Adding ASE LAMMPSlib Calculator

[stefanbringuier]

Was wondering if you have any guidance on how might be able to proceed with using ASE LAMMPSlib. I tried to simply pass the ASE LAMMPSlib calculator object but got deepcopy errors.

Heres the source to the ASE LAMMPSlib implementation:

https://wiki.fysik.dtu.dk/ase/_modules/ase/calculators/lammpslib.html#LAMMPSlib

[jochym]

I nevere tried other calculators but it should be possible. The siesta support was implemented by the otside contributor. The problem with LAMMPS is that it is an internal calculator not an external one (like vasp or abinit). You should be able to get around this problem by actually making a new objects for the deformed structures. Please post the error messages. Maybe I will be able to help.

[stefanbringuier]

I think I have an approach that will work for LAMMPSlib and possibly other internal calculators, it would only enable a serial execution thought. The implementation based on https://github.com/jochym/Elastic/blob/6117c9aeff8859a46860a34057e2f59c71c5a841/parcalc/parcalc.py#L578

Would look something like this:

def SerialCalculate(systems,calc,prefix="Calc_"):

    if type(systems) != type([]) :
        sysl=[systems]
    else :
        sysl=systems
    basedir=os.getcwd()
    res=[]
    for n,s in enumerate(sysl):
        if isinstance(calc, LAMMPSlib):
            s.set_calculator(calc)
        else:
            s.set_calculator(deepcopy(calc))
        s.get_calculator().block=False
        place=tempfile.mkdtemp(prefix=prefix, dir=basedir)
        os.chdir(place)
        s.get_calculator().working_dir=place
        s.get_potential_energy()
        os.chdir(basedir)
        time.sleep(0.2)
        res.append([n,s])

    return [r for ns,s in enumerate(sysl) for nr,r in res if nr==ns]

It seems to work, but I've only tested on a single LAMMPSlib example, would want to try ASE internal EMT and EAM calculators.

...
systems = get_elementary_deformations(structure, n=10, d=0.1)
result = SerialCalculate(systems, calculator)
cij, _ = get_elastic_tensor(structure, systems=result)

Can create a pull request if you think it makes sense.