README.md

January 23, 2025 ยท View on GitHub

Code for Multi-state MASH

Overview

This directory contains

  • A main Python script mash.py
  • A script model.py with functions for setting up the model, handling input/output, sampling initial variables, and running a debug trajectory
  • A Fortran source code in src/ that runs the trajectories These allow you to run MASH for various potentials, initial conditions and observables.

How to compile

Prerequisites:

  • gfortran (it also compiles and runs with ifort, but you will need to manually update the makefile)
  • lapack
  • numpy.f2py for Python3

Compile by typing make [option] and one of the following options

  • clean Remove compiled files and restart from clean directory
  • fast Produces fast parallelized code
  • debug Uses plenty of warning flags and should tell at which line the code breaks.

Also check that you have the required Python packages installed. One way to ensure this is to (preferably in a virtual environment) run pip install -r requirements.txt.

How to run

mash.py +[args].in where [args].in is an argument file with one line per argument (allowing commented lines). See examples for example input files. There are plenty of available option flags and you can add more to suit your system.

Hints:

  • Make sure mash.py is executable, otherwise chmod u+x mash.py
  • Create a symbolic link to mash.py in some place in your PATH, so that you can call it from anywhere.
  • You may need to add the directory to your PYTHONPATH.

What comes in

Run mash.py -h to see the available options. In general you should know about the following:

  • model String specifying your model system.
  • Model-specific parameters like beta, Delta etc.
  • init Integer that specifies initial state (index starts at 0)
  • dt Timestep
  • nt Number of timesteps
  • ntraj Number of trajectories
  • nucsamp Nuclear sampling option (see "sampling" in model.py)
  • elsamp Electronic sampling option (see "sampling" in model.py)
  • obstyp Specify what kind of observables you want to measure (e.g. pop for populations)

What comes out

Depends on what obstyp you specified, but for pop you should see a file pop.out which contains time in the first column and then the state populations in the following columns.

Add a new potential

The Fortran code contains a few potentials, e.g. linvib and tully. If you want to create a different potential, add another file with the subroutines pot and grad specifying the diabatic potential and gradient. Copy the initialization routine from one of the existing potentials, and add a potential-specific init subroutine to f2py.f90.

Problems?

Please don't hesitate contacting me if you discover any bugs. (Thanks to Eric Koessler, Rochester; Annina Lieberherr, Oxford; Bokang Huo, Berkeley)