Re: [AMBER] Query- Heating after Equilibration

From: Jagdeesh C <jagc666.gmail.com>
Date: Wed, 23 Jun 2010 16:58:15 +0530

Thanks Prof. Case.

I wished to check if both the minimization steps are generating proper
restart files that can be used for further simulations.

> I performed 2 sets of minimization and then went on to perform the MD.
>
> 1) 1000 steps of steepest descent
> steepest
> &cntrl
> imin=1, ntmin=2, maxcyc=1000,
> cut=10.0,ntb=1,
> ntc=2,ntf=2,
> ntpr=100,
> /
>
> 2) This is followed by 1000 steps of Polak-Ribiere conjugte gradient
> polak
> &cntrl
> imin=1, ntmin=3,
> irest=0, ntx=1,
> maxcyc=1000,
> cut=10.0,ntb=0,
> ntc=2,ntf=2,
> ntpr=100,
> drms = 0.0119,
> /
> &lmod
> xmin_method = prcg,
> /
>



I checked MD (constant volume, as pointed) from the rst file produced from
1st minimization and it works fine.

However when I use the restart file from steepest descent to perform 2nd
step of PRCG minimization (previously with ntb=0 shown above) the new rst
generated doesnt work (for MD)

Following your inputs I made the changes that were suggested in PRCG
minimization.step 2 (ntb=2, ntp= 1)

new_polak
&cntrl
imin=1, ntmin=3,
irest=0, ntx=1,
maxcyc=1000,
cut=8.0, ntb=2,
ntc=2, ntf=2,
ntpr=1,
ntp= 1,
/
&lmod
xmin_method = "PRCG",
/

However, I am getting a segmentation fault with this new input. How do I
deal with this error? I have also attached the output files from both
previous and new PRCG's.

Earlier, I wasnt getting any outputs file for subsequent MD (following old
PRCG) because of the error in rst file most likely. For further Md steps, I
shall make changes that have been suggested.

Best regards,
Jagdeesh



---------------------------------------------------------------------------------------------------



On Wed, Jun 23, 2010 at 10:13 AM, case <case.biomaps.rutgers.edu> wrote:

> On Wed, Jun 23, 2010, Jagdeesh C wrote:
>
> > Thank you very much for your inputs Professor Case and Jason! I seem to
> have
> > identified the problem. It had someting to do with the rst file. But, I
> > don't know the solution to it.
> >
> > I performed 2 sets of minimization and then went on to perform the MD.
> >
> > 1) 1000 steps of steepest descent
> > steepest
> > &cntrl
> > imin=1, ntmin=2, maxcyc=1000,
> > cut=10.0,ntb=1,
> > ntc=2,ntf=2,
> > ntpr=100,
> > /
> >
> > 2) This is followed by 1000 steps of Polak-Ribiere conjugte gradient
> > polak
> > &cntrl
> > imin=1, ntmin=3,
> > irest=0, ntx=1,
> > maxcyc=1000,
> > cut=10.0,ntb=0,
> > ntc=2,ntf=2,
> > ntpr=100,
> > drms = 0.0119,
> > /
> > &lmod
> > xmin_method = prcg,
> > /
> >
> > Now when I perform subsequent MD simulations (input file shown below)
> with
> > the rst generated from the first step they work fine. But, when I try
> them
> > with the rst generated from the PRCG (second) step no output files are
> > generated.
>
> The second minimization step sets ntb=0, which will probably do really
> weird
> things since you seem to have a periodic system. It's a little hard to see
> what would cause "no output files" however. One of the first things sander
> does is to write a header to the output file, along with a copy of the
> input.
> Is there a chance that errors going to standard error (stderr) are being
> lost?
> Are you sure you have a non-empty input coordinate file?
>
> Anyway, you certainly have to fix the value of ntb, if you indeed want to
> do
> the second minimization (it's not clear why this is desirable, however.)
>
>
> > Here is the input for subsequent simulations using rst's generated from
> 1
> > & 2
> > MD
> > &cntrl
> > imin=0, irest=0, ntx=1,
> > nstlim=25000,dt=0.002,
> > ntc=2,ntf=2,
> > cut=10.0, ntb=2,
> > ntpr=1000, ntwx=1000,
> > ntt=3, gamma_ln=2.0,
> > temp0=300.0,
> > ntp=1,taup=2.0,
> > /*
> > *
>
> There are lots of problems with the above. First, you are asking for 25000
> steps...when you are having problems, ask for 25 steps, set ntpr to 1, and
> carefully study the output.
>
> You should first equilibrate at constant volume (ntb=1) and only do
> constant
> pressure (ntb=2) as a second step. It is dangerous to use dt=0.002 for
> initial equilibration. There is probably no good reason to set the cutoff
> to 10 -- keep it at its default value.
>
> ....dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber


Received on Wed Jun 23 2010 - 04:30:03 PDT
Custom Search