I would save the first orbitals in a check file and then call them out for the next geometry.

Dear Prof. Bernstein,

this is what I am doing, but when I read the orbitals on the new geometry there is something strange going on, since the energy printed is completely wrong (hundreds of Hartree!).

I was wondering if I have to specify somethin special in the input (while generating the CAS on the first geometry and when I read the orbitals on the second geometry)

Best

Riccardo

Are you using symmetry? I  suppose you are doing CAS single point calculations, so in my experience I used the proposed strategy of Prof. Bernstein and it usually works well. Maybe removing symmetry could help you.

Dear Ana,

I am not using symmetry, as the molecule does not have any, and yes, it's all about single point calculations that have to be performed on a set of different geometries.

A typical input that i am using contains something like this

%Chk=ret+lyr_cas1213_ano-l-vdzp.2.chk

# CAS(12,13,Nroot=2,StateAverage, StateGuess=1)  Gen

# NoSymmetry Units=Ang

# P GFInput IOP(6/7=3,9/40=2)

# Pop=NPA Guess=Read,Alter

# SCF(MaxCycle=300, Conver=6)

Now, the code reads the guess orbitals stored in the .chk file (first line in this input), and overwites on the same file the converged orbitals one the CASSCF has finished.

If I use that orbitals on a different geometry, the energy is completely off.

I looked in the manual for several option for "Guess=" but didn't find a good solution.

Any Idea? Should I add/remove something to/from my input?

Best

Riccardo

I would try removing alter since I think orbital transposition that was useful for the first job (that for example started from a HF calculation for initial guess for the CAS) maybe is not needed if you start from a set of optimized CAS starting orbitals.

Dear Ana,

of course you are right, but my problem is not related to the use of "Alter", but rather to some kind of incompatibility between the binary generated on a certain geometry and then read on another one.

Riccardo