Bug in triclinic?

The attached SPC/E water system runs perfectly fine for an orthogonal box. However, if I tilt the box (even just a tiny amount) to make it triclinic, the code crashes and I get NaNs everywhere, which makes me think there is a bug…

Stan

data.spce (414 KB)

in.spce (590 Bytes)

Hi Stan,

I ran your input … if I or reduce timestep or minimize before dynamics, then it seems to run indefinitely. Maybe there are bad contact(s) despite the small tilt?

Tim

The attached SPC/E water system runs perfectly fine for an orthogonal box.
However, if I tilt the box (even just a tiny amount) to make it triclinic,
the code crashes and I get NaNs everywhere, which makes me think there is a
bug...

please note that you have coordinates with image flags != 0, so that even a
small tilt on them will result in a significant displacement and hence can
cause overlaps.

axel.

in addition a better way to change an orthogonal box to triclinic would be using the change box command instead of editing a data file. with the “remap” set, those overlaps will be completely avoided. e.g. you could set up your data file to have the tilt factors 0.0 0.0 0.0 and then use: