XDMC

XDMC is a code that implements eXchange-Diffusion Monte Carlo, as described here:
https://link.aps.org/doi/10.1103/PhysRevE.102.042105
An open access version is available here:
https://arxiv.org/abs/2001.02215

Compiling

To build the executable, run build.py (which can be found in the root directory). This will generate code, compile and link an executable (called xdmc) in the root directory. This will detect the number of cores on your system and use them all to speed up compilation. Unit tests can be ran (in a few seconds), either in serial or parallel, in the following manner:

    $ xdmc -t # serial
    $ mpirun xdmc -t # parallel

Example usage

The xdmc executable requires a single input file, called simply "input". The program must be executed in the same directory as this file. It can be invoked in serial, or in parallel with MPI:

    $ xdmc # serial
    $ mpirun xdmc # parallel with MPI

The input file contains a list of keywords, information about which can be obtained by invoking xdmc with the -h option

    $ xdmc -h
    ... info about input parameters

The input file contains the algorithm settings (number of walkers, timestep, iterations to carry out etc.) as well as the description of the system to simulate. The system to simulate consists (optionally) of the specification of an external potential and the specification of (at least one) quantum particle, the wavefunction of which will be sampled. An example input file for a lithium atom is shown below:

    # Lines starting with any of #!// will be treated as comments
    
    # Algorithm settings
    dimensions          3
    walkers             10000
    iterations          10000
    tau                 0.001
    tau_nodes           1.0
    diffusion_scheme    stochastic_nodes_mpi
    pre_diffusion       0.5
    coulomb_softening   0.00001
    max_weight          4

    # Atomic potential with charge 3 at x=0, y=0, z=0
    atomic_potential 3 0 0 0

    # Quantum particles to treat with monte carlo, described by
    # a mass, charge and spin (spin is in units of 1/2. The name is just a label).
    #        name      mass  charge  spin  Initial position
    particle electron  1     -1       1    0 0 0
    particle electron  1     -1       1    0 0 0
    particle electron  1     -1      -1    0 0 0

Running this input file will produce a variety of output files, listed below. Some types of output will be distributed to different files for each process. These have the PID of the process appended (e.g wavefunction_0 is the wavefunction file for the root process).

• progress File updated with a high-level report of the progress of the calculation (human readable).
  
• evolution File containing the evaluation of expectation values for each DMC iteration.
  
• wavefunction_n File containing all of the walker configurations for each iteration (large).
  
• nodal_surface_n File containing the configurations of walkers that were killed due to crossing a nodal surface (not written by default).
  

Analysis

Various analysis scripts can be found in the /src/scripts directory. The simplest of these is the plot_evolution.py script, which will plot the expectation values calculated (including the energy/walker population) vs. DMC timestep.