arXiv subjects
Publications and source records attributed to Lyuben Lichev.
We show that, for some absolute constants $c_1,c_2>0$, every $(n,d,\lambda)$-graph with $\lambda\le c_1 d$ contains an induced cycle of length at least $c_2n\log(d/\lambda)/d$. This is best possible up to the values of $c_1,c_2$. Our techniques include a multi-scale algorithmic analysis, an adapted depth-first exploration procedure, a link to percolation theory, and estimates for random row-and-column extraction in symmetric matrices.
We develop a framework for proving universality results in sparse random graphs. As a first application, we show that there exists an absolute constant $C>1$ such that, with high probability, for every fixed constant $\Delta$, the binomial random graph $G(n,C\ln n/n)$ contains every $n$-vertex tree with maximum degree at most $\Delta$. This answers a question of Montgomery (Advances in Mathematics, 2019). We also determine, for every $p$ satisfying $C\ln n/n\leq p=n^{-1+o(1)}$, the minimum girth $\ell$ (up to an absolute multiplicative constant) for which with high probability $G(n,p)$ contains all cycle factors of girth at least $\Omega(\ell)$. In particular, with high probability $G(n,C\ln n/n)$ contains all cycle factors of girth at least $100\ln n/\ln\ln n$, which is optimal up to a constant factor. This extends an earlier result of Ferber, Kronenberg, and Luh (Transaction of the American Mathematical Society, 2019) and significantly improves a corollary of a deep result of Kahn, Lubetzky, and Wormald (Communications on Pure and Applied Mathematics, 2017). One of the key ingredients in the proofs is establishing the optimal depth of linking systems in sparse random graphs.
We study the edge-expansion of the graph of a random $0/1$ polytope $P^d_p$, defined as the convex hull of a random subset of the points in $\{0,1\}^d$ where every point is retained independently and with probability $p$. This problem was introduced more than twenty years ago in a work of Gillmann and Kaibel, and has been extensively studied ever since. We prove that, for every fixed $\varepsilon>0$ and every $p\in(0,1-\varepsilon]$, with high probability the graph of $P^d_p$ has edge-expansion $\Theta(d)$. This improves the previously best known bound due to Ferber, Krivelevich, Sales and Samotij, and verifies, in a strong form, the celebrated Mihail-Vazirani conjecture for random $0/1$ polytopes. Although the expansion factor $\Theta(d)$ is typically best possible for $p\ge 1/2+\varepsilon$, we also show that the behaviour changes drastically at $p=1/2$. Namely, for every fixed $\varepsilon>0$ and every integer $k\ge 2$, if $p\le 1/2-\varepsilon$, then with high probability the graph of $P^d_p$ has edge-expansion $\Omega(d^k)$. Thus, random $0/1$ polytopes exhibit an interesting phase transition at $p=1/2$.
The depth-weighted tree DWT($f$) with weight function $f:\{0,1,2,\ldots\}\to (0,\infty)$ is a dynamic random tree grown from a root $r$ where vertices arrive consecutively and every new vertex attaches to a parent $u$ with probability proportional to $f$(distance between $u$ and $r$). This work is dedicated to a systematic analysis of the depth of DWT($f$). Namely, we provide precise analytic expressions of the typical depth of DWT($f$) for convergent, periodic, slowly growing, and (super-)exponentially growing weight functions. Furthermore, for bounded or exponentially growing $f$, we determine the typical depth up to a multiplicative constant, thus confirming and strengthening a conjecture of Leckey, Mitsche and Wormald.
We consider a random geometric graph process where random points $(X_i)_{i \ge 1}$ are embedded consecutively in the $d$-dimensional unit torus $\mathbb{T}^d$, and every two points at distance at most $r$ form an edge. As $r\to 0$, we confirm that well-known hitting time results for $k$-connectivity (with $k\ge 1$ fixed) and Hamiltonicity in the Erd\H{o}s-R\'enyi graph process also hold for the considered geometric analogue. Moreover, we exhibit a sort of probabilistic monotonicity for each of these properties. We also study a geometric analogue of the power of choice where, at each step, an agent is given two random points sampled independently and uniformly from $\mathbb{T}^d$ and must add exactly one of them to the already constructed point set. When the agent is allowed to make their choice with the knowledge of the entire sequence of random points (offline 2-choice), we show that they can construct a connected graph at the first time $t$ when none of the first $t$ pairs of proposed points contains two isolated vertices in the graph induced by $(X_i)_{i=1}^{2t}$, and maintain connectivity thereafter by following a simple algorithm. We also derive analogous results for $k$-connectivity and Hamiltonicity. This shows that each of the said properties can be attained two times faster (time-wise) and with four times fewer points in the offline 2-choice process compared to the 1-choice process. In the online version where the agent only knows the process until the current time step, we show that $k$-connectivity and Hamiltonicity cannot be significantly accelerated (time-wise) but may be realised on two times fewer points compared to the 1-choice analogue.
We consider bond percolation on the $d$-dimensional binary hypercube with $p=c/d$ for fixed $c>1$. We prove that the typical diameter of the giant component $L_1$ is of order $\Theta(d)$, and the typical mixing time of the lazy random walk on $L_1$ is of order $\Theta(d^2)$. This resolves long-standing open problems of Bollob\'as, Kohayakawa and {\L}uczak from 1994, and of Benjamini and Mossel from 2003. A key component in our approach is a new tight large deviation estimate on the number of vertices in $L_1$ whose proof includes several novel ingredients: a structural description of the residue outside the giant component after sprinkling, a tight quantitative estimate on the spread of the giant in the hypercube, and a stability principle which rules out the disintegration of large connected sets under thinning. This toolkit further allows us to obtain optimal bounds on the expansion in $L_1$.
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 Erd\H{o}s-R\'enyi random graph $G(n,p)$ when $np= \omega(\log n)$ and for random regular graphs $G(n,d)$ when $d= \omega(\log n)$. Moreover, we essentially confirm the conjecture up to a constant factor for sparser random regular graphs. This answers positively a question of Dragani\'{c} and Krivelevich. Our proof makes use of recent developments on sprinkling techniques in random regular graphs.
Let $Q^d_p$ be the random subgraph of the $d$-dimensional binary hypercube obtained after edge-percolation with probability $p$. It was shown recently by the authors that, for every $\varepsilon > 0$, there is some $c = c(\varepsilon)>0$ such that, if $pd\ge c$, then typically $Q^d_p$ contains a cycle of length at least $(1-\varepsilon)2^d$. We strengthen this result to show that, under the same assumptions, typically $Q^d_p$ contains cycles of all even lengths between $4$ and $(1-\varepsilon)2^d$.
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 determine the sharp threshold for the containment of all $n$-vertex trees of bounded degree in random geometric graphs with $n$ vertices. This provides a geometric counterpart of Montgomery's threshold result for binomial random graphs, and confirms a conjecture of Espuny D\'iaz, Lichev, Mitsche, and Wesolek. Our proof is algorithmic and adapts to other families of graphs, in particular graphs with bounded genus or tree-width.
Let $Q^d$ be the $d$-dimensional binary hypercube. We form a random subgraph $Q^d_p\subseteq Q^d$ by retaining each edge of $Q^d$ independently with probability $p$. We show that, for every constant $\varepsilon>0$, there exists a constant $C=C(\varepsilon)>0$ such that, if $p\ge C/d$, then with high probability $Q^d_p$ contains a cycle of length at least $(1-\varepsilon)2^d$. This confirms a long-standing folklore conjecture, stated in particular by Condon, Espuny D\'iaz, Gir\~ao, K\"uhn, and Osthus [Hamiltonicity of random subgraphs of the hypercube, Mem. Amer. Math. Soc. 305 (2024), No. 1534].
Fix a sequence of $d$-regular graphs $(G_d)_{d\in \mathbb{N}}$ and denote by $G_{d,p}$ the graph obtained from $G_d$ after edge-percolation with probability $p=c/d$, for a constant $c>0$. We prove a quantitative local convergence of $(G_{d,p})_{d\in \mathbb{N}}$. In combination with results of Bordenave, Lelarge and Salez, it implies that the rescaled matching number of $G_{d,p}$ is asymptotically equivalent to that of the binomial random graph $G(n,c/n)$.
Consider the process where the $n$ vertices of a square $2$-dimensional torus appear consecutively in a random order. We show that typically the size of the $3$-core of the corresponding induced unit-distance graph transitions from $0$ to $n-o(n)$ within a single step. Equivalently, by infecting the vertices of the torus in a random order, under two-neighbour bootstrap percolation, the size of the infected set transitions instantaneously from $o(n)$ to $n$. This hitting time result answers a question of Benjamini. We also study the much more challenging and general setting of bootstrap percolation on two-dimensional lattices for a variety of finite-range infection rules. In this case, powerful but fragile bootstrap percolation tools such as the rectangles process and the Aizenman-Lebowitz lemma become unavailable. We develop a new method complementing and replacing these standard techniques, thus allowing us to prove the above hitting time result for a wide family of threshold bootstrap percolation rules on the $2$-dimensional square lattice, including neighbourhoods given by large $\ell^p$ balls for $p\in[1,\infty]$.
Fix $d\ge2$ and consider a uniformly random set $P$ of $n$ points in $[0,1]^{d}$. Let $G$ be the Hasse diagram of $P$ (with respect to the coordinatewise partial order), or alternatively let $G$ be the Delaunay graph of $P$ with respect to axis-parallel boxes (where we put an edge between $u,v\in P$ whenever there is an axis-parallel box containing $u,v$ and no other points of $P$). In each of these two closely related settings, we show that the chromatic number of $G$ is typically $(\log n)^{d-1+o(1)}$ and the independence number of $G$ is typically $n/(\log n)^{d-1+o(1)}$. When $d=2$, we obtain bounds that are sharp up to constant factors: the chromatic number is typically of order $\log n/\log\log n$ and the independence number is typically of order $n\log\log n/\log n$. These results extend and sharpen previous bounds by Chen, Pach, Szegedy and Tardos. In addition, they provide new bounds on the largest possible chromatic number (and lowest possible independence number) of a $d$-dimensional box-Delaunay graph or Hasse diagram, in particular resolving a conjecture of Tomon.
For a tree $T$ and a function $f \colon E(T)\to \mathbb{S}^d$, the imbalance of a subtree $T'\subseteq T$ is given by $|\sum_{e \in E(T')} f(e)|$. The $d$-dimensional discrepancy of the tree $T$ is the minimum, over all functions $f$ as above, of the maximum imbalance of a subtree of $T$. We prove tight asymptotic bounds for the discrepancy of a tree $T$, confirming a conjecture of Krishna, Michaeli, Sarantis, Wang and Wang. We also settle a related conjecture on oriented discrepancy of subtrees by the same authors.
We address several related problems on combinatorial discrepancy of trees in a setting introduced by Erd\H{o}s, F\"{u}redi, Loebl and S\'{o}s. Given a fixed tree $T$ on $n$ vertices and an edge-colouring of the complete graph $K_n$, for every colour, we find a copy of $T$ in $K_n$ where the number of edges in that colour significantly exceeds its expected count in a uniformly random embedding. This resolves a problem posed by Erd\H{o}s, F\"{u}redi, Loebl and S\'{o}s by generalising their work from two to many colours. Furthermore, if $T$ has maximum degree $\Delta\leq\epsilon n$ for sufficiently small $\epsilon > 0$ and the edge-colouring of $K_n$ is both balanced and ``not too close'' to one particular instance, we show that, for every colour, there is a copy of $T$ in $K_n$ where that colour appears on linearly more edges than any other colour. Several related examples are provided to demonstrate the necessity of the introduced structural restrictions. Our proofs combine saturation arguments for the existence of particular coloured substructures and analysis of conveniently defined local exchanges. Using similar methods, we investigate the existence of copies of a graph $H$ with prescribed number of edges in each colour in $2$-edge-coloured dense host graphs. In particular, for a graph $H$ with bounded maximum degree and balanced $2$-edge-colourings $\mathbf{c}$ of a host graph $G$ with minimum degree at least $(1-\epsilon)n$ for some $\epsilon > 0$, we show that, for any sufficiently large $n$ and sufficiently small $\epsilon$, there exists a copy of $H$ where the number of edges in the two colours differ by at most $2$. Moreover, we completely characterise the pairs $(H,\mathbf{c})$ for which the difference of $2$ cannot be improved, refuting a conjecture by Mohr, Pardey, and Rautenbach.
A set $V$ is said to be separated by subsets $V_1,\ldots,V_k$ if, for every pair of distinct elements of $V$, there is a set $V_i$ that contains exactly one of them. Imposing structural constraints on the separating subsets is often necessary for practical purposes and leads to a number of fascinating (and, in some cases, already classical) graph-theoretic problems. In this work, we are interested in separating the vertices of a random graph by path-connected vertex sets $V_1,\ldots,V_k$, jointly forming a separating system. First, we determine the size of the smallest separating system of $G(n,p)$ when $np\to \infty$ up to lower order terms, and exhibit a threshold phenomenon around the sharp threshold for connectivity. Second, we show that random regular graphs of sufficiently high degree can typically be optimally separated by $\lceil \log_2 n\rceil$ sets. Moreover, we provide bounds for the minimum degree threshold for optimal separation of general graphs.
Given an increasing graph property $\mathcal{P}$, a graph $G$ is $\alpha$-resilient with respect to $\mathcal{P}$ if, for every spanning subgraph $H\subseteq G$ where each vertex keeps more than a $(1-\alpha)$-proportion of its neighbours, $H$ has property $\mathcal{P}$. We study the above notion of local resilience with $G$ being a random geometric graph $G_d(n,r)$ obtained by embedding $n$ vertices independently and uniformly at random in $[0,1]^d$, and connecting two vertices by an edge if the distance between them is at most $r$. First, we focus on connectivity. We show that, for every $\varepsilon>0$, for $r$ a constant factor above the sharp threshold for connectivity $r_c$ of $G_d(n,r)$, the random geometric graph is $(1/2-\varepsilon)$-resilient for the property of being $k$-connected, with $k$ of the same order as the expected degree. However, contrary to binomial random graphs, for sufficiently small $\varepsilon>0$, connectivity is not born $(1/2-\varepsilon)$-resilient in $2$-dimensional random geometric graphs. Second, we study local resilience with respect to the property of containing long cycles. We show that, for $r$ a constant factor above $r_c$, $G_d(n,r)$ is $(1/2-\varepsilon)$-resilient with respect to containing cycles of all lengths between constant and $2n/3$. Proving $(1/2-\varepsilon)$-resilience for Hamiltonicity remains elusive with our techniques. Nevertheless, we show that $G_d(n,r)$ is $\alpha$-resilient with respect to Hamiltonicity for a fixed constant $\alpha = \alpha(d)<1/2$.