Difference between molpro and mrcc fci

Hi everyone,

I made an fci/cc-pVDZ calculation from molpro for lithium atom, the fci energy
ROHF -7.432419879671
FCI -7.432636931047

is not the same as mrcc
ROHF -7.4324198796710963
CI(3) -7.432637291077

The convergence thresholds are set to 10^-16. The output files are included. I would like to know why the fci energies seem beyond the difference of convergence threshold.

Thank you

Thank you for reporting this.
1) Please note that the convergence threshold for CC/CI methods can be set by the cctol keyword (not citol).
2) I have found a bug in file ovirt.f which causes problems for CC/CI calculations with standard ROHF orbitals and few electrons as in the case of Li. You will find the corrected version in the download area. Please recompile the code with this.

Dear Prof. Kallay,

I am extremely grateful for your prompt reply.

I tried to download the ovirt file, replaced the old one in mrcc.2017-04-12.tar, compiled them, and used "cctol" command. It seems to me that the CI(3) energy is

-7.432637514965 a.u.

still differs with the molpro fci value considerably.

I am not sure if I have complied mrcc correctly, though. Please find enclosed output file and let me known what kind of additional information would be needed. I am very sorry for spending your valuable time.

Yes, this is exactly what I get. I have also reproduced this energy with other programs.
Probably you should also tighten some convergence criterion in molpro.

Dear Professor Kallay,

I got a reply from Professor Knowles from molpro mailing list. It seems that this issue is related to different exponents in the cc-pVDZ basis set in molpro and mrcc. I managed to reproduce mrcc's value, i.e. -7.432637514965 by modifying the basis set exponent in molpro. Please find enclosed file for the output of molpro.

Due to a technical issue related to compile different versions of mrcc, the inverse check (mrcc to molpro) has not been done.

Anyway, thank you so much for your help