# CASSCF occupation numbers

Dear guys,

I am trying to set up a case for a CASSCF calculation with Psi4, and so far it looks intriguing. I am using a test system and the input file like this one:

``````molecule{
TI    -0.0150616760   0.0005223151  -0.0088460157
O     1.6619541149  -0.1748776542  -0.6477026083
O    -0.5863580705   1.6951366408  -0.2155285424
O     0.0386116105  -0.4398317741   1.7395629568
O    -1.1457134213  -1.0938528281  -0.8958503769
H     2.2389860821   0.4859383075  -1.0298068211
H    -0.7144606180   2.3742027551   0.4467927151
H     0.8116342000  -0.6233596016   2.2733914551
H    -1.1195710325  -2.0485566872  -0.9597483836
}

set basis cc-pVDZ
set frozen_docc [ 27 ]
set active [ 4 ]
set num_roots 4
set mcscf_maxiter 60

en, wfn = energy('casscf', return_wfn=True)

molden(wfn, 'wfn.molden')
``````

This is a state-specific run, what is confirmed by: STATE AVERAGE = 0(1.00) in the output. In addition I find the following lines:

==> MCSCF root 0 information <==

``````MCSCF Root 0 energy =  -1150.426419689984868
``````

Active Space Natural occupation numbers:

``````     A   1.999216         A   1.994734         A   0.005262
A   0.000788
``````

and:

==> MCSCF root 1 information <==

``````MCSCF Root 1 energy =  -1149.646634938335183
``````

Active Space Natural occupation numbers:

``````     A   1.999601         A   1.016189         A   0.983873
A   0.000337
``````

If I add AVG_STATES = [0,1] to the input file, the convergence is slightly more difficult and the output says: 0(0.50) 1(0.50). So far so good. But in that case, shouldnt the occulation numbers be close to 2.0, 1.5, 0.5, 0.0 ? Instead I get the same (almost) values as for the state-specific run…