Now let's discuss coarsening. Imagine taking the system at a
temperature and rapidly (let us say instantaneously for
our purposes) quenching it at *t*=0 to a temperature . The
system begins in the paramagnetic state, with equal amounts of up and
down spins, randomly arranged. After the quench, the new equilibrium
state is ferromagnetic. It must, however, find a way to evolve
dynamically from the low temperature disordered state into the fully
polarized state. From the RG perspective, since the ferromagnetic
state is described by a *T*=0 fixed point, it's actually natural to
try and understand this coarsening process by completely ignoring
thermal fluctuations, i.e. working as if .

At *t*=0, there are then equal amounts of up and down spins in a
random pattern with a scale set by the correlation length at .
The system evolves from this point deterministically. If there is a
small bubble of down spins in a largely up spin domain, it will flip
to lower the local energy. Likewise, the large domain itself may grow
or shrink depending upon its environment. For simplicity, let's think
about a spherical droplet of down spins, of radius *r*, in the midst
of a much larger up spin domain. The energy of the droplet, relative
to a uniform domain, is

where the surface tension . From this, we can determine the total force on the bubble:

This is negative because the bubble tends to collapse. It represents the total force, so that the force per unit area

The local force is thus proportional to the local curvature of the surface. If we take a local equation of motion for the bubble wall,

we easily find that

The bubble thus fully collapses in a time .

In reality, the system has a much more complicated set of domains and
spatial structure, but roughly speaking, there is a characteristic
length scale *R*(*t*), below which the system appears ordered (i.e. all
domains with sizes below *R*(*t*) have collapsed). Since the local
forces are determined by the curvature , we still expect
that this characteristic scale . This is a fairly
rapid (power-law) coarsening, and we consequently expect, among other
things, a scaling for the spin-spin correlations after the quench:

where *F* is a ``universal'' scaling function.

There are many more interesting questions to be asked about coarsening, such as the nature of this function, and anyone who is interested should certainly look at the review article by Allan Bray[2].

Thu May 30 08:21:44 PDT 1996