THE JOURNAL OF CHEMICAL PHYSICS 130, 084102 共2009兲

Relativistic two-component geometric approximation of the electron-

positron contribution to magnetic properties in terms
of Breit–Pauli spinors
Daniel Zaccari,1 Juan I. Melo,2 Martín C. Ruiz de Azúa,2,a兲 and Claudia G. Giribet2
1
Departamento de Física, Facultad de Ciencias Exactas, Fisicoquímicas y Naturales, Universidad Nacional
de Río Cuarto, Ruta 36, km 601, 5800 Río Cuarto, Córdoba, Argentina
2
Departamento de Física, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Cdad.
Universitaria, Pab. I, 1428 Buenos Aires, Argentina and CONICET
共Received 7 October 2008; accepted 10 December 2008; published online 24 February 2009兲

An alternative approach for the calculation of the electron-positron 共e-p兲 contribution to magnetic
properties based on two-component Breit–Pauli spinors is presented. In it, the elimination of the
small component scheme is applied to the inverse propagator matrix of e-p pairs. The effect of the
positronic manifold is expressed as an operator acting on Breit–Pauli spinors. The operator form
thus obtained sums up the relativistic correction as a geometric series and as a result a totally
different behavior in the vicinity of a nucleus is obtained as compared to the one of the linear
response approximation. This feature has deep influence in numerical values of the e-p contribution
to the nuclear magnetic shielding of heavy atoms. Numerical calculations carried out for Kr, Xe, and
I show that with this approach, the e-p contributions to this property are in good agreement with
those of four-component methods. © 2009 American Institute of Physics.
关DOI: 10.1063/1.3063639兴

I. INTRODUCTION recent years.27–30 Approaches based on the Breit–Pauli ap-

Relativistic effects have proven to be of major impor-
tance on molecular magnetic properties of heavy atom con-
taining compounds, especially for those involving the field of
a magnetic nucleus, such as NMR parameters.1,2 Significant
advances have been achieved in recent years in the theoreti-
cal description of such effects, both within Dirac four-
component spinors and quasirelativistic two-component
approaches.3–20,27–29 The question of the diamagnetic-
paramagnetic separation of terms at the four-component
level has also been a subject of discussion in several papers
in the field.4,17–20 In the particular case of the NMR nuclear
magnetic shielding tensor, numerical results in model com-
pounds within four-component approaches stand as bench-
mark values to decide the accuracy of computationally less
expensive two-component approximations.
However, beyond the superiority of fully relativistic
four-component approaches from a theoretical point of view,
two-component approaches have shown to adequately de-
scribe qualitatively or even quantitatively different aspects of
relativistic effects. In fact, quasirelativistic theories are
nowadays available which are exactly equivalent to a four-
component theory and very promising results were obtained
in this context.21–26 Other two-component approaches are
also worthy to mention as they have shown to be very suc-
cessful in practical applications. For instance, the zeroth or-
der relativistic approximation 共ZORA兲 has been widely ap-
plied for the study of both nuclear magnetic shieldings and
spin spin couplings.10–12 Approaches based on the Douglas–
Kroll decomposition have received increasing attention in
a兲
Electronic mail: azua@df.uba.ar.
proximation are also in widespread use and produce particu-
larly chemical shift results that are very competitive com-
pared to other approaches available for both the light and
heavy nuclei of a molecule.2,6–9,31–36
In the linear response within the elimination of the small
component 共LRESC兲 approach,6 the elimination of the small
component 共ESC兲 reduction is applied directly to a four-
component Rayleigh–Schrödinger perturbation theory
共RSPT兲 expression of magnetic properties. In this sense, it is
wholly consistent with the most direct relativistic theory of
magnetic properties, as it is explicitly related to a perturba-
tional treatment of the four-component magnetic interaction
given by the operator W = ␣ · A in Dirac’s equation. The
equivalence of the total result of the usual Breit–Pauli and
LRESC schemes was theoretically proven, although the par-
ticular operators involved are different.37 The LRESC or
Breit–Pauli schemes are interesting because they present sev-
eral advantages compared to other approaches. First, from
the physical point of view, they are based on the Schrödinger
Hamiltonian and perturbation operators are amenable to
physical interpretation in terms of usual nonrelativistic spin
orbitals. Second, from the computational point of view, the
problems of generating the small component of spinors and
taking account of negative energy solutions are avoided, with
great savings of computational resources. Such approaches
are in widespread use currently and produce particularly
chemical shift results that are very competitive as compared
to other approaches available.31–33 It is therefore of great
interest to establish theoretically and numerically the reliabil-
ity of results of such schemes for elements of different rows
of the Periodic Table.

0021-9606/2009/130共8兲/084102/9/$25.00 130, 084102-1 © 2009 American Institute of Physics

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
084102-2 Zaccari et al.
One particular feature of the LRESC scheme is that
proper track of contributions coming from electronic excita-
tions and negative energy 共i.e., positronic兲 states can be kept
in the whole derivation in a way fully consistent with the
usual four-component random phase approximation 共RPA兲
scheme.3,4 In numerical applications, large negative relativ-
istic effects on the positronic contribution to the nuclear
magnetic shielding of a heavy element are found.7,43 Al-
though this result is in line with the corresponding one in
four-component calculations,4 it seems to be somewhat over-
estimated. A separation into atomic orbital contributions
shows that such effect is largely dominated by contributions
from inner shell s atomic orbitals.43 This fact suggests that
such contributions are overestimated and it might be inter-
esting to develop an alternative formulation which could cor-
rect this problem.
This is the aim of the present work. We present an alter-
native approach for considering the negative energy state
共i.e., positronic兲 contributions to magnetic properties within
the LRESC approach. We apply the ESC approximation to
the inverse electron-positron 共e-p兲 propagator matrix. How-
ever, we formally eliminate explicit evaluation of matrix el-
ements involving positronic states and therefore we obtain a
final expression which is evaluated in terms of Breit–Pauli
spinors.38 The propagator thus obtained adds up the relativ-
istic correction as a geometric series. Although this improve-
ment will have little effect on the calculated chemical shifts,
which are very well reproduced at this theoretical level, we
believe that it is of fundamental physical interest to introduce
this improvement in the LRESC scheme, as the total absolute
shielding results are largely affected. Numerical results are
presented for selected model compounds. A very good agree-
ment with four-component RPA results is recovered follow-
ing such scheme. It is explicitly shown that linearization of
the propagator yields precisely the LRESC expression of
previous work.
II. THEORY
The LRESC scheme for the calculation of magnetic
properties which are bilinear in the magnetic vector potential
is based on a second order RSPT expression in Dirac–Fock
space:6
具0兩␣ · A兩n典具n兩␣ · A兩0典
E共2兲 = 兺 E0 − En
n⫽0
具vac兩␣ · A兩n典具n兩␣ · A兩vac典
− 兺
, 共1兲
n⫽vac Evac − En
where the last term accounts for the correction to the
“vacuum” energy when adopting the quantum electrodynam-
ics 共QED兲 picture in which negative energy states are rein-
terpreted as positive energy states for positrons.39 In this
work, atomic units based on the Gaussian system of units for
the electromagnetic field are adopted. Therefore, the mag-
netic susceptibility of vacuum and other factors are taken
into account by a factor of c−1 accompanying the magnetic
vector potential A. Therefore, the mechanical momentum
takes the form
Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
J. Chem. Phys. 130, 084102 共2009兲
1
␲ = p + A. 共2兲
c
Such c factor should not be confused with a relativistic cor-
rection factor.
In Eq. 共1兲, states 兩0典, 兩n典 stand for eigenstates of the
Dirac–Coulomb–Breit Hamiltonian in Dirac–Fock space. As
a consequence, such states correspond to states with a well
defined charge Q = −eN, where N is the number of electrons
of the system in the nonrelativistic limit, but they do not
have a fixed number of particles.40 The consideration of such
Hamiltonian within the subspace of fixed N particle states
constitutes the no pair approximation.
In the presence of the magnetic perturbation W = ␣ · A,
different particle number manifolds are connected even if the
Coulomb–Breit interaction between particles is neglected.
The aim of the present work is to analyze in detail the pair
creation contribution to magnetic properties, which yields
the diamagnetic term in the nonrelativistic limit.41 Particle
states used to span the Dirac–Fock space are the solutions of
a one-body Hamiltonian. This Hamiltonian can be taken as
the Dirac Hamiltonian for a particle in the field of the nuclei
in the molecule or alternatively as that corresponding to the
Dirac–Hartree–Fock approximation.42 In the last case the in-
teraction between particles is included in the mean-field ap-
proximation, and the 共magnetically兲 nonperturbed Hamil-
tonian connects different particle number manifolds only
through the “fluctuation” potential of the interaction. How-
ever, such terms will be neglected in the present work. In this
approximation state 兩0典 is an N particle state, and it may be
connected to states 兩n典 of different particle numbers only by
the magnetic interaction.
Therefore, the RSPT expression of Eq. 共1兲 can be sepa-
rated into two terms according to the particle number of the
excited states involved:6
E2 = Ea + Eb , 共3兲
where
具0N兩␣ · A兩nN典具nN兩␣ · A兩0N典
Ea = 兺
n ⫽0 E0,N − EnN
, 共4兲
N N
具0N兩␣ · A兩nN+2典具nN+2兩␣ · A兩0N典
Eb = 兺 E0,N − En,N+2
nN+2
具vac兩␣ · A兩n2典具n2兩␣ · A兩vac典
−兺 . 共5兲
Evac − En,2
n=2
The vacuum correction is included in Eb, and in this
term the number of particles of the “excited” states is N = 2.
If an e-p pair is created by the magnetic interaction, the same
positron state must be destroyed to give a state overlapping
with the ground state 兩0典, but a different electronic state may
be involved. As a result, it may be written as
Eb = 兺 具e兩␣ · A兩p典Pep,e⬘p具p兩␣ · A兩e⬘典. 共6兲
e,e⬘occ,p
The sum runs over occupied electronic states only. P stands
for the propagator of pair creation excitations of the system.
084102-3 Geometric ESC approach of magnetic properties J. Chem. Phys. 130, 084102 共2009兲

Taking into account that the Hamiltonian under consideration
is a one-body Hamiltonian, the propagator has a straightfor-
ward expression in terms of the corresponding eigenstates
and eigenvalues:
Pep,e⬘p⬘ = ␦e,e⬘␦ p,p⬘共Ee + E p兲.
−1
In this case, Eb can be re-expressed as
Eb = 兺 兺 具e兩␣ · A兩p典Pep,ep具p兩␣ · A兩e典.
eocc p
This is our starting point to obtain the nonrelativistic
limit and the leading order relativistic correction coming
from these terms. The Dirac 共or Dirac–Fock兲 Hamiltonian
for electrons and positrons can be expressed as
共HD − mc2兲␺e = Ee⬘␺e ,
− 共HD + mc2兲␺ p = E⬘p␺ p .
The leading order in the energy E is given by the rest mass of
the electron and positron, and the difference, represented by
E⬘, which depends strictly on the states involved, will be
considered as a perturbation. As a consequence the propaga-
tor can be separated accordingly:
−1 共0兲−1 共1兲−1
Pep,e⬘p⬘ = Pep,e⬘p⬘ + Pep,e⬘p⬘ ,
共0兲−1
Pep,e⬘p⬘ = 2mc ␦e,e⬘␦ p,p⬘ ,
2
present work we intend to develop an approximation of order
1 / c2 to the inverse polarization propagator 共PP兲 matrix,
keeping it in the denominator of Eq. 共15兲. This procedure
sums the relativistic effects as a geometric series of the first
order correction, in close resemblance to the RPA approxi-
共7兲
mation. Therefore, it is expected to bring calculated results
closer to four-component RPA ones.
In what follows a fixed electronic state is assumed and
共8兲 the whole positronic contribution is added up. To this end we
consider matrix 共1 + M共e兲兲, Eq. 共17兲. Consistently to order
c−2 the eigenvalue E⬘p can be replaced by the nonrelativistic
limit of the Dirac equation for positrons. We start from
− 共HD + mc21兲␺ p = E⬘p␺ p . 共18兲
共9兲 Separating into upper 共U兲 and lower 共L兲 components, the
following is obtained:
共10兲
− 共共2mc2 + V兲␺Up + c共␴ · p兲␺Lp 兲 = E⬘p␺Up , 共19兲
− 共c共␴ · p兲␺Up + V␺Lp 兲 = E⬘p␺Lp . 共20兲
As it is seen, in this case ␺Up plays the role of the small
component and ␺Lp is the large one. Following the standard
ESC scheme the following is obtained at the lowest order:
共11兲
␴·p L
␺Up = − ␺ , 共21兲
共12兲 2mc p

共1兲−1 共0兲−1 and the following equation is obtained for ␺Lp :

Pep,e⬘p⬘ = ␦e,e⬘␦ p,p⬘共Ee⬘ + E⬘p兲 = Pep,e⬙p⬙ . M e⬙p⬙,e⬘p⬘ , 共13兲

E⬘ + E⬘
M ep,e⬘p⬘ = ␦e,e⬘␦ p,p⬘ e 2 p . 共14兲
冉 p2
2m
冊
− V ␺Lp = E⬘p␺Lp , 共22兲

2mc
which has the form of the Schrödinger equation for a particle
The definition of matrix M in Eq. 共14兲 allows one to write in a repulsive potential. This is the only component that re-
the propagator as mains at the order c0. As a consequence, consistently to the
P = 共1 + M兲−1P共0兲 . 共15兲 order c−2,

Taking into account electronic states on one hand and 1

M共e兲 p,p⬘ = 共E⬘ · 1 + H+兲 p,p⬘ , 共23兲
positronic states on the other, it is possible to write 2mc2 e
共1 + M兲ep,e⬘p⬘ = ␦e,e⬘共1 + M共e兲兲 p,p⬘ , 共16兲 where H+ stands for the operator in Eq. 共22兲. In order to
i.e., for each electronic state a corresponding matrix involv- obtain an explicit expression of this operator in four-
ing only positronic states can be defined. This matrix is dif- component Dirac space, it is useful to introduce the projec-
ferent for each electronic state. This is clear as matrix M共e兲 tors onto electronic and positronic state spaces and a unitary
contains the eigenvalue of the electronic state under consid- operator that expresses these projectors as diagonal
eration. The usefulness of such separation is given by the operators.
fact that the inverse matrix can be factorized in the same The projectors onto the subspace of electronic states 共Pe兲
way, i.e., and of positronic states 共P p兲 consistent with the ESC scheme
up to order c−2 can be expressed as38
共1 + M兲ep,e⬘p⬘ = ␦e,e⬘共1 + M共e兲兲 p,p⬘ .
冋 册
−1 −1
共17兲
1 − x2 x
It is readily seen that the lowest possible approximation Pe = 共24兲
x x2
to the propagator P of Eq. 共15兲 is of order 1 / c2, and only P共0兲
is retained in this case. This yields the nonrelativistic dia- and

冋 册
magnetic term of magnetic properties. Expansion of P in
terms of 1 / c requires the linearization of the inverse matrix x2 −x
as M is of order 1 / c2, and the proper definitions of the elec- Pp = , 共25兲
− x 1 − x2
tronic and positronic manifolds in the one-particle state
space. This was the way followed in previous work.6 In the where

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
084102-4 Zaccari et al. J. Chem. Phys. 130, 084102 共2009兲

␴·p Positronic states have nonzero amplitude only in the last

x= . 共26兲 two components of the corresponding four-component
2mc
spinor. This being the case, the projection of matrix 1
Neglecting terms of order higher than c−2 the transformation + M共e兲 关Eq. 共17兲兴 onto the positronic manifold has the form

冤 冥
x2

冤 冥
1− −x 0 0
2
U= 共27兲 1 + M共e兲 = 1 . 共31兲
x2 0 1+ 共E⬘1 + H+兲
x 1− 2mc2 e
2
The corresponding inverse is defined only for the lower 2
is unitary and has the properties ⫻ 2 block, as needed. It is readily obtained as

冋 册
1 0

冤 冥
U U† = Pe , 共28兲 0 0

冉 冊
0 0
共1 + M共e兲兲−1 = 1 −1
. 共32兲

冋 册
0 1+ 共E⬘1 + H+兲
0 0 2mc2 e
U U† = P p , 共29兲
0 1 As mentioned above, in this representation electronic
i.e., this transformation yields a representation such that elec- states are described by the corresponding Pauli spinor. In
tronic states have nonzero amplitude only in the first two order to evaluate the matrix element of Eq. 共8兲 the magnetic
components of the corresponding 4-spinor. In fact, this am- interaction operator must be transformed to this representa-
tion:
plitude is described by the Pauli two-component spinor ␾ ˜ e:

冋册 冋册
␺Le
␺Se
=U·
␾
˜e
0
. 共30兲
共␣ · A兲T = U† · ␣ · A · U.

The following is obtained:

共33兲

共␣ · A兲T = 冋 − 共x共␴ · A兲R + R共␴ · A兲x兲 R共␴ · A兲R − x共␴ · A兲x

R共␴ · A兲R − x共␴ · A兲x R共␴ · A兲x + x共␴ · A兲R
册 , 共34兲

where Eb = Eb,NR + Eb,A + Eb,P , 共39兲

x2 where
R=1− . 共35兲
2 1
Eb,NR = 兺 具␾e兩A2兩␾e典,
2mc2 eocc
共40兲
For the sake of brevity we define

a1 = − 共x共␴ · A兲R + R共␴ · A兲x兲, 1

共36兲
Eb,A = 兺 具␾e兩a22 − A2兩␾e典
2mc2 eocc
a2 = R共␴ · A兲R − x共␴ · A兲x,
1
+ E共2兲共Hmv + HDw ;A2兲, 共41兲
and thus 2mc2

共␣ · A兲T = 冋 a1
a2
a2
− a1
. 册 共37兲 Eb,P =
1
兺 具␾e兩A
2mc2 eocc
冉冉 1+
1
共E⬘1 + H+兲
2mc2 e
冊 −1

This result is now inserted in Eq. 共8兲, where the elec-

tronic state is represented by the corresponding Pauli spinor,
denoted by ␾ ˜ e:
冊
− 1 A兩␾e典, 共42兲

where 兩␾e典 are the Schrödinger spin orbitals; E共2兲共A , B兲

Eb =
eocc 2mc
1
兺 2 具␾˜ e兩a2 冉 1+
1
共E⬘1 + H+兲
2mc2 e
冊 −1
a 2兩 ␾
˜ e典. stands for a second order RSPT expression for operators A,
B:
共38兲 具0兩A兩n典具n兩B兩0典 + 具0兩B兩n典具n兩A兩0典
E共2兲共A,B兲 = 兺
n⫽0 E0 − En
, 共43兲
In order to separate explicitly the term containing the inverse
matrix of the propagator, Eb can be splitted as and the mass-velocity and Darwin operators are given by38

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
084102-5 Geometric ESC approach of magnetic properties J. Chem. Phys. 130, 084102 共2009兲

p4 具␾e兩␴ · A共Ee⬘1 + H+兲␴ · A兩␾e典, 共48兲

H mv = − , 共44兲
8m3c2 where Ee⬘ in the nonrelativistic limit is the Schrödinger ei-
genvalue for the electronic state under consideration. Ee⬘ and
ⵜ 2V C ␴ · A can be commuted to write
HDw = . 共45兲
4m2c2
具␾e兩␴ · A共Ee⬘1 + H+兲␴ · A兩␾e典
In Eqs. 共40兲–共42兲 a singlet ground state is assumed and there-
fore triplet operators are excluded from the final = 具␾e兩␴ · A共␴ · AH− + H+␴ · A兲兩␾e典, 共49兲
expressions. where H− stands for the Schrödinger Hamiltonian for the
Eb,A gathers relativistic effects originated in the c−2 ex- electronic state. Explicitly,

冉 冊冉 冊
pansion of the magnetic operators a1 and a2, Eq. 共36兲, and on
the molecular ground state 共second term of Eq. 共41兲. These p2 p2
␴ · AH− + H+␴ · A = ␴ · A +V + −V ␴·A
corrections keep the form of previous work. Eb,P, on the 2m 2m
other hand, contains relativistic corrections strictly originated
in the e-p propagator, which are added up as a geometric
series in the present work.
再
= ␴ · A,
p2
2m
冎 共50兲

The expression of Eb, Eq. 共38兲, is consistent to the order
−2
c . The effect of the positronic manifold is wholly taken
into account in this expression by means of operator H+. This
procedure sums the linear correction of previous work as a
geometric series and therefore it will be referred to as the
geometric ESC approximation 共GESC兲. The lowest order ap-
proximation is obtained by setting R = 1, neglecting the term
of order c−2 in the denominator, and replacing the Pauli
spinors by the Schrödinger ones. The nonrelativistic diamag-
netic term of magnetic properties is thus obtained. In the
relativistic result, it is seen that a different operator is defined
for each electronic state. However, this dependence is only
through the corresponding orbital energy. It can be taken out
from the operator either by neglecting its value, as in the
ZORA,10–12 or by linearizing its contribution, leaving a
unique operator for the positronic contribution for all elec-
tronic states. The full linearization of the operator yields the
diamagnetic term of previous work,6 as it will be explicitly
shown in Sec. II A. It is worthy to note that from the com-
putational point of view, in order to evaluate the diamagnetic
term of Eq. 共42兲, the matrix representation of p2 / 2m and V
are required, which are readily included as part of any
quantum chemistry program. Therefore, there is nearly no
need of extra computational effort to carry out numerical
calculations.
A. Linearization and LRESC expression
of the diamagnetic contribution
The leading order of the relativistic correction to the
diamagnetic contribution is now obtained by linearizing the
propagator Eq. 共38兲 to obtain
since V and ␴ · A commute. 兵 ,其 stands for the anticommuta-
tor. The explicit dependence on the energy eigenvalue Ee⬘ has
been eliminated and the whole contribution can be expressed
as an expectation value. The full result is
1 1
Eb = 兺 具␾e兩A2兩␾e典 + 2mc2 E共2兲共Hmv + HDw ;A2兲
2mc2 e
1
+ 兺 具␾e兩a22 − A2兩␾e典 − 具␾e兩A2x2
2mc2 e
+ ␴ · Ax2␴ · A兩␾e典. 共51兲
−2
Consistently to the order c and taking into account that
x2 and A2 can be commuted in the expectation value, the last
two term contributions can be rewritten as the expectation
value for the operator:
1
Odiam,LR = − 共兵A2,x2其 + 2␴ · Ax2␴ · A + 共x␴ · A兲2
2mc2
+ 共␴ · Ax兲2兲, 共52兲
which is exactly the expression found within the LRESC
approximation of previous work for the relativistic correction
of the diamagnetic contribution to magnetic properties.6,37
B. The electron-positron contribution to the nuclear
magnetic shielding tensor
Relativistic effects gathered in Eb,A are equivalent to
those calculated in previous work. The corresponding ex-
pressions for the specific case of the nuclear magnetic shield-
ing tensor are considered in this section. The nuclear mag-
netic shielding tensor of nucleus M can be obtained as

Eb = 兺
e
1 ˜
2mc2
冉
具␾e兩a2 1 −
1
2mc2
冊
共Ee1 + H+兲 a2兩␾
˜ e典. 共46兲 ␴共M兲i,j =
⳵ 2E
␮ =0,B=0 ,
⳵ ␮M,i ⳵ B j M
共53兲

Due to the factor 1 / 共2mc2兲2, the last term must be evaluated where E is the molecular energy in the presence of the mag-
for the nonrelativistic 共i.e., Schrödinger兲 electronic state ␾e netic fields of the nucleus and the uniform external field B0.
and the nonrelativistic approximation for a2: For a point dipole nucleus M:

a共0兲 ␮M ⫻ r M
2 = ␴ · A. 共47兲 AM = , 共54兲
r3M
Therefore, the expectation value for each electronic state in
the last term has the form and for the external magnetic field B0:

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
084102-6 Zaccari et al. J. Chem. Phys. 130, 084102 共2009兲

AB = 21 B0 ⫻ r. 共55兲 operator H+ were also defined on the basis of the Fock op-
erator F in the following way:
In what follows, the gauge origin is placed at the nucleus
position, i.e., r M = r. The explicit derivation of the final form H+ = T − VF = 2T − F, 共60兲
of the operators involved in Eb,A and LRESC Eb,P was car- where T is the kinetic energy operator and VF stands for the
ried out in Ref. 6. A brief account is presented in the Appen- Hartree–Fock potential. The corresponding matrix elements
dix. The following is obtained: are obtained from the SYSMO program files. Once the matrix
1 20␲ representation of the inverse propagator of Eq. 共59兲 is ob-
Eb,A = − 兺 具␾e兩 3 共B0 · ␮M兲␦共rM兲
8m3c4 eocc
tained, the calculation of Eb,P is carried out by inserting the
matrix representations of the nuclear magnetic potential op-

+4 冉 ␮M · L
r3
冊
共B0 · L兲 + B0 · B M,dip兩␾e典
erator and the uniform magnetic field potential operator,
which are included in the SYSMO program. For comparison
purposes, the linearized LRESC propagator is obtained with
1 共2兲 mv the same procedure, i.e., as the matrix products of the linear-
+ E 共H + HDw ;A M · AB兲, 共56兲
mc2 ized propagator and magnetic field operators. The calculated
values can be splitted into individual MO contributions, al-
Eb,P,LRESC = −
1
冉
8␲
兺 具␾e兩 3 共B0 · ␮M兲␦共rM兲
8m3c4 eocc
lowing interesting insight into the obtained values. The non-
relativistic diamagnetic contribution is also obtained with the
nonrelativistic propagator P共0兲, Eq. 共13兲, and the same mag-
冊
+ + 4共AM · AB兲p2 + B0 · BM,dip 兩␾e典, 共57兲
netic field operators. It is worthy to observe that both at the
relativistic and nonrelativistic levels, it is necessary to en-
large the atomic basis set with higher angular momentum
where basis functions in order to obtain appropriate matrix repre-
3共␮M · rM兲rM − ␮Mr2M sentations of the vector operators involved.
BM,dip = 共58兲
r5M
is the dipole component of the nucleus magnetic field. IV. RESULTS AND DISCUSSION
Within the formalism developed in the present work, the ex- In order to carry out numerical calculations the noble gas
pression of Eb,P is given by atoms Kr and Xe and the hydrogen halide IH were taken as
Eb,P =
1 ␮ ⫻r
兺 具␾e兩 Mr3
2c2 eocc
冉 冊 model compounds. The main reason for this choice is the
existence of a large amount of benchmark numerical results

冉冉 冊 冊冉 冊
−1
1 B0
⫻ 1+ 共E⬘1 + H+兲 −1 ⫻ r 兩␾e典.
2mc2 e 2 Kr and Xe and aug-ccpVTZ basis47–49 for I, enlarged in dif-
共59兲
The corresponding contributions to the magnetic shield-
ing tensor ␴M obtained according to Eq. 共53兲 will be referred
to as ␴b,A and ␴b,P, respectively.
III. METHOD OF CALCULATION
In order to carry out numerical calculations of the e-p
contribution to the nuclear magnetic shielding tensor with
the formalism of the previous section the following proce-
dure was adopted. Contributions contained in Eb,A involve
operators for which explicit numerical results were obtained
in previous works,7–9,43 both with the DALTON 共Ref. 44兲 and
SYSMO 共Ref. 45兲 programs. In order to calculate the newly
developed term Eb,P a computational program was imple-
mented which works as a subroutine of the SYSMO program.
The matrix representation of the inverse propagator of Eq.
共59兲 is carried out in the molecular orbital 共MO兲 basis set of
a Hartree–Fock calculation. The eigenvalue of the electronic
state E⬘ of each term of Eq. 共59兲 is replaced by the corre-
sponding orbital energy of the involved MO. This is consis-
tent with taking the Fock operator as the one-body Hamil-
tonian of the problem. Consistently, the matrix elements of
for these systems in the current bibliography. The basis sets
used in the calculations are based on Faegri’s basis sets46 for
ferent ways until saturation of results was achieved. On one
hand, tight functions were added in order to have flexibility
in the vicinity of the heavy nucleus. On the other, higher
angular momentum basis functions were added in order to be
able to reproduce the action of the vector operators involved
in Eq. 共59兲. In the latter case, it was verified that the nonrel-
ativistic value obtained with Eq. 共59兲 in the limit c → ⬁ is
coincident with the result of the diamagnetic operator. The
final results make use of uncontracted basis sets of quality:
共17s24p18d8f兲 for Kr, 共22s21p15d15f兲 for Xe,
共23s18p13d13f兲 for I, and 共6s4p3d兲 for H.
Results for the noble gas atoms Kr and Xe are presented
in Tables I and II, respectively. Values obtained with the
formalism developed in the present work are indicated as
“GESC” to indicate that the geometric approximation of the
propagator is used. Values of the LRESC approach are also
quoted for comparison. Contributions from individual con-
verged occupied orbitals are listed. As it was shown in pre-
vious work,50 the total relativistic effect is defined to a large
extent by contributions from the inner shell s-type orbitals. It
is seen that for these contributions the GESC and linearized
LRESC values differ significantly. As a consequence, contri-
bution ␴b,P to the relativistic effect on e-p rotations is re-
duced in absolute value by a factor of nearly 2 in both Kr and
Xe. The difference between both approaches becomes less

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
084102-7 Geometric ESC approach of magnetic properties J. Chem. Phys. 130, 084102 共2009兲

TABLE I. Atomic orbital contributions to relativistic effects on the nuclear and previous Breit–Pauli and LRESC two-component results
magnetic shielding constant of Kr for the isolated atom arising from e-p of the bibliography. In order to obtain the full e-p values of
rotations given by Eq. 共59兲. LRESC, GESC, and difference between both.
Values in ppm.
Table III, the contributions arising from Eb,A, Eq. 共56兲, must
be computed. As it is shown in the Appendix, the corre-
␴b,P sponding values can be obtained by adequate rescaling of
numerical data taken from Refs. 7 and 43 or Refs. 8 and 9. It
AO NR GESC LRESC Difference
is seen that the GESC result comes much closer to the four-
1s 1260.13 ⫺121.27 ⫺208.73 87.45 component ones than the LRESC/BPPT values in all cases.
2s 281.11 ⫺13.64 ⫺23.91 10.27 The differences in the cases of Xe and I are within 1% of the
2P 共⫻3兲 279.32 ⫺3.09 ⫺3.21 0.12 total absolute shielding.
3s 93.63 ⫺2.28 ⫺3.09 1.62
3p 共⫻3兲 89.56 ⫺0.53 ⫺0.54 0.02
3d 共⫻5兲 80.83 ⫺0.17 ⫺0.17 0.001 V. CONCLUDING REMARKS
4s 28.55 ⫺0.24 ⫺0.83 0.59
An alternative approach for the calculation of the e-p
4p 共⫻3兲 23.29 ⫺0.04 0.07 ⫺0.11
contribution to magnetic properties within the LRESC
Total 3244.08 ⫺149.26 ⫺249.28 100.02
scheme was developed. Its main features are as follows. The
ESC approximation is applied to the inverse propagator ma-
important as orbitals with zero density at the nucleus are trix in such a way that the relativistic effect is now summed
considered. It is worthy to note a significant difference in the up as a geometric series. However, no explicit evaluation of
stability and reliability of results of the GESC and LRESC positronic states is needed and the final form is expressed in
calculated values as obtained in the present work. As the terms of Breit–Pauli spinors corresponding to electronic
basis set is enlarged with tight s-type functions, the linear- states only. Due to the presence of operator H+ in the de-
ized propagator shows numerical problems, which are due to nominator, a singularity in the vicinity of the nucleus is
the 1 / r dependence of H+. This difficulty is only overcome avoided, yielding a better description of the sought relativis-
when the operator form of Eq. 共52兲 based in the elimination tic effect. This feature has a deep influence in numerical
of the potential V in Eq. 共50兲 is considered. On the contrary, results of the nuclear magnetic shielding tensor, especially on
in the GESC expressions, inclusion of tight s functions does those contributions corresponding to inner shell tight orbit-
not have a significant influence in the propagator matrix, as als. In fact, numerical results presented in this work show
in this case the large effect of the 1 / r dependence is in the that with the GESC approximation, the inner shell s-type
denominator. This is a situation in which the linearization in orbitals yield a smaller contribution in absolute value as
terms of c−2 yields an operator expression which is singular compared to the LRESC scheme. As a consequence, the total
in the vicinity of the nucleus. This singularity is spurious and nuclear shielding of the heavy nucleus is in close agreement
does not correspond to a good description of the relativistic with results of four-component approaches for atomic num-
effect. The GESC propagator overcomes this problem in a bers of the order Z = 50, i.e., in the fifth row of the Periodic
way similar to that of the ZORA10–12 approach. Table.
In Table III the total e-p contribution and the total abso-
lute value of the relativistic nuclear magnetic shielding con- ACKNOWLEDGMENTS
stants of the isolated Kr and Xe atoms and of I in IH ob-
Financial support from the University of Buenos Aires
tained in the present work are compared to four-component
共Grant No. UBACYT X222兲 and CONICET 共Grant No.
PIP5119兲 is gratefully acknowledged.
TABLE II. Atomic orbital contributions to relativistic effects on the nuclear
magnetic shielding constant of Xe for the isolated atom arising from e-p
rotations given by Eq. 共59兲. LRESC, GESC, and difference between both. APPENDIX: EXPLICIT FORM OF LRESC OPERATORS
Values in ppm. FOR THE E-P CONTRIBUTION TO THE
MAGNETIC SHIELDING TENSOR
␴b,P
The first term of Eq. 共56兲 is worked out as follows:6,37
AO NR GESC LRESC Difference
1
1s 1898.06 ⫺324.70 ⫺618.91 294.20 Eb,A
1 = 兺 具␾e兩a22 − A2兩␾e典
2c2 eocc
2s 436.98 ⫺39.39 ⫺63.39 24.01
2p 共⫻3兲 436.28 ⫺11.06 ⫺12.10 1.03
1
3s
3p 共⫻3兲
160.71
157.76
⫺8.24
⫺2.48
⫺16.80
⫺2.71
8.56
0.24
= 兺 具␾e兩共N共␴ . A兲N − x共␴ . A兲x兲2 − A2兩␾e典
2c2 eocc
3d 共⫻5兲 152.80 ⫺1.017 ⫺1.035 0.017
−1 1
4s
4p 共⫻3兲
65.41
61.82
⫺1.75
⫺0.504
⫺5.23
⫺0.546
3.48
0.042
= 兺 具␾e兩 2 p2A2 + Ap2A + 共␴ · A␴ · p兲2
8c4 eocc
4d 共⫻5兲 53.56 ⫺0.170 ⫺0.173 0.003
5s 23.00 ⫺0.218 ⫺0.304 0.086 1 1
+ ␴ · Ap2␴ · A + A2 p2 + 共␴ · A␴ · p兲2兩␾e典, 共A1兲
5p 共⫻3兲 19.47 ⫺0.059 ⫺0.026 ⫺0.033 2 2
Total 5641.98 ⫺422.56 ⫺756.82 334.27
where x and N were defined in Eqs. 共26兲 and 共35兲 and terms

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
084102-8 Zaccari et al. J. Chem. Phys. 130, 084102 共2009兲

TABLE III. Comparison of e-p contributions and total values of the nuclear magnetic shielding constant of Kr, Xe, and I obtained with present work GESC
formalism, LRESC, and four-component methods. Values in ppm.

Nucleus Kr Xe I

NR 3244.08, 3245.67a 5641.98, 5642.30b 5505.40b

␴b,Ac ⫺124.51 ⫺420.48 ⫺385.72
␴b,P LRESC ⫺249.28, ⫺251.12c ⫺756.82, ⫺882.57c ⫺819.05
␴b,P GESC ⫺149.26 ⫺422.56 ⫺400.61
Total e-p LRESC 2868.45 4338.93 4300.63
Total e-p GESC 2968.5 4798.94 4719.07
Four-component e-p 2990a 4992.8,b 4905a 4889.1b
LRESC e-ed 685.3 1379.6 1242.9
Total LRESC/BPPT 3553.8e 6718.5,e 6747.2f 5543.5g
Total GESC 3653.8 7052.8 5962.0
Total four component 3598,h 3525.5,i 3572.6,j 7040,h 6660.9,i 6982.2,j 5855.3,b 5913.7e
3577.3,a 3593.4k 6938,a 7011.6,b
7017.5,k 7019.8f
a
Reference 4.
b
Reference 5.
c
Reference 43. Rescaled as indicated in the Appendix.
d
Calculated as the difference of the total LRESC/BPPT value and present LRESC e-p contributions.
e
Reference 9.
f
Reference 31.
g
Reference 7.
h
Reference 52 共RRPA兲.
i
Reference 51. FPT-DF calculation with finite nucleus.
j
Reference 13.
k
Reference 20.

of order higher than c−4 have been neglected. The following 具␾e兩2共⳵iAN,j兲共⳵iAB,j兲兩␾e典 = 具␾e兩B0 · BN兩␾e典, 共A8兲
relations hold:
具␾e兩p2A2兩␾e典 = 具␾e兩A2 p2兩␾e典 共A2兲 具␾e兩 − ABⵜ2AN兩␾e典 = 4␲具␾e兩AB · 共ⵜ ⫻ ␮␦共rN兲兲兩␾e典,

for 兩␾e典 a real wave function, 共A9兲

共共␴ · p兲共␴ · A兲兲2 + 共共␴ · A兲共␴ · p兲兲2

具␾e兩AB · 共ⵜ ⫻ ␮␦共r兲兲兩␾e典 = 具␾e兩␮ · B0␦共rN兲兩␾e典, 共A10兲
= 共共␴ · p兲共␴ · A兲 + 共␴ · A兲共␴ · p兲兲2 − 共␴ · p兲A2共␴ · p兲
and, therefore,
− 共␴ · A兲p2共␴ · A兲 共A3兲
in the Coulomb gauge, Eb,A
1 =
−1
4 兺 具␾e兩4
8c eocc
冉 冊
␮ . LN
rN3
共B0 · L兲 + B0 · BN
共␴ · p兲共␴ · A兲 + 共␴ · A兲共␴ · p兲 = 2p · A + ␴ · B, 共A4兲
+ 4␲␮ · B0␦共rN兲兩␾e典. 共A11兲
pA2p = 共pA2兲 · p + A2 p2 , 共A5兲
Separating explicitly the Fermi contact term of the nucleus
where the brackets indicate that p acts only on A2. magnetic field the following is obtained:

冉 冊
Considering a singlet ground state wave function and
therefore neglecting spin-dependent operators, Eq. 共A1兲 can 1 ␮ · LN
Eb,A
1 =− 4 兺 具␾e兩4 共B0 · L兲 + B0 · BN,dip
be re-expressed as 8c eocc r3n
1 20␲
Eb,A
1 =− 兺 具␾e兩4共p · A兲2 + B2 − 共pA2兲 · p兩␾e典, 共A6兲
8c4 eocc
+
3
␮ · B0␦共r兲兩␾e典. 共A12兲

具␾e兩共pA2兲 · p兩␾e典 = 21 具␾e兩ⵜ2A2兩␾e典 The Eb,P,LRESC contribution contains the following matrix el-
ements:
= 具␾e兩共A · ⵜ2A兲 + 共⳵iA j兲共⳵iA j兲兩␾e典,
具␾e兩共␴ · A兲共共␴ · A兲p2 + p2共␴ · A兲兲兩␾e典
共A7兲
= 具␾e兩A2 p2 + Ap2 · A兩␾e典, 共A13兲
where i , j stand for Cartesian components and Einstein’s
convention of sum of repeated indices is applied. where only spin-independent operators are retained as a sin-
Inserting the magnetic potentials of the uniform and glet ground state wave function is assumed. Since
nucleus fields and retaining terms bilinear in ␮ and B0 the
following is obtained: p2A = 共p2A兲 + 2共piA兲pi + Ap2 , 共A14兲

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp
084102-9 Geometric ESC approach of magnetic properties J. Chem. Phys. 130, 084102 共2009兲

2A · 共piA兲pi = 共pA2兲 · p, 共A15兲 共2003兲.

9
P. Manninen, K. Ruud, P. Lantto, and J. Vaara, J. Chem. Phys. 122,
where once again brackets are used to indicate that operator 114107 共2005兲; P. Manninen, K. Ruud, P. Lantto, and J. Vaara, ibid. 124,
149901 共2006兲.
p acts only on functions contained in it and making use of 10
S. K. Wolff, T. Ziegler, E. van Lenthe, and E. J. Baerends, J. Chem. Phys.
Eq. 共A7兲 the following is obtained: 110, 7689 共1999兲.
11
J. Autschbach and T. Ziegler, J. Chem. Phys. 113, 936 共2000兲.
具␾e兩Ap2 · A兩␾e典 = 具␾e兩共⳵iA j兲共⳵iA j兲 + A2 p2兩␾e典. 共A16兲 12
J. Autschbach and T. Ziegler, J. Chem. Phys. 113, 9410 共2000兲.
13
R. Fukuda, M. Hada, and H. Nakatsuji, J. Chem. Phys. 118, 1015 共2003兲.
As a consequence, Eq. 共A13兲 can be expressed as 14
R. Fukuda, M. Hada, and H. Nakatsuji, J. Chem. Phys. 118, 1027 共2003兲.
具␾e兩A2 p2 + Ap2 · A兩␾e典 = 具␾e兩2A2 p2 + 共⳵iA j兲共⳵iA j兲兩␾e典.
15
K. Kudo and H. Fukui, J. Chem. Phys. 123, 114102 共2005兲; K. Kudo and
H. Fukui, ibid. 124, 209901 共2006兲.
共A17兲 16
W. Kutzelnigg, Phys. Rev. A 54, 1183 共1996兲.
17
W. Kutzelnigg, Phys. Rev. A 67, 032109 共2003兲.
Inserting the magnetic potentials of the uniform and nucleus 18
R. Szmytkowski, Phys. Rev. A 65, 032112 共2002兲.
magnetic fields and retaining terms bilinear in ␮ and B0 the 19
Y. Xiao, D. Peng, and W. Liu, J. Chem. Phys. 126, 081101 共2007兲.
20
following is obtained: Y. Xiao, W. Liu, L. Chang, and D. Peng, J. Chem. Phys. 126, 214101
共2007兲.
21
1 W. Kutzelnigg and W. Liu, J. Chem. Phys. 123, 241102 共2005兲.
Eb,P,LRESC = − 兺 具␾e兩4共AB · AN兲p2
8c4 eocc
22
23
W. Liu and D. Peng, J. Chem. Phys. 125, 044102 共2006兲.
M. Barysz and A. J. Sadlej, J. Chem. Phys. 116, 2696 共2002兲.
24
H. Maeda, Y. Ootani, and H. Fukui, J. Chem. Phys. 126, 174102 共2007兲.
8␲ 25
Y. Ootani, H. Maeda, and H. Fukui, J. Chem. Phys. 127, 084117 共2007兲.
+ 共␮ · B0␦共rN兲 + B0 · BN,dip兲兩␾e典. 共A18兲 26
3 H. Maeda, Y. Ootani, and H. Fukui, J. Chem. Phys. 128, 129903 共2008兲.
27
M. Reiher and A. Wolf, J. Chem. Phys. 121, 2037 共2004兲.
28
In Refs. 8, 9, and 43 numerical values for the relativistic A. Wolf and M. Reiher, J. Chem. Phys. 124, 064102 共2006兲.
29
corrections of the diamagnetic term are classified as ␴cont, A. Wolf and M. Reiher, J. Chem. Phys. 124, 064103 共2006兲.
30
J. I. Melo, M. C. Ruiz de Azúa, J. E. Peralta, and G. E. Scuseria, J. Chem.
␴d-KE, ␴L-PSO and ␴dip, which are obtained from the following Phys. 123, 204112 共2005兲.
energy expressions: 31
M. Hanni, P. Lantto, M. Ilias, H. J. Aa. Jensen, and J. Vaara, J. Chem.
Phys. 127, 164313 共2007兲.
2␲
兺 具␾e兩共␮ · B0␦共rN兲兲兩␾e典,
32
Econt = − 共A19兲 M. Straka, P. Lantto, M. Räsänen, and J. Vaara, J. Chem. Phys. 127,
3c4 eocc 234314 共2007兲.
33
P. Lantto and J. Vaara, J. Chem. Phys. 127, 084312 共2007兲.
34
M. M. Sternheim, Phys. Rev. 128, 676 共1962兲.
1
兺 具␾e兩4共AB · AN兲p2兩␾e典,
35
G. A. Aucar and J. Oddershede, Int. J. Quantum Chem. 47, 425 共1993兲.
Ed-KE = − 共A20兲 36
H. Fukui, T. Baba, and H. Inomata, J. Chem. Phys. 105, 3175 共1996兲.
2c4 eocc 37
D. G. Zaccari, M. C. Ruiz de Azúa, J. I. Melo, and C. G. Giribet, J.

冉 冊
Chem. Phys. 124, 054103 共2006兲.
1 ␮ · LN 38
R. E. Moss, Advanced Molecular Quantum Mechanics 共Chapman and
EL-PSO = − 4 兺 具␾e兩4 共B0 · L兲兩␾e典, 共A21兲 Hall, London, 1973兲.
2c eocc r3n 39
C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg, Photons and At-
oms 共Wiley, New York, 1997兲.
40
1 F. Gross, Relativistic Quantum Mechanics and Field Theory 共Wiley, New
Edip = − 兺 具␾e兩B0 · BN,dip兩␾e典.
4c4 eocc
共A22兲
41
York, 1993兲.
G. A. Aucar, T. Saue, L. Visscher, and H. J. Aa. Jensen, J. Chem. Phys.
110, 6208 共1999兲.
This last term makes no contribution to the isotropic shield- 42
P. G. Roura, J. I. Melo, M. C. Ruiz de Azua, and C. G. Giribet, J. Chem.
ing for the systems under study in the present work. Phys. 125, 064107 共2006兲.
43
Explicitly, the relations applied in order to obtain nu- S. S. Gómez, J. I. Melo, R. H. Romero, G. A. Aucar, and M. C. Ruiz de
Azúa, J. Chem. Phys. 122, 064103 共2005兲.
merical results from Ref. 43 are 44
T. Helgaker, H. J. Aa. Jensen, P. Joergensen, J. Olsen, K. Ruud, H.
5 cont Aagren, A. A. Auer, K. L. Bak, V. Bakken, O. Christiansen, S. Coriani, P.
Eb,A
1 = 4E + EL-PSO + 21 Edip , 共A23兲 Dahle, E. K. Dalskov, T. Enevoldsen, B. Fernandez, C. Haettig, K. Hald,
A. Halkier, H. Heiberg, H. Hettema, D. Jonsson, S. Kirpekar, R. Koba-
Eb,P,LRESC = 21 Econt + Ed-KE + 21 Edip . 共A24兲 yashi, H. Koch, K. V. Mikkelsen, P. Norman, M. J. Packer, T. B. Peder-
sen, T. A. Ruden, A. Sanchez, T. Saue, S. P. A. Sauer, B. Schimmelpfen-
nig, K. O. Sylvester-Hvid, P. R. Taylor, and O. Vahtras, DALTON, a
1 molecular electronic structure program, Release 1.2, 2001.
See, e.g., Calculation of NMR and EPR Parameters: Theory and Appli-
45
cations, edited by M. Kaupp, M. Bühl, and V. G. Malkin 共Wiley-VCH, P. Lazzeretti, J. Chem. Phys. 71, 2514 共1979兲.
46
Weinheim, 2004兲. See http://folk.uio.no/knutf/bases/one for Faegri’s basis sets used in the
2
L. B. Casabianca and A. C. de Dios, J. Chem. Phys. 128, 052201 共2008兲. calculation.
47
3
L. Visscher, T. Enevoldsen, T. Saue, H. J. Aa. Jensen, and J. Oddershede, T. H. Dunning, Jr., J. Chem. Phys. 90, 1007 共1989兲.
48
J. Comput. Chem. 20, 1262 共1999兲. L. Visscher and K. G. Dyall, J. Chem. Phys. 104, 9040 共1996兲.
4 49
J. Vaara and P. Pyykkö, J. Chem. Phys. 118, 2973 共2003兲. L. Visscher, J. Styszyñski, and W. C. Nieuwpoort, J. Chem. Phys. 105,
5
L. Visscher, Adv. Quantum Chem. 48, 369 共2005兲. 1987 共1996兲.
6 50
J. I. Melo, M. C. Ruiz de Azúa, C. G. Giribet, G. A. Aucar, and R. H. M. C. Ruiz de Azua, J. I. Melo, and C. G. Giribet, Mol. Phys. 101, 3103
Romero, J. Chem. Phys. 118, 471 共2003兲. 共2003兲.
7 51
J. I. Melo, M. C. Ruiz de Azúa, C. G. Giribet, G. A. Aucar, and P. F. Y. Ishikawa, T. Nakajima, M. Hada, and H. Nakatsuji, Chem. Phys. Lett.
Provasi, J. Chem. Phys. 121, 6798 共2004兲. 283, 119 共1998兲.
52
8
P. Manninen, P. Lantto, J. Vaara, and K. Ruud, J. Chem. Phys. 119, 2623 D. Kolb, W. R. Johnson, and P. Shorer, Phys. Rev. A 26, 19 共1982兲.

Downloaded 03 Mar 2009 to 157.92.44.71. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp