bizarre CCSD correlation energy


Click here for full thread
Clicked A Few Times
Dear Forum,

I have been trying to do some CCSD(T) calculations on some open-shell species and I am getting some correlation energies that don't seem to make a lot of sense.

So I have a (OH) substituted phenyl radical and I'm looking at the conformational energy difference between two OH rotomers. If I perform an ROCCSD(T) calculation on the two rotomers I get an energy difference:

c1-c2 = +9.6 kJ mol-1

with energies like the following:

c1:

CCSD correlation energy / hartree =        -1.138729684526594
CCSD(T) correction energy / hartree = -0.054336569695383
CCSD(T) total energy / hartree = -307.359516043510553

c2:

CCSD correlation energy / hartree =        -1.138739493778493
CCSD(T) correction energy / hartree = -0.054526511816299
CCSD(T) total energy / hartree = -307.363175100201033

However if I do this with UCCSD(T) then I get the following:

c1:

CCSD correlation energy / hartree =        -1.111323500626606
CCSD(T) correlation energy / hartree = -1.162727145664610
CCSD(T) total energy / hartree = -307.357498242822828

c2:

CCSD correlation energy / hartree =        -0.720195222066366
CCSD(T) correlation energy / hartree = -0.744704157414741
CCSD(T) total energy / hartree = -306.942326549647646

So you see that the correlation energy for the c2(UCCSD) calculation is about half as much as all the other calculations. How did this happen? Any solution here would be greatly appreciated. I have checked the following and have now run out of possible ideas:

(1) The SCF converges into the in/correct solution. This has been cross-checked against the same calculation with the a different program and was able to get the same SCFs in all 4 calculations.

(2) Ok so then it is something to do only with the correlation calculation. Is it the number of frozen orbitals? Well I have made sure it is the same number in each calculation...can't be this.

(3) Symmetry? This is were I think something is up. conf2 is in cs symmetry whereas conf1 is in c1 symmetry. I have tried: both with and without SYMMETRY CS in the geometry directive and I get the same answer. But because the ROCCSD(T) calculation is correct but the UCCSD(T) is not, is it that all the correlation energy is not being calculated? This is the input file that I used:

memory stack 5000 mb heap 1000 mb global 5000 mb noverify
PERMANENT_DIR /home/gpw501/tryptophan/oxidation/benzene_OH/bz_OH/
SCRATCH_DIR /scratch/gpw501/
GEOMETRY
SYMMETRY CS
C 0.99645638 -0.42648682 0.00000000
C -1.13763952 0.06248425 1.22330296
C -1.13763952 0.06248425 -1.22330296
C -1.83376503 0.20763768 0.00000000
C 0.19113955 -0.24170594 -1.24554896
C 0.19113955 -0.24170594 1.24554896
H 1.44466674 -1.43731034 0.00000000
H 0.73399717 -0.33302689 -2.18585706
H 0.73399717 -0.33302689 2.18585706
H -2.89396286 0.45223898 0.00000000
H -1.67575204 0.20666754 -2.15983510
H -1.67575204 0.20666754 2.15983510
O 2.14601254 0.42733523 0.00000000
H 1.80402493 1.33388138 0.00000000
END
BASIS large
* library 6-311++G(2d,2p)
END
BASIS small
* library 6-31G*
END
set "ao basis" small
SCF
maxiter 500
DOUBLET
UHF
vectors input atomic output small.movecs
THRESH 1e-7
END
task scf energy

set "ao basis" large
 SCF
maxiter 500
DOUBLET
UHF
vectors input project small small.movecs output large.mos
THRESH 1e-7
END
TCE
SCF
tilesize 15
io ga
maxiter 400
CCSD(T)
FREEZE 7
END
TASK TCE ENERGY