Project

General

Profile

Bug #2513

ref_t and temperature incorrect with coulomb-type = user

Added by Yu Du over 1 year ago. Updated over 1 year ago.

Status:
Feedback wanted
Priority:
Normal
Assignee:
Category:
-
Target version:
-
Affected version - extra info:
Affected version:
Difficulty:
uncategorized
Close

Description

Maybe, this is another bug of forcetables.cpp. Previous one is Bug #2286.

You can reproduce this bug with following command:
gmx grompp -quiet -n index.ndx -f nvt_change.mdp -c prev.gro -t prev.cpt -p topol6-12.top -o test.tpr
gmx mdrun -quiet -notunepme -pin on -pinoffset 0 -ntomp 1 -ntmpi 24 -deffnm test

Needed files attached and changed ff files can be found in Bug #2286.

coulomb-type_user_bug.rar (1.38 MB) coulomb-type_user_bug.rar Yu Du, 05/23/2018 09:10 AM
test_buffered.log (29.1 KB) test_buffered.log unix2doced Yu Du, 05/28/2018 04:21 PM
test_unbuffered.log (28.9 KB) test_unbuffered.log unix2doced Yu Du, 05/28/2018 04:21 PM
table_not_log_enoug_error.rar (1.43 MB) table_not_log_enoug_error.rar Yu Du, 05/28/2018 04:29 PM

History

#1 Updated by Berk Hess over 1 year ago

  • Status changed from New to Feedback wanted

What temperature difference do you get?

Since this uses the deprecated group scheme, it would be good to check if adding a list buffer helps, e.g.:
rlist = 1.2
vdw-modifier = potential-shift
coulomb-modifier = potential-shift

#2 Updated by Yu Du over 1 year ago

Berk Hess wrote:

What temperature difference do you get?

Since this uses the deprecated group scheme, it would be good to check if adding a list buffer helps, e.g.:
rlist = 1.2
vdw-modifier = potential-shift
coulomb-modifier = potential-shift

Thanks for Berk's reply and sorry for late reply because of being busy with group meeting.

I tried the three .mdp configs as you said, it seemed the temperature moves in the right direction to ref_t, but it still doesn't agree the ref_t.

Question NO.1
Because of the flaw of group scheme, is this the best result I can get in deprecated group scheme?

I attached the two log files (buffered and unbuffered). Please check if there is something wrong.

Question NO.2

When I used the original uploaded files (not using list buffer, rlist=1.1), the test.xvg and test_Protein_LIG.xvg end with 2.1000000000e+00 nm and this is OK.
But when I used the changed one (using list buffer, rlist=1.2), the the test.xvg and test_Protein_LIG.xvg must end with 2.2020000000e+00 nm or larger.
If xvg files end with 2.2000000000e+00 nm will give following error:
Fatal error:
Tables in file test.xvg not long enough for cut-off:
should be at least 2.200000 nm
What's the difference between the buffered and unbuffered ones? Does the potential-shift config need the extra 0.002 nm?

For reproducing this error, table_not_log_enoug_error.rar is attached.

Thanks for your reply again.

#3 Updated by Berk Hess over 1 year ago

The table values beyond the cut-off are actually not used with the potential-shift modifier, but we still require them.
If it goes in the right direction, please also try rlist=1.3, that should remove all cut-off artifacts, I think.

#4 Updated by Yu Du over 1 year ago

I tried rlist=1.3 and again it's better, log file excerpt copied here:

        <======  ###############  ==>
        <====  A V E R A G E S  ====>
        <==  ###############  ======>

        Statistics over 5001 steps using 51 frames

   Energies (kJ/mol)
           Bond          Angle    Proper Dih.  Improper Dih.          LJ-14
    1.38220e+03    3.57879e+03    4.56294e+03    2.42834e+02    1.56612e+03
     Coulomb-14        LJ (SR)   Coulomb (SR)      Potential    Kinetic En.
    1.97231e+04    5.27891e+04   -4.92940e+05   -4.09095e+05    7.73338e+04
   Total Energy  Conserved En.    Temperature Pressure (bar)   Constr. rmsd
   -3.31762e+05   -2.78727e+05    3.07302e+02   -8.55727e+01    0.00000e+00

   Total Virial (kJ/mol)
    2.66930e+04    7.83751e+01    3.51169e+02
    8.64849e+01    2.65006e+04    8.66612e+01
    3.49704e+02    8.48586e+01    2.64722e+04

   Pressure (bar)
   -1.06005e+02   -5.13490e+00   -4.06329e+01
   -6.02765e+00   -8.00210e+01   -1.15357e+01
   -4.04717e+01   -1.13372e+01   -7.06922e+01

  Epot (kJ/mol)        Coul-SR          LJ-SR        Coul-14          LJ-14
Protein-Protein   -3.55339e+04   -3.04419e+03    2.01068e+04    1.55589e+03
    Protein-LIG    0.00000e+00    1.92947e+00    0.00000e+00    0.00000e+00
   Protein-rest   -1.20865e+04   -1.18965e+03    0.00000e+00    0.00000e+00
        LIG-LIG    4.81730e+01   -6.68381e+00   -3.83678e+02    1.02351e+01
       LIG-rest   -9.74429e+02   -3.41499e+01    0.00000e+00    0.00000e+00
      rest-rest   -4.44394e+05    5.70619e+04    0.00000e+00    0.00000e+00

  T-Protein_LIGT-Water_and_ions
    3.03014e+02    3.07635e+02

For comparison, excerpt of rlist=1.2 log file:
        <======  ###############  ==>
        <====  A V E R A G E S  ====>
        <==  ###############  ======>

        Statistics over 5001 steps using 51 frames

   Energies (kJ/mol)
           Bond          Angle    Proper Dih.  Improper Dih.          LJ-14
    1.38357e+03    3.53176e+03    4.56410e+03    2.46057e+02    1.56504e+03
     Coulomb-14        LJ (SR)   Coulomb (SR)      Potential    Kinetic En.
    1.96845e+04    5.25421e+04   -4.87910e+05   -4.04393e+05    7.82045e+04
   Total Energy  Conserved En.    Temperature Pressure (bar)   Constr. rmsd
   -3.26188e+05   -2.63514e+05    3.10762e+02   -3.52796e+01    0.00000e+00

   Total Virial (kJ/mol)
    2.66439e+04   -1.89650e+02   -3.15863e+02
   -1.94354e+02    2.62090e+04    2.60091e+02
   -3.11830e+02    2.63453e+02    2.63130e+04

   Pressure (bar)
   -6.93904e+01    1.97168e+01    3.13309e+01
    2.02347e+01   -1.40037e+01   -2.52022e+01
    3.08870e+01   -2.55724e+01   -2.24448e+01

  Epot (kJ/mol)        Coul-SR          LJ-SR        Coul-14          LJ-14
Protein-Protein   -3.18375e+04   -3.03619e+03    2.00710e+04    1.55351e+03
    Protein-LIG    0.00000e+00    3.07512e+00    0.00000e+00    0.00000e+00
   Protein-rest   -1.23147e+04   -1.14675e+03    0.00000e+00    0.00000e+00
        LIG-LIG    4.95632e+01   -6.84378e+00   -3.86415e+02    1.15256e+01
       LIG-rest   -1.02760e+03   -3.14347e+01    0.00000e+00    0.00000e+00
      rest-rest   -4.42780e+05    5.67602e+04    0.00000e+00    0.00000e+00

  T-Protein_LIGT-Water_and_ions
    3.05669e+02    3.11157e+02

For comparison, excerpt of rlist=1.1 log file:

        <======  ###############  ==>
        <====  A V E R A G E S  ====>
        <==  ###############  ======>

        Statistics over 5001 steps using 51 frames

   Energies (kJ/mol)
           Bond          Angle    Proper Dih.  Improper Dih.          LJ-14
    1.39299e+03    3.59139e+03    4.61856e+03    2.51254e+02    1.56620e+03
     Coulomb-14        LJ (SR)   Coulomb (SR)      Potential    Kinetic En.
    1.97190e+04    5.23944e+04   -4.88823e+05   -4.05289e+05    7.94686e+04
   Total Energy  Conserved En.    Temperature Pressure (bar)   Constr. rmsd
   -3.25821e+05   -2.49528e+05    3.15785e+02    1.65515e+01    0.00000e+00

   Total Virial (kJ/mol)
    2.63054e+04   -1.11658e+02    1.80111e+02
   -1.20452e+02    2.65845e+04   -4.37282e+02
    1.74587e+02   -4.36210e+02    2.61277e+04

   Pressure (bar)
    1.58178e+01    1.98550e+01   -2.07894e+01
    2.08232e+01   -8.15119e+00    5.40439e+01
   -2.01813e+01    5.39259e+01    4.19878e+01

  Epot (kJ/mol)        Coul-SR          LJ-SR        Coul-14          LJ-14
Protein-Protein   -3.57440e+04   -2.98691e+03    2.01013e+04    1.55602e+03
    Protein-LIG    0.00000e+00    2.48770e+00    0.00000e+00    0.00000e+00
   Protein-rest   -1.14148e+04   -1.12670e+03    0.00000e+00    0.00000e+00
        LIG-LIG    8.43552e+01   -7.09268e+00   -3.82322e+02    1.01840e+01
       LIG-rest   -1.00285e+03   -3.26795e+01    0.00000e+00    0.00000e+00
      rest-rest   -4.40746e+05    5.65453e+04    0.00000e+00    0.00000e+00

  T-Protein_LIGT-Water_and_ions
    3.10253e+02    3.16214e+02


So, as Berk said, temperature variation is the result of deprecated group scheme, this issue could be closed now.

And I asked a question in the gmx-developer mail list, could you please answer it?
[gmx-developers] Still one question about pme-user with user potential table and energygroup[[https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-developers/2018-May/010073.html]]

Also available in: Atom PDF