SCS MP2 Geometries with df-mp2

Hi everyone

I’d like to get geometries optimized to the SCS method using df-mp2. I used the key-words

set mp2_os_scale 1.2
set mp2_ss_scale 0.33333333333333333

and actually got different structures. Is this the way to do it?

regards, jaf

The analytic df-mp2 gradient decomposed into same-spin and opposite-spin components is what would be needed to have analytic gradients for scs-mp2 (and friends). Unfortunately, the df-mp2 gradient as implemented doesn’t have these components available, so an optimize('scs-mp2') should fall back on finite difference. I have doubts that what you (very sensibly) specified would trigger the proper procedure. Will investigate.

Thank you for your quick answer. Indeed, the final structures with or without the SCS-keywords is the same. So obviously, the keywords don’t do anything. However, if there is a chance that the possibility to do SCS-geometry optimization is included into the program package, I would be very grateful. Regards, jaf