Chemistry Reference
In-Depth Information
Table 3
Interaction Energies (kcal mol
1
) for the Benzene Dimer
a
Method
Basis
S
T
PD
aug-cc-pVDZ
0
MP2
1.88
2.35
2.89
aug-cc-pVDZ*
b
2.83
3.00
4.12
aug-cc-pVDZ
2.90
3.07
4.22
aug-cc-pVTZ
3.25
3.44
4.65
aug-cc-pVQZ*
c
3.35
3.48
4.73
aug-cc-pVQZ
3.37
3.53
—
aug-cc-pVDZ*
b
CCSD(T)
1.33
2.24
2.22
aug-cc-pVDZ
d
1.39
2.31
2.30
estd. CCSD(T)/aug-cc-pVQZ*
2.63
a
Unless otherwise noted, all computations used intermonomer distances optimized at each level
of theory with rigid monomers (C
1.70
2.61
0800
˚
; Ref. 39). Data from Ref. 61
C
¼
1
:
3915
;
C
H
¼
1
:
except as noted.
b
This is aug-cc-pVDZ for carbon and cc-pVDZ for hydrogen.
c
This is aug-cc-pVQZ less
g
functions for carbon and less
f
functions for hydrogen.
d
Data from Ref. [66].
Stanton. Table 3 presents the corresponding interaction energies. Tables 2 and 3
contain some comparisons of these truncated basis sets against the correspond-
ing full basis sets. Examining MP2 geometries using the aug-cc-pVDZ basis or
its truncated versions, aug-cc-pVDZ* or aug-cc-pVDZ
0
, we see very little
change in the optimum intermolecular distances for each of the configurations
considered (we scanned the surface with a resolution of 0.1
˚
around the mini-
ma). The distances generally agree within 0.1
˚
. Similar results are observed
when comparing CCSD(T) geometries using the aug-cc-pVDZ or aug-cc-
pVDZ* basis sets. Again, when considering the aug-cc-pVQZ basis to its trun-
cated version aug-cc-pVQZ*, the MP2 geometries are virtually identical. This
suggests that, when necessary (and after careful benchmarking), some diffuse
functions and/or higher angular momentum functions may be safely removed
from the aug-cc-pVXZ basis sets for the prediction of intermolecular geome-
tries. However, the main point of Table 2 is that the intermolecular distances
for the benzene dimer are fairly insensitive to the basis set used. Electron corre-
lation also has only a minor effect, typically increasing the intermolecular dis-
tances by about 0.2
˚
upon improving the model from MP2 to CCSD(T).
When considering the interaction energies in Table 3, the truncated aug-
cc-pVDZ basis sets can lead to significant differences. Although the aug-cc-
pVDZ* basis (lacking only diffuse functions on hydrogens) remains within
0.1 kcal mol
1
for MP2 interaction energies, the smaller aug-cc-pVDZ
0
basis
(which excludes also diffuse
d
functions on carbon) is in error by as much as
1.3 kcal mol
1
compared to the MP2 values using the full basis. Thus, for
comparison to results for a given electron correlation model, care should be
exercised in any attempts to reduce the size of the basis by truncating certain
functions. This is not to say, however, that the MP2/aug-cc-pVDZ
0
results
are not useful. Quite the contrary—they seem to exhibit a very favorable
