[molpro-user] E2exdisp in CCSD-SAPT converges to the wrong value in DC+BS

[Leonid Shirkov]

Hello Molpro Community,

I was running some test jobs to calculate E2disp and E2exdisp energies
for He2 with
CCSD-SAPT at R=3.0 bohrs  (interatomic distance).  In order to speed
up the convergence
I employed midbond functions, and that works well for E2disp. However,
E2exdisp converges
to a smaller wrong value if DC+BS is used. Without the midbond
functions, i.e. with DCBS,
it converges to a value that  agrees well with  the previously
reported CCSD-SAPT calculations,
see 10.1063/1.2889006 and  10.1021/ct900232j.
These values are Edisp=2.937 and E2exdisp=0.479 (mhartree units).

Here is the outline of the results, '+' denotes the midbond functions,

without density-fitting:

             Edisp  Eexdisp
avtz       -2.599   0.391
avdz+    -2.884   0.165
avtz+     -2.907   0.235
avqz+    -2.940   0.261

with density-fitting (avXz/mp2fit used as the auxilary basis set):

           Edisp  Eexdisp
avtz     -2.600   0.383
avqz    -2.853   0.454
avdz+  -2.894   0.087
avtz+   -2.910   0.103
avqz+  -2.937   0.115

I am attaching the inputs that I used for DC+BS calculations, both df
and non-df versions.
Is there anything wrong with my inputs? Perhaps some input parameters
or thresholds ought
to be tweaked? Increasing the integration grid by the 'dispgg' parameter to
20 does not help. Just in case it matters, I am using the 2012.1 version.

I would appreciate any help and insight.

Leonid
-------------- next part --------------
A non-text attachment was scrubbed...
Name: he2_df-disp_dcpbs_avtz.inp
Type: application/octet-stream
Size: 5855 bytes
Desc: not available
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20160922/c6a1d586/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: he2_disp_dcpbs_avtz.inp
Type: application/octet-stream
Size: 1780 bytes
Desc: not available
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20160922/c6a1d586/attachment-0001.obj>