RHF energy inconsistency using built-in SBKJC and user-defined SBKJC for Pr

1234zou commented 3 years ago

[ ] Summary of the issue and expected results. For RHF/SBKJC calculation of a single cation Pr5+, the RHF energy using built-in SBKJC ECP/PP is -41.511677 a.u. But if I copy SBKJC ECP/PP data from GAMESS output, and paste it into the input file (as well as change PP=SBKJC to PP=READ), the RHF energy is -41.511723 a.u. The energy difference is not negligible.
[ ] Description of the run environment:
- operating system Centos 7.4
- compiler Intel 2019 Update 5
- math library Intel MKL
- communication library sockets
- GAMESS version 2020-Sep-30(R2)

[ ] An input file (if applicable) to reproduce the issue.
Using built-in SBKJC ECP/PP:

$CONTRL SCFTYP=RHF RUNTYP=ENERGY ICHARG=5 MULT=1 NOSYM=1 ICUT=11 PP=SBKJC
MAXIT=200 ISPHER=1 $END
$SYSTEM MWORDS=1000 $END
$BASIS GBASIS=SBKJC $END
$SCF DIIS=.T. SOSCF=.F. $END
$DATA
XIA XIE
C1   1
Pr   59.        0.0   0.0    0.0
$END

Using user-defined SBKJC ECP/PP:

 $CONTRL SCFTYP=RHF RUNTYP=ENERGY ICHARG=5 MULT=1 NOSYM=1 ICUT=11 PP=READ
  MAXIT=200 ISPHER=1 $END
 $SYSTEM MWORDS=1000 $END
 $BASIS GBASIS=SBKJC $END
 $SCF DIIS=.T. SOSCF=.F. $END
 $DATA
XIA XIE
C1   1
Pr   59.        0.0   0.0    0.0
 $END
 $ECP
1-ECP GEN   46   3
2     ----- f-ul potential -----
   -1.54435200E+01    1   9.82973000E+00
   -5.89611000E+00    1   1.98070000E+00
3     ----- s-f potential -----
   -2.23643980E+02    2   2.12956000E+00
    2.78134520E+02    2   2.22747000E+00
    1.26210700E+01    0   7.28372000E+00
3     ----- p-f potential -----
    1.25256400E+01    0   7.80928000E+00
    1.21952780E+02    2   1.92978000E+00
   -7.94047500E+01    2   1.76479000E+00
2     ----- d-f potential -----
    2.61926700E+01    2   1.79797000E+00
    9.87391000E+00    0   6.54805000E+00
 $END

Thank you!

vamironov commented 3 years ago

The precision of the ECP parameters in the output file is currently not enough to be used in input. See source/ecplib.src line 1886

9170 FORMAT(I5,F15.5,I5,F15.5) https://github.com/gms-bbg/gamess/blob/f4fed2e3c874b674b74fe2d72c59104ec26bf59e/source/ecplib.src#L1886

The ECP parameters which correspond to GAMESS internal basis set library are as follows:

 $ECP
1-ECP GEN   46   3
2     ----- f-ul potential -----
   -1.54435219D+01    1   9.82972856D+00
   -5.89611285D+00    1   1.98070077D+00
3     ----- s-f potential -----
   -2.23643982D+02    2   2.12955579D+00
    2.78134515D+02    2   2.22746546D+00
    1.26210718D+01    0   7.28371963D+00
3     ----- p-f potential -----
    1.25256390D+01    0   7.80928039D+00
    1.21952782D+02    2   1.92977861D+00
   -7.94047512D+01    2   1.76478925D+00
2     ----- d-f potential -----
    2.61926652D+01    2   1.79797432D+00
    9.87391206D+00    0   6.54805153D+00
 $END

1234zou commented 3 years ago

Thank you, @vamironov ! Problem solved.

gms-bbg / gamess-issues

RHF energy inconsistency using built-in SBKJC and user-defined SBKJC for Pr #47