Mixing time of critical Ising model on trees is polynomial in the height (Q981702): Difference between revisions
From MaRDI portal
ReferenceBot (talk | contribs) Changed an Item |
Normalize DOI. |
||
Property / DOI | |||
Property / DOI: 10.1007/s00220-009-0978-y / rank | |||
Property / DOI | |||
Property / DOI: 10.1007/S00220-009-0978-Y / rank | |||
Normal rank |
Latest revision as of 11:12, 10 December 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Mixing time of critical Ising model on trees is polynomial in the height |
scientific article |
Statements
Mixing time of critical Ising model on trees is polynomial in the height (English)
0 references
2 July 2010
0 references
This paper considers the single-site heath-bath Glauber dynamics for the Ising model on a regular finite \(b\)-ary tree with \(h\) levels, at or close to the inverse temperature \(\beta_c=\text{arctanh}(1/\sqrt b)\), which is the critical inverse temperature for the purity of the free-boundary condition state. It is known that for \(\beta<\beta_c\) the spectral gap of the generator of the dynamics is of order \(1\) (it does not grow with \(h\)) and is exponential in \(h\) if \(\beta>\beta_c\). Theorem 1 states that, for \(\beta=\beta_c\) and for arbitrary boundary conditions \(\tau\), the inverse spectral gap and the total-variation mixing time are upper bounded by a polynomial in \(h\) and the exponent of the polynomial is independent of both \(\tau\) and \(b\). Moreover, Theorem 2 extends this result to the critical window where \(\beta-\beta_c\) goes to zero when \(h\to\infty\). Loosely speaking, it is proven here that the inverse spectral gap and mixing time are upper bounded by \(\exp(const\times (\beta-\beta_c)h)\), which matches with the known exponential behavior when \(\beta>\beta_c\). Finally, in Theorem 3 is given a \textit{lower bound} on the inverse spectral gap (resp. mixing time), of order \(h^2\) (resp. \(h^3\)), always at \(\beta_c\) but with special boundary conditions (free b.c.). This disproves an earlier conjecture by N. Berger et al. (PTRF '05), according to which the inverse gap would be linear in \(h\).
0 references
Ising model on trees
0 references
gap
0 references
mixing time
0 references
critical temperature
0 references