[gmx-developers] Discrepancy with LAMMPS in Coulomb (SR) Coul. recip. energy

[David van der Spoel]

Den 2019-10-24 kl. 00:11, skrev Jeet Majumdar:
> Dear all,
> I was running a simulation with bulk water and CNT, with the protocol -> 
> Minimization, NPT, NVT. Keeping as much close as possible, I did the 
> same simulation in LAMMPS too. I observe that the LAMMPS and GROMACS log 
> file shows complete disagreement in sign and value in the coulomb energy 
> calculation in real space and reciprocal space; in all the steps. 
> Although upon adding the two parts (real+reciprocal) to get total 
> Coulomb energy, they are matching within error limit. Can anyone please 
> explain me why am i getting such a weird result? Is it some in-build 
> difference in calculation technique, or am I going wrong somewhere? I 
> couldn't figure it out.
Thanks for testing this. The difference comes from at what atom-atom 
distance one puts the transition from short range to long range but it 
is only the sum that matters.
> 
> GROMACS: (in kJ/mol)
> 
> Coulomb (SR) Coul. recip.
>   -3.59079e+05 8.41910e+02
> 
> in kcal/mol this translates to
> Potential Kinetic En. Total Energy Temperature Pres. DC
> -85821.94072 201.221319
> 
> LAMMPS: (in kcal/mol)
> 
> E_coul = 177903.2534 E_long = -270193.3273
> 
> 
> 


-- 
David van der Spoel, Ph.D., Professor of Biology
Head of Department, Cell & Molecular Biology, Uppsala University.
Box 596, SE-75124 Uppsala, Sweden. Phone: +46184714205.
http://www.icm.uu.se