[molpro-user] repeated CASSCF calculation - two different results:
Hans-Joachim Werner
werner at theochem.uni-stuttgart.de
Thu May 26 09:19:45 CEST 2016
Dear colleagues,
this is a known problem, and despite quite a lot of effort we were not yet able to find and fix the problem.
It is some numerical instability, which only occurs in the mpp case. At the moment, there is no other
solution than to run it on only 1 core. I am sorry for the trouble.
Best regards
Hans-Joachim Werner
---
Prof. Dr. Hans-Joachim Werner
Institute for Theoretical Chemistry
University of Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart
Tel: +49 711 / 685 64400
Fax: +49 711 / 685 64442
email: werner at theochem.uni-stuttgart.de
> Am 26.05.2016 um 06:12 schrieb Caroline Krauter <ckrauter at princeton.edu>:
>
> Dear experts, Prof. Burger,
>
> I am having exactly the same problem and joined this mailing list just today to report it. I have run several CASSCF calculations, all starting with an identical set of previously converged SCF orbitals, to test it. The following results were obtained with 5 runs:
> -1263.16187108
> -1263.16042269
> -1263.14364086
> -1263.17543796
> -1263.16355897
>
> I was running the jobs in parallel on 4 cores each.
>
> The relevant part of the input file:
> {matrop
> READ,scf,TYPE=ORB,SUBTYPE=CANONICAL,File=orbs
> SAVE,scf,2160.2,ORBITALS
> }
> {casscf
> start,2160.2
> closed,86
> occ,96
> config,csf
> wf,182,1,0
> state,1
> orbprint,100
> maxiter,40
> print,civector
> }
>
> I could also provide a full set of input files.
>
> I would be very thankful for any hints about what might be going wrong.
> Best regards,
> Caroline Krauter
>
>
> On 05/25/2016 04:19 PM, Peter Burger wrote:
>> Dear experts,
>>
>> I have repeated a CASSCF calculation and obtain two different results..
>>
>> Final CASSCF energies:
>>
>> -614.27913353 vs -614.29682976 H
>>
>> The reference states differ in the last digit (they converged with the same number of SCF cycles).
>>
>> !UHF STATE 1.2 Energy -614.174983749842
>> !UHF STATE 1.2 Energy -614.174983749841
>>
>> The codes were run in parallel on 16 cores on the exact same hardware (Molpro version 2015.1)
>>
>> Is this a sign of numerical instability - either code or hardware?
>>
>> I there a cure?
>>
>> Best regards
>>
>> Peter
>>
>> Number of orbital rotations: 5453 ( 104 Core/Active 3811 Core/Virtual 0 Active/Active 1538 Active/Virtual)
>> Total number of variables: 2582077
>>
>>
>> ITER. MIC NCI NEG ENERGY(VAR) ENERGY(PROJ) ENERGY CHANGE GRAD(0) GRAD(ORB) GRAD(CI) STEP TIME
>>
>> 1 110 27 0 -614.20752867 -614.25168006 -0.04415139 0.06581535 0.00265426 0.01172985 0.17D+01 176.90
>> 2 70 47 0 -614.25464720 -614.25941429 -0.00476708 0.05645649 0.00004046 0.00028798 0.52D+00 441.24
>> 3 90 42 0 -614.25993568 -614.26129568 -0.00136000 0.00660288 0.00000178 0.00446927 0.43D+00 678.86
>> 4 70 46 0 -614.26567248 -614.27766280 -0.01199032 0.04259532 0.00039830 0.00059713 0.68D+00 934.54
>> 5 70 44 0 -614.27881809 -614.27912994 -0.00031185 0.01865182 0.00000968 0.00003024 0.96D-01 1182.08
>> 6 60 31 0 -614.27913327 -614.27913353 -0.00000026 0.00029918 0.00000000 0.00000918 0.46D-02 1352.32
>> 7 60 16 0 -614.27913353 -614.27913353 0.00000000 0.00000253 0.00000000 0.00000000 0.00D+00 1444.36
>>
>> ** WVFN **** CONVERGENCE REACHED, FINAL GRADIENT: 0.25D-05
>>
>> Number of orbital rotations: 5453 ( 104 Core/Active 3811 Core/Virtual 0 Active/Active 1538 Active/Virtual)
>> Total number of variables: 2582077
>>
>>
>> ITER. MIC NCI NEG ENERGY(VAR) ENERGY(PROJ) ENERGY CHANGE GRAD(0) GRAD(ORB) GRAD(CI) STEP TIME
>>
>> 1 110 27 0 -614.20752867 -614.25165564 -0.04412697 0.06581535 0.00290952 0.01167685 0.17D+01 175.29
>> 2 70 47 0 -614.25457713 -614.25932641 -0.00474929 0.05596799 0.00004860 0.00033594 0.52D+00 432.76
>> 3 86 42 0 -614.25991802 -614.26002741 -0.00010939 0.00726826 0.00000001 0.00112142 0.13D+00 671.47
>> 4 90 42 0 -614.26005664 -614.26022502 -0.00016838 0.00457481 0.00000001 0.00037366 0.17D+00 906.91
>> 5 90 40 0 -614.26026701 -614.26035755 -0.00009054 0.00412629 0.00000002 0.00033484 0.13D+00 1131.90
>> 6 88 40 0 -614.26037277 -614.26041086 -0.00003809 0.00187738 0.00000002 0.00036759 0.11D+00 1354.14
>> 7 100 40 0 -614.26041739 -614.26045063 -0.00003324 0.00142076 0.00000000 0.00049945 0.12D+00 1578.86
>> 8 90 41 0 -614.26046420 -614.26063796 -0.00017376 0.00211835 0.00000006 0.00104825 0.25D+00 1808.91
>> 9 120 45 0 -614.26121765 -614.26689841 -0.00568076 0.01223538 0.00001136 0.01069074 0.11D+01 2032.57
>> 10 90 48 0 -614.27429321 -614.28858580 -0.01429259 0.04215226 0.00006428 0.00753717 0.97D+00 2281.39
>> 11 90 48 0 -614.29428392 -614.29664149 -0.00235757 0.04543245 0.00000335 0.00118828 0.35D+00 2546.96
>> 12 100 45 0 -614.29674677 -614.29682755 -0.00008079 0.00348054 0.00000006 0.00036145 0.15D+00 2792.74
>> 13 90 44 0 -614.29682976 -614.29683139 -0.00000164 0.00085366 0.00000015 0.00004194 0.21D-01 3030.52
>> 14 90 23 0 -614.29683140 -614.29683140 0.00000000 0.00003072 0.00000000 0.00000000 0.00D+00 3151.56
>>
>> ** WVFN **** CONVERGENCE REACHED, FINAL GRADIENT: 0.31D-04
>>
>> --
>> Univ.-Prof. Dr. Peter Burger
>> Institut fuer Anorganische und Angewandte Chemie
>> Universitaet Hamburg
>> Martin-Luther-King-Platz 6
>> D-20146 Hamburg
>> Tel.:+49 040 42838 3662
>> FAX 6097
>> email:
>> burger at chemie.uni-hamburg.de
>> http://www.chemie.uni-hamburg.de/ac/burger/index.html
>>
>>
>> _______________________________________________
>> Molpro-user mailing list
>>
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
More information about the Molpro-user
mailing list