MD results not consistent with older versions
Summary: Differences in MD calculations between Siesta versions 4.0 and 4.1.5 using identical setups.
Code version:
System information:
Steps to reproduce:
Setup: MD simulations on a 3x3 graphene sheet with a 20 Å separation in the z-direction between periodic images.
Temperature: Initial and final temperature set to 300K.
Dynamics: Nose Mass at 1 Ry*fs^2, run for 5000 steps at 1 fs/step.
Versions: The issue was observed comparing results between Siesta versions 4.1.5 and 4.0.
Files: Same pseudopotential files and identical FDF files used for both versions.
Expected behavior
In Siesta 4.0, the MD calculation results in a consistent average total energy with a slope < 0.5 eV/ns and a total energy variation < 50 meV.
In Siesta 4.1.5, there is a linear augmentation of the average total energy with a slope > 18.5 eV/ns and a total variation > 150 meV.
Relevant logs and/or screenshots
Possible fixes
Edited by ElieHAD5