<?xml version="1.0"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
	<id>https://tcs.nju.edu.cn/wiki/api.php?action=feedcontributions&amp;feedformat=atom&amp;user=172.21.1.240</id>
	<title>TCS Wiki - User contributions [en]</title>
	<link rel="self" type="application/atom+xml" href="https://tcs.nju.edu.cn/wiki/api.php?action=feedcontributions&amp;feedformat=atom&amp;user=172.21.1.240"/>
	<link rel="alternate" type="text/html" href="https://tcs.nju.edu.cn/wiki/index.php?title=Special:Contributions/172.21.1.240"/>
	<updated>2026-05-04T02:57:43Z</updated>
	<subtitle>User contributions</subtitle>
	<generator>MediaWiki 1.45.3</generator>
	<entry>
		<id>https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3518</id>
		<title>Combinatorics (Fall 2010)/Extremal graphs</title>
		<link rel="alternate" type="text/html" href="https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3518"/>
		<updated>2010-10-14T07:14:17Z</updated>

		<summary type="html">&lt;p&gt;172.21.1.240: /* Erdős–Stone theorem */&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;== Extremal Graph Theory ==&lt;br /&gt;
&lt;br /&gt;
=== Mantel&#039;s theorem ===&lt;br /&gt;
We consider a typical extremal problem for graphs: the largest possible number of edges of &#039;&#039;&#039;triangle-free&#039;&#039;&#039; graphs, i.e. graphs contains no &amp;lt;math&amp;gt;K_3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Theorem (Mantel 1907)|&lt;br /&gt;
:Suppose &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; is graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertice without triangles. Then &amp;lt;math&amp;gt;|E|\le\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (pigeonhole principle)|&lt;br /&gt;
We prove an equivalent theorem: Any &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;|E|&amp;gt;\frac{n^2}{4}&amp;lt;/math&amp;gt; must have a triangle.&lt;br /&gt;
&lt;br /&gt;
Use induction on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt;. The theorem holds trivially for &amp;lt;math&amp;gt;n\le 3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Induction hypothesis: assume the theorem hold for &amp;lt;math&amp;gt;|V|\le n-1&amp;lt;/math&amp;gt;. &lt;br /&gt;
&lt;br /&gt;
For &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices, without loss of generality, assume that &amp;lt;math&amp;gt;|E|=\frac{n^2}{4}+1&amp;lt;/math&amp;gt;, we will show that &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must contain a triangle. Take a &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, and let &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; be the subgraph of &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; induced by &amp;lt;math&amp;gt;V\setminus \{u,v\}&amp;lt;/math&amp;gt;. Clearly, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;n-2&amp;lt;/math&amp;gt; vertices.&lt;br /&gt;
:&#039;&#039;&#039;Case.1:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;&amp;gt;\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then by the induction hypothesis, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
:&#039;&#039;&#039;Case.2:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;\le\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then at least &amp;lt;math&amp;gt;\left(\frac{n^2}{4}+1\right)-\frac{(n-2)^2}{4}-1=n-1&amp;lt;/math&amp;gt; edges are between &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;\{u,v\}&amp;lt;/math&amp;gt;. By pigeonhole principle, there must be a vertex in &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; that is adjacent to both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. Thus, &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
}} &lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (Cauchy-Schwarz inequality)|(Mantel&#039;s original proof)&lt;br /&gt;
For any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, no vertex can be a neighbor of both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, or otherwise there will be a triangle. Thus, for any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, &amp;lt;math&amp;gt;d_u+d_v\le n&amp;lt;/math&amp;gt;. It follows that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)\le n|E|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Note that &amp;lt;math&amp;gt;d(v)&amp;lt;/math&amp;gt; appears exactly &amp;lt;math&amp;gt;d_v&amp;lt;/math&amp;gt; times in the sum, so that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)=\sum_{v\in V}d_v^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
Applying Chauchy-Schwarz inequality,&lt;br /&gt;
:&amp;lt;math&amp;gt;&lt;br /&gt;
n|E|\ge\sum_{v\in V}d_v^2\ge\frac{\left(\sum_{v\in V}d_v\right)^2}{n}=\frac{4|E|^2}{n},&lt;br /&gt;
&amp;lt;/math&amp;gt;&lt;br /&gt;
where the last equation is due to Euler&#039;s equality &amp;lt;math&amp;gt;\sum_{v\in V}d_v=2|E|&amp;lt;/math&amp;gt;. The theorem follows.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (inequality of the arithmetic and geometric mean)|&lt;br /&gt;
Assume that &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; vertices and is triangle-free.&lt;br /&gt;
&lt;br /&gt;
Let &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; be the largest independent set in &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\alpha=|A|&amp;lt;/math&amp;gt;. &lt;br /&gt;
Since &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; is triangle-free, for very vertex &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, all its neighbors must form an independent set, thus &amp;lt;math&amp;gt;d(v)\le \alpha&amp;lt;/math&amp;gt; for all &amp;lt;math&amp;gt;v\in V&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Take &amp;lt;math&amp;gt;B=V\setminus A&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\beta=|B|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Since &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; is an independent set, all edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; must have at least one endpoint in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;. Counting the edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; according to their endpoints in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;, we obtain &amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v&amp;lt;/math&amp;gt;. By the inequality of the arithmetic and geometric mean,&lt;br /&gt;
:&amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v\le\alpha\beta\le\left(\frac{\alpha+\beta}{2}\right)^2=\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Turán&#039;s theorem ===&lt;br /&gt;
{{Theorem|Theorem (Turán 1941)|&lt;br /&gt;
:Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a graph with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt;. If &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, &amp;lt;math&amp;gt;k\ge 2&amp;lt;/math&amp;gt;, then&lt;br /&gt;
::&amp;lt;math&amp;gt;|E|\le\frac{r-2}{2(r-1)}n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (induction)|(Turán&#039;s original proof)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (weight shifting)|(due to Motzkin and Straus)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (the probabilistic method)|(due to Alon and Spencer)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Fourth proof.|&lt;br /&gt;
Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices with a maximum number of edges.&lt;br /&gt;
:&#039;&#039;&#039;Claim:&#039;&#039;&#039; &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; does not contain three vertices &amp;lt;math&amp;gt;u,v,w&amp;lt;/math&amp;gt; such that &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt; but &amp;lt;math&amp;gt;uw\not\in E, vw\not\in E&amp;lt;/math&amp;gt;.&lt;br /&gt;
Suppose otherwise. There are two cases.&lt;br /&gt;
* &#039;&#039;&#039;Case.1:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt; or &amp;lt;math&amp;gt;d(w)&amp;lt;d(v)&amp;lt;/math&amp;gt;. Without loss of generality, suppose that &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt;. We duplicate &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; by creating a new vertex &amp;lt;math&amp;gt;u&#039;&amp;lt;/math&amp;gt; which has exactly the same neighbors as &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; (but &amp;lt;math&amp;gt;uu&#039;&amp;lt;/math&amp;gt; is not an edge). Such duplication will not increase the clique size. We then remove &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt;. The resulting graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is still &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free, and has &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices. The number of edges in &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+d(u)-d(w)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;,&lt;br /&gt;
:which contradicts the assumption that &amp;lt;math&amp;gt;|E(G)|&amp;lt;/math&amp;gt; is maximal.&lt;br /&gt;
* &#039;&#039;&#039;Case.2:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)\ge d(u)&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;d(w)\ge d(v)&amp;lt;/math&amp;gt;. Duplicate &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt; twice and delete &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. The new graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, and the number of edges is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+2d(w)-(d(u)+d(v)+1)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;.&lt;br /&gt;
:Contradiction again.&lt;br /&gt;
&lt;br /&gt;
The claim implies that &amp;lt;math&amp;gt;uv\not\in E&amp;lt;/math&amp;gt; defines an equivalence relation on vertices (to be more precise, it guarantees the transitivity of the relation, while the reflexivity and symmetry hold directly). Graph &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must be a complete multipartite graph &amp;lt;math&amp;gt;K_{n_1,n_2,\ldots,n_{r-1}}&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n_1+n_2+\cdots +n_{r-1}=n&amp;lt;/math&amp;gt;. Optimize the edge number, we have the Turán graph.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Erdős–Stone theorem ===&lt;br /&gt;
Let &amp;lt;math&amp;gt;K_s^r=K_{\underbrace{s,s,\cdots,s}_{r}}&amp;lt;/math&amp;gt; be the complete &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-partite graph with &amp;lt;math&amp;gt;s&amp;lt;/math&amp;gt; vertices in each class, i.e., the Turán graph &amp;lt;math&amp;gt;T(rs,r)&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Fundamental theorem of extremal graph theory (Erdős–Stone 1946)|&lt;br /&gt;
:For any integers &amp;lt;math&amp;gt;r\ge 2&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;s\ge 1&amp;lt;/math&amp;gt;, and any &amp;lt;math&amp;gt;\epsilon&amp;gt;0&amp;lt;/math&amp;gt;, if &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; is sufficiently large then every graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices and with at least &amp;lt;math&amp;gt;\left(\frac{r-2}{2(r-1)}+\epsilon\right)n^2&amp;lt;/math&amp;gt; edges contains &amp;lt;math&amp;gt;K_{r,s}&amp;lt;/math&amp;gt; as a subgraph, i.e.,&lt;br /&gt;
:::&amp;lt;math&amp;gt;\mathrm{ex}(n,K_s^r)= \left(\frac{r-2}{2(r-1)}+o(1)\right)n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Corollary|&lt;br /&gt;
:For every nonempty graph &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt;,&lt;br /&gt;
::&amp;lt;math&amp;gt;\lim_{n\rightarrow\infty}\frac{\mathrm{ex}(n,H)}{{n\choose 2}}=\frac{\chi(H)-2}{\chi(H)-1}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
== Cycle Structures ==&lt;br /&gt;
=== Girth ===&lt;br /&gt;
&lt;br /&gt;
=== Hamiltonian cycle ===&lt;/div&gt;</summary>
		<author><name>172.21.1.240</name></author>
	</entry>
	<entry>
		<id>https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3517</id>
		<title>Combinatorics (Fall 2010)/Extremal graphs</title>
		<link rel="alternate" type="text/html" href="https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3517"/>
		<updated>2010-10-14T07:04:04Z</updated>

		<summary type="html">&lt;p&gt;172.21.1.240: /* Erdős–Stone theorem */&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;== Extremal Graph Theory ==&lt;br /&gt;
&lt;br /&gt;
=== Mantel&#039;s theorem ===&lt;br /&gt;
We consider a typical extremal problem for graphs: the largest possible number of edges of &#039;&#039;&#039;triangle-free&#039;&#039;&#039; graphs, i.e. graphs contains no &amp;lt;math&amp;gt;K_3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Theorem (Mantel 1907)|&lt;br /&gt;
:Suppose &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; is graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertice without triangles. Then &amp;lt;math&amp;gt;|E|\le\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (pigeonhole principle)|&lt;br /&gt;
We prove an equivalent theorem: Any &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;|E|&amp;gt;\frac{n^2}{4}&amp;lt;/math&amp;gt; must have a triangle.&lt;br /&gt;
&lt;br /&gt;
Use induction on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt;. The theorem holds trivially for &amp;lt;math&amp;gt;n\le 3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Induction hypothesis: assume the theorem hold for &amp;lt;math&amp;gt;|V|\le n-1&amp;lt;/math&amp;gt;. &lt;br /&gt;
&lt;br /&gt;
For &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices, without loss of generality, assume that &amp;lt;math&amp;gt;|E|=\frac{n^2}{4}+1&amp;lt;/math&amp;gt;, we will show that &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must contain a triangle. Take a &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, and let &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; be the subgraph of &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; induced by &amp;lt;math&amp;gt;V\setminus \{u,v\}&amp;lt;/math&amp;gt;. Clearly, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;n-2&amp;lt;/math&amp;gt; vertices.&lt;br /&gt;
:&#039;&#039;&#039;Case.1:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;&amp;gt;\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then by the induction hypothesis, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
:&#039;&#039;&#039;Case.2:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;\le\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then at least &amp;lt;math&amp;gt;\left(\frac{n^2}{4}+1\right)-\frac{(n-2)^2}{4}-1=n-1&amp;lt;/math&amp;gt; edges are between &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;\{u,v\}&amp;lt;/math&amp;gt;. By pigeonhole principle, there must be a vertex in &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; that is adjacent to both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. Thus, &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
}} &lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (Cauchy-Schwarz inequality)|(Mantel&#039;s original proof)&lt;br /&gt;
For any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, no vertex can be a neighbor of both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, or otherwise there will be a triangle. Thus, for any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, &amp;lt;math&amp;gt;d_u+d_v\le n&amp;lt;/math&amp;gt;. It follows that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)\le n|E|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Note that &amp;lt;math&amp;gt;d(v)&amp;lt;/math&amp;gt; appears exactly &amp;lt;math&amp;gt;d_v&amp;lt;/math&amp;gt; times in the sum, so that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)=\sum_{v\in V}d_v^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
Applying Chauchy-Schwarz inequality,&lt;br /&gt;
:&amp;lt;math&amp;gt;&lt;br /&gt;
n|E|\ge\sum_{v\in V}d_v^2\ge\frac{\left(\sum_{v\in V}d_v\right)^2}{n}=\frac{4|E|^2}{n},&lt;br /&gt;
&amp;lt;/math&amp;gt;&lt;br /&gt;
where the last equation is due to Euler&#039;s equality &amp;lt;math&amp;gt;\sum_{v\in V}d_v=2|E|&amp;lt;/math&amp;gt;. The theorem follows.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (inequality of the arithmetic and geometric mean)|&lt;br /&gt;
Assume that &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; vertices and is triangle-free.&lt;br /&gt;
&lt;br /&gt;
Let &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; be the largest independent set in &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\alpha=|A|&amp;lt;/math&amp;gt;. &lt;br /&gt;
Since &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; is triangle-free, for very vertex &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, all its neighbors must form an independent set, thus &amp;lt;math&amp;gt;d(v)\le \alpha&amp;lt;/math&amp;gt; for all &amp;lt;math&amp;gt;v\in V&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Take &amp;lt;math&amp;gt;B=V\setminus A&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\beta=|B|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Since &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; is an independent set, all edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; must have at least one endpoint in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;. Counting the edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; according to their endpoints in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;, we obtain &amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v&amp;lt;/math&amp;gt;. By the inequality of the arithmetic and geometric mean,&lt;br /&gt;
:&amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v\le\alpha\beta\le\left(\frac{\alpha+\beta}{2}\right)^2=\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Turán&#039;s theorem ===&lt;br /&gt;
{{Theorem|Theorem (Turán 1941)|&lt;br /&gt;
:Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a graph with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt;. If &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, &amp;lt;math&amp;gt;k\ge 2&amp;lt;/math&amp;gt;, then&lt;br /&gt;
::&amp;lt;math&amp;gt;|E|\le\frac{r-2}{2(r-1)}n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (induction)|(Turán&#039;s original proof)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (weight shifting)|(due to Motzkin and Straus)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (the probabilistic method)|(due to Alon and Spencer)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Fourth proof.|&lt;br /&gt;
Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices with a maximum number of edges.&lt;br /&gt;
:&#039;&#039;&#039;Claim:&#039;&#039;&#039; &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; does not contain three vertices &amp;lt;math&amp;gt;u,v,w&amp;lt;/math&amp;gt; such that &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt; but &amp;lt;math&amp;gt;uw\not\in E, vw\not\in E&amp;lt;/math&amp;gt;.&lt;br /&gt;
Suppose otherwise. There are two cases.&lt;br /&gt;
* &#039;&#039;&#039;Case.1:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt; or &amp;lt;math&amp;gt;d(w)&amp;lt;d(v)&amp;lt;/math&amp;gt;. Without loss of generality, suppose that &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt;. We duplicate &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; by creating a new vertex &amp;lt;math&amp;gt;u&#039;&amp;lt;/math&amp;gt; which has exactly the same neighbors as &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; (but &amp;lt;math&amp;gt;uu&#039;&amp;lt;/math&amp;gt; is not an edge). Such duplication will not increase the clique size. We then remove &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt;. The resulting graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is still &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free, and has &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices. The number of edges in &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+d(u)-d(w)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;,&lt;br /&gt;
:which contradicts the assumption that &amp;lt;math&amp;gt;|E(G)|&amp;lt;/math&amp;gt; is maximal.&lt;br /&gt;
* &#039;&#039;&#039;Case.2:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)\ge d(u)&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;d(w)\ge d(v)&amp;lt;/math&amp;gt;. Duplicate &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt; twice and delete &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. The new graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, and the number of edges is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+2d(w)-(d(u)+d(v)+1)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;.&lt;br /&gt;
:Contradiction again.&lt;br /&gt;
&lt;br /&gt;
The claim implies that &amp;lt;math&amp;gt;uv\not\in E&amp;lt;/math&amp;gt; defines an equivalence relation on vertices (to be more precise, it guarantees the transitivity of the relation, while the reflexivity and symmetry hold directly). Graph &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must be a complete multipartite graph &amp;lt;math&amp;gt;K_{n_1,n_2,\ldots,n_{r-1}}&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n_1+n_2+\cdots +n_{r-1}=n&amp;lt;/math&amp;gt;. Optimize the edge number, we have the Turán graph.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Erdős–Stone theorem ===&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Fundamental theorem of extremal graph theory (Erdős–Stone 1946)|&lt;br /&gt;
:For any integers &amp;lt;math&amp;gt;r\ge 2&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;s\ge 1&amp;lt;/math&amp;gt;, and any &amp;lt;math&amp;gt;\epsilon&amp;gt;0&amp;lt;/math&amp;gt;, if &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; is sufficiently large then every graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices and with at least &amp;lt;math&amp;gt;\left(\frac{r-2}{2(r-1)}+\epsilon\right)n^2&amp;lt;/math&amp;gt; edges contains &amp;lt;math&amp;gt;K_{r,s}&amp;lt;/math&amp;gt; as a subgraph, i.e.,&lt;br /&gt;
:::&amp;lt;math&amp;gt;\mathrm{ex}(n,K_{r,s})= \left(\frac{r-2}{2(r-1)}+o(1)\right)n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Corollary|&lt;br /&gt;
:For every nonempty graph &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt;,&lt;br /&gt;
::&amp;lt;math&amp;gt;\lim_{n\rightarrow\infty}\frac{\mathrm{ex}(n,H)}{{n\choose 2}}=\frac{\chi(H)-2}{\chi(H)-1}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
== Cycle Structures ==&lt;br /&gt;
=== Girth ===&lt;br /&gt;
&lt;br /&gt;
=== Hamiltonian cycle ===&lt;/div&gt;</summary>
		<author><name>172.21.1.240</name></author>
	</entry>
	<entry>
		<id>https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3516</id>
		<title>Combinatorics (Fall 2010)/Extremal graphs</title>
		<link rel="alternate" type="text/html" href="https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3516"/>
		<updated>2010-10-14T07:00:05Z</updated>

		<summary type="html">&lt;p&gt;172.21.1.240: /* Erdős–Stone theorem */&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;== Extremal Graph Theory ==&lt;br /&gt;
&lt;br /&gt;
=== Mantel&#039;s theorem ===&lt;br /&gt;
We consider a typical extremal problem for graphs: the largest possible number of edges of &#039;&#039;&#039;triangle-free&#039;&#039;&#039; graphs, i.e. graphs contains no &amp;lt;math&amp;gt;K_3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Theorem (Mantel 1907)|&lt;br /&gt;
:Suppose &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; is graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertice without triangles. Then &amp;lt;math&amp;gt;|E|\le\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (pigeonhole principle)|&lt;br /&gt;
We prove an equivalent theorem: Any &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;|E|&amp;gt;\frac{n^2}{4}&amp;lt;/math&amp;gt; must have a triangle.&lt;br /&gt;
&lt;br /&gt;
Use induction on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt;. The theorem holds trivially for &amp;lt;math&amp;gt;n\le 3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Induction hypothesis: assume the theorem hold for &amp;lt;math&amp;gt;|V|\le n-1&amp;lt;/math&amp;gt;. &lt;br /&gt;
&lt;br /&gt;
For &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices, without loss of generality, assume that &amp;lt;math&amp;gt;|E|=\frac{n^2}{4}+1&amp;lt;/math&amp;gt;, we will show that &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must contain a triangle. Take a &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, and let &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; be the subgraph of &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; induced by &amp;lt;math&amp;gt;V\setminus \{u,v\}&amp;lt;/math&amp;gt;. Clearly, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;n-2&amp;lt;/math&amp;gt; vertices.&lt;br /&gt;
:&#039;&#039;&#039;Case.1:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;&amp;gt;\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then by the induction hypothesis, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
:&#039;&#039;&#039;Case.2:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;\le\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then at least &amp;lt;math&amp;gt;\left(\frac{n^2}{4}+1\right)-\frac{(n-2)^2}{4}-1=n-1&amp;lt;/math&amp;gt; edges are between &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;\{u,v\}&amp;lt;/math&amp;gt;. By pigeonhole principle, there must be a vertex in &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; that is adjacent to both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. Thus, &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
}} &lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (Cauchy-Schwarz inequality)|(Mantel&#039;s original proof)&lt;br /&gt;
For any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, no vertex can be a neighbor of both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, or otherwise there will be a triangle. Thus, for any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, &amp;lt;math&amp;gt;d_u+d_v\le n&amp;lt;/math&amp;gt;. It follows that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)\le n|E|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Note that &amp;lt;math&amp;gt;d(v)&amp;lt;/math&amp;gt; appears exactly &amp;lt;math&amp;gt;d_v&amp;lt;/math&amp;gt; times in the sum, so that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)=\sum_{v\in V}d_v^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
Applying Chauchy-Schwarz inequality,&lt;br /&gt;
:&amp;lt;math&amp;gt;&lt;br /&gt;
n|E|\ge\sum_{v\in V}d_v^2\ge\frac{\left(\sum_{v\in V}d_v\right)^2}{n}=\frac{4|E|^2}{n},&lt;br /&gt;
&amp;lt;/math&amp;gt;&lt;br /&gt;
where the last equation is due to Euler&#039;s equality &amp;lt;math&amp;gt;\sum_{v\in V}d_v=2|E|&amp;lt;/math&amp;gt;. The theorem follows.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (inequality of the arithmetic and geometric mean)|&lt;br /&gt;
Assume that &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; vertices and is triangle-free.&lt;br /&gt;
&lt;br /&gt;
Let &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; be the largest independent set in &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\alpha=|A|&amp;lt;/math&amp;gt;. &lt;br /&gt;
Since &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; is triangle-free, for very vertex &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, all its neighbors must form an independent set, thus &amp;lt;math&amp;gt;d(v)\le \alpha&amp;lt;/math&amp;gt; for all &amp;lt;math&amp;gt;v\in V&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Take &amp;lt;math&amp;gt;B=V\setminus A&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\beta=|B|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Since &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; is an independent set, all edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; must have at least one endpoint in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;. Counting the edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; according to their endpoints in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;, we obtain &amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v&amp;lt;/math&amp;gt;. By the inequality of the arithmetic and geometric mean,&lt;br /&gt;
:&amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v\le\alpha\beta\le\left(\frac{\alpha+\beta}{2}\right)^2=\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Turán&#039;s theorem ===&lt;br /&gt;
{{Theorem|Theorem (Turán 1941)|&lt;br /&gt;
:Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a graph with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt;. If &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, &amp;lt;math&amp;gt;k\ge 2&amp;lt;/math&amp;gt;, then&lt;br /&gt;
::&amp;lt;math&amp;gt;|E|\le\frac{r-2}{2(r-1)}n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (induction)|(Turán&#039;s original proof)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (weight shifting)|(due to Motzkin and Straus)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (the probabilistic method)|(due to Alon and Spencer)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Fourth proof.|&lt;br /&gt;
Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices with a maximum number of edges.&lt;br /&gt;
:&#039;&#039;&#039;Claim:&#039;&#039;&#039; &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; does not contain three vertices &amp;lt;math&amp;gt;u,v,w&amp;lt;/math&amp;gt; such that &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt; but &amp;lt;math&amp;gt;uw\not\in E, vw\not\in E&amp;lt;/math&amp;gt;.&lt;br /&gt;
Suppose otherwise. There are two cases.&lt;br /&gt;
* &#039;&#039;&#039;Case.1:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt; or &amp;lt;math&amp;gt;d(w)&amp;lt;d(v)&amp;lt;/math&amp;gt;. Without loss of generality, suppose that &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt;. We duplicate &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; by creating a new vertex &amp;lt;math&amp;gt;u&#039;&amp;lt;/math&amp;gt; which has exactly the same neighbors as &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; (but &amp;lt;math&amp;gt;uu&#039;&amp;lt;/math&amp;gt; is not an edge). Such duplication will not increase the clique size. We then remove &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt;. The resulting graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is still &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free, and has &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices. The number of edges in &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+d(u)-d(w)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;,&lt;br /&gt;
:which contradicts the assumption that &amp;lt;math&amp;gt;|E(G)|&amp;lt;/math&amp;gt; is maximal.&lt;br /&gt;
* &#039;&#039;&#039;Case.2:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)\ge d(u)&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;d(w)\ge d(v)&amp;lt;/math&amp;gt;. Duplicate &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt; twice and delete &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. The new graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, and the number of edges is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+2d(w)-(d(u)+d(v)+1)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;.&lt;br /&gt;
:Contradiction again.&lt;br /&gt;
&lt;br /&gt;
The claim implies that &amp;lt;math&amp;gt;uv\not\in E&amp;lt;/math&amp;gt; defines an equivalence relation on vertices (to be more precise, it guarantees the transitivity of the relation, while the reflexivity and symmetry hold directly). Graph &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must be a complete multipartite graph &amp;lt;math&amp;gt;K_{n_1,n_2,\ldots,n_{r-1}}&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n_1+n_2+\cdots +n_{r-1}=n&amp;lt;/math&amp;gt;. Optimize the edge number, we have the Turán graph.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Erdős–Stone theorem ===&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Fundamental theorem of extremal graph theory (Erdős–Stone 1946)|&lt;br /&gt;
:For any integers &amp;lt;math&amp;gt;r\ge 2&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;s\ge 1&amp;lt;/math&amp;gt;, and any &amp;lt;math&amp;gt;\epsilon&amp;gt;0&amp;lt;/math&amp;gt;, there exists an &amp;lt;math&amp;gt;N_0&amp;lt;/math&amp;gt; such that every graph with &amp;lt;math&amp;gt;n\ge N_0&amp;lt;/math&amp;gt; vertices and at least &amp;lt;math&amp;gt;\left(\frac{r-2}{2(r-1)}+\epsilon\right)n^2&amp;lt;/math&amp;gt; edges contains &amp;lt;math&amp;gt;K_{r,s}&amp;lt;/math&amp;gt; as a subgraph, i.e.,&lt;br /&gt;
:::&amp;lt;math&amp;gt;\mathrm{ex}(n,K_{r,s})= \left(\frac{r-2}{2(r-1)}+o(1)\right)n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Corollary|&lt;br /&gt;
:For every nonempty graph &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt;,&lt;br /&gt;
::&amp;lt;math&amp;gt;\lim_{n\rightarrow\infty}\frac{\mathrm{ex}(n,H)}{{n\choose 2}}=\frac{\chi(H)-2}{\chi(H)-1}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
== Cycle Structures ==&lt;br /&gt;
=== Girth ===&lt;br /&gt;
&lt;br /&gt;
=== Hamiltonian cycle ===&lt;/div&gt;</summary>
		<author><name>172.21.1.240</name></author>
	</entry>
	<entry>
		<id>https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3515</id>
		<title>Combinatorics (Fall 2010)/Extremal graphs</title>
		<link rel="alternate" type="text/html" href="https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3515"/>
		<updated>2010-10-14T06:59:33Z</updated>

		<summary type="html">&lt;p&gt;172.21.1.240: /* Erdős–Stone theorem */&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;== Extremal Graph Theory ==&lt;br /&gt;
&lt;br /&gt;
=== Mantel&#039;s theorem ===&lt;br /&gt;
We consider a typical extremal problem for graphs: the largest possible number of edges of &#039;&#039;&#039;triangle-free&#039;&#039;&#039; graphs, i.e. graphs contains no &amp;lt;math&amp;gt;K_3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Theorem (Mantel 1907)|&lt;br /&gt;
:Suppose &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; is graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertice without triangles. Then &amp;lt;math&amp;gt;|E|\le\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (pigeonhole principle)|&lt;br /&gt;
We prove an equivalent theorem: Any &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;|E|&amp;gt;\frac{n^2}{4}&amp;lt;/math&amp;gt; must have a triangle.&lt;br /&gt;
&lt;br /&gt;
Use induction on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt;. The theorem holds trivially for &amp;lt;math&amp;gt;n\le 3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Induction hypothesis: assume the theorem hold for &amp;lt;math&amp;gt;|V|\le n-1&amp;lt;/math&amp;gt;. &lt;br /&gt;
&lt;br /&gt;
For &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices, without loss of generality, assume that &amp;lt;math&amp;gt;|E|=\frac{n^2}{4}+1&amp;lt;/math&amp;gt;, we will show that &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must contain a triangle. Take a &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, and let &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; be the subgraph of &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; induced by &amp;lt;math&amp;gt;V\setminus \{u,v\}&amp;lt;/math&amp;gt;. Clearly, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;n-2&amp;lt;/math&amp;gt; vertices.&lt;br /&gt;
:&#039;&#039;&#039;Case.1:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;&amp;gt;\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then by the induction hypothesis, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
:&#039;&#039;&#039;Case.2:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;\le\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then at least &amp;lt;math&amp;gt;\left(\frac{n^2}{4}+1\right)-\frac{(n-2)^2}{4}-1=n-1&amp;lt;/math&amp;gt; edges are between &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;\{u,v\}&amp;lt;/math&amp;gt;. By pigeonhole principle, there must be a vertex in &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; that is adjacent to both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. Thus, &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
}} &lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (Cauchy-Schwarz inequality)|(Mantel&#039;s original proof)&lt;br /&gt;
For any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, no vertex can be a neighbor of both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, or otherwise there will be a triangle. Thus, for any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, &amp;lt;math&amp;gt;d_u+d_v\le n&amp;lt;/math&amp;gt;. It follows that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)\le n|E|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Note that &amp;lt;math&amp;gt;d(v)&amp;lt;/math&amp;gt; appears exactly &amp;lt;math&amp;gt;d_v&amp;lt;/math&amp;gt; times in the sum, so that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)=\sum_{v\in V}d_v^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
Applying Chauchy-Schwarz inequality,&lt;br /&gt;
:&amp;lt;math&amp;gt;&lt;br /&gt;
n|E|\ge\sum_{v\in V}d_v^2\ge\frac{\left(\sum_{v\in V}d_v\right)^2}{n}=\frac{4|E|^2}{n},&lt;br /&gt;
&amp;lt;/math&amp;gt;&lt;br /&gt;
where the last equation is due to Euler&#039;s equality &amp;lt;math&amp;gt;\sum_{v\in V}d_v=2|E|&amp;lt;/math&amp;gt;. The theorem follows.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (inequality of the arithmetic and geometric mean)|&lt;br /&gt;
Assume that &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; vertices and is triangle-free.&lt;br /&gt;
&lt;br /&gt;
Let &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; be the largest independent set in &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\alpha=|A|&amp;lt;/math&amp;gt;. &lt;br /&gt;
Since &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; is triangle-free, for very vertex &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, all its neighbors must form an independent set, thus &amp;lt;math&amp;gt;d(v)\le \alpha&amp;lt;/math&amp;gt; for all &amp;lt;math&amp;gt;v\in V&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Take &amp;lt;math&amp;gt;B=V\setminus A&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\beta=|B|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Since &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; is an independent set, all edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; must have at least one endpoint in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;. Counting the edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; according to their endpoints in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;, we obtain &amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v&amp;lt;/math&amp;gt;. By the inequality of the arithmetic and geometric mean,&lt;br /&gt;
:&amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v\le\alpha\beta\le\left(\frac{\alpha+\beta}{2}\right)^2=\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Turán&#039;s theorem ===&lt;br /&gt;
{{Theorem|Theorem (Turán 1941)|&lt;br /&gt;
:Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a graph with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt;. If &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, &amp;lt;math&amp;gt;k\ge 2&amp;lt;/math&amp;gt;, then&lt;br /&gt;
::&amp;lt;math&amp;gt;|E|\le\frac{r-2}{2(r-1)}n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (induction)|(Turán&#039;s original proof)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (weight shifting)|(due to Motzkin and Straus)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (the probabilistic method)|(due to Alon and Spencer)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Fourth proof.|&lt;br /&gt;
Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices with a maximum number of edges.&lt;br /&gt;
:&#039;&#039;&#039;Claim:&#039;&#039;&#039; &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; does not contain three vertices &amp;lt;math&amp;gt;u,v,w&amp;lt;/math&amp;gt; such that &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt; but &amp;lt;math&amp;gt;uw\not\in E, vw\not\in E&amp;lt;/math&amp;gt;.&lt;br /&gt;
Suppose otherwise. There are two cases.&lt;br /&gt;
* &#039;&#039;&#039;Case.1:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt; or &amp;lt;math&amp;gt;d(w)&amp;lt;d(v)&amp;lt;/math&amp;gt;. Without loss of generality, suppose that &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt;. We duplicate &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; by creating a new vertex &amp;lt;math&amp;gt;u&#039;&amp;lt;/math&amp;gt; which has exactly the same neighbors as &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; (but &amp;lt;math&amp;gt;uu&#039;&amp;lt;/math&amp;gt; is not an edge). Such duplication will not increase the clique size. We then remove &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt;. The resulting graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is still &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free, and has &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices. The number of edges in &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+d(u)-d(w)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;,&lt;br /&gt;
:which contradicts the assumption that &amp;lt;math&amp;gt;|E(G)|&amp;lt;/math&amp;gt; is maximal.&lt;br /&gt;
* &#039;&#039;&#039;Case.2:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)\ge d(u)&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;d(w)\ge d(v)&amp;lt;/math&amp;gt;. Duplicate &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt; twice and delete &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. The new graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, and the number of edges is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+2d(w)-(d(u)+d(v)+1)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;.&lt;br /&gt;
:Contradiction again.&lt;br /&gt;
&lt;br /&gt;
The claim implies that &amp;lt;math&amp;gt;uv\not\in E&amp;lt;/math&amp;gt; defines an equivalence relation on vertices (to be more precise, it guarantees the transitivity of the relation, while the reflexivity and symmetry hold directly). Graph &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must be a complete multipartite graph &amp;lt;math&amp;gt;K_{n_1,n_2,\ldots,n_{r-1}}&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n_1+n_2+\cdots +n_{r-1}=n&amp;lt;/math&amp;gt;. Optimize the edge number, we have the Turán graph.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Erdős–Stone theorem ===&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Fundamental theorem of extremal graph theory (Erdős–Stone 1946)|&lt;br /&gt;
:For any integers &amp;lt;math&amp;gt;r\ge 2&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;s\ge 1&amp;lt;/math&amp;gt;, and any &amp;lt;math&amp;gt;\epsilon&amp;gt;0&amp;lt;/math&amp;gt;, there exists an &amp;lt;math&amp;gt;N_0&amp;lt;/math&amp;gt; such that every graph with &amp;lt;math&amp;gt;n\ge N_0&amp;lt;/math&amp;gt; vertices and at least &amp;lt;math&amp;gt;\left(\frac{r-2}{2(r-1)}+\epsilon\right)n^2&amp;lt;/math&amp;gt; edges contains &amp;lt;math&amp;gt;K_{r,s}&amp;lt;/math&amp;gt; as a subgraph, i.e.,&lt;br /&gt;
:::&amp;lt;math&amp;gt;\mathrm{ex}(n,K_{s,r})= \left(\frac{r-2}{2(r-1)}+o(1)\right)n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Corollary|&lt;br /&gt;
:For every nonempty graph &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt;,&lt;br /&gt;
::&amp;lt;math&amp;gt;\lim_{n\rightarrow\infty}\frac{\mathrm{ex}(n,H)}{{n\choose 2}}=\frac{\chi(H)-2}{\chi(H)-1}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
== Cycle Structures ==&lt;br /&gt;
=== Girth ===&lt;br /&gt;
&lt;br /&gt;
=== Hamiltonian cycle ===&lt;/div&gt;</summary>
		<author><name>172.21.1.240</name></author>
	</entry>
	<entry>
		<id>https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3514</id>
		<title>Combinatorics (Fall 2010)/Extremal graphs</title>
		<link rel="alternate" type="text/html" href="https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3514"/>
		<updated>2010-10-14T06:58:37Z</updated>

		<summary type="html">&lt;p&gt;172.21.1.240: /* Erdős–Stone theorem */&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;== Extremal Graph Theory ==&lt;br /&gt;
&lt;br /&gt;
=== Mantel&#039;s theorem ===&lt;br /&gt;
We consider a typical extremal problem for graphs: the largest possible number of edges of &#039;&#039;&#039;triangle-free&#039;&#039;&#039; graphs, i.e. graphs contains no &amp;lt;math&amp;gt;K_3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Theorem (Mantel 1907)|&lt;br /&gt;
:Suppose &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; is graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertice without triangles. Then &amp;lt;math&amp;gt;|E|\le\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (pigeonhole principle)|&lt;br /&gt;
We prove an equivalent theorem: Any &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;|E|&amp;gt;\frac{n^2}{4}&amp;lt;/math&amp;gt; must have a triangle.&lt;br /&gt;
&lt;br /&gt;
Use induction on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt;. The theorem holds trivially for &amp;lt;math&amp;gt;n\le 3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Induction hypothesis: assume the theorem hold for &amp;lt;math&amp;gt;|V|\le n-1&amp;lt;/math&amp;gt;. &lt;br /&gt;
&lt;br /&gt;
For &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices, without loss of generality, assume that &amp;lt;math&amp;gt;|E|=\frac{n^2}{4}+1&amp;lt;/math&amp;gt;, we will show that &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must contain a triangle. Take a &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, and let &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; be the subgraph of &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; induced by &amp;lt;math&amp;gt;V\setminus \{u,v\}&amp;lt;/math&amp;gt;. Clearly, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;n-2&amp;lt;/math&amp;gt; vertices.&lt;br /&gt;
:&#039;&#039;&#039;Case.1:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;&amp;gt;\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then by the induction hypothesis, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
:&#039;&#039;&#039;Case.2:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;\le\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then at least &amp;lt;math&amp;gt;\left(\frac{n^2}{4}+1\right)-\frac{(n-2)^2}{4}-1=n-1&amp;lt;/math&amp;gt; edges are between &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;\{u,v\}&amp;lt;/math&amp;gt;. By pigeonhole principle, there must be a vertex in &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; that is adjacent to both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. Thus, &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
}} &lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (Cauchy-Schwarz inequality)|(Mantel&#039;s original proof)&lt;br /&gt;
For any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, no vertex can be a neighbor of both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, or otherwise there will be a triangle. Thus, for any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, &amp;lt;math&amp;gt;d_u+d_v\le n&amp;lt;/math&amp;gt;. It follows that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)\le n|E|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Note that &amp;lt;math&amp;gt;d(v)&amp;lt;/math&amp;gt; appears exactly &amp;lt;math&amp;gt;d_v&amp;lt;/math&amp;gt; times in the sum, so that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)=\sum_{v\in V}d_v^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
Applying Chauchy-Schwarz inequality,&lt;br /&gt;
:&amp;lt;math&amp;gt;&lt;br /&gt;
n|E|\ge\sum_{v\in V}d_v^2\ge\frac{\left(\sum_{v\in V}d_v\right)^2}{n}=\frac{4|E|^2}{n},&lt;br /&gt;
&amp;lt;/math&amp;gt;&lt;br /&gt;
where the last equation is due to Euler&#039;s equality &amp;lt;math&amp;gt;\sum_{v\in V}d_v=2|E|&amp;lt;/math&amp;gt;. The theorem follows.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (inequality of the arithmetic and geometric mean)|&lt;br /&gt;
Assume that &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; vertices and is triangle-free.&lt;br /&gt;
&lt;br /&gt;
Let &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; be the largest independent set in &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\alpha=|A|&amp;lt;/math&amp;gt;. &lt;br /&gt;
Since &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; is triangle-free, for very vertex &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, all its neighbors must form an independent set, thus &amp;lt;math&amp;gt;d(v)\le \alpha&amp;lt;/math&amp;gt; for all &amp;lt;math&amp;gt;v\in V&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Take &amp;lt;math&amp;gt;B=V\setminus A&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\beta=|B|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Since &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; is an independent set, all edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; must have at least one endpoint in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;. Counting the edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; according to their endpoints in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;, we obtain &amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v&amp;lt;/math&amp;gt;. By the inequality of the arithmetic and geometric mean,&lt;br /&gt;
:&amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v\le\alpha\beta\le\left(\frac{\alpha+\beta}{2}\right)^2=\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Turán&#039;s theorem ===&lt;br /&gt;
{{Theorem|Theorem (Turán 1941)|&lt;br /&gt;
:Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a graph with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt;. If &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, &amp;lt;math&amp;gt;k\ge 2&amp;lt;/math&amp;gt;, then&lt;br /&gt;
::&amp;lt;math&amp;gt;|E|\le\frac{r-2}{2(r-1)}n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (induction)|(Turán&#039;s original proof)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (weight shifting)|(due to Motzkin and Straus)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (the probabilistic method)|(due to Alon and Spencer)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Fourth proof.|&lt;br /&gt;
Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices with a maximum number of edges.&lt;br /&gt;
:&#039;&#039;&#039;Claim:&#039;&#039;&#039; &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; does not contain three vertices &amp;lt;math&amp;gt;u,v,w&amp;lt;/math&amp;gt; such that &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt; but &amp;lt;math&amp;gt;uw\not\in E, vw\not\in E&amp;lt;/math&amp;gt;.&lt;br /&gt;
Suppose otherwise. There are two cases.&lt;br /&gt;
* &#039;&#039;&#039;Case.1:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt; or &amp;lt;math&amp;gt;d(w)&amp;lt;d(v)&amp;lt;/math&amp;gt;. Without loss of generality, suppose that &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt;. We duplicate &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; by creating a new vertex &amp;lt;math&amp;gt;u&#039;&amp;lt;/math&amp;gt; which has exactly the same neighbors as &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; (but &amp;lt;math&amp;gt;uu&#039;&amp;lt;/math&amp;gt; is not an edge). Such duplication will not increase the clique size. We then remove &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt;. The resulting graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is still &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free, and has &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices. The number of edges in &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+d(u)-d(w)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;,&lt;br /&gt;
:which contradicts the assumption that &amp;lt;math&amp;gt;|E(G)|&amp;lt;/math&amp;gt; is maximal.&lt;br /&gt;
* &#039;&#039;&#039;Case.2:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)\ge d(u)&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;d(w)\ge d(v)&amp;lt;/math&amp;gt;. Duplicate &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt; twice and delete &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. The new graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, and the number of edges is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+2d(w)-(d(u)+d(v)+1)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;.&lt;br /&gt;
:Contradiction again.&lt;br /&gt;
&lt;br /&gt;
The claim implies that &amp;lt;math&amp;gt;uv\not\in E&amp;lt;/math&amp;gt; defines an equivalence relation on vertices (to be more precise, it guarantees the transitivity of the relation, while the reflexivity and symmetry hold directly). Graph &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must be a complete multipartite graph &amp;lt;math&amp;gt;K_{n_1,n_2,\ldots,n_{r-1}}&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n_1+n_2+\cdots +n_{r-1}=n&amp;lt;/math&amp;gt;. Optimize the edge number, we have the Turán graph.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Erdős–Stone theorem ===&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Fundamental theorem of extremal graph theory (Erdős–Stone 1946)|&lt;br /&gt;
:For any integers &amp;lt;math&amp;gt;r\ge 2&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;s\ge 1&amp;lt;/math&amp;gt;, and any &amp;lt;math&amp;gt;\epsilon&amp;gt;0&amp;lt;/math&amp;gt;, there exists an &amp;lt;math&amp;gt;N_0&amp;lt;/math&amp;gt; such that every graph with &amp;lt;math&amp;gt;n\ge N_0&amp;lt;/math&amp;gt; vertices and at least &amp;lt;math&amp;gt;\left(\frac{r-2}{2(r-1)}+\epsilon\right)n^2&amp;lt;/math&amp;gt; edges contains &amp;lt;math&amp;gt;K_{r,s}&amp;lt;/math&amp;gt; as a subgraph, i.e.,&lt;br /&gt;
:::&amp;lt;math&amp;gt;\mathrm{ex}(n,K_{s,r})\le \left(\frac{r-2}{2(r-1)}+o(1)\right)n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Corollary|&lt;br /&gt;
:For every nonempty graph &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt;,&lt;br /&gt;
::&amp;lt;math&amp;gt;\lim_{n\rightarrow\infty}\frac{\mathrm{ex}(n,H)}{{n\choose 2}}=\frac{\chi(H)-2}{\chi(H)-1}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
== Cycle Structures ==&lt;br /&gt;
=== Girth ===&lt;br /&gt;
&lt;br /&gt;
=== Hamiltonian cycle ===&lt;/div&gt;</summary>
		<author><name>172.21.1.240</name></author>
	</entry>
	<entry>
		<id>https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3513</id>
		<title>Combinatorics (Fall 2010)/Extremal graphs</title>
		<link rel="alternate" type="text/html" href="https://tcs.nju.edu.cn/wiki/index.php?title=Combinatorics_(Fall_2010)/Extremal_graphs&amp;diff=3513"/>
		<updated>2010-10-14T06:50:52Z</updated>

		<summary type="html">&lt;p&gt;172.21.1.240: /* Erdős–Stone theorem */&lt;/p&gt;
&lt;hr /&gt;
&lt;div&gt;== Extremal Graph Theory ==&lt;br /&gt;
&lt;br /&gt;
=== Mantel&#039;s theorem ===&lt;br /&gt;
We consider a typical extremal problem for graphs: the largest possible number of edges of &#039;&#039;&#039;triangle-free&#039;&#039;&#039; graphs, i.e. graphs contains no &amp;lt;math&amp;gt;K_3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Theorem (Mantel 1907)|&lt;br /&gt;
:Suppose &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; is graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertice without triangles. Then &amp;lt;math&amp;gt;|E|\le\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (pigeonhole principle)|&lt;br /&gt;
We prove an equivalent theorem: Any &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;|E|&amp;gt;\frac{n^2}{4}&amp;lt;/math&amp;gt; must have a triangle.&lt;br /&gt;
&lt;br /&gt;
Use induction on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt;. The theorem holds trivially for &amp;lt;math&amp;gt;n\le 3&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Induction hypothesis: assume the theorem hold for &amp;lt;math&amp;gt;|V|\le n-1&amp;lt;/math&amp;gt;. &lt;br /&gt;
&lt;br /&gt;
For &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices, without loss of generality, assume that &amp;lt;math&amp;gt;|E|=\frac{n^2}{4}+1&amp;lt;/math&amp;gt;, we will show that &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must contain a triangle. Take a &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, and let &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; be the subgraph of &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; induced by &amp;lt;math&amp;gt;V\setminus \{u,v\}&amp;lt;/math&amp;gt;. Clearly, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;n-2&amp;lt;/math&amp;gt; vertices.&lt;br /&gt;
:&#039;&#039;&#039;Case.1:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;&amp;gt;\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then by the induction hypothesis, &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
:&#039;&#039;&#039;Case.2:&#039;&#039;&#039; If &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;\le\frac{(n-2)^2}{4}&amp;lt;/math&amp;gt; edges, then at least &amp;lt;math&amp;gt;\left(\frac{n^2}{4}+1\right)-\frac{(n-2)^2}{4}-1=n-1&amp;lt;/math&amp;gt; edges are between &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;\{u,v\}&amp;lt;/math&amp;gt;. By pigeonhole principle, there must be a vertex in &amp;lt;math&amp;gt;H&amp;lt;/math&amp;gt; that is adjacent to both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. Thus, &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has a triangle.&lt;br /&gt;
}} &lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (Cauchy-Schwarz inequality)|(Mantel&#039;s original proof)&lt;br /&gt;
For any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, no vertex can be a neighbor of both &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, or otherwise there will be a triangle. Thus, for any edge &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt;, &amp;lt;math&amp;gt;d_u+d_v\le n&amp;lt;/math&amp;gt;. It follows that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)\le n|E|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Note that &amp;lt;math&amp;gt;d(v)&amp;lt;/math&amp;gt; appears exactly &amp;lt;math&amp;gt;d_v&amp;lt;/math&amp;gt; times in the sum, so that&lt;br /&gt;
:&amp;lt;math&amp;gt;\sum_{uv\in E}(d_u+d_v)=\sum_{v\in V}d_v^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
Applying Chauchy-Schwarz inequality,&lt;br /&gt;
:&amp;lt;math&amp;gt;&lt;br /&gt;
n|E|\ge\sum_{v\in V}d_v^2\ge\frac{\left(\sum_{v\in V}d_v\right)^2}{n}=\frac{4|E|^2}{n},&lt;br /&gt;
&amp;lt;/math&amp;gt;&lt;br /&gt;
where the last equation is due to Euler&#039;s equality &amp;lt;math&amp;gt;\sum_{v\in V}d_v=2|E|&amp;lt;/math&amp;gt;. The theorem follows.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (inequality of the arithmetic and geometric mean)|&lt;br /&gt;
Assume that &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; has &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt; vertices and is triangle-free.&lt;br /&gt;
&lt;br /&gt;
Let &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; be the largest independent set in &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\alpha=|A|&amp;lt;/math&amp;gt;. &lt;br /&gt;
Since &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; is triangle-free, for very vertex &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;, all its neighbors must form an independent set, thus &amp;lt;math&amp;gt;d(v)\le \alpha&amp;lt;/math&amp;gt; for all &amp;lt;math&amp;gt;v\in V&amp;lt;/math&amp;gt;.&lt;br /&gt;
&lt;br /&gt;
Take &amp;lt;math&amp;gt;B=V\setminus A&amp;lt;/math&amp;gt; and let &amp;lt;math&amp;gt;\beta=|B|&amp;lt;/math&amp;gt;.&lt;br /&gt;
Since &amp;lt;math&amp;gt;A&amp;lt;/math&amp;gt; is an independent set, all edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; must have at least one endpoint in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;. Counting the edges in &amp;lt;math&amp;gt;E&amp;lt;/math&amp;gt; according to their endpoints in &amp;lt;math&amp;gt;B&amp;lt;/math&amp;gt;, we obtain &amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v&amp;lt;/math&amp;gt;. By the inequality of the arithmetic and geometric mean,&lt;br /&gt;
:&amp;lt;math&amp;gt;|E|\le\sum_{v\in B}d_v\le\alpha\beta\le\left(\frac{\alpha+\beta}{2}\right)^2=\frac{n^2}{4}&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Turán&#039;s theorem ===&lt;br /&gt;
{{Theorem|Theorem (Turán 1941)|&lt;br /&gt;
:Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a graph with &amp;lt;math&amp;gt;|V|=n&amp;lt;/math&amp;gt;. If &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, &amp;lt;math&amp;gt;k\ge 2&amp;lt;/math&amp;gt;, then&lt;br /&gt;
::&amp;lt;math&amp;gt;|E|\le\frac{r-2}{2(r-1)}n^2&amp;lt;/math&amp;gt;.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|First proof. (induction)|(Turán&#039;s original proof)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Second proof. (weight shifting)|(due to Motzkin and Straus)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Third proof. (the probabilistic method)|(due to Alon and Spencer)&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Prooftitle|Fourth proof.|&lt;br /&gt;
Let &amp;lt;math&amp;gt;G(V,E)&amp;lt;/math&amp;gt; be a &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free graph on &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices with a maximum number of edges.&lt;br /&gt;
:&#039;&#039;&#039;Claim:&#039;&#039;&#039; &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; does not contain three vertices &amp;lt;math&amp;gt;u,v,w&amp;lt;/math&amp;gt; such that &amp;lt;math&amp;gt;uv\in E&amp;lt;/math&amp;gt; but &amp;lt;math&amp;gt;uw\not\in E, vw\not\in E&amp;lt;/math&amp;gt;.&lt;br /&gt;
Suppose otherwise. There are two cases.&lt;br /&gt;
* &#039;&#039;&#039;Case.1:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt; or &amp;lt;math&amp;gt;d(w)&amp;lt;d(v)&amp;lt;/math&amp;gt;. Without loss of generality, suppose that &amp;lt;math&amp;gt;d(w)&amp;lt;d(u)&amp;lt;/math&amp;gt;. We duplicate &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; by creating a new vertex &amp;lt;math&amp;gt;u&#039;&amp;lt;/math&amp;gt; which has exactly the same neighbors as &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; (but &amp;lt;math&amp;gt;uu&#039;&amp;lt;/math&amp;gt; is not an edge). Such duplication will not increase the clique size. We then remove &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt;. The resulting graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is still &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique-free, and has &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices. The number of edges in &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+d(u)-d(w)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;,&lt;br /&gt;
:which contradicts the assumption that &amp;lt;math&amp;gt;|E(G)|&amp;lt;/math&amp;gt; is maximal.&lt;br /&gt;
* &#039;&#039;&#039;Case.2:&#039;&#039;&#039; &amp;lt;math&amp;gt;d(w)\ge d(u)&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;d(w)\ge d(v)&amp;lt;/math&amp;gt;. Duplicate &amp;lt;math&amp;gt;w&amp;lt;/math&amp;gt; twice and delete &amp;lt;math&amp;gt;u&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;v&amp;lt;/math&amp;gt;. The new graph &amp;lt;math&amp;gt;G&#039;&amp;lt;/math&amp;gt; has no &amp;lt;math&amp;gt;r&amp;lt;/math&amp;gt;-clique, and the number of edges is&lt;br /&gt;
::&amp;lt;math&amp;gt;|E(G&#039;)|=|E(G)|+2d(w)-(d(u)+d(v)+1)&amp;gt;|E(G)|\,&amp;lt;/math&amp;gt;.&lt;br /&gt;
:Contradiction again.&lt;br /&gt;
&lt;br /&gt;
The claim implies that &amp;lt;math&amp;gt;uv\not\in E&amp;lt;/math&amp;gt; defines an equivalence relation on vertices (to be more precise, it guarantees the transitivity of the relation, while the reflexivity and symmetry hold directly). Graph &amp;lt;math&amp;gt;G&amp;lt;/math&amp;gt; must be a complete multipartite graph &amp;lt;math&amp;gt;K_{n_1,n_2,\ldots,n_{r-1}}&amp;lt;/math&amp;gt; with &amp;lt;math&amp;gt;n_1+n_2+\cdots +n_{r-1}=n&amp;lt;/math&amp;gt;. Optimize the edge number, we have the Turán graph.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
=== Erdős–Stone theorem ===&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Fundamental theorem of extremal graph theory (Erdős–Stone 1946)|&lt;br /&gt;
:For any &amp;lt;math&amp;gt;r\ge 2&amp;lt;/math&amp;gt; and &amp;lt;math&amp;gt;s\ge 1&amp;lt;/math&amp;gt;, and every sufficiently large &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt;, every graph with &amp;lt;math&amp;gt;n&amp;lt;/math&amp;gt; vertices and at least &amp;lt;math&amp;gt;\left(\frac{r-2}{2(r-1)}+\epsilon\right)n^2&amp;lt;/math&amp;gt; edges contains &amp;lt;math&amp;gt;K_{r,s}&amp;lt;/math&amp;gt; as a subgraph.&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
{{Theorem|Corollary|&lt;br /&gt;
&lt;br /&gt;
}}&lt;br /&gt;
&lt;br /&gt;
== Cycle Structures ==&lt;br /&gt;
=== Girth ===&lt;br /&gt;
&lt;br /&gt;
=== Hamiltonian cycle ===&lt;/div&gt;</summary>
		<author><name>172.21.1.240</name></author>
	</entry>
</feed>