Upload
inmaculada-garcia
View
212
Download
0
Embed Size (px)
Citation preview
CCSD-CTOCD Static Dipole Shielding Polarizability forQuantification of the Chiral NMR Effects in OxaziridineDerivatives
Stefano Pelloni[a] and Inmaculada Garc�ıa Cuesta*[b]
Chiral discrimination by nuclear magnetic resonance (NMR)
spectroscopy might be achieved through the pseudo-scalar
derived from the dipole shielding polarizability tensor. Coupled
Cluster Singles and Doubles-Quadratic Response (CCSD-QR) cal-
culations inside the continuous translation of the origin of the
current density formalism have been carried out to determine
the effects of basis set, electron correlation, and gauge transla-
tion on the determination of this magnitude in oxaziridine deriva-
tives. Inclusion of electronic correlation is needed for adequately
describing the pseudo-scalar for the heavier nuclei, making CCSD
a rigorous and affordable method to compute these high order
properties in medium-sized molecules. The observable magni-
tudes for chiral discrimination (produced RF voltage and required
electric field) are calculated. Half of the considered molecules
show values of the observable magnitudes near the lower limit
for experimental detection. Nuclei 19F, 31P, and 79Br produce
the largest values of RF voltage (50–80 nV). Moreover, 31P and79Br are the nuclei requiring smallest electric fields (3 MVm21) to
separate the NMR signals, being then suitable for both the
techniques. VC 2014 Wiley Periodicals, Inc.
DOI: 10.1002/jcc.23689
Introduction
It is known that nuclear magnetic resonance (NMR) spectroscopy
has a “chiral blindness” as the two mirror images of a chiral mol-
ecule cannot be distinguished. Magnetic properties like nuclear
magnetic shieldings or coupling constants do not change as
they have the same sign and magnitude on inverting the coordi-
nate system, that is, passing from one enantiomer to another.
Conversely, resolution of racemic mixtures, recognition of enan-
tiomers, and absolute configuration assignment are central prob-
lem in synthetic biochemistry. Some years ago, Buckingham[1]
tried to give a solution with the introduction of a pseudo-scalar
related to the magnetoelectric third-rank tensor dipole shielding
polarizability (DSP). The DSP describes the interaction of a mag-
netic and electric field with matter, that is, an NMR experiment
with an external electric field that can be static or dynamic. The
important point is that this tensor changes sign passing from a
system to its mirror image and, therefore, makes it possible, at
least in principle, to recognize two enantiomers of a molecule.
Chiral discrimination via NMR spectroscopy could be a quick and
affordable method, but most of theoretical calculations of DSP pre-
dicts magnitudes too small to be experimentally detected and,
hence, to differentiate between two enantiomers of a molecule by
NMR measures was not feasible.[2–4] However, recent works have
encountered[5] at least two order of magnitude larger than those
reported so far; therefore, the experimental detection is starting to
be possible. In this context, an accurate theoretical description of
the magnitude is of fundamental importance to be susceptible of
being unequivocally confirmed by experiment.
The effects of an electric field on NMR spectral parameters
have been theoretically determined since some time ago,[3,6–14]
either in a purely analytically manner through quadratic response
(QR) methods or by numerical differentiation of analytical shield-
ing tensors. Even though there are predominate calculations at
the coupled Hartree–Fock (CHF) level, correlation contributions
may be important as the nuclear magnetic shieldings are, in
many molecules, strongly affected by electron correlation.[15–21]
Indeed, it is possible to find a variety of examples in the literature
of calculations of DSP with different correlated methods, includ-
ing Complete Active Space Self Consistent Field (CASSCF), MP2
and MP3, coupled cluster, and several parameterizations of the
density functional theory (DFT). Unfortunately, such studies have
mainly focused on small molecular models not necessarily pre-
senting chiral behavior. However, chiral molecules of chemical
interest are usually of larger size and contain several heavy atoms
often even from the third or fourth row of the periodic system.
Therefore, we find it convenient to carry out a systematic investi-
gation of DSP in some medium-sized chiral molecules account-
ing also for correlation effects.
As the nuclear magnetic shieldings are origin sensitive, it is
advantageous to use a gauge-independent method as the
continuous translation of the origin of the current density
(CTOCD).[22] We use in this work, a coupled cluster implemen-
tation of the CTOCD method in its diamagnetic zero (DZ)
[a] S. Pelloni
Dipartimento di Chimica Universit�a degli Studi di Modena e Reggio Emilia.
via Campi 183, 41100 Modena, Italy
[b] I. G. Cuesta
Instituto de Ciencia Molecular, Universidad de Valencia, P.O. Box 22085,
46071 Valencia, Spain
E-mail: [email protected]
Contract grant sponsor: Spanish MICINN; Contract grant number:
CTQ2010–19738; Contract grant sponsor: Italian MIUR (Ministero
dell’Istruzione, dell’Universit�a e della Ricerca; PRIN 2009 scheme)
VC 2014 Wiley Periodicals, Inc.
Journal of Computational Chemistry 2014, 35, 1815–1823 1815
FULL PAPERWWW.C-CHEM.ORG
formulation, the only CTOCD method in which analytical QR is
available, while the others need numerical integration. A study
on (R)-N-fluor-C-methyl-oxa-aziridine with a hierarchy of
increasing basis sets and of coupled cluster methods has been
carried out to investigate for the first time the performance of
the CTOCD-DZ approach to compute DSP. Furthermore, differ-
ent gauge origins have been considered, namely the center of
mass (CM) or each one of the atoms themselves.
The choice of such aziridine derivative has three principal
reasons: (i) oxaziridine molecule is often used as substrate in a
wide range of enantioselective organic oxidation reactions,
and there is a lot of publications on the chemistry of oxaziri-
dine;[23,24] (ii) in a previous article,[4] the (2R)N-methyloxaziri-
dine shows the highest value of DSP found in the literature
(for oxygen atom) and so oxaziridine was a good starting
point to build a new molecule that may show interesting fea-
tures; (iii) this system contains C, O, N, and F atoms, that is,
the nonmetals of the second row of the periodic table, those
are the bricks to build the 90% of the known molecules.
From the conclusions of the previous assessment, the study
has been extended to include also a set of derivatives of the
(R)-N-fluor-C-methyl-oxa-aziridine obtained by substitution of
oxygen by sulfur, nitrogen by phosphorus, and/or fluorine by
chlorine or bromine. As we are mainly interested in the chiral
discrimination problem, we also computed the chiral NMR
effects for the whole set of considered molecules using two of
the strategies previously proposed by Buckingham:[1,2] the cal-
culation of the RF voltage produced by the rotating chiral
electric polarization induced by permanent magnetic nuclear
momentum and the required electric field to produce a differ-
ence of 0.01 ppm between the NMR chemical shifts of the
two enantiomers.
This article is organized as follows: in Calculation Method
section, we summarize the theory that leads to the definitions
of the response properties of interest as well as the technical
details of the reported calculations. The first part of Results
and Discussion section analyzes the effects of atomic basis,
gauge origin, and correlation level on the computed CTOCD-
DZ DSP of (R)-N-fluor-C-methyl-oxa-aziridine. The second part
of this section is devoted to the theoretical determination of
observable magnitudes for chiral discrimination for the whole
set of considered derivatives. Finally, some concluding remarks
are given in Conclusions section.
Calculation Method
Theoretical background
From a Taylor-series expansion of the total energy of a mole-
cule with nuclei with nuclear magnetic dipoles mI in the pres-
ence of external magnetic field B and electric field E (see for
instance, eq. (365) of Ref. [25]), it is trivial to identify the
nuclear magnetic shielding as
rIab5
@2W
@mIa@Bb
����mI;B!0
(1)
and the nuclear magnetic shielding polarizability as
rIabc5
@3W
@mIa@Bb@Ec
����mI;B!0;E!0
(2)
For an isotropic medium, it is expedient to define molecular
average properties by introducing the pseudo-scalar[26–28]
rð1ÞI5
1
6�abcr
Iabc (3)
where �abc is the skew-symmetric Levi-Civita tensor. Therefore,
the tensor rIabc has a chirally sensitive isotropic part rð1Þ
I. Fur-
thermore, the pseudo-scalar allows for compact expressions
for the magnetic field induced at nucleus I
BnI 52rIB2r 1ð Þ I
B3E (4)
the electric dipole induced in the electron cloud
M5 2r 1ð Þ ImI3B (5)
and the induced magnetic dipole
M 5 2rImI2r 1ð Þ IE3mI (6)
Being a third-order property, the shielding polarizability can
be expressed in terms of QR functions. Using the notation of
Bishop, Orr, and Ward[6–8] for the propagator,[29,30] eqs. (7) and
(8) give, respectively, the expressions for the frequency-
dependent paramagnetic ðrpIabcÞ and diamagnetic (rdI
abcÞ contri-
butions to the magnetic-shielding polarizability
rpIabc 2xr; x1;x2ð Þ52hhBn
Ia; mb; lciix1;x2
521
�h2
XP
Xj;k 6¼a
hajBn
Iajjih jjlc jki hkjmbjaixja2xr� �
xka2x1ð Þ(7)
rdIabc 2x; xð Þ52<hhrdI
ab; lciix
51
�h
Xj 6¼a
2xja
x2ja2x2
� �< hajrdIabjji hjjlcjai
� �(8)
where, in addition to the standard definitions of the electric and
magnetic dipole operators, we have used xr 5 x1 1 x2, over-
lined operators are defined, for instance, as la5la2hajlajaiand RP stands for the sum over all permutations of the pairs
Bn
Ia=2xr
� �, mb=x1
� �, lc=x2
� �. The operator for the diamag-
netic contribution is written as
rdIab5
e
2mec2
Xn
i51
ricEi
Icdab2riaEi
Ib
� �" #(9)
Ei
I is the operator for the electric field exerted on nucleus I by
electron i, such that the operator for the total field of n elec-
trons is
En
I 5Xn
i51
Ei
I5Xn
i51
1
4p�0e
ri2RI
jri2RIj3(10)
FULL PAPER WWW.C-CHEM.ORG
1816 Journal of Computational Chemistry 2014, 35, 1815–1823 WWW.CHEMISTRYVIEWS.COM
while Bn
I is the operator of the magnetic field of n electrons
on nucleus I[25,27,31]
Bn
I 52l0
4pe
meM
n
I 52l0
4pe
me
Xn
i51
ri2RI
jri2RIj33pi (11)
In the CTOCD-DZ[25] scheme, the diamagnetic term is
replaced by a D contribution (rDIabcÞ:
rDIabc 2xr; x1;x2ð Þ5�bklhhPk; Rc; T
n
Ilaiix1;x2
5�bkl1
�h2
XP
Xj;k 6¼a
hajPkjjihjjRc jki hkjTn
Ilajaixja2xr� �
xka2x1ð Þ(12)
In eq. (12), P is the linear momentum operator and T is a
Hermitian operator needed in the CTOCD-DZ approach:
Tn
Iab51
2
Xn
i51
ria2r0a� �
Mi
Ib1Mi
Ib ria2r0a� �
(13)
An indirect measure of the pseudo-scalar rð1ÞI
can be
obtained by taking into account[1,2] that the precessing
nuclear magnetization MIx induces a rotating chiral electric
polarization PIy ,
PIy5rð1Þ
IMI
x Bz (14)
which generates a RF voltage in a capacitor with plates at 6 d2
on the y axis incorporated into the resonance circuit
VðIÞ5PI
y d
e21ð Þe05
NIjrð1ÞIj gIlNB0
z
� �2II II11ð Þ
3kT
d
e21ð Þe0(15)
where NI is the number of nuclei per unit volume, gI is the
g-factor of the nucleus II, mN is the nuclear magneton, B0z the
magnetic field, II is the nuclear spin, e is the dielectric constant
of the medium and e0 is the permitivity of free space.
An alternative strategy is based on the chiral chemical shift
(i.e., the difference of nuclear magnetic shieldings between the
two enantiomers) produced by the external electric
field.[2,4,5,32] The shift depends on the magnitude of the field
and on the pseudo-scalar rð1ÞI
[see eq. (4)]. Consequently, it is
possible to determine the electric field required to observe an
induced magnetic field equivalent to an experimentally detect-
able shift. Note, however, that the induced field would be per-
pendicular to the strong magnetic field B of an NMR
spectrometer and, therefore, with no effect on the chemical
shifts at first order. Experimental designs overpassing this
problem have been discussed in Ref. [2].
Computational details
All geometries were optimized using the Gaussian03 pro-
gram[33] at the B3LYP/aug-cc-pVTZ level. The paramagnetic
[eq. (7)] and D [eq. (12)] contributions to static DSP tensor
components were calculated using the Dalton[34] program via
QR for both correlated and uncorrelated calculations. As cus-
tomary in CC response theory, we use the unrelaxed CC
approach, although—obviously—full relaxation is included in
the CHF estimations. To reach the Hartree–Fock limit in the
CHF calculations, four basis set of increasing size were used:
aug-cc-pVDZ, aug-cc-pVTZ, aug-cc-pVQZ, and aug-cc-pV5Z.[35]
Likewise, Sadlej-pVTZ[36–40] developed for electric polarizabil-
ities was also checked for latter use in the correlated calcula-
tions. Electron correlation effects in the calculation of the
shielding polarizability pseudo-scalar are analyzed in the
CTOCD-DZ scheme at different coupled cluster levels (CCS,
CC2, CCSD) inside the QR formalism. In addition, the men-
tioned pseudo-scalar was also computed through finite field’s
(FF) calculations at the CC2 and CCSD levels using analytically
evaluated nuclear magnetic shieldings. According to previous
literature,[9,41,42] we have used electric fields of 0.001 au of
magnitude.
Results and Discussion
Calculation assessment
To validate the used theoretical approach, we start by carrying
out a preliminary study of the effects of basis sets, gauge ori-
gin, and correlation contributions in the parent molecule (R)-
N-fluor-C-methyl-oxa-aziridine (see Fig. 1). This system is a chi-
ral molecule containing four heavy atoms with an asymmetric
carbon in a three-member ring.
It is well known that the CTOCD-DZ approach requires in
many cases rather large basis to guarantee reliable results.
Conversely, the N6 scaling—N is the size of the atomic basis
Figure 1. (R)-N-fluor-C-methyl-oxa-aziridine.
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 1815–1823 1817
set—of CCSD calculations makes it convenient from a compu-
tational point of view to reduce as much as possible the afore-
mentioned basis set size. Clearly, for size-enlarged and less-
symmetric molecules, the use of extended basis becomes
more and more problematic. It is, therefore, desirable to use
basis set of relatively small size but still able to reproduce the,
in principle, more accurate results form larger basis. Conse-
quently, we have analyzed the convergence of the calculated
value of the pseudo-scalar rð1ÞI
with a hierarchy of Dunning’s
correlation consistent basis sets for the considered (R)-N-fluor-
C-methyl-oxa-aziridine. In Table 1, our results of rð1ÞI
for heavy
nuclei 13C, 14N, 17O, and 19F are shown. We present the para-
magnetic, D, and total contributions, which have been com-
puted at the CHF-QR(CTOCD-DZ) level of theory. In addition,
the results corresponding to the complete basis set (CBS)
extrapolation,[44–46] as well as those using Sadlej’s TZV basis
are also presented. The former were calculated as
lim X!1 a1bexp 2c Xð Þf g.To calculate the CBS limit, the extrapolation was carried out
independently for each one of the QR functions in eqs. (7) and
(12) using Dunning’s aug-cc-pVXZ with X ranging from D to 5.
In most of the cases, a monotonic behavior was encountered
although in some instances the aug-cc-pVDZ basis results
break the monotonicity. At any rate, differences from including
or not the aug-cc-pVDZ basis results are not relevant, as they
appear only in the smallest contributions to the pseudo-scalar,
the dominant terms in it showing a monotonous trend. How-
ever, those small components—together with the opposite
sign of the large components—are responsible of the lack of
monotonicity of the overall pseudo-scalar. Only the nitrogen
atom behaves differently, because of a very slow convergence
of the XZY element of the paramagnetic DSP tensor, which is
the one dominating rð1ÞN
. To illustrate this fact, consider that
the CBS limit of rpIxzy varies from 231.0719 au to 229.9382 au
depending whether the aug-cc-pVDZ basis is included or not
in the extrapolation, in spite of the perfectly monotonic
behavior of the quadratic function with respect of basis set
size in this case.
For all atoms, the values of rð1ÞI
total are rather similar with
all basis sets; indeed, the computed value for the aug-cc-pV5Z
is nearly identical in all the cases to the limit value with the
exception mentioned above. Nonetheless, the paramagnetic
and D contributions vary a little bit more, even though the
paramagnetic contribution—that is, the leading one—shows a
fast convergence. The convergence of the D contribution is
slightly slower, but we note that it converges to zero in
accordance to the fact that the diamagnetic part of the DSP
tensor rIabc is symmetric in bc and, therefore, does not contrib-
ute to the pseudo-scalar if the origin of the coordinate system
is taken on the nucleus in question.[2]
As mentioned before, we have also tested Sadlej’s TZV
basis looking for a reduced-size atomic basis set. Having
been optimized for the computation of electric properties,
notably it shows a performance close to that of the aug-cc-
pVQZ for C1, C2, O, and F atoms as can be checked by
comparison to the limit values. Due to the mentioned slow
convergence of the large XZY element of the DSP tensor in
the case of N atom, Sadlej’s basis results are not so close
to those from aug-cc-pVQZ basis, as the overestimation of
the paramagnetic part is not cancelled by an adequate
compensation from the D contribution, as happens for the
fluorine atom. Remarkably, Sadlej’s basis gives for nitrogen
the results closest to the CBS limit. Moreover, Sadlej’s TZV
basis provides results that are inside a margin of 2 ppm
with respect to the values extrapolated to the complete
basis.
Even though the convergence of the basis at the CPHF level
is a necessary condition to achieve convergence also at CCSD,
it is obviously not sufficient, as correlated methods require
larger basis sets, especially for the QR functions that we calcu-
late. An estimation of the quality of Sadlej’s basis for CCSD-QR
calculations can be found by comparing the results using it to
those using aug-cc-pVQZ basis. We focus in the fluorine
atom, for it is the most sensitive to the basis set size in (R)-N-
fluor-C-methyl-oxa-aziridine. In particular, we obtained a value
of rð1ÞF5 226:8 ppm with Dunning’s and a slightly higher
value of rð1ÞF5 228:5 ppm with Sadlej’s basis set. Summariz-
ing, Sadlej’s TZV basis set represents a fair compromise
between accuracy and computational cost of the DPS pseudo-
scalar rð1ÞI.
Table 1. Basis set dependence of static dipole-shielding polarizability
pseudo-scalar rð1ÞI
for heavy atoms of (R)-N-fluor-C-methyl-oxa-aziridine
molecule in ppm au.
Atom Basis set Paramagnetic D Total
C1 aug-cc-pVDZ 222.346 20.531 222.876
aug-cc-pVTZ 224.218 0.110 224.108
aug-cc-pVQZ 224.001 0.146 223.855
aug-cc-pV5Z 224.254 0.072 224.182
CBS 224.79 0.04 224.75
Sadlej-VTZ 223.793 0.051 223.742
C2 aug-cc-pVDZ 27.848 0.766 27.082
aug-cc-pVTZ 28.360 0.313 28.048
aug-cc-pVQZ 28.338 0.070 28.268
aug-cc-pV5Z 28.355 0.010 28.344
CBS 28.41 20.17 28.57
Sadlej-VTZ 28.310 0.354 27.955
O aug-cc-pVDZ 230.792 1.362 232.154
aug-cc-pVTZ 231.994 21.478 230.516
aug-cc-pVQZ 233.090 20.798 232.292
aug-cc-pV5Z 234.111 20.286 233.825
CBS 234.09 20.26 233.83
Sadlej-VTZ 232.343 0.071 232.414
N aug-cc-pVDZ 243.706 1.746 241.961
aug-cc-pVTZ 238.402 0.639 237.762
aug-cc-pVQZ 237.956 0.166 237.790
aug-cc-pV5Z 237.976 0.140 237.835
CBS 244.09 20.05 244.14
Sadlej-VTZ 245.077 1.032 244.045
F aug-cc-pVDZ 223.280 6.364 216.916
aug-cc-pVTZ 221.097 1.247 219.849
aug-cc-pVQZ 220.865 0.442 220.423
aug-cc-pV5Z 220.823 20.024 220.847
CBS 220.14 21.68 221.82
Sadlej-VTZ 224.365 5.008 219.357
The conversion factors for nuclear magnetic shielding polarizability, from
atomic units to SI and cgs units, obtained from the CODATA recom-
mended values (Ref. [43]) are 1 ppm au 5 1026(ea0/Eh) 5 1.94469057 3
10218 m V21 5 5.83003566 3 10214 cm stat V21, respectively.
FULL PAPER WWW.C-CHEM.ORG
1818 Journal of Computational Chemistry 2014, 35, 1815–1823 WWW.CHEMISTRYVIEWS.COM
To confirm the intrinsic gauge-origin independence of the
CTOCD-DZ approach, as well as to indirectly check the cor-
rectness of the variety of contractions of four index tensors
carried out, five CTOCD-DZ QR shifted origin calculations
have been performed for each atoms putting the origin of
the coordinate system in all heavy atoms. The results are
summarized in Table 2.
The gauge variation causes a change in the paramagnetic
contribution and an opposite, with equal magnitude, shift for
the D term, and as a result the sum does not vary signifi-
cantly. In fact, the maximum difference found because of the
displacement of the origin of gauge is approximately 0.4
ppm for the fluorine atom. Such a difference is very much
smaller than the deviations introduced for not taking a com-
plete (and clearly unfeasible) basis set. Again, it is convenient
to recheck the quality of our results when correlation is
switched on. The last section of Table 2 shows that gauge
origin invariance is maintained at the CCSD level. The
obtained results, then, prove the reliability of the chosen
method to calculate the DSP. Thus, the rest of reported calcu-
lations have been done using Sadlej’s TZV basis and CTOCD-
DZ approach.
We finish the study on (R)-N-fluor-C-methyl-oxa-aziridine
considering the effects of electron correlation on the CTOCD-
DZ estimations of the pseudo-scalar at different coupled clus-
ter levels, specifically CCS, CC2, and CCSD. As customary in
coupled cluster response theory, we use unrelaxed QR func-
tions, that is, Hartree–Fock orbitals are not allowed to adapt to
the perturbation in the post-Hartree–Fock part of the calcula-
tions. Anyway, relaxation effects are partially accounted for as
single amplitudes are treated to infinite order.[47,48] Further-
more, we recall that for static properties the unrelaxed
coupled cluster hierarchies converges toward the full-
configuration interaction limit, if the basis sets are perturba-
tion independent as we do here.[49] For the sake of complete-
ness, we have also performed CCS calculations, as they do
include neither correlation nor relaxation effects. Table 3
shows the computed values of rð1ÞI
for the heavy nuclei of
the studied aziridine at the different considered levels of
theory.
The same that in the case of CTOCD-DZ nuclear magnetic
shieldings,[16] the first point to remark on the numbers in
Table 3 is the absolute lack of accuracy of the CCS esti-
mates for rð1ÞI, which are consistently lower than the pre-
dicted by the other methods. Parallel again to the NMR
shieldings, the inadequacy of the CCS approach is caused
by not including at all the orbital relaxation contributions
as proven by the reasonable quality of the values com-
puted inside the CHF formalism, which implicitly includes
orbital relaxation.
Table 2. Origin dependence of static dipole-shielding polarizability
pseudo-scalar rð1ÞI
for heavy atoms of (R)-N-fluor-C-methyl-oxa-aziridine
molecule in ppm au.
Atom Gauge Paramagnetic D Total
CHF-QR
C1 C.M. 223.793 0.051 223.742
C1 223.777 0.028 223.748
C2 224.722 0.974 223.748
O 225.966 2.229 223.737
N 223.624 20.115 223.739
F 224.418 22.330 223.748
C2 C.M. 28.310 0.354 27.955
C1 28.292 0.341 27.952
C2 28.049 0.095 27.954
O 26.273 21.686 27.960
N 28.613 0.660 27.953
F 29.978 2.020 27.958
O C.M. 232.343 0.070 232.414
C1 231.776 0.692 232.467
C2 227.280 5.167 232.447
O 232.530 20.057 232.473
N 232.223 0.125 232.349
F 236.645 24.341 232.304
N C.M. 245.077 1.032 244.045
C1 245.173 1.159 244.014
C2 242.814 21.116 243.930
O 247.979 4.108 243.871
N 244.062 20.012 244.074
F 245.027 0.861 244.166
F C.M. 224.365 5.008 219.357
C1 229.555 10.173 219.382
C2 238.788 19.450 219.337
O 214.239 24.779 219.019
N 221.357 2.006 219.351
F 219.377 20.064 219.441
CCSD-QR
C1 C.M. 229.040 20.110 229.150
F 226.908 22.241 229.150
O C.M. 203.955 0.709 204.664
F 201.688 2.982 204.670
N C.M. 259.380 20.303 259.683
F 261.560 1.876 259.684
F C.M. 232.595 5.749 226.845
F 226.817 20.029 226.846
All calculations used the Sadlej-pVTZ basis set. CM is the center of
mass.
Table 3. CTOCD-DZ Coupled Cluster estimates of dipole-shielding polariz-
ability pseudo-scalar rð1ÞI
in ppm au for (R)-N-fluor-C-methyl-oxa-aziridine
molecule using the Sadlej-pVTZ basis set.
Atom Method Paramagnetic D Total
C1 CHF-QR 223.793 0.051 223.742
CCS-QR 219.338 0,715 218.624
CC2-QR(FF[a]) 233.728 0.347 233.382 (233.639)
CCSD-QR(FF) 229.040 20.110 229.150 (229.154)
C2 CHF-QR 28.310 0.354 27.955
CCS-QR 27.541 21.242 28.782
CC2-QR(FF) 213.370 0.900 212.470 (212.500)
CCSD-QR(FF) 211.227 20.132 211.359 (211.361)
O CHF-QR 232.343 0.070 232.414
CCS-QR 165.708 9.838 175.546
CC2-QR(FF) 245.984 23.804 242.180 (241.256)
CCSD-QR(FF) 203.955 0.709 204.664 (204.681)
N CHF-QR 245.077 1.032 244.045
CCS-QR 232.440 8.400 224.040
CC2-QR(FF) 281.002 20.581 281.583 (280.555)
CCSD-QR(FF) 259.380 20.303 259.683 (259.686)
F CHF-QR 224.365 5.008 219.357
CCS-QR 219.257 13.323 25.934
CC2-QR(FF) 244.977 5.279 239.698 (239.185)
CCSD-QR(FF) 232.595 5.749 226.845 (226.843)
[a] Finite Fields.
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 1815–1823 1819
From the analysis above, it is evident that orbitals’ relaxation
effects are required to get a trustable picture of the electron
correlation contribution to the DSP pseudo-scalar. The lowest
level coupled cluster method is CC2 model, where as in CCSD,
the t1 amplitudes are treated to infinite order with the aim
of providing an approximate implicit description of orbital
relaxation, but some approximations are introduced in the
doubles part of CCSD equations. Moreover, in cases where
the correlation contributions are essential, the CC2 numbers
are much closer than CHF ones to the highly correlated
CCSD(T) results.[16,50] For the molecules considered in this
article, it is very likely that CC2 strongly overestimates corre-
lation effects, giving values that can exceed the CPH results
up to 50%.
This indicates the necessity of an improved treatment of
dynamical correlation using the CCSD method. CCSD is a rigor-
ous, correlated ab initio level of theory, which is, generally, a
good systematic level available for these high-order properties,
for which inclusion of linked triples would be extremely costly.
However, it must be realized that correlation may become more
important in chiral molecule with large p-electrons systems. In
these systems, an explicit treatment of triples excitation—at
least in a perturbative manner—could be fundamental.
Our best estimates for rð1ÞI
have been computed using
CCSD-QR. Comparing these results to the uncorrelated CHF
ones, it is observed that for C, N, and F atom correlation effects
increase the absolute value of the property from a minimum of
�19% (C1 atom) to a maximum of �30% (C2 atom). The oxygen
case behaves oppositely, as there is a decrease of �14% when
including electron correlation. At any instance, it is evident that
electron correlation effects are very important to get an
adequate description of the pseudo-scalar rð1ÞIin (R)-N-fluor-C-
methyl-oxa-aziridine.
The encountered concordance between the analytical CCQR
numbers and those from unrelaxed FF approach (see Table 3) in
the calculations of the shielding polarizability pseudo-scalar is
excellent for these systems, confirming previous results.[9,20,41,42]
Additionally, using an electric field with 0.001 au magnitude has
been proven to be adequate for FF procedure also in the study
of the rð1ÞI
pseudo-scalar. The good agreement indicates that
this approach is viable for the calculation of the DSP in systems
with larger size, where CCSD-QR calculations may be prohibitive.
Estimation of chiral-discrimination observables in oxaziridine
derivatives
DSP pseudo-scalar has been calculated for a set of derivatives
of the parent (R)-N-fluor-C-methyl-oxa-aziridine molecule by
systematically changing nitrogen, oxygen, and fluorine atoms
by other atoms of the same group in the periodic system
(CH3ACHAZAYAX, for X@F, Cl, and Br; Y@N and P; Z@O
and S). According to the results from the previous section, we
use a CCSD-QR approach and Sadlej’s TZV basis along the
whole set of calculations. The obtained results for the heavy
atoms (13C, 14N, 17O, 19F, 31P, 33S, 35Cl and 79Br) of the consid-
ered derivatives are collected in Table 4.
It can be observed in Table 3 that the computed values of
rð1ÞI
are very different for each atom in the parent molecule.
Thus, while for the asymmetric carbon the value of the
pseudo-scalar is only of 29 ppm, rð1ÞI
is duplicated for the
nitrogen bonded to it. Notably, the oxygen atom that com-
pletes the ring shows a pseudo-scalar magnitude that is
approximately seven times larger, to be precise 205 ppm. Con-
versely, the fluorine atom bonded to nitrogen present a mag-
nitude very similar to that of the asymmetric carbon. It is to
remark that the large value of rð1ÞI
that we have calculated for
the oxygen atom stands out against those found in similar
molecular systems (2R)-N-methyloxaziridine (40 ppm) or
(2R)22-methyloxirane (10 ppm).[4]
The change of heavy atom causes different consequences
on rð1ÞI
depending on the environment. In this way, the DSP
pseudo-scalars for nitrogen and phosphorus are analogous
when these atoms are bonded to oxygen, and both are very
sensitive to the change of either the chalcogen or the halo-
gen. We observe an enlargement of the pseudo-scalar when
going down in the group. Oppositely, oxygen and sulfur show
very dissimilar magnitudes of rð1ÞI, the largest values corre-
sponding to sulfur. For these two atoms, the change of the
neighbor nitrogen by phosphorus does not produce a signifi-
cant variation, and even a lesser influence has the substitution
of the fluorine atom, which is not directly joined to the refer-
ence atom. Finally, among the halogens, fluorine presents the
smaller magnitude of the DSP pseudo-scalar; additionally, fluo-
rine is also less sensitive to the substitution of nitrogen by
phosphorus. It is noteworthy, that the change of oxygen by
sulfur increases in a factor of �2.5 the value of rð1ÞF, even
though they are not directly bonded. The chlorine and bro-
mine atoms show the largest values of rð1ÞI
among the halo-
gens, both of them being also exceedingly affected by
changes in the environment in all the considered cases.
The highest values of rð1ÞI
that we have found are those
corresponding to sulfur and phosphorus in the compounds of
larger molecular weight, that is, CH3ACHASANBr and
CH3ACHASAPBr. For those cases, the value of the DSP
pseudo-scalar becomes as high as 800 ppm, a value sensibly
larger than most of the previously reported. We stress that the
presence of a sulfur nucleus significantly enhances the DSP
pseudo-scalar of the neighboring heavy atoms, especially in the
case of phosphorus in which it can become four times bigger.
Anyway, it is worthy to mention that larger values of the
Table 4. CCSD-QR(CTOCD-DZ) dipole-shielding polarizability pseudo-
scalar rð1ÞI
in ppm using the Sadlej-pVTZ basis set.
Molecule N/P O/S F/Cl/Br
CH3ACHAOANF 259.68 204.66 226.84
CH3ACHAOANCl 2171.10 224.71 227.23
CH3ACHAOANBr 2232.39 276.68 280.01
CH3ACHAOAPF 269.68 210.74 248.19
CH3ACHAOAPCl 2215.70 276.63 2156.12
CH3ACHAOAPBr 2282.02 340.94 2374.43
CH3ACHASANF 2184.33 623.01 290.65
CH3ACHASANCl 2392.93 724.12 2127.03
CH3ACHASANBr 2518.91 830.39 2266.65
CH3ACHASAPF 2287.40 547.15 2118.34
CH3ACHASAPCl 2612.04 694.81 2281.68
CH3ACHASAPBr 2771.10 802.56 2650.09
FULL PAPER WWW.C-CHEM.ORG
1820 Journal of Computational Chemistry 2014, 35, 1815–1823 WWW.CHEMISTRYVIEWS.COM
pseudo-scalar do not imply per se higher chance of measurabil-
ity, as (i) the larger the spin, the lower the accuracy of an NMR
measurement and (ii) the larger the natural abundance, the
stronger the intensity of the NMR bands. Therefore, we empha-
size that the best candidate for chiral discrimination by NMR is
the fluorine in CH3ACHASPF and CH3ACHASNF followed by
the phosphorous in CH3ACHASPCl and CH3ACHASPBr.
The presence of bromine in some of the molecules consid-
ered makes it opportune to look for a gross estimation of the
effect of relativity on the computed values of the DSP pseudo-
scalar. As relativistic calculations require very large basis specif-
ically designed for them, CCSD-QR calculations are too costly
to be carried out and, therefore, relativistic effects have been
approximately evaluated at the DFT-KT2 level of theory in a
finite differences approach using triple-zeta Dyall’s basis for
bromine and Sadlej’s elsewhere. To this end, we have use
DIRAC[51] program to calculate rð1ÞBr
in CH3ACHAONBr and
CH3ACHAOPBr and compared fully relativistic values of the
pseudo-scalar to those at the nonrelativistic limit. Such calcula-
tions indicate that relativity increases around 7% the nonrela-
tivistic result.
Having available accurate values of the CCSD shielding polariz-
ability pseudo-scalar, we make an estimation of the values that
could be experimentally measured and used as chiral discrimina-
tors. As stated before, two methodologies previously discussed
by Buckingham and coworkers are used to this end. First, we
consider the RF voltage generated in an appropriated experi-
mental setup. According to eq. (15), in addition to intrinsic prop-
erties of the considered nucleus and some universal constants,
voltage depends on the density of the liquid, the temperature,
the applied magnetic field, and the characteristics of the capaci-
tor. In Table 5, we report the generated RF voltage assuming a
magnetic field with flux density B 5 23 T, a pure liquid of density
1.0 g cm23, a capacitor with ðd=ðe21ÞÞ50:016 m, an isotopic
abundance of 100% and room temperature (T 5 300 K).
The largest RF voltage we have found is 76 nV and it is induced
by the 79Br nucleus in CH3ACHASAPBr. Relativistic effects are pre-
sumed to increase this magnitude, and therefore, 76 nV is the mini-
mum expected value. Note, nonetheless, that 79Br has a spin of
I 5 3/2, which can cause problems in the precision of the measure.
Somewhat smaller values, around 50 nV, are induced by 19F and by31P in CH3ACHASAYAF and CH3ACHASAPAX molecules. Notice
that both 19F and 31P have spin I 5 1/2 and 100% of abundance,
what in principle favors the experimental determination. It is also
worth to mention that all the “measurable favorable” compounds
contain sulfur in the three-member ring and a pair of them phos-
phorus too. As previously reported by Monaco and Zanasi,[32] fluo-
rine has proven to be a good candidate for detecting the discussed
RF voltage as its high value of g-factor provides values of that mag-
nitude above average. Nevertheless, the elevate values of the RF
voltage encountered for phosphorus are due to a large rð1ÞPin the
studied molecules.
The detection of the RF voltage depends on the experimen-
tal conditions, contrary to the determination of the NMR
chemical shift displacement in the presence of an external
static electric field. Table 6 collects the magnitude of the exter-
nal electric field required to induce a magnetic field perpen-
dicular to that of the NMR spectrophotometer and equivalent
to a difference of shifts of 0.01 ppm between the two enan-
tiomers. Anyway, this technique can be applicable only to
nuclei of low spin because the larger the spin, the lower the
precision of NMR measurement, due to band broadening aris-
ing for nuclear quadrupolar interaction.
In the studied molecules, the electric fields required to pro-
duce detectable chiral shifts have intensities of the order of
MV/m, which is maybe affordable but still too high for routine
measurements. At any rate, our results provide information on
the characteristics of compounds and nuclei more adequate
for a first experimental confirmation. Again, the “best”—that is,
the smallest—values are found for the nuclei of sulfur, phos-
phorus, bromine, and nitrogen in CH3ACHASANX and
CH3ACHASAPX. Even though the study of sulfur and bromine
could need signal-separation techniques for their high nuclear
spin, phosphorus and nitrogen should not face this difficulty.
Conclusions
NMR spectroscopy can distinguish diastereoisomers through a
pseudo-scalar obtained from the static DSP tensor, which has
Table 5. rf-voltage, V(I)(nV), in a capacitor with plates at 6d/2 on a direc-
tion perpendicular to the spectrometer’s magnetic field, induced by the
precessing nuclei of the species I, at 300 K, assuming density of 1 g
cm23 and 100% of isotopic abundance.[a]
Molecule N/P O/S F/Cl/Br
CH3ACHAOANF 0.6 30.7 16.6
CH3ACHAOANCl 1.4 27.9 0.8
CH3ACHAOANBr 1.3 23.3 11.5
CH3ACHAOAPF 6.5 25.9 24.4
CH3ACHAOAPCl 17.3 29.0 3.7
CH3ACHAOAPBr 16.2 25.6 48.0
CH3ACHASANF 1.5 10.6 46.4
CH3ACHASANCl 2.7 10.6 3.0
CH3ACHASANBr 2.5 8.6 34.4
CH3ACHASAPF 23.1 7.9 51.3
CH3ACHASAPCl 42.9 8.8 5.8
CH3ACHASAPBr 40.1 7.5 75.5
[a] Natural abundance: 1.108% (13C), 99.635% (14N), 0.037% (17O), 100%
(19F), 100% (31P), 0.750% (33S), 75.50% (35Cl), 50.60% (79Br), 49.40%
(80Br). All calculations were carried out at the CCSD-QR(CTOCD-DZ) level
with the Sadlej-pVTZ basis set.
Table 6. Electric field in 106 (Vm21) to be applied to yield a 0.01 ppm dif-
ference between NMR chemical shift of two enantiomers.
Molecule N/P O/S F/Cl/Br
CH3ACHAOANF 43.08 12.56 95.79
CH3ACHAOANCl 15.03 11.44 94.41
CH3ACHAOANBr 11.06 9.29 32.14
CH3ACHAOAPF 36.90 12.20 53.35
CH3ACHAOAPCl 11.92 9.29 16.47
CH3ACHAOAPBr 9.12 7.54 6.87
CH3ACHASANF 13.93 4.13 28.36
CH3ACHASANCl 6.54 3.55 20.24
CH3ACHASANBr 4.95 3.10 9.64
CH3ACHASAPF 8.95 4.70 21.73
CH3ACHASAPCl 4.20 3.70 9.13
CH3ACHASAPBr 3.33 3.20 3.95
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 1815–1823 1821
opposite sign in each enantiomer and is zero for achiral mole-
cules. A systematic study has been carried out for the first
time to investigate the ability of the CTOCD-DZ approximation
to determine the DSP pseudo-scalar.
Using the parent molecule (R)-N-fluor-C-methyl-oxa-aziridine
as reference, the dependence of the computed pseudo-scalar
on basis set, gauge origin, and electron correlation has been
extensively investigated. Furthermore, the CBS limit at the CHF
level of theory has been determined, showing that basis satu-
ration is achieved with Dunning’s aug-ccpV5Z basis set and
that a reasonable compromise between numerical perform-
ance and the number of basis function is obtained with
Sadlej’s TZ that yields results that are in satisfactory agreement
with aug-ccpV5Z and CBS basis. The gauge invariance of
CTOCD-DZ method was checked and confirmed.
The reported calculations show that correlation may cause
large changes in the property value—its inclusion being
required for an accurate picture of the DSP pseudo-scalar—
even though it does not change the order of magnitude of
the calculated property. It seems that CHF and CCS generally
underestimate DSP, while CC2 overestimate it, exaggerating
electron correlation effects in its typical way. Conversely, CCSD
appears as a rigorous and affordable method to theoretically
determine accurate DSP in medium-sized molecules where the
inclusion on linked triples excitation would be too expensive.
All the molecules considered in this study show large values
of the DSP pseudo-scalar in one or more nuclei. Moreover, in at
least half of the studied systems the RF voltage generated by
the rotating chiral electric polarization, induced by the perma-
nent magnetic dipole moment and the chiral chemical shift are
inside the experimentally detectable range, allowing for chiral
discrimination by NMR techniques. The largest computed DSP
pseudo-scalar corresponds in all the considered examples to
the sulfur atom, but this is not the case for the measurable chi-
ral effects. However, its presence in the ring substantially
increases the value of the pseudo-scalar on the nearby atoms
N, F, Br, and especially P, whose isotopic abundance and nuclear
spin make it very favorable for NMR spectroscopy. Summarizing,
among the studied molecular systems, we suggest the fluorine
in CH3ACHASPF and CH3ACHASNF as well as the phosphorous
in CH3ACHASPCl and CH3ACHASPBr as the best candidates
for experimental verification of NMR chiral discrimination.
Acknowledgment
The authors are grateful to Profs. P. Lazzeretti and A. S�anchez de
Mer�as for helpful discussions.
Keywords: shielding polarizability � CCSD � complete basis
set � chiral discrimination
How to cite this article: S. Pelloni, I. G. Cuesta. J. Comput.
Chem. 2014, 35, 1815–1823. DOI: 10.1002/jcc.23689
[1] A. D. Buckingham, Chem. Phys. Lett. 2004, 398, 1.
[2] A. D. Buckingham, P. Fischer, Chem. Phys. 2006, 324, 111.
[3] A. D. Buckingham, Lecture Series on Computer and Computational Sci-
ences, Vol. 6; Brill Academic Publisher: Leiden, 2006; pp. 1–5.
[4] R. Zanasi, S. Pelloni, P. Lazzeretti, J. Comput. Chem. 2007, 28, 2159.
[5] S. Pelloni, F. Fagioni, P. Lazzeretti, Rend. Fis. Acc. Lincei 2013, 24,
283.
[6] D. M. Bishop, S. M. Cybulski, J. Chem. Phys. 1990, 93, 590.
[7] B. J. Orr, J. F. Ward, Mol. Phys. 1971, 20, 513.
[8] D. M. Bishop, Rev. Mod. Phys. 1990, 62, 343.
[9] A. Rizzo, T. Helgaker, K. Ruud, A. Barszczewicz, M. Jaszu�nski,
P. J�rgensen, J. Chem. Phys. 1995, 102, 8953.
[10] S. M. Cybulski and D. M. Bishop, Mol. Phys. 1998, 93, 739.
[11] M. Grayson, Int. J. Mol. Sci. 2003, 4, 218.
[12] S. P. A. Sauer, W. T. Raynes, J. Chem. Phys. 2001, 115, 5994.
[13] H. Kjær, S. P. A. Sauer, J. Kongsted, J. Comput. Chem. 2011, 32, 3168.
[14] H. Kjær, M. R. Nielsen, G. I. Pagola, M. B. Ferraro, P. Lazzeretti, S. P. A.
Sauer, J. Comput. Chem. 2012, 33, 1845.
[15] A. Ligabue, S. P. A. Sauer, P. Lazzeretti, J. Chem. Phys. 2003, 118, 6830.
[16] I. G. Cuesta, J. S�anchez Mar�ın, A. M. S�anchez de Mer�as, F. Pawlowski,
P. Lazzeretti, Phys. Chem. Chem. Phys. 2010, 12, 6163.
[17] I. G. Cuesta, A. M. S�anchez de Mer�as, S. Pelloni, P. Lazzeretti, J. Comput.
Chem. 2009, 30, 551.
[18] J. Gauss, J. Stanton, J. Chem. Phys. 1996, 104, 2574.
[19] D. Sundholm, J. Gauss, A. Sch€afer, J. Chem. Phys. 1996, 105, 11051.
[20] A. Rizzo, J. Gauss, J. Chem. Phys. 2002, 116, 869.
[21] H. Kjær, S. P. A. Sauer, J. Kongsted, J. Chem. Phys. 2011, 134, 044514.
[22] P. Lazzeretti, M. Malagoli, R. Zanasi, Chem. Phys. Lett. 1994, 220, 299.
[23] F. A. Davis, J. Lamendola, Jr., U. Nadir, E. W. Kluger, T. C. Sedergran, T.
W. Panunto, R. Billmers, R. Jenkins, Jr., I. J. Turchi, W. H. Watson, J. S.
Chen, M. Kimura, J. Am. Chem. Soc. 1980, 102, 2000.
[24] F. A. Davis, M. C. Weismiller, C. K. Murphy, R. T. Reddy, B. C. Chen,
J. Org. Chem. 1992, 57, 7274.
[25] P. Lazzeretti, Handbook of Molecular Physics and Quantum Chemistry,
Vol. 3, Part 1, Chapter 3; S. Wilson, Ed.; Wiley: Chichester, 2003; pp.
53–145.
[26] S. Pelloni, P. Lazzeretti, R. Zanasi, J. Chem. Theory Comput. 2007, 3,
1691.
[27] P. Lazzeretti, A. Soncini, R. Zanasi, Theor. Chem. Acc. 2008, 119, 99.
[28] G. I. Pagola, M. B. Ferraro, S. Pelloni, P. Lazzeretti, S. P. A. Sauer, Theor.
Chem. Acc. 2011, 129, 359.
[29] J. Olsen, P. J�rgensen, J. Chem. Phys. 1985, 82, 3235.
[30] S. Coriani, M. Pecul, A. Rizzo, P. Jorgensen, M. Jaszu�nski, J. Chem. Phys.
2002, 117, 6417.
[31] P. Lazzeretti, Adv. Chem. Phys. 1987, 75, 507.
[32] G. Monaco, R. Zanasi, Chirality 2011, 23, 752.
[33] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb,
J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery, R. E. Stratmann,
J. C. Burant, S. Dapprich, J. M. Millam, A. D. Daniels, K. N. Kudin, M. C.
Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R. Cammi, B. Mennucci,
C. Pomelli, C. Adamo, S. Clifford, J. Ochterski, G. A. Petersson, P. Y.
Ayala, Q. Cui, K. Morokuma, D. K. Malick, A. D. Rabuck,
K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V. Ortiz, A. G. Baboul,
B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi,
R. Gomperts, R. L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y.
Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. G. Johnson,
W. Chen, M. W. Wong, J. L. Andres, C. Gonzalez, M. Head-Gordon,
E. S. Replogle, J. A. Pople, Gaussian 2003, Revision B.05; Gaussian:
Pittsburgh, PA, 2003.
[34] DALTON2013, A molecular electronic structure program, 2011. Avail-
able at: http://www.daltonprogram.org.
[35] T. H. Dunning, Jr., J. Chem. Phys. 1989, 90, 1007.
[36] A. J. Sadlej, Theor. Chim. Acta 1991, 79, 123.
[37] A. J. Sadlej, Theor. Chim. Acta 1991, 81, 45.
[38] A. J. Sadlej, Theor. Chim. Acta 1992, 81, 339.
[39] A. J. Sadlej, M. Urban, J. Mol. Struct. (THEOCHEM) 1991, 234, 147.
[40] A. J. Sadlej, Collect. Czech. Chem. Commun. 1988, 53, 1995.
[41] S. Coriani, A. Rizzo, K. Ruud, T. Helgaker, Mol. Phys. 1996, 88, 931.
[42] S. Coriani, A. Rizzo, K. Ruud, T. Helgaker, Chem. Phys. 1997, 216, 53.
[43] R. Zanasi, P. Lazzeretti, Chem. Phys. Lett. 1998, 286, 240.
[44] (a) D. Feller, J. Chem. Phys. 1992, 96, 6104; (b) D. Feller, J. Chem. Phys.
1993, 98, 7059.
FULL PAPER WWW.C-CHEM.ORG
1822 Journal of Computational Chemistry 2014, 35, 1815–1823 WWW.CHEMISTRYVIEWS.COM
[45] T. Helgaker, W. Klopper, H. Koch, J. Noga, J. Chem. Phys. 1997, 106,
9639.
[46] A. Halkier, W. Klopper, T. Helgaker, P. J�rgensen, J. Chem. Phys. 1999,
111, 4424.
[47] (a) O. Christiansen, C. H€attig, J. Gauss, J. Chem. Phys, 1998, 109, 4745; (b)
K. Hald, P. J�rgensen, J. Olsen, M. Jaszu�nski, J. Chem. Phys. 2001, 115, 671.
[48] H. Koch, O. Christiansen, P. J�rgensen, A. M. S�anchez de Mer�as,
T. Helgaker, J. Chem. Phys. 1997, 106, 1808.
[49] (a) A. Rizzo, M. Kallay, J. Gauss, F. Pawlowski, P. J�rgensen, C. H€attig,
J. Chem. Phys. 2004, 121, 9461; (b) H. Larsen, J. Olsen, C. H€attig,
P. J�rgensen, O. Christiansen, J. Gauss, J. Chem. Phys. 1999, 111, 1917.
[50] O. Christiansen, J. Gauss, J. Stanton, Chem. Phys. Lett. 1997, 266, 53.
[51] DIRAC, A relativistic ab initio electronic structure program, Release
DIRAC13 (2013), written by L. Visscher, H. J. Aa. Jensen, R. Bast, T.
Saue, with contributions from V. Bakken, K. G. Dyall, S. Dubillard, U.
Ekstr€om, E. Eliav, T. Enevoldsen, E. Faßhauer, T. Fleig, O. Fossgaard, A.
S. P. Gomes, T. Helgaker, J. K. Lærdahl, Y. S. Lee, J. Henriksson, M. Ilia�s,
Ch. R. Jacob, S. Knecht, S. Komorovsk�y, O. Kullie, C. V. Larsen, H. S.
Nataraj, P. Norman, G. Olejniczak, J. Olsen, Y. C. Park, J. K. Pedersen, M.
Pernpointner, K. Ruud, P. Sałek, B. Schimmelpfennig, J. Sikkema, A. J.
Thorvaldsen, J. Thyssen, J. van Stralen, S. Villaume, O. Visser, T.
Winther, S. Yamamoto. Available at: http://www.diracprogram.org.
Received: 28 February 2014Revised: 23 June 2014Accepted: 7 July 2014Published online on 22 July 2014
FULL PAPERWWW.C-CHEM.ORG
Journal of Computational Chemistry 2014, 35, 1815–1823 1823