[gmx-users] Fwd: Odd behavior in log files for two simulations with only different nstlog for Gromacs 5.0.4
Mark Abraham
mark.j.abraham at gmail.com
Sat Feb 14 11:48:31 CET 2015
On Sat, Feb 14, 2015 at 4:02 AM, Justin Lemkul <jalemkul at vt.edu> wrote:
>
>
> On 2/13/15 9:13 PM, Christopher Neale wrote:
>
>> Not sure about the differences, but focusing on what is similar: you're
>> not
>> getting anything close to the desired temperature (in one case half and in
>> one case 3 times after only 10 integration steps). I suspect that there is
>> something else going wrong. I've never combined the sd integrator with
>> another temperature coupling method, though I have read that the
>> nose-hoover
>> thermostat is more sensitive than other thermostats to unequilibrated
>> initial
>> conditions and tends to oscillate, so I suspect that you're just seeing
>> two
>> different oscillations that have more to do with statistical differences
>> than
>> with the log output frequency.
>>
>> Try running without nose-hoover and let sd to the temperature coupling for
>> you for a while and then turn on the nose-hoover if you think you need it.
>>
>>
> The tcoupl keyword is ignored with the sd integrator, so it's not
> Nose-Hoover
> having an issue. I agree with your assessment that there is something else
> going on here. There is a randomness to sd, of course, so the quoted
> results
> simply suggest likely an unstable system diverging in two different runs.
> To
> the question in the OP, nstlog has no implication for physics, only
> output, so
> this is just coincidence.
>
Not quite. nstlog = 1 requires mdrun to compute the energies every time
step, which it does not do by default (because that's work in the kernels,
and communication afterwards). The default behaviour leading to the
internal value of nstcalcenergy is undocumentable - too many parts of the
code with fingers in the pie. Some of the offenders have been removed from
mdrun already, and others are on the chopping block (yes, group scheme, I'm
looking at you).
In consequence, I have no idea whether there is a feedback mechanism
between the latest kinetic energy and the SD update that changes with
nstlog, but there could/should be (and it could be broken, too).
However, I would look at the apparent 100-fold increase in KE and assume
there is some other more drastic problem here, such as lack of
equilibration.
Mark
-Justin
>
>
> Chris.
>>
>> ________________________________________ From:
>> gromacs.org_gmx-users-bounces at maillist.sys.kth.se
>> <gromacs.org_gmx-users-bounces at maillist.sys.kth.se> on behalf of
>> Shuxiang Li
>> <xiang693 at gmail.com> Sent: 13 February 2015 18:08 To:
>> gromacs.org_gmx-users at maillist.sys.kth.se Subject: [gmx-users] Fwd: Odd
>> behavior in log files for two simulations with only different nstlog for
>> Gromacs 5.0.4
>>
>> Hi Gromacs users, I’ve recently been running stochastic dynamics
>> simulations
>> of a coarse-grained DNA system using Gromacs version 5.0.4 (time step =
>> 10fs)
>> and I’ve encountered some odd behavior. I’ve performed two separate
>> simulations that were set up and run identically, with the exception that
>> in
>> one simulation I am writing to the log file every simulation step
>> (nstlog=1)
>> and in the other simulation I’m writing to the log file every 10
>> simulation
>> steps (nstlog=10). At the start of each simulation (step 0), the potential
>> energy, kinetic energy, and temperature are identical to each other.
>> However
>> by the 10th simulation step, the results written to the log file are very
>> different. For example, for the simulation where nstlog=1 the kinetic
>> energy
>> is 3.82607e+05 and the temperature is 1.44300e+02 while in the simulation
>> where nstlog=10, the kinetic energy is 1.67058e+07 and the temperature is
>> 9.60676e+03. However, the potential energies are identical for the two
>> simulations, so I think that the trajectories are also probably
>> identical. In
>> other words, I think the only problem is with the kinetic energies and
>> temperatures as written to the log file. I was wondering if anyone in the
>> community could help me explain this behavior. I'm running Gromacs 5.0.4,
>> but
>> I have seen the same odd behavior both in Gromacs 5.0 and Gromacs 4.6.5.
>> I’ve
>> included a portion of the log files and the grompp.mdp from my two
>> simulations below. Thank you very much for your assistance, Shuxiang Li
>>
>> log file 1 (nstlog=1): ################################### Step
>> Time Lambda 0 0.00000 0.00000 Energies (kJ/mol) Bond
>> Tab. Bonds Angle LJ (SR) Coulomb (SR) 1.12668e+06
>> -1.10569e+07 9.97976e+04 6.40874e+05 0.00000e+00 Dih. Rest.
>> Potential Kinetic En. Total Energy Temperature 8.54999e+03
>> -9.18102e+06 7.90352e+05 -8.39067e+06 2.98080e+02 Pressure (bar)
>> 0.00000e+00 ...... Step Time Lambda 10 0.10000
>> 0.00000
>>
>> Energies (kJ/mol) Bond Tab. Bonds Angle LJ (SR)
>> Coulomb
>> (SR) 1.25536e+06 -1.09081e+07 1.20898e+05 6.47746e+05
>> 0.00000e+00
>> Dih. Rest. Potential Kinetic En. Total Energy Temperature
>> 1.17815e+05 -8.76625e+06 3.82607e+05 -8.38364e+06 1.44300e+02
>> Pressure (bar) 0.00000e+00
>>
>> log file 2 (nstlog=10): ################################### Step
>> Time Lambda 0 0.00000 0.00000 Energies (kJ/mol) Bond
>> Tab. Bonds Angle LJ (SR) Coulomb (SR) 1.12668e+06
>> -1.10569e+07 9.97976e+04 6.40874e+05 0.00000e+00 Dih. Rest.
>> Potential Kinetic En. Total Energy Temperature 8.54999e+03
>> -9.18102e+06 7.90352e+05 -8.39067e+06 2.98080e+02 Pressure (bar)
>> 0.00000e+00
>>
>> Step Time Lambda 10 0.10000 0.00000
>> Energies
>> (kJ/mol) Bond Tab. Bonds Angle LJ (SR) Coulomb (SR)
>> 1.25536e+06 -1.09081e+07 1.20898e+05 6.47746e+05 0.00000e+00
>> Dih.
>> Rest. Potential Kinetic En. Total Energy Temperature
>> 1.17814e+05
>> -8.76625e+06 2.54721e+07 1.67058e+07 9.60676e+03 Pressure (bar)
>> 0.00000e+00
>>
>> grompp.mdp file1 (nstlog=1): ####################### title
>> = 298 kelvin BD cpp = /lib/cpp define
>> = integrator = sd comm-mode = Angular dt
>> = 0.010 nsteps = 1000 nstxout = 100000
>> nstvout = 100000 nstlog = 1 nstenergy
>> = 100000 nstxtcout = 100000 xtc_grps =
>> system
>> energygrps = system nstlist = 25 ns_type
>> = grid pbc = no vdwtype = user
>> coulombtype = user cutoff_scheme = group dispcorr
>> = no tcoupl = nose-hoover tc_grps =
>> system tau_t = 10 ref_t = 298
>> gen_vel
>> = yes gen_temp = 298 gen_seed = 6789
>> ld_seed
>> = 6789 bd-fric = 0.0 rlist = 4.0
>> rcoulomb
>> = 4.0 rvdw = 4.0
>>
>> grompp.mdp file2 (nstlog=10): ####################### title
>> = 298 kelvin BD cpp = /lib/cpp define
>> = integrator = sd comm-mode = Angular dt
>> = 0.010 nsteps = 1000 nstxout = 100000
>> nstvout = 100000 nstlog = 10 nstenergy
>> = 100000 nstxtcout = 100000 xtc_grps =
>> system
>> energygrps = system nstlist = 25 ns_type
>> = grid pbc = no vdwtype = user
>> coulombtype = user cutoff_scheme = group dispcorr
>> = no tcoupl = nose-hoover tc_grps =
>> system tau_t = 10 ref_t = 298
>> gen_vel
>> = yes gen_temp = 298 gen_seed = 6789
>> ld_seed
>> = 6789 bd-fric = 0.0 rlist = 4.0
>> rcoulomb
>> = 4.0 rvdw = 4.0 -- Gromacs Users mailing list
>>
>> * Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_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-users or
>> send a
>> mail to gmx-users-request at gromacs.org.
>>
>>
> --
> ==================================================
>
> Justin A. Lemkul, Ph.D.
> Ruth L. Kirschstein NRSA Postdoctoral Fellow
>
> Department of Pharmaceutical Sciences
> School of Pharmacy
> Health Sciences Facility II, Room 629
> University of Maryland, Baltimore
> 20 Penn St.
> Baltimore, MD 21201
>
> jalemkul at outerbanks.umaryland.edu | (410) 706-7441
> http://mackerell.umaryland.edu/~jalemkul
>
> ==================================================
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at http://www.gromacs.org/
> Support/Mailing_Lists/GMX-Users_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-users or
> send a mail to gmx-users-request at gromacs.org.
>
More information about the gromacs.org_gmx-users
mailing list