Aarhus University Seal

Infinite WARM graphs III: Strong reinforcement regime

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaperJournal articleResearchpeer-review

DOI

  • Christian Hirsch
  • Mark Holmes, University of Melbourne
  • ,
  • Victor Kleptsyn, Universite de Rennes 1

We study a reinforcement process on graphs G of bounded degree. The model involves a parameter α > 0 governing the strength of reinforcement, and Poisson clock rates λ v at the vertices v of the graph. When the Poisson clock at a vertex v rings, one of the edges incident to it is reinforced, with edge e being chosen with probability proportional to its current count (counts start from 1) raised to the power α. The main problem in such models is to describe the (random) subgraph, consisting of edges that are reinforced infinitely often. In this paper, we focus on the finite connected components of in the strong reinforcement regime (α > 1) with clock rates that are uniformly bounded above. We show here that when α is sufficiently large, all connected components of are trees. When the firing rates λ v are constant, we show that all components are trees of diameter at most 3 when α is sufficiently large, and that there are infinitely many phase transitions as α "1 . For example, on the triangular lattice, increasingly large (odd) cycles appear as α "1 (while on the square lattice no finite component of contains a cycle for any α > 1). Increasingly long paths and other structures appear in both lattices when taking α "1 . In the special case where G=Z and α > 1, all connected components of are finite and we show that the possible cluster sizes are non-monotone in α. We also present several open problems.

Original languageEnglish
JournalNonlinearity
Volume36
Issue6
Pages (from-to)3013-3042
Number of pages30
ISSN0951-7715
DOIs
Publication statusPublished - Jun 2023

    Research areas

  • 37C10, 60J25, 60K35, dynamical systems, probability, reinforcement

See relations at Aarhus University Citationformats

ID: 322436773