Issue with isolated systems

[mfherbst]

Empirical potentials does not seem to work with isolated systems. Here is a reproducer. The first example work the second errors out with an InexactError Int(NaN).

using AtomsBase
using AtomsCalculators
using EmpiricalPotentials
using Unitful
using UnitfulAtomic

zH = 1
emins = Dict((zH, zH) => -1.17u"hartree", )
rmins = Dict((zH, zH) => 0.743u"Å", )
calc = LennardJones(emins, rmins, 5.0u"Å")

let
    cell_vectors = ([10.0, 0.0, 0.0]u"Å", [0.0, 10.0, 0.0]u"Å", [0.0, 0.0, 10.0]u"Å")
    periodic = periodic_system([:H => [0, 0, 1.]u"bohr",
                                :H => [0, 0, 3.]u"bohr"],
                               cell_vectors)
    AtomsCalculators.potential_energy(periodic, calc)
end
println("---  ok  ---")

let
    isolated = isolated_system([:H => [0, 0, 0.0]u"bohr",
                                :H => [0, 0, 1.9]u"bohr"])
    AtomsCalculators.potential_energy(isolated, calc)  # issue
end
println("---  this will not be reached  ---")

[tjjarvinen]

This is related to cell.

The current pairlist implementation needs well defined cell and that is not true for isolated systems. My plan for this was to add proper isolated system support for the new neigborlist package, but that is work in progress still.

[cortner]

In JuLIP when I had no PBC I simply created a minimal cubic cell that contained all the atoms in the system. Should be very quick to do here too.

[tjjarvinen]

You should do it at Neighbourlists.jl with AtomsBase interface. That would then propagate the change to all the packages that use AB interface.

[cortner]

JuliaMolSim/NeighbourLists.jl#32

[cortner]

new version of neighbourlists is now registered. Please check and close this if it is now ok.

[cortner]

This has been open since Jan 13, it should be fixed. So I will close. Reopen if the issue arises again