Low accuracy default settings yield incorrect liquid densities - Redmine #2071
Following mail discussions and running many test simulations it is found that certain default settings from gromacs 4.6 onward lead to incorrect/different liquid densities. Examples are:
Caleman2012 gromacs-5.1
chloroform (GAFF) 1375 1448
1,1,2-trichloroethane (OPLS) 1424 1438
1-bromopropane (CGenFF) 1389 1385
dibromomethane (OPLS) 2474 2570
methanol (OPLS) 777 774
The worst cases are really off. Preliminary tests with other codes show that the gromacs-5.1 result above for chloroform is roughly correct. However the following updates should be made in gromacs defaults which are too weak:
verlet-buffer-tolerance = 1e-4 (default 5e-3)
lincs-iter = 2 (default 1)
constraints = h-bonds (give warning for all-bonds?)
delta_t = 0.001 ps (give warning for combinations?)
shake-tol = 1e-5 (single precision, default 1e-4)
(from redmine: issue id 2071, created on 2016-11-07 by dspoel)
- Changesets:
- Revision 483a7a11 by Berk Hess on 2017-02-07T10:37:31Z:
Add mdp option check with decoupled modes
When atoms involved in an angle with constrained bonds have very
different masses, there can be very weakly coupled dynamics modes.
Default mdp settings are often not sufficiently accurate to obtain
equipartitioning. This change adds a grompp check for this issue.
Part of #2071.
Change-Id: I64323154c38abe23b8d38d9d539d2a713a5c35e0
- Uploads:
- chloroform.tgz Input for running chloroform simulations described.
- simulation-results.xls Spreadsheet with simulation results.
- LIPID14_validation_gromacs.xls
- simulation-results.xls
- LIPID14_validation_gromacs.xls