Chemistry Reference
In-Depth Information
Table 7
MP2 and CCSD(T) Interaction Energies of the Stacked and T-shaped Dimers
of Acetylene and Diacetylene
a
Orientation
CSD
ð
T
Þ
Dimer
CCSD(T)
MP2
d
MP2
C
C
H)
2
b
Stacked
(H
C
C
1
:
31
2
:
38
þ
1
:
07
C
C
H)
2
b
T-shaped
(H
C
C
1
:
63
1
:
94
þ
0
:
31
C
H)
2
c
Stacked
(H
C
1
:
72
1
:
99
þ
0
:
17
C
H)
2
c
T-shaped
(H
C
2
:
32
2
:
45
þ
0
:
13
a
All values are in kcal mol
1
.
b
Reference 93
c
Reference 120
C
dimer (H
H)
2
. In agreement with the ''gold standard'' of quantum
chemistry [i.e., the CCSD(T) method], MP2 calculations correctly predict an
attractive interactions in each case. (See data in Table 7.) However, for the
dimers composed of fragments containing a delocalized
C
electron network,
the MP2 method substantially overestimates the interaction energy (by more
than 1 kcal mol
1
, which represents a relative error in excess of 80%) while
the difference between the CCSD(T) and MP2 results (
p
CCSD
ð
T
Þ
d
) is modest for
MP2
17 kcal mol
1
or 10%). Matters are made even
worse by the fact that the MP2 error is not uniform across the entire potential
surface, which can lead to qualitatively incorrect conclusions about the nature
of delocalized
C
(H
C
H)
2
(less than 0
:
-type interactions. With MP2 computations, one would con-
clude that the stacked configuration of (H
p
C
C
H)
2
is more stable
than the T-shaped structure. However, the opposite (and correct) conclusion
is reached when the effects of higher order excitations are included via CCSD
(T) calculations.
C
C
Spin-Scaled MP2
Several approaches have been introduced that attempt to address this
shortcoming of MP2 for delocalized
interactions. In 2003, Grimme intro-
duced the spin-component-scaled second-order Møller-Plesset perturbation
theory (SCS-MP2) method where the parallel spin (
p
""
) and antiparallel spin
(
) pair correlation energies (which are related to the singlet and triplet com-
ponents of the correlation energy) were assigned different weights.
145
The
empirical scaling parameters (
p
""
¼
"#
6
1
3
) improve MP2 results signif-
icantly not only for a variety of reaction energies and atomization energies but
also for interaction energies between delocalized
5
and
p
"#
¼
systems (including the ben-
zene dimer). The approach has no additional computational overhead relative
to MP2, and it preserves the size consistency of the MP2 method. Other var-
iations of these scaling parameters have since been introduced in an attempt to
further improve the description of noncovalent interactions and/or reduce the
computational demands.
146-149
Of particular interest is the SCSN-MP2 meth-
od of Platts and Hill in which the scaling parameters were optimized using
p
