[molpro-user] PNO-LMP2 optimizations fail with basis set errors

[Hans-Joachim Werner]

Dear Matthias,
you need to give the DF-basis sets on the command lines:

df-hf,basis=jkfit
df-lmp2,basis=mp2fit

Otherwise it will not use what you specified in the basis block.

The pno-lmp2 program has no analytical gradients and your calculation would compute the gradient by finite energy differences. 
However, the pno-lmp2 does not fix the domains, and this can lead to completely wrong gradients.
In fact, the geometry optimization with your input does not converge (this problem will be fixed in the next version, molpro2018.1). 
For the time being, I recommend to use the old pao-lmp2 program, which does have analytical gradients:

df-lmp2,basis=mp2fit

And finally: unless you have a recently patched version, you should not use „orient,mass“ in geometry optimizations. This is a tough case to optimize. Use

optg,method=bmat 

and it converges in 40 iterations (final energy -604.75843996). bmat is usually the best optimization method, but if there are redundant
internal coordinates it can get into an infinite loop. We are trying to fix this problem.

Best wishes
Joachim


---
Prof. Dr. Hans-Joachim Werner
Institut für Theoretische Chemie
Universität Stuttgart
Pfaffenwaldring 55
70569 Stuttgart, Germany
e-mail: werner at theochem.uni-stuttgart.de




> Am 24.10.2017 um 20:32 schrieb Matthias Heger <heger at ualberta.ca>:
> 
> Hello,
> 
> I'm trying to use PNO-LMP2 for geometry optimizations, which consistently fails with an error message:
> 
> ? Error
> ? Cannot find basis set data
> ? The problem occurs in index_basis_set
> 
> In the log file, the last lines are:
> 
> Looking for basis set CC-PVDZ-JKFIT
> The list of defined basis sets is ORBITAL
> JKFIT                   MP2FIT                  CC-PVDZ-MP2FIT
> MINAO-AUTO-PP
> 
> Even if I explicitly define the basis sets in a basis block in the input, this keeps happening. I'm attaching my input and output files below. What's interesting is that the optimization actually seems to make one successful step - the error occurs at the start of the second step. So in principle it seems to work, but apparently the program gets confused somewhere along the way.
> 
> I hope there's an easy way around these problems.
> 
> Best,
> Matthias
> 
> 
> 
> -- 
> Dr. Matthias Heger
> Department of Chemistry, University of Alberta
> 11227 Saskatchewan Drive NW
> Edmonton, AB T6G 2G2, Canada
> 
> Office: CCIS 4-151
> Lab: CCIS 4-108
> Email: heger at ualberta.ca
> <L-Arg.inp><L-Arg.log><L-Arg.out><L-Arg_0.xyz>_______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user