arXiv++ Combinatorics

We study the accumulation of zeros of a polynomial arising from the enumeration of edge-coloured graphs along certain limit curves. The polynomial is a variant of an edge-chromatic polynomial, which specialises to the partition function of the ferromagnetic Ising model on a random regular graph. We call this accumulation behaviour a Lee-Yang phenomenon in analogy with the Lee-Yang theorem. The limiting loci are semialgebraic and arise from anti-Stokes curves of an exponential integral.

We establish new couplings among several random graph and multigraph models related to the random regular graph $G(n,d)$, including the configuration model and unions of random perfect matchings. As a main result, we verify the Kim-Vusandwich conjecture for all large degrees $d=n-O(\log^4 n)$ and prove a weakened version for $d=O(\log^4 n)$, which are the only remaining open cases. Our approach introduces a coupling framework that links $G(n,d)$ and $G(n,p)$ through a chain of intermediate models.

In 2004, Kim and Vu conjectured that, when $d=ω(\log n)$, the random $d$-regular graph $G_d(n)$ can be sandwiched with high probability between two random binomial graphs $G(n,p)$ with edge probabilities asymptotically equal to $\frac{d}{n}$. That is, there should exist $p_*=(1-o(1))\frac{d}{n}$, $p^*=(1+o(1))\frac{d}{n}$ and a coupling $(G_*,G,G^*)$ such that $G_*\sim G(n,p_*)$, $G\sim G_d(n)$, $G^*\sim G(n,p^*)$, and $\mathbb{P}(G_*\subset G\subset G^*)=1-o(1)$. Known as the sandwich conjecture, such a coupling is desirable as it would allow properties of the random regular graph to be inferred from those of the more easily studied binomial random graph. The conjecture was recently shown to be true when $d\gg\log^4n$ by Gao, Isaev and McKay. In this paper, we prove the sandwich conjecture in full. We do so by analysing a natural coupling procedure introduced in earlier work by Gao, Isaev and McKay, which had only previously been done when $d\gg n/\sqrt{\log n}$.

Determining the asymptotic independence ratio of random regular graphs is a key challenge in the area of sparse random graphs. Due to the interpolation method, very good upper bounds are known, which are actually known to be sharp for sufficiently large degrees. This paper provides explicit lower bounds for given degrees $d$, beating the previous best bounds for $d \geq 10$. The starting point is a second moment argument that we can boost by arguing that the obtained independent set has a certain spatial Markov property. One can then exploit this property by making local modifications to the independent set, resulting in substantial improvements. Our approach can also be used to prove the existence of other objects in random regular graphs. To demonstrate this, we consider the problem of decomposing random regular graphs into stars.

In 1989, Zehavi and Itai conjectured that every $k$-connected graph contains $k$ independent spanning trees rooted at any prescribed vertex $r$. That is, for each vertex $v$, the unique $r$-$v$ paths within these $k$ spanning trees are internally disjoint. This fundamental problem has received much attention, in part motivated by its applications to network reliability, but despite that has only been resolved for $k \le 4$ and certain restricted graph families. We establish the conjecture for almost all graphs of essentially any relevant density. Specifically, we prove that there exists a constant $C > 1$ such that, with high probability, the random graph $G(n,p)$ contains $δ(G)$ independent spanning trees rooted at any vertex whenever $C \log n/n \leq p < 0.99$. Since the lower bound on $p$ coincides (up to the constant $C$) with the connectivity threshold of $G(n,p)$, this result is essentially optimal. In addition, we show that $(n,d,λ)$-graphs with fairly mild bounds on the spectral ratio $d/λ$ contain $(1-o(1))d$ independent spanning trees rooted at each vertex, thereby settling the conjecture asymptotically for random $d$-regular graphs as well.

We introduce a new technique to prove bounds for the spectral radius of a random matrix, based on using Jensen's formula to establish the zerofreeness of the associated characteristic polynomial in a region of the complex plane. Our techniques are entirely non-asymptotic, and we instantiate it in three settings: (i) The spectral radius of non-asymptotic Girko matrices -- these are asymmetric matrices $\mathbf{M} \in \mathbb{C}^{n \times n}$ whose entries are independent and satisfy $\mathbb{E} \mathbf{M}_{ij} = 0$ and $\mathbb{E} |\mathbf{M}_{ij}^2| \le \frac{1}{n}$. (ii) The spectral radius of non-asymptotic Wigner matrices -- these are symmetric matrices $\mathbf{M} \in \mathbb{C}^{n \times n}$ whose entries above the diagonal are independent and satisfy $\mathbb{E} \mathbf{M}_{ij} = 0$, $\mathbb{E} |\mathbf{M}_{ij}^2| \le \frac{1}{n}$, and $\mathbb{E} |\mathbf{M}_{ij}^4| \le \frac{1}{n}$. (iii) The second eigenvalue of the adjacency matrix of a random $d$-regular graph on $n$ vertices, as drawn from the configuration model. In all three settings, we obtain constant-probability eigenvalue bounds that are tight up to a constant. Applied to specific random matrix ensembles, we recover classic bounds for Wigner matrices, as well as results of Bordenave--Chafaï--García-Zelada, Bordenave--Lelarge--Massoulié, and Friedman, up to constants.

In this paper, we present a new factor of IID process based on the local algorithm introduced by Díaz, Serna, and Wormald (2007). This new approach allows us to improve the previously known upper bounds on the minimum and maximum bisection width and the maximum cut of random d-regular graphs for d > 4 by introducing a new recoloring phase after the termination of the original algorithm. As an application, we show that random 5-regular graphs asymptotically almost surely admit an internal partition, i.e., a partition of the vertex set into two nonempty classes so that every vertex has at least half of its neighbors in its own class.

This paper introduces two canonical constructions that transform arbitrary finite graphs into perfect graphs: the symmetric lift $\mathrm{HL}'_2(G)$, which is purely structural and label-invariant, and the ordered lift $\mathrm{HL}_2(G)$, which depends explicitly on vertex labeling and encodes directional information. Both lifts arise as line graphs of bipartite double covers and are box-perfect. The symmetric lift $\mathrm{HL}'_2(G)$ forms a canonical 2-cover of the line graph $L(G)$. This involution decomposes $\mathrm{HL}'_2(G)$ into symmetric and antisymmetric components: the symmetric part recovers $L(G)$, while the antisymmetric part yields a signed graph $L^-(G)$, the antisymmetric line graph, with +1/-1 edges encoding consistent vs. crossed overlaps. Thus, all adjacency and Laplacian eigenvalues of $L(G)$, with multiplicities, appear within those of $\mathrm{HL}'_2(G)$, despite $L(G)$ typically not being a subgraph. For regular graphs such as Paley graphs, this yields infinite families of sparse, highly structured regular and box-perfect expanders that also retain large cliques. The lift retains much of the spectral expansion of the base while improving the combinatorial expansion. Much the same behavior is observed with random regular base graphs, allowing for the possibility of the study of box-perfect random regular graphs. Finally, we generalize these constructions to parameterized lifts $\mathrm{HL}_{r,d}(G)$ and $\mathrm{HL}_{r,d}'(G)$ defined on ordered $r$-tuples connected by Hamming distance constraints, which structurally encode the base graph and remain box-perfect.

We study how eigenvectors of random regular graphs behave when projected onto fixed directions. For a random $d$-regular graph with $N$ vertices, where the degree $d$ grows slowly with $N$, we prove that these projections follow approximately normal distributions. Our main result establishes a Berry-Esseen bound showing convergence to the Gaussian with error $O(\sqrt{d} \cdot N^{-1/6+\varepsilon})$ for degrees $d \leq N^{1/4}$. This bound significantly improves upon previous results that had error terms scaling as $d^3$, and we prove our $\sqrt{d}$ scaling is optimal by establishing a matching lower bound. Our proof combines three techniques: (1) refined concentration inequalities that exploit the specific variance structure of regular graphs, (2) a vector-based analysis of the resolvent that avoids iterative procedures, and (3) a framework combining Stein's method with graph-theoretic tools to control higher-order fluctuations. These results provide sharp constants for eigenvector universality in the transition from sparse to moderately dense graphs.

We establish the first quantitative Berry-Esseen bounds for edge eigenvector statistics in random regular graphs. For any $d$-regular graph on $N$ vertices with fixed $d \geq 3$ and deterministic unit vector $\mathbf{q} \perp \mathbf{e}$, we prove that the normalized overlap $\sqrt{N}\langle \mathbf{q}, \mathbf{u}_2 \rangle$ satisfies \[ \sup_{x \in \mathbb{R}} \left|\mathbb{P}\left(\sqrt{N}\langle \mathbf{q}, \mathbf{u}_2 \rangle \leq x\right) - Φ(x)\right| \leq C_d N^{-1/6+\varepsilon} \] where $\mathbf{u}_2$ is the second eigenvector and $C_d \leq \tilde{C}d^3\varepsilon^{-10}$ for an absolute constant $\tilde{C}$. This provides the first explicit convergence rate for the recent edge eigenvector universality results of He, Huang, and Yau \cite{HHY25}. Our proof introduces a single-scale comparison method using constrained Dyson Brownian motion that preserves the degree constraint $\tilde{H}_t\mathbf{e} = 0$ throughout the evolution. The key technical innovation is a sharp edge isotropic local law with explicit constant $C(d,\varepsilon) \leq \tilde{C}d\varepsilon^{-5}$, enabling precise control of eigenvector overlap dynamics. At the critical time $t_* = N^{-1/3+\varepsilon}$, we perform a fourth-order cumulant comparison with constrained GOE, achieving optimal error bounds through a single comparison rather than the traditional multi-scale approach. We extend our results to joint universality for the top $K$ edge eigenvectors with $K \leq N^{1/10-δ}$, showing they converge to independent Gaussians. Through analysis of eigenvalue spacing barriers, critical time scales, and comparison across multiple proof methods, we provide evidence that the $N^{-1/6}$ rate is optimal for sparse regular graphs. All constants are tracked explicitly throughout, enabling finite-size applications in spectral algorithms and network analysis.

The cycle space of a graph $G$, denoted $C(G)$, is a vector space over ${\mathbb F}_2$, spanned by all incidence vectors of edge-sets of cycles of $G$. If $G$ has $n$ vertices, then $C_n(G)$ is the subspace of $C(G)$, spanned by the incidence vectors of Hamilton cycles of $G$. We prove that asymptotically almost surely $C_n(G_{n,d}) = C(G_{n,d})$ holds whenever $n$ is odd and $d$ is a sufficiently large (even) integer. This extends (though with a weaker bound on $d$) the well-known result asserting that $G_{n,d}$ is asymptotically almost surely Hamiltonian for every $d \geq 3$ (but not for $d < 3$). Since $n$ being odd mandates that $d$ be even, somewhat limiting the generality of our result, we also prove that if $n$ is even and $d$ is any sufficiently large integer, then asymptotically almost surely $C_{n-1}(G_{n,d}) = C(G_{n,d})$. An influential result of Bohman, Frieze, and Martin asserts that if $H$ is an $n$-vertex graph with minimum degree at least $δn$ for some constant $δ> 0$, and $G \sim \mathbb{G}(n, C/n)$, where $C := C(δ)$ is a sufficiently large constant, then $H \cup G$ is asymptotically almost surely Hamiltonian. We strengthen this result by proving that the same assumptions on $H$ and $G$ ensure that $C_n(H \cup G) = C(H \cup G)$ holds asymptotically almost surely.

The metric dimension reduction modulus $k^α_n(\ell_\infty)$ is the smallest $k$ such that every $n$--point metric space can be embedded into some $k$-dimensional normed space, with bi--Lipschitz distortion at most $α$. Determining sharp asymptotics for $k^α_n(\ell_\infty)$ is a fundamental task in metric geometry, with $α=Θ(\log n)$ bearing particular interest. A line of advances over the past decades has led to an upper bound on $k^α_n(\ell_\infty)$ for $α= Ω(\log n)$, but a matching lower bound has remained open. We close this gap, establishing: for every fixed $β> 0$, $$ k^α_n(\ell_\infty) =Θ\bigg(\frac{\log n}{\log(\fracα{\log n}+1)}\bigg)\quad \mbox{for every $α\geq β\log n$}. $$ This resolves a question from Naor's 2018 ICM plenary lecture. Our result is obtained by characterizing the minimum dimension $d$ for which, with high probability, a random regular graph admits an $α$--embedding into some $d$--dimensional normed space.

Let $G$ be a graph on $n$ vertices and $(H,+)$ be an abelian group. What is the minimum size ${\sf S}_H(G)$ of the set of all sums $A(u)+A(v)$ over all injections $A:V(G)\to H$? In 2012, the first author, Angel, the second author, and Lubetzky proved that, for expander graphs and $H=\mathbb{Z}$, this minimum is at least $Ω(\log n)$, and this bound is tight -- there exists a regular expander $G$ with ${\sf S}_{\mathbb{Z}}(G)=O(\log n)$. We prove that, for every constant $d\geq 3$, the random $d$-regular graph $\mathcal{G}_{n,d}$ has significantly larger sum-sets: with high probability, for every abelian group $H$, ${\sf S}_H(\mathcal{G}_{n,d})=Ω(n^{1-2/d})$. In particular, this proves that, for every $\varepsilon>0$, there exists a regular graph with $O(n)$ edges and with sum-sets of size at least $n^{1-\varepsilon}$, for all abelian groups. The bound ${\sf S}_H(\mathcal{G}_{n,d})=Ω(n^{1-2/d})$ is tight up to a polylogarithmic factor: We show that, for every $3\leq d\leq \ln n/ \ln \ln n$, there exists an abelian group $H$ such that, for every graph $G$ on $n$ vertices with maximum degree at most $d$, ${\sf S}_H(G) \leq n^{1-2/d}(\log n)^{O(1)}$. We also prove that, for $d\gg\ln^2 n$, with high probability, for every abelian group $H$, ${\sf S}_H(\mathcal{G}_{n,d})=n(1-o(1))$ and determine the second-order term, up to a polylogarithmic factor.

A famous conjecture by Itai and Zehavi states that, for every $d$-vertex-connected graph $G$ and every vertex $r$ in $G$, there are $d$ spanning trees of $G$ such that, for every vertex $v$ in $G\setminus \{r\}$, the paths between $r$ and $v$ in different trees are internally vertex-disjoint. We show that with high probability the Itai-Zehavi conjecture holds asymptotically for the Erdős-Rényi random graph $G(n,p)$ when $np= ω(\log n)$ and for random regular graphs $G(n,d)$ when $d= ω(\log n)$. Moreover, we essentially confirm the conjecture up to a constant factor for sparser random regular graphs. This answers positively a question of Draganić and Krivelevich. Our proof makes use of recent developments on sprinkling techniques in random regular graphs.

In the context of random regular graphs, the size of the maximum cut is probably the second most studied graph parameter after the independence ratio. Zdeborová and Boettcher used the cavity method, a non-rigorous statistical physics technique, to predict one-step replica symmetry breaking (1-RSB) formulas. Coja-Ohglan et al. confirmed these predictions as rigorous upper bounds using the interpolation method. While these upper bounds were not expected to be exact, they may be very close to the true values. In this paper, we establish 2-RSB upper bounds and fine-tune their parameters to beat the aforementioned 1-RSB bounds.

A $k$-star decomposition of a graph is a partition of its edges into $k$-stars (i.e., $k$ edges with a common vertex). The paper studies the following problem: given $k \leq d/2$, does the random $d$-regular graph have a $k$-star decomposition (asymptotically almost surely, provided that the number of edges is divisible by $k$)? Delcourt, Greenhill, Isaev, Lidický, and Postle proved the a.a.s. existence for every odd $k$ using earlier results regarding orientations satisfying certain degree conditions modulo $k$. In this paper we give a direct, self-contained proof that works for every $d$ and every $k<d/2-1$. In fact, we prove stronger results. Let $s\geq 1$ denote the integer part of $d/(2k)$. We show that the random $d$-regular graph a.a.s. has a $k$-star decomposition such that the number of stars centered at each vertex is either $s$ or $s+1$. Moreover, if $k < d/3$ or $k \leq d/2 - 2.6 \log d$, we can even prescribe the set of vertices with $s$ stars, as long as it is of the appropriate size.

In this work we establish several monotonicity and decomposition results in the framework of random regular graphs. Among other results, we show that, for a wide range of parameters $d_1 \leq d_2$, there exists a coupling of $G(n,d_1)$ and $G(n,d_2)$ satisfying that $G(n,d_1) \subseteq G(n,d_2)$ with high probability, confirming a conjecture of Gao, Isaev and McKay in a new regime. Our contributions include new tools for analysing contiguity and total variation distance between random regular graph models, a novel procedure for generating unions of random edge-disjoint perfect matchings, and refined estimates of Gao's bounds on the number of perfect matchings in random regular graphs. In addition, we make progress towards another conjecture of Isaev, McKay, Southwell and Zhukovskii.

We establish the asymptotic behaviour of $μ(G(n,p))$, the number of unlabelled induced subgraphs in the binomial random graph $G(n,p)$, for almost the entire range of the probability parameter $p=p(n)\in[0,1]$. In particular, we show that typically the number of subgraphs becomes exponential when $p$ passes $1/n$, reaches maximum possible base of exponent (asymptotically) when $p\gg 1/n$, and reaches the asymptotic value $2^n$ when $p$ passes $2\ln n/n$. For $p\gg \ln n/n$, we get the first order term and asymptotics of the second order term of $μ(G(n,p))$. We also prove that random regular graphs $G_{n,d}$ typically have $μ(G_{n,d})\geq 2^{c_d n}$ for all $d\geq 3$ and some positive constant $c_d$ such that $c_d\to 1$ as $d\to\infty$.

A $k$-star decomposition of a graph is a partition of its edges into $k$-stars (i.e., $k$ edges with a common vertex). The paper studies the following problem: for what values of $k>d/2$ does the random $d$-regular graph have a $k$-star decomposition (asymptotically almost surely, provided that the number of edges is divisible by $k$)? Delcourt, Greenhill, Isaev, Lidický, and Postle proposed the following conjecture. It is easy to see that a $k$-star decomposition necessitates the existence of an independent set of density $1-d/(2k)$. So let $k^{\mathrm{ind}}_d$ be the largest $k$ for which the random $d$-regular graph a.a.s. contains an independent set of this density. Clearly, $k$-star decompositions cannot exist for $k>k^{\mathrm{ind}}_d$. The conjecture suggests that this is essentially the only restriction: there is a threshold $k^\star_d$ such that $k$-star decompositions exist if and only if $k \leq k^\star_d$, and it (basically) coincides with the other threshold, i.e., $k^\star_d \approx k^{\mathrm{ind}}_d$. We confirm this conjecture for sufficiently large $d$ by showing that a $k$-star decomposition exists if $d/2< k < k^{\mathrm{ind}}_d$. In fact, we prove the existence even if $k=k^{\mathrm{ind}}_d$ for degrees $d$ with asymptotic density $1$.

The Sudoku number $s(G)$ of graph $G$ with chromatic number $χ(G)$ is the smallest partial $χ(G)$-colouring of $G$ that determines a unique $χ(G)$-colouring of the entire graph. We show that the Sudoku number of the random $3$-regular graph $\mathcal{G}_{n,3}$ satisfies $s(\mathcal{G}_{n,3}) \leq (1+o(1))\frac{n}{3}$ asymptotically almost surely. We prove this by analyzing an algorithm which $3$-colours $\mathcal{G}_{n,3}$ in a way that produces many locally forced vertices, i.e., vertices which see two distinct colours among their neighbours. The intricacies of the algorithm present some challenges for the analysis, and to overcome these we use a non-standard application of Wormald's differential equations method that incorporates tools from finite Markov chains.