The MRCC interface ignores the MAXIT directive

[Tibor Győri]

Hello everyone,

For testing purposes, I wanted to do a CISD calculation using MRCC. The CISD has not converged within 50 iterations (the MRCC default limit), so I tried to increase the limit, like this:

{mrcc,method=cisd,memory=1400
maxit,150}

The MRCC program unfortunately still quits after 50 iterations, as far as I can tell the maxit directive is silently ignored.

I am aware that the CI non-convergence is a sign that the HF reference is not great, I just want to make sure that the Molpro and MRCC CISD energies match.

I also tried this: {mrcc,method=cisd,memory=1400;maxit,150} with exactly the same results.

Is this a bug in one of the programs, or is my input file wrong? I am using Molpro 2015.1.41 and MRCC 2019-02-09.

[Peterson, Kirk]

Dear Tibor,

Unfortunately maxit is not an option available in the molpro interface to mrcc. 

Regards,  -Kirk

Sent from my iPhone

--
You received this message because you are subscribed to the Google Groups "molpro-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to molpro-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/molpro-user/167feadd-91bd-4c8c-a2a8-5eb34f5fec67%40googlegroups.com.

[Tibor Győri]

That's a bummer. This however also means that the documentation is incorrect.

Here ( https://www.molpro.net/info/2015.1/doc/manual/node378.html ) it clearly states that "the CORE, OCC, ORBITAL, MAXIT, directives work as in the MOLPRO CCSD program".

Sincerely,

Tibor Győri

2019. július 22., hétfő 12:29:56 UTC+2 időpontban Peterson, Kirk a következőt írta:

Dear Tibor,

Unfortunately maxit is not an option available in the molpro interface to mrcc. 

Regards,  -Kirk

Sent from my iPhone

On Jul 20, 2019, at 2:35 PM, Tibor Győri <tibo...@gmail.com> wrote:

Hello everyone,

For testing purposes, I wanted to do a CISD calculation using MRCC. The CISD has not converged within 50 iterations (the MRCC default limit), so I tried to increase the limit, like this:

{mrcc,method=cisd,memory=1400
maxit,150}

The MRCC program unfortunately still quits after 50 iterations, as far as I can tell the maxit directive is silently ignored.

I am aware that the CI non-convergence is a sign that the HF reference is not great, I just want to make sure that the Molpro and MRCC CISD energies match.

I also tried this: {mrcc,method=cisd,memory=1400;maxit,150} with exactly the same results.

Is this a bug in one of the programs, or is my input file wrong? I am using Molpro 2015.1.41 and MRCC 2019-02-09.

--
You received this message because you are subscribed to the Google Groups "molpro-user" group.

To unsubscribe from this group and stop receiving emails from it, send an email to molpr...@googlegroups.com.

[Nike Dattani]

Dear Tibor,

A possible workaround might be to print the integrals in FCIDUMP format in MOLPRO, then convert the file into fort.55 format (MRCC's integral format, which is exactly the same, except for the pre-amble at the top), then run MRCC standalone with the keyword "iface=cfour" which tricks MRCC into thinking that CFOUR had written the fort.55 already and that MRCC is ready to go. the MAXIT keyword will certainly work from MRCC standalone.

With best wishes!
Nike

To unsubscribe from this group and stop receiving emails from it, send an email to molpro-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/molpro-user/14aaecea-e1ff-48a2-847b-668f90c011dc%40googlegroups.com.

[Peterson, Kirk]

Umm, you’re right about the documentation. I’ve run into this problem with the CCSDT and CCSDTQ code where I found the max iteration count in MRCC was hardwired in their code. 

Best,

Kirk

Sent from my iPhone

To unsubscribe from this group and stop receiving emails from it, send an email to molpro-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/molpro-user/14aaecea-e1ff-48a2-847b-668f90c011dc%40googlegroups.com.

[Tibor Győri]

As far as I can tell this is no longer the case, MRCC seems to have a keyword for changing it ("ccmaxit").

At least according to their manual: https://www.mrcc.hu/MRCC/manual/html/multiple/S12.xhtml

[Peterson, Kirk]

Thanks Tibor, I’ll see if I can get this working later this week when I’m back in the office.

best,

-Kirk

[Nike Dattani]

Dear Tibor,

At first I thought the "Maxit" keyword was just not working for you via the molpro interface... so I suggested you just get the integrals from molpro (the fort.55 file) and then run MRCC standalone. However now I notice that you were trying to go past 50 iterations. For that you need to change the "maxit" value in MRCCCOMMON and recompile: https://www.mrcc.hu/index.php/forum/ageneral-questions/99-keywords-ignored#455

Most people will tell you that if you need more than 50 iterations, there's something deeper that is going wrong in your calculation (bad initial guess, wrong symmetry, etc.) but I agree with you that the user should be allowed to make this number whatever they want. Furthermore, I do have experience in some calculations of excited states where I needed about 55 iterations to converge the CI calculations, so I do sympathize with you that you may actually need more than 50 for your particular application.

With best wishes!

Nike

To view this discussion on the web, visit https://groups.google.com/d/msgid/molpro-user/794989DB-BCD5-4108-A322-66A65DDE861C%40wsu.edu.