[gmx-developers] DIfference in Coulomb terms 5.0/4.6

[David van der Spoel]

On 2014-11-16 16:24, Berk Hess wrote:
> Hi,
>
> The exclusion correction for both PME and RF is calculated in the non-bonded kernel with the Verlet scheme. As you mailed before, the energy matches exactly.
>
OK, but why is there such a large difference between group and verlet 
schemes.

For 512 SWM4-NDP molecules, 0 steps, niter = 0:

Cut-off scheme = Verlet
Energy                      Average   Err.Est.       RMSD  Tot-Drift
-------------------------------------------------------------------------------
LJ (SR)           2316.47         --          0          0  (kJ/mol)
Coulomb (SR)     -14480.6         --          0          0  (kJ/mol)
Coul. recip.      3378.39         --          0          0  (kJ/mol)
Polarization   1.5361e-07         --          0          0  (kJ/mol)
Potential        -8785.71         --          0          0  (kJ/mol)
Kinetic En.       6829.52         --          0          0  (kJ/mol)

Cut-off scheme = Group
Energy                      Average   Err.Est.       RMSD  Tot-Drift
-------------------------------------------------------------------------------
LJ (SR)           2316.65         --          0          0  (kJ/mol)
Coulomb (SR)     -13430.5         --          0          0  (kJ/mol)
Coul. recip.     -861.894         --          0          0  (kJ/mol)
Polarization            0         --          0          0  (kJ/mol)
Potential        -11975.8         --          0          0  (kJ/mol)
Kinetic En.       6727.81         --          0          0  (kJ/mol)


> Cheers,
>
> Berk
>
> On Nov 16, 2014 2:53 PM, David van der Spoel <spoel at xray.bmc.uu.se> wrote:
>>
>> On 2014-11-15 19:34, Berk Hess wrote:
>>> Hi,
>>>
>>> Is this group vs Verlet cutoff scheme? With the Verlet Cutoff's scheme the subtraction of the reciprocal component for excluded pair has been moved from the reciprocal term to the SR term.
>>> That is not really debugging, since the division is not based on physics and thus there is no bug.
>>> Additionally the automated PME load balancing can change beta, thereby changing the distribution of the total Coulomb energy over the two terms.
>>>
>>
>> Thanks. There is more though. I find that the exclusion/self corrections
>> are turned off when using Verlet scheme.
>>
>> The routine ewald_LRcorrection is called (in both 4.6.x and 5.0.x) such
>> that the variable calc_excl_corr is set as ir->cutoff_scheme !=
>> ecutsVERLET. This means in practice that with the Verlet scheme the self
>> and exclusion corrections are 0.
>>
>> Is this intentional? The effect is a large energy difference between
>> systems in the two versions (we are trying to get the polarizable drude
>> water working).
>>
>>
>> --
>> David van der Spoel, Ph.D., Professor of Biology
>> Dept. of Cell & Molec. Biol., Uppsala University.
>> Box 596, 75124 Uppsala, Sweden. Phone: +46184714205.
>> spoel at xray.bmc.uu.se    http://folding.bmc.uu.se
>> --
>> Gromacs Developers mailing list
>>
>> * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before posting!
>>
>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>> * For (un)subscribe requests visit
>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers or send a mail to gmx-developers-request at gromacs.org.


-- 
David van der Spoel, Ph.D., Professor of Biology
Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
spoel at xray.bmc.uu.se    http://folding.bmc.uu.se