作者: Lucas Nicolao , Giorgio Parisi , Federico Ricci-Tersenghi
DOI: 10.1103/PHYSREVE.89.032127
关键词: Physics 、 Statistical physics 、 Exponent 、 Quantum mechanics 、 Measure (mathematics) 、 Monte Carlo method 、 Spin glass 、 Symmetry breaking 、 Spatial correlation 、 Phase (waves) 、 Gaussian
摘要: The study of the low temperature phase spin glass models by means Monte Carlo simulations is a challenging task, because very slow dynamics and severe finite-size effects they show. By exploiting at best capabilities standard modern CPUs (especially streaming single instruction, multiple data extensions), we have been able to simulate four-dimensional Edwards-Anderson model with Gaussian couplings up sizes $L=70$ for times long enough accurately measure asymptotic behavior. quenching systems different critical temperatures in whole phase, identify regime where are negligible: $\ensuremath{\xi}(t)\ensuremath{\lesssim}L/7$. Our estimates dynamical exponent ($z\ensuremath{\simeq}1/T$) replicon ($\ensuremath{\alpha}\ensuremath{\simeq}1.0$ $T$ independent), that controls decay spatial correlation zero overlap sector, consistent replica symmetry breaking theory, but latter differs from theoretically conjectured value.