Re: "Jump" in Energy and Volume graphs of my MD

From: Peter Freddolino (petefred_at_umich.edu)
Date: Thu Mar 31 2016 - 09:13:46 CDT

Ahhh, I third that explanation. I hadn’t looked closely enough at the stem of your inputname — I thought you were using the coor/vel/xsc files written at the end of the run, but missed the fact that you were actually using the .restart.* files because that was defined in inputname

> On Mar 31, 2016, at 9:44 AM, Jeff Comer <jeffcomer_at_gmail.com> wrote:
>
> Dear D
>
> I second Norman Geist. This is likely the issue:
>
> outputEnergies 500
> restartfreq 1000000
>
> The firsttimestep seems to suggest that the last energy was written at 28763000 steps. But, the last restart files were written at step 28000000. The restart files are 763000 steps backwards in time from the end of the run. You should be comparing the energy at step 28000000 of the first run with the energy at step 0 of the second run.
>
> Since you are writing the dcd every 500 steps, there is probably no advantage to having such a large restartFreq. You will probably see no performance loss if you change with restartFreq 5000.
>
> Jeff
>
>
>
> –––––––––––––––––––––––––––––––––––———————
> Jeffrey Comer, PhD
> Assistant Professor
> Institute of Computational Comparative Medicine
> Nanotechnology Innovation Center of Kansas State
> Kansas State University
> Office: P-213 Mosier Hall
> Phone: 785-532-6311
>
> On Thu, Mar 31, 2016 at 7:56 AM, Brian Radak <bradak_at_anl.gov> wrote:
> Perhaps unrelated, but the kinetic energy and instantaneous temperature should always jump slightly after a restart due to a change in the precision of the velocities (these are saved in single precision at the binary level, if I'm not mistaken). This is generally noticeable at multiple decimal places (up to 1 kcal/mol for kinetic energy) whereas changes in the potential energy are often limited to just the 4th place.
>
> Brian
>
>
> On 03/31/2016 01:00 AM, Norman Geist wrote:
>> The data that causes the jump might just be data that was written after the last restart files have been written. So, as you restarted the simulation from a point in time, earlier than your simulation already was, you get a jump of course, as your system made a time travel. If this is the issue, you don't need to worry about: For analyzation purpose you could remove that abundant frames (e.g. catdcd) and data from the log files after the last restart writing (text editor).
>>
>> Norman Geist
>>
>>
>>> -----Ursprüngliche Nachricht-----
>>> Von:
>>> owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu
>>> ] Im
>>> Auftrag von Peter Freddolino
>>> Gesendet: Donnerstag, 31. März 2016 02:34
>>> An: Daniela Rivas
>>> <dani.rivas.r_at_gmail.com>
>>>
>>> Cc:
>>> namd-l_at_ks.uiuc.edu
>>>
>>> Betreff: Re: namd-l: "Jump" in Energy and Volume graphs of my MD
>>>
>>> Hi Daniela,
>>> Just so we’re clear, using wrapNearest is not the right answer just because
>>> you’re simulating a membrane protein, but rather, if you want to use a
>>> hexagonal unit cell.
>>>
>>> Could you post a diff of the original config file vs. the restart config file? The
>>> periodic cell dimensions have to be getting messed up somehow, I’m just
>>> not quite sure how. The last few lines of the .xst file from the first run, and
>>> the output .xsc file, would also be helpful.
>>> Thanks,
>>> Peter
>>>
>>>
>>>> On Mar 30, 2016, at 5:01 PM, Daniela Rivas <dani.rivas.r_at_gmail.com>
>>> wrote:
>>>
>>>> Hi Peter,
>>>>
>>>> I'm attaching the graphs so you can take a better look and I'm using
>>>>
>>> wrapNearest on because it is a membrane protein simulation.
>>>
>>>> Thanks a lot.
>>>> Daniela.
>>>>
>>>> On Wed, Mar 30, 2016 at 5:56 PM, Peter Freddolino
>>>>
>>> <petefred_at_umich.edu>
>>> wrote:
>>>
>>>> Hi Daniela,
>>>> How big is the jump that you see in potential energy and volume?
>>>>
>>> (absolute and relative)
>>>
>>>> Could you provide a diff between the config file below and then one that
>>>>
>>> you used to generate the previous run?
>>>
>>>> Are you sure you want wrapNearest on, by the way? Are you trying to
>>>>
>>> simulate an unusually shaped unit cell?
>>>
>>>> Best,
>>>> Peter
>>>>
>>>>
>>>>> On Mar 30, 2016, at 4:49 PM, Daniela Rivas <dani.rivas.r_at_gmail.com>
>>> wrote:
>>>
>>>>> Yes, I'm very sure. I used this:
>>>>>
>>>>> # INPUT FILES #
>>>>>
>>>>> structure Modelo23_memb_ion.psf
>>>>> coordinates Modelo23_memb_ion.pdb
>>>>>
>>>>> # Continuing a job from the restart files #
>>>>>
>>>>> if {1} {
>>>>> set inputname dinamica_modelo.restart ;# only need to edit this
>>>>>
>>> in one place!
>>>
>>>>> binCoordinates $inputname.coor ;# coordinates from last
>>>>>
>>> run (binary)
>>>
>>>>> binVelocities $inputname.vel ;# velocities from last run
>>>>>
>>> (binary)
>>>
>>>>> extendedSystem $inputname.xsc ;# cell dimensions
>>>>>
>>> from last run
>>>
>>>>> firsttimestep 28763000 ;# last step of previous run
>>>>> numsteps 100000000 ;# run stops when this step is
>>>>>
>>> reached
>>>
>>>>> }
>>>>>
>>>>> # FORCE-FIELD PARAMETERS #
>>>>>
>>>>> paraTypeCharmm on
>>>>> parameters par_all27_prot_lipid.inp
>>>>>
>>>>> exclude scaled1-4
>>>>> 1-4scaling 1.0
>>>>> switching on
>>>>> switchDist 8.5
>>>>> cutoff 10
>>>>> pairlistdist 11.5
>>>>>
>>>>> # INTEGRATOR PARAMETERS #
>>>>>
>>>>> timestep 1
>>>>> stepsPerCycle 20
>>>>> nonBondedFreq 2
>>>>> fullElectFrequency 4
>>>>>
>>>>> # FIXED ATOMS #
>>>>>
>>>>> fixedAtoms on
>>>>> fixedAtomsForces off
>>>>> fixedAtomsFile myfixedatoms.pdb
>>>>> fixedAtomsCol B
>>>>>
>>>>> # CONSTRAINTS #
>>>>>
>>>>> #constraints off
>>>>> #consRef off
>>>>> #consKFile off
>>>>> #consKCol off
>>>>>
>>>>> # PRESSURE AND TEMPERATURE CONTROL #
>>>>>
>>>>> langevin on
>>>>> langevinDamping 10
>>>>> langevinTemp 310
>>>>> langevinHydrogen no
>>>>>
>>>>> langevinPiston on
>>>>> langevinPistonTarget 1.01325
>>>>> langevinPistonPeriod 200
>>>>> langevinPistonDecay 100
>>>>> langevinPistonTemp 310
>>>>>
>>>>> useGroupPressure yes # smaller fluctuations
>>>>> useFlexibleCell yes # allow dimensions to fluctuate
>>>>>
>>> independently
>>>
>>>>> useConstantRatio yes # fix shape in x-y plane
>>>>>
>>>>> # OUTPUT FILES #
>>>>>
>>>>> binaryoutput off
>>>>> outputname dinamica_modelo2
>>>>> outputEnergies 500
>>>>> outputTiming 500
>>>>> restartfreq 1000000
>>>>> xstFreq 500
>>>>> dcdFreq 500
>>>>> wrapAll on
>>>>> wrapNearest on
>>>>>
>>>>> # DYNAMICS #
>>>>>
>>>>> run 100000000
>>>>> output Modelo_final
>>>>>
>>>>> On Wed, Mar 30, 2016 at 5:23 PM, Peter Freddolino
>>>>>
>>> <petefred_at_umich.edu>
>>> wrote:
>>>
>>>>> Are you sure that on restart you are reading the extended system
>>>>>
>>> information from the restart file and not resetting the cell basis vectors?
>>>
>>>>> Best,
>>>>> Peter
>>>>>
>>>>>
>>>>>> On Mar 30, 2016, at 4:04 PM, Daniela Rivas <dani.rivas.r_at_gmail.com>
>>> wrote:
>>>
>>>>>> Dear NAMD users,
>>>>>>
>>>>>> When analyzing the basic data (T, P, V, Energy and protein RMSD) of
>>>>>>
>>> my MD I found that there was a "jump" in my Volume and Potential Energy
>>> graphs. This jump occurs just when I had to stop the simulation and re-
>>> started it from the restart files from the last step. I used all the restart files to
>>> re-start my simulation, and there's no problem with the Temperature,
>>> Pressure, Kinetic Energy or protein RMSD. Any thoughts of what could have
>>> happened? Has anyone else had this problem when restarting a MD?
>>>
>>>>>> I hope you can help me solve this riddle. Thanks in advance!
>>>>>>
>>>> <Energia_Modelo.png><Volumen_Modelo.png>
>>>>
>>
>
> --
> Brian Radak
> Postdoctoral Appointee
> Leadership Computing Facility
> Argonne National Laboratory
>
> 9700 South Cass Avenue, Bldg. 240
> Argonne, IL 60439-4854
> (630) 252-8643
> brian.radak_at_anl.gov
>

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:56 CST