребро цвета $d$. Тогда последняя хроматическая компонента также принадлежит классу $F$. ## После промежуточного финиша. ## $C I S$ - $d$-графы Галлаи. 18. Проверьте, является ли класс $C I S$ - $d$-графов замкнутым относительно подстановки, точно замкнутым относительно подстановки и является ли он наследственным. 19. Если среди подграфов данного $d$-графа нет $\Delta$-графов и П-графов, то он является $C I S$ - $d$-графом. 20. Покажите, что $d$-граф, сопоставленный игре с полной информацией (см. введение), не содержит $\Delta$-графа и П-графа, и что любой такой граф сопоставляется некоторой игре. 21. Приведите пример $d$-графа Галлаи, у которого есть рёбра хотя бы трёх цветов, но не являющегося $C I S$ - $d$ графом. 22. Приведите пример $C I S$ - $d$-графа, у которого есть рёбра хотя бы трёх цветов. 23. $d$-граф Галлаи является $C I S$ - $d$-графом, если и только если все его хроматические компоненты являются $C I S$-графами. 24. Если все, кроме (может быть) одной, хроматические компоненты $d$-графа Галлаи являются $C I S$-графами, то этот $d$-граф является $C I S$ - $d$-графом. 25. Предположим, что все $C I S$-3-графы являются 3 -графами Галлаи. Тогда все $C I S$ - $d$-графы являются $d$ графами Галлаи. 26. Гипотеза. Любой $C I S$ - $d$-граф является $d$-графом Галлаи. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-154.jpg?height=534&width=472&top_left_y=93&top_left_x=792) Обозначим через $2 G_{10}$ граф на двадцати вершинах $v_{0}, v_{1}, \ldots, v_{9}, v_{0}^{\prime}, v_{1}^{\prime}, \ldots, v_{9}^{\prime}$, составленный из двух копий $G_{10}$ (первый граф $G_{10}$ на вершинах $v_{0}, v_{1}, \ldots, v_{9}$, другой - на остальных), в котором для любых различных $i$ и $j$ вершины $v_{i}$ и $v_{j}^{\prime}$ соединены, если и только если вершины $v_{i}$ и $v_{j}$ не соединены. Любое ребро между вершинами $v_{i}$ и $v_{i}^{\prime}$ может быть как проведено, так и нет. Под $2 G_{10}$ будем подразумевать любой из этих 1024-ёх графов. Граф, изображённый на следующем рисунке, называется графом Петерсена. Обозначим его $P$. Возьмём две его копии с вершинами $u_{1}, u_{2}, \ldots, u_{10}$ и $v_{1}, v_{2}, \ldots, v_{10}$ соответственно. Соединим вершины $u_{i}$ и $v_{j}$ ребром, если вершины $u_{i}$ и $u_{j}$ не соединены ребром для $i \neq j$. При этом вершины $u_{i}$ и $v_{i}$ можем соединить ребром, а можемнет. В результате получим $2^{10}=1024$ графов. Запись "граф $2 P$ " подразумевает любой из них. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-154.jpg?height=479&width=485&top_left_y=1017&top_left_x=794) Теорема. Пусть среди подграфов данного графа нет 3-расчёски и её дополнения, и все 2-расчёски дополнены до 2-гребешков. Тогда этот граф является $C I S$-графом. В оставшихся задачах доказываем эту теорему. 27. В поисках $G_{10}$. Пусть в графе $G$ есть хотя бы одна пара непересекающихся максимальной клики $C$ и независимого множества $S$, среди подграфов данного графа нет 3 -расчёски и её дополнения, и все 2 -расчёски дополнены до 2-гребешков. a) Докажите, что есть вершины, не лежащие ни в $C$, ни в $S$. b) Покажите, что граф $G$ содержит $G_{10}$. ## 28. Найди второй! Пусть в графе $G$ все 2 -расчёски содержатся в 2 -гребешках, он содержит $G_{10}$ на вершинах $v_{0}, v_{2}, \ldots, v_{9}$ и не содержит 3-расчёску и её дополнение. a) Найдите в графе $G$ десять вершин $v_{0}^{\prime}, v_{1}^{\prime}, \ldots, v_{9}^{\prime}$ такие, что для любых различных $i$ и $j$ вершины $v_{i}$ и $v_{j}^{\prime}$ соединены ребром, если и только если вершины $v_{i}$ и $v_{j}$ не соединены. b) Найдите в графе $G$ подграф $2 G_{10}$. А где расчёски? Заметим, что в любом графе $2 G_{10}$ нет 3-расчёски и её дополнения, но зато есть 2 -расчёски, не вложенные в 2 -гребешки, если предположить, что, например, вершины $v_{2}$ и $v_{2}^{\prime}$ соединены. Далее предполагаем, что в графе $G$ все 2 -расчёски содержатся в 2 -гребешках, он содержит $2 G_{10}$ и не содержит 3 -расчёску и её дополнение. 29. а) Покажите, что вершины подграфа $2 G_{10} v_{2}$ и $v_{2}^{\prime}$ не соединены. b) Покажите, что вершины $2 G_{10}$ можно перенумеровать так, что получится $2 P$. c) Докажите теорему. ## РешЕния Для краткости будем иногда называть независимое множество антикликой, множество вершин графа $X$ будем обозначать через $V(X)$. Решение задачи 1. Возьмем в несвязном объединении графов $G_{1}$ и $G_{2}$ максимальное независимое множество $S$ и максимальную клику $C$. Очевидно, что $S$ пересекается с каждым из графов $G_{1}$ и $G_{2}$, а $C$ содержится в одном из них, скажем, в $G_{1}$. Тогда $S \cap G_{1}$ и $C \cap G_{1}$ пересекаются как максимальное независимое множество и максимальная клика в $G_{1}$, а значит, пересекаются $S$ и $C$. Решение задачи 2. Если в $C I S$-графе нет треугольников, то в нем нет П-подграфов: иначе две средние вершины этого подграфа образуют максимальную клику (обозначим ее $C$ ), а две крайние вершины образуют независимое множество, которое не пересекается с $C$ и содержится в каком-то максимальном независимом множестве, также не пересекающемся с $C$. Рассмотрим любую компоненту связности нашего графа. Если в ней есть две вершины на расстоянии не меньше 3 , то есть и две вершины на расстоянии ровно 3 . Кратчайший путь, их соединяющий, состоит из четыpex вершин, и подграф на этих вершинах является П-графом, что невозможно. Значит в каждой компоненте связности любые две вершины соединены путем, в котором не более двух ребер. Предположим, что в графе есть цикл нечетной длины, возьмем тогда наименьший из них. В нем будет хотя бы 5 вершин, и пары его несоседних вершин не будут соединены ребрами (иначе цикл делится на два меньших, один из которых нечетный). Но тогда любые четыре подряд идущие вершины цикла образуют П-граф, что невозможно. Значит, нечетных циклов нет. Несложно доказать, что такой граф является двудольным: его вершины можно разделить на две части так, что у любого ребра концы будет лежать в разных частях. Итак, каждая компонента связности является двудольным графом. Он должен быть полным (любые две вершины из разных долей соединены): ведь длина кратчайшего пути между двумя несоединенными вершинами из разных долей не меньше 3. Таким образом, наш изначальный граф был несвязным объединением нескольких полных двудольных графов. Покрасить вершины такого графа в два цвета требуемым образом очень легко: в каждой компоненте связности красим одну долю в первый цвет, а другую - во второй. Для завершения решения осталось проверить, что любой такой граф удовлетворяет условию задачи. Заметим, что раз полный граф является $C I S$-графом, то и полный двудольный граф (как дополнение к несвязному объединению двух полных графов) тоже. Но тогда и любое несвязное объединение полных двудольных графов является $C I S$-графом. В этом месте уместно сделать следующее очевидное замечание: граф является $C I S$-графом тогда и только тогда, когда любые клику и антиклику в нем можно дополнить до пересекающихся клики и антиклики. Решение задачи 3. Граф, получающийся из графа $G$ удалением вершины $v$ (и выходящих из нее ребер), обозначим через $G-v$. Пусть $G$ был $C I S$-графом, а $G-v$ нет, причем вершина $v$ была соединена в графе $G$ с единственной вершиной $u$. Тогда в $G-v$ есть непересекающиеся максимальные клика $C$ и антиклика $S$. Дополним их соответственно до максимальной клики $C^{\prime}$ и максимальной антиклики $S^{\prime}$ в графе $G$. Ясно, что $C^{\prime}$ и $S^{\prime}$ будут пересекаться только по вершине $v$. Но тогда в $C$ не может быть никаких вершин кроме $u$ (так как из $v$ выходит ребро лишь в вершину $u$ ). С другой стороны, если максимальная клика состоит из одной вершины, то эта вершина изолирована (то есть не соединена с другими вершинами) и значит содержится в любой максимальной антиклике. Поэтому $C$ и $S$ пересекаются по вершине $u$. Противоречие. Решение задачи 4. а) Предположим противное, тогда в графе $G$ есть непересекающиеся максимальные клика $C$ и независимое множество $S$, а в $C$ есть хорошая вершина $v$. Если существует вершина $u \in V(S)$, соединеная с $v$, то все вершины $C$ смежны с $u$ (т.к. $v$ хорошая), что противоречит максимальности $C$. Значит $v$ не соединена ни с одной вершиной $S$, но это противоречит максимальности $S$. b) Примером служит цикл длины 4. Решение задачи 5. Пусть граф является объединением пересекающихся клики $C$ и независимого множества $S$. Ясно, что $C$ и $S$ пересекаются ровно по одной вершине (назовем ее $u$ ). Рассмотрим любую максимальную клику $C^{\prime}$. Если она содержит вершину из $S$, то ровно одну, и эта вершина является хорошей (поскольку все ее соседи лежат в клике $C$ ). Если же все вершины $C^{\prime}$ лежат в $C$, но не в $S$, то в силу максимальности $C^{\prime}$ получаем $C^{\prime}=C$, и значит в $C^{\prime}$ все-таки есть вершина из $S$, а именно $u$ - противоречие. Итак, в любой клике нашего графа есть хорошая вершина - значит он является $C I S$-графом по задаче 4. Решение задачи 6. Пусть $G$ - граф, полученный подстановкой $G_{1}$ в $G_{2}$. Разберем два случая. Пусть $G_{1}$ не был $C I S$-графом, тогда в нем найдутся непересекающиеся максимальные клика $C$ и антиклика $S$. Рассмотрим в $G$ любую максимальную клику, содержащую $C$, и любую максимальную антиклику, содержащую $S$. Они пересекаются (так как $G$ является $C I S$-графом), но их общая вершина должна быть соединена с вершинами из $C$ и не соединена с вершинами из $S$, что противоречит определению подстановки. Пусть $G_{2}$ не был $C I S$-графом, тогда в нем найдутся непересекающиеся максимальная клика $C$ и независимое множество $S$. Заметим, что вершина, которую мы заменяем на граф $G_{1}$, не может лежать сразу и в $C$, и в $S$. Поэтому после подстановки один из графов $C$ и $S$ (например $C$ ), не изменится (и останется максимальной кликой), а другой ( $S$ ) либо тоже не изменится (и останется максимальной антикликой), либо в нем одна вершина заменится на $G_{1}$ (тогда остальные вершины $S$ вместе с любой максимальной антикликой $G_{1}$ дадут максимальную антиклику $S^{\prime}$ ). В любом случае находим непересекающиеся клику и антиклику в $G$ - противоречие. Значит и $G_{1}$, и $G_{2}$ были $C I S$-графами, что и требовалось доказать. Решение задачи 7. Рассмотрим произвольный граф $G$ и для каждой его максимальной клики дорисуем к графу вершину и соединим её со всеми вершинами этой клики и только с ними (процедуру следует проделать один раз). В полученном графе (по построению) каждая максимальная клика содержит ровно одну добавленную вершину, которая, очевидно, хорошая. Таким образом (по задаче 4) полученный граф является $C I S$-графом, а исходный является его подграфом ## Решение задачи 8. a) Рассмотрим все возможные графы, являющиеся объединением непересекающихся максимальных клики и независимого множества. Все такие графы получаются следующим способом: рисуем клику $C$, рисуем независимое множество $S$, после чего каждую вершину $C$ соединяем хотя бы с одной вершиной $S$ (но не со всеми сразу). Занумеруем эти графы натуральными числами (например, так: сначала нумеруем (в произвольном порядке) такие графы из трех вершин, потом - из четырех, и так далее; на каждом шаге число нумеруемых графов будет конечным). Граф с номером $i$ обозначим через $G_{i}$. Тогда $H_{i}$ строим так: добавляем к $G_{i}$ одну вершину и соединяем ее со всеми вершинами максимальной клики $G_{i}$ (и только с ними). При этом никакой $C_{i}$ не будет $C I S$-графом, зато каждый $H_{i}$ будет $C I S$-графом. Ясно, что любой $C I S$-граф $G$, содержащий $G_{i}$, содержит и $H_{i}$ - достаточно дополнить клику и антиклику в $G_{i}$ до максимальных в $G$, и мы получим вершину, соединенную с кликой из $G_{i}$ и не соединенную с антикликой из $G_{i}$ (то есть получим $H_{i}$ ). И наоборот, если есть граф $G$, не являющийся $C I S$-графом, то в нем есть максимальные клика и антиклика, которые не пересекаются. Подграф на их вершинах будет одним из $G_{i}$ и очевидно не будет дополняться до $H_{i}$. b) Предположим, что существует такой конечный набор, будем считать, что максимальное количество вершин в графах $G_{i}, H_{i}$ равно $n$. Рассмотрим тогда $n$-расческу и $n$-гребешок. Несложно проверить, что если в гребешке любой $G_{i}$ дополняется до $H_{i}$, то и в расческе тоже. Однако гребешок явялется $C I S$-графом, а расческа - нет. Противоречие. Решение задачи 9. а) Пусть клика $C$ и независимое множество $S$ в объединении дают весь граф $G$, кроме вершины $v$. Пусть $C^{\prime}$ - это множество смежных с $v$ вершин $S$, а $S^{\prime}$ - это множество несмежных с $v$ вершин $S$. Тогда клика $C^{\prime} \cup v$ должна дополняться до максимальной клики, пересекающейся с $S$, следовательно, существует вершина $s \in S^{\prime}$, смежная со всеми вершинами $C^{\prime}$ и с вершиной $v$. Аналогично, существует вершина $c \in C^{\prime}$, не смежная со всеми вершинами $S^{\prime}$ и не смежная с вершиной $v$. Не умаляя общности, ребро $(s, c)$ проведено, тогда найдем в $C$ вершину $c^{\prime}$, не смежную с $s$ (она по построению не будет соединена с $v$ ). Теперь осталось посмотреть на максимальную клику, состоящую из всех смежных с $s$ вершинами и на любое максимальное независимое множество, содержащее $c^{\prime}$ и $v$, убедиться в том, что они не могут пересекаться ни по вершине $v$ (потому, что она не соединена с лежащей в клике $c^{\prime}$ ), ни внутри $C$ (потому, что $c^{\prime}$ не смежна с $s$ ), ни внутри $S$ (потому, что $s$ смежна с $v$ ). b),с) См. статью cis1.pdf. Решение задачи 10. а) (Решение предложено командой Василия Мокина, Виктора Омельяненко и Виктора Садкова.) Назовем эту пару множеств $C$ и $S$ и рассмотрим полный ориентированный двудольный граф с множеством вершин $V(C) \cup V(S)$ и ребрами, проведенными по правилу: если в исходном графе вершины $c \in C$ и $s \in S$ были смежны, то в новом графе ребро будет идти от $c$ к $s$, в противном случае от $s$ к $c$. Так, как $S$ была максимальной антикликой, любая вершина остального графа была смежна хотя бы с одной вершиной $S$, в частности это означает, что в новом графе из любой вершины $C$ выходит (в ориентированном смысле!) хотя бы одно ребро. Аналогично, из любой вершины $S$ так же выходит хотя бы одно ребро. Таким образом, в нашем ориентированном графе исходящяя степень каждой вершины не меньше 1 , следовательно, в нем есть цикл (под циклом в ориентированном графе понимаем ориентированный цикл). Тогда рассмотрим цикл наименьшей длины $A_{1} A_{2} \ldots A_{2 n}$ (длина четная так, как граф был двудолен), если $2 n>4$, то в зависимости от ориентации ребра $A_{1} A_{4}$ в графе можно найти либо цикл $A_{1} A_{2} A_{3} A_{4}$, либо $A_{1} A_{4} A_{5} \ldots A_{2 n}$, меньший минимального. Значит, $2 n=4$. Тогда несложно проверить, что в исходном графе на вершины $A_{1}, A_{2}$, $A_{3}, A_{4}$ индуцировался П-граф. b) Например, $A$-граф, он же 2-гребешок. Решение задачи 11. а) Является частным случаем нижеследующего пункта b) рассмотрим любую максимальную клику, содержащую первую группу вершин расчески (они образуют клику) и любую максимальную антиклику, содержащую вторую группу вершин расчески (они образуют антиклику). В их пересечении будет хотя бы одна вершина, которая при добавлении к расческе даст веник. Решение задачи 12. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-156.jpg?height=368&width=1416&top_left_y=1912&top_left_x=320) Рис. 6 ## Основное свойство графов Галлаи. Решение задачи 13. Пусть в условии дан $d$-граф $G$, а $v$ - некоторая его вершина. Заметим, что найдётся не более одного цвета такого, что удаление из $G$ вершины $v$ влечёт несвязность на рёбрах не этого цвета: если при удалении первого цвета $G$ станет несвязным, то все ребра между компонентами связности должны быть обязательно первого цвета, а полный двудольный граф связен, и тогда при удалении любого другого цвета оставшийся граф будет связным. Предположим, что для любой вершины найдётся цвет такой, что граф без этой вершины на рёбрах не этого цвета несвязен. Таких цветов не может быть больше одного, поэтому покрасим вершину в этот цвет. Заметим, что в любом связном графе есть хотя бы две вершины, удаление которых не изменяет связности. Для этого рассмотрим остов графа (остовное дерево), а в нём можно удалить висячие вершины. В нашем случае исходный граф без ребер какого-то фиксированного цвета связен, поэтому для этого цвета обязательно найдутся хотя бы две вершины, удаление которых не нарушает связности графа, то есть покрашенные не в этот цвет. Пусть есть вершина $u$ первого цвета. Пусть $v_{1}, v_{2}, \ldots, v_{n}$ и $w_{1}, w_{2}, \ldots, w_{m}$ - какое-то разбиение остальных вершин графа на два множества, между которыми потеряется связь при выкидывании первого цвета (если вершину $u$ уже выкинули). Все ребра между этими двумя группами первого цвета. Эти два множества будем называть кусками; кусок, в котором хотя бы две вершины, будем называть большим. Если оба куска маленькие, то либо данный $d$-граф - это разноцветный треугольник, либо он не удовлетворяет свойству в условии. Пусть есть хотя бы один большой кусок. Вершин не первого цвета как минимум две, а маленький кусок может содержать максимум одну вершину не первого цвета, поэтому хотя бы один большой кусок содержит вершину не первого цвета. Не умаляя общности, предположим, что это второй кусок, то есть $m \geqslant 2$ и вершина $w_{1}$ второго цвета. Теперь проделаем ту же процедуру для второго цвета. Вершины графа разбиваются на два куска, между которыми все рёбра второго цвета. Так как рёбра $\left(w_{2}, v_{1}\right),\left(w_{2}, v_{2}\right), \ldots,\left(w_{2}, v_{n}\right),\left(v_{1}, w_{3}\right),\left(v_{1}, w_{4}\right), \ldots,\left(v_{1}, w_{m}\right)$ первого цвета, то вершины $w_{2}, v_{1}, v_{2}, \ldots, v_{n}, w_{3}, \ldots, w_{m}$ в одном куске относительно $w_{1}$. Значит, вершина $u$ в другом куске, и из вершины $u$ во все вершины, кроме $w_{1}$, выходят рёбра второго цвета. Рассмотрим вторую вершину $w_{1}^{\prime}$ цвета $k \neq 1$. Проделаем для нее то же самое, что и для $w_{1}$, то есть разобьем остальные вершины на два куска. Посмотрим еще, в какой кусок относительно $u$ она попала. Если в большой, то аналогично вершина $u$ соединена со всеми вершинами, кроме $w_{1}^{\prime}$, рёбрами цвета $k$, но в графе есть еще вершины, кроме $u, w_{1}, w_{1}^{\prime}$ (т.к. в графе хотя бы 4 вершины). Так как в них из $u$ ведут ребра одновременно цвета 2 и $k$, то $k=2$. Получается, что из вершины $u$ выходят рёбра только второго цвета. Значит, данный $d$-граф несвязен на рёбрах не второго цвета. Получаем, что $w_{1}^{\prime}=v_{1}, n=1$ и вершины $w_{2}, w_{3}, \ldots, w_{m}$ первого цвета. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-157.jpg?height=286&width=466&top_left_y=905&top_left_x=795) Рис. 7 Удалим $w_{2}$. Вершины графа разбились на две группы вершин, между которыми рёбра первого цвета. Так как рёбра $\left(u, v_{1}\right),\left(u, w_{2}\right),\left(u, w_{3}\right), \ldots,\left(u, w_{m}\right)$ второго цвета, то вершины $u, v_{1}, w_{3}, \ldots, w_{m}$ в одной группе. Значит, $w_{1}$ в другой. Получаем, что вершина $w_{1}$ соединена с вершинами $u, v_{1}, w_{3}, w_{4}, \ldots, w_{m}$ рёбрами первого цвета. Если $m>2$, то аналогично получаем, что ребро $\left(w_{1}, w_{2}\right)$ первого цвета, то есть из вершины $w_{1}$ выходят рёбра только первого цвета - противоречие. Тогда $m=2$. Если ребро $\left(w_{1}, w_{2}\right.$ ) второго цвета, то данный $d$-граф является П-графом. Если ребро ( $w_{1}, w_{2}$ ) первого цвета, то данный $d$-граф на рёбрах не первого цвета несвязен противоречие. Решение задачи 14. Индукция по количеству вершин. База, скажем, для $n=3$ : очевидно, один из цветов несвязен. Переход. Пусть $n>3$ и все цвета связны. Выбросим одну из вершин (назовем ее $a$ ); если связность не нарушилась, то по предположению индукции треугольник нашелся. Значит, какой-то (скажем, 1) цвет потерял связность. Пусть $C_{1}, \ldots, C_{k}$ - компоненты связности первого цвета в полученном графе; заметим, что $a$ была соединена с любой из них ребром первого цвета. Рассмотрим $C_{i}$ и $C_{j}$ и ребро, соединяющее две их вершины $c_{i}$ и $c_{j}$; оно, б.о.о., имеет цвет 2 . Тогда, если вершина $c_{i}^{\prime}$ (из $C_{i}$ ) связана с $c_{j}$ ребром 3 цвета, то образовался разноцветный треугольник; значит, ребро $c_{i}^{\prime}-c_{j}$ также второго цвета. Продолжая так далее, получаем, что для любых двух компонент $C_{i}$ и $C_{j}$ все ребра между ними одноцветны. Вернемся к исходному графу. Из связности следует, что $a$ соединена ребрами цветов 2 хоть с кем-то. Пусть она соединена ребром 2 с вершиной $c_{r} \in C_{r}$. Если из $C_{r}$ хоть в какую-то $C_{s}$ ведут ребра 3 цвета, то найдем вершину $c_{s} \in C_{s}$ такую, что $a-c_{s}$ цвета 1 (такая, напомним, найдется из связности); тогда $a, c_{r}, c_{s}$ образуют искомый треугольник. Итак, из $C_{r}$ во все другие компоненты ведут ребра только 2 цвета. Аналогично, рассмотрев компоненту $C_{t}$, в которую из $a$ ведет ребро 3 цвета, мы получим, что из $C_{t}$ в другие компоненты ведут только ребра 3 цвета. Это невозможно ни в случае $r=t$ (так как есть еще какие-то компоненты, и непонятно, ребра какого цвета туда идут из $C_{r}$ ), ни в случае $r \neq t$ (тогда непонятно, какого цвета ребра между $C_{r}$ и $C_{t}$ ). Решение задачи 15. Аналогично решению задачи 6 получаем, что данный класс замкнут относительно подстановки. Заметим, что для любой вершины $d$-графа можно рассмотреть набор из $d$ максимальных множеств, каждое из которых свободно от ребер какого-то своего цвета. Из этого выводится, что данный класс точно замкнут. При $d=2$, очевидно, этот класс не наследственный. Решение задачи 16. Утверждение. Пусть в $d$-графе Галлаи есть хотя бы одно ребро $d$-го цвета и хроматическая компонента $d$-го цвета несвязна. Тогда все рёбра между двумя любыми фиксированными компонентами связности этой хроматической компоненты покрашены в один и тот же цвет. См. решение задачи 14. Решение задачи. Будем доказывать спуском по $k(k4$ then depending on orientation of the edge $A_{1} A_{4}$ it is possible to find a cycle smaller than minimal in our graph: either $A_{1} A_{2} A_{3} A_{4}$ or $A_{1} A_{4} A_{5} \ldots A_{2 n}$. So $2 n=4$. Then it is not difficult to check that the $\Pi$-graph was induced on the vertices $A_{1}, A_{2}, A_{3}, A_{4}$ in the initial graph. b) For instance, the $A$-graph, in other words, the settled 2-comb. Solution of Problem 11. a) It is a partial case of the next subproblem. b) Consider an arbitrary maximal clique containing the first group of vertices in the comb (they form a clique) and any maximal anticlique containing the second group of vertices in the comb (they form an anticlique). Their intersection contains at least one vertex, which forms a settled comb by being added to the comb. ## Solution of Problem 12. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-170.jpg?height=389&width=1422&top_left_y=91&top_left_x=320) Fig. 6 ## The main property of Gallai graphs. Solution of Problem 13. Denote the given $d$-graph by $G$, and let $v$ be one of its vertices. Notice that there exists at most one colour such that if we exclude $v$ from $G$ then the subgraph of not this colour is not connected: If $G$ becomes non-connected when we omit the edges of the 1st colour, then all the edges between its connected components are of the 1st colour. But the full bipartite graph is connected, hence if we omit some other colour then the remaining graph will be connected. Suppose that for any vertex $v$ there exists a colour such that $G-v$ is not connected without this colour. We showed that for each vertex there exists at most one such colour (if for a vertex there are no such colours, then we are done). Colour this vertex in this colour. Notice that in any connected graph there exist at least two vertices such that deleting of them does not break connectedness. To show this, take the spanning tree of this graph. Its terminal vertices can be deleted. In our case, the initial graph without any fixed colour is connected, this means that for this colour there exist at least two vertices such that deleting of them does not break the connectedness of the graph, i.e. coloured not in this colour. Suppose $u$ is a vertex of the 1 st colour. Let $v_{1}, v_{2}, \ldots, v_{n}$ and $w_{1}, w_{2}, \ldots, w_{m}$ be a splitting of all the other vertices of the graph such that these components are not connected without the 1st colour (in $G-u$ ). All the edges between these two groups are of the 1st colour. We will name these groups by parts; a part containing at least two vertices will be named big. If both parts are small, then either the given $d$-graph is a 3-coloured triangle or it does not satisfy the conditions of the problem. Suppose that there exists at least one big part. We have at least two vertices not of the 1st colour, while a small part can contain at most one of these vertices. This means that at least one big part contains a vertex not of the 1st colour. Without loss of generality, suppose it is the second part, i.e. $m \geqslant 2$, and $w_{1}$ is of the 2 nd colour. Perform the same procedure for the 2nd colour. The vertices of $G$ split into two parts, such that all the edges between these parts are of the 2nd colour. But the edges $\left(w_{2}, v_{1}\right),\left(w_{2}, v_{2}\right), \ldots,\left(w_{2}, v_{n}\right),\left(v_{1}, w_{3}\right),\left(v_{1}, w_{4}\right), \ldots,\left(v_{1}, w_{m}\right)$ are of the 1 st colour, this means that $w_{2}, v_{1}, v_{2}, \ldots, v_{n}, w_{3}, \ldots, w_{m}$ are in one part with respect to $w_{1}$. Hence, $u$ is in the other part, and all the edges from $u$ to all the other vertices, excluding $w_{1}$, are of the 2 nd colour. Consider the 2 nd vertex $w_{1}^{\prime}$ of colour $k \neq 1$. Perform the same procedure as with $w_{1}$ : Split all the other vertices into two parts and look in which part with respect to $u$ it belongs. If it belongs to a big part, then, similarly, $u$ is connected with all vertices but $w_{1}^{\prime}$ with the edges of colour $k$. Notice that this graph contains some other vertices besides $u, w_{1}$, and $w_{1}^{\prime}$ (since $G$ contains at least 4 vertices). They are connected with $u$ with the edges of colours 2 and $k$ simultaneously, hence, $k=2$. It follows that all the edges incident to $u$ are of the 2 nd colour. Hence, the initial $d$-graph is not connected if we exclude the edges of the 2 nd colour. Contradiction. We obtain that $w_{1}^{\prime}=v_{1}, n=1$, and the vertices $w_{2}, w_{3}, \ldots, w_{m}$ are all of the 1 st colour. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-170.jpg?height=294&width=468&top_left_y=2040&top_left_x=794) Fig. 7 Delete $w_{2}$. The vertices of the graph split into two parts in such a way that all the edges between them are of the 1st colour. Since all the edges $\left(u, v_{1}\right),\left(u, w_{2}\right),\left(u, w_{3}\right), \ldots,\left(u, w_{m}\right)$ are of the 2 nd colour, all the vertices $u, v_{1}, w_{3}, \ldots, w_{m}$ are in one part. Hence, $w_{1}$ is in the other part. We obtain that $w_{1}$ is connected with the vertices $u, v_{1}, w_{3}, w_{4}, \ldots, w_{m}$ with the edges of the 1 st colour. If $m>2$, we obtain in the same way that the edge $\left(w_{1}, w_{2}\right)$ is of the 1 st colour, which means that all the edges incident to $w_{1}$ are of the 1 st colour. Contradiction. Then $m=2$. If the edge $\left(w_{1}, w_{2}\right)$ is of the 2 nd colour, then the given $d$-graph is a $\Pi$-graph. If the edge $\left(w_{1}, w_{2}\right)$ is of the 1 st colour, then the initial $d$-graph without the edges of the 1 st colour is not connected. This is also a contradiction. Solution of Problem 14. Induction on the number of vertices in the graph. The base for $n=3$ is obvious: at least one chromatic component is not connected. Perform the step. Suppose that $n>3$ and all chromatic components are connected. Exclude one vertex (let $a$ ); if all the chromatic components remain connected, then we use the statement of induction and find the 3-coloured triangle. Otherwise one of the chromatic components (suppose the 1st) became disconnect. Let $C_{1}, \ldots, C_{k}$ be the connected components of the 1st colour in the obtained graph; notice that for each of these components there exists an edge from $a$ to this component of the 1st colour. Consider two components $C_{i}$ and $C_{j}$ and an edge connecting two their vertices $c_{i}$ and $c_{j}$; without loss of generality, let it be of 2 nd colour. Take another vertex $c_{i}^{\prime}$ from $C_{i}$. If it is connected with $c_{j}$ with an edge of the $3 \mathrm{rd}$ colour, then we get a 3 -coloured triangle. It means that the edge $c_{i}^{\prime}-c_{j}$ is also of 2 nd colour, and so on. One gets that for any two components $C_{i}$ and $C_{j}$ all the edges between them are of the same colour. Now return to the initial graph. All its chromatic components are connected, this means that $a$ has an edge of colour 2. Let this edge join $a$ with a vertex $c_{r} \in C_{r}$. If there exists a component $C_{s}$ such that the edges between $C_{s}$ and $C_{r}$ are of another colour (denote this colour 3), then find a vertex $c_{s} \in C_{s}$ such that the edge $a-c_{s}$ is of colour 1 (this is possible because the 1st connected component of the initial graph is connected); then $a, c_{r}, c_{s}$ form a 3 -coloured triangle. We showed that all the edges joining $C_{r}$ with all the other components are of colour 2. Similarly, if we take an edge of colour 3 incident to $a$ and consider the component $C_{t}$ incident to the second end of this edge, then we obtain that all the edges connecting $C_{t}$ with other components are of colour 3. But this is impossible: if $r=t$, notice that we have more than one component, and it is not clear edges of which colour link this component with $C_{r}$. If, otherwise, $r \neq t$, then it is not clear edges of which colour join $C_{r}$ and $C_{t}$. Solution of Problem 15. As in the solution of problem 6, we show that this family is closed under substitution. To show the exact closeness, notice that for any vertex from the $d$-graph we can find $d$ sets each of which will be independent of the edges of its colour and will be maximal with this property. When $d=2$, as we noticed above, this family is not hereditary. Solution of Problem 16. An obvious Lemma.Suppose that a Gallai $d$-graph has an edge of $d$ th colour and its $d$ th chromatic component is not connected. Then all the edges between any two fixed connected components of this chromatic component are of the same colour. Its proof is a part of proof of 14 . The solution of the problem. We prove, using descent on $k(kVladimir Gurvich ${ }^{c}$ RRR 17-2006, JuLY 2006 RUTCOR Rutgers Center for Operations Research Rutgers University 640 Bartholomew Road Piscataway, New Jersey 08854-8003 Telephone: $732-445-3804$ Telefax: 732-445-5472 Email: rr@rutcor.rutgers.edu http://rutcor.rutgers.edu/ rrr[^7] # RUtCOR RESEARch Report RRR 17-2006, July 2006 ## ON GRAPHS WHOSE MAXIMAL CLIQUES AND STABLE SETS INTERSECT Diogo V. Andrade Endre Boros Vladimir Gurvich #### Abstract We say that a graph $G$ has the CIS-property and call it a CIS-graph if every maximal clique and every maximal stable set of $G$ intersect. By definition, $G$ is a CIS-graph if and only if the complementary graph $\bar{G}$ is a CISgraph. Let us substitute a vertex $v$ of a graph $G^{\prime}$ by a graph $G^{\prime \prime}$ and denote the obtained graph by $G$. It is also easy to see that $G$ is a CIS-graph if and only if both $G^{\prime}$ and $G^{\prime \prime}$ are CIS-graphs. In other words, CIS-graphs respect complementation and substitution. Yet, this class is not hereditary, that is, an induced subgraph of a CIS-graph may have no CIS-property. Perhaps, for this reason, the problems of efficient characterization and recognition of CIS-graphs are difficult and remain open. In this paper we only give some necessary and some sufficient conditions for the CIS-property to hold. There are obvious sufficient conditions. It is known that $P_{4}$-free graphs have the CIS-property and it is easy to see that $G$ is a CIS-graph whenever each maximal clique of $G$ has a simplicial vertex. However, these conditions are not necessary. There are also obvious necessary conditions. Given an integer $k \geq 2$, a comb (or $k$-comb) $S_{k}$ is a graph with $2 k$ vertices $k$ of which, $v_{1}, \ldots, v_{k}$, form a clique $C$, while others, $v_{1}^{\prime}, \ldots, v_{k}^{\prime}$, form a stable set $S$, and $\left(v_{i}, v_{i}^{\prime}\right)$ is an edge for all $i=1, \ldots, k$, and there are no other edges. The complementary graph $\overline{S_{k}}$ is called an anti-comb (or $k$-anti-comb). Clearly, $S$ and $C$ switch in the complementary graphs. Obviously, the combs and anti-combs are not CIS-graphs, since $C \cap S=\emptyset$. Hence, if a CIS-graph $G$ contains an induced comb or anti-comb then it must be settled, that is, $G$ must contain a vertex $v$ connected to all vertices of $C$ and to no vertex of $S$. However, these conditions are only necessary. The following sufficient conditions are more difficult to prove: $G$ is a CIS-graph whenever $G$ contains no induced 3-combs and 3-anti-combs, and every induced 2comb is settled in $G$. It is an open question whether $G$ is a CIS-graph if $G$ contains no induced 4-combs and 4-anti-combs, and all induced 3-combs, 3-anti-combs, and 2-combs are settled in $G$. We generalize the concept of CIS-graphs as follows. For an integer $d \geq 2$ we define a $d$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ as a complete graph whose edges are colored by $d$ colors (that is, partitioned into $d$ sets). We say that $\mathcal{G}$ is a CIS- $d$-graph (has the CIS- $d$-property) if $\bigcap_{i=1}^{d} C_{i} \neq \emptyset$ whenever for each $i=1, \ldots, d$ the set $C_{i}$ is a maximal color $i$-free subset of $V$, that is, $\left(v, v^{\prime}\right) \notin E_{i}$ for any $v, v^{\prime} \in C_{i}$. Clearly, in case $d=2$ we return to the concept of CIS-graphs. (More accurately, CIS-2-graph is a pair of two complementary CIS-graphs.) We conjecture that each CIS-d-graph is a Gallai graph, that is, it contains no triangle colored by 3 distinct colors. We obtain results supporting this conjecture and also show that if it holds then characterization and recognition of CIS- $d$-graphs are easily reduced to characterization and recognition of CIS-graphs. We also prove the following statement. Let $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ be a Gallai $d$-graph such that at least $d-1$ of its $d$ chromatic components are CIS-graphs, then $\mathcal{G}$ has the CIS- $d$-property. In particular, the remaining chromatic component of $\mathcal{G}$ is a CIS-graph too. Moreover, all $2^{d}$ unions of $d$ chromatic components of $\mathcal{G}$ are CISgraphs. Key words: CIS-graphs, CIS-property, clique, clique-kernel intersection property, graph, independent set, simplicial vertex, stable graph, stable set, substitution. ## 1 Introduction. ### 1.1 CIS-graphs and simplicial vertices Given a graph $G$, we say that it has the $C I S$-property, or equivalently that $G$ is a $C I S$-graph, if every maximal clique $C$ and every maximal stable set $S$ in $G$ intersect. Obviously, they may have at most one common vertex and hence $|C \cap S|=1$. It is convenient to represent a CIS-graph $G$ as a 2-dimensional box partition, that is, a matrix whose rows and columns are labeled respectively by the maximal cliques and stable sets of $G$ and whose entries are the (unique) vertices of the corresponding intersections. For example, Figure 1 shows four CIS-graphs and their intersection matrices. More examples are given in Figures 6, 7 and 10. The CIS-property appears in the survey [6] (under the name clique-kernel intersection property) but no related results are mentioned. Indeed, natural problems of efficient characterization and recognition of the CIS-graphs look difficult and remain open. Perhaps, one of the reasons is that the CIS-property is not hereditary. Indeed, if $C \cap S=\{v\}$ then $C \backslash\{v\}$ and $S \backslash\{v\}$ may become disjoint maximal clique and stable set after $v$ is deleted. On the positive side, by definition, the CIS-property is self-complementary, that is, $G$ is a CIS-graph if and only if the complementary graph $\bar{G}$ is a CIS-graph. We start with a simple sufficient condition. Given a graph $G=(V, E)$, a vertex $v \in V$ is called simplicial if its neighborhood $N[v]$ is a clique. Clearly, if a maximal clique $C$ of $G$ contains a simplicial vertex $v$ then it is a private vertex of $C$, that is, $v$ cannot belong to any other maximal clique, except $C$. Vice versa, every private vertex $v$ of a maximal clique $C$ is simplicial, since in this case $N[v]=C$. Moreover, in this case $C \cap S \neq \emptyset$ for every maximal stable set $S$ in $G$. Indeed, if $S \cap(C \backslash\{v\})=\emptyset$ then $v \in S$, since $S$ is maximal. Thus, we obtain the following statement. Proposition 1. Graph $G$ is a CIS-graph whenever (s) every maximal clique of $G$ has a simplicial vertex. Let us remark that condition (s): is only sufficient but not necessary. For example, (s) holds for the first graph in Figure 1 but not for the other three graphs. Let us also remark that (s) does not hold for both graphs in Figure 2. Furthermore, (s) holds for the graphs of Figures 6, 7, and 10 and (s) does not hold for the graphs of Figures 4, 5, and 9, because they are not CIS-graphs. By Proposition 1, given an arbitrary graph $G$, we can get a CIS-graph $G^{s}$ just adding a simplicial (private) vertex $v_{C}$ to each maximal clique $C$ of $G$ that does not have one. Let us remark that we have to add such a vertex to $C$ even when $C \cap S \neq \emptyset$ for each maximal stable set $S$ in $G$, since otherwise $C$ may become disjoint from a new maximal stable set of $G^{s}$; consider, for example, $G=\overline{C_{6}}$. Obviously, the size of $G^{s}$ may be exponential in the size of $G$. Corollary 1. Any graph $G$ is an induced subgraph of a CIS-graph. | | 2468 | 15 | 37 | 146 | 368 | 582 | 724 | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 123 | 2 | 1 | 3 | 1 | 3 | 2 | 2 | | 345 | 4 | 5 | 3 | 4 | 3 | 5 | 4 | | 567 | 6 | 5 | 7 | 6 | 6 | 5 | 7 | | 781 | 8 | 1 | 7 | 1 | 8 | 8 | 7 | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-180.jpg?height=369&width=374&top_left_y=455&top_left_x=1269) | | 258 | 147 | 369 | 249 | 537 | 816 | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 123 | 2 | 1 | 3 | 2 | 3 | 1 | | 456 | 5 | 4 | 6 | 4 | 5 | 6 | | 789 | 8 | 7 | 9 | 9 | 7 | 8 | | 267 | 2 | 7 | 6 | 2 | 7 | 6 | | 591 | 5 | 1 | 9 | 9 | 5 | 1 | | 834 | 8 | 4 | 3 | 4 | 3 | 8 | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-180.jpg?height=423&width=479&top_left_y=845&top_left_x=1211) | | $13579 a$ | 1470 | $369 b$ | $258 a$ | 260 | $48 b$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $12 b$ | 1 | 1 | $\mathrm{~b}$ | 2 | 2 | $\mathrm{~b}$ | | 234 | 3 | 4 | 3 | 2 | 2 | 4 | | 456 | 5 | 4 | 6 | 5 | 6 | 4 | | 678 | 7 | 7 | 6 | 8 | 6 | 8 | | 890 | 9 | 0 | 9 | 8 | 0 | 8 | | $0 a b$ | $\mathrm{a}$ | 0 | $\mathrm{~b}$ | $\mathrm{a}$ | 0 | $\mathrm{~b}$ | | 168 | 1 | 1 | 6 | 8 | 6 | 8 | | 249 | 9 | 4 | 9 | 2 | 2 | 4 | | 380 | 3 | 0 | 3 | 8 | 0 | 8 | | $46 a$ | $\mathrm{a}$ | 4 | 6 | $\mathrm{a}$ | 6 | 4 | | $50 b$ | 5 | 0 | $\mathrm{~b}$ | 5 | 0 | $\mathrm{~b}$ | | $27 b$ | 7 | 7 | $\mathrm{~b}$ | 2 | 2 | $\mathrm{~b}$ | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-180.jpg?height=571&width=508&top_left_y=1279&top_left_x=1202) | | 159 | 168 | 249 | 276 | 348 | 357 | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 123 | 1 | 1 | 2 | 2 | 3 | 3 | | 456 | 5 | 6 | 4 | 6 | 4 | 5 | | 789 | 9 | 8 | 9 | 7 | 8 | 7 | | 147 | 1 | 1 | 4 | 7 | 4 | 7 | | 258 | 5 | 8 | 2 | 2 | 8 | 5 | | 369 | 9 | 6 | 9 | 6 | 3 | 3 | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-180.jpg?height=368&width=382&top_left_y=1872&top_left_x=1268) Figure 1: Four CIS-graphs and their intersection matrices. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-181.jpg?height=476&width=968&top_left_y=270&top_left_x=570) Figure 2: Complements to the first two graphs in the previous Figure. (Obviously, for every graph $G$ the intersection matrix of $\bar{G}$ is the transposed of the intersection matrix of $G$.) Proof. Indeed, for any graph $G$ the CIS-graph $G^{s}$ contains $G$ as an induced subgraph. Thus, CIS-graphs cannot be characterized in terms of forbidden induced subgraphs. This is not surprising, since the CIS-property is not hereditary. Remark 1. Interestingly, this mapping $f: G \rightarrow G^{\text {s }}$ can be viewed as a "bridge" between perfect graphs and cooperative games [2]. Given a graph $G=(V, E)$, let $\mathcal{C}=\mathcal{C}_{G}$ and $\mathcal{S}=\mathcal{S}_{G}$ be, respectively, the families of all maximal cliques and stable sets of $G$. Let us assign a player (voter) $i_{C}$ to each maximal clique $C \in \mathcal{C}_{G}$ and an outcome (candidate) $a_{S}$ to each maximal stable set $S \in \mathcal{S}_{G}$. Furthermore, to every vertex $v \in V$ let us assign a coalition of players $K_{v}=\left\{i_{C} \mid v \in C\right\} \subseteq \mathcal{C}_{G}$ and block of outcomes $B_{v}=\left\{a_{S} \mid v \in S\right\} \subseteq \mathcal{S}_{G}$. Then let us introduce a family of coalitions $\mathcal{K}_{G}=\left\{K_{v} \mid v \in V\right\}$ and define an effectivity function $\mathcal{E}_{G}: 2^{\mathcal{C}} \times 2^{\mathcal{S}} \rightarrow\{0,1\}$ by formula $\mathcal{E}_{G}(K, B)=1$ iff $K_{v} \subseteq K$ and $B_{v} \subseteq B$ for some $v \in V$. It is proved in [2, 3, 4] that the following claims are equivalent: (i) Graph $G$ is perfect; (ii) Effectivity function $\mathcal{E}_{G}$ is stable; (iii) Family of coalitions $\mathcal{K}_{G}$ is stable. (iv) Family of coalitions $\mathcal{K}_{G^{s}}$ is partitionable. A family of sets is called partitionable if every its minimal balanced subfamily is a partition. A family of coalitions or an effectivity function is called stable if the corresponding core is not empty for any utility function. We refer to [2, 3, 4] for accurate definitions. ### 1.2 Almost CIS-graphs and split graphs We will call a graph $G=(V, E)$ an almost $C I S$-graph if every (maximal) clique $C$ and stable set $S$ in $G$ intersect, except for a unique pair $\left(C_{0}, S_{0}\right)$. By definition, almost CIS-graphs are closed under complementation. However, unlike CIS-graphs, they are not closed under substitution. Notice that, by definition, the families of CIS- and almost CIS-graphs are disjoint. Let us recall that $G=(V, E)$ is a split graph if $V=C_{0} \cup S_{0}$, where $C_{0}$ and $S_{0}$ are (maximal) clique and stable set, respectively. Foldes and Hammer [15] showed that split graphs are exactly $\left(C_{4}, \overline{C_{4}}, C_{5}\right)$-free graphs. It is not difficult to see that every split graph is either a CIS-graph or an almost CISgraph. More precisely, the following claim holds. Proposition 2. Let $G=(V, E)$ be a split graph in which $C_{0}$ and $S_{0}$ are maximal and $V=C_{0} \cup S_{0}$. If $C_{0} \cap S_{0} \neq \emptyset$ then $G$ is a CIS-graph, otherwise, if $C_{0} \cap S_{0}=\emptyset$, then $G$ is an almost CIS-graph in which $\left(C_{0}, S_{0}\right)$ is the only disjoint pair. Proof. Obviously, for each maximal clique $C$ and stable set $S$ in $G$ we have: $C_{0} \cap S \neq \emptyset$ unless $S=S_{0}$ and $C \cap S_{0} \neq \emptyset$ unless $C=C_{0}$. Let us assume that both intersections are non-empty (then, clearly, each of them consists of a single vertex) and denote $C_{0} \cap S$ by $v_{S}$ and $C \cap S_{0}$ by $v_{C}$. If $v_{C}=v_{S}$ then $C \cap S=\left\{v_{S}\right\}=\left\{v_{S}\right\}$. Otherwise, if $\left(v_{C}, v_{S}\right) \in E$ then $C \cap S=\left\{v_{S}\right\}$; if $\left(v_{C}, v_{S}\right) \notin E$ then $C \cap S=\left\{v_{C}\right\}$. In any case $C \cap S \neq \emptyset$. Thus, if $C \cap S=\emptyset$ then $C=C_{0}, S=S_{0}$, and $C_{0} \cap S_{0}=\emptyset$. Conjecture 1. Every almost CIS-graph is a split graph. By [15], it would be sufficient to show that almost CIS graphs are ( $\left.C_{4}, \overline{C_{4}}, C_{5}\right)$-free. It would be also sufficient to prove that if $\left(C_{0}, S_{0}\right)$ is the unique disjoint pair of an almost CIS-graph $G=(V, E)$ then $V=C_{0} \cup S_{0}$, that is, $V^{\prime}=V \backslash\left(C_{0} \cup S_{0}\right)=\emptyset$. However, we can only show that the corresponding induced subgraph $G\left[V^{\prime}\right]$ cannot be a split graph. In other words, $G$ is not almost CIS-graph whenever $V=C_{0} \cup S_{0} \cup C \cup S$, where $C$ and $C_{0}$ are cliques, $S$ and $S_{0}$ are stable sets, $C_{0}$ and $S_{0}$ are maximal, and $C_{0} \cap S_{0}=\emptyset$. However, we cannot prove that $V^{\prime}=\emptyset$. Obviously, given a split graph $G$ with a unique disjoint pair $C_{0} \cap S_{0}=\emptyset$, we can get a split CIS-graph $G_{0}$ by adding to $G$ the new vertex $v_{0}$ which is connected to each vertex of $C_{0}$ and to no vertex of $S_{0}$. In other words, $v_{0}$ is a simplicial vertex of $C \cup\left\{v_{0}\right\}$ in $G_{0}$ and of $\bar{S} \cup\left\{v_{0}\right\}$ in $\overline{G_{0}}$. ## $1.3 \quad P_{4}$-free CIS-graphs We proceed with the following simple observation: every $P_{4}$-free graph is a CIS-graph; see e.g. $[12,13,16,18,19,21,25,37]$. In fact, a stronger claim holds. We say that a set $T \subseteq V$ is a transversal of the hypergraphs $\mathcal{H} \subseteq 2^{V}$ if $T \cap H \neq \emptyset$ for all hyperedges $H \in \mathcal{H}$. The family of minimal transversals of $\mathcal{H}$ is denoted by $\mathcal{H}^{d}$ and is called the dual of $\mathcal{H}$. Given a graph $G=(V, E)$ we assign to it two hypergraphs, $\mathcal{C}=\mathcal{C}_{G}$ the collection of all maximal cliques of $G$, and $\mathcal{S}=\mathcal{S}_{G}$ the collections of all its maximal stable sets. Proposition 3 ([18, 21, 25]). A graph $G$ has no induced $P_{4}$ if and only if the hypergraphs $\mathcal{C}$ and $\mathcal{S}$ of all maximal cliques and stable sets of $G$ are dual hypergraphs. Furthermore, $P_{4}$-free graphs are closely related to read-once Boolean functions and 2person positional games, see for definitions, e.g., [17, 20, 21, 25]. Remark 2. Read-once Boolean functions can be efficiently characterized, since their cooccurrence graphs are $P_{4}$-free, $[12,13,18,19,21,25]$. Moreover, the normal forms of positional 2-person games with perfect information can be characterized by Proposition 3 $[19,20,21]$. Such a normal form is exactly the intersection matrix of the maximal cliques and stable sets of the corresponding graph, where the final positions (outcomes) of the game are in one-to-one correspondence with the vertices of this graph. See an example in Figure 3, where the monotone Boolean functions $F_{S}=13 \vee 24$ and $F_{C}=(1 \vee 3)(2 \vee 4)$ corresponding to the hypergraphs $\mathcal{S}=\{(1,3),(2,4)\}$ and $\mathcal{C}=\{(1,2),(2,3),(3,4),(4,1)\}$ are read-once. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-183.jpg?height=277&width=297&top_left_y=775&top_left_x=594) $$ \begin{aligned} \mathcal{C} & =\{(1,2)(2,3)(3,4)(4,1)\} \\ \mathcal{S} & =\{(1,3)(2,4)\} \end{aligned} $$ ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-183.jpg?height=383&width=633&top_left_y=1085&top_left_x=426) | | 12 | 23 | 34 | 41 | | :---: | :---: | :---: | :---: | :---: | | 13 | 1 | 3 | 3 | 1 | | 24 | 2 | 2 | 4 | 4 | Figure 3: A $P_{4}$-free graph and the corresponding positional and normal game forms However, the absence of induced $P_{4} \mathrm{~S}$ is only sufficient but not necessary for the CISproperty to hold. Let a graph $G$ contain an induced $P_{4}$ defined by $\left(v_{1}, v_{1}^{\prime}\right),\left(v_{2}, v_{2}^{\prime}\right),\left(v_{1}, v_{2}\right)$. The clique $\left\{v_{1}, v_{2}\right\}$ and stable set $\left\{v_{1}^{\prime}, v_{2}^{\prime}\right\}$ are disjoint. Hence, they cannot be maximal in $G$ if it is a CIS-graph. In other words, $G$ must contain a fifth vertex $v_{0}$ such that $\left(v_{0}, v_{1}\right),\left(v_{0}, v_{2}\right)$ are edges, while $\left(v_{0}, v_{1}^{\prime}\right),\left(v_{0}, v_{2}^{\prime}\right)$ are not. In this case we will say that $P_{4}$ is settled by $v_{0}$, cf. [32]. Let us note that the graph induced by $\left\{v_{0}, v_{1}, v_{2}, v_{1}^{\prime}, v_{2}^{\prime}\right\}$ is a CIS-graph, see Figure 6. Thus, every induced $P_{4}$ in a CIS-graph must be settled. This condition is necessary, as we argued above, yet, it is not sufficient, according to the following examples. ### 1.4 Combs and anti-combs Given an integer $k \geq 2$, a comb (or $k$-comb) $S_{k}$ is defined as a graph with $2 k$ vertices $k$ of which form a clique $C=\left\{v_{1}, \ldots, v_{k}\right\}$, while the remaining $k$ form a stable set $S=$ $\left\{v_{1}^{\prime}, \ldots, v_{k}^{\prime}\right\}$. In addition, $S_{k}$ contains the perfect matching $\left(v_{i}, v_{i}^{\prime}\right)$ for $i=1, \ldots, k$, and there are no more edges in $S_{k}$. Let us note that graphs $S_{2}$ and $P_{4}$ are isomorphic. Furthermore, $S_{3}$ contains 3 induced $S_{2}$ and all 3 are settled. More generally, $S_{k}$ contains $k$ induced $S_{k-1}$ and they all are settled. Figure 4 shows $S_{k}$, for $k=2,3$, and 4 . ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-184.jpg?height=434&width=1262&top_left_y=434&top_left_x=386) Figure 4: Combs $S_{k}$, for $k=2,3$ and 4 The complementary graph $\overline{S_{k}}$ is called an anti-comb (or $k$-anti-comb). Figure 5 shows $\overline{S_{k}}$ for $k=2,3$, and 4 . ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-184.jpg?height=442&width=1262&top_left_y=1139&top_left_x=410) Figure 5: Anti-combs $\overline{S_{k}}$, for $k=2,3$ and 4 . Clearly, the sets $S$ and $C$ are switched in $S_{k}$ and $\overline{S_{k}}$. It is also clear that combs and anti-combs are not CIS-graphs, since they contain a maximal clique $C$ and stable set $S$ that are disjoint. Hence, if a CIS-graph $G$ contains an induced comb $S_{k}$ (respectively, anti-comb $\left.\overline{S_{k}}\right)$ then it must be settled, that is, $G$ must contain a vertex $v_{0}$ adjacent to each vertex of $C$ and to no vertex of $S$. Thus, the following condition is necessary for the CIS-property to hold. (COMB) Every induced comb and anti-comb must be settled in $G$. Figures 6 and 7 show settled combs and anti-combs. It is easy to verify that they are complementary CIS-graphs. Hence, the corresponding intersection matrices are mutually transposed. The following obvious properties of combs and anti-combs are worth summarizing: - The 2-comb $S_{2}$ and 2-anti-comb $\overline{S_{2}}$ are isomorphic, while the $k$-comb $S_{k}$ and $k$-anticomb $\overline{S_{k}}$ are not isomorphic for $k>2$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-185.jpg?height=385&width=280&top_left_y=293&top_left_x=451) | | $01^{\prime} 2^{\prime}$ | $12^{\prime}$ | $1^{\prime} 2$ | | :---: | :---: | :---: | :---: | | 012 | 0 | 1 | 2 | | $11^{\prime}$ | $1^{\prime}$ | 1 | $1^{\prime}$ | | $22^{\prime}$ | $2^{\prime}$ | $2^{\prime}$ | 2 | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-185.jpg?height=443&width=443&top_left_y=681&top_left_x=361) | | $01^{\prime} 2^{\prime} 3^{\prime}$ | $12^{\prime} 3^{\prime}$ | $1^{\prime} 23^{\prime}$ | $1^{\prime} 2^{\prime} 3$ | | :---: | :---: | :---: | :---: | :---: | | 0123 | 0 | 1 | 2 | 3 | | $11^{\prime}$ | $1^{\prime}$ | 1 | $1^{\prime}$ | $1^{\prime}$ | | $22^{\prime}$ | $2^{\prime}$ | $2^{\prime}$ | 2 | $2^{\prime}$ | | $33^{\prime}$ | $3^{\prime}$ | $3^{\prime}$ | $3^{\prime}$ | 3 | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-185.jpg?height=437&width=483&top_left_y=1141&top_left_x=364) | | $01^{\prime} 2^{\prime} 3^{\prime} 4^{\prime}$ | $12^{\prime} 3^{\prime} 4^{\prime}$ | $1^{\prime} 23^{\prime} 4^{\prime}$ | $1^{\prime} 2^{\prime} 34^{\prime}$ | $1^{\prime} 2^{\prime} 3^{\prime} 4$ | | :---: | :---: | :---: | :---: | :---: | :---: | | 01234 | 0 | 1 | 2 | 3 | 4 | | $11^{\prime}$ | $1^{\prime}$ | $1^{\prime}$ | $1^{\prime}$ | $1^{\prime}$ | $1^{\prime}$ | | $22^{\prime}$ | $2^{\prime}$ | $2^{\prime}$ | 2 | $2^{\prime}$ | $2^{\prime}$ | | $33^{\prime}$ | $3^{\prime}$ | $3^{\prime}$ | $3^{\prime}$ | 3 | $3^{\prime}$ | | $44^{\prime}$ | $4^{\prime}$ | $4^{\prime}$ | $4^{\prime}$ | $4^{\prime}$ | 4 | Figure 6: Settled combs $S_{k}$, for $k=2,3$ and 4. - The $k$-comb $S_{k}$ contains $\binom{k}{m}$ induced $m$-combs $S_{m}$ that are all settled in $S_{k}$, yet, it contains no induced $m$-anti-combs $\overline{S_{m}}$ for $m>2$; respectively, the $k$-anti-comb $\overline{S_{k}}$ contains $\binom{k}{m}$ induced $m$-anti-combs $\overline{S_{m}}$ that are all settled in $\overline{S_{k}}$, yet, it contains no induced $m$-combs $S_{m}$ for $m>2$. - The settled $k$-comb and anti-comb are complementary CIS-graphs. Obviously, COMB is a necessary condition for the CIS-property to hold. Yet, it is not sufficient, as we will see in Section 1.5. Let us introduce the following stronger condition. COMB(3,3) There is no induced 3-comb or 3-anti-comb, and every induced 2-comb is settled in $G$. Our main result claims that this stronger condition already implies the CIS-property. Theorem 1. A graph G is a CIS-graph whenever it satisfies COMB(3,3). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-186.jpg?height=438&width=1214&top_left_y=294&top_left_x=434) Figure 7: Settled anti-combs $\overline{S_{k}}$, for $k=2,3$ and 4 . We give the proof in Section 2. It contains a complicated case analysis in which one of the cases is especially interesting and results in a remarkable graph that is "almost" a counterexample to Theorem 1. This graph $2 \mathcal{P}$ (see Figure 8) consists of two identical copies of the Petersen graph induced by the vertices $v_{0}, \ldots, v_{9}$ and $v_{0}^{\prime}, \ldots, v_{9}^{\prime}$ respectively. Furthermore, $\left(v_{i}^{\prime}, v_{j}\right)$ is an edge if and only if $\left(v_{i}, v_{j}\right)$ is not, for all $i \neq j$. Ten remaining pairs, $\left(v_{i}, v_{i}^{\prime}\right), i=0, \ldots, 9$, are uncertain, that is, configuration $2 \mathcal{P}$ represents in fact $2^{10}$ possible graphs rather than one graph. The following properties of $2 \mathcal{P}$ are easy to see. (a) $2 \mathcal{P}$ is isomorphic to its complement. (b) $2 \mathcal{P}$ is regular of "degree 9.5 ", that is, each vertex is incident to 9 edges and belongs to one uncertain pair. (c) For every two vertices $u, v$ there is an automorphism $\alpha$ of $2 \mathcal{P}$ such that $\alpha(u)=v$. (d) None of the $2^{10}$ graphs of $2 \mathcal{P}$ contains an induced 3-comb or 3-anti-comb. (e) Every induced 2-comb in all $2^{10}$ graphs of $2 \mathcal{P}$ involves a pair $v_{i}, v_{i}^{\prime}$ for some $i=0, \ldots, 9$. In fact, 36 induced 2-combs appear, whenever we substitute a pair $v_{i}, v_{i}^{\prime}$ by an edge (or by a non-edge). It is easy to see that none of these 2 -combs can be settled by a vertex of $2 \mathcal{P}$, and if it is settled by a new vertex then an unsettled 3-comb or 3-anti-comb always appears. Thus, the case under consideration does not lead to a counterexample, and a complete case analysis yields the proof of Theorem 1, see Section 2. Four examples of CIS-graphs satisfying condition $\operatorname{COMB}(3,3)$ are given in Figure 1. It would be interesting to analyze the following relaxations of condition $\operatorname{COMB}(3,3)$ that are still stronger than COMB. Given integers $i, j \geq 2$, we say that a graph $G$ satisfies condition COMB $(i, j)$ if all induced combs and anti-combs in $G$ are settled and, moreover, $G$ contains no induced $S_{i}$ and $\overline{S_{j}}$. By a natural convention we have $\operatorname{COMB}=\operatorname{COMB}(\infty, \infty)$. Clearly, condition $\operatorname{COMB}(2,2)$ implies the CIS-property, since it means that the graph is $P_{4}$-free. In fact, we have $\operatorname{COMB}(2,2) \equiv \operatorname{COMB}(2, i) \equiv \operatorname{COMB}(i, 2)$ for every $i \geq 2$, ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-187.jpg?height=448&width=1168&top_left_y=267&top_left_x=478) Figure 8: Graph 2P . since the 2-comb $S_{2} \equiv P_{4}$ is self-complementary and every comb and anti-comb contains an induced 2-comb. Furthermore, condition $\operatorname{COMB}(i, j)$ is monotone in the sense that it implies $\operatorname{COMB}\left(i^{\prime}, j^{\prime}\right)$ for all $i \leq i^{\prime}$ and $j \leq j^{\prime}$, and symmetric, in the sense that $\operatorname{COMB}(i, j)$ implies the CIS-property if and only if $\operatorname{COMB}(j, i)$ does (due to the fact that $G$ is a CISgraph if and only if its complement $\bar{G}$ is a CIS-graph). According to Theorem 1, condition COMB(3,3) implies the CIS-property. However, it is not known whether $\operatorname{COMB}(4,4)$ or $\operatorname{COMB}(3, j)$ for some $j \geq 4$ imply the CIS-property or not. Certainly, condition $\operatorname{COMB}(5,4)$ does not, as the next section shows. ## $1.5(n, k, \ell)$-graphs and their complements The following graph $G=(V, E)$ was suggested by Ron Holzman in 1994. It has $\binom{5}{1}+\binom{5}{2}$ $=5+10=15$ vertices, where subsets $S=\left\{v_{1}, \ldots, v_{5}\right\}$ and $C=\left\{v_{12}, \ldots, v_{45}\right\}$ induce a stable set and clique, respectively; $V=C \cup S$ (hence, $G$ is a split graph); furthermore, every pair $\left(v_{i}, v_{i j}\right)$, where $i, j=1, \ldots, 5$ and $i \neq j$, is an edge, and there are no more edges. Let us denote this graph by $G(5,1,2)$, see Figure 9 . ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-187.jpg?height=274&width=962&top_left_y=1782&top_left_x=570) Figure 9: Graph $G(5,1,2)$ was constructed by Ron Holzman in 1994. It is easy to verify that $G(5,1,2)$ contains no induced 5 -combs and 4 -anti-combs. In section 3 we will show that all induced combs and anti-combs in $G(5,1,2)$ are settled. For example, the 4 -comb induced by vertices $\left(v_{12}, v_{13}, v_{14}, v_{15}, v_{2}, v_{3}, v_{4}, v_{5}\right)$ is settled by $v_{1}$ and the 3 -anti-comb induced by $\left(v_{12}, v_{13}, v_{23}, v_{1}, v_{2}, v_{3}\right)$ is settled by $v_{45}$, etc. Thus, the graph $G(5,1,2)$ satisfies condition $\operatorname{COMB}(5,4)$, however, it is not a CIS-graph, since $C \cap S=\emptyset$. Let us notice that the settled extension of $G(5,1,2)$ is a CIS-graph, see Figure 10. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-188.jpg?height=417&width=962&top_left_y=425&top_left_x=570) | | 01213141523 | | | | | | | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | | 2425343545 | 112131415 | 212232425 | 313233435 | 414243445 | 515253545 | | | 012345 | 0 | 1 | 2 | 3 | 4 | 5 | | | 12345 | 12 | 12 | 12 | 3 | 4 | 5 | | | 13245 | 13 | 13 | 2 | 13 | 4 | 5 | | | 14235 | 14 | 14 | 2 | 3 | 14 | 5 | | | 15234 | 15 | 15 | 2 | 3 | 4 | 15 | | | 23145 | 23 | 1 | 23 | 23 | 4 | 5 | | | 24135 | 24 | 1 | 24 | 3 | 24 | 5 | | | 25134 | 25 | 1 | 25 | 3 | 4 | 25 | | | 34125 | 34 | 1 | 2 | 34 | 34 | 5 | | | 35124 | 35 | 1 | 2 | 35 | 4 | 35 | | | 45123 | 45 | 1 | 2 | 3 | 45 | 45 | | Figure 10: Settled $G(5,1,2)$. We generalize the above example as follows. Given integers $n, k, \ell$ such that $n>k \geq 1$ and $n>\ell \geq 1$, consider a set $S$ (respectively, $C$ ) consisting of $\binom{n}{k}$ (respectively, $\binom{n}{\ell}$ ) vertices labeled by $k$-subsets (respectively, by $\ell$-subsets) of a ground $n$-set. Let us introduce the graph $G(n, k, \ell)$ on the vertex-set $C \cup S$ such that $S$ is a stable set, $C$ is a clique, and a vertex of $S$ is adjacent to a vertex of $C$ if and only if the corresponding $k$-set is either a subset or a superset of the corresponding $\ell$-set. Obviously, $G(n, k, \ell)$ is not a CIS-graph, since $C \cap S=\emptyset$. However, some of these graphs satisfy the condition COMB, for example, $G(5,1,2)$. Moreover, $G(5,1,2)$ satisfies the stronger condition $\operatorname{COMB}(5,4)$. By definition, $G(n, 1,1)=S_{n}$ is an $n$-comb and $G(n, n-1,1)=\overline{S_{n}}$ is an $n$-anti-comb. Furthermore, it is easy to see that (i) the graphs $G(n, k, \ell)$ and $G(n, n-k, n-\ell)$ are isomorphic. Hence, without loss of generality we can assume that $k \leq \ell$ and even that $k<\ell$, since $G(n, k, k)$ is just a comb $S_{\binom{n}{k}}$. Then, from the simple fact that a set contains an element if and only if the complementary set does not contain it, we derive (ii ) the graphs $G(n, k, 1)$ and $G(n, 1, n-k)$ are complementary. Thus, the graphs $G(n, k, n-1)$ and $G(n, n-k, 1)$ are isomorphic by (i) and complementary to $G(n, 1, k)$ by (ii). Hence, without loss of generality we can assume that $\ell \leq n-2$. Summarizing, we will assume in the sequel that $$ 1 \leq k<\ell \leq n-2 $$ In section 3 we will prove the following two claims analyzing the existence of unsettled anti-combs and combs in $G(n, k, \ell)$. ## Theorem 2. (i) Each induced anti-comb in $G(n, k, \ell)$ is settled whenever $$ n>\frac{k+1}{k} \ell $$ (ii) An unsettled induced anti-comb exists in $G(n, k, \ell)$ whenever $$ k+\ell \leq n \leq \frac{k+1}{k} \ell $$ ## Theorem 3. (a) Each induced comb is settled in $G(n, 1, \ell)$, and it is settled in $G(n, 2, \ell)$ whenever $$ n<2 \ell-3 $$ (b) An unsettled induced comb exists in $G(n, k, \ell)$ for $k \geq 2$ whenever $$ n \geq \frac{k}{k-1} \ell-\frac{r}{k-1} \quad \text { or } \quad n=\frac{k}{k-1} \ell-\frac{r}{k-1}-1 \quad \text { and } \ell>r+k^{2}-k $$ where $r \equiv \ell(\bmod k-1)$ and $r \in\{2,3, \ldots, k\}$. Let us denote by $\mathbf{G}$ the subfamily of graphs $G(n, k, \ell)$ whose induced combs and anticombs are all settled and $n, k, \ell$ satisfy (1.1). Corollary 2. For $k=1$ and $k=2$ the membership in $\mathbf{G}$ is characterized as follows: $$ \begin{gathered} G(n, 1, \ell) \in \mathbf{G} \quad \text { iff } \quad n>2 \ell \\ G(n, 2, \ell) \in \mathbf{G} \quad \text { iff } \quad 2 \ell-3>n>(3 / 2) \ell \end{gathered} $$ Proof. By (1.1) we have $n \geq \ell+2 \geq \ell+k$, whenever $k \leq 2$, and thus, by Theorem 2 , all induced anti-combs are settled in $G(n, k, \ell)$ for $k \leq 2$ if and only if $n>\frac{k+1}{k} \ell$. This and (a) of Theorem 3 then implies the claim for $k=1$. If $k=2$ then $G(n, 2, \ell)$ has an unsettled comb, by (b) of Theorem 3, if $n \geq 2 \ell-2$ or if $n=2 \ell-3$ and $\ell>4$, since $r=2$ in this case. However, if $n=2 \ell-3$ then $\ell \geq 5$ by (1.1). Hence, the second condition holds automatically, and therefore by (a) and (b) of Theorem 3 , we can conclude that $G(n, 2, \ell)$ has an unsettled comb if and only if $n \geq 2 \ell-3$. Thus, for $k=1$ we get $\{G(5,1,2), G(6,1,2), G(7,1,2), G(7,1,3), \ldots\} \subseteq \mathbf{G}$ and for $k=2$ we get $\{G(14,2,9), G(16,2,10), G(17,2,11), G(18,2,11), G(19,2,12), G(20,2,13), \ldots\} \subseteq \mathbf{G}$. Remark 3. Notice that conditions (i) and (ii) of Theorem 2 provide an almost complete characterization of the existence of unsettled anti-combs in $G(n, k, \ell)$. However, it is not clear if condition $n \geq k+\ell$ in part (ii) is necessary. Note that if $k \leq 2$, then this condition holds automatically by (1.1). For instance, we do not know if $G(8,3,6)$ has an unsettled anticomb. Computer experiments show that there are no unsettled $m$-anti-combs for $m \leq 10$. In any case, $G(8,3,6)$ has an unsettled 6 -comb, by Theorem 3. Let us also note that we know much less about combs. For instance, we could only treat the case of $k \leq 2$ in (a) of Theorem 3, though we conjecture that a similar claims can hold for all $k$. For example, $G(10,3,8)$ is the smallest graph for which we do not know if it contains an unsettled comb or anti-comb. Based on the proofs of the above theorems and on several numerical examples we conjecture that membership in $\mathbf{G}$ can be characterized by inequalities of the approximate form $$ \frac{k}{k-1} \ell+O(k) \geq n \geq \frac{k+1}{k} \ell-O(k) $$ This is certainly the case for $k \leq 2$, by Corollary 2. By definition, in a graph $G=G(n, k, \ell) \in \mathbf{G}$, as well as in its complement $\bar{G}$, all induced combs and anti-combs are settled, that is, both $G$ and $\bar{G}$ satisfy the condition COMB. Let us notice however that $\bar{G}$ is not an $(n, k, \ell)$-graph unless $k=1$. (Recall that $G(n, 1, \ell)$ and $G(n, n-\ell, 1)$ are complementary.) It seems that every non-CIS-graph satisfying COMB contains either an induced $G(n, k, \ell) \in$ $\mathbf{G}$ or its complement. At least, we have no counterexample for this claim. Let us add that, unlike the case of combs and anti-combs, one graph from $\mathbf{G}$ may contain another as an unsettled induced subgraph. For example, $G(6,1,2)$ contains an unsettled induced $G(5,1,2)$, while in $G(7,1,2)$ all induced $G(5,1,2)$ are settled. Yet, in $G(7,1,2)$ there is an unsettled induced $G(6,1,2)$. Vice versa, in $G(7,1,3)$ each induced $G(6,1,2)$ is settled but there are unsettled induced $G(5,1,2)$. Further, in $G(8,1,3)$, all induced $G(5,1,2)$ and $G(7,1,2)$ are settled but there are unsettled induced $G(6,1,2)$ and $G(7,1,3)$. Due to this "non-transitivity", in order to enforce the CIS-property for a graph $G$, it seems easier to assume that all induced subgraphs from $\mathbf{G}$ as well as their complements are settled in $G$. Of course, it is even simpler to assume that $G$ does not contain such subgraphs at all. Conjecture 2. If $G$ contains no induced $G(5,1,2)$ nor its complement $G(5,3,1)$ and all induced combs and anti-combs are settled in $G$ then $G$ is a CIS-graph. We remark here that $G(n, k, l)$ contains an induced $G\left(n^{\prime}, k^{\prime}, l^{\prime}\right)$ whenever $n^{\prime} \leq n, k^{\prime} \leq k$, and $l^{\prime} \leq l$. Remark 4. Let us note that CIS-graphs and perfect graphs look somewhat similar. Both classes are closed with respect to complementation and substitution. Odd holes and antiholes are similar to combs and anti-combs. The following two tests look similar too: whether $G$ contains an induced odd hole or anti-hole and whether $G$ contains an induced unsettled comb or anti-comb. It seems that CIS-graphs, like perfect graphs, may allow a simple characterization and a polynomial recognition algorithm (that may be very difficult to obtain, though). However, there are dissimilarities, too. The property of perfectness is hereditary, unlike the CIS-property. Also, there are non-CIS-graphs in which all induced combs and anticombs are settled. (By Conjecture 2, every such graph contains an induced $G(5,1,2)$ or its complement $G(5,3,1)$.) Remark 5. CIS-graphs were recently mentioned (under the name of stable graphs) in [38], where it is shown that recognition of stable graphs is a special case of a difficult problem (strongly bipartite bihypergraph recognition problem) introduced in this paper. Based on this observation, the authors conjecture that recognition of stable graphs is co-NP-complete. However, we conjecture that this problem is polynomial. The following relaxation of the CIS-property was considered in [26] and [36]. Triangle condition: for every maximal stable set $S$ and every edge $(u, v)$ such that $u, v \notin S$ there exists a vertex $w \in S$ such that vertices $u, v, w$ induce a clique. Obviously, each CIS-graph has this property. ### 1.6 Gallai's and CIS- $d$-graphs. Let us generalize the concept of a CIS-graph as follows. For a given integer $d \geq 2$, a complete graph whose edges are colored by $d$ colors $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ is called a $d$-graph. To a given $d$-graph $\mathcal{G}$ let us assign a family of $d$ hypergraphs $\mathcal{C}=\mathcal{C}(G)=\left\{\mathcal{C}_{i} \mid i=1, \ldots, d\right\}$ on the common vertex-set $V$, where the hyperedges of $\mathcal{C}_{i}$ are all inclusion maximal subsets of $V$ containing no edges of color $i$. We say that $\mathcal{G}$ is a CIS-d-graph (has the CIS-d-property) if $\bigcap_{i=1}^{d} C_{i} \neq \emptyset$ for all selections $C_{i} \in \mathcal{C}_{i}$ for $i=1, \ldots, d$. Obviously, such an intersection can contain at most one vertex. If $d=2$ then we obtain the original concept of $C I S$-graphs. (More accurately, CIS-2-graph is a pair of two complementary CIS-graphs.) Similarly to CIS-graphs, CIS-d-graphs also satisfy a natural requirement that can be considered as a generalization of settling. Assume that $X_{i}$ is a clique in the subgraph $G_{i}=\left(V, \cup_{j \neq i} E_{j}\right)$ for $i=1, \ldots, d$, and that $\cap_{i=1}^{d} X_{i}=\emptyset$. Then, these cliques cannot all be maximal and, hence, there must be a vertex $x \in V$ such that for every $i=1, \ldots, d$ and $y \in X_{i}$ we have $(x, y) \notin E_{i}$. We will say in this case that $\left\{X_{1}, \ldots, X_{d}\right\}$ are settled by $x$. Given a CIS- $d$-graph $\mathcal{G}$, let us assign to it a $d$-dimensional table $g=g(\mathcal{G})$, that is, a mapping $g: \mathcal{C}_{1} \times \cdots \times \mathcal{C}_{d} \rightarrow V$ defined by the rule: $g\left(C_{1}, \ldots, C_{d}\right)=v$ whenever $\{v\}=\cap_{i=1}^{d} C_{i}$. Let us observe that this $d$-dimensional array is partitioned by the elements of $V$ into $n=|V|$ sub-arrays called boxes, since the following implication holds: if $g\left(C_{1}^{\prime}, \ldots, C_{d}^{\prime}\right)=g\left(C_{1}^{\prime \prime}, \ldots, C_{d}^{\prime \prime}\right)=v$, then $v$ belongs to all these $2 d$ sets, and hence, $g\left(C_{1}, \ldots, C_{d}\right)=v$ for all $2^{d}$ choices $C_{i} \in\left\{C_{i}^{\prime}, C_{i}^{\prime \prime}\right\}, i=1, \ldots, d$. Let us further introduce two special edge colored graphs. Let $\Pi$ denote the 2-colored graph whose both chromatic components form a $P_{4}$, that is, $V=\left\{v_{1}, v_{2}, v_{3}, v_{4}\right\} ; E_{1}=$ $\left\{\left(v_{1}, v_{2}\right),\left(v_{2}, v_{3}\right),\left(v_{3}, v_{4}\right)\right\}$, and $E_{2}=\left\{\left(v_{2}, v_{4}\right),\left(v_{4}, v_{1}\right),\left(v_{1}, v_{3}\right)\right\}$. Furthermore, let $\Delta$ denote the 3-colored triangle, for which $V=\left\{v_{1}, v_{2}, v_{3}\right\}, E_{1}=\left\{\left(v_{1}, v_{2}\right)\right\}, E_{2}=\left\{\left(v_{2}, v_{3}\right)\right\}$, and $E_{3}=\left\{\left(v_{3}, v_{1}\right)\right\}$. Figure 12 illustrates these graphs. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-192.jpg?height=266&width=614&top_left_y=878&top_left_x=707) Figure 11: Colored $\Pi$ and $\Delta$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-192.jpg?height=276&width=616&top_left_y=1272&top_left_x=706) Figure 12: Colored $\Pi$ and $\Delta$ (in black and white for printing). Proposition 4 ([19,21]). Every $\Pi$ - and $\Delta$-free d-graph is a CIS-d-graph. In fact, a stronger claim holds. Proposition 5 ( $[19,20,21]) . A$ d-graph $\mathcal{G}$ is $\Pi$ - and $\Delta$-free if and only if the corresponding mapping $g(\mathcal{G})$ defines the normal form of a positional d-person game with perfect information whose final positions (outcomes of the game) are in one-to-one correspondence with the vertices of $\mathcal{G}$. For example, let us consider the $\Pi$ - and $\Delta$-free 3 -graph $\mathcal{G}$ given in Figure 13. For this graph we have $\mathcal{C}_{1}=\{(1,3),(2,4)\}, \mathcal{C}_{2}=\{(1,2,4),(2,3,4)\}$, and $\mathcal{C}_{3}=\{(1,2,3),(1,3,4)\}$. The mapping $g(\mathcal{G})$ and the corresponding positional game are shown in Figure 13. Another example of a $\Pi$ - and $\Delta$-free 3 -graph is given in Figure 14. In this case $\mathcal{C}_{1}=$ $\{(1),(2,3,4)\}, \mathcal{C}_{2}=\{(1,3),(1,2,4)\}$, and $\mathcal{C}_{3}=\{(1,2,3),(1,3,4)\}$. The mapping $g(\mathcal{G})$ and the corresponding positional game are shown in Figure 14. Of course, the condition that a $d$-graph $\mathcal{G}$ must be $\Pi$ - and $\Delta$-free is only sufficient but not necessary for the CIS- $d$-property to hold. On the other hand, the following condition is clearly ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-193.jpg?height=274&width=297&top_left_y=300&top_left_x=662) $$ \begin{aligned} \mathcal{C}_{1} & =\{(13)(24)\} \\ \mathcal{C}_{2} & =\{(124)(234)\} \\ \mathcal{C}_{3} & =\{(123)(134)\} \end{aligned} $$ | 13 | | 24 | | | | :---: | :---: | :---: | :---: | :---: | | 1 | 1 | 124 | 2 | 74 | | 3 | 3 | 234 | 2 | 7 | | 1 | 1 | | 1 | 1 | | 2 | 3 | | 2 | 3 | | 3 | 4 | | 3 | | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-193.jpg?height=380&width=628&top_left_y=615&top_left_x=497) Figure 13: A $\Pi$ - and $\Delta$-free 3 -graph and the corresponding positional and normal game forms. necessary. Given a $d$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ and a partition $P_{1} \cup \ldots \cup P_{\delta}=\{1, \ldots, d\}$ of its colors, let us define a $\delta$-graph $\mathcal{G}^{\prime}=\left(V ; E_{1}^{\prime}, \ldots, E_{\delta}^{\prime}\right)$ by setting $E_{i}^{\prime}=\cup_{j \in P_{i}} E_{j}, i=1, \ldots, \delta$ and call $\mathcal{G}^{\prime}$ the $\delta$-projection of $\mathcal{G}$. Proposition 6. Let $\mathcal{G}$ be a CIS-d-graph whose set of colors $\{1, \ldots, d\}$ is partitioned into $\delta$ non-empty subsets $(2 \leq \delta \leq d)$ then the corresponding $\delta$-graph $\mathcal{G}^{\prime}$ is a CIS- $\delta$-graph. In particular, in case $\delta=2$ we must get two complementary CIS-graphs. The following conjecture is open since 1978 . Conjecture 3. ([19]) Every CIS-d-graph is $\Delta$-free. By Proposition 6, it would suffice to prove this conjecture for $d=3$. In this case, it was verified up to $n=12$ vertices by a computer code written by Steven Jaslar in 2003. We will consider this conjecture in Section 4 and show that, similarly to combs and anti-combs, all $\Delta$ s in a CIS-d-graph must be settled, and it takes two vertices to settle a $\Delta$ (see Section 4.2). Although there are $d$-graphs in which all $\Delta \mathrm{s}$ are settled, yet, it seems impossible to have settled simultaneously all combs and anti-combs in all 2-projections of these $d$-graphs, a condition that is necessary by Proposition 6 . In the literature $\Delta$-free $d$-graphs are called Gallai's graphs, since they were introduced by Gallai in [16]. We will call them Gallai's $d$-graphs which is more accurate. They are well studied $[1,8,9,10,14,24,27,28]$. Conjecture 3 means that CIS- $d$-graphs form a subfamily of Gallai's $d$-graphs. Next, we will characterize Gallai's CIS-d-graphs in terms of CIS-graphs. Hence, to characterize CIS- $d$-graphs it would suffice to do it for $d=2$ and prove Conjecture 3. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-194.jpg?height=274&width=294&top_left_y=300&top_left_x=664) $$ \begin{aligned} & \mathcal{C}_{1}=\{(1)(234)\} \\ & \mathcal{C}_{2}=\{(13)(124)\} \\ & \mathcal{C}_{3}=\{(123)(134)\} \end{aligned} $$ | 1 | | 234 | | | | :---: | :---: | :---: | :---: | :---: | | 1 | 1 | 13 | 3 | | | 1 | 1 | 124 | 2 | 4 | | 1 | 1 | | 1 | 1 | | 2 | 3 | | 2 | 3 | | 3 | 4 | | 3 | | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-194.jpg?height=314&width=528&top_left_y=648&top_left_x=547) Figure 14: A $\Pi$ - and $\Delta$-free 3-graph and the corresponding positional and normal game forms. First, let us note that both Gallai's and CIS-d-graphs are closed under substitution. (For Gallai's $d$-graphs this is well known [8, 24].) Moreover, the inverse claims hold too. Proposition 7. Let us substitute a d-graph $\mathcal{G}^{\prime \prime}$ for a vertex $v$ of a $d$-graph $\mathcal{G}^{\prime}$ and denote the obtained d-graph by $\mathcal{G}=\mathcal{G}\left(\mathcal{G}^{\prime}, v, \mathcal{G}^{\prime \prime}\right.$ ). Then $\mathcal{G}$ is a Gallai (respectively, CIS-) d-graph if and only if both $\mathcal{G}^{\prime}$ and $\mathcal{G}^{\prime \prime}$ are Gallai's (respectively, CIS-) d-graphs. In case $d=2$ this proposition implies the similar property for CIS-graphs. Proposition 8. Let us substitute a graph $G^{\prime \prime}$ for a vertex $v$ of a graph $G^{\prime}$ and denote the obtained graph by $G=G\left(G^{\prime}, v, G^{\prime \prime}\right)$. Then $G$ is a CIS-graph if and only if both $G^{\prime}$ and $G^{\prime \prime}$ are CIS-graphs. Let us recall, however, that CIS- $d$-property is not hereditary, that is, an induced subgraph of a CIS-d-graph may have no CIS- $d$-property. In particular, for $d=2$, this means that an induced subgraph of a CIS-graph may have no CIS-property. Here and in the sequel we assume that the set of colors $[d]=\{1, \ldots, d\}$ is the same for all considered $d$-graphs, while some chromatic components may be trivial (edge-empty). For example, by a 2 -graph we mean a $d$-graph with at most 2 non-trivial chromatic components. It is known that each Gallai $d$-graph can be obtained from 2-graphs by substitutions. More precisely, the following claim holds. Proposition 9 (Cameron and Edmonds, [8]; Gyárfás and Simonyi, [24]). For each Gallai $d$-graph $\mathcal{G}$ there exist a 2 -graph $\mathcal{G}_{0}$ with $n$ vertices and $n$ Gallai $d$-graphs $\mathcal{G}_{1}, \ldots, \mathcal{G}_{n}$ such that $\mathcal{G}$ is obtained by substituting $\mathcal{G}_{1}, \ldots \mathcal{G}_{n}$ for $n$ vertices of $\mathcal{G}_{0}$. In [24], this claim is derived from the following Lemma. Lemma 1 ([16], [8], and [24]). Every Gallai $d$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ with $d \geq 3$ has a color $i \in[d]$ that does not span $V$, or in other words, the graph $G_{i}=\left(V, E_{i}\right)$ is not connected. Remark 6. It is interesting to compare Lemma 1 with the following Lemma from [19, 21]. If a d-graph $\mathcal{G}$ is $\Pi$ - and $\Delta$-free then there exists a unique color $i \in[d]$ such that the complement of the $i$-th chromatic component, $\overline{G_{i}}$, is disconnected. Gyárfás and Simonyi remark that Lemma 1 "is essentially a content of Lemma (3.2.3) in [16]" and they derive Proposition 9 from it as follows. If $d \leq 2$ we are done. Otherwise, we have a color $i \in[d]$ such that graph $G_{i}=\left(V, E_{i}\right)$ is disconnected. It is not difficult to show that for each two of its connected components all edges between them are of the same color $j$ (clearly, $j \neq i$ ), since otherwise a $\Delta$ appears. Collapsing these components into vertices we get a smaller $(d-1)$-graph which is still $\Delta$-free, by Proposition 7 . By induction, $\mathcal{G}_{1}, \ldots, \mathcal{G}_{n}$ and $\mathcal{G}_{0}$ can be constructed as required. Moreover, applying the above decomposition recursively, we can represent an arbitrary Gallai $d$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ by a substitution-tree $T(\mathcal{G})$ whose leaves are associated to 2-graphs. If $d \leq 2$ then $\mathcal{G}$ itself is a 2-graph and $T(\mathcal{G})$ is reduced to one vertex. If $d \geq 3$ then, by Lemma 1 , there is a color $i \in[d]$ such that the i-th component $G_{i}=\left(V, E_{i}\right)$ does not span $V$, or in other words, it is disconnected. Let $W \subset V$ be a connected component of $G_{i}$. Furthermore, let $G^{\prime \prime}=G[W]$ be the subgraph of $\mathcal{G}$ induced by $W$, while $G^{\prime}$ be obtained from $\mathcal{G}$ by contracting $W$ to a single new vertex $v$. Then, as it was shown above, substituting $\mathcal{G}^{\prime \prime}$ for $v$ in $\mathcal{G}^{\prime}$ we get $\mathcal{G}=\mathcal{G}\left(\mathcal{G}^{\prime}, v, \mathcal{G}^{\prime \prime}\right.$ ); see Figure 15. If $\mathcal{G}^{\prime}$ (or $\mathcal{G}^{\prime \prime}$ ) is a 2-graph then it becomes a leaf of $T(\mathcal{G})$. Otherwise, if $\mathcal{G}^{\prime}$ (or $\mathcal{G}^{\prime \prime}$ ) has more than 2 non-trivial chromatic components, we decompose it further in the same way until only 2-graphs remain. They are the leaves of the obtained decomposition tree $T(\mathcal{G})$, as required. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-195.jpg?height=388&width=529&top_left_y=1479&top_left_x=792) Figure 15: Decomposing $\mathcal{G}$ by the tree $T(\mathcal{G})$; substituting $\mathcal{G}^{\prime \prime}$ for $v$ in $\mathcal{G}^{\prime}$ to get $\mathcal{G}$. It is well-known that decomposing a given graph into connected components can be executed in linear time. Hence, given $\mathcal{G}$, its decomposition tree $T(\mathcal{G})$ can be constructed in linear time, too. Remark 7. As defined above, tree $T(\mathcal{G})$ is not unique, since several chromatic components of $\mathcal{G}$ may be disconnected and any connected component of any chromatic component can be chosen as $W$ for the decomposition. Let us note, however, that the corresponding vertex sets are nested. More precisely, if $E_{i}^{a}, E_{j}^{b}$ are connected components of colors $i, j \in[d]$ then the corresponding vertex-sets $V_{i}^{a}, V_{j}^{b} \subseteq V$ are either disjoint, or one of them is a subset of the other. Yet, the latter case cannot take place when $i=j$. Let us also note that in general $T(\mathcal{G})$ can be extended further, since some 2-graphs also can be decomposed by substitution. Obviously, the decomposition of a 2-graph $\mathcal{G}=\left(V ; E_{1}, E_{2}\right)$ is reduced to a decomposition of a graph, namely, of a chromatic component, $G_{1}=G\left(V, E_{1}\right)$ or $G_{2}=G\left(V, E_{2}\right)$. In general, decomposing graphs (as well as d-graphs, digraphs, Boolean functions, etc.) by substitution is known as their modular decomposition. $A$ module is a set $X \subseteq V$ such that no member of $V \backslash X$ distinguishes members of $X . A$ set family $\mathcal{F}$ is called decomposable if $X \cap Y, X \cup Y, X \backslash Y, Y \backslash X$, and $X \Delta Y=(X \backslash Y) \cup(Y \backslash X)$ are in $\mathcal{F}$ whenever $X, Y \in \mathcal{F}$ and $X \cap Y \neq \emptyset$. Möring [34] proved that the family of modules is decomposable and hence, there is a unique canonical modular decomposition tree. In general, modular decomposition is more complicated than decomposition of Gallai's $d$ graphs. There have been a number of $O\left(n^{4}\right), O\left(n^{3}\right), O(m n), O\left(n^{2}\right), O(n+m \log n)$ algorithms. Finally, $O(m+n)$ algorithms were given by Cournier and Habib [11] and McConnell and Spinrad [33]. Some linear time algorithms work for graphs, d-graphs, digraphs, and Boolean functions. See [6, 7, 33, 34, 35] for a survey on modular decomposition. We make use of the decomposition tree $T(\mathcal{G})$ to recognize whether $\mathcal{G}$ is a CIS-d-graph. Obviously, by Propositions 7, we can extend Proposition 9 as follows. Proposition 10. A Gallai $d$-graph $\mathcal{G}$ has the CIS-d-property if and only if all $n+1 d$-graphs $\mathcal{G}_{1}, \ldots, \mathcal{G}_{n}$ and $\mathcal{G}_{0}$ from Proposition 9 have this property. Thus, every Gallai's $C I S$ - $d$-graph can be obtained from $C I S$-2-graphs by recursive substitutions, and hence, a characterization or polynomial recognition algorithm of CIS-graphs would provide one for the Gallai $C I S$ - $d$-graphs too. From Propositions 6, 9, and 10 we will derive the following two claims. Proposition 11. A Gallai d-graph $\mathcal{G}$ is a CIS-d-graph if and only if all $d$ of its chromatic components are CIS-graphs. The "only if" part follows from Proposition 6 and "if" part can be strengthened as follows. Proposition 12. Given a Gallai d-graph $\mathcal{G}$ such that at least $d-1$ of its $d$ chromatic components are CIS-graphs, then $\mathcal{G}$ is a CIS-d-graph. In particular, the remaining chromatic component of $\mathcal{G}$ must be a CIS-graph. In the next subsection we generalize the last claim by showing that it holds not only for CIS-graphs but also for perfect graphs and, in fact, for every family of graphs satisfying some simple requirements. Yet, of course, it is essential that $\mathcal{G}$ is a Gallai $d$-graph. For example, let us consider a 3-graph $\mathcal{G}$ in Figure 16. Graphs $G_{1}$ and $G_{2}$ are isomorphic, each of them is a settled 2-comb with one isolated vertex. Hence, they are CIS-graphs. Yet, $G_{3}$ is not, since the stable set $S=\{2,3,5,6\}$ and clique $C=\{1,4\}$ are disjoint. However, $\mathcal{G}$ is not Gallai's 3-graph, for example, $\{5,6,1\}$ as well as $\{1,2,3\}$ is a $\Delta$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-197.jpg?height=420&width=514&top_left_y=270&top_left_x=794) Figure 16: A non-Gallai 3-graph in which $G_{1}$ and $G_{2}$ are CIS-graphs, while $G_{3}$ is not. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-197.jpg?height=414&width=529&top_left_y=815&top_left_x=792) Figure 17: A non-Gallai 3-graph in which $G_{1}$ and $G_{2}$ are CIS-graphs, while $G_{3}$ is not (in black and white for printing). ### 1.7 Extending Cameron-Edmonds-Lovász' Theorem Cameron, Edmonds, and Lovász [9] proved the statement similar to Proposition 12 for perfect graphs: given a Gallai $d$-graph, if at least $d-1$ of its chromatic components are perfect graphs, then the remaining component is a perfect graph, too. Later, Cameron and Edmonds [8] showed that, in fact, the statement holds for any family of graphs that is closed under: (i) substitution, (ii) complementation, and (iii) taking induced subgraphs. For example, it holds for $P_{4}$-free graphs, or in other words, for the components of $\Pi$ - and $\Delta$-free $d$-graphs [19]. However, CIS-graphs satisfy only (i) and (ii) but not (iii). Nevertheless, the statement holds for them too; see Proposition 12. In general, one can substitute the following property for (iii). Let us say that a family of graphs (or $d$-graphs) $\mathcal{F}$ is exactly closed under substitution $G=G\left(G^{\prime}, v, G^{\prime \prime}\right)$ whenever $G \in \mathcal{F}$ if and only if both $G^{\prime}$ and $G^{\prime \prime}$ belong to $\mathcal{F}$. For example, CIS-graphs are exactly closed under substitution, by Propositions 8 , and both, Gallai's and CIS- $d$-graphs, by Propositions 7 . Proposition 13. If $\mathcal{F}$ is closed under substitution and taking induced subgraphs then $\mathcal{F}$ is exactly closed under substitution. Proof. Indeed, if $G=G\left(G^{\prime}, v, G^{\prime \prime}\right)$ then both $G^{\prime}$ and $G^{\prime \prime}$ are induced subgraphs of $\mathcal{G}$. We say that the family of graphs $\mathcal{F}$ has the $C E S$-property and call it a $C E S$-family if $\mathcal{F}$ is closed under complementation and exactly closed under substitution. For example, the families of perfect graphs and CIS-graphs have the CES-property. We strengthen Cameron-Edmonds' theorem as follows. Theorem 4. Let $\mathcal{F}$ be a CES-family of graphs and $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ be a Gallai d-graph such that at least $d-1$ of its chromatic components, say, $G_{i}=\left(V, E_{i}\right)$ for $i=1, \ldots, d-1$, belong to $\mathcal{F}$. Then (a) the last component $G_{d}=\left(V, E_{d}\right)$ is in $\mathcal{F}$ too, and moreover, (b) all $2^{d}$ projections of $\mathcal{G}$ belong to $\mathcal{F}$, that is, for each subset $I \subseteq[d]=\{1, \ldots, d\}$ the graph $G_{I}=\left(V, \cup_{i \in I} E_{i}\right)$ is in $\mathcal{F}$. We will prove this Theorem in Section 4.1. By Proposition 13, part (a) implies CameronEdmonds' theorem. Since CIS-graphs form a CES-family, we obtain the following claim. Corollary 3. Let $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ be a Gallai d-graph such that at least $d-1$ of its chromatic components are CIS-graphs. Then the remaining chromatic component of $\mathcal{G}$ is a CIS-graph too; hence, $\mathcal{G}$ is a CIS-d-graph and all its $2^{d}$ projections are CIS-graphs. ### 1.8 On families of graphs closed with respect to substitution To get more examples of CES-families let us, first, consider hereditary classes. Each such class is a family of graphs $\mathcal{F}$ defined by a family, finite or infinite, of forbidden subgraphs $\mathcal{F}^{\prime}$. By definition, $G \in \mathcal{F}$ if and only if $G$ contains no induced subgraph isomorphic to a $G^{\prime} \in \mathcal{F}^{\prime}$. Let us call a graph (or $d$-graph) $G$ substitution-prime (or just, prime, for brevity) if it is not decomposable by substitution, or more precisely, if $G=G\left(G^{\prime}, v, G^{\prime \prime}\right)$ for no $G^{\prime}, G^{\prime \prime}$ and $v$, except for two trivial cases: $\left(G=G^{\prime}\right.$ and $\left.V\left(G^{\prime \prime}\right)=\{v\}\right)$ or $\left(G=G^{\prime \prime}\right.$ and $\left.V\left(G^{\prime}\right)=\{v\}\right)$. Suppose that $G$ is decomposable, $G=G\left(G^{\prime}, v, G^{\prime \prime}\right)$. Then, as we already mentioned, both $G^{\prime}$ and $G^{\prime \prime}$ are induced subgraphs of $G$. Hence, if $G^{\prime}$ or $G^{\prime \prime}$ contains an induced subgraph $G_{0}$ then $G$ also contains it. However, $G$ may contain $G_{0}$ even if $G^{\prime}$ and $G^{\prime \prime}$ do not. Yet, clearly, in this case $G_{0}$ is not substitution-prime. Thus, for both, graphs and $d$-graphs, we obtain the following statement. Proposition 14. Family $\mathcal{F}$ is exactly closed under substitution if all (d-)graphs in $\mathcal{F}^{\prime}$ are substitution-prime. The inverse holds too if we assume (by the way, without any loss of generality) that no $\left(d\right.$-)graph of $\mathcal{F}^{\prime}$ contains another one as an induced subgraph. Thus, $\mathcal{F}$ is a CES-family (and, hence, it satisfies all conditions of Theorem 4) whenever $\mathcal{F}^{\prime}$ is closed under complementation $\left(G \in \mathcal{F}^{\prime}\right.$ if and only if $\left.\bar{G} \in \mathcal{F}^{\prime}\right)$ and $\mathcal{F}^{\prime}$ contains only substitution-prime graphs. For example, these two properties hold for the odd holes and anti-holes. In this case, $\mathcal{F}$ is the family of Berge graphs. Thus, Theorem 4 and the Strong Perfect Graph Theorem imply the Cameron-Edmonds-Lovász Theorem [9]. Of course, it can be proved simpler: first, show that perfect graphs are exactly closed under substitution, [30], and then apply Lovász' perfect graph theorem, $[29,30]$, instead of the strong one. Another example is provided by the family $\mathcal{F}$ of $P_{4}$-free graphs. In this case $\mathcal{F}^{\prime}=\left\{P_{4}\right\}$ and all conditions of Theorem 4 hold, since $P_{4}$ is self-complementary and prime. Remark 8. Moreover, in this case, it is easy to verify directly claims (a) and (b) of Theorem 4, see [19]. The following implication is instrumental: if a d-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ is $\Pi$ - and $\Delta$-free then every its 2 -projection $\mathcal{G}^{\prime}=\left(V ; E_{1}^{\prime}, E_{2}^{\prime}\right)$ is $\Pi$-free too. A similar example is given by the family $\mathcal{F}$ of $A$-free graphs. In this case $\mathcal{F}^{\prime}=\{A\}$, where $A$ is the settled $P_{4}$ (or in other words, settled 2-comb, or bull-graph). Like $P_{4}$, it is also self-complementary and substitution-prime. However, if $\mathcal{F}^{\prime}$ contains a decomposable graph, e.g., $C_{4}$, then $\mathcal{F}$ may be not closed under substitution. For example, let $\mathcal{F}^{\prime}=\left\{C_{4}, \overline{C_{4}}\right\}$ and consider the Gallai 3-graph in Figure 13. Two of its chromatic components belong to $\mathcal{F}$, while the third one, $C_{4}$, does not. As another example, let us consider $\mathcal{F}^{\prime}=\left\{C_{4}, \overline{C_{4}}, C_{5}\right\}$. By [15], $\mathcal{F}$ is the family of split graphs. This family is self-comlementary, yet, it is not closed under substitution. Indeed, substituting, for example, a non-edge for the middle vertex of $P_{3}$ we get $C_{4}$. There are also non-hereditary families of graphs (respectively, $d$-graphs) closed under substitution; for example, CIS-graphs (respectively, CIS-d-graphs). It is not difficult to give more examples of such families and even to characterize them. Given a family $\mathcal{F}^{\prime}$, finite or infinite, of $\left(d\right.$-)graphs, let us denote by $\operatorname{cl}\left(\mathcal{F}^{\prime}\right)$ its closure with respect to substitution. Proposition 15. A family $\mathcal{F}$ of (d-)graphs is exactly closed under substitution if and only if $\mathcal{F}=\operatorname{cl}\left(\mathcal{F}^{\prime}\right)$, where $\mathcal{F}^{\prime}$ is a family, finite or infinite, of substitution-prime (d-)graphs. Furthermore, $\mathcal{F}$ is closed under complementation whenever $\mathcal{F}^{\prime}$ is. Proof. The second claim makes sense only for graphs and it is obvious. The first one follows from uniqueness of the canonical modular decomposition [34]. The obtained family $\mathcal{F}=\operatorname{cl}\left(\mathcal{F}^{\prime}\right)$ is not hereditary if and only if there are substitutionprime (d)-graphs $G \in \mathcal{F}^{\prime}$ and $G^{\prime} \notin \mathcal{F}^{\prime}$ such that $G^{\prime}$ is an induced subgraph of $G$. For example, let $\mathcal{F}^{\prime}=\{A\}$ contains only the bull-graph $A$ then $\mathcal{F}=\operatorname{cl}\left(\mathcal{F}^{\prime}\right)$ contains no 2-comb. However, the characterization of the CES-families by Proposition 15 is not constructive. For example, the substitution-prime perfect or CIS-graphs form infinite families that are difficult to describe explicitly. ### 1.9 Almost CIS- $d$-graphs A $d$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ will be called an almost CIS-d-graph if $\bigcap_{i=1}^{d} C_{i}=\emptyset$ for exactly one $d$-tuple $C_{1}, \ldots, C_{d}$, where $C_{i} \subseteq V$ is an inclusion maximal vertex-set containing no edges of color $i$, that is, for each $i \in[d]=\{1, \ldots, d\}$, we have $\left(v, v^{\prime}\right) \in E_{i}$ for no $v, v^{\prime} \in C_{i}$, Notice that, by definition, the families of CIS- and almost CIS- $d$-graphs are disjoint. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-200.jpg?height=331&width=374&top_left_y=294&top_left_x=864) Figure 18: Almost CIS-3-graph $\Delta^{\prime}$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-200.jpg?height=322&width=374&top_left_y=707&top_left_x=864) Figure 19: Almost CIS-3-graph $\Delta^{\prime}$ (in black and white for printing). For $d=2$ we return to the definition of almost CIS-graphs. More precisely, an almost CIS-2-graph is a pair of two complementary almost CIS-graphs. By Proposition 2, any split graph is either a CIS- or almost CIS-graph. Moreover, we conjecture that, except split graphs, there are no other almost CIS-graphs. The latter are in a natural one-to-one correspondence with the split almost CIS-2-graphs. Let us recall that we may have $d \geq 2$ for a 2-graph. In particular, for an arbitrary $d \geq 2$ and almost CIS graph $G=(V, E)$ let us define a $d$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ by setting $E_{1}=E, E_{2}=\bar{E}$, and $E_{i}=\emptyset$ for each $i>2$. It is easy to see that $\mathcal{G}$ is an almost CIS-d-graph. Let us note that already the 3-graph $\Delta$ is not almost CIS, since it has two distinct triplets $C_{1}=\left\{v_{2}, v_{3}\right\}, C_{2}=\left\{v_{3}, v_{1}\right\}, C_{3}=\left\{v_{1}, v_{2}\right\}$ and $C_{1}^{\prime}=\left\{v_{3}, v_{1}\right\}, C_{2}^{\prime}=\left\{v_{1}, v_{2}\right\}, C_{3}^{\prime}=\left\{v_{2}, v_{3}\right\}$ such that $C_{1} \cap C_{2} \cap C_{3}=\emptyset$ and $C_{1}^{\prime} \cap C_{2}^{\prime} \cap C_{3}^{\prime}=\emptyset$; see Section 4.2 for more details. However, it is not difficult to extend $\Delta$ to an almost CIS-3-graph. Indeed, let us add to $\Delta$ a new vertex $v_{4}$ such that $\left(v_{1}, v_{4}\right) \in E_{1},\left(v_{2}, v_{4}\right) \in E_{2},\left(v_{3}, v_{4}\right) \in E_{3}$, and denote the obtained 3 -graph by $\Delta^{\prime}$. In other words, $\Delta^{\prime}=\left(V ; E_{1}, E_{2}, E_{3}\right)$, where $$ V=\left\{v_{1}, v_{2}, v_{3}, v_{4}\right\} ; E_{1}=\left\{\left(v_{1}, v_{2}\right),\left(v_{1}, v_{4}\right)\right\}, E_{2}=\left\{\left(v_{2}, v_{3}\right),\left(v_{2}, v_{4}\right)\right\}, E_{3}=\left\{\left(v_{3}, v_{1}\right),\left(v_{3}, v_{4}\right)\right\} $$ It is easy to see that in $\Delta^{\prime}$ vertices $v_{1}, v_{2}, v_{3}$ induce $\Delta$ and that $$ \mathcal{C}_{1}=\left\{\left(v_{3}, v_{1}\right),\left(v_{2}, v_{3}, v_{4}\right)\right\}, \mathcal{C}_{2}=\left\{\left(v_{1}, v_{2}\right),\left(v_{3}, v_{1}, v_{4}\right)\right\}, \mathcal{C}_{3}=\left\{\left(v_{2}, v_{3}\right),\left(v_{1}, v_{2}, v_{4}\right)\right\} $$ Thus, $\Delta^{\prime}$ is an almost CIS-3-graph, since only one of its eight triplets has the empty intersection: $\left\{v_{3}, v_{1}\right\} \cap\left\{v_{1}, v_{2}\right\} \cap\left\{v_{2}, v_{3}\right\}=\emptyset$. Remark 9. We can say that vertex $v_{4}$ settles one of the above two triplets of $\Delta$, namely, $\left(C_{1}, C_{2}, C_{3}\right)$. However, if we introduce one more vertex $v_{5}$ to settle $\left(C_{1}^{\prime}, C_{2}^{\prime}, C_{3}^{\prime}\right)$, too, then we have to choose a color for $\left(v_{4}, v_{5}\right)$. It is easy to verify that for each coloring of it a new $\Delta$ appears that should be, in its turn, settled, etc.; see Section 4.2 for more details. Furthermore, it is not difficult to verify that $\Delta^{\prime}$ is a unique almost CIS-3-graph with four vertices. Let us recall that there are also two CIS-3-graphs given in Figures 13 and 14. Standardly, for any $d \geq 3$ we obtain an almost CIS-3-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}\right)$ setting $E_{1}, E_{2}, E_{3}$, and $V$ by (1.2) and $E_{i}=\emptyset$ for each $i>3$. Let us also remark that, unlike CIS-d-graphs, almost CIS-d-graphs (and, in particular, almost CIS-graphs) are not closed under substitution. Nevertheless, we get an almost CIS3-graph substituting $\Delta^{\prime}$ for a vertex of a 1-graph $\mathcal{G}$. More precisely, $\mathcal{G}$ is a monochromatic clique whose all edges are colored by one of the three colors of $\Delta^{\prime}$. However, if all edges of $\mathcal{G}$ are colored by a new, fourth, color then the obtained 4-graph is not almost CIS. Similarly, we won't get an almost CIS-3- or 4-graph by substituting $\Delta^{\prime}$ for more than one vertex of $\mathcal{G}$, nor, vice-versa, by substituting $\mathcal{G}$ for a vertex of $\Delta^{\prime}$. Finally, let us mention that we are not aware of any other almost CIS-d-graphs. ## 2 Proof of Theorem 1 In this section we prove Theorem 1 which claims that graphs satisfying condition $\operatorname{COMB}(3,3)$ are CIS-graphs. First we describe the structure of our proof and a few main lemmas, then we give the complete proofs which are technical, long, and partially computer assisted. ### 2.1 Plan of the proof of Theorem 1 Let us assume by contradiction that there is a graph $G$ such that (i) it contains no induced 3 -combs and 3 -anti-combs, (ii) each induced 2-comb is settled in $G$, and (iii) there exist a maximal clique $C$ and a maximal stable set $S$ in $G$ such that $S \cap C=\emptyset$. First, we will prove that $G$ must contain an induced subgraph $G_{10}$, shown in Figure 20. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-201.jpg?height=477&width=471&top_left_y=1829&top_left_x=815) Figure 20: Graph $G_{10}$. Lemma 2. If $G$ satisfies conditions (i), (ii), and (iii), then $G$ must contain an induced $G_{10}$. Graph $G_{10}$ contains no induced 3-combs and 3-anti-combs, yet it contains several unsettled induced 2-combs. To settle them we have to introduce 10 new vertices that, somewhat surprisingly, induce a graph isomorphic to $G_{10}$ itself (since otherwise an induced 3-comb or 3-anti-comb would appear). Moreover, the obtained 20-vertex graph is the sum of two Petersen graphs, that is, the graph $2 \mathcal{P}$ described in section 1.4, Figure 8. Lemma 3. If $G$ contains an induced $G_{10}$ and satisfies conditions (i) and (ii), then $G$ must contain an induced $2 \mathcal{P}$. Let us recall that $2 \mathcal{P}$ contains 10 uncertain pairs of vertices each of which can be either an edge or non-edge. Hence in fact, $2 \mathcal{P}$ represent $2^{10}=1024$ graphs. We will show that all these 1024 graphs contain no induced 3-combs and 3-anti-combs and, moreover, each induced 2-comb in $2 \mathcal{P}$ (that contains no uncertain pair) is settled. However, 36 induced 2 -combs appear in $2 \mathcal{P}$ whenever we fix any uncertain pair either as an edge or as a non-edge. It is easy to see that none of these 2 -combs are settled in $2 \mathcal{P}$. We will show that they cannot be settled in $G$ either, because if a vertex of $G$ were settling one of them then an induced 3 -comb or 3-anti-comb would exist in $G$. We can reformulate this result as follows. Lemma 4. If $G$ satisfies conditions (i) and (ii), then it cannot contain an induced $2 \mathcal{P}$. Obviously, the above 3 lemmas prove Theorem 1 by contradiction. We will prove Lemmas 2,3 , and 4 below in Sections 2.2,2.3, and 2.4, respectively. The last two proofs are computer assisted. We use two procedures, one for generating all induced 2-combs, 3-combs, and 3-anti-combs of a given graph $G$, and a second one for testing if all induced 2-combs are settled in $G$, and outputting all non-settled ones. ### 2.2 Proof of Lemma 2 Let us consider a pair of disjoint maximal clique $C$ and maximal stable set $S$ of $G$, as in condition (iii). Let $N_{S}(v)$ be the set of neighbors of $v$ in $S$. Notice that $$ \bigcap_{v \in C} N_{S}(v)=\emptyset $$ because $C$ is maximal. Moreover, $$ N_{S}(v) \neq \emptyset \quad \text { for all } \quad v \in C $$ because $S$ is maximal. We assume that $G$ satisfies conditions (i), (ii), and (iii). The following series of claims will imply the lemma. Claim 4.1. Given a maximal clique $C$ and a (not necessarily maximal) stable set $S$ in $G$ such that $C \cap S=\emptyset$, there exists vertices $u, v \in C$ such that $N_{S}(u) \cap N_{S}(v)=\emptyset$. Proof. Assume by contradiction that for all pairs of vertices $u, v \in C$, we have $N_{S}(u) \cap$ $N_{S}(v) \neq \emptyset$. By this assumption, $|C| \geq 3$, otherwise $C$ would not be maximal. So let $I=\left\{v_{1}, v_{2}, \ldots, v_{k}\right\}$ be a minimal subset of $C$ such that $\bigcap_{v \in I} N_{S}(v)=\emptyset$. Such a minimal subset of $C$ exists according to (2.3). Furthermore, by our assumption $|I| \geq 3$. Now, define $u_{i} \in \bigcap_{j \neq i}^{k} N_{S}\left(v_{j}\right)$ for $i=1, \ldots, k$. Note that $u_{i} \neq u_{j}$, due to the minimality of $I$. Thus, any 3 vertices $v_{1}, v_{2}, v_{3} \in I$ with the corresponding $u_{1}, u_{2}, u_{3}$ form an $\overline{S_{3}}$ (see Figure 21), contradicting condition (i). Note that for this claim we only need that $G$ is $S_{3}$-free. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-203.jpg?height=511&width=468&top_left_y=790&top_left_x=794) Figure 21: Illustration of the proof of Claim 4.1. From Claim 4.1, it follows that there are some pairs of vertices $u, v \in C$ such that $N_{S}(u) \cap N_{S}(v)=\emptyset$. Hence, there exist $x \in N_{S}(u)$ and $y \in\left(N_{S}(v)\right)$ such that $x, u, v, y$ form an $S_{2}$ not settled by any vertex of $S$. The following claim states a useful property of any vertex $w \in V(G)$ settling such an $S_{2}$. Claim 4.2. We have $N_{S}(w) \subseteq N_{S}(u) \cup N_{S}(v)$. Proof. First notice that $x, y \notin N_{S}(w)$ because $w$ is a settling vertex. Then, assume by contradiction that there is a vertex $z \in N_{S}(w) \backslash\left(N_{S}(u) \cup N_{S}(v)\right)$. Then, vertices $u, v, w, x, y, z$ form an $S_{3}$ (see Figure 22), contradicting condition (i). For the remainder of the proof we fix a maximal clique $C$, a maximal stable set $S$, and vertices $u, v \in C$ such that (iv) $C \cap S=\emptyset, N_{S}(u) \cap N_{S}(v)=\emptyset$, and $N_{S}(u) \cup N_{S}(v)$ is minimal, among all possible choices of such sets $C, S$ and vertices $u, v \in C$ satisfying the conditions of (iv). Let us note that by (2.3) and (2.4), we have such a selection of $C, S$, u, and $v$ for which $N_{S}(u) \neq \emptyset, N_{S}(v) \neq \emptyset$, and hence $u \neq v$. Claim 4.3. Let $x \in N_{S}(u), y \in N_{S}(v)$, and $w$ be a vertex of $V(G)$ that settles $S_{2}=$ $\{x, u, v, y\}$. Then, $N_{S}(w) \cap N_{S}(u) \neq \emptyset$ and $N_{S}(w) \cap N_{S}(v) \neq \emptyset$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-204.jpg?height=494&width=468&top_left_y=324&top_left_x=794) Figure 22: Illustration of the proof of Claim 4.2. Proof. From Claim 4.2, we know that $N_{S}(w) \subseteq N_{S}(u) \cup N_{S}(v)$. Assume by contradiction that e.g., $N_{S}(w) \cap N_{S}(u)=\emptyset$. This implies that $N_{S}(w) \subseteq N_{S}(v) \backslash\{y\}$ (since $w$ is settling $S_{2}$ ). Then, consider a maximal clique $C^{\prime} \supseteq\{u, w\}$. Notice that $C^{\prime} \cap S=\emptyset$ because $N_{S}(w) \cap$ $N_{S}(u)=\emptyset$. But $N_{S}(u) \cup N_{S}(w) \subsetneq N_{S}(u) \cup N_{S}(v)$, since $y \notin N_{S}(u) \cup N_{S}(w)$, contradicting property (iv), that is, the minimality of $N_{S}(u) \cup N_{S}(v)$. We define next a minimal collection of settling vertices $\mathcal{W}$. Given a maximal clique $C$, a maximal stable set $S$, and vertices $u, v \in C$ satisfying property (iv), let us consider all possible 2-combs induced by $\{x, u, v, y\}$ in $G$, where $x \in N_{S}(u)$ and $y \in N_{S}(v)$. Let us call a settling vertex a vertex $w$ of $G$ that settles such a 2-comb. If $w$ is a settling vertex, then we have by Claims 4.2 and 4.3 that $X(w)=N_{S}(w) \cap N_{S}(u)$ and $Y(w)=N_{S}(w) \cap N_{S}(v)$ are subsets, uniquely defined by $w$, satisfying the following properties: $$ X(w) \neq \emptyset \quad Y(w) \neq \emptyset \quad \text { and } \quad N_{S}(w)=X(w) \cup Y(w) $$ Note that we may have $X(w)=X\left(w^{\prime}\right)$ and $Y(w)=Y\left(w^{\prime}\right)$ for two distinct settling vertices. Note further that if $X(w) \subseteq X\left(w^{\prime}\right)$ and $Y(w) \subseteq Y\left(w^{\prime}\right)$ hold for two vertices $w$ and $w^{\prime}$, then the set of $S_{2}$ subgraphs settled by $w^{\prime}$ are also settled by $w$. Let us consider now all pairs of subsets $(X, Y)$ such that $X=X(w)$ and $Y=Y(w)$ for some settling vertex $w$. Let us call such a pair $(X, Y)$ minimal, if for there is no settling vertex $w^{\prime}$ such that $X\left(w^{\prime}\right) \subseteq X, Y\left(w^{\prime}\right) \subseteq Y$ and $X\left(w^{\prime}\right) \cup Y\left(w^{\prime}\right) \subseteq X \cup Y$, and let $\mathcal{X} \mathcal{Y}$ denote the collection of all such minimal pairs. For each pair $(X, Y) \in \mathcal{X} \mathcal{Y}$ let us choose one settling vertex $w=w_{X Y}$ for which $X=X(w)$ and $Y=Y(w)$, and denote by $\mathcal{W}=\left\{w_{X Y} \mid(X, Y) \in\right.$ $\mathcal{X Y}\}$ the collection of these vertices. Claim 4.4. There are at least two distinct vertices in $\mathcal{W}$. Proof. The statement follows from the definition of $\mathcal{W}$ and (2.5). Indeed, if $w_{X Y} \in \mathcal{W}$, then by (2.5) there are vertices $x \in X$ and $y \in Y$, and hence the 2-comb $S_{2}$ induced by ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-205.jpg?height=445&width=640&top_left_y=317&top_left_x=731) Figure 23 $\{x, u, v, y\}$ is not settled by $w_{X Y}$. Let $w$ be a vertex settling this 2 -comb. By the minimality of $(X, Y)$ the pair $(X(w), Y(w))$ is not comparable to $(X, Y)$, and hence we must have a pair $\left(X^{\prime}, Y^{\prime}\right) \in \mathcal{X} \mathcal{Y}$ such that $X^{\prime} \subseteq X$ and $Y^{\prime} \subseteq Y$. Consequently, $w_{X^{\prime} Y^{\prime}} \in \mathcal{W}$ and $w_{X Y} \neq w_{X^{\prime} Y^{\prime}}$. In the sequel we consider pairs of vertices from $\mathcal{W}$ and derive some containment relations for the corresponding sets. First we consider pairs which are edges of $G$. Claim 4.5. If $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \in E(G)$ and $X \cap X^{\prime} \neq \emptyset$, then $Y \subseteq Y^{\prime}$ or $Y^{\prime} \subseteq Y$. Proof. Assume by contradiction that there is a vertex $x \in X \cap X^{\prime}$, but $Y \nsubseteq Y^{\prime}$ and $Y^{\prime} \nsubseteq$ $Y$, that is, there are vertices $y_{1} \in Y \backslash Y^{\prime}$ and $y_{2} \in Y^{\prime} \backslash Y$. Then, an $\overline{S_{3}}$ is formed by $w_{X Y}, w_{X^{\prime} Y^{\prime}}, v, x, y_{1}, y_{2}$ (see Figure 24), in contradiction to (i). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-205.jpg?height=539&width=734&top_left_y=1695&top_left_x=684) Figure 24: Illustration of the proof of Claim 4.5. We next show a stronger version of the above claim, by proving proper containments. Claim 4.6. If $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \in E(G)$ and $X \cap X^{\prime} \neq \emptyset$, then either $Y \subsetneq Y^{\prime}$ or $Y^{\prime} \subsetneq Y$. Proof. Assume by contradiction that $X \cap X^{\prime} \neq \emptyset$ and $Y=Y^{\prime}$. By this assumption $Y \cap Y^{\prime} \neq \emptyset$. Hence, we can apply Claim 4.5 (with the roles of $X$ and $Y$ exchanged), and conclude that $X \subseteq X^{\prime}$ or $X^{\prime} \subseteq X$. Say e.g., that $X \subseteq X^{\prime}$. Then, $X \cup Y \subseteq X^{\prime} \cup Y^{\prime}$, and consequently we would not have both $w_{X, Y}$ and $w_{X^{\prime}, Y^{\prime}}$ in $\mathcal{W}$, by its definitions. Claim 4.7. If $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \in E(G)$, then exactly one of the following holds: (a) $X \cap X^{\prime}=Y \cap Y^{\prime}=\emptyset$, (b) $\left(X \subsetneq X^{\prime}\right.$ and $\left.Y^{\prime} \subsetneq Y\right)$, (c) $\left(X^{\prime} \subsetneq X\right.$ and $\left.Y \subsetneq Y^{\prime}\right)$. Proof. This follows from Claim 4.6 by applying it twice: once directly and once exchanging the roles of $X$ and $Y$. Since $X, Y, X^{\prime}$ and $Y^{\prime}$ are nonempty sets by (2.5), cases (a), (b) and (c) are pairwise exclusive. Next we consider pairs of settling vertices that are not edges of $G$. Claim 4.8. If $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \notin E(G)$, then either $X \subseteq X^{\prime}$ or $Y \subseteq Y^{\prime}$. Proof. If not, then there are vertices $x \in X \backslash X^{\prime}$ and $y \in Y \backslash Y^{\prime}$ such that $\left\{w_{X Y}, u, v, x, y, w_{X^{\prime} Y^{\prime}}\right\}$ form a 3-anti-comb $\overline{S_{3}}$ (see Figure 25), in contradiction to condition (i). Note that we cannot have both containments in the claim, because of the minimality of pairs in $\mathcal{X Y}$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-206.jpg?height=528&width=731&top_left_y=1678&top_left_x=685) Figure 25: Illustration of the 3-anti-comb $\overline{S_{3}}$ induced by $\left\{w_{X Y}, u, v, x, y, w_{X^{\prime} Y^{\prime}}\right\}$. Claim 4.9. If $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \notin E(G)$, then exactly one of the following must hold: (a) $X \subsetneq X^{\prime}$ and $Y^{\prime} \subsetneq Y$, (b) $X^{\prime} \subsetneq X$ and $Y \subsetneq Y^{\prime}$, (c) $X=X^{\prime}$, (d) $Y=Y^{\prime}$. Proof. Since the roles of $(X, Y)$ and $\left(X^{\prime}, Y^{\prime}\right)$ are symmetric, it follows directly by Claim 4.8 that one of (a), (b), (c), or (d) holds. To see that exactly one of them holds, it is enough to note that (c) and (d) together would contradict the minimality of the pairs $(X, Y) \in \mathcal{X Y}$. We are going to show next that if (c) or (d) holds in the previous claim for some vertices $w_{X Y}, w_{X^{\prime} Y^{\prime}} \in \mathcal{W}$, then $G$ contains an induced $G_{1} 0$, as claimed in Lemma 2. For this end, let us first observe that if e.g., (d) holds, then we cannot have $X \subseteq X^{\prime}$ or $X^{\prime} \subseteq X$, by the minimality and uniqueness of pairs in $\mathcal{X} \mathcal{Y}$. Consequently, we can choose vertices $x \in X \backslash X^{\prime}$, and $x^{\prime} \in X^{\prime} \backslash X$. Let us also choose an arbitrary vertex $y \in Y=Y^{\prime}$ (which exists by (2.5)), and consider first the 2-comb $S_{2}$ induced by $\{x, u, v, y\}$. This 2 -comb is settled by neither $w_{X Y}$ nor $w_{X^{\prime} Y^{\prime}}$, and therefore there must be a vertex $w_{A B} \in \mathcal{W}$ settling it, since all 2-combs, containing $(u, v)$ as their middle edge, are settled by some vertices in $\mathcal{W}$. Claim 4.10. If $Y=Y^{\prime}$, then $\left(w_{A B}, w_{X Y}\right) \in E(G)$. Proof. Since $x \notin A$ and $y \notin B$ we have $$ X \nsubseteq A \quad \text { and } \quad Y \nsubseteq B $$ implied. Assume indirectly that $\left(w_{A B}, w_{X Y}\right) \notin E(G)$, then the previous observation implies that in Claim 4.9 applied to $w_{X Y}$ and $w_{A B}$ none of (a), (b), (c) or (d) could hold. This contradiction proves the claim. Claim 4.11. If $Y=Y^{\prime}$, then $A \cap X=B \cap Y=\emptyset, A \cup X=N_{S}(u)$ and $B \cup Y=N_{S}(v)$. Proof. Due to (2.6) only (a) of Claim 4.7 is possible, that is $A \cap X=B \cap Y=\emptyset$ is implied. Therefore the neighborhoods of $w_{A B}$ and $w_{X Y}$ within $S$ are disjoint, and since they are subsets of the neighborhoods of $u$ and $v$, they cannot be proper subsets by property (iv), implying the statement. Claim 4.12. If $Y=Y^{\prime}$, then $\left(w_{A B}, w_{X^{\prime} Y^{\prime}}\right) \notin E(G)$. Proof. Since $y \in Y^{\prime} \backslash B$ and $x \in X \backslash A$ (since $w_{A B}$ is settling $\{x, u, v, y\}$ ), cases (b) and (c) of Claim 4.7 cannot hold for the pair $w_{A B}$ and $w_{X^{\prime} Y^{\prime}}$. Thus, if $\left(w_{A B}, w_{X^{\prime} Y^{\prime}}\right) \in E(G)$ then $A \cap X^{\prime}=B \cap Y^{\prime}=\emptyset$ would follow by Claim 4.7. Therefore, the neighborhoods of $w_{A B}$ and $w_{X^{\prime} Y^{\prime}}$ in $S$ are disjoint, and their union is a proper subset of $N_{S}(u) \cup N_{S}(v)$, in contradiction with property (iv). This contradiction proves the claim. Claim 4.13. If $Y=Y^{\prime}$, then $A=X^{\prime}=N_{S}(u) \backslash X$ and $Y=Y^{\prime}=N_{S}(v) \backslash B$. Proof. Claim 4.11 and Claim 4.9 applied to $w_{A B}$ and $w_{X^{\prime} Y^{\prime}}$ implies that only (c) of Claim 4.9 can hold. Thus, the statement implied by Claim 4.11 and (c) of Claim 4.9. Let us still assume $Y=Y^{\prime}$ and consider next the 2-comb induced by $\left\{x^{\prime}, u, v, y\right\}$ (where $\left.x^{\prime} \in X^{\prime} \backslash X\right)$. None of the vertices $w_{X Y}, w_{X^{\prime} Y^{\prime}}$ and $w_{A B}$ settle this 2 -comb, hence, there is a vertex $w_{A^{\prime} B^{\prime}} \in \mathcal{W}$ that settles it. By exchanging the roles of $w_{X Y}$ and $w_{X^{\prime} Y^{\prime}}$ in Claims 4.10 - 4.13, we can conclude that $$ \left(w_{A^{\prime} B^{\prime}}, w_{X Y}\right) \notin E(G), \quad\left(w_{A^{\prime} B^{\prime}}, w_{X^{\prime} Y^{\prime}}\right) \in E(G), \quad A^{\prime}=X^{\prime} \quad \text { and } \quad B=B^{\prime} $$ Claim 4.14. If $Y=Y^{\prime}$ or $X=X^{\prime}$, then $G$ contains an induced $G_{10}$. Proof. Note that the roles of conditions (c) and (d) in Claim 4.9 are perfectly symmetric, thus we could arrive to the same conclusions from both assumptions. Starting with $Y=Y^{\prime}$ we arrived to the equalities of Claim 4.13 and (2.7). Choosing one vertex from each of the sets $X, Y, A$, and $B$, these four vertices together with $u, v, w_{X Y}, w_{X^{\prime} Y^{\prime}}, w_{A B}$, and $w_{A^{\prime} B^{\prime}}$ form an induced $G_{10}$ by the above claims and definitions (see Figure 26). ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-208.jpg?height=482&width=491&top_left_y=1067&top_left_x=811) Figure 26: Illustration of the induced $G_{10}$ that appears by adding the settling vertices $w_{X Y}, w_{X^{\prime} Y^{\prime}}, w_{A B}, w_{A^{\prime} B^{\prime}}$. For the rest of the proof, we assume that (a) or (b) of Claim 4.9 holds for every nonedge $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \notin E(G)$. We are going to derive a contradiction from this assumption, completing the proof of Lemma 2. First, we show that under the above assumption, case (a) of Claim 4.7 never holds. Claim 4.15. If $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \in E(G)$, then either $X \cap X^{\prime} \neq \emptyset$ or $Y \cap Y^{\prime} \neq \emptyset$. Proof. Assume by contradiction that (a) of Claim 4.9 holds, that is that $X \cap X^{\prime}=Y \cap Y^{\prime}=\emptyset$. Then, by the minimality of $N_{S}(u) \cup N_{S}(v)$ as stated in property (iv), and by Claim 4.2, we know that $N_{S}(u)=X \cup X^{\prime}$ and $N_{S}(v)=Y \cup Y^{\prime}$. Let us consider vertices $x \in X$ and $y \in Y^{\prime}$ such that the set $\{x, u, v, y\}$ forms a 2-comb. This 2 -comb is settled neither by $w_{X Y}$ nor by $w_{X^{\prime} Y^{\prime}}$. Since every 2 -comb with $(u, v)$ as a middle edge is settled by a vertex of $\mathcal{W}$, this 2 -comb is also settled by one, say by a vertex $w_{A B} \in \mathcal{W}$. Let us now check the connections of this vertex to to $w_{X Y}$ and $w_{X^{\prime} Y^{\prime}}$. We consider two cases: Case 1. If $\left(w_{A B}, w_{X Y}\right) \notin E(G)$, then by Claim 4.9 we must have $A \subset X$ and $Y \subset B$, because $x \notin A$, and because we assumed that only cases (a) or (b) are possible in Claim 4.9. If $\left(w_{A B}, w_{X^{\prime} Y^{\prime}}\right) \notin E(G)$, then by similar reasoning based on by Claim 4.9 and the fact that $y \notin B$ we can conclude that $X^{\prime} \subset A$ and $B \subset Y^{\prime}$. This however leads to a contradiction, since $A \subseteq X$ and $X \cap X^{\prime}=\emptyset$. Hence, we must have $\left(w_{A B}, w_{X^{\prime} Y^{\prime}}\right) \in E(G)$ in this case. Then by Claim 4.7 either $X^{\prime} \cap A=Y^{\prime} \cap B=\emptyset$ or $A, X^{\prime}$ and $B, Y^{\prime}$ are inversely nested. However, the latter is not possible, since $A \subset X$ and $X \cap X^{\prime}=\emptyset$. In this case the neighborhoods of $w_{A B}$ and $w_{X^{\prime} Y^{\prime}}$ are disjoint in $S$, and their union is a proper subset of $N_{S}(u) \cup N_{S}(v)$ (since $x \notin A$ ), in contradiction with property (iv). Case 2. If $\left(w_{A B}, w_{X Y}\right) \in E(G)$, then (b) of Claim 4.7 is not possible, since $x \in X \backslash A$. If (a) holds, that is if $X \cap A=Y \cap B=\emptyset$, then the neighborhoods of $w_{A B}$ and $w_{X Y}$ are disjoint in $S$, and their union is a proper subset of $N_{S}(u) \cup N_{S}(v)$ (since $y \in Y^{\prime} \backslash B$ ), contradicting to property (iv). Consequently, case (c) holds, that is $A \subset X$ and $Y \subset B$, and consequently we can proceed as in Case 1. In both cases we arrived to a contradiction, completing the proof of the claim. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-209.jpg?height=532&width=731&top_left_y=1276&top_left_x=685) Figure 27 The above claim implies that if $\left(w_{X Y}, w_{X^{\prime} Y^{\prime}}\right) \in E(G)$, then the sets $X, X^{\prime}$ and $Y, Y^{\prime}$ are inversely nested (cases (b) or (c) in Claim 4.7). Since we also assumed that only cases (a) or (b) are possible in Claim 4.9, we can conclude that for all pairs of settling vertices $w_{X Y}, w_{X^{\prime} Y^{\prime}} \in \mathcal{W}$ we have $$ \text { either } X \subset X^{\prime} \text { and } Y^{\prime} \subset X \quad \text { or } \quad X^{\prime} \subset X \text { and } Y \subset Y^{\prime} $$ Now we are ready to complete the proof of the lemma. Let us consider an arbitrary vertex $w_{X Y} \in \mathcal{W}$. Since $w_{X Y}$ is settling a 2-comb with $(u, v)$ as its middle edge, we must have $Y \neq N_{S}(v)$, and consequently we can choose a vertex $y \in N_{S}(v) \backslash Y$. Furthermore, we have $X \neq \emptyset$ by (2.5), thus we can also choose a vertex $x \in X$. Then, the 2-comb $S_{2}$ induced by $\{x, u, v, y\}$ is not settled by $w_{X Y}$, and therefore there is a vertex $w_{X^{\prime} Y^{\prime}} \in \mathcal{W}$ settling this 2 -comb. Then, by (2.8) we must have $X^{\prime} \subseteq X \backslash\{x\}$ and $Y \subset Y^{\prime}$, since $x \notin X^{\prime}$. Then, $X^{\prime} \neq \emptyset$ by (2.5), so we can choose a vertex $x^{\prime} \in X^{\prime} \subsetneq X$. The 2 -comb induced by $\left\{x^{\prime}, u, v, y\right\}$ is not settled by either $w_{X Y}$ or $w_{X^{\prime} Y^{\prime}}$, and therefore there is a vertex $w_{X^{\prime \prime} Y^{\prime \prime}} \in \mathcal{W}$ settling this 2-comb. Clearly, we can repeat the same arguments, and choose a vertex $x^{\prime \prime} \in X^{\prime \prime} \subsetneq X^{\prime} \subsetneq X$, etc., resulting in an infinite chain $X \supsetneq X^{\prime} \supsetneq X^{\prime \prime} \supsetneq \cdots$ of strictly nested nonempty subsets, contradicting the finiteness of $G$. This concludes the proof of the lemma. ### 2.3 Proof of Lemma 3 In this section we present the proof of Lemma 3, claiming that if $G$ contains $G_{10}$ as an induced subgraph and satisfies conditions (i) and (ii) of Section 2.1, then it must have an induced $2 \mathcal{P}$ configuration (see Figures 20 and 8 ). The proof is a case analysis that was assisted by a computer program. We assume by contradiction that there is a graph that has an induced $G_{10}$, has all 2-combs settled and does not contain 3-combs and 3-anti-combs. The graph $G_{10}$ itself contains neither 3-combs nor 3 -anti-combs, but it has several 2-combs that are not settled in it. For instance, such 2 -combs are induced by $\left\{v_{2}, v_{1}, v_{5}, v_{4}\right\},\left\{v_{6}, v_{7}, v_{3}, v_{4}\right\},\left\{v_{1}, v_{2}, v_{3}, v_{7}\right\}$, etc. Therefore, some other vertices of $G$ must settle these 2-combs. We show that in order to settle all 2-combs of $G_{10}$, the graph $G$ must contain a disjoint copy of $G_{10}$ such that the 20 vertices of these two $G_{10}$ subgraphs form an induced $2 \mathcal{P}$ configuration. Since we do not know $G$, we try to extend $G_{10}$, and we show that this can be done essentially in a unique way. We use a computer program to find all unsettled 2-combs of $G_{10}$. For each, one by one, we introduce a new vertex to settle it. After adding a settling vertex $v^{\prime} \notin V\left(G_{10}\right)$, we consider the pairs $\left(v^{\prime}, v_{j}\right)$ for all $v_{j} \in V\left(G_{10}\right)$. Some of these pairs are forced to be edges or non-edges, since $G$ contains no induced 3 -combs and 3 -anti-combs. Some other pairs, however, may remain uncertain, that is those pairs may be either edges or non-edges of $G$. Surprisingly, all but one of the pairs are forced. We can discover the forced edge assignments by excluding all other possible assignments. This can be accomplished by exhibiting an induced 3-comb or 3 -anti-comb. This task is also assisted by a computer program. Another property which simplifies our case analysis is the symmetry of $G_{10}$. In particular, we reduce significantly the number of cases in our proof by means of the following three automorphisms: $A_{1}:(3)(7)(1,5)(2,4)(6,8)(0,9)$ $A_{2}:(1)(5)(2,8)(3,7)(4,6)(0,9)$ $A_{3}:(7,5,3,1)(8,6,4,2)(0,9)$ They are given in the cycle notation, that is $\left(i_{1}, i_{2}, \ldots, i_{n}\right)$ means the cyclic mapping $i_{1} \mapsto i_{2}$, $i_{2} \mapsto i_{3}, \ldots, i_{n} \mapsto i_{1}$. Figure 28 shows the graphs after the application of these automorphisms. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-211.jpg?height=480&width=1464&top_left_y=534&top_left_x=320) Figure 28: Graphs $A_{1}\left(G_{10}\right), A_{2}\left(G_{10}\right)$, and $A_{3}\left(G_{10}\right)$. From now on we will choose some of the unsettled 2-combs to be settled, and try to fix as many edges and non-edges as possible. Even though the order that we pick the 2-combs may seem arbitrary, we follow an order that reduces the number of cases to be considered. Let us choose first the 2 -comb induced by $\left\{v_{2}, v_{3}, v_{7}, v_{8}\right\}$, and denote by $v_{1}^{\prime}$ the vertex that settles it. The pairs $\left(v_{1}^{\prime}, v_{3}\right)$ and $\left(v_{1}^{\prime}, v_{7}\right)$ are forced to be edges, while $\left(v_{1}^{\prime}, v_{2}\right)$ and $\left(v_{1}^{\prime}, v_{8}\right)$ are forced to be non-edges, by the definition of settling. There are six more pairs, connecting $v_{1}^{\prime}$ with $v_{0}, v_{1}, v_{4}, v_{5}, v_{6}$ and $v_{9}$, that remain uncertain. Let us note first that $\left(v_{1}^{\prime}, v_{5}\right)$ has to be a non-edge, since otherwise the vertices $\left\{v_{3}, v_{7}, v_{1}^{\prime}, v_{2}, v_{8}, v_{5}\right\}$ form a 3 -comb. Unlike $\left(v_{1}^{\prime}, v_{5}\right)$, the pairs $\left(v_{1}^{\prime}, v_{0}\right),\left(v_{1}^{\prime}, v_{4}\right),\left(v_{1}^{\prime}, v_{6}\right),\left(v_{1}^{\prime}, v_{9}\right)$ cannot be fixed if treated individually. But analyzing them together, we conclude that $\left(v_{1}^{\prime}, v_{4}\right)$ and $\left(v_{1}^{\prime}, v_{6}\right)$ are edges, while $\left(v_{1}^{\prime}, v_{0}\right)$ and $\left(v_{1}^{\prime}, v_{9}\right)$ are non-edges. Table 1 shows that in any other case there is an induced 3 -comb or 3 -anti-comb. Only one pair $\left(v_{1}^{\prime}, v_{1}\right)$ remains uncertain, since no induced $S_{3}$ nor $\overline{S_{3}}$ appears whether this pair is an edge or not. Table 2 shows the connections between $v_{1}^{\prime}$ and the vertices of $G_{10}$. Next, we use automorphisms to simplify case analysis for the three 2-combs induced by $\left\{v_{4}, v_{3}, v_{7}, v_{6}\right\},\left\{v_{6}, v_{5}, v_{1}, v_{8}\right\}$, and $\left\{v_{2}, v_{1}, v_{5}, v_{4}\right\}$ respectively, and not settled by $v_{1}^{\prime}$. Let us denote by $v_{5}^{\prime}$ the vertex that settles $\left\{v_{4}, v_{3}, v_{7}, v_{6}\right\}$. By applying the automorphism $A_{1}$ to $G_{10}$, the 2 -comb $\left\{v_{2}, v_{3}, v_{7}, v_{8}\right\}$ settled by $v_{1}^{\prime}$ becomes $\left\{v_{4}, v_{3}, v_{7}, v_{6}\right\}$. Consequently, $v_{5}^{\prime}$ should have the same connections as $v_{1}^{\prime}$ has after applying $A_{1}$. Table 3 shows the connections between $v_{5}^{\prime}$ and $G_{10}$. Analogously, let us denote by $v_{3}^{\prime}$ the vertex that settles $\left\{v_{2}, v_{1}, v_{5}, v_{4}\right\}$. By applying $A_{3}$ to $G_{10},\left\{v_{2}, v_{3}, v_{7}, v_{8}\right\}$ becomes $\left\{v_{2}, v_{1}, v_{5}, v_{4}\right\}$. Therefore, $v_{3}^{\prime}$ should have the same connections as $v_{1}^{\prime}$ after transformation $A_{3}$. Table 4 shows the connections between $v_{3}^{\prime}$ and $G_{10}$. Next, let us denote by $v_{7}^{\prime}$ the vertex that settles $\left\{v_{8}, v_{1}, v_{5}, v_{6}\right\}$. By applying $A_{3}$ then $A_{2}$ to $G_{10},\left\{v_{2}, v_{3}, v_{7}, v_{8}\right\}$ becomes $\left\{v_{8}, v_{1}, v_{5}, v_{6}\right\}$. Thus, $v_{3}^{\prime}$ should have the same connections as | $\left(v_{1}^{\prime}, v_{4}\right)$ | $\left(v_{1}^{\prime}, v_{6}\right)$ | $\left(v_{1}^{\prime}, v_{0}\right)$ | $\left(v_{1}^{\prime}, v_{9}\right)$ | $S_{3}$ or $\overline{S_{3}}$ | | :---: | :---: | :---: | :---: | :---: | | 0 | 0 | 0 | 0 | $S_{3}:\left\{v_{3}, v_{0}, v_{9}, v_{1}^{\prime}, v_{6}, v_{8}\right\}$ | | 0 | 0 | 0 | 1 | $\overline{S_{3}}:\left\{v_{2}, v_{5}, v_{1}^{\prime}, v_{3}, v_{0}, v_{9}\right\}$ | | 0 | 0 | 1 | 0 | $\overline{S_{3}}:\left\{v_{4}, v_{8}, v_{1}^{\prime}, v_{3}, v_{7}, v_{9}\right\}$ | | 0 | 0 | 1 | 1 | $\overline{S_{3}}:\left\{v_{4}, v_{6}, v_{1}^{\prime}, v_{5}, v_{0}, v_{9}\right\}$ | | 0 | 1 | 0 | 0 | $S_{3}:\left\{v_{5}, v_{6}, v_{0}, v_{2}, v_{4}, v_{1}^{\prime}\right\}$ | | 0 | 1 | 0 | 1 | $S_{3}:\left\{v_{3}, v_{9}, v_{1}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | | 0 | 1 | 1 | 0 | $\overline{S_{3}}:\left\{v_{4}, v_{8}, v_{1}^{\prime}, v_{3}, v_{7}, v_{9}\right\}$ | | 0 | 1 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{9}, v_{1}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | | 1 | 0 | 0 | 0 | $S_{3}:\left\{v_{3}, v_{0}, v_{9}, v_{6}, v_{8}, v_{1}^{\prime}\right\}$ | | 1 | 0 | 0 | 1 | $\overline{S_{3}}:\left\{v_{2}, v_{5}, v_{1}^{\prime}, v_{3}, v_{0}, v_{9}\right\}$ | | 1 | 0 | 1 | 0 | $S_{3}:\left\{v_{4}, v_{5}, v_{9}, v_{6}, v_{8}, v_{1}^{\prime}\right\}$ | | 1 | 0 | 1 | 1 | $S_{3}:\left\{v_{7}, v_{0}, v_{1}^{\prime}, v_{2}, v_{4}, v_{8}\right\}$ | | 1 | 1 | 0 | 0 | none | | 1 | 1 | 0 | 1 | $S_{3}:\left\{v_{3}, v_{9}, v_{1}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | | 1 | 1 | 1 | 0 | $S_{3}:\left\{v_{7}, v_{0}, v_{1}^{\prime}, v_{2}, v_{4}, v_{8}\right\}$ | | 1 | 1 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{9}, v_{1}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | Table 1: Case analysis for the pairs $\left(v_{1}^{\prime}, v_{0}\right),\left(v_{1}^{\prime}, v_{4}\right),\left(v_{1}^{\prime}, v_{6}\right),\left(v_{1}^{\prime}, v_{9}\right)$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{1}^{\prime}$ | $*$ | 0 | 1 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | Table 2: Connections between $v_{1}^{\prime}$ and $G_{10}$. An entry 1 for $v_{i}$ means that there is an edge between $v_{1}^{\prime}$ and $v_{i}$, while 0 means that there is no edge between them. Finally, $*$ means an uncertain pair. $v_{1}^{\prime}$ after transformations $A_{3}$ then $A_{2}$ (or the same connections as $v_{3}^{\prime}$ after $A_{2}$ ). Table 5 shows the connections between $v_{7}^{\prime}$ and $G_{10}$. Let us next consider four 2 -combs induced by $\left\{v_{5}, v_{1}, v_{2}, v_{3}\right\},\left\{v_{1}, v_{5}, v_{4}, v_{3}\right\},\left\{v_{7}, v_{3}, v_{4}, v_{5}\right\}$, and $\left\{v_{1}, v_{2}, v_{3}, v_{7}\right\}$. They are not settled by any of the vertices of $G_{10}$, nor by $v_{1}^{\prime}, v_{3}^{\prime}, v_{5}^{\prime}, v_{7}^{\prime}$. Let $v_{2}^{\prime}$ denote the vertex settling $\left\{v_{3}, v_{4}, v_{5}, v_{1}\right\}$. By definition of settling, the pairs $\left(v_{2}^{\prime}, v_{4}\right)$ and $\left(v_{2}^{\prime}, v_{5}\right)$ are edges, while $\left(v_{2}^{\prime}, v_{1}\right)$ and $\left(v_{2}^{\prime}, v_{3}\right)$ are non-edges. The pair $\left(v_{2}^{\prime}, v_{9}\right)$ must be an edge, since otherwise $\left\{v_{1}, v_{3}, v_{2}^{\prime}, v_{4}, v_{5}, v_{9}\right\}$ forms a 3 -anti-comb. Table 6 shows the case analysis for the pairs $\left(v_{2}^{\prime}, v_{6}\right),\left(v_{2}^{\prime}, v_{7}\right),\left(v_{2}^{\prime}, v_{8}\right)$, and $\left(v_{2}^{\prime}, v_{0}\right)$. The only possible configuration is that $\left(v_{2}^{\prime}, v_{6}\right),\left(v_{2}^{\prime}, v_{7}\right),\left(v_{2}^{\prime}, v_{8}\right)$ are edges, and $\left(v_{2}^{\prime}, v_{0}\right)$ is not. The pair $\left(v_{2}^{\prime}, v_{2}\right)$ remains uncertain. Table 7 shows the connections between $v_{2}^{\prime}$ and the vertices of $G_{10}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{5}^{\prime}$ | 0 | 1 | 1 | 0 | $*$ | 0 | 1 | 1 | 0 | 0 | Table 3: Connections between $v_{5}^{\prime}$ and $G_{10}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{3}^{\prime}$ | 1 | 0 | $*$ | 0 | 1 | 1 | 0 | 1 | 0 | 0 | Table 4: Connections between $v_{3}^{\prime}$ and $G_{10}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{7}^{\prime}$ | 1 | 1 | 0 | 1 | 1 | 0 | $*$ | 0 | 0 | 0 | Table 5: Connections between $v_{7}^{\prime}$ and $G_{10}$. Let $v_{4}^{\prime}$ denote the vertex settling $\left\{v_{5}, v_{1}, v_{2}, v_{3}\right\}$. By applying $A_{1}$ to $G_{10}$, the subgraph $\left\{v_{1}, v_{5}, v_{4}, v_{3}\right\}$ becomes $\left\{v_{5}, v_{1}, v_{2}, v_{3}\right\}$. Therefore, vertex $v_{4}^{\prime}$ must have the same connections as $v_{2}^{\prime}$ after transformation $A_{1}$. Table 8 shows the connections between $v_{4}^{\prime}$ and $G_{10}$. Next, let $v_{6}^{\prime}$ denote the vertex settling $\left\{v_{7}, v_{3}, v_{4}, v_{5}\right\}$. By applying transformations, first $A_{1}$ and then $A_{3}$, to $G_{10}$, the subgraph $\left\{v_{1}, v_{5}, v_{4}, v_{3}\right\}$ becomes $\left\{v_{7}, v_{3}, v_{4}, v_{5}\right\}$. Thus, $v_{6}^{\prime}$ must have the same connections as $v_{2}^{\prime}$ after the transformation $A_{3} \circ A_{1}$. Table 9 shows the connections between $v_{6}^{\prime}$ and $G_{10}$. Let us next denote by $v_{8}^{\prime}$ the vertex that settles $\left\{v_{1}, v_{2}, v_{3}, v_{7}\right\}$. By applying $A_{3}^{-1}$, to $G_{10}$, the subgraph $\left\{v_{1}, v_{5}, v_{4}, v_{3}\right\}$ becomes $\left\{v_{1}, v_{2}, v_{3}, v_{7}\right\}$. Therefore, $v_{8}^{\prime}$ should have the same connections as $v_{2}^{\prime}$ after $A_{3}^{-1}$. Table 10 shows the connections between $v_{8}^{\prime}$ and $G_{10}$. At this point, all $S_{2}$ subgraphs of $G_{10}$ are settled by some of the vertices $v_{1}^{\prime}, v_{2}^{\prime}, \ldots$, $v_{8}^{\prime}$. Yet, nothing was said about the connections between those vertices. Nevertheless, all 3 -combs and 3-anti-combs that appeared to indicate contradictions were independent from those connections; in other words, each of those subgraphs contains only one vertex $v_{i}^{\prime}$ and the remaining five vertices are in $G_{10}$. Interestingly, the connections between these eight vertices are uniquely implied. Table 11 shows the only possible assignments of edges and non-edges between the vertices $v_{i}^{\prime}$ and $v_{j}^{\prime}$, for $i, j=1, \ldots, 8, i \neq j$. Each entry of the table contains the assignment, and the corresponding 3 -comb or 3 -anti-comb that would appear if the entry was reversed. Let us notice that the pairs $\left(v_{i}, v_{i}^{\prime}\right)$ still remain uncertain. This means that all $2^{8}$ possible graphs have no induced 3-combs and 3-anti-combs. Yet, they contain some unsettled induced 2 -combs. Next, we introduce the automorphism $A_{4}$ of the current configuration, induced by the 18 vertices $V\left(G_{10}\right) \cup\left\{v_{1}^{\prime}, \ldots, v_{8}^{\prime}\right\}$. $$ A_{4}:(1,3,5,7)(2,4,6,8)(0,9)\left(1^{\prime}, 3^{\prime}, 5^{\prime}, 7^{\prime}\right)\left(2^{\prime}, 4^{\prime}, 6^{\prime}, 8^{\prime}\right) $$ Let us further consider the unsettled 2-comb induced by $\left\{v_{2}, v_{1}^{\prime}, v_{5}^{\prime}, v_{6}\right\}$, and denote by $v_{0}^{\prime}$ the vertex that settles it. By definition, $\left(v_{0}^{\prime}, v_{1}^{\prime}\right)$ and $\left(v_{0}^{\prime}, v_{5}^{\prime}\right)$ are edges, while $\left(v_{0}^{\prime}, v_{2}\right)$ and $\left(v_{0}^{\prime}, v_{6}\right)$ are non-edges. The pair $\left(v_{0}^{\prime}, v_{9}\right)$ cannot be an edge, since otherwise $\left\{v_{1}^{\prime}, v_{5}^{\prime}, v_{0}^{\prime}, v_{2}, v_{6}, v_{9}\right\}$ forms a 3 -comb. Table 16 shows that $\left(v_{0}^{\prime}, v_{4}\right)$ and $\left(v_{0}^{\prime}, v_{8}\right)$ must be edges, while $\left(v_{0}^{\prime}, v_{1}\right)$, | $\left(v_{2}^{\prime}, v_{6}\right)$ | $\left(v_{2}^{\prime}, v_{7}\right)$ | $\left(v_{2}^{\prime}, v_{8}\right)$ | $\left(v_{2}^{\prime}, v_{0}\right)$ | $S_{3}$ or $\overline{S_{3}}$ | | :---: | :---: | :---: | :---: | :---: | | 0 | 0 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}, v_{3}, v_{8}, v_{2}^{\prime}\right\}$ | | 0 | 0 | 0 | 1 | $\overline{S_{3}}:\left\{v_{6}, v_{8}, v_{2}^{\prime}, v_{7}, v_{0}, v_{9}\right\}$ | | 0 | 0 | 1 | 0 | $S_{3}:\left\{v_{4}, v_{5}, v_{2}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 0 | 0 | 1 | 1 | $S_{3}:\left\{v_{4}, v_{5}, v_{2}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 0 | 1 | 0 | 0 | $\overline{S_{3}}:\left\{v_{1}, v_{5}, v_{0}, v_{3}, v_{8}, v_{2}^{\prime}\right\}$ | | 0 | 1 | 0 | 1 | $\overline{S_{3}}:\left\{v_{1}, v_{4}, v_{7}, v_{5}, v_{0}, v_{2}^{\prime}\right\}$ | | 0 | 1 | 1 | 0 | $S_{3}:\left\{v_{4}, v_{5}, v_{2}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 0 | 1 | 1 | 1 | $S_{3}:\left\{v_{4}, v_{5}, v_{2}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 1 | 0 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}, v_{3}, v_{8}, v_{2}^{\prime}\right\}$ | | 1 | 0 | 0 | 1 | $\overline{S_{3}}:\left\{v_{1}, v_{4}, v_{6}, v_{0}, v_{9}, v_{2}^{\prime}\right\}$ | | 1 | 0 | 1 | 0 | $\overline{S_{3}}:\left\{v_{1}, v_{7}, v_{2}^{\prime}, v_{5}, v_{6}, v_{0}\right\}$ | | 1 | 0 | 1 | 1 | $\overline{S_{3}}:\left\{v_{1}, v_{4}, v_{6}, v_{0}, v_{9}, v_{2}^{\prime}\right\}$ | | 1 | 1 | 0 | 0 | $\overline{S_{3}}:\left\{v_{1}, v_{5}, v_{0}, v_{3}, v_{8}, v_{2}^{\prime}\right\}$ | | 1 | 1 | 0 | 1 | $\overline{S_{3}}:\left\{v_{1}, v_{4}, v_{6}, v_{0}, v_{9}, v_{2}^{\prime}\right\}$ | | 1 | 1 | 1 | 0 | $\overline{\text { none }}$ | | 1 | 1 | 1 | 1 | $\overline{S_{3}}:\left\{v_{1}, v_{4}, v_{6}, v_{0}, v_{9}, v_{2}^{\prime}\right\}$ | Table 6: Case analysis for the pairs $\left(v_{2}^{\prime}, v_{6}\right),\left(v_{2}^{\prime}, v_{7}\right),\left(v_{2}^{\prime}, v_{8}\right),\left(v_{2}^{\prime}, v_{0}\right)$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{2}^{\prime}$ | 0 | $*$ | 0 | 1 | 1 | 1 | 1 | 1 | 1 | 0 | Table 7: Connections between $v_{2}^{\prime}$ and $G_{10}$. $\left(v_{0}^{\prime}, v_{3}\right),\left(v_{0}^{\prime}, v_{5}\right)$ and $\left(v_{0}^{\prime}, v_{7}\right)$, must be non-edges. Furthermore, the pairs $\left(v_{0}^{\prime}, v_{2}^{\prime}\right),\left(v_{0}^{\prime}, v_{3}^{\prime}\right)$, $\left(v_{0}^{\prime}, v_{6}^{\prime}\right)$ and $\left(v_{0}^{\prime}, v_{7}^{\prime}\right)$ must be edges, since otherwise one of the following 3 -combs would appear: $\left\{v_{4}, v_{5}, v_{2}^{\prime}, v_{1}, v_{7}, v_{0}^{\prime}\right\},\left\{v_{1}, v_{8}, v_{3}^{\prime}, v_{2}, v_{6}, v_{0}^{\prime}\right\},\left\{v_{1}, v_{8}, v_{6}^{\prime}, v_{3}, v_{5}, v_{0}^{\prime}\right\}$, or $\left\{v_{1}, v_{8}, v_{9}, v_{3}, v_{7}^{\prime}, v_{0}^{\prime}\right\}$. The pairs $\left(v_{0}^{\prime}, v_{4}^{\prime}\right)$ and $\left(v_{0}^{\prime}, v_{8}^{\prime}\right)$ cannot be edges, since otherwise the 3 -combs induced by $\left\{v_{1}, v_{2}, v_{4}^{\prime}, v_{3}, v_{5}, v_{0}^{\prime}\right\}$ and $\left\{v_{2}, v_{3}, v_{8}^{\prime}, v_{1}, v_{7}, v_{0}^{\prime}\right\}$ would appear. Finally, the pair $\left(v_{0}^{\prime}, v_{0}\right)$ remains uncertain. Table 12 shows the connections between $v_{0}^{\prime}$ and $G_{10}$. Next, let us consider the 2 -comb induced by $\left\{v_{3}^{\prime}, v_{7}^{\prime}, v_{4}, v_{8}\right\}$ and denote by $v_{9}^{\prime}$ the vertex settling it. Notice that this 2 -comb can be obtained from $\left\{v_{1}^{\prime}, v_{5}^{\prime}, v_{2}, v_{6}\right\}$ by applying transformation $A_{4}$. Therefore $v_{9}^{\prime}$ must have the same connections as $v_{0}^{\prime}$ after applying $A_{4}$. Table 13 shows the connections between $v_{9}^{\prime}$ and $G_{10}$. We summarize the connections between vertices $v_{1}^{\prime}, \ldots, v_{9}^{\prime}, v_{0}^{\prime}$ in Table 14, and between | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{4}^{\prime}$ | 1 | 1 | 0 | $*$ | 0 | 1 | 1 | 1 | 0 | 1 | Table 8: Connections between $v_{4}^{\prime}$ and $G_{10}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{6}^{\prime}$ | 1 | 1 | 1 | 1 | 0 | $*$ | 0 | 1 | 1 | 0 | Table 9: Connections between $v_{6}^{\prime}$ and $G_{10}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{8}^{\prime}$ | 0 | 1 | 1 | 1 | 1 | 1 | 0 | $*$ | 0 | 1 | Table 10: Connections between $v_{8}^{\prime}$ and $G_{10}$. | ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-215.jpg?height=56&width=201&top_left_y=957&top_left_x=405) | $\overline{S_{3}}$ | Edge | $\overline{S_{3}}$ | | :---: | :---: | :---: | :---: | | $\left(v_{1}^{\prime}, v_{2}^{\prime}\right)=1$ | $S_{3}:\left\{v_{4}, v_{5}, v_{2}^{\prime}, v_{8}, v_{0}, v_{1}^{\prime}\right\}$ | $\left(v_{1}^{\prime}, v_{3}^{\prime}\right)=0$ | $S_{3}:\left\{v_{6}, v_{1}^{\prime}, v_{3}^{\prime}, v_{4}, v_{8}, v_{0}\right\}$ | | $=0$ | $S_{3}:\left\{v_{6}, v_{1}^{\prime}, v_{4}^{\prime}, v_{3}, v\right.$ | $\left(v_{1}^{\prime}, v_{5}^{\prime}\right)=1$ | $\overline{S_{3}}:$ | | $=0$ | $S_{3}:\left\{v_{4}\right.$ | $=0$ | $S_{3}:$ | | | $\left.v_{8}^{\prime}, v_{2}, v_{9}, v_{1}^{\prime}\right\}$ | $\left.b^{\prime}\right)=1$ | $S_{3}:\left\{v_{7}\right.$ | | | $\overline{S_{3}}:\left\{v_{1}, v_{3}, v_{2}^{\prime}, v_{7}, v_{0}, v_{4}^{\prime}\right\}$ | $\left(v_{2}^{\prime}, v\right.$ | $S_{3}:\left\{v_{8}, v_{5}^{\prime}\right.$ | | | $\left.v_{7}, v_{8}, v_{2}^{\prime}, v_{6}^{\prime}\right\}$ | $\left(v_{2}^{\prime}\right.$ | $S_{3}:\left\{v_{4}, v_{7}^{\prime}\right.$ | | | $\overline{S_{3}}:\left\{v_{1}\right.$ | | $S_{3}:\left\{v_{6}, v_{7}, v_{4}^{\prime}\right.$ | | 0 | $v_{3}^{\prime}, v_{2}, v$ | $=0$ | $S_{3}:\left\{v_{8}, v_{3}^{\prime}, v_{6}^{\prime}\right.$ | | | $\left.v_{3}^{\prime}, v_{1}, v_{5}, v_{7}^{\prime}\right\}$ | $=0$ | $S_{3}$ | | | $S_{3}:\left\{v_{1}, 2\right.$ | | $\overline{S_{3}}:$ | | -0 | $S_{3}:\left\{v_{2}, v_{7}^{\prime}, v_{4}^{\prime}, v_{3}, v_{5}, v_{8}\right\}$ | $1-0$ | $\overline{S_{2}}$ | | | $S_{3}:\left\{v_{1}, v_{8}, v_{6}^{\prime}, v_{4}, v_{0}, v_{5}^{\prime}\right\}$ | $=0$ | $S_{3}:\left\{v_{2}, v_{5}^{\prime}, v_{7}^{\prime},\right\}$ | | $(8)=0$ | : $:\left\{v_{2}, v_{5}^{\prime}, v_{8}^{\prime}, v_{1}, v_{4}\right.$ | 1 | $S_{3}:\left\{v_{3}\right.$ | | | | | | Table 11: Case analysis for the connections between $v_{1}^{\prime}, \ldots, v_{8}^{\prime}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{0}^{\prime}$ | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 | $*$ | Table 12: Connections between $v_{0}^{\prime}$ and $G_{10}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{9}^{\prime}$ | 0 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | $*$ | Table 13: Connections between $v_{9}^{\prime}$ and $G_{10}$. | | $v_{1}^{\prime}$ | $v_{2}^{\prime}$ | $v_{3}^{\prime}$ | $v_{4}^{\prime}$ | $v_{5}^{\prime}$ | $v_{6}^{\prime}$ | $v_{7}^{\prime}$ | $v_{8}^{\prime}$ | $v_{9}^{\prime}$ | $v_{0}^{\prime}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{1}^{\prime}$ | - | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 1 | 1 | | $v_{2}^{\prime}$ | 1 | - | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | | $v_{3}^{\prime}$ | 0 | 1 | - | 1 | 0 | 0 | 1 | 0 | 1 | 1 | | $v_{4}^{\prime}$ | 0 | 0 | 1 | - | 1 | 0 | 0 | 0 | 1 | 0 | | $v_{5}^{\prime}$ | 1 | 0 | 0 | 1 | - | 1 | 0 | 0 | 1 | 1 | | $v_{6}^{\prime}$ | 0 | 0 | 0 | 0 | 1 | - | 1 | 0 | 0 | 1 | | $v_{7}^{\prime}$ | 0 | 0 | 1 | 0 | 0 | 1 | - | 1 | 1 | 1 | | $v_{8}^{\prime}$ | 1 | 0 | 0 | 0 | 0 | 0 | 1 | - | 1 | 0 | | $v_{9}^{\prime}$ | 1 | 0 | 1 | 1 | 1 | 0 | 1 | 1 | - | 1 | | $v_{0}^{\prime}$ | 1 | 1 | 1 | 0 | 1 | 1 | 1 | 0 | 1 | - | Table 14: Connections between vertices $v_{1}^{\prime}, \ldots, v_{9}^{\prime}, v_{0}^{\prime}$. | | $v_{1}$ | $v_{2}$ | $v_{3}$ | $v_{4}$ | $v_{5}$ | $v_{6}$ | $v_{7}$ | $v_{8}$ | $v_{9}$ | $v_{0}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | $v_{1}^{\prime}$ | $*$ | 0 | 1 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | | $v_{2}^{\prime}$ | 0 | $*$ | 0 | 1 | 1 | 1 | 1 | 1 | 1 | 0 | | $v_{3}^{\prime}$ | 1 | 0 | $*$ | 0 | 1 | 1 | 0 | 1 | 0 | 0 | | $v_{4}^{\prime}$ | 1 | 1 | 0 | $*$ | 0 | 1 | 1 | 1 | 0 | 1 | | $v_{5}^{\prime}$ | 0 | 1 | 1 | 0 | $*$ | 0 | 1 | 1 | 0 | 0 | | $v_{6}^{\prime}$ | 1 | 1 | 1 | 1 | 0 | $*$ | 0 | 1 | 1 | 0 | | $v_{7}^{\prime}$ | 1 | 1 | 0 | 1 | 1 | 0 | $*$ | 0 | 0 | 0 | | $v_{8}^{\prime}$ | 0 | 1 | 1 | 1 | 1 | 1 | 0 | $*$ | 0 | 1 | | $v_{9}^{\prime}$ | 0 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | $*$ | 0 | | $v_{0}^{\prime}$ | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 | $*$ | Table 15: Connections between vertices $v_{1}, \ldots, v_{9}, v_{0}$ and $v_{1}^{\prime}, \ldots, v_{9}^{\prime}, v_{0}^{\prime}$. $v_{1}, \ldots, v_{9}, v_{0}$ and $v_{1}^{\prime}, \ldots, v_{9}^{\prime}, v_{0}$ in Table 15 . Table 16: Case analysis for the pairs $\left(v_{0}^{\prime}, v_{1}\right),\left(v_{0}^{\prime}, v_{3}\right),\left(v_{0}^{\prime}, v_{4}\right),\left(v_{0}^{\prime}, v_{5}\right),\left(v_{0}^{\prime}, v_{7}\right)$ and $\left(v_{0}^{\prime}, v_{8}\right)$. | $\left(v_{0}^{\prime}, v_{1}\right)$ | $\left(v_{0}^{\prime}, v_{3}\right)$ | $\left(v_{0}^{\prime}, v_{4}\right)$ | $\left(v_{0}^{\prime}, v_{5}\right)$ | $\left(v_{0}^{\prime}, v_{7}\right)$ | $\left(v_{0}^{\prime}, v_{8}\right)$ | $S_{3}$ or $\overline{S_{3}}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 0 | 0 | 0 | 0 | 0 | 0 | $S_{3}:\left\{v_{2}, v_{3}, v_{5}^{\prime}, v_{1}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 0 | 0 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{9}, v_{2}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 0 | 0 | 1 | 0 | $S_{3}:\left\{v_{2}, v_{3}, v_{5}^{\prime}, v_{1}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 0 | 0 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{9}, v_{2}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 0 | 1 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 0 | 1 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 0 | 1 | 1 | 0 | $S_{3}:\left\{v_{2}, v_{3}, v_{5}^{\prime}, v_{1}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 0 | 1 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{9}, v_{2}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 1 | 0 | 0 | 0 | $S_{3}:\left\{v_{3}, v_{4}, v_{9}, v_{2}, v_{8}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 1 | 0 | 0 | 1 | none | | 0 | 0 | 1 | 0 | 1 | 0 | $S_{3}:\left\{v_{3}, v_{4}, v_{9}, v_{2}, v_{8}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 1 | 0 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{7}, v_{9}, v_{2}, v_{5}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 1 | 1 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 1 | 1 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 1 | 1 | 1 | 0 | $S_{3}:\left\{v_{3}, v_{4}, v_{9}, v_{2}, v_{8}, v_{0}^{\prime}\right\}$ | | 0 | 0 | 1 | 1 | 1 | 1 | $S_{3}:\left\{v_{4}, v_{5}, v_{0}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 0 | 1 | 0 | 0 | 0 | 0 | $S_{3}:\left\{v_{3}, v_{7}, v_{9}, v_{1}, v_{6}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 0 | 0 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{9}, v_{2}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 0 | 0 | 1 | 0 | $\overline{S_{3}}:\left\{v_{4}, v_{8}, v_{0}^{\prime}, v_{3}, v_{7}, v_{9}\right\}$ | | 0 | 1 | 0 | 0 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{9}, v_{2}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 0 | 1 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 0 | 1 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 0 | 1 | 1 | 0 | $S_{3}:\left\{v_{3}, v_{7}, v_{0}^{\prime}, v_{2}, v_{5}, v_{8}\right\}$ | | 0 | 1 | 0 | 1 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{9}, v_{2}, v_{4}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 1 | 0 | 0 | 0 | $S_{3}:\left\{v_{3}, v_{7}, v_{9}, v_{1}, v_{6}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 1 | 0 | 0 | 1 | $S_{3}:\left\{v_{3}, v_{4}, v_{0}^{\prime}, v_{2}, v_{5}, v_{8}\right\}$ | | 0 | 1 | 1 | 0 | 1 | 0 | $S_{3}:\left\{v_{4}, v_{5}, v_{9}, v_{6}, v_{8}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 1 | 0 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{4}, v_{0}^{\prime}, v_{2}, v_{5}, v_{8}\right\}$ | | 0 | 1 | 1 | 1 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 1 | 1 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{2}, v_{7}, v_{0}^{\prime}\right\}$ | | 0 | 1 | 1 | 1 | 1 | 0 | $S_{3}:\left\{v_{3}, v_{7}, v_{0}^{\prime}, v_{2}, v_{5}, v_{8}\right\}$ | | 0 | 1 | 1 | 1 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{1}^{\prime}, v_{0}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | | 1 | 0 | 0 | 0 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 0 | 0 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 0 | 0 | 1 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 0 | 0 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 0 | 1 | 0 | 0 | $S_{3}:\left\{v_{3}, v_{7}, v_{1}^{\prime}, v_{2}, v_{8}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 0 | 1 | 0 | 1 | $S_{3}:\left\{v_{3}, v_{7}, v_{5}^{\prime}, v_{4}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 0 | 1 | 1 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}^{\prime}, v_{2}, v_{4}, v_{7}\right\}$ | | $\left(v_{0}^{\prime}, v_{1}\right)$ | $\left(v_{0}^{\prime}, v_{3}\right)$ | $\left(v_{0}^{\prime}, v_{4}\right)$ | $\left(v_{0}^{\prime}, v_{5}\right)$ | $\left(v_{0}^{\prime}, v_{7}\right)$ | $\left(v_{0}^{\prime}, v_{8}\right)$ | $S_{3}$ or $\overline{S_{3}}$ | | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | 1 | 0 | 0 | 1 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}^{\prime}, v_{2}, v_{4}, v_{7}\right\}$ | | 1 | 0 | 1 | 0 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 1 | 0 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 1 | 0 | 1 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 1 | 0 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{9}, v_{3}, v_{6}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 1 | 1 | 0 | 0 | $S_{3}:\left\{v_{3}, v_{4}, v_{9}, v_{2}, v_{8}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 1 | 1 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{0}^{\prime}, v_{2}, v_{4}, v_{7}\right\}$ | | 1 | 0 | 1 | 1 | 1 | 0 | $S_{3}:\left\{v_{3}, v_{4}, v_{9}, v_{2}, v_{8}, v_{0}^{\prime}\right\}$ | | 1 | 0 | 1 | 1 | 1 | 1 | $S_{3}:\left\{v_{4}, v_{5}, v_{0}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 1 | 1 | 0 | 0 | 0 | 0 | $S_{3}:\left\{v_{6}, v_{7}, v_{1}^{\prime}, v_{5}, v_{8}, v_{0}^{\prime}\right\}$ | | 1 | 1 | 0 | 0 | 0 | 1 | $S_{3}:\left\{v_{3}, v_{1}^{\prime}, v_{0}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | | 1 | 1 | 0 | 0 | 1 | 0 | $S_{3}:\left\{v_{3}, v_{7}, v_{0}^{\prime}, v_{1}, v_{4}, v_{6}\right\}$ | | 1 | 1 | 0 | 0 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{7}, v_{0}^{\prime}, v_{1}, v_{4}, v_{6}\right\}$ | | 1 | 1 | 0 | 1 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 1 | 1 | 0 | 1 | 0 | 1 | $S_{3}:\left\{v_{3}, v_{1}^{\prime}, v_{0}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | | 1 | 1 | 0 | 1 | 1 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}^{\prime}, v_{2}, v_{4}, v_{7}\right\}$ | | 1 | 1 | 0 | 1 | 1 | 1 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}^{\prime}, v_{2}, v_{4}, v_{7}\right\}$ | | 1 | 1 | 1 | 0 | 0 | 0 | $S_{3}:\left\{v_{4}, v_{5}, v_{9}, v_{6}, v_{8}, v_{0}^{\prime}\right\}$ | | 1 | 1 | 1 | 0 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{0}^{\prime}, v_{2}, v_{4}, v_{7}\right\}$ | | 1 | 1 | 1 | 0 | 1 | 0 | $S_{3}:\left\{v_{4}, v_{5}, v_{9}, v_{6}, v_{8}, v_{0}^{\prime}\right\}$ | | 1 | 1 | 1 | 0 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{4}, v_{0}^{\prime}, v_{2}, v_{5}, v_{8}\right\}$ | | 1 | 1 | 1 | 1 | 0 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 1 | 1 | 1 | 1 | 0 | 1 | $S_{3}:\left\{v_{1}, v_{8}, v_{0}^{\prime}, v_{2}, v_{4}, v_{7}\right\}$ | | 1 | 1 | 1 | 1 | 1 | 0 | $S_{3}:\left\{v_{1}, v_{5}, v_{0}^{\prime}, v_{3}, v_{6}, v_{8}\right\}$ | | 1 | 1 | 1 | 1 | 1 | 1 | $S_{3}:\left\{v_{3}, v_{1}^{\prime}, v_{0}^{\prime}, v_{2}, v_{6}, v_{8}\right\}$ | Interestingly, the graph induced by $v_{1}^{\prime}, \ldots, v_{9}^{\prime}, v_{0}^{\prime}$ is an isomorphic copy of $G_{10}$. Moreover, $\left(v_{i}, v_{j}^{\prime}\right)$ for $i \neq j$ is an edge if and only if $\left(v_{i}, v_{j}\right)$ is not an edge, while the pairs $\left(v_{i}, v_{i}^{\prime}\right)$, $i=0,1, \ldots, 9$ are uncertain. Thus, this configuration is the sum of two copies of $G_{10}$, that is, the graph $2 G_{10}$ (see Figure 29). Let us recall that to any graph $G$ we can apply the same operation and obtain the sum $G+G=2 G$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-218.jpg?height=480&width=1162&top_left_y=1870&top_left_x=478) Figure 29: The sum of two graphs $G_{10}$ or $2 G_{10}$. Another remarkable property of the obtained configuration is as follows: if we exchange $v_{0}$ with $v_{0}^{\prime}$ and $v_{9}$ with $v_{9}^{\prime}$ then the resulting graph becomes the sum of two Petersen graphs, that is, $2 \mathcal{P} \equiv 2 G_{10}$, as shown in Figure 30 . ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-219.jpg?height=486&width=1179&top_left_y=471&top_left_x=473) Figure 30: The graph $2 \mathcal{P}$, isomorphic to $2 G_{10}$ by exchanging $v_{0}, v_{0}^{\prime}$ and $v_{9}, v_{9}^{\prime}$. This completes the proof of Lemma 3. ### 2.4 Proof of Lemma 4 We prove that if a graph $G$ contains an induced $2 \mathcal{P}$, then it must have either an unsettled 2-comb, or an induced 3-comb or 3-anti-comb. Let us recall that $2 \mathcal{P}$ still has 10 uncertain edges. Hence, it gives us in fact 1024 possible graphs, one of which is an induced subgraph of $G$. Since we do not know which one, we will prove the statement by considering each such possible subgraphs. Remarkably, none of these 1024 graphs contains an induced 3-comb or 3-anti-comb, as verified by computer. Furthermore, $2 \mathcal{P}$ itself contains no induced 2 -combs either. (Since $2 \mathcal{P}$ contains uncertain pairs, we call a subgraph of $2 \mathcal{P}$ an induced one only if it does not involve any uncertain pair.) However, each of the 1024 graphs obtained from $2 \mathcal{P}$ contains many 2-combs each of which involves exactly one pair of vertices $v_{i}$ and $v_{i}^{\prime}$ for some index $i$. Now we will fix one of the uncertain pairs (once as an edge and once as a non-edge), while keeping all others uncertain. Several (36) unsettled induced 2-combs appear that contain the fixed uncertain pair. Each of these 2-combs must be settled in $G$ by our assumption (i), thus there exists a vertex $x$ settling it. There are 16 pairs $(x, y)$, where $y$ is a vertex of $2 \mathcal{P}$, not belonging to the unsettled 2-comb. We check all $2^{16}$ possible edge/non-edge assignments to these 16 pairs, and find by computer search that for each of them an induced 3-comb or 3 -anti-comb exists. More precisely, let us fix the uncertain pair $\left(v_{0}, v_{0}^{\prime}\right)$ and consider two cases: 1. If $\left(v_{0}, v_{0}^{\prime}\right)$ is an edge then the 2 -comb induced by the vertices $\left\{v_{1}, v_{0}, v_{0}^{\prime}, v_{4}\right\}$ is unsettled in $2 \mathcal{P}$, because no vertex in $2 \mathcal{P}$ is connected to both $v_{0}$ and $v_{0}^{\prime}$ by the definition of the sum of two graphs. Let $x$ be a settling vertex. Then, by definition, $\left(x, v_{0}\right),\left(x, v_{0}^{\prime}\right)$ must be edges of $G$, and the pairs $\left(x, v_{1}\right)$ and $\left(x, v_{4}\right)$ must be non-edges. There are 16 other pairs of the form $(x, y)$, where $y$ is a vertex of $2 \mathcal{P}$. Hence, there are $2^{16}$ possible assignments of edges/non-edges between $x$ and $2 \mathcal{P}$. We check by computer all $2^{16}$ possible assignments and find that in each $2^{16}$ graphs there is an induced (without uncertain pairs) 3-comb or 3 -anti-comb. 2. If $\left(v_{0}, v_{0}^{\prime}\right)$ is not an edge of $G$ then the 2 -comb induced by the vertices $\left\{v_{0}, v_{1}, v_{8}, v_{0}^{\prime}\right\}$ is not settled in $2 \mathcal{P}$. Since it must be settled in $G$ by condition (i), there is a vertex $x$ of $G$ that settles it. Similarly to the previous case, we again consider all $2^{16}$ graphs, and find by computer search that all of them contain an induced 3-comb or 3-anti-comb. This concludes the proof of Lemma 4. ## 3 Proof of Theorems 2 and 3 Proof of Theorem 2: Recall by (1.1) that we can reduce the case analysis by assuming that $1 \leq k<\ell \leq n-2$. We start by proving (i). Assume by contradiction that there exists an unsettled $\overline{S_{m}}=\left\{B_{1}, \ldots, B_{m}, A_{1}, \ldots, A_{m}\right\},\left|B_{i}\right|=k,\left|A_{i}\right|=\ell$. Then, by assumption we must have $$ A_{i} \supset B_{j} \text { for all } j \neq i \quad \text { and } \quad A_{i} \not \supset B_{i} $$ Let us recall that $\overline{S_{m}}$ is settled by a $k$-set $K$ iff $K \subseteq \bigcap_{j=1}^{m} A_{j}$, and it is settled by an $\ell$-set $L$ iff $L \nsupseteq B_{i}$ for $i=1, \ldots, m$. Let $\mathcal{B}=\left\{B_{1}, \ldots, B_{m}\right\}$, and let $X \subseteq[n]$ be the set that contains the elements that are in more that one of the $B_{i}$ 's, i.e. $X=\left\{x \in[n] \mid \operatorname{deg}_{\mathcal{B}}(x)>1\right\}$. Notice that $X \subseteq \bigcap_{j=1}^{m} A_{j}$ because by (3.9) we have that every vertex belonging to two or more of the sets from $\mathcal{B}$ must belong to all sets $A_{i}, i=1, \ldots, m$. Clearly $|X|n-\ell$, since otherwise there exists an $\ell$-set settling $\overline{S_{m}}$. Thus, we can conclude that $$ \tau(\mathcal{H})+a_{0} \geq n-\ell+1 $$ Assume w.l.o.g. that $\left|B_{1} \cap X\right| \leq\left|B_{2} \cap X\right| \leq \ldots \leq\left|B_{m} \cap X\right|$. Since we know by (3.9) that $\bigcup_{j=1}^{m-1} B_{j} \subseteq A_{m}$, we have: $$ \left|\bigcup_{i=1}^{m-1} B_{j}\right|=|X|+\sum_{p=0}^{q}(k-p) a_{p}-(k-q) \leq \ell $$ Let us now take away $k$ times equation (3.10) from (3.11) and obtain $$ |X|+\sum_{p=0}^{q}(k-p) a_{p}-(k-q)-k\left(\tau(\mathcal{H})+a_{0}\right) \leq \ell-k(n-\ell+1) $$ which can be simplified to $$ |X|+\sum_{p=1}^{q}(k-p) a_{p}+q-k \tau(\mathcal{H}) \leq(k+1) \ell-k n $$ Notice that the right had side of (3.12) is negative by our initial assumption of $k n>$ $(k+1) \ell$. Thus, to arrive to a contradiction, it is enough to prove that $$ k \tau(\mathcal{H}) \leq|X|+\sum_{p=1}^{q}(k-p) a_{p}+q $$ Let us observe next that $\sum_{p=1}^{q}\left(k a_{p}\right)=k|\mathcal{H}|$, and that $\sum_{p}\left(p a_{p}\right)=\sum_{H \in \mathcal{H}}|H|$. Thus, we can equivalently rewrite inequality (3.13) as: $$ k(|\mathcal{H}|-\tau(\mathcal{H})) \geq \sum_{H \in \mathcal{H}}|H|-|X|-q $$ To show (3.14), let us construct a cover $C$ of $\mathcal{H}$ as follows. First we choose into $C$ a vertex of the highest degree in $\mathcal{H}$. This vertex covers at least $\frac{\sum_{H \in \mathcal{H}}|H|}{X]}$ hyperedges of $\mathcal{H}$. We cover the remaining edges by choosing one vertex from each. This simple procedure shows that $$ \tau(\mathcal{H}) \leq|C| \leq|\mathcal{H}|-\frac{\sum_{H \in \mathcal{H}}|H|}{|X|}+1 $$ From this simple inequality we can derive the following: $$ \begin{aligned} k(|\mathcal{H}|-\tau(\mathcal{H})) & \geq \frac{k}{|X|} \sum_{H \in \mathcal{H}}|H|-k \\ & =\sum_{H \in \mathcal{H}}|H|+\frac{k-|X|}{|X|} \sum_{H \in \mathcal{H}}|H|-k \\ & \geq \sum_{H \in \mathcal{H}}|H|-|X| \end{aligned} $$ where, the second inequality follows from $|X| \leq \sum_{H \in \mathcal{H}}|H|$, which is true, since every vertex of $X$ has degree at least 2 in $\mathcal{B}$. The above inequalities then prove (3.14), since $q \geq 0$, which then yields the desired contradiction, completing the proof of (i). We prove next (ii). We will show, by a construction that an unsettled $\overline{S_{m}}$ exists in $G(n, k, \ell)$, whenever $k n \leq(k+1) \ell$ and $n \geq k+\ell$. For this let us set $r \equiv \ell(\bmod k), 0 \leq rr$ and our assumption that $k n \leq(k+1) \ell$. Thus, $|L|<\ell$ follows, completing the proof of (ii). This completes the proof of Theorem 2. ## Proof of Theorem 3: We prove first (a). Even though this claim is only for $k \leq 2$, let us first disregard this restriction. Assume by contradiction that there exists an unsettled $S_{m}$ in $G(m, k, \ell)$ defined by the sets $\left\{B_{1}, \ldots, B_{m}, A_{1}, \ldots, A_{m}\right\}$, where $\left|B_{i}\right|=k,\left|A_{i}\right|=\ell$, for $i=1, \ldots, m$, and $B_{j} \subseteq A_{i}$, iff $i=j$. Set $\mathcal{B}=\left\{B_{1}, \ldots, B_{m}\right\}$ and $\mathcal{A}=\left\{A_{1}, \ldots, A_{m}\right\}$. By definitions, an $\ell$-set $L$ can settle $S_{m}$ only if $[n] \backslash L$ is a vertex cover of the hypergraph $\mathcal{B}$. Furthermore, a $k$-set $K$ can settle $S_{m}$, only if $K \subseteq A_{i}$ for all $i=1, \ldots, m$. Since $S_{m}$ is assumed to be unsettled in $G(n, k, \ell)$, we must have the following properties. (i) $\tau(\mathcal{B}) \geq n-\ell+1$, since otherwise the complement of a minimum vertex cover of $\mathcal{B}$ would contain a settling $\ell$-set. (ii) $\left|\bigcap_{i=1}^{m} A_{i}\right|k$ then for all $v \in Q_{j}$, there is an index $i$ such that $B_{i} \subset Q_{j} \backslash\{v\}$, implying by the definitions that $v \notin A_{i}$. Hence, (3.20) follows. Case 2. If $\left|Q_{j}\right|=k$ and $m \geq k+1$, then we have $Q_{j}=B_{i^{*}}$ for exactly one index $i^{*} \in$ $\{1, \ldots, m\}$. For all other indices $i$ we have $Q_{j} \cap A_{i}=R_{i j}$ of size $k-1$. Thus, since $m \geq k+1$, we can choose for each $v \in Q_{j}$ an index $i \neq i^{*}$ such that $v \notin A_{i}$, implying (3.20). Case 3. If $m \leq k$ then we must have $\left|Q_{1}\right|=\left|Q_{2}\right|=\ldots=\left|Q_{p}\right|=k, m=p \leq k$, since we already know that $\left|Q_{j}\right| \geq k$ for all $j=1, \ldots, p$, and if $\left|Q_{j}\right|>k$ for at least one index $j$, then $m \geq k+1$ would be implied. Thus, we have $$ \begin{aligned} n & =|X|+\sum_{i=1}^{p}\left|Q_{i}\right| \\ & =r-1+p k \\ & =r-1+k \frac{\ell-r}{k-1} \\ & =\ell \frac{k}{k-1}-\frac{r}{k-1}-1 \end{aligned} $$ and hence, by our assumption, we must have $\ell \geq r+k^{2}-k+1$. However, $p \leq k$ implies that $p=\frac{\ell-r}{k-1} \leq k$ from which $\ell \leq r+k^{2}-k$ follows. This completes the proof of Theorem 3. ## 4 More about CIS- $d$-graphs ### 4.1 Proofs of Propositions 6,7, 11 and Theorem 4 Proof of Proposition 6. Obviously, every partition of colors can be realized by successive identification of two colors. Hence, the following Lemma implies Proposition 6. Given a $(d+1)$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}, E_{d+1}\right)$, let us identify the last two colors $d$ and $d+1$ and consider the $d$-graph $\mathcal{G}^{\prime}=\left(V ; E_{1}, \ldots, E_{d-1}, E_{\mathbf{d}}\right)$, where $E_{\mathbf{d}}=E_{d} \cup E_{d+1}$. Lemma 5. If $\mathcal{G}$ is a $\operatorname{CIS}-(d+1)$-graph then $\mathcal{G}^{\prime}$ is a CIS-d-graph. Proof. Suppose that $\mathcal{G}^{\prime}$ does not have the CIS- $d$-property, that is, there are $d$ vertex-sets $C_{1}, \ldots, C_{d-1}, C_{\mathrm{d}} \subseteq V$ such that they have no vertex in common, where $C_{i}$ is a maximal subset of $V$ avoiding color $i$ for $i=1, \ldots, d-1$, and $C_{\mathbf{d}}$ is a maximal subset of $V$ avoiding both colors $d$ and $d+1$. Clearly, there exist maximal vertex-sets $C_{d}$ and $C_{d+1}$ avoiding colors $d$ and $d+1$ respectively and such that $C_{d} \cap C_{d+1}=C_{\mathrm{d}}$. Then $C_{1}, \ldots, C_{d-1}, C_{d}, C_{d+1} \subseteq V$ are maximal vertex-sets avoiding colors $1, \ldots, d-1, d, d+1$ respectively and with no vertex in common. Hence, the $(d+1)$-graph $\mathcal{G}^{\prime}$ does not have the CIS- $(d+1)$-property, either. A little later we will need the following similar claim. Lemma 6. If $\mathcal{G}$ is a Gallai $(d+1)$-graph then $\mathcal{G}^{\prime}$ is a Gallai d-graph. Proof. It is obvious. If $\mathcal{G}^{\prime}$ contains a $\Delta$ then the same three vertices form a $\Delta$ in $\mathcal{G}$ too. Proof of Proposition 7. It follows by a routine case analysis from the definitions. First, let us consider Gallai's property. Suppose that $\mathcal{G}$ has a $\Delta$. Clearly, it cannot contain exactly one edge in $\mathcal{G}^{\prime \prime}$, since then two remaining edges are of the same color. If this $\Delta$ contains 2 edges in $\mathcal{G}^{\prime \prime}$ then the third one is there, too, and hence $G^{\prime \prime}$ contains a $\Delta$. If all 3 edges are in $\mathcal{G}^{\prime}$ then $G^{\prime}$ contains a $\Delta$. Conversly, If $\mathcal{G}^{\prime}$ or $\mathcal{G}^{\prime \prime}$ contains a $\Delta$ then, clearly, $\mathcal{G}$ contains it too, since both $\mathcal{G}^{\prime}$ and $\mathcal{G}^{\prime \prime}$ are induced subgraphs of $\mathcal{G}$. Now let us consider the CIS-property. To simplify the notation we restrict ourselves by the case $d=2$, though exactly the same arguments work in general. It is easy to see that any maximal cliques (respectively, stable sets) of $\mathcal{G}^{\prime}$ which do not contain $v$ remain unchanged in $\mathcal{G}$, while a maximal clique $C^{\prime}$ (respectively, a maximal stable set $S^{\prime}$ ) of $\mathcal{G}^{\prime}$ which contains $v$ and for every maximal clique $C^{\prime \prime}$ (respectively, every maximal stable set $S^{\prime \prime}$ ) of $\mathcal{G}^{\prime \prime}$ the set $C=C^{\prime} \cup C^{\prime \prime} \backslash\{v\}$ (respectively, $S=S^{\prime} \cup S^{\prime \prime} \backslash\{v\}$ is a maximal clique (respectively, a maximal stable set) of $\mathcal{G}$ and moreover, there are no other maximal cliques (respectively, maximal stable sets) in $\mathcal{G}$. It is not difficult to verify that every maximal clique $C=C^{\prime} \cup C^{\prime \prime} \backslash\{v\}$ and every maximal stable set $S=S^{\prime} \cup S^{\prime \prime} \backslash\{v\}$ in $\mathcal{G}$ intersect if and only if every maximal clique $C^{\prime}$ intersects every a maximal stable set $S^{\prime}$ of $G^{\prime}$ and every maximal clique $C^{\prime \prime}$ intersects every a maximal stable set $S^{\prime \prime}$ of $G^{\prime \prime}$. Indeed, if $C^{\prime} \cap S^{\prime}=\left\{v^{\prime}\right\} \neq\{v\}$ then $C \cap S=\left\{v^{\prime}\right\}$ for any $C^{\prime \prime}$ and $S^{\prime \prime}$. If $C^{\prime} \cap S^{\prime}=\{v\}$ then $C \cap S=C^{\prime \prime} \cap S^{\prime \prime}$ and hence $C \cap S \neq \emptyset$ if and only if $C^{\prime \prime} \cap S^{\prime \prime} \neq \emptyset$. If $C \cap S \neq \emptyset$ then both $C^{\prime} \cap S^{\prime}$ and $C^{\prime \prime} \cap S^{\prime \prime}$ must be non-empty. ## Proof of Theorem 4. Part (a). By Proposition 7, $\mathcal{G}$ is exactly closed under substitution. By Proposition 9, $\mathcal{G}$ can be obtained from 2-graphs by substitutions. Such a decomposition of $\mathcal{G}$ is given by a tree $T(\mathcal{G})$ whose leaves correspond to 2-graphs. It is easy to see that by construction each chromatic component of $\mathcal{G}$ is decomposed by the same tree $T(\mathcal{G})$. Hence, all we have to prove is that both chromatic components of every 2 -graph belong to $\mathcal{F}$. For colors $1, \ldots, d-1$ this holds, since $\mathcal{F}$ is exactly closed under substitution, and for the color $d$ it holds, too, since $\mathcal{F}$ is also closed under complementation. Part (b). It follows easily from from part (a). As in Lemma 5, given a $(d+1)$-graph $\mathcal{G}=\left(V ; E_{1}, \ldots, E_{d}, E_{d+1}\right)$, let us identify the last two colors $d$ and $d+1$ and consider the $d$-graph $\mathcal{G}^{\prime}=\left(V ; E_{1}, \ldots, E_{d-1}, E_{\mathbf{d}}\right)$, where $E_{\mathbf{d}}=E_{d} \cup E_{d+1}$. We assume that $\mathcal{G}$ is $\Delta$-free and that $G_{i}=\left(V, E_{i}\right) \in \mathcal{F}$ for $i=1, \ldots, d-1$. Then, by Lemma $6, \mathcal{G}^{\prime}$ is $\Delta$-free, too, and it follows from part (a) that $G_{\mathbf{d}}=\left(V, E_{\mathbf{d}}\right)$ is also in $\mathcal{F}$. Hence, the union of any two colors is in $\mathcal{F}$. From this by induction we derive that the union of any set of colors is in $\mathcal{F}$. Proof of Proposition 11. Given $\mathcal{G}$, let us again consider the decomposition tree $T(\mathcal{G})$, fix an arbitrary its leaf $v$, and consider the corresponding 2-graph $\mathcal{G}_{v}$. Both its chromatic components are CIS-graphs, by Proposition 8. Hence, $\mathcal{G}_{v}$ is a CIS-d-graph. Thus, $\mathcal{G}$ is a CIS- $d$-graph, too, by Proposition 7. ### 4.2 Settling $\Delta$ Let $V=\left\{v_{1}, v_{2}, v_{3}\right\}$ and assume that $E_{1}=\left\{\left(v_{1}, v_{2}\right)\right\}, E_{2}=\left\{\left(v_{2}, v_{3}\right)\right\}$, and $E_{3}=\left\{\left(v_{3}, v_{1}\right)\right\}$ form a $\Delta$, see Figure 12. Obviously, $\Delta$ is not a CIS-3-graph. Indeed, let us consider $C_{1}=$ $\left\{v_{2}, v_{3}\right\}, C_{2}=\left\{v_{3}, v_{1}\right\}$, and $C_{3}=\left\{v_{1}, v_{2}\right\}$. There is no edge from $E_{i}$ in $C_{i}$ for $i=1,2,3$ and $C_{1} \cap C_{2} \cap C_{3}=\emptyset$. Hence, if a CIS-3-graph $\mathcal{G}=\left(V ; E_{1}, E_{2}, E_{3}\right)$ contains a $\Delta$ then it must contain a vertex $v_{4}$ such that the sets $C_{1}^{\prime}=\left\{v_{2}, v_{3}, v_{4}\right\}, C_{2}^{\prime}=\left\{v_{3}, v_{1}, v_{4}\right\}$, and $C_{3}^{\prime}=\left\{v_{1}, v_{2}, v_{4}\right\}$ contain no edges from $E_{1}, E_{2}$, and $E_{3}$, respectively. Similarly, let us consider the sets $C_{1}=\left\{v_{3}, v_{1}\right\}, C_{2}=\left\{v_{1}, v_{2}\right\}$, and $C_{3}=\left\{v_{2}, v_{3}\right\}$. Again, there is no edge from $E_{i}$ in $C_{i}$ for $i=1,2,3$ and $C_{1} \cap C_{2} \cap C_{3}=\emptyset$. Hence, if a CIS-3-graph $\mathcal{G}=\left(V ; E_{1}, E_{2}, E_{3}\right)$ contains a $\Delta$ then it must contain a vertex $v_{5}$ such that $C_{1}^{\prime}=\left\{v_{3}, v_{1}, v_{5}\right\}$, $C_{2}^{\prime}=\left\{v_{1}, v_{2}, v_{5}\right\}$, and $C_{3}^{\prime}=\left\{v_{2}, v_{3}, v_{5}\right\}$ contain no edges from $E_{1}, E_{2}$, and $E_{3}$, respectively. It is easy to check that $v_{4} \neq v_{5}$ and that we must have $\left(v_{4}, v_{1}\right),\left(v_{1}, v_{2}\right),\left(v_{2}, v_{5}\right) \in E_{1}$, $\left(v_{4}, v_{2}\right),\left(v_{2}, v_{3}\right),\left(v_{3}, v_{5}\right) \in E_{2},\left(v_{4}, v_{3}\right),\left(v_{3}, v_{1}\right),\left(v_{1}, v_{5}\right) \in E_{3}$, see Figure 12 . This leaves only one pair $\left(v_{4}, v_{5}\right)$ whose color is not implied. Yet, let us note that for any coloring of $\left(v_{4}, v_{5}\right)$ a new $\Delta$ appears. For example, if $\left(v_{4}, v_{5}\right) \in E_{1}$ then vertices $\left(v_{3}, v_{4}, v_{5}\right)$ form a $\Delta^{\prime}$. ### 4.3 A stronger conjecture We say that two vertices $v_{4}$ and $v_{5}$ settle $\Delta$. Note however that $v_{1}$ and $v_{2}$ do not settle $\Delta^{\prime}$. So we need more vertices to settle it. Nevertheless, there are $d$-graphs whose all $\Delta \mathrm{s}$ are settled. First such example was given by Andrey Gol'berg in 1984, see Figure 34. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-228.jpg?height=304&width=780&top_left_y=340&top_left_x=635) Figure 31: Settling $\Delta$. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-228.jpg?height=304&width=776&top_left_y=777&top_left_x=638) Figure 32: Settling $\Delta$ (in black and white for printing). We call this construction a 4-cycle. It has $4 \Delta \mathrm{s}$ and they are all settled. Yet, if we partition its three colors into two sets we will get 44 2-combs none of which is settled. Hence, by Proposition 6, the 4-cycle is not a CIS-3-graph. Moreover, in the next section we give examples of 3 -graphs whose all $\Delta \mathrm{s}$ and 2 -combs are settled, however, their 2-projections have unsettled induced 3-combs or 3-anti-combs. Conjecture 4. Let $\mathcal{G}$ be a non-Gallai 3-graph with chromatic components $G_{1}, G_{2}, G_{3}$, then there is an unsettled $\Delta$ in $\mathcal{G}$ or $G_{i}$ has an unsettled induced comb or anti-comb for some $i=1,2,3$. Obviously, Proposition 6 and Conjecture 4 imply Conjecture 3. Remark 11. It is not difficult to show that for every fixed $k, d \in \mathbb{Z}_{+}$and $\epsilon>0$ there is $n=n(k, d, \epsilon)$ such that in a random $d$-graph $\mathcal{G}$ with a fixed $|V(G)| \geq k$ all $\Delta s$ as well as all induced $m$-combs and $m$-anti-combs for $m \leq k$ in all projections of $\mathcal{G}$ are settled with probability greater than $1-\epsilon$. Yet, for $m>k$, unsettled induced $m$-combs and $m$-anti-combs exist with high probability. ### 4.4 Even cycles and flowers In this section we describe some interesting 3-graphs in support of Conjecture 4. They have all $\Delta \mathrm{s}$ settled, and sometimes even all 2-combs are settled in their 2-projections. However, then unsettled 3-combs, or 3-anti-combs, or 4-combs appear. Let us consider four $\Delta \mathrm{s}$ in Figure 33. They form a cycle. This construction can be extended (uniquely) to a 3-graph, shown in Figure 35, in which all four $\Delta \mathrm{s}$ are settled "counterclockwise" (i.e., $\Delta$ s induced by the triplets $\{0,1,2\},\{2,3,4\}$, ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-229.jpg?height=877&width=874&top_left_y=367&top_left_x=611) Figure 33: Initial 4-cycle structure. $\{4,5,6\}$, and $\{6,7,0\}$ are settled by the pairs $\{3,4\},\{5,6\},\{7,1\}$, and $\{1,2\}$, respectively), and no new $\Delta$ appears. However, 2-projections of this 3-graph contain 44 unsettled 2-combs (induced by the quadruples $\{0,5,1,4\},\{3,2,6,7\},\{4,1,2,3\},\{0,5,6,7\}$, etc.) as shown in Figure 35 . Level 1: GBBGGBBG Level 2: RGRBRGRB Level 3: RBRGRBRG Level 4: GBBRGBBR 4 settled $\Delta \mathrm{s}$ $44 S_{2}$ : 0 settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-230.jpg?height=866&width=871&top_left_y=1126&top_left_x=615) Figure 34: 4-cycle with all $\Delta \mathrm{s}$ settled. Level 1: GBBGGBBG Level 2: RGRBRGRB Level 3: RBRGRBRG Level 4: GBBRGBBR 4 settled $\Delta \mathrm{s}$ $44 S_{2}: 0$ settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-231.jpg?height=1262&width=1265&top_left_y=908&top_left_x=410) Figure 35: 4-cycle (in black and white for printing). This 3-graph was constructed by Andrey Gol'berg in 1984. ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-232.jpg?height=877&width=874&top_left_y=932&top_left_x=611) Figure 36: Initial 4-flower structure. Now, let us consider four $\Delta \mathrm{s}$ with one common vertex as shown in Figure 36. This construction we call a 4-flower. It can be extended to a 3-graph, as shown in Figure 38, in which all four $\Delta \mathrm{s}$ are settled "counterclockwise" (i.e., $\Delta \mathrm{s}$ induced by the triplets $\{0,1,2\}$, $\{0,3,4\},\{0,5,6\}$, and $\{0,7,8\}$ are settled by the pairs $\{3,4\},\{5,6\},\{7,8\}$, and $\{1,2\}$, respectively). Although four more $\Delta \mathrm{s}$ (induced by the triplets $\{0,1,6\},\{0,2,5\},\{0,4,7\}$, and $\{0,3,8)\}$ appear in this extension), yet they are settled too. Moreover, 2-projections of this 3-graph contain twenty induced 2-combs that are all settled. However, there exist also eight induced 3 -combs that are not settled. Using a computer, we analyzed also some larger flowers (namely, $2 j$-flowers for $j=3,4,5$, and 6) shown below. In all these examples all $\Delta \mathrm{s}$ are settled. However, in agreement with Conjecture 4, for each of these 3-graphs always there is a 2-projection that contains an unsettled comb or anti-comb. We have to explain the notation used in the figures. The three colors are red $R$, blue $B$, and green $G$, and we denote them by solid, dashed, and dotted lines, respectively. In a $2 j$-flower we denote the central vertex by 0 and other vertices are labeled by $1,2, \ldots, 2 j-1,2 j$. Due to the symmetry, we can describe this 3 -graph in terms of a list of colors $L$ present in level $i$, where level $i$ contain all edges $(a, b)$ such that $a-b= \pm i$ $(\bmod n)$. Clearly, we only need to provide the color lists from level 1 to $j$, since level $i$ gives the same assignment as level $2 j-i$. Finally Level 0 shows the coloring of the radial edges. For example, the 4 -flower on Figure 38 is colored as follows: Level 0 : the edges $(0,1),(0,2),(0,3)(0,4),(0,5),(0,6),(0,7),(0,8)$ are colored by $R G R G R G R G$. Level 1: the edges $(1,2),(2,3),(3,4),(4,5),(5,6),(6,7),(7,8),(8,1)$ are colored by $B G B G B G B G$; Level 2: the edges $(1,3),(2,4),(3,5),(4,6),(5,7),(6,8),(7,1),(8,2)$ are all colored by $B B B B B B B B$; Level 3: the edges $(1,4),(2,5),(3,6),(4,7),(5,8),(6,1),(7,2),(8,3)$ are colored by $R B R B R B R B$; Level 4 : the edges $(1,5),(2,6),(3,7),(4,8),((5,1),(6,2),(7,3),(8,4))$ are colored by $R G R G(R G R G)$. Level 0: RGRGRGRG Level 1: BGBGBGBG Level 2: BBBBBBBB Level 3: RBRBRBRB Level 4: RGRGRGRG $8 \Delta \mathrm{s}: 8$ settled $20 S_{2}: 20$ settled $8 S_{3}$ : 0 settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-234.jpg?height=1262&width=1268&top_left_y=977&top_left_x=411) Figure 37: 4-flower example. Level 0: RGRGRGRG Level 1: BGBGBGBG Level 2: BBBBBBBB Level 3: RBRBRBRB Level 4: RGRGRGRG $8 \Delta \mathrm{s}: 8$ settled $20 S_{2}: 20$ settled $8 S_{3}: 0$ settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-235.jpg?height=1268&width=1279&top_left_y=971&top_left_x=411) Figure 38: 4-flower example (in black and white for printing). Level 0: RGRGRGRGRGRG Level 1: BGBGBGBGBGBG Level 2: BBBBBBBBBBBB Level 3: RBRBRBRBRBRB Level 4: RGRGRGRGRGRG Level 5: BRBRBRBRBRBR Level 6: BBBBBBBBBBBB $18 \Delta \mathrm{s}$ : 18 settled $66 S_{2}$ : 66 settled $38 S_{3}: 20$ settled $6 S_{4}$ : 0 settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-236.jpg?height=1265&width=1265&top_left_y=1044&top_left_x=410) Figure 39: 6-flower example. Level 0: RGRGRGRGRGRG Level 1: BGBGBGBGBGBG Level 2: BBBBBBBBBBBB Level 3: RBRBRBRBRBRB Level 4: RGRGRGRGRGRG Level 5: BRBRBRBRBRBR Level 6: BBBBBBBBBBBB $18 \Delta \mathrm{s}$ : 18 settled $66 S_{2}: 66$ settled $38 S_{3}$ : 20 settled $6 S_{4}: 0$ settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-237.jpg?height=1268&width=1268&top_left_y=1022&top_left_x=408) Figure 40: 6-flower example (in black and white for printing). This 3-graph was constructed by Bianca Viray in 2004. Level 0: RGRGRGRGRGRGRGRG Level 1: BGBGBGBGBGBGBGBG Level 2: BBBBBBBBBBBBBBBB Level 3: RBRBRBRBRBRBRBRB Level 4: RGRGRGRGRGRGRGRG Level 5: BRBRBRBRBRBRBRBR Level 6: BBBBBBBBBBBBBBBB Level 7: GBGBGBGBGBGBGBGB Level 8: RRRRRRRRRRRRRRRR $32 \Delta \mathrm{s}: 32$ settled $192 S_{2}$ : 192 settled $256 S_{3}$ : 0 settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-238.jpg?height=1262&width=1254&top_left_y=1071&top_left_x=421) Figure 41: 8-flower example. Level 0: RGRGRGRGRGRGRGRG Level 1: BGBGBGBGBGBGBGBG Level 2: BBBBBBBBBBBBBBBB Level 3: RBRBRBRBRBRBRBRB Level 4: RGRGRGRGRGRGRGRG Level 5: BRBRBRBRBRBRBRBR Level 6: BBBBBBBBBBBBBBBB Level 7: GBGBGBGBGBGBGBGB Level 8: RRRRRRRRRRRRRRRR $32 \Delta \mathrm{s}: 32$ settled $192 S_{2}$ : 192 settled $256 S_{3}$ : 0 settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-239.jpg?height=1259&width=1250&top_left_y=1072&top_left_x=423) Figure 42: 8-flower example (in black and white for printing). Level 0: RGRGRGRGRGRGRGRGRGRG Level 1: BGBGBGBGBGBGBGBGBGBG Level 2: : $B B B B B B B B B B B B B B B B B B B$ Level 3: RBRBRBRBRBRBRBRBRBRB Level 4: RGRGRGRGRGRGRGRGRGRG Level 5: BRBRBRBRBRBRBRBRBRBR Level 6: $\mathrm{BBBBBBBBBBBBBBBBBBBB}$ Level 7: RBRBRBRBRBRBRBRBRBRB Level 8: RGRGRGRGRGRGRGRGRGRG Level 9: BRBRBRBRBRBRBRBRBRBR Level 10: BBBBBBBBBBBBBBBBBBBB $50 \Delta \mathrm{s}: 50$ settled $290 S_{2}: 290$ settled $220 S_{3}: 120$ settled $110 S_{4}$ : 0 settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-240.jpg?height=1259&width=1282&top_left_y=1072&top_left_x=407) Figure 43: 10-flower example. Level 0: RGRGRGRGRGRGRGRGRGRG Level 1: BGBGBGBGBGBGBGBGBGBG Level 2: : $B B B B B B B B B B B B B B B B B B B$ Level 3: RBRBRBRBRBRBRBRBRBRB Level 4: RGRGRGRGRGRGRGRGRGRG Level 5: BRBRBRBRBRBRBRBRBRBR Level 6: $\mathrm{BBBBBBBBBBBBBBBBBBBB}$ Level 7: RBRBRBRBRBRBRBRBRBRB Level 8: RGRGRGRGRGRGRGRGRGRG Level 9: BRBRBRBRBRBRBRBRBRBR Level 10: BBBBBBBBBBBBBBBBBBBB $50 \Delta \mathrm{s}: 50$ settled $290 S_{2}: 290$ settled $220 S_{3}: 120$ settled $110 S_{4}$ : 0 settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-241.jpg?height=862&width=1248&top_left_y=1468&top_left_x=427) Figure 44: 10-flower example (in black and white for printing). Level 0: RGRGRGRGRGRGRGRGRGRGRGRG Level 1: BGBGBGBGBGBGBGBGBGBGBGBG Level 2: BBBBBBBBBBBBBBBBBBBBBBBB Level 3: RBRBRBRBRBRBRBRBRBRBRBRB Level 4: RRRRRRRRRRRRRRRRRRRRRRRR Level 5: BRBRBRBRBRBRBRBRBRBRBRBR Level 6: BBBBBBBBBBBBBBBBBBBBBBBB Level 7: RBRBRBRBRBRBRBRBRBRBRBRB Level 8: RGRGRGRGRGRGRGRGRGRGRGRG Level 9: BGBGBGBGBGBGBGBGBGBGBGBG Level 10: BBBBBBBBBBBBBBBBBBBBBBBB Level 11: RBRBRBRBRBRBRBRBRBRBRBRB Level 12: RRRRRRRRRRRRRRRRRRRRRRRR 72 settled $\Delta \mathrm{s}$ $600 S_{2}: 600$ settled $184 S_{3}: 76$ settled $24 S_{4}: 0$ settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-242.jpg?height=1265&width=1268&top_left_y=1121&top_left_x=408) Figure 45: 12-flower example. Level 0: RGRGRGRGRGRGRGRGRGRGRGRG Level 1: BGBGBGBGBGBGBGBGBGBGBGBG Level 2: BBBBBBBBBBBBBBBBBBBBBBBB Level 3: RBRBRBRBRBRBRBRBRBRBRBRB Level 4: RRRRRRRRRRRRRRRRRRRRRRRR Level 5: BRBRBRBRBRBRBRBRBRBRBRBR Level 6: BBBBBBBBBBBBBBBBBBBBBBBB Level 7: RBRBRBRBRBRBRBRBRBRBRBRB Level 8: RGRGRGRGRGRGRGRGRGRGRGRG Level 9: BGBGBGBGBGBGBGBGBGBGBGBG Level 10: BBBBBBBBBBBBBBBBBBBBBBBB Level 11: RBRBRBRBRBRBRBRBRBRBRBRB Level 12: RRRRRRRRRRRRRRRRRRRRRRRR 72 settled $\Delta \mathrm{s}$ $600 S_{2}: 600$ settled $184 S_{3}: 76$ settled $24 S_{4}: 0$ settled ![](https://cdn.mathpix.com/cropped/2024_04_28_21af81b42f704f50c439g-243.jpg?height=831&width=1267&top_left_y=1549&top_left_x=423) Figure 46: 12-flower example (in black and white for printing). ## References [1] R.N. Ball, A. Pultr and P. Vojtěchovský, Colored graphs without colorful cycles, to appear in Combinatorica, http://www.math.du.edu/data/preprints/m0506.pdf [2] E. Boros and V. Gurvich, Perfect graphs are kernel-solvable, RUTCOR Research Report, RRR-16-1994, Rutgers University, Discrete Math. 159 (1996), 35-55. [3] E. Boros and V. Gurvich, Stable effectivity functions and perfect graphs, RUTCOR Research Report, RRR-23-1995, Rutgers University, Mathematical Social Sciences 39 (2000), 175-194. [4] E. Boros, V. Gurvich and A. Vasin, Stable families of coalitions and normal hypergraphs, RUTCOR Research Report, RRR-22-1995, Rutgers University, Mathematical Social Sciences 34 (1997), 107-123. [5] E.Boros, V.Gurvich and P.L.Hammer (1993), Dual subimplicants of positive Boolean functions, RUTCOR Research Report, RRR-11-1993, Rutgers University, Optimization Methods and Software 10, 1998, 147-156. [6] A Brandstädt, V.B. Le, and J.P. Spinrad, Graph classes: a survey, SIAM, Philadelphia, 1999. [7] H. Buer and R. Möring, A fast algorithm for decomposition of graphs and posets. Math. Oper. Res. 3 (1983), 170-184. [8] K. Cameron and J. Edmonds, Lambda composition, J. Graph Theory 26 (1997) 9-16. [9] K. Cameron, J. Edmonds, and L. Lovász, A note on perfect graphs, Periodica Math. Hungar. 17(3) (1986), 441-447. [10] F.R.K. Chung and R.L. Graham, Edge-colored complete graphs with precisely colored subgraphs, Combinatorica 3 (1983) 315-324. [11] A. Cournier and M. Habib, A new linear algorithm for modular decomposition, in Proc. of 19th Internat. Coll. on Trees in Algebra and Programming (CAAP-94), Edinburgh, Scotland (S. Tison ed.), LNCS 787, Springer, Berlin (1994), 68-82. [12] T. Eiter, Exact transversal hypergraphs and application to Boolean $\mu$-functions. J. Symbolic Computation 17 (1994) 215-225. [13] T. Eiter, Generating Boolean $\mu$-expressions. Acta Informatica, 32 (1995) 171-187. [14] P. Erdős, M. Simonovits, and V.T. Sos, Anti-Ramsey theorems, Coll. Math. Soc. J. Bolyai 10 (1973), 633-643. [15] S. Foldes and P.L. Hammer, Split graphs; in Proc. of the 8th Southeastern Conference on Combinatorics, Graph Theory, and Computing (Louisiana State Univ., Baton Rouge, La, 1977) Congressus Numerantium XIX (1977), 311-315. [16] T. Gallai, Transitiv orientierbare graphen, Acta Math. Acad. Sci. Hungar., 18 (1-2) (1967), 25-66. English translation by F. Maffray and M. Preissmann, in: Perfect Graphs (Chapter 3) eds. J.L.R Alfonsin and B.A. Reed, John Wiley and Sons. [17] M.C. Golumbic and V. Gurvich, Read-once functions, chapter in: Boolean Functions (Y. Crama and P.L. Hammer, eds.), to appear. [18] V.Gurvich, On repetition-free Boolean functions, Uspechi mat. nauk (Russian Math. Surveys), 32, no 1 (1977), 183-184 (in Russian). [19] V. Gurvich, Applications of Boolean functions and contact schemes in game theory, section 5, Repetition-free Boolean functions and normal forms of positional games (1978), Ph.D. thesis, Moscow Inst. of Physics and Technology, Moscow, USSR (in Russian). [20] V. Gurvich, On the normal form of positional games, Soviet math. dokl., 25, no 3 (1982) $572-575$. [21] V. Gurvich, Some properties and applications of complete edge-chromatic graphs and hypergraphs, Soviet math. dokl., 30, no 3 (1984) 803-807. [22] V. Gurvich, Criteria for repetition-freeness of functions in the Algebra of Logic, Russian Acad. of Sci. Dokl. Math., 43, no 3 ( 1991) 721- 726. [23] V. Gurvich, 1992, Positional game forms and edge-chromatic graphs, Russian Acad. of Sci. Dokl. Math., 45, no 1 ( 1992) 168-172. [24] A. Gyárfás and G. Simonyi, Edge coloring of complete graphs without tricolored triangles, J. Graph Theory 46 (2004), 211-216. [25] M.Karchmer, N.Linial, L. Newman, M. Saks, A. Wigderson, Combinatorial characterization of read-once formulae, Discrete Math. 114 (1993) 275-282. [26] T. Kloks, C.-M. Lee, J. Liu, and H. Müller, On the recognition of general partition graphs, in: Graph-theoretic concepts of computer science, Lecture Notes in Comput. Sci. 2880 (Springer, Berlin, 2003) 273-283. [27] J. Körner and G. Simonyi, Graph pairs and their entropies: modularity problems, Combinatorica 20 (2000) 227-240. [28] J. Körner, G. Simonyi, and Zs. Tuza, Perfect couples of graphs, Combinatorica 12 (1992) 179-192. [29] L. Lovász, Normal hypergraphs and the weak perfect graph conjecture, Discrete Math., 2 (3) (1972), 253-267. [30] L. Lovász A characterization of perfect graphs, J. Combinatorial Theory, (Ser. B) 13 (2) (1972), 95-98. [31] L. Lovász, Combinatorial problems and Exercises, North-Holland Publishing Co., Amsterdam, New York, Oxford, 1979. [32] McAvaney, J. Robertson, and D. DeTemple, A characterization and hereditary properties for partition graphs, Discrete Math. 113 (1-3) (1993) 131-142. [33] R.M. McCollel and J.P. Spinrad, Modular decomposition and transitive orientation, Discrete Math. 201 (1999), 189-241. [34] R. Möring, Algorithmic aspects of the substitution decomposition in optimization over relations, set systems, and Boolean functions, Ann. Oper. Res. 4 (1985/6), 195-225. [35] J. Muller and J. Spinrad, Incremental modular decomposition, J. ACM 36 (1) (1989), $1-19$. [36] Yu.L. Orlovich and I.E. Zverovich, Independent domination and the triangle condition, 12 pp. To appear. [37] G. Ravindra, Strongly perfect line graphs and total graphs, in Finite and infinite sets, 6-th Hungarian Combinatorial Colloquium, vol. 2, Eger, Hungary, 1981, Colloq. Math. Soc. Janos Bolyai 37 (1984) 621-633. [38] I.Zverovich and I.Zverovich, Bipartite hypergraphs: A survey and new results, Discrete Math. 306 (2006) 801-811 [^0]: *) Если у вас нет ткземнляра игрушки, советуем нарисовать развертку куба (чернилами на хорошем листе о́умаги) и. пользуясь карандашом и ластиком, следить за передвижением клеток на развертке. [^1]: *) Автор Приложения - В. Дубровскияи. **) То есть в состоянии, полученном из пра. вильного поворотами плиток (без перекрашивання клеток или разборки кубика). [^2]: ${ }^{1}$ Проблема устойчивости самопересечений путей интересна не только с точки зрения теории графов, но и с точки зрения топологии: она является частным случаем проблемы реализации отображений графов в плоскости [Si69, RS98, Ak00, Sk03]. Настоящий цикл задач основан на статьях [Mi97, Sk03] и пересекается с [RS00, §2, S, глава 7] только по задачам $0, \mathrm{~B} 1$ и D4. ${ }^{2}$ Многие задачи настоящего цикла можно начать решать экспериментально. ${ }^{3}$ Формальное определение приведено на стр. 5 [^3]: ${ }^{4}$ Формальное определение приведено на стр. 5 ${ }^{5}$ Проблема устойчивости самопересечений путей похожа на классическую проблему планарности графов (т.е. реализуемости графов в плоскости без самопересечений) и даже сводится к распознаванию планарности графов (однако число графов, планарность которых надо выяснить, для одного данного пути, велико). Проблема реализуемости графов решается, например, критерием Куратовского. Для проблемы аппроксимируемости вложениями аналогичного критерия не существует [Sk03], см. задачу D8. ${ }^{6}$ Формальное определение подпути приведено на стр. 5 [^4]: ${ }^{1}$ Problem of the stability of intersections of paths interesting not only with the point of view of graph theory but also with the point of view of topology. It is a particular case of problem on realization of mapping graphs in the plane. This sequence of problems is based on the articles [Mi97, Sk03] and overlaps with [RS00, §2, S, chapter 7] only in problems 0 , B1 and D4. ${ }^{2}$ It is possible to solve many problems of this sequence by experiment. To carry out the experiments that involve dogs ask adults for permission. ${ }^{3}$ The formal definition is given in page 5 . [^5]: ${ }^{4}$ The formal definition is given in page 5 . ${ }^{5}$ Problem on the stability of self-intersections of paths is similar to the classical problem on planarity of graphs (i.e. realization of graphs in the plane without self-intersections) and even is reduced to recognition of the planarity of graphs (however the number of graphs required for one path is large). Problem of the realization of graphs is solved, for example, by the Kuratowski criterion. For the problem on approximation by embeddings such a criterion does not exist [Sk03], see problem D8. [^6]: ${ }^{6}$ The formal definition of a subpath is given in page 5 [^7]: ${ }^{\text {a} R U T C O R, ~ R u t g e r s ~ U n i v e r s i t y, ~} 640$ Bartholomew Road, Piscataway NJ 08854-8003; email: dandrade@rutcor.rutgers.edu ${ }^{\text {b} R U T C O R, ~ R u t g e r s ~ U n i v e r s i t y, ~} 640$ Bartholomew Road, Piscataway NJ 08854-8003; email: boros@rutcor.rutgers.edu ${ }^{\text {} R}$ RUTCOR, Rutgers University, 640 Bartholomew Road, Piscataway NJ 08854-8003; email: gurvich@rutcor.rutgers.edu The author is grateful to DIMACS, the NSF Center for Discrete Mathematics and Theoretical Computer Science, for partial support