[gmx-developers] Conserved energy drifts more with shorter time steps
Anton Feenstra
k.a.feenstra at vu.nl
Tue Jun 2 15:20:26 CEST 2015
On 02-06-15 15:14, Berk Hess wrote:
> On 2015-06-02 15:08, David van der Spoel wrote:
>> On 02/06/15 14:30, Berk Hess wrote:
>>> On 2015-06-02 14:26, David van der Spoel wrote:
>>>> On 02/06/15 14:05, Berk Hess wrote:
>>>>> I assume David is looking at the conserved energy quantity.
>>>> Indeed, averages and drift over 900 ps, 1000 molecules:
>>>>
>>>> Energy Average Err.Est. RMSD Tot-Drift
>>>> -------------------------------------------------------------------------------
>>>>
>>>>
>>>> Total Energy 20460.7 2.5 778.213 12.4265 (kJ/mol)
>>>> Conserved En. -4384.7 380 778.283 2665.29 (kJ/mol)
>>>>
>>>>>
>>>>> But I just noticed that Verlet - buffer - tolerance is not mentioned,
>>>>> so I assume it is default. The drift estimate is an overestimate
>>>>> which is tighter for shorter pair list lifetimes. I guess that is the
>>>>> explanation. Set it very low, so you are sure the energy drift is not
>>>>> affected by the (too small) buffer.
>>>>>
>>>> This may be the default.
>>>> verlet-buffer-tolerance = 0.005
>>>> should we set it to 1e-5 or something like that?
>>> It depends on how low drift you want to be able to measure. You need to
>>> set it lower than the drift you want to be able to measure.
>> Ok, will try. What about the time step dependence? Any clue?
> I explained that before, see 4 steps above.
>
> Berk
>>
>>
>>>
>>> Berk
>>>>
>>>>> Berk
>>>>>
>>>>> On Jun 2, 2015 1:45 PM, "Shirts, Michael R. (mrs5pt)"
>>>>> <mrs5pt at eservices.virginia.edu> wrote:
>>>>>>
>>>>>>> How do we run more NVE than with the settings below?
>>>>>>
>>>>>> tcoupl = Nose-Hoover
>>>>>>
>>>>>>
>>>>>> Would make it not NVE.
>>>> Turning tcoupl off would not keep the temperature constant of course,
>>>> seeing that there is a drift in the conserved energy this means that
>>>> the temperature will change.
>>>>
>>>> However, the main question was why does "conserved-energy
>>>> conservation" get worse when the time step is reduced? Integration
>>>> should presumably become more accurate, or not? Or does one need
>>>> double precision for short time steps?
Yes, at least that is what I saw for my and Berk's 1999 JCC paper.
At single precision, drift would not go down below a certain time step.
I have no idea anymore which timestep precisely, and I'm sure this would
be system specific.
I also recall drift indeed going up at (extreme) short timesteps. My
intuition for that at the time was that arithmetic errors would
accumulate faster for calculating many very small delta's.
>>>>
>>>>
>>>>>>
>>>>>> Do you mean the conserved quantity isn't conserved? Energy certainly
>>>>>> won't be conserved with Nose-Hoover.
>>>>>>
>>>>>> Best,
>>>>>> ~~~~~~~~~~~~
>>>>>> Michael Shirts
>>>>>> Associate Professor
>>>>>> Department of Chemical Engineering
>>>>>> University of Virginia
>>>>>> michael.shirts at virginia.edu
>>>>>> (434) 243-1821
>>>>>>
>>>>>>
>>>>>>
>>>>>> On 6/2/15, 7:22 AM, "David van der Spoel" <spoel at xray.bmc.uu.se>
>>>>>> wrote:
>>>>>>
>>>>>>> On 02/06/15 12:47, Berk Hess wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> So this is not using SETTLE or LINCS?
>>>>>>>> With SETTLE this is a known issue.
>>>>>>> No this is flexible organic molecules at contant volume.
>>>>>>>
>>>>>>>>
>>>>>>>> Did you try running NVE?
>>>>>>> How do we run more NVE than with the settings below?
>>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>> Cheers,
>>>>>>>>
>>>>>>>> Berk
>>>>>>>>
>>>>>>>> On 2015-06-02 12:43, David van der Spoel wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> this is in between a user and a developer query. We find that for
>>>>>>>>> flexible liquid simulations using gromacs 5.0.4/single
>>>>>>>>> precision the
>>>>>>>>> energy conservation gets WORSE with decreasing time step. A plot
>>>>>>>>> showing this is in
>>>>>>>>> http://folding.bmc.uu.se/images/Econserved-timestep.xvg or
>>>>>>>>> http://folding.bmc.uu.se/images/Econserved-timestep.pdf
>>>>>>>>>
>>>>>>>>> MDP settings
>>>>>>>>> coulombtype = PME
>>>>>>>>> coulomb-modifier = Potential-shift
>>>>>>>>> rcoulomb-switch = 0
>>>>>>>>> rcoulomb = 1.1
>>>>>>>>> epsilon-r = 1
>>>>>>>>> epsilon-rf = inf
>>>>>>>>> vdw-type = PME
>>>>>>>>> vdw-modifier = Potential-shift
>>>>>>>>> rvdw-switch = 0
>>>>>>>>> rvdw = 1.1
>>>>>>>>> tcoupl = Nose-Hoover
>>>>>>>>> nsttcouple = 10
>>>>>>>>> nh-chain-length = 1
>>>>>>>>> pcoupl = No
>>>>>>>>> ref-t: 298.15
>>>>>>>>> tau-t: 0.5
>>>>>>>>>
>>>>>>>>> Any clues whether we are doing something wrong? Or is there a bug?
>>>>>>>>>
>>>>>>>>> Cheers,
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> 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.
>>>>>>
>>>>>> --
>>>>>> 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.
>>>>
>>>>
>>>
>>
>>
>
--
Groetjes,
Anton
_____________ _______________________________________________________
| | |
| _ _ ___,| K. Anton Feenstra |
| / \ / \'| | | IBIVU/Bioinformatics - Vrije Universiteit Amsterdam |
|( | )| | | De Boelelaan 1081 - 1081 HV Amsterdam - Netherlands |
| \_/ \_/ | | | Tel +31 20 59 87783 - Fax +31 20 59 87653 - Room P136 |
| | Feenstra at few.vu.nl - www.few.vu.nl/~feenstra/ |
| | "Proceed, With Fingers Crossed" (TeX) |
|_____________|_______________________________________________________|
More information about the gromacs.org_gmx-developers
mailing list