Thanks to Paul Tangney for a very comprehensive and useful explanation. I believe that you are absolute right wrt to the issues to the electrostatic issues that arise in DFT vs. classical MD calculations.
I would like to expand upon the issue that I am running into in the hopes that somebody on the list may have an idea. I am studying liquid nitromethane in a VERY small unit cell. There are only 8 nitromethane molecules. Already a problematic systems because the predominat interaction is vdW. It is also at a very low density. I run an NVT simulation for about 8 ps. I wrote a seperate program that takes each nitromethane molecular and calculation the translational kinetic energy, rotation kinetic energy of the molecular (I recalculate the moment of inertia at each step) and outputs a temperature based on those quantities, e.g. T_trans=KE_trans/(3*N_mol) Anyhow, if the system is properly equilibrated, I would think that the average T_trans and T_rot would be equal to the imposed temperature by the thermostat. However, I am finding that it is lower by 20-40% lower. My only ideas are that: 1. That this system requires ALOT more time to equilibrate. 2. That initially all the energy must be getting locked into intramolecular vibrations. Ideas anyone? On 1/4/07, Paul Tangney <tangney at civet.berkeley.edu> wrote: > > > Hi Nichols, > > Dipole corrections to what quantity ? > > There is now a large literature on macroscopic polarization > in the context of DFT (Resta, Vanderbilt and others) but if > I understand your problem correctly, it is not necessary for > you to delve into this. > > Slabs with net dipole moments under periodic boundary > conditions are problematic because, in a slab calculation, > you don't generally want the periodic images to 'see' each other, but > a slab with a strong moment interacts with its images. > If a dipole approximation is valid (if the spatial separation > of charges is small compared to your unit cell) the dipole-dipole > interaction decays with 1/r^5. > > For a bulk liquid, the problem should be much less serious because > it is less ordered and there is no vacuum. > There shouldn't be any strong multipole moments and any moments that > do exist should be well screened and transient. > > Finite size effects are *much* *much* more serious for simple > classical MD - particularly when only point-charge electrostatics > are included. The reason is that, even in strongly ionic materials > such as NaCl, water or silica, very simple electronic screening mechanisms > (such as screening from polarizable atoms) are sufficient > to kill electrostatic interactions within a few nanometers..and usually > within 1 nm. On the other hand, finite size effects in MD with simple > classical potentials are large because the only screening is from > the ions themselves and occurs on ionic time scales. > As a result, some physical properties (e.g. thermal expansion) require > 1000 to 10000 atoms for convergence while the same property with > DFT is converged with 50 to 100 atoms. > > I have been waving my hands while typing this - I'm going by > my experience but these issues are poorly controlled, and > should be investigated more thoroughly. > Hopefully somebody will have the time some day. > > Best regards, > > Paul > > > Hi, > > > > This is a question regarding the NVT MD simulation of a polar liquid, > i.e. a > > molecular fluid with a dipole moment (e.g. water, nitromethane, etc.) > > > > When dealing with a polar slab, there are dipole corrections to > consider. > > This comes from a net dipole moment. Is there a similar issue with the > NVT > > MD simulation of a polar liquid? My co-works here (Army Research Lab) > have > > told me that there are such issues with there classical (non-DFT) > simulation > > but that the corrections become smaller as one increases the system > size. > > > > Does anyone no of such discussion at the DFT level which may be found in > > the literature? > > > > Bests, > > > > -- > ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo > Paul Tangney > Theory of Nanostructured Materials Facility > The Molecular Foundry > Lawrence Berkeley National Lab. E-mail: PTTangney at lbl.gov > 1 Cyclotron Road, Bldg 67 Phone: (510) 495-2769 > Berkeley, CA 94720 > ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- Nichols A. Romero, Ph.D. 1613 Denise Dr. Apt. D Forest Hill, MD 21050 443-567-8328 (C) 410-306-0709 (O) -------------- next part -------------- An HTML attachment was scrubbed... URL: /pipermail/attachments/20070117/10ad35ca/attachment.htm
