During the summers of 2006 and 2007, I worked as an undergraduate research assistant with Dr. Paul Harper at Calvin College. I was funded by an undergrad research fellowship from Dr. Jack Kuipers (of quaternion fame). Our goal was to perform ultra coarse-grained simulations of lipid bilayers to see if realistic dynamics could be acheived with minimal compute power.
A lipid was represented by a "chain" of 4-8 point particles in our 2D models. Links in the "chain" were represented by simple harmonic oscillators ($\vec{F} = -k \left(\vec{x} - \vec{x}_0 \right)$) with different stiffnesses and rest lengths. The "head" was a hydrophillic particle and the tail particles were hydrophobic. We simulated inter-particle forces using Lennard-Jones potentials with appropriate coeffcients. We also included individual water particles to simulate immersion in a water bath. Our hope was to achieve realistic phase transitions (lamellar to hexagonal) by adjusting several parameters of this model (rest lengths, number of tail particles, Lennard-Jones coefficients).
We initialized the system in the lamellar phase and gave all particles initial velocities according to a Maxwell-Boltzmann distribution. We first let the system relax at the initial temperature for some time, then very slowly increased the ambient temperature, scaling all velocities so that the net kinetic energy matched the thermal energy for that iteration's prescribed temperature.
The simulations were written in Java (Summer 2006) and C++ (Summer 2007) and carried out on a desktop as well as on a computer science lab. When we ran the simulation on the lab computers, we initialized the simulations slightly differently on each machine to explore parameter space more efficiently.
In our best runs, we were able to achieve phase-transition behavior, though it was too rough to make firm conclusions with the time we had available.