I am sorry for you. :-)

You have to do the calculation for a significant number of orientations, and you have 3 orientation angles for a pair of dimers. This means a lot of computations. If your dimers are symmetric, you can reduce the number of orientations to compute.

Best write a shell script that creates and submits Gaussian jobs, and then take a vacation!

… and in addition to Victor's response, remember to use basis sets which are sufficiently diffuse to cover all configurations of your dimer, e.g., augccpvXz (X = D, T, q, 5). If van der Waals interactions play a major role in your dimer, you have to use post-Hartree-Fock methods with Gaussian, e.g., coupled-cluster methods.

Thanks to you both ...

hmm Ulrich i guess use of such higher basis set will take more time to process the job. Also, there will be infinite number of orientations possible based on distances and angles and so to study all possible configurations will take more than a year.

I have read a few publications where people have reported that they have done computational studies on dimeric, trimeric interaction but method was not mentioned in detail.

So i think there must be some other way to model dimers, trimers, and higher aggregates.

You have not told us what kind of molecules you are considering. If the dimer is held together by covalent bonds or hydrogen bonds, the Hartree–Fock method will capture most of the interaction energy; also, you will be able to identify a relatively small number of conformers. Then the computation does not require very much time.

But if you are considering van der Waals dimers even of something simple like nitrogen, you are facing a major computational task. But it is not hopeless. You need many orientations, but not an infinite number. If you have access to a good multi-processor server, the effort is manageable. In fact, good pair potentials have already been computed ab initio for noble gases, H2, N2, CO, CO2, CH4, H2S, and several more. But this is not something that one can accomplish over a weekend …

Dear Hongguang,

perhaps we should first ask what you need the dimer energy for. Do you merely need the structure and energy of the most stable dimer [A]? Do you want to calculate 2nd virial coefficients, dilute-gas viscosities etc. ab initio [B]? Or do you want to perform computer simulations of fluids and/or solids [C]?

[A]: You can often obtain a decent first estimate of the dimer structure if you do a Hartree-Fock calculation and then use empirical Lennard-Jones pair potentials to account for vdW interactions; then you can optimize that structure with a full coupled-cluster calculation; you will not have to try out too many configurations.

[B]: You need a full pair potential, which means a lot of CPU hours.

[C]: You need a full pair potential, but not necessarily an ab initio potential. In dense phases, three-body effects contribute significantly to the thermophysical properties, and simulations based on ab initio *pair* potentials will not give good results. Depending on your objectives, it may be sufficient to use effective Lennard-Jones potentials for the vdW interactions (e.g., TraPPE potentials) plus a charge distribution obtained from quantum mechanical calculations. Alternatively, you can try to construct a full ab initio pair potential and add some three-body potentials – but this would really mean a major effort.

in weakly bonded dimers do not forget to think of dispersion effect. Situation is even more complicated when solvent plays a significant role (e.g. stabilizing the dimer)

One should also keep in mind whether the dimer in a gas phase or in a matrix or maybe part of a polymer. In gas phase you really may have to calculate energies for all relative orientations. In a matrix, there may be restrictions, eg., the dimer is required to maintain a planar geometry. In a polymer situation, the omitted part of the polymer may also put constraints on the geometry. In both situation, experimental geometries may be available for the dimer and the monomers. Also these may be used to put constraints on the relative orientations and distances of the monomers. As already said, there may also be symmetry constraints. I guess that most dimeric, trimeric, .. studies did not consider all relative orientations but only varied some relative orientations, and distances of the monomers in a certain range. Of course, one can afford to calculate the more geometries the more simple a model is used. Thus, a simple model calculation can be used to find the relative orientations that should be investigated by more expensive models.

> One should also keep in mind whether the dimer in a gas phase or in a matrix or maybe part of a polymer …

… or in a solution. Solvent/matrix effects can be surpising. A simple, yet very important example is the “hydrophobic association” of nonpolar molecules dissolved in water.

Dear Hongguang,

60 atoms is a lot! Unless you happen to be the admin of the new Tianhe2 supercomputer, I would not recommend a high-level coupled-cluster calculation. If you merely want to compute charge transfer integrals, you probably do not need to take electron correlation; then Hartree–Fock calculations should be sufficient. But you should use diffuse basis sets (e.g., augcc…) and try to extrapolate to infinite basis set size.

Dear Ulrich,

I am working on molecules where hydrogen bonding and pie-pie stacking plays a major role so now i have initiated Hartree-Fock calculations by selecting smaller basis sets. I am looking for optimized geometry of dimers and higher aggregates but now as you mentioned solvent effect also plays significant role so now i am looking for basis sets and computational models which will take all these factors into account. I am using supercomputer for jobs to run but i do i really need to go for coupled -cluster calculations.

I am new to dimeric structure optimization so it will be really nice if somebody suggest me good review or books to begin with.

Thanks to all

For initial optimization process use DFT with small basis sets. It's a lot faster. Basic HF calculations have a greater deviation than the DFT stractures.

Dear Jean

I was also thinking to perform DFT-D calculation but latter on i changed my idea as i didn't find many reports on it. For dimeric systems, people have reported mainly Hartree-Fock and MP2 calculations by simultaneously considering BSSE. As far as i know, the problem of effect of solvation (in case of aggregates) is still not encountered to the levels, so i have also putted my head to come out of this labyrinth of theories and models.

Once you get the optimized structure with DFT compare it with MP2 and MP4. Also try cluster methods as well. They are very slow but once you have a good structure they go much faster.

Dear Malta,

Thanks for very detailed and informative suggestion and i have initiated my work like the way you suggested with little modifications. and yup i will first begin both vacuum level calculations and latter on will switch to different solvent system. I will recommend Turbolmole to our computer center where we have supercomputer for computation. And finally thanks for suggesting QZVP basis set as i was also searching for a proper basis set which minimizes BSES.

You need to consider the dispersive interaction (electron-electron correlation) with dispersive interaction corrected DFT or MPn or CCSD(T) methods.

Before developing the DFT-D, MPn and CCSD(T) generally used to describe dispersive interaction.

I have been studied pi-pi interaction between benzene and benzene. In this case, you need to consider the three types of interaction modes, which can illustrate stacking, displaced, and T-shaped benzene dimer.

If you need to calculate dispersive interaction of small molecules, you need to use MPn or CCSD(T)

However, DFT-D method is essential in .large molecules

Thanks Sunwoo Kang for sharing the information.

Another way to analyse interactions between molecules is perturbation theory. In symmetry-adapted perturbation theory (SAPT) you obtain directly the interaction energy as a sum of various physically sound contributions: electrostatics, induction, dispersion and their so-called exchange counterparts, which arise from imposing antisymmetry on approximate wave functions. SAPT is programmed as a freeware program (http://www.physics.udel.edu/~szalewic/SAPT/programs.html), it is also available in Molpro, and some other packages. The SAPT version where interacting molecules are described on a TD-DFT level is capable to handle large cases.

There is a very interesting method to analyse weak non-covalent interaction regions, the Reduce Density Gradient method (RDG). This methodology can distinguish between H-bonds, Van deer Waals and steric repulsion interactions, between molecules in a dimer. All you have to know is the atomic coordiantes of your dimer. There is a ver y good and detailed paper describing this method,

http://pubs.acs.org/doi/abs/10.1021/ja100936w

With the RDG calculations, you can generate isosurfaces that reveal the existing interactions. Then, using a color scale, similar to a MEP surface, you can identificate the different kind of interactions that your dimer exhibits. There are some free softwares, such as Multiwfn (http://multiwfn.codeplex.com) which is very easy to use, to implement the RDG calculations.

Interesting: somebody does not like SAPT. Or maybe Molpro? I got one "downvote" for my answer from 17 July. Perturbation theory is an appropriate tool to study "different types of interaction" for a "dimeric" system, and can really improve understanding of physics of the interaction, so I really do not understand what one can "dislike" here. I would prefer that this person gives some comment etc., so that I can address this "dislike" and possibly improve my answer. Making just a downvote without a comment is useless in my opinion, unless the original comment is really bad or completely off-topic.

Good afternoon, Dr. Korona. I am not a person who dislike SAPT, but i am really interested in it (sorry if i choose a wrong topic, i am pretty new here). How good (or how bad) is SAPT-DFT for charge-transfer systems? I already read "Charge-transfer from Regularized Symmetry-Adapted Perturbation Theory" of Alston J. Misquitta and papers before it, but still don't get an answer for this question.

@Tatiana Korona

I would like to use SAPT, but I do not figured out the software.

Is there some tutorials, examples of how to use it?

@Alexey Nikitin & Gess Sharapa

Travis has already answered your questions. I can only add that CT is a tricky issue indeed. In the moment of the definition of a pertubation theory (PT) operator V [where H(X) is the electronic Hamiltonian of molecule X]:

H(AB)=H(A)+H(B)+V

we start to assume that we can divide AB into A and B and that both new molecules keep their electrons for themselves, only electron densities of A and B get deformed. Therefore, PT cannot handle reactions like A+B->A(-)+B(+) directly, because. in PT we calculate what quantity we should add to the sum of the total energies of A and B in order to obtain the total energy of AB, and this quantity is called the interaction energy E_int(AB):

E(AB)=E(A)+E(B)+E_int(AB).

If this division is not sensible enough, we obtain anomalously large interaction energies - in an ideal case when PT is convergent, or rubbish - if PT has difficulties to converge.

We've developed Dispersion-Correcting Potentials specifically to improve the ability of B3LYP to predict non-covalent interactions. They are easy to use with the Gaussian program, see: www.ualberta.ca/~gdilabio. We've demonstrated that the B3LYP-DCP approach is one of the best for computing the interaction energy between molecular dimers. Presently we have DCPs for H, C, N, and O, with additional atoms coming soon.