[gmx-users] Pressure very high with chloroform and gromacs-4.5.5

Mark Abraham Mark.Abraham at anu.edu.au
Thu Oct 13 00:21:23 CEST 2011


On 13/10/2011 12:35 AM, Nuno Azoia wrote:
> I know that from the beginning. That's why I found very strange the
> increasing on the system volume. My initial setup have a density of
> ~400g/L, and for liquid chloroform is ~1400g/L.
> Using gromacs-4.5 I get densities of about 200 after 750ps simulation
> time, and using 4.0 I get almost 500 after 1ns.

Your chloroform parameters are probably only valid for a condensed 
phase, and your initial density is nowhere near that. Please follow the 
approach here 
http://www.gromacs.org/Documentation/How-tos/Non-Water_Solvation.

>
> And I was using such low densities because I build the box using
> genconf with the options -nbox and -dist, starting with one molecule
> of chloroform. If I used small distance values I get LINCS warnings in
> the energy minimization step. Using genbox to solvate one empty box
> with chloroform, I get very strange behaviors, with a very large
> amount of molecule overlap.

Shrug. Without command lines and descriptions of what is in your input, 
we are helpless. Please follow the above advice.

>
> But as you said before, I now think that this should be also a
> paralelization problem, not to mention the numerical instability
> caused by this very low density.
> And I think this is a paralelization problem associated with this
> numerical instability judging from the system behavior.
> In 4.5, using the option -nt in mdrun, I get some small droplets of
> liquid chloroform, and in 4.0, using mpi, I get one large droplet of
> chloroform. It looks like if 4.5 is treating the system like
> independent small systems, while 4.0 is able to divide the system into
> small, but interdependent systems.

No, the implementation of parallelization might have an effect here if 
you were starting from a reasonable approximation of a condensed phase. 
These two observations are just two of the possible things that could 
happen to a liquid at extremely low density.

Mark

>
> On Wed, Oct 12, 2011 at 12:26 PM, Mark Abraham<Mark.Abraham at anu.edu.au>  wrote:
>> On 12/10/2011 8:04 PM, Nuno Azoia wrote:
>>> Thank you Mark for your answer!
>>>
>>> I agree with you when you say that the high pressure is not itself a
>>> problem. To avoid that problem I change my system from 1000 molecules
>>> to 27000 molecules, and the pressure change from several thousands to
>>> several hundreds bar.
>>> What I found very strange was the increasing volume (and pressure). My
>>> system is very unstable, I know, because the density is very low, and
>>> not the opposite, so I was expecting to see the box getting smaller.
>>> Looking to the trajectory I can see a box almost empty (with empty I
>>> mean empty space, with chloroform molecules aggregating in small
>>> droplets), and that's why I found the system behavior very strange.
>> Sounds like your initial density might be far too low. grompp reports it, so
>> do check.
>>
>> Mark
>>> I will follow your suggestion concerning nsttcouple and nstpcouple and
>>> the other initial conditions.
>>>
>>> Nuno Azoia
>>>
>>>
>>> On Wed, Oct 12, 2011 at 6:24 AM, Mark Abraham<Mark.Abraham at anu.edu.au>
>>>   wrote:
>>>> On 12/10/2011 2:22 AM, Nuno Azoia wrote:
>>>>> Hello!
>>>>>
>>>>> I found something very strange while making a CHCl3 box using
>>>>> gromacs-4.5.5.
>>>>> A look the mailing list, the manual and some release notes for
>>>>> gromacs-4.5 and I couldn't found the answer for my problem. It's
>>>>> possible that I'm doing something wrong, but I can not find what, so
>>>>> I'm describe my problem.
>>>>>
>>>>> I start a chloroform box from scratch, using genconf, and I get o
>>>>> chloroform box with 1000 molecules. I get energy minimization without
>>>>> problems.
>>>>> Then I've run some equilibration steps in a NVT ensemble and in the
>>>>> end I get pressure in the order of hundreds of bar.
>>>> The high pressure is not itself a problem - small numbers of molecules
>>>> and
>>>> short simulation times lead to doubtful statistics for pressure.
>>>>
>>>>>   Then I change to
>>>>> NPT conditions and both the pressure and the volume keep increasing
>>>>> with time.
>>>> Be sure to visualize your trajectory to confirm its behaviour matches
>>>> what
>>>> you expect from the trends in P and V.
>>>>
>>>>> To discard the possibility of a size problem, I repeat everything with
>>>>> a box of 27000 molecules, with a volume of ~13800 nm^3. The problem
>>>>> was the same. Very high pressures (150-200 bar) and very low densities
>>>>> (<      200 g/L) after 750 ps simulation time. And both volume and
>>>>> pressure
>>>>> increasing with time.
>>>>>
>>>>>
>>>>> I'm doing now the same procedure, but using gromacs-4.0.7 and I'm
>>>>> getting very different (and better) results. After energy minimization
>>>>> I run 50000 steps in a nvt ensemble and I got pressure around -30 bar
>>>>> (Ok for me). After that I start to run the simulations in npt ensemble
>>>>> and the pressure start to increase slowly, with negative values
>>>>> because the system have very low densities (~400 g/L), and the volume
>>>>> is decreasing. So I'm getting the normal reaction from the system.
>>>>>
>>>>>
>>>>>
>>>>> Where is the problem? There are some different parameters to set in
>>>>> the mdp file and I didn't realize that, or is this a problem in
>>>>> gromacs-4.5?
>>>> It seems you are generating some numerical instability with your choice
>>>> of
>>>> initial and simulation conditions, and that gromacs-4.0 was luckier in
>>>> avoiding the problem. If this is a parallel simulation, then the way
>>>> communication is managed in order to organize T and P coupling changed
>>>> going
>>>> up to 4.5. For equilibration, I would encourage setting nsttcouple and
>>>> nstpcouple to 1. Using a shorter time step can help with numerical
>>>> instability. Berendsen T coupling or a larger coupling time constant
>>>> might
>>>> also help. After gentler NVT, then switch to NPT, let the box size
>>>> stabilize, and finally move to larger coupling periods, time steps and
>>>> v-rescale T coupling.
>>>>
>>>> Mark
>>>>
>>>>> In both cases I used this parameters:
>>>>>
>>>>> -------------------------------------------------------------------------
>>>>> integrator          =  md
>>>>> dt                    =  0.002
>>>>> nsteps              =  50000
>>>>> nstcomm          =  1
>>>>> nstxtcout           =  500
>>>>> xtc-precision      =  1000
>>>>> nstxout             =  0
>>>>> nstvout             =  0
>>>>> nstfout             =  0
>>>>> nstlog              =  500
>>>>> nstenergy           =  500
>>>>>
>>>>> nstlist               =  5
>>>>> ns_type             =  grid
>>>>>
>>>>> ;Reaction field
>>>>> rlist               = 0.8
>>>>> coulombtype         = Reaction-field
>>>>> rcoulomb            = 1.4
>>>>> epsilon_r           = 1.0
>>>>> epsilon_rf          = 4.8
>>>>> vdwtype             = cut-off
>>>>> rvdw                = 1.4
>>>>>
>>>>> ; temperature coupling
>>>>> Tcoupl              =  v-rescale
>>>>> tc-grps             =  CHCL3
>>>>> tau_t               =  0.05
>>>>> ld_seed             =  -1
>>>>> ref_t               =  300
>>>>>
>>>>> ; Energy monitoring
>>>>> energygrps          =  CHCl3
>>>>>
>>>>> ; Isotropic pressure
>>>>> Pcoupl              =  no
>>>>> Pcoupltype          =  isotropic
>>>>> tau_p               =  0.5
>>>>> compressibility     =  4.5e-5
>>>>> ref_p               =  1.0
>>>>>
>>>>> ; Generate velocites is on at 300 K.
>>>>> gen_vel             =  yes
>>>>> gen_temp            =  300
>>>>> gen_seed            =  -1
>>>>>
>>>>> constraint_algorithm=lincs
>>>>> lincs_order         =  4
>>>>> lincs-warnangle     = 90
>>>>> constraints         =  all-bonds
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>> and of course, for the npt ensemble I just change
>>>>> _________________________________________
>>>>> Pcouple     = berendsen
>>>>>
>>>>>
>>>>> gen_vel      = no
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>> Thank you in advance
>>>>>
>>>> --
>>>> gmx-users mailing list    gmx-users at gromacs.org
>>>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>>>> Please search the archive at
>>>> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>>>> Please don't post (un)subscribe requests to the list. Use the www
>>>> interface
>>>> or send it to gmx-users-request at gromacs.org.
>>>> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>>
>>>
>> --
>> gmx-users mailing list    gmx-users at gromacs.org
>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>> Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>> Please don't post (un)subscribe requests to the list. Use the www interface
>> or send it to gmx-users-request at gromacs.org.
>> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>
>




More information about the gromacs.org_gmx-users mailing list