MP2 converging to strange energies

Hello,

I have been running some MP2 aug-cc-pVQZ, aug-cc-pVTZ and cc-pVTZ single point calculations, some of which are returning strange energies. An example of the type of output is shown below:

                 Total energy      Delta E      RMS

@DF-ROHF iter 0: -582.96967123898082 -5.82970e+02 0.00000e+00
@DF-ROHF iter 1: -138.94956589736944 4.44020e+02 0.00000e+00
@DF-ROHF iter 2: -138.94956589736944 0.00000e+00 0.00000e+00 DIIS

I’ve tried using canonical orbitals, and 3-21G, cc-pVDZ and cc-pVTZ basis set guesses, as well as gwh and sad.

Do you have any suggestions as to how I can get these calculations to converge? I’ve included an example input file below.

Thank you for your help.
David.

molecule complex {
0 2
C 0.299446 0.145069 0.025712
C 0.320471 0.165328 1.574454
C 1.781238 -0.061340 1.972966
C 1.758909 -0.006201 -0.396109
N 2.422185 -0.700513 0.766624
C 2.116510 -2.159187 0.739954
C 3.893116 -0.488533 0.756353
H 1.032411 -2.298019 0.779404
H 2.587221 -2.625504 1.608947
H 2.521753 -2.582747 -0.181859
H 4.304576 -0.954062 -0.143319
H 4.321887 -0.952096 1.649192
H 4.080079 0.588110 0.753293
H 1.925986 -0.723182 2.830469
H 2.301026 0.885885 2.130909
H -0.326524 -0.617315 1.981396
H -0.025973 1.123679 1.971456
H -0.305289 -0.689523 -0.342675
H -0.119181 1.066055 -0.389096
H 1.924692 -0.600093 -1.298226
H 2.246668 0.966568 -0.487488
N 2.375903 2.951944 0.911713
C 1.311220 3.441076 0.859309
N 0.177240 4.098386 0.811739
C -0.949999 3.433162 0.699065
N -2.000901 2.925452 0.594825
C -1.575354 -2.920804 0.925918
H -1.296743 -3.446389 0.017534
H -2.302464 -2.116091 0.879407
H -1.269567 -3.314237 1.890683
}
set globals {
basis aug-cc-pvqz
scf_type df
freeze_core True
guess gwh
basis_guess cc-pvdz
reference rohf
s_orthogonalization canonical
}
set cachelevel 0
set print 2
energy(‘df-mp2’)

I tried out your molecule and agree that convergence is shaky. The default guess is perhaps best for this system– I was able to get ROHF to converge to -568.410685 with cc-pvdz and -568.5732969 with aug-cc-pvtz, while sad guess was going awry. Was that the value expected?

set basis cc-pvdz
set reference rohf

energy('scf')

Then mp2 was -570.33081 with cc-pvdz and -571.16394 with aug-cc-pvtz.

set basis aug-cc-pvtz
set reference rohf

energy('mp2')

Will that work for you for now or are there particular basis sets you need? At minimum, this seems a good case to file away for testing scf convergence.

Thank you for your help.
The default guess worked for most cases and returned energies that were consistent with those of similar calculations.
Some of my aug-cc-pVQZ calculations did not converge using any of the default, sad or gwh guesses. I’m not going to continue with these - the results we have should be sufficient.

Kind regards,
David.