SCS-CCSD is bogus on restart


Gets Around
It seems that SCS-CCSD is not written to RTDB correctly or at all, which means that the printout on restart is bogus.

This post is for tracking this bug, because I'm not aware of an other way to do it, other than https://github.com/jeffhammond/nwchem/issues/5, which of course is only useful to me.

Start

-------------------------------------------------------------------------
 iter     correlation     delta       rms       T2     Non-T2      Main
             energy      energy      error      ampl     ampl      Block
                                                time     time      time
-------------------------------------------------------------------------
 g_st2 size:     325 MB
 mem. avail     1891 MB
Memory based method: ST2 is allocated
 ST2 array is replicated      0.32s
   1     -1.3848491782 -1.385D+00  1.603D-07  2920.35     0.65  2796.27
 g_st2 size:     325 MB
 mem. avail     1891 MB
Memory based method: ST2 is allocated
 ST2 array is replicated      0.16s
   2     -1.3848491790 -7.499D-10  1.427D-07  1655.97     0.63  1531.99
                  *************converged*************
-------------------------------------------------------------------------

 -----------
 CCSD Energy
 -----------
 Reference energy:            -380.339222153928404
 CCSD corr. energy:             -1.384849178954665
 Total CCSD energy:           -381.724071332883057


 --------------------------------
 Spin Component Scaled (SCS) CCSD
 --------------------------------
 Same spin contribution:                 -0.301173310594643
 Same spin scaling factor:                1.130000000000000
 Opposite spin contribution:             -1.083675868360022
 Opposite spin scaling fact.:             1.270000000000000
 SCS-CCSD correlation energy:            -1.716594193789175
 Total SCS-CCSD energy:                -382.055816347717609


Restart

   ***** ccsd parameters *****
   iprt   =     0
   convi  =  0.100E-05
   maxit  =    20
   mxvec  =     5
 memory             274965444
  Restarting from old vector in 
 /home/admin/NWCHEM/scratch/w5_rccsd-t_aug-cc-pvtz_energy.t2
 CCSD is already converged
 -----------
 CCSD Energy
 -----------
 Reference energy:            -380.339222153928858
 CCSD corr. energy:             -1.384849176076638
 Total CCSD energy:           -381.724071330005472


 --------------------------------
 Spin Component Scaled (SCS) CCSD
 --------------------------------
 Same spin contribution:                 -1.384849176076638
 Same spin scaling factor:                1.130000000000000
 Opposite spin contribution:              0.000000000000000
 Opposite spin scaling fact.:             1.270000000000000
 SCS-CCSD correlation energy:            -1.564879568966601
 Total SCS-CCSD energy:                -381.904101722895462

Forum Vet
Thanks for reporting the bug.
I have just checked in a fix in the SVN repository (but I have not fully tested it yet)

Gets Around
Thanks! That was fast. The machine where I discovered this does not have internet access so I'm running NWChem 6.6 rather than SVN. I will switch to a new machine and attempt to verify the fix later in the week.


Forum >> NWChem's corner >> General Topics