I am trying to disable all the empirical scalings and execute a “pure” OOMP2 with both direct and exchange terms. Is there a proper way of conducting such calculations? According to the manual page, this can be achieved with
do_o2 = 3
I am attaching an output with two jobs. The first one runs OOMP2; the second one reads orbitals, skips SCF, and evaluates MP2 on top of them. The final energies of the two runs are different, which makes me think that the empirical stuff is not dropped somewhere.
Somehow the forum web interface does not let me attach text files. So, this is the input:
$molecule
0 1
N -1.578718 -0.046611 0.000000
H -2.158621 0.136396 -0.809565
H -2.158621 0.136396 0.809565
H -0.849471 0.658193 0.000000
$end
$rem
INPUT_BOHR FALSE
MAX_SCF_CYCLES 30
exchange hf
do_o2 3
BASIS aug-cc-pvtz
SYM_IGNORE TRUE
JOBTYPE SP
SCF_ALGORITHM gdm
N_FROZEN_CORE 0
N_FROZEN_VIRTUAL 0
AUX_BASIS_CORR RIMP2-aug-cc-pVTZ
SCF_CONVERGENCE 9
SCS 0
purecart 1111
$end
@@@
$molecule
read
$end
$rem
MAX_SCF_CYCLES 0
scf_guess = read
method mp2
BASIS aug-cc-pvtz
SYM_IGNORE TRUE
JOBTYPE SP
N_FROZEN_CORE 0
N_FROZEN_VIRTUAL 0
AUX_BASIS_CORR RIMP2-aug-cc-pVTZ
SCS 0
purecart 1111
$end
The output is pretty long. This is the end of the first job:
Components of the RI-MP2 Correlation Energy:
RI-MP2 ENERGY (aa|aa) = -0.0287264898 au
RI-MP2 ENERGY (aa|bb) = -0.1017526610 au
RI-MP2 ENERGY (bb|aa) = -0.1017526610 au
RI-MP2 ENERGY (bb|bb) = -0.0287264898 au
RI-MP2 SINGLES ENERGY = 0.0000000000 au
RI-MP2 CORRELATION ENERGY = -0.2609583015 au
RI-MP2 TOTAL ENERGY = -0.2609583015 au
Skipping Job=167....
Skipping Job=164....
Skipping remaining construction of the OS gradient
ETot before SOS = -56.2183073494
ETot after SOS = -56.4792656509
15 -56.4792656509 9.55E-10 Convergence criterion met
---------------------------------------
SCF time: CPU 70.22 s wall 10.35 s
SCF energy in the final basis set = -56.47926565
Total energy in the final basis set = -56.47926565
This is the relevant part of the second job:
Components of the RIMP2 correlation energy:
aaaa correlation energy = -0.0287264898 a.u.
abab correlation energy = -0.2035053220 a.u.
bbbb correlation energy = -0.0287264898 a.u.
non-Brillouin singles = -0.0016085332 a.u.
total same-spin energy = -0.0574529795 a.u.
total opposite-spin energy = -0.2035053220 a.u.
Total RIMP2 correlation energy = -0.2625668347 a.u.
RIMP2 total energy = -56.4808741840 a.u.
Notice that both correlation energies and total energies are different.
Eventually I gave up and obtained the values from PySCF by overloading the FCI solver. The resulting total energy is -56.4791150348922 a.u.