A backbone transfer principle for cluster-size decay beyond power laws in spatial random graphs
We give a finite-box transfer principle for spatial random graphs with independent continuous marks and an increasing pair rule. A connected backbone at the volume-k scale gives a second-largest-component bound at speed kF̄(w), where F̄(w) is the mark tail at the backbone level. The transfer needs neither regular variation nor local seeds or upper kernel-growth bounds. For rapidly assortative kernels of the form (a ∨ b) exp{g(a ∧ b)}, with g regularly varying and g(t) = O(t), the matching finite-cluster speed is asymptotic to k times the root-degree tail at k. For g(t) = t^δ and Pareto marks it has order k/(log k)^(b/δ). We also obtain an exact cutoff criterion in the strict high–high regime beyond regular variation of g, sharp logarithmic corrections for stretched-exponential spatial profiles, and high-density surface upper bounds uniform over arbitrary finite mark assignments and bounded Cox media.
Authors
- Shardul Mishra (ORCID: https://orcid.org/0009-0005-4164-7569)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-14
- DOI
- https://doi.org/10.5281/zenodo.24224303
- Primary Topic
- Stochastic processes and statistical mechanics
- Type
- preprint