Uniqueness, Spatial Mixing, and Approximation for Ferromagnetic 2-Spin Systems

Abstract

For anti-ferromagnetic 2-spin systems, a beautiful connection has been established, namely that the following three notions align perfectly: the uniqueness in infinite regular trees, the decay of correlations (also known as spatial mixing), and the approximability of the partition function. The uniqueness condition implies spatial mixing, and an FPTAS for the partition function exists based on spatial mixing. On the other hand, non-uniqueness implies some long range correlation, based on which NP-hardness reductions are built. These connections for ferromagnetic 2-spin systems are much less clear, despite their similarities to anti-ferromagnetic systems. The celebrated Jerrum-Sinclair Markov chain [JS93] works even if spatial mixing or uniqueness fails.

We provide some partial answers. We use (β,γ) to denote the (+,+) and (−,−) edge interactions and λ the external field, where βγ>1. If all fields satisfy λ<λ_c (assuming β≤γ), where λ_c=(γ/β)^{(Δ_c+1)/2} and Δ_c=(\sqrt{βγ}+1)/(\sqrt{βγ}−1), then a weaker version of spatial mixing holds in all trees. Moreover, if β≤1, then λ<λ_c is sufficient to guarantee strong spatial mixing and FPTAS. This improves the previous best algorithm, which is an FPRAS based on Markov chains and works for λ<γ/β [LLZ14a]. The bound λ_c is almost optimal. When β≤1, uniqueness holds in all infinite regular trees, if and only if λ≤λ^int_c, where λ^int_c=(γ/β)(⌈Δc⌉+1)/2. If we allow fields λ>λ^int′_c, where λ^int′_c=(γ/β)(⌊Δc⌋+2)/2, then approximating the partition function is #BIS-hard.