Ceilidh ... Poor SCF convergence with re-read orbitals

Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

We are NATO-free zone

Poor SCF convergence with re-read orbitals

Slawomir Janicki
slawomir.janicki@comcast.net

I run into SCF convergence trouble with re-read orbitals. Is there a way to increase the accuracy of punched orbitals beyond WIDE=.T. ?

In the first pass I construct the guess orbitals by:

- $GUESS GUESS=MOREAD NORB=904 KDIAG=0 EXTRA=.T. $END from a previous calculation with QZVPD on all atoms except one H with QZVP;
- adding the p diffuse functions from QZVPD to that unique H with $EXTRAF NEXTRA(1)=0,0,0,0,0,0,0,0,0,0,0,0,0,3 $END
- declaring the basis set in $DATA

This SCF calculation converges fine in 12 iterations.

NONZERO BLOCKS
ITER EX DEM TOTAL ENERGY E CHANGE DENSITY CHANGE DIIS ERROR VIR. SHIFT DAMPING INTEGRALS SKIPPED
* * * INITIATING DIIS PROCEDURE * * *
1 0 0 -434.359184038 -434.359184038 26.000472601 0.000765616 0.000000000 1.000000000 -1 231093005
2 1 0 -434.359184889 -0.000000851 21.033149847 0.000696271 0.000000000 1.000000000 -1 230685217
3 2 0 -434.359185951 -0.000001062 2.264576385 0.000341393 0.000000000 1.000000000 -1 230830209
4 3 0 -434.359186134 -0.000000183 1.741446876 0.000100257 0.000000000 1.000000000 -1 230858906
5 4 0 -434.359186149 -0.000000015 0.110725963 0.000003935 0.000000000 1.000000000 -1 230829709
6 5 0 -434.359186149 0.000000000 0.000703852 0.000000093 0.000000000 1.000000000 -1 230828030
7 6 0 -434.359186149 0.000000000 0.000949385 0.000000032 0.000000000 1.000000000 -1 230828021
8 7 0 -434.359186149 0.000000000 0.000186930 0.000000010 0.000000000 1.000000000 -1 230828001
9 8 0 -434.359186149 0.000000000 0.000155360 0.000000006 0.000000000 1.000000000 -1 230828007
10 9 0 -434.359186149 0.000000000 0.000225597 0.000000011 0.000000000 1.000000000 -1 230828011
11 10 0 -434.359186149 0.000000000 0.000018039 0.000000012 0.000000000 1.000000000 -1 230828006
12 11 0 -434.359186149 0.000000000 0.000035272 0.000000021 0.000000000 1.000000000 -1 230828006

Next, I use the orbitals from the first dat file and construct the guess orbitals for the second input file by:

- $GUESS GUESS=MOREAD NORB=907 KDIAG=0 EXTRA=.F. $END
- declaring the basis set in $DATA (the $DATA is the same as in the first input file)

The second SCF fails to converge and the energy and density oscillate wildly.

NONZERO BLOCKS
ITER EX DEM TOTAL ENERGY E CHANGE DENSITY CHANGE DIIS ERROR VIR. SHIFT DAMPING INTEGRALS SKIPPED
* * * INITIATING DIIS PROCEDURE * * *
1 0 0 -434.359184142 -434.359184142************** 0.001427558 0.000000000 1.000000000 -1 230828006
2 1 0 -220.595295982 213.763888161 2.036093324 112.110529823 0.191843150 1.000000000 -1 219960281
3 2 0 -561.039293414 -340.443997432 0.609648347 34.077380819 0.091843148 1.000000000 -1 219960281
4 3 0 -586.729178088 -25.689884675 1.225957379 68.635216291 0.091843148 1.000000000 -1 219960281
5 4 0 -506.720698955 80.008479134 0.264003476 50.804093873 0.091843148 1.000000000 -1 219960281
6 5 0 -436.096265988 70.624432967 0.493343875 33.901854332 0.091843148 1.000000000 -1 219960281
etc.

The SCF convergence options are the same in both cases:

$SCF DIRSCF=.T. $END
$CONTRL MAXIT=100 $END
$SCF FDIFF=.F. $END
$SCF NCONV=5 $END
$SCF EXTRAP=.F. $END
$SCF SODIIS=0 DIONCE=.F. $END
$SCF SOSCF=.F. $END
$SCF DIIS=.T. ETHRSH=2500 MAXDII=10 $END
$SCF DEM=.F. $END
$SCF DAMP=.T. DMPCUT=0.0 $END
$SCF SHIFT=.T. $END
$SCF RSTRCT=.F. $END

Any idea how to fix this?

Slawomir

Fri Dec 6 '13 4:08pm

This message read 973 times