Currently I’m working on some simulations and it strikes me as rather difficult to obtain a generally “realistic” model, despite most of it being quite easy to understand.

The system in question is a series of interacting magnetic particles. The basic ingredients are all there; a couple of terms that favor alignment of like spins near by, a term that favors opposite alignment for spins far away, a term to control the external applied field and a few different schemes for introducing disorder, changing the spins, and such. All in all, it’s not too difficult to understand and even to get it into a computer program that does the basics.

Along the way you may encounter a few difficulties if you’re new at writing numerical models (such as how to account for boundaries in your simulation, or how to solve various equations and in which order to do them). But with some practice and a decent library built up over time (or through the use of other people’s libraries) you can have the basic thing going.

What strikes me as hard at the moment is settling on an area of “parameter space” for which the model can describe what we’ve seen experimentally. Ideally what I’d like is the following: to find a set of parameters such that I only need to vary one of them to properly describe things. Now, I’ve got 3 different parameters (well, 4 actually, but I’m forcing one of them to be equal to 1 since I can divide all the other parameters by it). Two of those parameters I think I can link together by a particular relationship and then only have to vary one of them. However, right now everything seems very, very “touchy.” Small adjustments in even one of the 3 parameters can take it from describing things pretty well to looking very meaningless.