[gmx-developers] Inconsistency in Coulomb 1-4 energy term scaled by lambda

[Matt Wyczalkowski]

I am running some sanity checks on a system where I scale the charges
on a peptide in water.  Specifically, I can modify the charges in one
of two ways:
A) with free-energy = yes, and having lambda scale the charges
B) with free-energy = no, and modifying the charges directly in the
itp file.  To do this, I multiply the OPLS-AA charges by the factor
lambda, and write the modified itp file prior to processing with
grompp.

In principle, these two approaches should yield the same energies --
am I correct in assuming this?

In fact, I find that the coulomb 1-4 energy are different, depending
on the lambda value used.  For lambda = 0 and 1, approaches A and B
give me the same energies (as read in the log file).  For lambda =
0.5, most of the energy terms are identical (or very similar), but the
Coulomb 1-4 terms differ depending on whether A or B are used.  I list
the relevant lines of the two log files below.

SCALING TYPE A
           Bond          Angle    Proper Dih. Ryckaert-Bell.
LJ-14
      2.96803e+00    7.06877e+01    2.24532e+01    7.20134e+02
1.14116e+02
       Coulomb-14        LJ (SR)   Coulomb (SR)   Coulomb (LR)
Potential
      1.12253e+03    5.36139e+04   -3.49154e+05    1.12833e+03
-2.92359e+05
      Kinetic En.   Total Energy    Temperature Pressure (bar)
dVpot/dlambda
      5.30803e+04   -2.39278e+05    2.97292e+02    1.17644e+03
-2.01533e+03

SCALING TYPE B
           Bond          Angle    Proper Dih. Ryckaert-Bell.
LJ-14
      2.96803e+00    7.06877e+01    2.24532e+01    7.20134e+02    1.14116e+02
       Coulomb-14        LJ (SR)   Coulomb (SR)   Coulomb (LR)      Potential
      5.61266e+02    5.36139e+04   -3.48176e+05    1.12811e+03   -2.91942e+05
      Kinetic En.   Total Energy    Temperature Pressure (bar)
      5.30805e+04   -2.38862e+05    2.97294e+02    1.18713e+03

This issue is observed for both Reaction Field and Coulomb
electrostatics, and in versions 3.3.1, 3.3.2 and
gromacs-3.3.99_development_20071120.

I'll be happy to provide the relevant files.

Regards,

Matt

-- 
Matt Wyczalkowski
Doctoral Candidate, Biomedical Engineering
Pappu Lab: http://lima.wustl.edu
Washington University in St. Louis