tolerance in CPHF module


  • Guest -
Greetings,

I was wondering if there was a way to change the tolerance and convergence criteria in the CPHF module. Specifically, I want to drop "tolerance = 0.10D-03" and "Convergence 1.0D-04" by a few orders of magnitude. Thank you.

Forum Vet
You can set the tolerance by "set cphf:thresh 1.0d-3". The convergence is by default 0.1*tolerance, but can be set by "set cphf:acc 1.0d-4".

Bert


Quote: Sep 16th 12:54 pm
Greetings,

I was wondering if there was a way to change the tolerance and convergence criteria in the CPHF module. Specifically, I want to drop "tolerance = 0.10D-03" and "Convergence 1.0D-04" by a few orders of magnitude. Thank you.

  • Guest -
Bert, I tried your suggestion. I'm still a bit stuck. I'm doing a calculation within the property module, and it calls the dynamic CPHF module followed by the CPHF module. The "set cphf: thresh" commands sets the correct tolerance for the dynamic CPHF module, but it defaults back to 1d-4 for the second call. I apologize if I'm missing something obvious.

Thank you.

Quote:Bert Sep 16th 4:23 pm
You can set the tolerance by "set cphf:thresh 1.0d-3". The convergence is by default 0.1*tolerance, but can be set by "set cphf:acc 1.0d-4".

Bert


Quote: Sep 16th 12:54 pm
Greetings,

I was wondering if there was a way to change the tolerance and convergence criteria in the CPHF module. Specifically, I want to drop "tolerance = 0.10D-03" and "Convergence 1.0D-04" by a few orders of magnitude. Thank you.

Forum Vet
You may also be able to set convergence in the property block "convergence 1.0d-4"

Can you send me the input and output at bert.dejong@pnnl.gov. It's unclear to me why it would reset.

Bert



Quote: Sep 16th 6:17 pm
Bert, I tried your suggestion. I'm still a bit stuck. I'm doing a calculation within the property module, and it calls the dynamic CPHF module followed by the CPHF module. The "set cphf: thresh" commands sets the correct tolerance for the dynamic CPHF module, but it defaults back to 1d-4 for the second call. I apologize if I'm missing something obvious.

Thank you.

Quote:Bert Sep 16th 4:23 pm
You can set the tolerance by "set cphf:thresh 1.0d-3". The convergence is by default 0.1*tolerance, but can be set by "set cphf:acc 1.0d-4".

Bert


Quote: Sep 16th 12:54 pm
Greetings,

I was wondering if there was a way to change the tolerance and convergence criteria in the CPHF module. Specifically, I want to drop "tolerance = 0.10D-03" and "Convergence 1.0D-04" by a few orders of magnitude. Thank you.

Forum Regular
Can you send us your input and output ?

Thanks.
Niri

niri.govind@pnnl.gov

Forum Vet
The “cphf:thresh” is a NWChem 6.0 quirk, fixed in the developed release as far as I can see, and the “cphf:acc” is a fat finger from you to “cphf:accu”. The dynamic cphf uses the “convergence” keyword from the property block whereas the older cphf uses “cphf:precond_tol” and “cphf:acc”.

Simply add:

  set cphf:precond_tol 1.0d-7
set cphf:thresh 1.0d-7
set cphf:acc 1.0d-7

That will give you what you need.

Bert





Quote:Bert Sep 16th 7:46 pm
You may also be able to set convergence in the property block "convergence 1.0d-4"

Can you send me the input and output at bert.dejong@pnnl.gov. It's unclear to me why it would reset.

Bert



Quote: Sep 16th 6:17 pm
Bert, I tried your suggestion. I'm still a bit stuck. I'm doing a calculation within the property module, and it calls the dynamic CPHF module followed by the CPHF module. The "set cphf: thresh" commands sets the correct tolerance for the dynamic CPHF module, but it defaults back to 1d-4 for the second call. I apologize if I'm missing something obvious.

Thank you.

Quote:Bert Sep 16th 4:23 pm
You can set the tolerance by "set cphf:thresh 1.0d-3". The convergence is by default 0.1*tolerance, but can be set by "set cphf:acc 1.0d-4".

Bert


Quote: Sep 16th 12:54 pm
Greetings,

I was wondering if there was a way to change the tolerance and convergence criteria in the CPHF module. Specifically, I want to drop "tolerance = 0.10D-03" and "Convergence 1.0D-04" by a few orders of magnitude. Thank you.


Forum >> NWChem's corner >> General Topics