LCCSD jobs failing

More
5 months 1 week ago #1083 by sodohsunf
LCCSD jobs failing was created by sodohsunf
Hello,
 I am trying to run a sample LCCSD job after seeing this paper: pubs.acs.org/doi/pdf/10.1021/acs.jctc.9b00511 

My file looks thus:
unit=angs
basis=atomtype
H:cc-pVDZ
C:cc-pVDZ
dfbasis_cor=none
calc=LCCSD
lcorthr=tight
mem=110GB
ccprog=mrcc
ccmaxit=500
charge=0
verbosity=3
scftype=rhf
mult=1
ccmaxit=1000
geom=xyz
5

C       -0.000000001      0.000000000      0.000000000
H        0.365397423     -0.516749994      0.895037238
H        0.365397444     -0.516749979     -0.895037238
H        0.365397418      1.033499984      0.000000013
H       -1.096192272     -0.000000011     -0.000000013


I am getting errors:
 ************************ 2021-04-07 05:13:06 *************************
 Executing drpa...

 Three-index integral transformation

 Allocation of 110.0 Gbytes of memory...
 Corrupted file VARS!
 Variable dfipra_cor is missing!
 Please run integ first!

 Fatal error in drpa.
 Program will stop.


Could you give me some help with this?
Thanks.

 

Please Log in or Create an account to join the conversation.

More
5 months 1 week ago #1084 by kallay
Replied by kallay on topic LCCSD jobs failing
Local CC calculations are not possible without density fitting if localcc=2018 (default). If you want to perform a local CC calculation with four-center Coulomb integrals, you should set localcc=2013.
Nevertheless, unless you have overriding reasons not to do so, you should use density fitting and the default local CC algorithm. That is, you should delete the dfbasis_cor=none and ccprog=mrcc lines from your input.

Best regards,
Mihaly Kallay
The following user(s) said Thank You: sodohsunf

Please Log in or Create an account to join the conversation.

More
5 months 1 week ago #1085 by nagypeter
Replied by nagypeter on topic LCCSD jobs failing
Hello,

thank you for the interest.

Local correlation methods are only implemented with density fitting, so dfbasis_cor=none is not compatible with LNO-CC*.
The program will try to select a proper density fitting basis, so in most of the
cases you do not need to set dfbasis_cor in the input.

You deviate quite a lot from usual use cases so I cannot guess what is your
targeted application. For CCSD and CCSD(T) you should use the default and much more efficient ccprog=ccsd, although ccprog=mrcc works in your case as well.
But ccprog=mrcc is primarily for higher-order (local) CC method, e.g., (LNO-)CCSDT, (LNO-)CCSDT(Q).

I hope this helps, let me know.
Best wishes,
Peter
The following user(s) said Thank You: sodohsunf

Please Log in or Create an account to join the conversation.

More
5 months 1 week ago #1086 by sodohsunf
Replied by sodohsunf on topic LCCSD jobs failing
I removed the ccprog and dfbasis_cor lines.

Ultimately, I would like to use LCSSDT to generate benchmarks for heterogeneous catalysis mechanisms in zeolites (truncated cluster model). The hope is to use that as a reference for DFT and CASSCF+DFT methods.

However, I likely need to read the LNO papers again. Right now though, this is what I am getting:

MO Natoms Completeness Atoms
1 5 1.000000 1 2 3 4 5
2 5 1.000000 1 2 3 4 5
3 5 1.000000 1 2 3 4 5
4 5 1.000000 1 2 3 4 5

Executing the Boughton-Pulay algorithm...
with occ MO completeness: 0.9850000000

MO Natoms Completeness Atoms
1 2 0.998720 1 5
2 2 0.998720 1 2
3 2 0.998720 1 4
4 2 0.998720 1 3
CPU time [min]: 0.037 Wall time [min]: 0.006

Dimensions of the occupied LMO domains: Minimum / Average / Maximum

Number of atoms in domain: 5 5.0 5
Number of atomic orbitals: 34 34.0 34

Total number of pairs: 6
Number of distant pairs: 0 (0.0%)
Sum of multipole based pair energies: 0.0000000000

CPU time [min]: 0.037 Wall time [min]: 0.006
Number of strong pairs: 6 (100.0%)
Corrupted file VARS!
Variable savedtimes is missing!
Please run integ first!

Please Log in or Create an account to join the conversation.

More
5 months 1 week ago #1087 by nagypeter
Replied by nagypeter on topic LCCSD jobs failing
Removing ccprog and dfbasis_cor from your first input file works for me with the latest release.
Please, upload the complete input/output file.
Which version of MRCC do you use?

Please Log in or Create an account to join the conversation.

More
5 months 1 week ago #1088 by sodohsunf
Replied by sodohsunf on topic LCCSD jobs failing
Sorry about this. I was using the 2017 version. Ooof!
I downloaded this (mrcc.2020-02-22.binary.tar) just now.

Now it works.

Thanks so much for your kind and patient help.


Final local CC results:

Minimum number of correlated electrons: 8 (100.0%)
Maximum number of correlated electrons: 8 (100.0%)
Average number of correlated electrons: 8.0 (100.0%)
Minimum number of correlated orbitals: 33 (100.0%)
Maximum number of correlated orbitals: 33 (100.0%)
Average number of correlated orbitals: 33.0 (100.0%)

CPU time for integral calculations [min]: 0.090
Wall time for integral calculations [min]: 0.037

CPU time for SCF calculation [min]: 0.146
Wall time for SCF calculation [min]: 0.013

CPU time for orbital localization [min]: 0.001
Wall time for orbital localization [min]: 0.001

CPU time for integral transformations [min]: 0.212
Wall time for integral transformations [min]: 0.015

Reference energy [au]: -40.198638043859

LMP2 correlation energy [au]: -0.161450652362
Total LMP2 energy [au]: -40.360088696220

Total MP2 correction for dropped NAFs [au]: -0.000056495184
Total MP2 correction [au]: -0.000056495184

CPU time for CCSD calculations [min]: 0.173
Total CPU time for CCSD [min]: 0.622
Wall time for CCSD calculations [min]: 0.011
Total wall time for CCSD [min]: 0.078

CCSD correlation energy [au]: -0.185181730388
Total CCSD energy [au]: -40.383819774247
CCSD correlation energy + MP2 corrections [au]: -0.185238225572
Total LNO-CCSD energy with MP2 corrections [au]: -40.383876269431

======================================================================
======================================================================

************************ 2021-04-07 09:47:41 *************************
Normal termination of mrcc.

Please Log in or Create an account to join the conversation.

Time to create page: 0.020 seconds
Powered by Kunena Forum