Hi @GSQ. Could you please upload your full code and the GAMESS input and output files to a publicly-accessible github page? I will run the code and will debug it.
Thanks @GSQ. The problem is very likely due to a mismatch between the way that GAMESS and ORCA report the final results. I am working on your code and will let you know as soon as I find the problem with ORCA.
Hi @GSQ.
I looked at the data extracted from GAMESS and ORCA and it seems that your code uses the normal modes of ORCA without mass weighting. You should double check the ORCA manual to see if the normal modes are already mass-weighted or not. You must also check the consistency of the units in which the normal mode displacements are printed.
Also, it seems that GAMESS and ORCA give somehow different vibrational frequencies and normal modes. You can visualize the vibrational normal modes and see if they are significantly different or not. In principle, with the same electronic structure method in your calculations, you must get almost identical vibrational frequencies and normal modes.
You can also compare the outputs of GAMESS and ORCA for a smaller molecule such as hydrogen and see if you get identical results. It might help to resolve the inconsistency in the vibrational frequencies and normal modes.
1 Like
I finally solved this problem. Noticed that
when use the normal modes from ORCA, I need to mass-weight it to make it orthogonal and then normalize it. Finally it give me the correct result. Thank you very much!
1 Like
Hello Alain,
you said
However, the Duschinsky matrix I got is not block diagonal. 
Do I need to do some transformations on it?
Hi @GSQ, thank you for your question. We will be taking a look at it and coming back with an answer hopefully on Tuesday. Have a good weekend!
Hi @GSQ. This could be because of the electronic structure method you are using to compute the vibrational normal modes for formic acid, which very likely does not provide two sets of fully decoupled modes. If the shape of the spectrum matches the reference ones given in this demo, no need to perform any transformations.
Thanks for your reply. I wonder how to get two sets of fully decoupled modes. Because it can reduce the time of calculation and the hardware demand efficiently.
Hi @GSQ, the discussion on the Duschinsky matrix being block diagonal, was based on data taken from this paper. You might follow the procedure in the paper for performing the electronic structure calculations to obtain the vibrational frequencies and normal modes. It might require going beyond DFT though.
1 Like