7
$\begingroup$

I'm finishing my Ph.D. in Inorganic Chemistry and having a debate with my supervisor. The work in question have two organometallic compounds A and B. The structures are based on X-Ray Diffraction, so I already have a good molecular frame and just need to optimize the hydrogens. For that I'm using PBEh-3c that usually work well for organometallics. In the next steps, using the now pre-optimized structure, I have changed the level of theory for each task, selecting the (supposedly) best level for each, based on benchmarks:

  • Broken-symmetry (magnetic coupling): TPSS/Def2-TZVP
  • Bond Orders and spin distribution: SCS-MP2/Def2-TZVP
  • Energy Decomposition: DLPNO-CCSD(T)/Def2-TZVP
  • TDDFT: wB2PLYP/Def2-TZVP

My Supervisor is not comfortable with the mix of levels of theory, but I believe that it is OK if I'm sure that I have the correct ground-state geometry and I'm not comparing the same property for A and B with different methods (for example TDDFT using PBE for compound A and wB2PLYP for compound B).

So, the question is: is it OK to change the level of theory of my calculations to select the best setup for the property I'm trying to calculate if I know that I have a good ground-state geometry and that my calculation is not displacing the geometry?

Improve this question
$\endgroup$
6
  • $\begingroup$ Hi, @Buttonwood. I have not added H, just optimized the H positions using DFT because, as you said, their positions are idealized. I have listed the properties I'm interested in: broken-symmetry, absorption using TDDFT, bond orders... $\endgroup$
    – HCSthe2nd
    May 27, 2021 at 14:00
  • 2
    $\begingroup$ I am slightly curious as to why you are not using DLPNO-CCSD(T) for all of the methods if you can afford it. I might be wrong, but I thought that DLPNO-CCSD(T) would be the most accurate of all of the methods. $\endgroup$
    – S R Maiti
    May 27, 2021 at 21:23
  • $\begingroup$ @ShoubhikRMaiti I'm using it to get the Local Energy Decomposition calculated by the ORCA software. It is based on DLPNO-CCSD(T), that gives an almost full CCSD(T) result while being less expensive. And, actually, I can only afford a single point. $\endgroup$
    – HCSthe2nd
    May 27, 2021 at 22:32
  • 2
    $\begingroup$ The way you are asking this question it appears that you are doing full structure optimisations for each property which is probably not the case. There should be some level of calibration of the ground state structure. I'm not sure that PBEh-3c is a good enough guess on the structure, so that should be made crystal clear. After that, if you qualify your results well enough, I see absolutely no issue with using the best tool for the job. I'm sure you've done your research and found benchmarks (or similar systems computed similarly); quote that and you should be fine. $\endgroup$
    – Martin - マーチン ♦
    May 27, 2021 at 23:17
  • 3
    $\begingroup$ I find this a very useful chemistry question and totally on topic here. (And I prefer to keep it here.) However, it would be totally unfair to not tell you about Matter Modeling. It might have been asked there already, so it's probably worth checking out. $\endgroup$
    – Martin - マーチン ♦
    May 27, 2021 at 23:23

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.