Re: AMBER: NVE energy and temperature drift

From: Robert Duke <rduke.email.unc.edu>
Date: Thu, 9 Aug 2007 22:15:19 -0400

Hi Nikola -
State variables do fluctuate, with fluctuation dependent on system size (the
smaller the system, the larger the fluctuation), so pressure fluctuation is
expected; I can't tell you whether you are seeing more fluctuation than
expected here without more info, but doubt that is it. Case and Cheatham
are much more qualified to answer questions on the thermostats; Dave has
been saying less than positive things lately about ntt 1, but I know of lots
of systems where it seems to work fine (kind of the old standard, at least
for some, but there are some known problems apparently), so I am a bit
skeptical about that being the problem. I am inclined to believe something
subtle is going on here, but don't have a clue at the moment. If you want
to send me some files (directly to me), I'll look for anything we may have
missed; no promises, but I am curious.
Best Regards - Bob

----- Original Message -----
From: "Nikola Trbovic" <nt2146.columbia.edu>
To: <amber.scripps.edu>
Sent: Thursday, August 09, 2007 8:11 PM
Subject: RE: AMBER: NVE energy and temperature drift


> Hi Bob,
>
> The density looks perfectly equilibrated around 1.01 (rmsd of 0.0013 over
> the relevant 1 ns of equilibration). But the PRESS value in the output
> varies wildly around zero with an rmsd of 127. Is that the pressure and is
> that behavior normal?
>
> On the note of mden values I found the relevant thread shortly after I
> sent
> out my last email, so I know now that amber8 and amber9 simply don't
> distinguish between solute and solvent temperature anymore. But is there
> still a simple way to calculate these values? I'm just curious whether
> both
> solute and solvent follow the overall heating equally, since I only used a
> weak thermostat (ntt=1).
>
> Many thanks again,
> Nikola
>
> -----Original Message-----
> From: owner-amber.scripps.edu [mailto:owner-amber.scripps.edu] On Behalf
> Of
> Robert Duke
> Sent: Thursday, August 09, 2007 7:44 PM
> To: amber.scripps.edu
> Subject: Re: AMBER: NVE energy and temperature drift
>
> Hi Nikola -
> Tom, or some of the other guys with a bit more history behind them than I
> have will have to give you the lowdown on the mden values, but with regard
> to these three values, I see values like that fairly typically, and the
> pmemd behaviour replicates sander behaviour. Okay, one last
> consideration -
>
> are you sure you were pressure-equilibrated before starting nve? (ie.,
> the
> system density was not still slowly increasing). That's the only other
> wild
>
> guess I can come up with off the top of my head. This is really bad temp
> conservation, so something is just not right, and I don't see a problem
> with
>
> your run params.
> Best Regards - Bob
>
> ----- Original Message -----
> From: "Nikola Trbovic" <nt2146.columbia.edu>
> To: <amber.scripps.edu>
> Sent: Thursday, August 09, 2007 6:50 PM
> Subject: Re: AMBER: NVE energy and temperature drift
>
>
>> Dear Bob and Tom,
>>
>> first of all thanks for getting back to me so quickly. I'm not aware of
>> having made any changes to the ewald parameters. Here's my input file
>> for the nve production run:
>>
>> MD run
>> &cntrl
>> ntx=5, irest=1, ntpr=500, ntwr=500, ntwx=500, ntwe=500, ioutfm=1,
>> nstlim=2500000, dt=0.002,
>> ntt=0, ntp=0, ntc=2, ntf=2,
>> ntb=1, cut=10.0,
>> /
>>
>> To Bob's question I can say that the system doesn't seem to be doing
>> anything interesting over the 20 ns when visualized. The rmsd over the
>> trajectory looks fine, the protein stays folded... is there something
>> else I could look for?
>>
>> To Tom's final question I can say that yes, the potential energy also
>> increases linearly from ~-95300 to ~-94800 over the 20 ns. What does
>> that indicate?
>>
>> I will try out Tom's suggestions right away, but I wanted to answer your
>> questions and also share another observation that surprised me, to say
>> the least: the mden file shows the solute temperature always to be equal
>> to the overall temperature, while the solvent temperature value is
>> around -.77E+11. Is that normal? Is it just a bug, or is it something I
>> have to worry about?
>>
>> Thanks for your help,
>> Nikola
>>
>> On Thu, 2007-08-09 at 18:15 -0400, Robert Duke wrote:
>>> Hi Nikola -
>>> Yes, that seems sort of big. A few months ago I ran a small solvated
>>> dna
>>> sample under nve for 5 nsec; I was really comparing other methods to
>>> pme,
>>> but for pme (probably with cut 8 or 9, everything else fairly standard)
>>> I
>>> only saw a downward temp drift of about 0.2 K in 5 nsec with a total
>>> Etot
>>> drop from -87902 to -87948. Did you perhaps dink with the default ewald
>>> params or something? What is the system doing over the 20 nsec if you
>>> visualize it - perhaps something interesting? Others may perhaps want
>>> to
>>> share their experiences, but well-adjusted pme simulations are supposed
>>> to
>>> be pretty darn good at energy conservation.
>>> Regards - Bob Duke
>>>
>>> ----- Original Message -----
>>> From: "Nikola Trbovic" <nt2146.columbia.edu>
>>> To: <amber.scripps.edu>
>>> Sent: Thursday, August 09, 2007 4:24 PM
>>> Subject: AMBER: NVE energy and temperature drift
>>>
>>>
>>> > Dear amber community,
>>> >
>>> > I have run a couple of 20 ns NVE simulations of a protein in tip3p
>>> > water
>>> > with a parallel installation of the amber9 pmemd. I first equilibrated
>>> > the system in an NPT simulation (ntt=1) and then started the NVE
>>> > production runs from different starting structures extracted from the
>>> > equilibrated part of the simulation with
>>> >
>>> > ntt=ntp=0, ntc=ntf=2, ntb=1 and cut=10.
>>> >
>>> > Over 20 ns Etot rose absolutely linearly and with barely any random
>>> > fluctuations from ~-76660 to ~-75880. Over the same period the
>>> > temperature rose from an average value of ~300 K to an average value
>>> > of
>>> > ~305 K. The rms fluctuations of the temperature were fairly constant
>>> > at
>>> > around 1.5-2 K, but the moving average rose again astonishingly
>>> > linearly, as if I had unwittingly set it to be heated slowly.
>>> >
>>> > I have combed through the archives and found several threads on energy
>>> > drifts in nve simulations, but none seemed to report such a strong
>>> > drift
>>> > or a solution that would be applicable in my case.
>>> >
>>> > If I have missed something obvious I apologize for having wasted your
>>> > time in advance, but I would appreciate any insight on what I am doing
>>> > wrong.
>>> >
>>> > Many thanks in advance,
>>> >
>>> > Nikola Trbovic
>>> >
>>> > -----------------------------------------------------------------------
>>> > The AMBER Mail Reflector
>>> > To post, send mail to amber.scripps.edu
>>> > To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>>> >
>>>
>>>
>>> -----------------------------------------------------------------------
>>> The AMBER Mail Reflector
>>> To post, send mail to amber.scripps.edu
>>> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber.scripps.edu
>> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>>
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>


-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Aug 12 2007 - 06:07:44 PDT
Custom Search