An isothermal-isobaric ensemble Metropolis Monte Carlo code for the Citrine Technical Challenge (#2). Written in C++ by an obviously Java trained programmer.
Two main class clusters: Moves and Shapes
The driver class MCDriver populates a list of Shapes, initializes a Cell, and performs MC Moves.
All collision detection is handled in the derived shape classes (Sphere and Tetrahedron) with help from "Collisions.cpp" pulled from the challenge site.
Results of the simulation are printed to files with the following name convention: output/best%04d. The configuration for the optimal packing found will be in the last file in this list.
Best recorded Sphere packing (6 spheres): 73.33%
Best recorded Tetrahedron packing (4 tetrahedra): 82.5% - a bit below the best known packing of ~85% (and not the same structure).
There are a number of cmd line arguments for setting runtime variables:
n_particles - Number of particles in the cell
n_steps - Number of MC moves to perform
n_drivers - Number of parallel MCDrivers to run (each system runs simultaneously, with the occasional parallel tempering move attempted to swap the pressures of two randomly selected systems)
p{i} - For each driver (of which there are n_drivers), a corresponding p{i} (p0, p1, p2, etc) must be supplied which sets the initial pressure of each system.
p_cell_move - Probability of choosing a cell move v.s. a particle move
ProjectionThreshold - A stability parameter limiting the internal angles of the fundamental cell from becoming too acute. If your MC trajectories diverge, use a smaller value. If you don't find dense solutions, try a larger value. I've found .7 to be a good compromise.
dcell - Maximum size of a cell shape move. (Larger values lead to less efficient MC sampling and high move rejection, smaller values lead to poor phase space sampling).
dr - Maximum particle displacement/rotation move size (measured in radians for rotations, edge lengths for displacements).
First, you'll have to compile it. Assuming you have the standard libraries installed with gcc 4.7 or higher, the project should compile by just typing make in the root.
Example Usage (with suggested values):
./main n_particles 4 n_steps 3000000 n_drivers 4 p0 50 p1 250 p2 500 p3 1000 dcell .01 dr .02 ProjectionThreshold 0.7
note the code can be run with Tetrahedra or Spheres, but this choice must be made by changing ChosenShape in src/main.cpp and recompiling.