After I sent this earlier email, I happened to be in the process of running some
calculations on XeF4, also using an ECP (although small-core). If I included tight f
functions in the basis set (for a core-valence calculation), the RHF program completely
failed even when given the correct occupation. I was able to successfully do the calculation by
first running a RHF without the extra tight functions and then following this with another
RHF with the tight functions and the RHF then easily converged.
Perhaps Nick or someone could look into this?
-Kirk
On Monday, August 18, 2003, at 09:30 AM, Kirk Peterson wrote:
Dear Jacek,
in 2002.6 the default initial guess in the RHF program is often quite poor when ECP's are
used. If you include explicit "occ" and "closed" commands it will of course work
fine. I believe the molpro folks are aware of this problem and are working on a
fix.
-Kirk
On Monday, August 18, 2003, at 06:51 AM, Jacek Klos wrote:
Dear Molpro Users,
I hit such a problem when using effective core potential for Xe atom in interaction with NO molecule. In molpro 2002.3 and molpro 2000.1 input I use works and gives RCCSD(T) energy of the Xe-NO molecule. But when the same input is used in molpro 2002.6 I get different solution of RHF program and RCCSD is exiting with UNREASONABLE NORM error. The CCSD exits because of something wrong is going in RHF I presume. Here is the input:
***,Xe-NO memory,15,m gthresh,energy=1.d-9; gprint,basis; geometry={noorient; q1; q2,q1,1.0 N,q1,rnx,q2,90.0; O,q1,rox,q2,90.0,N,180.0; Xe,q1,r,q2,beta,N,0.0 Be,q1,rbe,q2,beta,N,0.0; } rox=0.5371799322 ang rnx=0.6135900677 ang theta=0.0 beta=90.0-theta basis={ ecp,Xe,ECP46MWB; spdf,Xe,ECP46MWB; spdf,n,avtz;c; spdf,o,avtz;c; sp,be,0.9,0.3,0.1; df,be,0.6,0.2; g,be,0.3;} r=20.0 ang rbe=r*0.5 ang text,calculation for complex BIS ; dummy,be rhf;wf,23,2,1;maxit,90 orbprint,0 rccsd(t);maxit,90; ---
In Molpro 2002.3 RHF energy is:-144.56731345 but in 2002.6 version RHF is:-143.91398149
First iterations in RHF in 2002.3 and 2002.6 versions:
RHF iterations from Molpro 2002.6:
ITERATION DDIFF GRAD ENERGY 2-EL.EN. DIPOLE MOMENTS DIIS 1 .000D+00 .000D+00 -70.56488111 444.580663 .000000 301.451947 .000000 0 2 .000D+00 .221D+00 -73.15131316 170.093612 .000000 ********** .000000 0 3 .628D+00 .803D-01 -87.90178385 402.689735 .000000 301.531151 .000000 1 4 .364D+00 .179D+00 -73.02868137 169.716549 .000000 ********** .000000 1 5 .365D+00 .934D-01 -102.53801608 355.594529 .000000 301.571132 .000000 2 6 .245D+00 .125D+00 -111.74460806 317.557998 .000000 302.670775 .000000 3 7 .901D-01 .884D-01 -134.11525158 234.103093 .000000 189.863457 .000000 4 8 .928D-01 .400D-01 -139.43368935 146.100659 .000000 ********** .000000 5 9 .692D-01 .222D-01 -142.37680048 200.240607 .000000 74.352007 .000000 6 10 .516D-01 .199D-01 -142.61964648 196.806101 .000000 74.632483 .000000 7
RHF iterations from Molpro 2002.3:
ITERATION DDIFF GRAD ENERGY 2-EL.EN. DIPOLE MOMENTS DIIS 1 .000D+00 .000D+00 -70.56488111 444.580663 .000000 301.451947 .000000 0 2 .000D+00 .221D+00 -75.05389406 167.962428 .000000 ********** .000000 0 3 .627D+00 .849D-01 -87.90178428 402.689735 .000000 301.531150 .000000 1 4 .364D+00 .179D+00 -77.68623161 161.516536 .000000 ********** .000000 1 5 .581D+00 .924D-01 -102.27524477 356.553891 .000000 301.551912 .000000 2 6 .515D+00 .126D+00 -112.19090149 317.041387 .000000 303.057258 .000000 3 7 .110D+00 .866D-01 -139.77210755 219.054895 .000000 112.749107 .000000 4 8 .104D+00 .322D-01 -144.47518819 169.737969 .000000 .217617 .000000 5 9 .452D-01 .816D-02 -144.52805483 170.443554 .000000 -.231275 .000000 6 10 .803D-02 .508D-02 -144.54885496 171.414992 .000000 .079383 .000000 7
First iteration is the same, third very similar, but then it starts to
deviate and final solution is different.
I checked in case of Kr, which interest me more, there is no problem
when I choose aug-cc-pv*z basis, but when I choose ECP potential for Kr
the Molpro 2002.6 exits with similar error like in case of Xe atom.
I checked if it can be the reason of bond-functions. No, this behavior
repeats without bond functions as well. I checked if it happens only for
Xe atom. No, Xe atom alone finishes in the same way in both revisions of
Molpro using ECP potential.
Why the same input for Xe-NO dimer behaves differently in these 2 revisions of Molpro 2002?
Platform used: IBM AIX 5.1
Anybody with similar problem before?
Best regards -- Jacek A. Klos --------------------------------- Quantum Chemistry Laboratory | Department of Chemistry | University of Warsaw | Pasteura 1 Street | 02-093 Warsaw | Poland | (+4822)8220211 ext. 213 | e-mail: [EMAIL PROTECTED]| ---------------------------------
