Thursday, June 01, 2023

What is a spin glass?

As mentioned previously, structural glasses are materials in which there is no periodic lattice (no long-range spatial order) and the building blocks get "stuck" in some configuration, kinetically unable to get to the true energetic minimum state which would almost certainly be a periodic crystal.  Upon cooling from the liquid state, their viscosity increases by many orders of magnitude (in various ways) until they act like rigid solids.  Distinguishing "glassy" physics includes strongly interacting building blocks, a complicated energy landscape with many local minima, spatial disorder leading to hugely varying interaction strengths and a very broad distribution of relaxation times (so that responses to perturbations aren't simple exponentials in time, but are more slowly decaying functions such as \(-\log t\)).  These slow relaxations are called "aging", and when the system is perturbed (e.g., a sudden stress is applied, or a sudden temperature change is applied and removed), the system's response picks back up ("rejuvenation") before aging again.

Analogs of all of these properties are also seen in spin glasses, which I wrote about a bit in this post about the 2021 Nobel in Physics.  In a spin glass, the degrees of freedom aren't atoms or groups of atoms, but instead are the magnetic moments of particular atoms, such as isolated Fe atoms in a Cu bulk.   The analog of the periodic crystal would be some version of long-range magnetic order.  In a typical spin glass, the magnetic atoms are positioned randomly in a non-magnetic host, so that the magnetic interactions between neighbors are strong, but often random in sign and strength due to disorder.  As a result, the magnetic system has a complicated energy landscape with many minima (corresponding to configurations with similar energies but it would cost significant energy to rearrange the spins to get from one local energy minimum configuration to another).  These systems show aging, rejuvenation, etc.

The universality of glassy dynamics across such microscopically different systems is one of those remarkable emergences that crops up in condensed matter.  Despite the very different microscopic physics, there is some deeper organizing principle at work that leads to these properties.  

Spin glasses have attracted quite a bit of interest for a couple of reasons.  First, they are comparatively easy to study, since magnetic properties and their time evolution are usually easier to measure than detailed microscopic structural arrangements in structural glasses.  Second, it is possible to create models of spin glasses in a variety of systems, including using qubits.  Spin glasses can also be mapped to certain kinds of optimization problems (see this pdf news article).

Interestingly, a recent paper in Nature (arxiv version) by folks at D-Wave has used their 5000 qubit gadget to do a quantum simulation of a spin glass.  They can program the interactions among the qubits and make them random and frustrated as in a spin glass.  In small test configurations, they show that they can see (at short times, anyway) quantum coherent dynamics that agree with calculations.  They can then look at much larger systems, well beyond traditional calculational practicality, and see what happens.  I don't know enough about the system to evaluate this critically, but it looks like a very nice platform.  (They’ve come along way from when their founder used to argue and insult in blog comments.  They now show as anonymous, but the one from Geordie Rose is clear from context.)

1 comment:

Anonymous said...

The Ising model from a computer science standpoint is really useful because it sets up a general constraint satisfiability-type problem (SAT). Many versions of SAT type problems are NP complete, as is the Ising model in 3D. This really comes down to non-planarity, so Ising models with next nearest coupling and beyond are also NP complete.

The mind boggling aspect of DWave is that they've somehow stayed in business for 20+ yrs without solving a "useful" example of spin physics or optimization. Can anyone explain how that's possible? This is not purely criticism, I'm honestly curious how a company can manage to do that for so long