AMBER Archive (2003)

Subject: Re: Etot is not constant in NVE ensemble

From: A. Hungie (hungie01_at_hotmail.com)
Date: Fri May 09 2003 - 11:08:56 CDT


Dear Prof. Tom Darden and amber users,

Thank you very much for your response. However, I am still confused. I also
have no idea what value of ewald coefficient (ew_coeff ) should be used or
increased. By default, ew_coeff is determine by dsum_tol and cutoff. So how
to calculate ew_coeff? In my case, if dsum_tol = 0.00001 and cutoff = 8.0
Angstrom, what is the value of ew_coeff? If I know ew_coeff in this case, I
may know how large of ew_coeff should be explicitly entered in my input file
(to get rid of energy drift).

Thank you again in advance.

Best Regards,
Hungie

>For comparison I have run a water box of 216 molecules for 5ns NVE with
>either a 2fs or 1fs time step with standard default. I got a drift of
>~2kcals per mole out of ~1750 kcals with a 2fs time step and ~1/2 kcal per
>mole with a 1fs time step. This latter number is comparable to yours
>relative to total energy. Also note that the energy conservation behaved
>quadratically with respect to timestep. This does not happen when
>you have a serious energy drift problem. However if you're unhappy with
>your results....
>Tom Cheatham (he can agree/disagree) had a similar problem and found that
>by increasing the default ewald coefficient a bit he got rid of the
>drift. It might be that for highly charged systems the discontinuity in
>direct sum energy/forces at the cutoff [i.e. erfc(beta*r_cutoff)/r_cutoff
>is too large] could cause heating. By increasing beta (ewald coefficient)
>this problem is alleviated. I don't recall how much Tom C increased beta.
>To be correct [i.e. keep a low error vs exact ewald] you then need to
>increase reciprocal sum parameters slightly. However since the
>reciprocal pme sum is smooth (no discontinuities) it doesn't seem to me
>that the reciprocal sum can cause heating. That is your problem is
>probably due to direct sum discontinuities and can be helped by increasing
>beta.
>
>Hope this help
>Tom Darden
>
> On Fri, 9 May 2003, A. Hungie wrote:
>
> > Dear All,
> >
> > I have run MD simulation of 14 basepairs DNA (in water and nuetralized
>by
> > Na+ ions) using protocol as in Amber tutorial. After doing 6 rounds of
> > minimizations (reducing restrain energy from 25 to 0 kcal/mol.A^2), I
>ran MD
> > in NPT ensemble for 100 ps which I obtained canstant box size. Then I
> > switched to NVE ensemble for 5.6 ns (800 ps for each run). I found that
> > total energy is not constant, i.e. it increases about 6 kcal/mol (from
> > -37828 kacl/mol to -37822 kcal/mol) within 5.6 ns. I used Amber 6 on
>Linux
> > cluster and input file as below.
> >
> > Is the obtained total energy normal and acceptable? Is it equilibrium?
> >
> > #my input file.
> > ------------------------------------------
> > &cntrl
> > ntx = 7, irest = 1, nmropt = 0,
> > imin = 0,
> > maxcyc = 1000, ncyc = 5000,
> > ntmin = 1, dx0 = 0.1, dxm = 0.5, drms = 0.0001,
> >
> > nstlim = 800000,
> > dt = 0.001, ndfmin = 0, t = 0.0,
> > timlim = 999999., ntcm = 1, nscm = 2500,
> >
> > ntpr = 1000, ntwr = 1000,
> > ntwx = 1000, ntwv = 0, ntwe = 0,
> > ntwprt = 0, ioutfm = 0, ntrx = 1, ntxo = 1,
> >
> > cut = 8.0, dielc = 1.0, nsnb = 10,
> > scnb = 2.0, scee = 1.2, iwrap = 0,
> >
> > ntb = 1, ntp = 0, npscal = 1,
> > pres0 = 1.0, comp = 44.6, taup = 2.0,
> >
> > ntc = 2, tol = 0.000001,
> > ntf = 2,
> >
> > ibelly = 0, ntr = 0,
> >
> > temp0 = 300.0, tempi = 300.0,
> > ig = 71277, heat = 0.0, tautp = 4.0,
> > ntt = 0, vlimit = 20.0, dtemp = 1.0,
> > &end
> >
> > &ewald
> > a = 42.0151161, b = 43.0746222, c = 69.8275997,
> > alpha = 90.000, beta = 90.000, gamma = 90.000,
> > nfft1 = 45, nfft2 = 45, nfft3 = 72,
> > order = 4, dsum_tol = 0.00001, eedmeth = 1,
> > opt_infl = 1, vdwmeth = 1, use_pme = 1,
> > frc_int = 0, nbflag = 1, skinnb = 1.0
> > &end
> > -----------------------------------
> >
> > Thank you very much in advance.
> >
> > Hungie

_________________________________________________________________
MSN 8 helps eliminate e-mail viruses. Get 2 months FREE*.
http://join.msn.com/?page=features/virus