arXiv++ Combinatorics

The Ising antiferromagnet is an important statistical physics model with close connections to the {\sc Max Cut} problem. Combining spatial mixing arguments with the method of moments and the interpolation method, we pinpoint the replica symmetry breaking phase transition predicted by physicists. Additionally, we rigorously establish upper bounds on the {\sc Max Cut} of random regular graphs predicted by Zdeborová and Boettcher [Journal of Statistical Mechanics 2010]. As an application we prove that the information-theoretic threshold of the disassortative stochastic block model on random regular graphs coincides with the Kesten-Stigum bound.

In this paper we give new bounds on the bisection width of random 3-regular graphs on $n$ vertices. The main contribution is a new lower bound of $0.103295n$ based on a first moment method together with a structural analysis of the graph, thereby improving a 27-year-old result of Kostochka and Melnikov. We also give a complementary upper bound of $0.139822n$ by combining a result of Lyons with original combinatorial insights. Developping this approach further, we obtain a non-rigorous improved upper bound with the help of Monte Carlo simulations.

We study the set ${\cal L}(G)$ of lengths of all cycles that appear in a random $d$-regular $G$ on $n$ vertices for a fixed $d\geq 3$, as well as in Erdős--Rényi random graphs on $n$ vertices with a fixed average degree $c>1$. Fundamental results on the distribution of cycle counts in these models were established in the 1980's and early 1990's, with a focus on the extreme lengths: cycles of fixed length, and cycles of length linear in $n$. Here we derive, for a random $d$-regular graph, the limiting probability that ${\cal L}(G)$ simultaneously contains the entire range $\{\ell,\ldots,n\}$ for $\ell\geq 3$, as an explicit expression $θ_\ell=θ_\ell(d)\in(0,1)$ which goes to $1$ as $\ell\to\infty$. For the random graph ${\cal G}(n,p)$ with $p=c/n$, where $c\geq C_0$ for some absolute constant $C_0$, we show the analogous result for the range $\{\ell,\ldots,(1-o(1))L_{\max}(G)\}$, where $L_{\max}$ is the length of a longest cycle in $G$. The limiting probability for ${\cal G}(n,p)$ coincides with $θ_\ell$ from the $d$-regular case when $c$ is the integer $d-1$. In addition, for the directed random graph ${\cal D}(n,p)$ we show results analogous to those on ${\cal G}(n,p)$, and for both models we find an interval of $c ε^2 n$ consecutive cycle lengths in the slightly supercritical regime $p=\frac{1+ε}n$.

We study the problem of efficiently refuting the k-colorability of a graph, or equivalently certifying a lower bound on its chromatic number. We give formal evidence of average-case computational hardness for this problem in sparse random regular graphs, showing optimality of a simple spectral certificate. This evidence takes the form of a computationally-quiet planting: we construct a distribution of d-regular graphs that has significantly smaller chromatic number than a typical regular graph drawn uniformly at random, while providing evidence that these two distributions are indistinguishable by a large class of algorithms. We generalize our results to the more general problem of certifying an upper bound on the maximum k-cut. This quiet planting is achieved by minimizing the effect of the planted structure (e.g. colorings or cuts) on the graph spectrum. Specifically, the planted structure corresponds exactly to eigenvectors of the adjacency matrix. This avoids the pushout effect of random matrix theory, and delays the point at which the planting becomes visible in the spectrum or local statistics. To illustrate this further, we give similar results for a Gaussian analogue of this problem: a quiet version of the spiked model, where we plant an eigenspace rather than adding a generic low-rank perturbation. Our evidence for computational hardness of distinguishing two distributions is based on three different heuristics: stability of belief propagation, the local statistics hierarchy, and the low-degree likelihood ratio. Of independent interest, our results include general-purpose bounds on the low-degree likelihood ratio for multi-spiked matrix models, and an improved low-degree analysis of the stochastic block model.

A spanning tree $T$ in a graph $G$ is a sub-graph of $G$ with the same vertex set as $G$ which is a tree. In 1981, McKay proved an asymptotic result regarding the number of spanning trees in random $k$-regular graphs. In this paper we prove a high-dimensional generalization of McKay's result for random $d$-dimensional, $k$-regular simplicial complexes on $n$ vertices, showing that the weighted number of simplicial spanning trees is of order $(ξ_{d,k}+o(1))^{\binom{n}{d}}$ as $n\to\infty$, where $ξ_{d,k}$ is an explicit constant, provided $k> 4d^2+d+2$. A key ingredient in our proof is the local convergence of such random complexes to the $d$-dimensional, $k$-regular arboreal complex, which allows us to generalize McKay's result regarding the Kesten-McKay distribution.

We prove that a random $d$-regular graph, with high probability, is a cut sparsifier of the clique with approximation error at most $\left(2\sqrt{\frac 2 π} + o_{n,d}(1)\right)/\sqrt d$, where $2\sqrt{\frac 2 π} = 1.595\ldots$ and $o_{n,d}(1)$ denotes an error term that depends on $n$ and $d$ and goes to zero if we first take the limit $n\rightarrow \infty$ and then the limit $d \rightarrow \infty$. This is established by analyzing linear-size cuts using techniques of Jagannath and Sen derived from ideas in statistical physics, and analyzing small cuts via martingale inequalities. We also prove new lower bounds on spectral sparsification of the clique. If $G$ is a spectral sparsifier of the clique and $G$ has average degree $d$, we prove that the approximation error is at least the "Ramanujan bound'' $(2-o_{n,d}(1))/\sqrt d$, which is met by $d$-regular Ramanujan graphs, provided that either the weighted adjacency matrix of $G$ is a (multiple of) a doubly stochastic matrix, or that $G$ satisfies a certain high "odd pseudo-girth" property. The first case can be seen as an "Alon-Boppana theorem for symmetric doubly stochastic matrices," showing that a symmetric doubly stochastic matrix with $dn$ non-zero entries has a non-trivial eigenvalue of magnitude at least $(2-o_{n,d}(1))/\sqrt d$; the second case generalizes a lower bound of Srivastava and Trevisan, which requires a large girth assumption. Together, these results imply a separation between spectral sparsification and cut sparsification. If $G$ is a random $\log n$-regular graph on $n$ vertices, we show that, with high probability, $G$ admits a (weighted subgraph) cut sparsifier of average degree $d$ and approximation error at most $(1.595\ldots + o_{n,d}(1))/\sqrt d$, while every (weighted subgraph) spectral sparsifier of $G$ having average degree $d$ has approximation error at least $(2-o_{n,d}(1))/\sqrt d$.

The modularity of a graph is a parameter that measures its community structure; the higher its value (between $0$ and $1$), the more clustered the graph is. In this paper we show that the modularity of a random $3$-regular graph is at least $0.667026$ asymptotically almost surely (a.a.s.), thereby proving a conjecture of McDiarmid and Skerman. We also improve the a.a.s. upper bound given therein to $0.789998$. For a uniformly chosen graph $G_n$ over a given bounded degree sequence with average degree $d(G_n)$ and with $|CC(G_n)|$ many connected components, we distinguish two regimes with respect to the existence of a giant component. In the subcritical regime, we compute the second term of the modularity. In the supercritical regime, we prove that there is $\varepsilon > 0$, for which the modularity is a.a.s. at least \begin{equation*} \dfrac{2\left(1 - μ\right)}{d(G_n)}+\varepsilon, \end{equation*} where $μ$ is the asymptotically almost sure limit of $\dfrac{|CC(G_n)|}{n}$.

The main goal of this note is to illustrate the advantage of analyzing the non-backtracking spectrum of a regular graph rather than the ordinary spectrum. We show that by switching to non-backtracking spectrum, the method of proof used in [Puder 2015, arXiv::1212.5216] yields a bound of $2\sqrt{d-1}+\frac{2}{\sqrt{d-1}}$ instead of the original $2\sqrt{d-1}+1$ on the second largest eigenvalue of a random $d$-regular graph.

An independent set of size $k$ in a finite undirected graph $G$ is a set of $k$ vertices of the graph, no two of which are connected by an edge. Let $x_{k}(G)$ be the number of independent sets of size $k$ in the graph $G$ and let $α(G)=\max\{k\geq0\colon x_{k}(G)\neq0\}$. In 1987, Alavi, Malde, Schwenk and Erdös asked if the independent set sequence $x_{0}(G),x_{1}(G),\ldots,x_{α(G)}(G)$ of a tree is unimodal (the sequence goes up and then down). This problem is still open. In 2006, Levit and Mandrescu showed that the last third of the independent set sequence of a tree is decreasing. We show that the first 46.8\% of the independent set sequence of a random tree is increasing with (exponentially) high probability as the number of vertices goes to infinity. So, the question of Alavi, Malde, Schwenk and Erdös is ``four-fifths true'', with high probability. We also show unimodality of the independent set sequence of Erdös-Renyi random graphs, when the expected degree of a single vertex is large (with (exponentially) high probability as the number of vertices in the graph goes to infinity, except for a small region near the mode). A weaker result is shown for random regular graphs. The structure of independent sets of size $k$ as $k$ varies is of interest in probability, statistical physics, combinatorics, and computer science.

We consider the spectral gap of a uniformly chosen random $(d_1,d_2)$-biregular bipartite graph $G$ with $|V_1|=n, |V_2|=m$, where $d_1,d_2$ could possibly grow with $n$ and $m$. Let $A$ be the adjacency matrix of $G$. Under the assumption that $d_1\geq d_2$ and $d_2=O(n^{2/3}),$ we show that $λ_2(A)=O(\sqrt{d_1})$ with high probability. As a corollary, combining the results from Tikhomirov and Youssef (2019), we showed that the second singular value of a uniform random $d$-regular digraph is $O(\sqrt{d})$ for $1\leq d\leq n/2$ with high probability. Assuming $d_2$ is fixed and $d_1=O(n^2)$, we further prove that for a random $(d_1,d_2)$-biregular bipartite graph, $|λ_i^2(A)-d_1|=O(\sqrt{d_1})$ for all $2\leq i\leq n+m-1$ with high probability. The proofs of the two results are based on the size biased coupling method introduced in Cook, Goldstein, and Johnson (2018) for random $d$-regular graphs and several new switching operations we defined for random bipartite biregular graphs.

An emerging trend in approximate counting is to show that certain `low-temperature' problems are easy on typical instances, despite worst-case hardness results. For the class of regular graphs one usually shows that expansion can be exploited algorithmically, and since random regular graphs are good expanders with high probability the problem is typically tractable. Inspired by approaches used in subexponential-time algorithms for Unique Games, we develop an approximation algorithm for the partition function of the ferromagnetic Potts model on graphs with a small-set expansion condition. In such graphs it may not suffice to explore the state space of the model close to ground states, and a novel feature of our method is to efficiently find a larger set of `pseudo-ground states' such that it is enough to explore the model around each pseudo-ground state.

We describe a new method to remove short cycles on regular graphs while maintaining spectral bounds (the nontrivial eigenvalues of the adjacency matrix), as long as the graphs have certain combinatorial properties. These combinatorial properties are related to the number and distance between short cycles and are known to happen with high probability in uniformly random regular graphs. Using this method we can show two results involving high girth spectral expander graphs. First, we show that given $d \geq 3$ and $n$, there exists an explicit distribution of $d$-regular $Θ(n)$-vertex graphs where with high probability its samples have girth $Ω(\log_{d - 1} n)$ and are $ε$-near-Ramanujan; i.e., its eigenvalues are bounded in magnitude by $2\sqrt{d - 1} + ε$ (excluding the single trivial eigenvalue of $d$). Then, for every constant $d \geq 3$ and $ε> 0$, we give a deterministic poly$(n)$-time algorithm that outputs a $d$-regular graph on $Θ(n)$-vertices that is $ε$-near-Ramanujan and has girth $Ω(\sqrt{\log n})$, based on the work of arXiv:1909.06988 .

A coupling of random walkers on the same finite graph, who take turns sequentially, is said to be an avoidance coupling if the walkers never collide. Previous studies of these processes have focused almost exclusively on complete graphs, in particular how many walkers an avoidance coupling can include. For other graphs, apart from special cases, it has been unsettled whether even two non-colliding simple random walkers can be coupled. In this article, we construct such a coupling on (i) any $d$-regular graph avoiding a fixed subgraph depending on $d$; and (ii) any square-free graph with minimum degree at least three. A corollary of the first result is that a uniformly random regular graph on $n$ vertices admits an avoidance coupling with high probability.

Consider random $d$-regular graphs, i.e., random graphs such that there are exactly $d$ edges from each vertex for some $d\ge 3$. We study both the configuration model version of this graph, which has occasional multi-edges and self-loops, as well as the simple version of it, which is a $d$-regular graph chosen uniformly at random from the collection of all $d$-regular graphs. In this paper, we discuss mixing times of Glauber dynamics for the Ising model with an external magnetic field on a random $d$-regular graph, both in the quenched as well as the annealed settings. Let $β$ be the inverse temperature, $β_c$ be the critical temperature and $B$ be the external magnetic field. Concerning the annealed measure, we show that for $β> β_c$ there exists $\hat{B}_c(β)\in (0,\infty)$ such that the model is metastable (i.e., the mixing time is exponential in the graph size $n$) when $β> β_c$ and $0 \leq B < \hat{B}_c(β)$, whereas it exhibits the cut-off phenomenon at $c_\star n \log n$ with a window of order $n$ when $β< β_c$ or $β> β_c$ and $B>\hat{B}_c(β)$. Interestingly, $\hat{B}_c(β)$ coincides with the critical external field of the Ising model on the $d$-ary tree (namely, above which the model has a unique Gibbs measure). Concerning the quenched measure, we show that there exists $B_c(β)$ with $B_c(β) \leq \hat{B}_c(β)$ such that for $β> β_c$, the mixing time is at least exponential along some subsequence $(n_k)_{k\geq 1}$ when $0 \leq B < B_c(β)$, whereas it is less than or equal to $Cn\log n$ when $B>\hat{B}_c(β)$. The quenched results also hold for the model conditioned on simplicity, for the annealed results this is unclear.

This is the first in a series of six articles devoted to showing that a typical covering map of large degree to a fixed, regular graph has its new adjacency eigenvalues within the bound conjectured by Alon for random regular graphs. Many of the techniques we develop hold whether or not the base graph is regular. Our first main theorem in this series of articles is that if the base graph is $d$-regular, then for any $ε>0$, as the degree, $n$, of the covering map tends to infinity, some new adjacency eigenvalue of the map is larger in absolute value that $2(d-1)^{1/2}+ε$ with probability at most order $1/n$. Our second main theorem is that if, in addition, the base graph is Ramanujan, then this probability is bounded above and below by $1/n$ to the power of a positive integer that we call the {\em tangle power} of the model, i.e., of the probability spaces of random covering maps of degree $n$. The tangle power is fairly easy to bound from below, and at times to compute exactly; it measures the probability that certain {\em tangles} appear in the random covering graph, where a {\em tangle} is a local event that forces the covering graph to have a new eigenvalue strictly larger than $2(d-1)^{1/2}$. Our main theorems are relativizations of Alon's conjecture on the second eigenvalue of random regular graphs of large degree. In this first article of the series, we introduce all the terminology needed in this series, motivate this terminology, precisely state all the results in the remaining articles, and make some remarks about their proofs. As such, this article provides an overview of the entire series of articles; furthermore, the rest of the articles in this series may be read independently of one another.

This is the second in a series of articles devoted to showing that a typical covering map of large degree to a fixed, regular graph has its new adjacency eigenvalues within the bound conjectured by Alon for random regular graphs. The first main result in this article concerns the function $f(k,n)$ defined as the number of SNBC (strictly non-backtracking closed) walks of length $k$ of a given homotopy type in a random covering graph of degree $n$ of a fixed graph. We prove the existence of asymptotic expansions in powers of $1/n$ for $f(k,n)$, where the coefficients---functions of $k$---are proven to have some desirable properties; namely, these coefficients are approximately a sum of polynomials times exponential functions. The second main result is a generalization of the first, where the number of SNBC walks of length $k$ is multiplied by an indicator function that the covering graph contains a certain type of {\em tangle}; the second result requires more terminology, although its proof uses the same basic tools used to prove the first result. % The motivation for the second main result will be clear in % the third article in this series of articles. The results in this article are mostly straightforward generalizations of methods used in previous works. However, this article (1) "factors" these methods into a number of short, conceptually simple, and independent parts, (2) writes each independent part in more general terms, and (3) significantly simplifies of one of the previous computations. As such we expect that this article will make it easier to apply trace methods to related models of random graphs.

This is the third in a series of articles devoted to showing that a typical covering map of large degree to a fixed, regular graph has its new adjacency eigenvalues within the bound conjectured by Alon for random regular graphs. In this paper we consider random graphs that are random covering graphs of large degree $n$ of a fixed base graph. We prove the existence of asympototic expansion in $1/n$ for the expected value of the number of strictly non-backtracking closed walks of length $k$ times the indicator function that the graph is free of certain {\em tangles}; moreover, we prove that the coefficients of these expansions are "nice functions" of $k$, namely approximately equal to a sum of polynomials in $k$ times exponential functions of $k$. Our results use the methods of Friedman used to resolve Alon's original conjecture, combined with the results of Article~II in this series of articles. One simplification in this article over the previous methods of Friedman is that the "regularlized traces" used in this article, which we call {\em certified traces}, are far easier to define and work with than the previously utilized {\em selective traces}.

This is the fourth in a series of articles devoted to showing that a typical covering map of large degree to a fixed, regular graph has its new adjacency eigenvalues within the bound conjectured by Alon for random regular graphs. In this paper we prove a {\em Sidestepping Theorem} that is more general and easier to use than earlier theorems of this kind. Such theorems concerns a family probability spaces $\{\mathcal{M}_n\}$ of $n\times n$ matrices, where $n$ varies over some infinite set, $N$, of natural numbers. Many trace methods use simple "Markov bounds" to bound the expected spectral radius of elements of ${\mathcal{M}}_n$: this consists of choosing one value, $k=k(n)$, for each $n\in N$, and proving expected spectral radius bounds based on the expected value of the trace of the $k=k(n)$-power of elements of ${\mathcal{M}}_n$. {\em Sidestepping} refers to bypassing such simple Markov bounds, obtaining improved results using a number of values of $k$ for each fixed $n\in N$. In more detail, if the $M\in {\mathcal{M}}_n$ expected value of ${\rm Trace}(M^k)$ has an asymptotic expansion in powers of $1/n$, whose coefficients are "well behaved" functions of $k$, then one can get improved bounds on the spectral radius of elements of ${\mathcal{M}}_n$ that hold with high probability. Such asymptotic expansions are shown to exist in the third article in this series for the families of matrices that interest us; in the fifth and sixth article in this series we will apply the Sidestepping Theorem in this article to prove the main results in this series of articles. This article is independent of all other articles in this series; it can be viewed as a theorem purely in probability theory, concerning random matrices or, equivalently, the $n$ random variables that are the eigenvalues of the elements of ${\mathcal{M}}_n$.

This is the fifth in a series of articles devoted to showing that a typical covering map of large degree to a fixed, regular graph has its new adjacency eigenvalues within the bound conjectured by Alon for random regular graphs. In this article we use the results of Articles~III and IV in this series to prove that if the base graph is regular, then as the degree, $n$, of the covering map tends to infinity, some new adjacency eigenvalue has absolute value outside the Alon bound with probability bounded by $O(1/n)$. In addition, we give upper and lower bounds on this probability that are tight to within a multiplicative constant times the degree of the covering map. These bounds depend on two positive integers, the \emph{algebraic power} (which can also be $+\infty$) and the \emph{tangle power} of the model of random covering map. We conjecture that the algebraic power of the models we study is always $+\infty$, and in Article~VI we prove this when the base graph is regular and \emph{Ramanujan}. When the algebraic power of the model is $+\infty$, then the results in this article imply stronger results, such as (1) the upper and lower bounds mentioned above are matching to within a multiplicative constant, and (2) with probability smaller than any negative power of the degree, the some new eigenvalue fails to be within the Alon bound only if the covering map contains one of finitely many "tangles" as a subgraph (and this event has low probability).

This is the sixth in a series of articles devoted to showing that a typical covering map of large degree to a fixed, regular graph has its new adjacency eigenvalues within the bound conjectured by Alon for random regular graphs. In this article we show that if the fixed graph is regular Ramanujan, then the {\em algebraic power} of the model of random covering graphs is $+\infty$. This implies a number of interesting results, such as (1) one obtains the upper and lower bounds---matching to within a multiplicative constant---for the probability that a random covering map has some new adjacency eigenvalue outside the Alon bound, and (2) with probability smaller than any negative power of the degree of the covering map, some new eigenvalue fails to be within the Alon bound without the covering map containing one of finitely many "tangles" as a subgraph (and this tangle containment event has low probability).