[gmx-users] charge group moved too far between two domain decomposition step
Justin Lemkul
jalemkul at vt.edu
Fri Oct 24 04:28:17 CEST 2014
On 10/23/14 10:19 PM, Kester Wong wrote:
> Hi Justin and all,
>
> > >
> > > Thanks for the input, I thought a time step of 2 fs is small enough?
> > >
> >
> > Not always for unconstrained bonds. The largest stable time step is
> > determined by the size of the period of the fastest oscillation, which is
> > vibrations of bonds to hydrogen. 2fs stretches the friendship there, and
> > you would certainly want to equilibrate thoroughly at a smaller time step
> > if you are doing messy things like freezing nanotubes and using pressure
> > coupling (which would not be my go-to method without a clear demonstration
> > that it produces valid results). More generally, equilibration of a
> > not-quite-right structure works better with a smaller timestep, because
> > that copes better with the artificially large forces you get...
> >
> > Thanks for the very informative input, this should serve me well as a rule of
> > thumb for other systems as well.
> >
> >
> > > Should I change my constraint to constraints = h-bonds instead?
> > >
> >
> > That would be normal.
> >
> >
> > > Or,
> > >
> > >
> > > Should I use:
> > >
> > > dt = 0.001 ; 1 fs
> > >
> > > constraints = none
> > >
> > > constraint-algorithm = shake
> > >
> >
> > Plausible, but you want to run the final simulation with constraints for
> > the higher ns/day you obtain...
> >
> > I think I got a little confused here.
> > For energy minimisation, I should impose no constraints to obtain the most
> > realistic starting structure for NVT.
> > As for NVT, one should use constraints to achieve a much quicker equilibration,
> > is that correct?
> >
>
> Constraints generally do not have any real effect on how fast the target
> properties reach an equilibrated state. One thing to note - if you're going to
> run with constraints, you should minimize with constraints. If you have
> problems during EM, setting the water to flexible can help, but you should
> minimize again with constraints, otherwise you often end up with distorted
> geometries that cannot be constrained during dynamics.
>
>
> Yes, I do want to use constraints in later calculations, so I will do the
> two-step minimisation as mentioned.
>
> Isn't setting water to flexible effectively calls for [ settles ], which means
> the calculation will be run with constraints,
> shouldn't the [ bonds ] and [ angles ] restraints be used instead? For this
> reason, I asked "For constraints in a system with tips3p.ipt, and hydronium.itp,
> how does one define "FLEXIBLE" or "CONSTRAINTS"" in the previous email.
>
Note that the topology uses different constructs. One is #ifndef, the other is
#ifdef. With #ifndef, it says "if FLEXIBLE is not defined, use SETTLE," in
contrast to the more direct use of #ifdef.
> Always a good thing to be cautious about the distorted geo. that cannot be
> constrained during MD can be the root of all evil.
>
> > For instance (TIPS3P.itp as shown below), to achieve a higher ns/day in my NVT
> > run, I should have constraints = none in my mdp setting (using [bonds] and
> > [angles]);
>
> If you want greater performance in terms of ns/day, you need a larger time step,
> in which case you do want constraints. Now, the thing you have to consider is
> that if there is a [settles] block, "constraints = none" has no real effect
> since the SETTLE algorithm will be used (constraints present in the topology
> always override "constraints = none" in the .mdp file).
>
>
> This is exactly what I wanted to find out, it is clear to me now that
> constraints = none will have no effect since GROMACS will use what ever
> constraints defined in the topology (if any), as stated in the manual. Thank you
> for that.
>
LINCS is more stable than SHAKE, anyway, so it is preferable.
> According to the manual, SHAKE cannot be used with energy minimisation.
> How does one minimise the water (settles) and ions system without using SHAKE,
> given that the maxwarn-angle for LINCS is 90 degree and the H-O-H angle of
> hydronium (H3O) is 112 degree?
>
Bond rotations and topology angles are unrelated.
> Can both LINCS and SHAKE be turned off during minimisation without constraints
> (constraints = none, i.e. using [ bonds ] and [ angles ] for water and hydronium),
> and then constraint-algorithm = SHAKE for the second minimisation with
> constraints (i.e. using [ settles ] and [ constraints ] for water and hydronium,
> respectively)?
>
Not if the statement in the manual is accurate, but you can constrain with LINCS.
>
> > or have define = -DFLEXIBLE in my .mdp setting to achieve a better energy
> > minimisation (using SETTLES)?
> >
>
> See above.
>
> > [ moleculetype ]
> > ; molname nrexcl
> > SOL 2
> >
> > [ atoms ]
> > ; id at type res nr residu name at name cg nr charge
> > 1 OT 1 SOL OW 1 -0.834
> > 2 HT 1 SOL HW1 1 0.417
> > 3 HT 1 SOL HW2 1 0.417
> >
> > #ifndef FLEXIBLE
> > [ settles ]
> > ; i j funct length
> > 1 1 0.09572 0.15139
> >
> > [ exclusions ]
> > 1 2 3
> > 2 1 3
> > 3 1 2
> > #else
> > [ bonds ]
> > ; i j funct length force.c.
> > 1 2 1 0.09572 376560.0 0.09572 376560.0
> > 1 3 1 0.09572 376560.0 0.09572 376560.0
> >
> > [ angles ]
> > ; i j k funct angle force.c.
> > 2 1 3 1 104.52 460.24 104.52 460.24
> > #endif
> >
> >
> > I would appreciate it if anyone could point me to the right direction, once and for all.
> >
> > For constraints in a system with tips3p.itp, and hydronium.itp, how does one define "FLEXIBLE" or "CONSTRAINTS"?
> >
>
> define = -DFLEXIBLE in the .mdp file. See the manual.
>
> I ought to write the question differently, I was meant to ask, how does one tell
> GROMACS not to impose constraints on both water and hydronium.
>
That depends on the topology construct. You have one case where #ifdef FLEXIBLE
is used to achieve that, in which case you need "define = -DFLEXIBLE" in the
.mdp file. In the case of #ifndef FLEXIBLE, do not use the "define" keyword.
-Justin
> > Below is the .itp for the hydronium model:
>
> >
> > [ moleculetype ]
> > ; molname nrexcl
> > H3O 2
> >
> > [ atoms ]
> > ; id at type res nr residu name at name cg nr charge mass
> > 1 H3O-O 1 H3O OW 1 -0.59 15.9994
> > 2 H3O-H 1 H3O H31 1 0.53 1.008
> > 3 H3O-H 1 H3O H32 1 0.53 1.008
> > 4 H3O-H 1 H3O H33 1 0.53 1.008
> >
> > ; use #ifdef FLEXIBLE or #ifdef CONSTRAINTS
> > #ifdef FLEXIBLE
> > [ bonds ]
> > ;i j funct length (nm) force.c. (kJ/mol)
> > 1 2 1 0.102 ; Hydronium OH bond
> > 1 3 1 0.102 ; Wolf used 1000? check this
> > 1 4 1 0.102 ; Hub used 400,000
> >
> > [ angles ]
> > ; i j k funct angle force.c.
> > 2 1 3 1 112 ; Hub used 400 force
> > 4 1 3 1 112
> > 2 1 4 1 112
> >
> > #else
> > [ constraints ]
> > ; i funct doh dhh
> > 1 2 2 0.102
> > 1 3 2 0.102
> > 1 4 2 0.102
> > 2 3 2 0.169124
> > 3 4 2 0.169124
> > 2 4 2 0.169124
> > #endif
> >
> > [ exclusions ]
> > 1 2 3 4
> > 2 1 3 4
> > 3 1 2 4
> > 4 1 2 3
> >
> >
> > I tried the two approaches, by either having define = -DFLEXIBLE in my .mdp file, or #define FLEXIBLE in my topol.top file,
> >
>
> Either using the "define" keyword or explicitly using #define in the topology
> will trigger the same behavior. The more flexible approach is to use .mdp
> settings, lest you perhaps forget one day and end up doing the run incorrectly
> because you forgot you hacked the topology!
>
> Thanks for the tip! Yes, I do make changes more regularly on .mdp, less so on
> the topology; one can forget to amend the topology for that matter.
>
> Thank you for your input Justin!
>
> Regards,
> Kester
>
>
> -Justin
>
> > and received this error: Incorrect number of parameters - found 1, expected 2 or 4 for Bond.
> >
> > Setting the LD random seed to 3234307754
> > Generated 207 of the 210 non-bonded parameter combinations
> > Generating 1-4 interactions: fudge = 0.5
> > Generated 210 of the 210 1-4 parameter combinations
> >
> >
> > Thank you for your time and patience.
> >
> > Regards,
> >
> > Kester
> >
> >
> > Mark
> >
> > Cheers,
> > >
> > > Kester
> > >
> > >
> > > You should start by using a time step more appropriate to your use of
> > > constraints = none, or vice versa.
> > >
> > > Mark
> > >
> > > On Wed, Oct 22, 2014 at 4:26 AM, Kester Wong wrote:
> > >
> > > > Dear all,
> > > >
> > > >
> > > > I am intrigued by the error that I have received this morning, as follows:
> > > >
> > > >
> > > >
> > > > The charge group starting at atom 37920 moved more than the distance
> > > > allowed by the domain decomposition (1.400000) in direction Y
> > > >
> > > > distance out of cell 7.144197
> > > >
> > > > Old coordinates: 54.499 39.697 14.058
> > > >
> > > > New coordinates: 54.499 39.697 14.058
> > > >
> > > > Old cell boundaries in direction Y: 19.468 32.846
> > > >
> > > > New cell boundaries in direction Y: 18.667 32.553
> > > >
> > > >
> > > > If the charge group moved, if any, then shouldn't the new coordinates be
> > > > showing different values?
> > > >
> > > > It was a new NPT calculation that I continued from a 20ns NVT run, so it
> > > > should be equilibrated well enough, for a simple water on graphene system.
> > > >
> > > >
> > > > Using GROMACS 5.0, I was unable to start the NPT with Berendsen
> > > > thermostat, so I had to go with Parrinello-Rahman with a tau-P at 5.0 (my
> > > > tau-T was 0.5, following the general rule of thumb, that tau-P should be
> > > > larger than tau-T to allow thermal degrees of freedom to equilibrate faster
> > > > than the box size).
> > > >
> > > >
> > > > Should I reduce tau-P?
> > > >
> > > >
> > > >
> > > >
> > > > Below is my NPT.mdp parameters:
> > > >
> > > > ; RUN CONTROL PARAMETERS
> > > >
> > > > integrator = md
> > > >
> > > > tinit = 20000.000
> > > >
> > > > dt = 0.002 ; 2 fs
> > > >
> > > > nsteps = 5000000 ; 10,000 ps
> > > >
> > > > comm-mode = Linear
> > > >
> > > > nstcomm = 10
> > > >
> > > >
> > > > emtol = 0.001
> > > >
> > > > niter = 30
> > > >
> > > >
> > > > ; OUTPUT CONTROL OPTIONS
> > > >
> > > > nstxout = 5000 ; save coordinates every 8 ps
> > > >
> > > > nstvout = 5000 ; save velocities every 8 ps
> > > >
> > > > nstfout = 0
> > > >
> > > > nstlog = 5000 ; update log file every 8 ps
> > > >
> > > > nstenergy = 5000 ; save energies every 8 ps
> > > >
> > > > ;nstxtcout is replaced with nstxout-compressed
> > > >
> > > > nstxout-compressed = 5000 ; no. of steps between writing
> > > > coordinates using lossy compression
> > > >
> > > > ;xtc-precision is replaced with compressed-x-precision
> > > >
> > > > compressed-x-precision = 3000 ; precision with which to write to the
> > > > compressed trajectory file
> > > >
> > > >
> > > > ; Periodic boundary conditions: xyz (default), no (vacuum)
> > > >
> > > > cutoff-scheme = Verlet ; default for GROMACS-5.0
> > > >
> > > > nstlist = 1 ; A smaller nstlist may reduce LINCS
> > > > warnings
> > > >
> > > > ns-type = grid ;
> > > >
> > > > pbc = xyz
> > > >
> > > > ;periodic_molecules = yes ; cannot be used with SHAKE
> > > >
> > > > ; nblist cut-off
> > > >
> > > > rlist = 1.40 ; short-range neighbourlist cutoff
> > > >
> > > > nstcalclr = 1
> > > >
> > > >
> > > > coulombtype = PME
> > > >
> > > > coulomb-modifier = Potential-shift-Verlet
> > > >
> > > > r_coulomb = 1.40 ; short-range electrostatic cutoff
> > > >
> > > > rcoulomb-switch = 0 ; used when potential-switch
> > > > (coulomb-modifier) is specified
> > > >
> > > > epsilon-r = 1 ; default value=1
> > > >
> > > > pme_order = 4
> > > >
> > > > fourierspacing = 0.12
> > > >
> > > > fourier_nx = 0
> > > >
> > > > fourier_ny = 0
> > > >
> > > > fourier_nz = 0
> > > >
> > > > ewald_rtol = 1e-05
> > > >
> > > > ewald_geometry = 3d
> > > >
> > > > epsilon_surface = 0 ; Use 0 when simulation is done with PME
> > > >
> > > >
> > > >
> > > > ;vdw-type = switch is replaced with the following two flags in GROMACS-5.0
> > > >
> > > > vdwtype = Cut-off
> > > >
> > > > vdw_modifier = Potential-switch
> > > >
> > > > ; cut-off lengths
> > > >
> > > > rvdw-switch = 1.00
> > > > rvdw = 1.40 ; short-range van der Waals cutoff
> > > >
> > > > constraints = none
> > > > constraint-algorithm = SHAKE
> > > > continuation = yes ; starting from NVT run
> > > > Shake-SOR = no
> > > > shake-tol = 0.0001
> > > > ;lincs-order = 4 ; accuracy of LINCS
> > > > ;lincs-iter = 24 ; also related to accuracy
> > > > ; Lincs will write a warning to the stderr if in one step a bond
> > > > ; rotates over more degrees than
> > > > ;lincs-warnangle = 30
> > > >
> > > > Tcoupl = v-rescale
> > > > nh-chain-length = 1 ; the leapfrog md integrator only supports
> > > > 1
> > > > ; Groups to couple separately
> > > > tc-grps = GRA SOL
> > > > ; Time constant (ps) and reference temperature (K)
> > > > tau-t = 0.5 0.5
> > > > ref-t = 300 300
> > > > nsttcouple = 20 ; added by k-wong ; not specified in
> > > > Yanbin's MDP
> > > >
> > > > Pcoupl = Parrinello-Rahman
> > > > Pcoupltype = isotropic
> > > > ; Time constant (ps), compressibility (1/bar) and reference P (bar)
> > > > tau-p = 5.0 ; time constant in ps
> > > > compressibility = 4.5e-5 ; isothermal compressibility of water,
> > > > bar^-1
> > > > ref-p = 1.0 ; reference pressure, in bar
> > > >
> > > >
> > > > ; Non-equilibrium MD stuff
> > > > freezegrps = GRA
> > > > freezedim = Y Y Y
> > > >
> > > > gen_vel = no
> > > > ;gen_temp = 300
> > > > ;gen_seed = 173529
> > > >
> > > >
> > > > Thanks in advance!
>
> > > > Regards,
> > > >
> > > > Kester
>
>
>
--
==================================================
Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow
Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201
jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul
==================================================
More information about the gromacs.org_gmx-users
mailing list