On the tail of the branching random walk local time (Q2032431)

The paper under review is to study the tail of the total number of particles that ever visit a vertex \(x\) under suitable conditions on the offspring distribution. Authors lay main results in Theorem 1.1 and Theorem 1.2 in the introduction and background in the second section. For a branching random walk \((B_n)_{n\ge 0}\) on \(\mathbb Z^d\) (for dimension \(d\ge 1\)) whose offspring distribution \(\mu\) is critical (mean equals 1), nontrivial (as \(\mu (1) \le 1\)), and subexponential (\(\mu (n) \le Ce^{-cn}\) for every \(n\ge 1\)) starting with a single particle at the origin, then \[ cg(n, x, d) \le P_{\mu, 0} (L(x) \ge n)\le C g(n, x, d), \] where \(L(x)\) is the total number of particles that visit x, the explicit function \(g(n, 0, d)\) is given in Theorem 1.1 and \(g(n, x, d)\) for \(x\ne 0\) is given in Theorem 1.2. The work is motivated from analogous questions for the abelian sampling model by \textit{D. Dhar} [Phys. Rev. Lett. 64, No. 14, 1613--1616 (1990; Zbl 0943.82553)] and \textit{S.N. Majumdar} and \textit{D. Dhar} [``Equivalence between the abelian sandpile model and the \(q \to 0\) limit of the Potts model'', Phys. A 185, 129--145 (1992)], where the total number of times the origin topples in an avalanche at equilibrium is expected to behave in a roughly analogous way to the branching random walk local time at the origin. Theorem 1.2 relies on the asymptotics of the hitting probability which belongs to \textit{J.-F. Le Gall} and \textit{S. Lin} [Ann. Probab. 43, No. 5, 2701--2728 (2015; Zbl 1335.60070); J. Inst. Math. Jussieu 15, No. 2, 271--317 (2016; Zbl 1339.60049)] for \(d\ne 4\), and lower bound and upper bound for \(d=4\) goes to LeGall and Lin [loc. cit.] and \textit{Q. Zhu} [Ann. Inst. Henri Poincaré, Probab. Stat. 57, No. 1, 73--93 (2021; Zbl 1491.60061); Electron. Commun. Probab. 24, Paper No. 32, 6 p. (2019; Zbl 1488.60121)], respectively. \textit{J.-F. Le Gall} and \textit{M. Merle} [Electron. Commun. Probab. 11, 252--265 (2006; Zbl 1110.60083)] studied the conditional distribution of the occupation measure of the unit ball for large \(x\), which is very close to Theorem 1.2 of the paper under review. Define the local time \(L_n(x) = \sum_{m=0}^n B_m (x)\) to be the total number of particles that occupy the site \(x\) up to time \(n\), and \(L(x) = \lim_{n\to \infty}L_n (x)\). Section 3 discusses how the moments of the branching random walk local time may be expanded in terms of diagrammatic sums, to shift the problem into some methodology well known to experts. A rooted plane tree is a locally finite tree with a distinguished root vertex and a distinguished linear ordering of the children of each vertex. Proposition 3.1 states diagrammatic expansion of moments for the law of a branching random walk with offspring distribution \(\mu\) started with a single particle \(x_0\), are the sum of each \(k\)-skeleton (\(k\)-labelled rooted plane tree with at least two children) diagram with offspring distribution for all children of its included vertexes (with upper bound up to time \(n\)). The proof follows from the span of the \(k\)-labelled rooted plane tree by unions of geodesics among the root of \(T\) and the vertices with labeling function, and the truncated case from the inclusion of the \(k\)-labelled rooted plane trees with height at most \(n\) to related with modified Green's function. Lemma 3.2 presents a recursive inequality for the maximal diagram from the symmetry feature of the random walk. The proof of Lemma 3.2 follows from understanding three cases with labeling function not a leaf, a leaf and the parent of the labeling function lying on the boundary, and a leaf and the parent of the labeling function lying in the interior. The proof of main results for dimension less than 4 is given in Section 4. The easiest way for the local time to be large in low dimensions is for the genealogical tree to be sufficiently large. If the genealogical tree survives to generation \(k\) with probability \(\Theta (k^{-1})\), then the tree contains roughly \(k^2\) vertices and the locations of the corresponding particles are roughly uniformly distributed on the ball of radius \(\sqrt{k}\). The estimation starts with Lemma 4.3 for second moment case and finite third moment case with using Lemma 3.2. Notice that there are typos in the article, e.g., to refer the non-existing Lemma 3.1, for instance on page 478 in the proof of the second moment case and page 486 on the upper bound in Proposition 5.1. Hence, the proof of Proposition 4.1 (low dimensional case of Theorem 1.1 and 1.2) follows from the Kolmogorov estimate under a third moment assumption for the genealogical tree of the branching process, and the Paley-Zygmund inequality to get the lower bound. The proof of main results for dimension bigger than or equal to 5 is given in Section 5 in Proposition 5.1. The proof for \(d\ge 5\) follows from Lemma 3.2, Lemma 5.1 (the skeleton partition function) and Lemma 5.3 (contribution of a single skeleton) and Markov inequality for the upper bound, and lower bound from the Markov property. The technical Lemma 5.2 is related to a standard generating function by \textit{H. S. Wilf} [Generatingfunctionology. Wellesley, MA: A K Peters (2006; Zbl 1092.05001)] and the Catalan number given the counting of isomorphism classes of rooted plane trees with exactly \(n\) vertices; Lemma 5.3 follows from the recursive inequality in Lemma 3.2 and the fact that the simple random walk bubble diagram converges when dimension \(d\ge 5\). The proof of main results for the critical dimension 4 is given in Section 6 in Proposition 6.1 which relies on the work of Zhu [loc. cit.]. The upper bound follows from Zhu's theorem and Stirling's approximation with Markov inequality, and the lower bound from the Paley-Zygmund inequality, and a technical lemma for dimension 4 analogue of Lemma 5.4 for \(d\ge 5\).