On the nontrivial extremal eigenvalues of graphs

Wenbo Li and Shiping Liu

arXiv: 2310.17520v2 · PDF · 30 October 2023
Categories: math.CO, math.SP
Licence: CC BY 4.0 The site shows the text.

Abstract

We present a finer quantitative version of an observation due to Breuillard, Green, Guralnick and Tao which tells that for finite non-bipartite Cayley graphs, once the nontrivial eigenvalues of their normalized adjacency matrices are uniformly bounded away from 1, then they are also uniformly bounded away from -1. Unlike previous works which depend heavily on combinatorial arguments, we rely more on analysis of eigenfunctions. We establish a new explicit lower bound for the gap between -1 and the smallest normalized adjacency eigenvalue, which improves previous lower bounds in terms of edge-expansion, and is comparable to the best known lower bound in terms of vertex-expansion.

Items 18

Theorem

Theorem 1

KNHY

Let G=(V,E) be a finite connected graph. Then we have

1+\mu_{n-1}\geq \frac{(1-\mu_2)^2}{2h^2} \left(\sqrt{1+\frac{h^2}{1-\mu_2}}-1\right)^2.

Proof (Proof of Theorem 1) of Theorem 1

Let f,g be eigenfunctions of the eigenvalues \mu_n and \mu_{n-1} such that \Vert f\Vert_2=\Vert g\Vert_2=1 and \langle f, g\rangle=0, i.e., \sum_{u\in V}f(u)g(u)d_u=0. Applying Proposition 1 to the function fg and inserting the estimates from Lemma 2 and Lemma 4, we derive the following inequality

\frac{\sqrt{2}}{2}h\leq \frac{ \sqrt{1+\mu_n}+\sqrt{1+\mu_{n-1}}}{\sqrt{1-\frac{1+\mu_n}{1-\mu_2}}\sqrt{1-\frac{1+\mu_{n-1}}{1-\mu_2}}-\sqrt{\frac{1+\mu_n}{1-\mu_2}}\sqrt{\frac{1+\mu_{n-1}}{1-\mu_2}}},

when 1+\mu_{n-1} < (1-\mu_2)/2. Notice that the following function

f(x,y)=\frac{\sqrt{x}+\sqrt{y}}{\sqrt{1-\frac{1}{1-\mu_2}x}\sqrt{1-\frac{1}{1-\mu_2}y}-\sqrt{\frac{1}{1-\mu_2}x}\sqrt{\frac{1}{1-\mu_2}y}}

is continuous on [0,\frac{1}{2}(1-\mu_2))\times[0,\frac{1}{2}(1-\mu_2)), monotonically increasing with respect to either x or y and f(0,0)=0. Let c=c(h,1-\mu_2) be the number such that

f(c,c)=\frac{\sqrt{2}}{2}h.

Then we have

1+\mu_{n-1}\geq c,

since, otherwise,

f(1+\mu_n,1+\mu_{n-1})< f(c,c)=\frac{\sqrt{2}}{2}h

is a contradiction. A direct computation shows that

\sqrt{c}=\frac{1-\mu_2}{\sqrt{2}h} \left(\sqrt{1+\frac{h^2}{1-\mu_2}}-1\right).

This proves the inequality (1) when 1+\mu_{n-1}<(1-\mu_2)/2.

For the case that 1+\mu_{n-1} \geq (1-\mu_2)/2, the inequality (1) still holds since

1+\mu_{n-1} \geq \frac{1-\mu_2}{2}\geq \frac{(1-\mu_2)^2}{2h^2} \left(\sqrt{1+\frac{h^2}{1-\mu_2}}-1\right)^2,

while the last inequality comes from the fact that 1-\mu_2 \geq 0.

Uses Proposition 1, Lemma 2, Lemma 4

Corollary

Corollary 1

UXJ3

Let G be a finite, non-bipartite, vertex-transitive graph. There hold

1+\mu_n \geq {\rm min} \left\{\frac{2}{d}, \frac{(\sqrt{3}-1)^2}{8}h^2\right\},

and

1+\mu_n \geq {\rm min} \left\{\frac{2}{d}, 2\left(\sqrt{1+\frac{1-\mu_2}{4}}-1\right)^2 \right \}.

Lemma

Lemma 1

Y22B

Let G=(V,E) be a finite graph and f be a function on V. We have

\sum_{\{u,v\}\in E}(f(u)-f(v))^2=\ \langle f,(I-D^{-1}A)f\rangle

and

\sum_{\{u,v\}\in E}(f(u)+f(v))^2=\ \langle f,(I+D^{-1}A)f \rangle,

where I is the n\times n identity matrix.

Used by Proof, Proof

Proposition

Proposition 1

X735

Let G=(V,E) be a finite graph. For any nonzero function f on V satisfying \sum_{u \in V}f(u)d_u=0, we have

\frac{\sum_{\{u,v\}\in E}|f(u)-f(v)|}{\sum_{u \in V}|f(u)|d_u}\geq \frac{1}{2}h.

Used by Proof

Theorem

Theorem 2 (Cheeger inequality)

IRNI

Let G=(V,E) be a finite graph. We have

2h \geq 1-\mu_2 \geq \frac{1}{2}h^2.

Lemma

Lemma 2

JH14

Let G=(V,E) be a finite graph and f,g be two eigenfunctions of the eigenvalues \mu and \nu of the normalized adjacency matrix D^{-1}A, respectively. We have the following estimate

\sum_{\{u,v\}\in E}|f(u)g(u)-f(v)g(v)|\leq \frac{\sqrt{2}}{2} \left(\sqrt{1+\mu}+\sqrt{1+\nu}\right)\Vert f\Vert_2\Vert g\Vert_2.

Proof.

By Cauchy-Schwarz inequality and Lemma 1, we calculate

\begin{aligned} & \sum_{\{u,v\}\in E}2|f(u)g(u)-f(v)g(v)|\\ = &\sum_{\{u,v\}\in E}|(f(u)-f(v))(g(u)+g(v))+(f(u)+f(v))(g(u)-g(v))|\\ \leq & \sum_{\{u,v\}\in E}|(f(u)-f(v))(g(u)+g(v))|+\sum_{\{u,v\}\in E}|(f(u)+f(v))(g(u)-g(v))|\\ \leq &\sqrt{\sum_{\{u,v\}\in E}(f(u)-f(v))^2}\sqrt{\sum_{\{u,v\}\in E}(g(u)+g(v))^2} \\ &\hspace{3.5cm} +\sqrt{\sum_{\{u,v\}\in E}(f(u)+f(v))^2}\sqrt{\sum_{\{u,v\}\in E}(g(u)-g(v))^2}\\ =& \left(\sqrt{(1-\mu)(1+\nu)}+\sqrt{(1+\mu)(1-\nu)}\right) \Vert f\Vert_2\Vert g\Vert_2 \\ \leq & \sqrt{2} (\sqrt{1+\mu}+\sqrt{1+\nu})\Vert f\Vert_2\Vert g\Vert_2. \end{aligned}

This completes the proof.

Used by Proof

Proof

64NZ

By Cauchy-Schwarz inequality and Lemma 1, we calculate

This completes the proof.

Uses Lemma 1

Lemma

Lemma 3

DJ2T

Let G=(V,E) be a finite connected graph. Assume that f is an eigenfunction of the eigenvalue \mu of the normalized adjacency matrix D^{-1}A with \Vert f\Vert_2=1. Let \boldsymbol{c} is the constant function such that \Vert c\Vert_2=1 and f_1:=|f|-\langle |f|,\boldsymbol{c}\rangle\boldsymbol{c}. Then we have

\Vert f_1\Vert_2^2 \leq \frac{1+\mu}{1-\mu_2}

and

\langle |f|, \boldsymbol{c}\rangle^2 \geq 1-\frac{1+\mu}{1-\mu_2}.

Used by Proof

Proof of Lemma 3

Oberserving that \langle f_1, \boldsymbol{c} \rangle=0, we have

(1-\mu_2) \langle f_1,f_1\rangle \leq \langle f_1,(I-D^{-1}A)f_1\rangle.

Therefore, we estimate by Lemma 1

\begin{split} \Vert f_1\Vert_2^2 &\leq \frac{1}{1-\mu_2}\langle f_1,(I-D^{-1}A)f_1\rangle= \frac{1}{1-\mu_2}\langle |f|,(I-D^{-1}A)|f|\rangle\\ &= \frac{1}{1-\mu_2} \sum_{\{u,v\}\in E}(|f(u)|-|f(v)|)^2\\ & \leq \frac{1}{1-\mu_2} \sum_{\{u,v\}\in E}(f(u)+f(v))^2\\ & =\frac{1}{1-\mu_2} \langle f,(I+D^{-1}A)f \rangle\\ & =\frac{1+\mu}{1-\mu_2}. \end{split}

The other inequality holds since

1=\Vert f\Vert_2^2=\langle |f|,\boldsymbol{c}\rangle^2+\Vert f_1\Vert_2^2.

This concludes the proof.

Uses Lemma 1

Lemma

Lemma 4

B4JN

Let G=(V,E) be a finite connected graph and f,g be eigenfunctions of the eigenvalues \mu and \nu of D^{-1}A, respectively, such that \Vert f\Vert_2 =\Vert g\Vert_2=1. Then

\langle |f|,|g|\rangle \geq \sqrt{1-\frac{1+\mu}{1-\mu_2}}\sqrt{1-\frac{1+\nu}{1-\mu_2}}-\sqrt{\frac{1+\mu}{1-\mu_2}}\sqrt{\frac{1+\nu}{1-\mu_2}}.

Proof.

Let \boldsymbol{c} be a positive constant function with \Vert c\Vert_2=1. Decompose |f| and |g| such that

|f|=\langle |f|,\boldsymbol{c}\rangle\boldsymbol{c}+f_1,\,\,\,\,\text{and}\,\,\,\,|g|=\langle |g|,\boldsymbol{c}\rangle\boldsymbol{c}+g_1.

Then we compute

\begin{split} \langle |f|,|g|\rangle&=\langle |f|,\boldsymbol{c}\rangle\langle |g|,\boldsymbol{c}\rangle+\langle f_1,g_1\rangle\\ &\geq \langle |f|,\boldsymbol{c}\rangle\langle |g|,\boldsymbol{c}\rangle-\Vert f_1\Vert_2\Vert g_1\Vert_2\\ &\geq\sqrt{1-\frac{1+\mu}{1-\mu_2}}\sqrt{1-\frac{1+\nu}{1-\mu_2}}-\sqrt{\frac{1+\mu}{1-\mu_2}}\sqrt{\frac{1+\nu}{1-\mu_2}} \end{split}

where the last inequality comes from Lemma 3.

Used by Proof

Proof

UCKG

Let \boldsymbol{c} be a positive constant function with \Vert c\Vert_2=1. Decompose |f| and |g| such that

|f|=\langle |f|,\boldsymbol{c}\rangle\boldsymbol{c}+f_1,\,\,\,\,\text{and}\,\,\,\,|g|=\langle |g|,\boldsymbol{c}\rangle\boldsymbol{c}+g_1.

Then we compute

where the last inequality comes from Lemma 3.

Uses Lemma 3

Corollary

Corollary 2

RG4X

Let G=(V,E) be a finite graph. Let n, \mu_2, \mu_{n-1} and h be as in Theorem 1. Then, we have

1+\mu_{n-1}\geq 2\left(\sqrt{1+\frac{1-\mu_2}{4}}-1\right)^2,

1+\mu_{n-1}\geq \frac{(\sqrt{3}-1)^2}{8}h^2.

Remark

Remark 1

IL7F

If we do not care about the constant, the inequality (2) can also be derived from the higher order dual Cheeger inequalities [liu2015multi, Theorem 1.2]. Indeed, we have 1+\mu_{n-1}\geq C(1-\overline{h}(2))^2, where \overline{h}(2) stands for the two-way dual Cheeger constant. Then the inequality (2) follows directly from the observation that 1-\overline{h}(2)\geq h. The proof of higher order dual Cheeger inequalities involves applying deep results from the random partition theory. Our proof here is much simpler and more elementary. We also obtain a better constant here.

Theorem

Theorem 3

RMY5

Let G be a finite connected vertex-transitive graph and \mu a simple eigenvalue of its normalized adjacency matrix. Then

\mu=\frac{2k}{d}-1

where 0\leq k \leq d is an integer and d is the degree of G.

Used by Proof

Proof of Theorem 3

Fix a\in V. Suppose that \mu is a simple eigenvalue of D^{-1}A with f being its eigenfunction. Assume further that f(a)=-1 or 1. For every g \in \mathrm{Aut}(G), the function f_g defined as f_g(\cdot)=f(g(\cdot)) is still an eigenfunction of the same eigenvalue \mu. Since \mu is simple, there exits \lambda_g such that f_g=\lambda_g f. Since \Vert f\Vert_2=\Vert f_g\Vert_2, we have \lambda_g \in \{-1,1\}. From the fact that G is vertex transitive, \mathrm{Aut}(G) acts transitively on G, we have for any u\in V that f(u)\in \{-1,1\}. Calculating D^{-1}Af(a) yields the result.

Corollary

Corollary 3

2WH1

Let (G,S) be a finite connected Cayley graph with a finite group G and a gernerating set S. Assume that G is a simple group or the size |G| is odd. Let

-1< \mu_{n} \leq \mu_{n-1} \leq ... \leq \mu_{2} \leq \mu_{1}=1

be the eigenvalues of its normalized adjacency matrix where n=|G|. Denote by h its edge-expansion. Then we have

1+\mu_n \geq \frac{(\sqrt{3}-1)^2}{8}h^2

and

1+\mu_{n}\geq 2\left(\sqrt{1+\frac{1-\mu_2}{4}}-1\right)^2.

Proof of Corollary 3

The map \lambda_g: G \rightarrow \{\pm 1\} in the proof of Theorem 3 is in fact a homomorphism of groups since for any g_1,g_2 \in G, it holds that

\lambda_{g_1g_2}f(e)=f_{g_1g_2}(e)=f(g_1g_2)=f_{g_1}(g_2)=\lambda_{g_1}f(g_2)=\lambda_{g_1}\lambda_{g_2}f(e).

Here we denote by e the identity element of the group G. If G is a simple group or the size |G| is odd, then such a homomorphism must be trivial otherwise there will be a subgroup of G of index 2. As a result all non-trivial eigenvalues of (G,S) will be of multiplicity great than 1 and the above estimates hold.

Uses Theorem 3

Source and licence

The text on this page is from arXiv:2310.17520v2 by Wenbo Li and Shiping Liu, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2310/2310.17520.jsonl.