arXiv subjects
Publications and source records attributed to Oliver Riordan.
The Erd\H os--Sós Conjecture was recently proved by GPT-6 Astra, using a very ingenious and surprising argument. In this note, we present a simplified version of this argument in an (arguably) more natural form. We also determine the extremal graphs for the Erd\H os--Sós Conjecture, and prove a related conjecture of Addario-Berry, Havet, Linhares Sales, Reed and Thomassé, again determining the extremal graphs.
Achlioptas processes such as the Bohman--Frieze process are much harder to analyse than the classical Erdős--Rényi process, due to the dependence between edges added at different stages. This dependence means that most analysis so far is dynamic, often based on the differential equation method. In the Erdős--Rényi case there is an alternative static approach, pioneered by Erdős, Rényi and Bollobás, based on evaluating the expectation (and higher moments) of various subgraph counts, and using this to study the component structure. Here we show that this latter approach can be applied (with some complications) to the Bohman--Frieze process. For example, we are able to show that the expected number $μ_{k,t,n}$ of $k$-vertex tree components after $tn$ steps satisfies (essentially) $μ_{k,t,n}=c_{k,t}n(1+O(k/\sqrt{n}))$. Our method gives a very complicated formula for $c_{k,t}$, which seems to be unusable. However, since $c_{k,t}$ does not depend on $n$, we may use recent results obtained by the differential equation method and branching process analysis to find the asymptotics of $c_{k,t}$ as $k\to\infty$. The latter results also give a formula for $μ_{k,t,n}$ of the form $c_{k,t}n$ plus an error term, with a much more usable description of $c_{k,t}$ but a much worse error term. We combine the best of both worlds to prove a number of new results about the process near criticality. In particular, we obtain extremely sharp bounds on the size of the largest non-giant component near criticality, including the limiting distribution of its fluctuations.
We study the distribution of the set of copies of some given graph $H$ in the random graph $G(n,p)$, focusing on the case when $H = K_r$. Our main results capture the 'leading term' in the difference between this distribution and the 'independent hypergraph model', where (in the case $H = K_r$) each copy is present independently with probability $π= p^{\binom{r}{2}}$. As a concrete application, we derive a new upper bound on the number of $K_r$-factors in $G(n,p)$ above the threshold for such factors to appear. We will prove our main results in a much more general setting, so that they also apply to random hypergraphs, and also (for example) to the case when $p$ is constant and $r = r(n) \sim 2\log_{1/p}(n)$.
Perhaps the best understood phase transition is that in the component structure of the uniform random graph process introduced by Erdős and Rényi around 1960. Since the model is so fundamental, it is very interesting to know which features of this phase transition are specific to the model, and which are `universal', at least within some larger class of processes (a `universality class'). Achlioptas process, a class of variants of the Erdős--Rényi process that are easy to define but difficult to analyze, have been extensively studied from this point of view. Here, settling a number of conjectures and open problems, we show that all `bounded-size' Achlioptas processes share (in a strong sense) all the key features of the Erdős--Rényi phase transition. We do not expect this to hold for Achlioptas processes in general.
Let $p_\mathrm{c}$ and $q_\mathrm{c}$ be the threshold and the expectation threshold, respectively, of an increasing family $\mathcal{F}$ of subsets of a finite set $X$, and let $l$ be the size of a largest minimal element of $\mathcal{F}$. Recently, Park and Pham proved the Kahn-Kalai conjecture, which says that $p_\mathrm{c} \leqslant K q_\mathrm{c} \log_2 l$ for some universal constant $K$. Here we slightly strengthen their result by showing that $p_\mathrm{c} \leqslant 1 - \mathrm{e}^{-K q_\mathrm{c} \log_2 l}$. The idea is to apply the Park-Pham Theorem to an appropriate `cloned' family $\mathcal{F}_k$, reducing the general case (of this and related results) to the case where the individual element probability $p$ is small.
How does the chromatic number of a graph chosen uniformly at random from all graphs on $n$ vertices behave? This quantity is a random variable, so one can ask (i) for upper and lower bounds on its typical values, and (ii) for bounds on how much it varies: what is the width (e.g., standard deviation) of its distribution? On (i) there has been considerable progress over the last 45 years; on (ii), which is our focus here, remarkably little. One would like both upper and lower bounds on the width of the distribution, and ideally a description of the (appropriately scaled) limiting distribution. There is a well known upper bound of Shamir and Spencer of order $\sqrt{n}$, improved slightly by Alon to $\sqrt{n}/\log n$, but no non-trivial lower bound was known until 2019, when the first author proved that the width is at least $n^{1/4-o(1)}$ for infinitely many $n$, answering a longstanding question of Bollobás. In this paper we have two main aims: first, we shall prove a much stronger lower bound on the width. We shall show unconditionally that, for some values of $n$, the width is at least $n^{1/2-o(1)}$, matching the upper bounds up to the error term. Moreover, conditional on a recently announced sharper explicit estimate for the chromatic number, we improve the lower bound to order $\sqrt{n} \log \log n /\log^3 n$, within a logarithmic factor of the upper bound. Secondly, we will describe a number of conjectures as to what the true behaviour of the variation in $χ(G_{n,1/2})$ is, and why. The first form of this conjecture arises from recent work of Bollobás, Heckel, Morris, Panagiotou, Riordan and Smith. We will also give much more detailed conjectures, suggesting that the true width, for the worst case $n$, matches our lower bound up to a constant factor. These conjectures also predict a Gaussian limiting distribution.
We show that for each $r\ge 4$, in a density range extending up to, and slightly beyond, the threshold for a $K_r$-factor, the copies of $K_r$ in the random graph $G(n,p)$ are randomly distributed, in the (one-sided) sense that the hypergraph that they form contains a copy of a binomial random hypergraph with almost exactly the right density. Thus Jeff Kahn's recent asymptotically sharp bound for the threshold in Shamir's hypergraph matching problem implies a corresponding bound for the threshold for $G(n,p)$ to contain a $K_r$-factor. The case $r=3$ is more difficult, and has been settled by Annika Heckel. We also prove a corresponding result for $K_r^{(t)}$-factors in random $t$-uniform hypergraphs, as well as (in some cases weaker) generalizations replacing $K_r$ by certain other (hyper)graphs.
Most social, technological and biological networks are embedded in a finite dimensional space, and the distance between two nodes influences the likelihood that they link to each other. Indeed, in social systems, the chance that two individuals know each other drops rapidly with the distance between them; in the cell, proteins predominantly interact with proteins in the same cellular compartment; in the brain, neurons mainly link to nearby neurons. Most modeling frameworks that aim to capture the empirically observed degree distributions tend to ignore these spatial constraints. In contrast, models that account for the role of the physical distance often predict bounded degree distributions, in disagreement with the empirical data. Here we address a long-standing gap in the spatial network literature by deriving several key network characteristics of spatial networks, from the analytical form of the degree distribution to path lengths and local clustering. The mathematically exact results predict the existence of two distinct phases, each governed by a different dynamical equation, with distinct testable predictions. We use empirical data to offer direct evidence for the practical relevance of each of these phases in real networks, helping better characterize the properties of spatial networks.
We present a short, self-contained, and purely combinatorial proof of Linnik's theorem: for any $\varepsilon > 0$ there exists a constant $C_\varepsilon$ such that for any $N$, there are at most $C_\varepsilon$ primes $p \leqslant N$ such that the least positive quadratic non-residue modulo $p$ exceeds $N^\varepsilon$.
In 1990 Bender, Canfield and McKay gave an asymptotic formula for the number of connected graphs on $[n]=\{1,2,\ldots,n\}$ with $m$ edges, whenever $n\to\infty$ and $n-1\le m=m(n)\le \binom{n}{2}$. We give an asymptotic formula for the number $C_r(n,m)$ of connected $r$-uniform hypergraphs on $[n]$ with $m$ edges, whenever $r\ge 3$ is fixed and $m=m(n)$ with $m/n\to\infty$, i.e., the average degree tends to infinity. This complements recent results of Behrisch, Coja-Oghlan and Kang (the case $m=n/(r-1)+Θ(n)$) and the present authors (the case $m=n/(r-1)+o(n)$, i.e., `nullity' or `excess' $o(n)$). The proof is based on probabilistic methods, and in particular on a bivariate local limit theorem for the number of vertices and edges in the largest component of a certain random hypergraph. The arguments are much simpler than in the sparse case; in particular, we can use `smoothing' techniques to directly prove the local limit theorem, without needing to first prove a central limit theorem.
We consider branching processes consisting of particles (individuals) of two types (type L and type S) in which only particles of type L have offspring, proving estimates for the survival probability and the (tail of) the distribution of the total number of particles. Such processes are in some sense closer to single- than to multi-type branching processes. Nonetheless, the second, barren, type complicates the analysis significantly. The results proved here (about point and survival probabilities) are a key ingredient in the analysis of bounded-size Achlioptas processes in a recent paper by the last two authors.
Jigsaw percolation is a model for the process of solving puzzles within a social network, which was recently proposed by Brummitt, Chatterjee, Dey and Sivakoff. In the model there are two graphs on a single vertex set (the `people' graph and the `puzzle' graph), and vertices merge to form components if they are joined by an edge of each graph. These components then merge to form larger components if again there is an edge of each graph joining them, and so on. Percolation is said to occur if the process terminates with a single component containing every vertex. In this note we determine the threshold for percolation up to a constant factor, in the case where both graphs are Erdős--Rényi random graphs.
A rack on $[n]$ can be thought of as a set of maps $(f_x)_{x \in [n]}$, where each $f_x$ is a permutation of $[n]$ such that $f_{(x)f_y} = f_y^{-1}f_xf_y$ for all $x$ and $y$. In 2013, Blackburn showed that the number of isomorphism classes of racks on $[n]$ is at least $2^{(1/4 - o(1))n^2}$ and at most $2^{(c + o(1))n^2}$, where $c \approx 1.557$; in this paper we improve the upper bound to $2^{(1/4 + o(1))n^2}$, matching the lower bound. The proof involves considering racks as loopless, edge-coloured directed multigraphs on $[n]$, where we have an edge of colour $y$ between $x$ and $z$ if and only if $(x)f_y = z$, and applying various combinatorial tools.
As proved by Régnier and Rösler, the number of key comparisons required by the randomized sorting algorithm QuickSort to sort a list of $n$ distinct items (keys) satisfies a global distributional limit theorem. Fill and Janson proved results about the limiting distribution and the rate of convergence, and used these to prove a result part way towards a corresponding local limit theorem. In this paper we use a multi-round smoothing technique to prove the full local limit theorem.
Recently, we adapted exploration and martingale arguments of Nachmias and Peres, in turn based on ideas of Martin-Löf, Karp and Aldous, to prove asymptotic normality of the number $L_1$ of vertices in the largest component $C$ of the random $r$-uniform hypergraph throughout the supercritical regime. In this paper we take these arguments further to prove two new results: strong tail bounds on the distribution of $L_1$, and joint asymptotic normality of $L_1$ and the number $M_1$ of edges of $C$. These results are used in a separate paper "Counting connected hypergraphs via the probabilistic method" to enumerate sparsely connected hypergraphs asymptotically.
In 1990 Bender, Canfield and McKay gave an asymptotic formula for the number of connected graphs on $[n]$ with $m$ edges, whenever $n$ and the nullity $m-n+1$ tend to infinity. Asymptotic formulae for the number of connected $r$-uniform hypergraphs on $[n]$ with $m$ edges and so nullity $t=(r-1)m-n+1$ were proved by Karoński and Łuczak for the case $t=o(\log n/\log\log n)$, and Behrisch, Coja-Oghlan and Kang for $t=Θ(n)$. Here we prove such a formula for any $r\ge 3$ fixed, and any $t=t(n)$ satisfying $t=o(n)$ and $t\to\infty$ as $n\to\infty$. This leaves open only the (much simpler) case $t/n\to\infty$, which we will consider in future work. ( arXiv:1511.04739 ) Our approach is probabilistic. Let $H^r_{n,p}$ denote the random $r$-uniform hypergraph on $[n]$ in which each edge is present independently with probability $p$. Let $L_1$ and $M_1$ be the numbers of vertices and edges in the largest component of $H^r_{n,p}$. We prove a local limit theorem giving an asymptotic formula for the probability that $L_1$ and $M_1$ take any given pair of values within the `typical' range, for any $p=p(n)$ in the supercritical regime, i.e., when $p=p(n)=(1+ε(n))(r-2)!n^{-r+1}$ where $ε^3n\to\infty$ and $ε\to 0$; our enumerative result then follows easily. Taking as a starting point the recent joint central limit theorem for $L_1$ and $M_1$, we use smoothing techniques to show that `nearby' pairs of values arise with about the same probability, leading to the local limit theorem. Behrisch et al used similar ideas in a very different way, that does not seem to work in our setting. Independently, Sato and Wormald have recently proved the special case $r=3$, with an additional restriction on $t$. They use complementary, more enumerative methods, which seem to have a more limited scope, but to give additional information when they do work.
Graph bootstrap percolation is a simple cellular automaton introduced by Bollobás in 1968. Given a graph $H$ and a set $G \subseteq E(K_n)$ we initially "infect" all edges in $G$ and then, in consecutive steps, we infect every $e \in K_n$ that completes a new infected copy of $H$ in $K_n$. We say that $G$ percolates if eventually every edge in $K_n$ is infected. The extremal question about the size of the smallest percolating sets when $H = K_r$ was answered independently by Alon, Kalai and Frankl. Here we consider a different question raised more recently by Bollobás: what is the maximum time the process can run before it stabilizes? It is an easy observation that for $r=3$ this maximum is $\lceil \log_2 (n-1) \rceil $. However, a new phenomenon occurs for $r=4$ when, as we show, the maximum time of the process is $n-3$. For $r \geq 5$ the behaviour of the dynamics is even more complex, which we demonstrate by showing that the $K_r$-bootstrap process can run for at least $n^{2-\varepsilon_r}$ time steps for some $\varepsilon_r$ that tends to $0$ as $r \to \infty$.
In 1998, Molloy and Reed showed that, under suitable conditions, if a sequence of degree sequences converges to a probability distribution $D$, then the size of the largest component in corresponding $n$-vertex random graph is asymptotically $ρ(D)n$, where $ρ(D)$ is a constant defined by the solution to certain equations that can be interpreted as the survival probability of a branching process associated to $D$. There have been a number of papers strengthening this result in various ways; here we prove a strong form of the result (with exponential bounds on the probability of large deviations) under minimal conditions.