Implement BaseEnsemble

Description

One of the core elements of the Monte Carlo module is the BaseEnsemble, from which all other ensembles are derived. It must provide basic functionality for running a Monte Carlo simulation as illustrated by the following snippet (see #113 (closed) for a more extensive use case). Here, Ensemble is a child of BaseEnsemble.


# set up simulation
ce = ClusterExpansion.read('mymodel.ce')
atoms = ce.primitive_structure.repeat(4)
mc = Ensemble(atoms, calculator=ce, data_container='myrun.dc', ...)

# attach an observer (child of BaseObserver)
obs = ShortRangeOrder(interval=10, mode='Warren-Cowley', neighbor_shells=3)
mc.attach_observer(obs, tag='sro')

# run MC
mc.run(steps=15000)

Upon initialization (and possibly also after changes to the interval parameters) the BaseEnsemble class must determine the greatest common denominator (gcd) of all the interval values and call the observers accordingly during the run phase.

The BaseEnsemble must initialize and handle the internal DataContainer (#158 (closed)) object. This includes adding metadata

date/time
optional: user account name, machine name

and restart information

complete set of input parameters (latest set in case the parameters are changed during simulation, see below)
calculator (ClusterExpansion)

as well as appending data as the simulation progresses (see #158 (closed)).

The ensemble could also be used to loop over ensemble parameter (e.g., the temperature)

for temperature in range(1000, 100, -10):
    mc.reset_data_container()
    mc.temperature = temperature
    mc.run(steps=15000)

Here, temperature() is a property that is specific to e.g., the canonical, SGC, and VCSGC ensembles. Care must be taken to ensure that observer tags are unique.

Other basic properties/variables to include:

counters for total and accepted trials
interval for storing elementary observables (possibly in sub MC cycle, float?)

Sub-tasks

Demonstration

all tests pass
docstrings complete

Edited Mar 21, 2018 by Mattias Ångqvist