AMBER Archive (2006)

Subject: Re: AMBER: ntt=1 in simulated annealing

From: Tanya Johannsen (tanyajohannsen_at_yahoo.com)
Date: Tue Nov 14 2006 - 17:59:04 CST


I tried gamma_ln=1.0 and my simulated annealing run
didn't work at all--but I doubt that the gamma_ln is
the culprit. For some reason the &wt section was
completely ignored and the temperature simply rose to
300 K and stayed there. I compared my input file to
the Nudged Elastic Band simulated annealing tutorial
which uses ntt=3 and I can't find meaningful
differences. If anyone could take a look at my input
file and tell me what I did wrong to make the &wt
section ignored I would be really grateful. Here it
is:

This simulated annealing run heats the system to 700 K
over the course of 5 ps,
leaves it there for 5 ps, and then cools over a period
of 90 ps. It uses a Langevin temperature scaling
rather than Berendsen.
 &cntrl
  imin=0,
  nmropt=1,
  ntx=1,
  irest=0,
  ntpr=200,
  ntwx=500,
  ntf=2,
  ntb=1,
  cut=8,
  ipol=0,
  ntr=1,
  restraint_wt=10,
  restraintmask=':1-382,2293-2674',
  nstlim=200000,
  dt=0.0005,
  ntt=3,
  tempi=0,
  gamma_ln=1.0,
  ntc=2,
 /
 &wt type='TEMP0', istep1=0, istep2=10000, value1=0.0,
value2=700.0,/
 &wt type='TEMP0', istep1=10001, istep2=20000,
value1=700.0, value2=700.0,/
 &wt type='TEMP0', istep1=20001, istep2=30000,
value1=700.0, value2=665.0,/
 &wt type='TEMP0', istep1=30001, istep2=40000,
value1=665.0, value2=630.0,/
 &wt type='TEMP0', istep1=40001, istep2=50000,
value1=630.0, value2=595.0,/
 &wt type='TEMP0', istep1=50001, istep2=60000,
value1=595.0, value2=560.0,/
 &wt type='TEMP0', istep1=60001, istep2=70000,
value1=560.0, value2=525.0,/
 &wt type='TEMP0', istep1=70001, istep2=80000,
value1=525.0, value2=490.0,/
 &wt type='TEMP0', istep1=80001, istep2=90000,
value1=490.0, value2=455.0,/
 &wt type='TEMP0', istep1=90001, istep2=100000,
value1=455.0, value2=420.0,/
 &wt type='TEMP0', istep1=100001, istep2=110000,
value1=420.0, value2=385.0,/
 &wt type='TEMP0', istep1=110001, istep2=120000,
value1=385.0, value2=350.0,/
 &wt type='TEMP0', istep1=120001, istep2=130000,
value1=350.0, value2=315.0,/
 &wt type='TEMP0', istep1=130001, istep2=140000,
value1=315.0, value2=280.0,/
 &wt type='TEMP0', istep1=140001, istep2=150000,
value1=280.0, value2=245.0,/
 &wt type='TEMP0', istep1=150001, istep2=160000,
value1=245.0, value2=210.0,/
 &wt type='TEMP0', istep1=160001, istep2=170000,
value1=210.0, value2=175.0,/
 &wt type='TEMP0', istep1=170001, istep2=180000,
value1=175.0, value2=125.0,/
 &wt type='TEMP0', istep1=180001, istep2=200000,
value1=125.0, value2=0.0,/
 &wt type='END'/

Thanks,
Tanya

--- Steven Winfield <saw44_at_cam.ac.uk> wrote:

> This may or may not be helpful:
>
> JChemPhys 120(24) 11432-11441 (2004)
>
> This paper states that the optimum gamma is found by
> integrating the
> memory function of an NVE simulation of your system,
> but as a rule of
> thumb the thermostat frequency (= 2*pi / gamma)
> should be "a few tens of
> times smaller than the smallest characteristic
> frequency of the system
> to be studied"
>
> Steve.
>
> Ross Walker wrote:
> > Hi Tanya,
> >
> > The use of such a high gamma_ln in that tutorial
> has nothing to do with
> > the high temperatures it is a function of the way
> nudged elastic band
> > simulations work and the tutorial is specific to
> this method.
> >
> > For a regular MD simulation you should not use
> such a high value of
> > gamma_ln. I suspect a value of 1.0 should be
> perfectly good for your
> > simulated annealing. You could perhaps set it a
> bit higher if you find
> > that the system is taking too long to reach the
> target temperature but
> > if all is working fine this is probably okay. One
> could perhaps argue
> > that at higher temperatures the number of
> collisions that would be
> > occuring per ps would be higher and so a higher
> value of gamma_ln might
> > be appropriate but I have no idea of the
> magnitudes of the changes here.
> >
> > Perhaps others can comment some more.
> >
> > All the best
> > Ross
> >
> > /\
> > \/
> > |\oss Walker
> >
> > | HPC Consultant and Staff Scientist |
> > | San Diego Supercomputer Center |
> > | Tel: +1 858 822 0854 | EMail:-
> ross_at_rosswalker.co.uk |
> > | http://www.rosswalker.co.uk
> <http://www.rosswalker.co.uk/> | PGP Key
> > available on request |
> >
> > Note: Electronic Mail is not secure, has no
> guarantee of delivery, may
> > not be read every day, and should not be used for
> urgent or sensitive
> > issues.
> >
> >
> >
> >
>
------------------------------------------------------------------------
> > *From:* owner-amber_at_scripps.edu
> [mailto:owner-amber_at_scripps.edu] *On
> > Behalf Of *Tanya Johannsen
> > *Sent:* Friday, November 03, 2006 16:39
> > *To:* amber_at_scripps.edu
> > *Subject:* Re: AMBER: ntt=1 in simulated
> annealing
> >
> > Thank you for your reply. One question: does
> the use of ntt=3 for
> > simulated annealing (and hence high
> temperatures) affect how I
> > should define gamma_ln? Based on tutorials and
> the literature that
> > I've looked at I thought that gamma_ln=1.0 was
> a pretty reasonable
> > choice for room temperature simulations.
> However, in the only
> > tutorial that I've been able to find that uses
> ntt=3 for simulated
> > annealing
> >
>
(http://www.rosswalker.co.uk/tutorials/amber_workshop/Tutorial_seven/section5.htm),
> > gamma_ln is equal to 1000.0. Does this have
> something to do with the
> > high temperatures? I have consulted the
> literature and it is not
> > clear to me how gamma_ln should be affected
> (if at all) by
> > performing simulated annealing
> > rather than a room temperature simulation. I
> would be very grateful
> > if anyone could offer their suggestions on
> defining gamma_ln.
> >
> > Thank you in advance.
> >
> > Tanya
> >
> >
> > */Carlos Simmerling <carlos_at_csb.sunysb.edu>/*
> wrote:
> >
> > you can use the weight changes to make
> gradual changes to TEMP0
> > and it will work fine. Check page 106 of
> the Amber9 manual and it
> > shows you how to do weight changes.
> >
> > Tanya Johannsen wrote:
> >
> > > Dear AMBER community:
> > >
> > > I am working on a simulated annealing
> run. My understanding
> > was that
> > > ntt=1 is traditionally used for
> simulated annealing since it is
> > > possible to use the tautp parameter to
> induce slow and gradual
> > > temperature changes. However, I saw that
> in the AMBER 9 manual
> > it says
> > > "Unless you are sure you know what you
> are doing, please don't
> > use
> > > ntt=1!" If not using ntt=1, what is the
> best method to use in
> > > simulated annealing? My understanding is
> that ntt=3 leads to a
> > > relatively quick adjustment to the
> target temperature, so even
> > if I
> > > included many small adjustments in
> TEMP0, I'm concerned that
> > it would
> > > be impossible to create a smooth
> temperature curve. I would
> > appreciate
> > > any advice on the best choice for "ntt"
> in simulated annealing
> > runs.
> > >
> > > Thank you in advance for any
> suggestions.
> > >
> > > Tanya
> > >
> > >
> >
>
------------------------------------------------------------------------
> > > Access over 1 million songs - Yahoo!
> Music Unlimited Try it
> > today.
> > >
> >
> >
> >
>
-----------------------------------------------------------------------
> > The AMBER Mail Reflector
> > To post, send mail to amber_at_scripps.edu
> > To unsubscribe, send "unsubscribe amber"
> to majordomo_at_scripps.edu
> >
> >
> >
>
------------------------------------------------------------------------
> > Get your email and see which of your friends
> are online - Right on
> > the new Yahoo.com
> >
>
<http://us.rd.yahoo.com/evt=42973/*http://www.yahoo.com/preview>
>
>
-----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to
> majordomo_at_scripps.edu
>

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu