Repository logo
Log in(current)
Repository logoMIT Open ScholarshipDSpace@MIT
  1. Home
  2. MIT Open Access Articles
  3. MIT Open Access Articles
  4. Sequential Cavity Method for Computing Free Energy and Surface Pressure

Sequential Cavity Method for Computing Free Energy and Surface Pressure

Thumbnail Image
Download
Name

0807.1551v1.pdf

Description
http://arxiv.org/abs/0807.1551
Size

460.65 KB

Format

Adobe PDF

Checksum (MD5)

0a41a064bdeb67169e7a899ce8dc3d52

Author(s)
Gamarnik, David
•
Katz, Dmitriy
Date Issued
October 2009
Journal
Journal of Statistical Physics
Citation
Gamarnik, David, and Dmitriy Katz. “Sequential Cavity Method for Computing Free Energy and Surface Pressure.” Journal of Statistical Physics 137.2 (2009) : 205-232. Copyright © 2009, Springer Science+Business Media, LLC
Version
Author's final manuscript
Abstract
We propose a new method for the problems of computing free energy and surface pressure for various statistical mechanics models on a lattice Zd [Z superscript d]. Our method is based on representing the free energy and surface pressure in terms of certain marginal probabilities in a suitably modi ed sublattice of Zd [Z superscript d]. Then recent deterministic algorithms for computing marginal probabilities are used to obtain numerical estimates of the quantities of interest. The method works under the assumption of Strong Spatial Mixing (SSP), which is a form of a correlation decay. We illustrate our method for the hard-core and monomer-dimer models, and improve several earlier estimates. For example we show that the exponent of the monomer-dimer coverings of Z3 [Z superscript 3] belongs to the interval [0:78595; 0:78599], improving best previously known estimate of (approx- imately) [0:7850; 0:7862] obtained in [FP05],[FKLM]. Moreover, we show that given a target additive error e [epsilon] > 0, the computational e ffort of our method for these two models is (1=e)O(1) [(1 / epsilon) superscript 0(1)] both for free energy and surface pressure. In contrast, prior methods, such as transfer matrix method, require exp ((1/e)0/1) [((1 / epsilon) superscript 0(1))] computation e ffort.
MIT Department
Massachusetts Institute of Technology. Operations Research Center
Sloan School of Management
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
http://hdl.handle.net/1721.1/65918
DOI of Published Version
https://doi.org/10.1007/s10955-009-9849-3
Repository logo
PrivacyPermissionsAccessibilityContact us
Repository logo
Notify us about copyright concerns.