download report - Sapienza
download report - Sapienza
download report - Sapienza
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Scientific Report 2007-2009<br />
Theoretical physics<br />
T12. Statistical mechanics of disordered systems and renormalization<br />
group<br />
The behavior of strongly disordered systems is very<br />
different from the one of pure, homogeneous systems.<br />
Experimentally, the most dramatic effects are observed<br />
in the dynamics. Very slow relaxation and aging, severe<br />
nonequilibrium effects, memory and oblivion, generalizations<br />
of the usual fluctuation-dissipation relations are all<br />
specific features of disordered systems. These dynamic<br />
effects have a static counterpart. For example, in spin<br />
glasses the onset of the slow relaxation is associated with<br />
the divergence of a static quantity, the nonlinear susceptibility.<br />
Unfortunately, our understanding of the static behavior<br />
of strongly disordered systems is rather limited with<br />
the notable exceptions of the Sherrington-Kirkpatrick<br />
model and of Derrida’s random energy model. In most of<br />
the cases, Monte Carlo simulations provide the only tool<br />
to determine the critical behavior and to sort out the<br />
different theories which have been proposed to describe<br />
these systems.<br />
In the last few years we have performed several numerical<br />
studies of the three-dimensional Edwards-Anderson<br />
model. The use of the most advanced numerical techniques<br />
— the random-exchange or parallel-tempering<br />
method, multi-spin coding, cluster algorithms, etc. —<br />
and of very fast computers allowed us to address longstanding<br />
problems and to obtain several new and important<br />
results.<br />
Numerical simulations of random systems are notoriously<br />
very difficult and, in spite of significant algorithmic<br />
progress, numerical simulations are limited to relatively<br />
small system sizes. A significant improvement has<br />
been obtained by developing a new dedicated machine<br />
(JANUS) in collaboration with the University of Ferrara<br />
and several Spanish research groups. JANUS is a<br />
modular, massively parallel, and reconfigurable FPGAbased<br />
computing system. JANUS is tailored for, but<br />
not limited to, the requirements of a class of hard scientific<br />
applications characterized by regular code structure,<br />
unconventional data manipulation instructions, and nottoo-large<br />
database size. In particular, the machine is<br />
well suited for numerical simulations of spin glasses.<br />
On this class of applications JANUS achieves impressive<br />
performances: in some cases one JANUS processing<br />
element outperfoms high-end PCs by a factor of approximately<br />
1000. Several simulations have been performed<br />
on Janus. The critical behavior of the four-state<br />
commutative random-permutation glassy Potts model in<br />
three and four dimensions and of the four-state threedimensional<br />
Potts model have been carefully studied.<br />
More importantly, the use of JANUS allowed us to<br />
study carefully the relaxational dynamics in the threedimensional<br />
Edwards-Anderson model [1], for a time<br />
spanning 11 orders of magnitude, thus approaching the<br />
experimentally relevant scale (i.e., seconds).<br />
One of the most peculiar properties of the mean-field<br />
solution of the Edwards-Anderson model is the so-called<br />
ultrametricity: In the low-temperature phase thermodynamic<br />
states are organized in a hierarchical structure.<br />
One of the long-standing questions is whether such a<br />
structure also holds in the three-dimensional model or<br />
instead is a peculiarity of the mean-field solution as predicted<br />
by the droplet theory. In [2] we studied numerically<br />
the issue and found good evidence for the presence<br />
of an ultrametric structure also in the three-dimensional<br />
case.<br />
Given the difficulty in obtaining clear-cut results for<br />
the three-dimensional spin glass, we also investigated<br />
several spin-glass models which share some of the<br />
properties of the finite-dimension Edwards-Anderson<br />
model, but, at the same time, are significantly simpler<br />
to simulate numerically. For this purpose we introduced<br />
a one-dimensional spin-glass model with long-range interactions.<br />
The interaction between two spins a distance<br />
r apart is either ±1 with a probability that decays with<br />
r as 1/r ρ , or zero. Depending on the exponent ρ, the<br />
model may or may not show mean-field behavior: for<br />
ρ ≤ 4/3 the mean-field approximation is exact, for ρ > 2<br />
no phase transition occurs, while in between the behavior<br />
is nontrivial. Since this model is one-dimensional<br />
and, in spite of the presence of long-range interactions,<br />
each spin only interacts with a finite number of different<br />
(may be far) spins, it is possible to simulate quite large<br />
systems and carefully investigate finite-size effects. In<br />
[3] we studied numerically the model in the absence of<br />
magnetic field for values of p in the intermediate range,<br />
identified the paramagnetic-glassy phase transition, and<br />
characterized the low-temperature phase. We found<br />
both static and dynamic indications in favor of the<br />
so-called replica-symmetry breaking theory. In [4] we<br />
considered the behavior in an external magnetic field<br />
h. The results, obtained by means of a new analysis<br />
method, strongly suggest the presence of a finite-h<br />
transition, as also observed in the mean-field solution of<br />
the Edwards-Anderson model.<br />
References<br />
1. F. Belletti et al., Phys. Rev. Lett. 101, 157201 (2008).<br />
2. P. Contucci et al., Phys. Rev. Lett. 99, 057206 (2007).<br />
3. L. Leuzzi et al., Phys. Rev. Lett. 101, 107203 (2008).<br />
4. L. Leuzzi et al., Phys. Rev. Lett. 103, 267201 (2009).<br />
Authors<br />
G. Parisi, E. Marinari, A. Pelissetto, F. Ricci-Tersenghi, A.<br />
Cavagna, 3 I. Giardina, 3 L. Leuzzi, 3 A. Maiorano, S. Perez<br />
<strong>Sapienza</strong> Università di Roma 35 Dipartimento di Fisica