Add test for global Nosé-Hoover thermostat + refactor #96
Conversation
danielhollas
added
refactor
danielhollas
force-pushed
the
nhc-global
branch
from
a0d324c to
c925420
Compare
Codecov Report
@@ Coverage Diff @@
## master #96 +/- ##
==========================================
+ Coverage 84.74% 85.55% +0.80%
==========================================
Files 41 41
Lines 5783 5780 -3
==========================================
+ Hits 4901 4945 +44
+ Misses 882 835 -47
|
danielhollas
force-pushed
the
nhc-global
branch
from
c925420 to
6c1e177
Compare
| call fatal_error(__FILE__, __LINE__, & | ||
| 'Invalid NHC thermostat parameter') | ||
| end if | ||
| end subroutine check_nhc_parameters | ||
|
|
||
| subroutine nhc_init() |
There was a problem hiding this comment.
This init function was quite messy so I've broken in down to several pieces in individual functions. I also moved keyword sanity checks related to NHC here from init.F90.
| end do | ||
| if (tau0 > 0) Qm(1) = ams ! see tuckermann,stat.mech. |
There was a problem hiding this comment.
Here I made an actual functional change for PIMD with staging transformation. Previously, if tau0 was not set in the input file, the NHC mass for the staging centroid would be the same as for the other beads. Not sure whether this has real consequence, I'll do some testing later. Need to read the Tuckermann again to better understand where these formulas are coming from.
| if (ams < 0 .and. tau0 < 0) then | ||
| call set_yoshida_weights(nyosh) | ||
|
|
||
| ! Not sure if we should have this default value |
There was a problem hiding this comment.
Štěpán pointed out that the default tau0=0.001 ps was designed primarily for PIMD thermostatting. However, everyone uses it for regular MD simulations which is not ideal ... But this comes to a need for better education of our users. Maybe we should write down some pamphlet explaining global/local NHC or recommend using Langevine.
There was a problem hiding this comment.
That is a very good point, I've been worried about that for some time, but didn't know how to approach it.
I would suggest for now to remove the default value for classical MD and require it to be set explicitly in the input file, which should at least alert the users for it's importance a bit. What do you think? We could maybe keep the default value for PIMD.
Writing something up would be great, perhaps I could do a seminar in our group here and than share the slides.
There was a problem hiding this comment.
The removal of default value might be the way. It's an additional input section so the user should know what he wants. I would personally set default values for both PIMD (as is) and MD (e.g. global thermostat with tau 1.0 ps), however doing it just for PIMD is better - if somebody tries to recalculate old input, it will be backwards compatible in a sense that ABIN won't change the constant on its own.
Any material would be great, also adding the task to my to-do list.
danielhollas
force-pushed
the
nhc-global
branch
from
b08f69e to
1d1aede
Compare
No description provided.