Project

General

Profile

Bug #1146

free energy kernel does not treat exact cut-off's correctly

Added by Berk Hess over 4 years ago. Updated over 3 years ago.

Status:
Closed
Priority:
High
Assignee:
Category:
mdrun
Target version:
Affected version - extra info:
4.6
Affected version:
Difficulty:
uncategorized
Close

Description

With PME and soft-core, the free energy kernel incorrectly subtracts the PME mesh interaction beyond the cut-off.
With VdW no cut-off is applied at all.
The PME mesh interaction subtraction can give NaN energies and forces with r close to 0.

Associated revisions

Revision 0968f3ab (diff)
Added by Mark Abraham over 4 years ago

Update warning reference files

Some fixes to FEP soft-core changed the warnings that
were issued.

Refs #1146

Change-Id: I88f7210a2abddab30c051faec53c52e3aea62ccb

Revision a46b7f1f (diff)
Added by Berk Hess over 4 years ago

fixed issues with FEP soft-core and cut-off's

Now the cut-off is applied to the PME mesh correction for perturbed
pairs with soft-core. Now the cut-off is applied to VdW with soft-core.
Fixes possible NaN in free energy kernel with r close to 0.
Replaced a confusing grompp warning with PME and soft-core by a note.
Removed an invalid grompp warning about soft-core and twin-range.
Fixes #1146

Change-Id: I79a06b20158df2bce575808b6e31e660163bd307

Revision 96b98744 (diff)
Added by Berk Hess about 4 years ago

fixed issues with FEP soft-core and cut-off's

Now the cut-off is applied to the PME mesh correction for perturbed
pairs with soft-core. Now the cut-off is applied to VdW with soft-core.
Fixes possible NaN in free energy kernel with r close to 0.
Replaced a confusing grompp warning with PME and soft-core by a note.
Removed an invalid grompp warning about soft-core and twin-range.
Fixes #1146

Change-Id: I79a06b20158df2bce575808b6e31e660163bd307

History

#1 Updated by Berk Hess over 4 years ago

Most of these issues are trivial to fix.
But soft-core and exact cut-offs is somewhat problematic, as there with typical setups with a cut-off of 1 nm the soft-core r_sc is maximally around 1.0001 at r=1 nm. For most interactions we can simply put the cut-off on r_sc. But for Ewald we shouldn't do this, as real and reciprocal space contributions should (almost) cancel. So there I now put the cut-off on r.
I added a note in grompp for the fact that the at the cut-off the potential will usually be 1e-4 iso ewald_rtol for perturbed interactions. Note that this is almost never problematic, as this only affect intermediate, unphysical, lambda states and the Hamiltonian is well defined. But we should avoid PME load balancing, which changes the cut-off.

#2 Updated by Mark Abraham about 4 years ago

  • Affected version set to 4.6

Is this fully resolved, Berk?

#3 Updated by Mark Abraham about 4 years ago

  • Target version changed from 4.6.1 to 4.6.3

#4 Updated by Berk Hess about 4 years ago

  • Status changed from In Progress to Resolved
  • % Done changed from 0 to 100

#5 Updated by Rossen Apostolov over 3 years ago

  • Status changed from Resolved to Closed

Also available in: Atom PDF