search

CederGroupHub / chgnet

Pretrained universal neural network potential for charge-informed atomistic modeling https://chgnet.lbl.gov
https://doi.org/10.1038/s42256-023-00716-3
Other
220 stars 60 forks source link

1. Fix `taut` + `taup` units 2. Add missing time in NPT #67

Closed tsihyoung closed 11 months ago

janosh commented 11 months ago

Thanks, could you add a unit test?

tsihyoung commented 11 months ago

Thanks, could you add a unit test?

This PR can still pass the original test_md.py. To make the difference caused by taut and taup observable, MD simulation duration should be longer.

Another thing is that for each run, Ekin can deviate up to 2e-3 relatively. This high threshold might cause false positive in testing.

For Nose-Hoover part, ase is buggy when external stress is not given. I have also made a PR there. It might be better to add a unit test for N-H later?

BowenD-UCB commented 11 months ago

Thanks for fixing this, there is indeed a bug that we passed taut with unit of units.fs**2 to berendsen thermostat.

BowenD-UCB commented 11 months ago

The Nose-Hoover thermostat is just added. Currently the ASE implementation only supports upper triangular lattice, with no numerical tolerance. As a result it's not useful at this point, and we have only tested it preliminarily. The transformation of lattice to upper-triangular matrix will be implemented later