A few questions regarding calculation of molecules

[Chieh-Min]

Dear Edward,

I have a few questions:

1. Is there a way to restart an abnormally terminated calculation? For example, my calculation was terminated in "calc_alpha" when some orbitals are not yet calculated?

2. One of my calculation, when performing conjugate gradient minimization of empty states in "init", the energy is not converged (delta E went down to 0.5E-6 at iteration 299). Which parameter can I change to prevent this problem?

3. How many empty states should one consider? Does that make a difference of the final result of EA and IP?

Thanks and best regards,

Chieh-Min

Chieh-Min Hsieh

PhD student

Institute for Physical and Theoretical Chemistry

University of Bremen

[De Gennaro Riccardo]

Dear Chieh-Min,

I took the freedom to answer your questions.

1. There is a logical keyword named from_scratch, as part of the workflow block, which is set by default to true.
   After setting it to false, you can run again koopmans <your-system>.json and the workflow will restart from the last incomplete calculation (in your case the first unfinished alpha calc).
   
   
2. the threshold on the energy minimization for the occupied and empty states is given by the QE keyword conv_thr. You can modify the default value in the JSON file by adding conv_thr to the calculator_parameters block (followed by the value that you wish, e.g. 1e-6). When you then restart the workflow (as indicated in my previous answer), remember to delete the output file where the energy minimization reached iteration 299. If that calculation had indeed reached the end (indicated by a "JOB DONE" at the end of the corresponding .cpo file) it will be recognised as COMPLETED by the workflow and it will not be rerun.
   
   
3. Regarding the IP, the number of empty states does not affect its value. The IP is calculated as the opposite of the HO energy eigenvalue and it is determined by the densities of the occupied orbitals only. No matter how many empty states you decide to include, the value of the IP should not change.
   Regarding the EA, this depends only on the total charge density (i.e. the total density of the occupied states), and on the density of the LU orbital. While the total density does not change with the number of empty states, understanding whether the LU state might be affected is less trivial. During the optimization of the empty states -- either this is done by minimizing the KI "fake" energy for the empty states (standard procedure for molecules), or by performing a disentanglement of the empty bands and computing the Wannier functions (standard procedure for solids) -- the LU state could be, in principle, modified by changing the number of states. Practically, the change in the LU energy usually is not significant (e.g. I just run a quick test on the ozone molecule and the LU energy does not seem to change by including 1, 3, or 5 empty states).
   However, this is a feature that we did not investigate much, so if you really want to be 100% sure you could run a couple of tests on your system by increasing the value of nbnd.

Hope this helps.

Best regards,

Riccardo

---

From: koopman...@googlegroups.com <koopman...@googlegroups.com> on behalf of Chieh-Min <a0989...@gmail.com>
Sent: Tuesday, October 25, 2022 11:50:36 AM
To: koopmans-users
Subject: [koopmans-users] A few questions regarding calculation of molecules

 

--
You received this message because you are subscribed to the Google Groups "koopmans-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to koopmans-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/koopmans-users/773c1cbf-3fce-4767-9ceb-3a95cf6d50d6n%40googlegroups.com.