Dear users,
I am performing MD simulations of a material at T=300K. I've noticed that if I use the following options:
pot_extrapolation = 'second_order'
pot_extrapolation = 'second_order'
SCF steps are quite unstable (but converge anyway), starting from first MD step:
scf convergence threshold = 1.0E-10
estimated scf accuracy < 0.00000080 Ry
estimated scf accuracy < 0.00000044 Ry
estimated scf accuracy < 0.00000016 Ry
estimated scf accuracy < 9.5E-09 Ry
estimated scf accuracy < 2.1E-09 Ry
estimated scf accuracy < 2.0E-10 Ry
estimated scf accuracy < 1.0E-10 Ry
estimated scf accuracy < 2.0E-11 Ry
estimated scf accuracy < 0.00982726 Ry
estimated scf accuracy < 0.00335630 Ry
estimated scf accuracy < 0.00036212 Ry
estimated scf accuracy < 0.00002729 Ry
estimated scf accuracy < 0.00001055 Ry
estimated scf accuracy < 0.00001287 Ry
estimated scf accuracy < 0.00000066 Ry
estimated scf accuracy < 0.00000097 Ry
estimated scf accuracy < 0.00000007 Ry
estimated scf accuracy < 0.00000007 Ry
estimated scf accuracy < 0.00000002 Ry
estimated scf accuracy < 8.7E-09 Ry
estimated scf accuracy < 2.2E-09 Ry
estimated scf accuracy < 1.1E-10 Ry
estimated scf accuracy < 9.8E-12 Ry
estimated scf accuracy < 124.55557104 Ry
estimated scf accuracy < 566.20308384 Ry
estimated scf accuracy < 24.52678095 Ry
estimated scf accuracy < 11.49168590 Ry
estimated scf accuracy < 4.74770768 Ry
estimated scf accuracy < 2.84584072 Ry
estimated scf accuracy < 1.82872297 Ry
estimated scf accuracy < 1.13633834 Ry
estimated scf accuracy < 0.85163312 Ry
estimated scf accuracy < 0.75225475 Ry
estimated scf accuracy < 0.23630576 Ry
estimated scf accuracy < 0.06657040 Ry
estimated scf accuracy < 0.12356423 Ry
estimated scf accuracy < 0.10745706 Ry
estimated scf accuracy < 0.07377183 Ry
estimated scf accuracy < 0.03204377 Ry
estimated scf accuracy < 0.01108520 Ry
estimated scf accuracy < 0.00237517 Ry
estimated scf accuracy < 0.00051729 Ry
estimated scf accuracy < 0.00010551 Ry
estimated scf accuracy < 0.00001862 Ry
estimated scf accuracy < 0.00000216 Ry
estimated scf accuracy < 0.00000081 Ry
estimated scf accuracy < 0.00000112 Ry
estimated scf accuracy < 0.00000038 Ry
estimated scf accuracy < 0.00000017 Ry
estimated scf accuracy < 0.00000002 Ry
estimated scf accuracy < 2.1E-09 Ry
estimated scf accuracy < 5.2E-09 Ry
estimated scf accuracy < 7.1E-10 Ry
estimated scf accuracy < 2.1E-11 Ry
estimated scf accuracy < 82.01884485 Ry
estimated scf accuracy < 541.29488792 Ry
estimated scf accuracy < 30.68489908 Ry
estimated scf accuracy < 9.32622744 Ry
estimated scf accuracy < 5.74410066 Ry
estimated scf accuracy < 2.78523601 Ry
...
This is a how SCF accuracy looks without those steps:
scf convergence threshold = 1.0E-10
estimated scf accuracy < 0.00000012 Ry
estimated scf accuracy < 0.00000003 Ry
estimated scf accuracy < 4.4E-09 Ry
estimated scf accuracy < 3.9E-10 Ry
estimated scf accuracy < 6.1E-11 Ry
estimated scf accuracy < 0.00832406 Ry
estimated scf accuracy < 0.00236207 Ry
estimated scf accuracy < 0.00032964 Ry
estimated scf accuracy < 0.00002279 Ry
estimated scf accuracy < 0.00000535 Ry
estimated scf accuracy < 0.00000077 Ry
estimated scf accuracy < 0.00000017 Ry
estimated scf accuracy < 0.00000003 Ry
estimated scf accuracy < 0.00000001 Ry
estimated scf accuracy < 5.6E-10 Ry
estimated scf accuracy < 4.6E-10 Ry
estimated scf accuracy < 3.3E-11 Ry
estimated scf accuracy < 0.00491362 Ry
.....
When other system (magnetic) with second_order extrapolation for pot/wfc, it crashes after few steps (too many bands are not converged, at the beginning of SCF cycle), while without those options it works fine.
I wonder if this is an expected behavior of MD simulations. I thought pot/wfc extrapolation was supposed to help.
Thanks,
Sergey
USF
_______________________________________________ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users