[gmx-users] Coarse-graining and tabulated non-bonded potentials- will write

Berk Hess gmx3 at hotmail.com
Mon Dec 3 16:59:49 CET 2007

>From: Steven Kirk <Steven.Kirk at hv.se>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: gmx-users at gromacs.org
>Subject: RE: [gmx-users] Coarse-graining and tabulated non-bonded 
>potentials- will write
>Date: Mon, 03 Dec 2007 16:17:03 +0100
>Berk Hess kindly answered my original query below. I have some more 
>questions that might be of general interest to the list, so I'm conducting 
>another step of the conversation in the list.
>Date: Mon, 26 Nov 2007 14:18:36 +0100
>From: "Berk Hess" <gmx3 at hotmail.com>
>Subject: RE: [gmx-users] Coarse-graining and tabulated non-bonded
>	potentials -	will write
>To: gmx-users at gromacs.org
>Message-ID: <BAY110-F32D5E7B7CD6C268A5699518E750 at phx.gbl>
>Content-Type: text/plain; format=flowed
>>>From: Steven Kirk <Steven.Kirk at hv.se>
>>>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>>>To: gmx-users at gromacs.org
>>>Subject: [gmx-users] Coarse-graining and tabulated non-bonded potentials 
>>>- will write up on the wiki
>>>Date: Mon, 26 Nov 2007 13:57:37 +0100
>>>Hello all,
>>>Firstly, many thanks to everyone who has contributed useful advice to me 
>>>over a number of years using GROMACS.
>>>I have performed a large number of potential of mean force calculations 
>>>for the forces acting between two approximately spherical amorphous 
>>>silica particles (various sizes < 5 nm diameter) in TIP4P water with PME 
>>>electrostatics and varying concentrations of background ions.
>>>Now I want to 'coarse-grain' the simulation, treating each silica 
>>>particle as a single point mass, and use the interaction potential 
>>>between the particles obtained from the PMF results as a tabulated 
>>>potential in mdrun, to allow longer time- and size-scales to be 
>>>investigated (I want to examine colloidal aggregation behaviour of these 
>>>I have tabulated the PMF potential and its derivatives as a function of 
>>>centre-of-mass separation of the particles as suggested in the manual 
>>>(but I can only tabulate for COM-COM distances greater than the contact 
>>>distance [ = 2r in the hard-sphere approximation] out to some cutoff).
>>>Will I have to add a short-distance 'hard-sphere' wall to my tabulated 
>>I don't really understand the issue here.
>>You (would) also directly get the repulsive part of the PMF from a 
>>simulation. I assume that you mean that you have not done simulations
>>to obtain the PMF at shorter distances?
>>If not, just do so.
>Yes, I have calculated the PMF at shorter distances, and it becomes 
>repulsive at shorter distances.
>My confusion was caused by my uncertainty as to what should be in the 
>tabulated potential file for very short distances (tabulated potentials 
>must be tabulated for distances r >= 0, even if I only have real tabulated 
>potential data from some r_min (>0) upward). A quick check of the example 
>'table6-9.xvg' file for the directory referenced in the standard GROMACS 
>distribution using the environment variable GMXLIB has shown me that I can 
>safely 'pad out' all the columns for distances 0 <= r < r_min with zeroes.
>>>I have read the appropriate section of the manual on tabulated 
>>>interactions, and am working on building an appropriate topology. IU am 
>>>assuming that my coarse-grained particles consist of the silica particles 
>>>plus a number of surface counterions in such a way that the particles 
>>>will be electrically neutral, so presumably I can set all the entries in 
>>>the tabulated potential file for the Coulomb terms to zero. If my 
>>>understanding of the manual is correct, I can then introduce my PMF 
>>>potential in one or other of the g() and h() columns, along with its 
>>>appropriate derivatives.
>>>The plan is to randomly place a number of the coarse-grained particles in 
>>>a simulation box and choose an appropriate time step and thermostat to 
>>>run aggregation simulations. Some of the literature I have read suggests 
>>>timesteps of around 10^-6 s and Brownian dynamics - can anyone comment on 
>>>the advisability of these choices? I anticipate significant aggregation 
>>>within ~seconds.
>>>Another issue is whether or not to include coarse-grained water and 
>>>explicit ions in the simulation box. Recent postings on this list have 
>>>suggested that Lagrangian dynamics should not be done in a vacuum, so 
>>>presumably the same is true for Brownian dynamics? Has anyone on the list 
>>>used 'coarse-grained' water in their GROMACS simulations (references 
>>>If I have extracted different tabulated potentials for each background 
>>>counterion concentration, this information is presumably 'built in' to my 
>>>extracted PMF interparticle potentials, and I shouldn't need to include 
>>>explicit counterions in my coarse-grained simulation, correct?
>>Because of the way you did things, everything is included in the PMF,
>>both water and counterions.
>>If this is an accurate approach is a completely different matter.
>>Now you have the 2-body coarse-grained term correct, but there
>>could of course be multi-body non-additive effects.
>>For such effects you might need coarse-grained water or counterions,
>>but then things get much more complicated.
>I'm uncertain that if my particles have a persistent (though fluctuating) 
>electric dipole moment throughout the PMF runs, this will also be fully 
>accounted for by the PME electrostatics used during my PMF runs. I have 
>read a paper
>1. R. Blaak, M. A. Miller, and J.-P. Hansen, “Reversible gelation and 
>dynamical arrest of dipolar colloids,” Europhysics Letters (EPL) 78, no. 2 
>(2007). doi: 10.1209/0295-5075/78/26002
>where the authors have used GROMACS and added an *explicit* dipole to 
>colloidal particles.
>Can I use such an approach in further work based on my PMF potentials, or 
>by doing so would I be counting the electric dipole effects twice?
>You mention that things get complicated in this situation. Do you (or 
>anyone reading this on the list) know how to implement such 3-body terms in 
>GROMACS (presumably using *assumed* and unchanging bonded angular terms 
>where each particle becomes an 'atom' in a bigger 'molecule'), or 
>alternatively another free MD code that can handle both dynamic bonding 
>*and* multibody potentials ?
>Again, many thanks in advance for all responses.

If there are important dipole contributions, the best thing is to implement
dipoles in the coarse-grained model. You would then explicitly take the most
important multi-body contributions into account.
But with explicit dipoles you can no longer directly use the PMF for your
coarse-grained simulations. You will have to subtract the dipole-dipole
interactions of the coarse-grained model to avoid double counting.
If this is a simple dipole-dipole term. you can analytically determine
the (Boltzmann averaged) dipole-dipole interaction to the PMF
of the coarse-grained model


Live Search, for accurate results! http://www.live.nl

More information about the gromacs.org_gmx-users mailing list