2
$\begingroup$

My Main purpose is using quantum chemistry for simulation of some reactions. Actually we are working on making some new polymers and I need to simulate some molecular properties of these polymers after adding some molecules to them. These include bond breaking and bond making as well as evaluation of optical properties.
That's why I started studying quantum chemistry models. I started with the Schödinger Mechanical Model and then Hartree Fock Theory, but as everyone knows these models are not that efficient for molecules larger than lithium. Then I started learning DFT and got interested in semi-empirical models, but could not learn them yet. Finally I am absolutely confused about which method should I select.
This is my question: As I have not enough time to learn all those quantum chemistry models/techniques, which method is currently best option if I just want to learn ONE method. (I have access to some super-computers in our university.)

$\endgroup$
  • 3
    $\begingroup$ To start a meaningful discussion you need to provide at least the following information. 1) How big your polymers are, i.e. how many atoms do they consist of? 2) What kind of properties do you want to predict? 3) How much computational resources do you have (how many CPUs, how much memory, and how much time)? $\endgroup$ – Wildcat Jul 21 '14 at 10:57
  • 2
    $\begingroup$ Roughly speaking (i.e. without any information provided by the request above), you could go with DFT if the polymers consist of hundreds of atoms. For larger molecules (i.e. thousands of atoms) you are limited to semi-empirical and QM/MM approaches. $\endgroup$ – Wildcat Jul 21 '14 at 11:00
  • $\begingroup$ @Wildcat My polymer is a ring , consisting of around 65 atoms. Mostly we are interested in ring opening reaction feasibility and optical properties of the polymer before and after that ring opening. We have access to super-computers of University of Colorado (USA) but I don't know how powerful they are. I just know they are some of the best available. $\endgroup$ – Aug Jul 21 '14 at 18:31
8
$\begingroup$

The question is by far not an elementary one. Short answer is: unfortunately, there is no silver bullet.

For a molecule of around 65 atoms I would go with DFT as the methodology which provides probably the best quality/cost ratio for molecules of this size.

Now, you need to choose functional and basis set, and this is a tricky business. But to start from:

  1. Do not blindly use B3LYP.
  2. Try to search literature for the same types of calculations (energetics of ring opening reactions, optical properties) for similar compounds. Maybe someone already did the so-called benchmark calculations, i.e. compared on how good different functionals in combination with different basis sets are on reproducing the properties (say, specific rotation) either measured experimentally or calculated at some high level of theory (say, CCSD(T) with reasonably sized basis set).
  3. Do your own benchmark calculations with different functionals and basis sets. First, benchmark different functionals with the same basis set (say, Dunning correlation-consistent basis sets, cc-pVTZ or better cc-pVQZ), vary the functional, compare results with experimental values, choose the best one. Once method is chosen, try to decrease and also increase (if possible) the basis set to see the effect of basis set size (with the above mentioned bases it is quite simple: if cc-pVTZ was used to benchmark different functionals, try the best functional in combination with cc-pVDZ and cc-pVQZ). Choose the smallest basis set that reproduces experimental results with desired accuracy.

P.S. You could also ask Google, there are some resources of information here and there on how to choose method and basis set, though, as I said, this question is not elementary, so only general advice can be given.

$\endgroup$
  • $\begingroup$ Thank you ! I am going to dig into that and will write about here. Thank you again ! $\endgroup$ – Aug Jul 21 '14 at 22:50
  • $\begingroup$ Just as another question: How do you prefer to work with DFT ? do you write your own code or use available softwares? $\endgroup$ – Aug Jul 21 '14 at 22:52
  • 2
    $\begingroup$ @Aug writing your own DFT code would require enormous amounts of time and intellectual efforts, which, taking into account the existence of many extensively debugged effectively parallel DFT codes, is pointless. Just use available software packages. $\endgroup$ – Wildcat Jul 22 '14 at 8:13
  • 2
    $\begingroup$ The choice of software package is another nonelementary question. Basically, people use one which they are mostly familiar with and which is, of course, available on a supercomputer. But since you just start and since (according to your questions) your background in quantum chemistry is rather limited, you would better use a program which someone can help you to use properly and effectively. You need an assistance, at lest at the very beginning. $\endgroup$ – Wildcat Jul 22 '14 at 8:21
  • 1
    $\begingroup$ Strike blindly from number one. There are much more sophisticated and efficient hybrids nowadays - the wonder punsch was a nice addition in the early years... $\endgroup$ – Martin - マーチン Jul 22 '14 at 12:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.