Jahn-Teller, pseudo-Jahn-Teller, and spin-orbit coupling Hamiltonian of a d

electron in an octahedral environment

Leonid V. Poluyanov and Wolfgang Domcke
Citation: J. Chem. Phys. 137, 114101 (2012); doi: 10.1063/1.4751439
View online: http://dx.doi.org/10.1063/1.4751439
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v137/i11
Published by the American Institute of Physics.

Additional information on J. Chem. Phys.

Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

THE JOURNAL OF CHEMICAL PHYSICS 137, 114101 (2012)

Jahn-Teller, pseudo-Jahn-Teller, and spin-orbit coupling Hamiltonian

of a d electron in an octahedral environment
Leonid V. Poluyanov1 and Wolfgang Domcke2
1
2

Institute of Chemical Physics, Academy of Sciences, Chernogolovka, Moscow 142432, Russia

Department of Chemistry, Technische Universitt Mnchen, D-85747 Garching, Germany

(Received 16 May 2012; accepted 24 August 2012; published online 17 September 2012)
Starting from the model of a single d-electron in an octahedral crystal environment, the Hamiltonian
for linear and quadratic Jahn-Teller (JT) coupling and zeroth order as well as linear spin-orbit (SO)
coupling in the 2 T2g + 2 Eg electronic multiplet is derived. The SO coupling is described by the
microscopic Breit-Pauli operator. The 10 10 Hamiltonian matrices are explicitly given for all
linear and quadratic electrostatic couplings and all linear SO-induced couplings. It is shown that the
2
T2g manifold exhibits, in addition to the well-known electrostatic JT effects, linear JT couplings
which are of relativistic origin, that is, they arise from the SO operator. While only the eg mode is JTactive in the 2 Eg state in the nonrelativistic approximation, the t2g mode becomes JT-active through
the SO operator. Both electrostatic as well as relativistic forces contribute to the 2 T2g 2 Eg pseudoJT coupling via the t2g mode. The relevance of these analytic results for the static and dynamic JT
effects in octahedral complexes containing heavy elements is discussed. 2012 American Institute
of Physics. [http://dx.doi.org/10.1063/1.4751439]
I. INTRODUCTION

The Jahn-Teller (JT) effect is an extremely widespread

phenomenon in molecular and solid-state spectroscopy.16
It is reflected, for example, in the absorption and photoluminescence spectra of transition-metal or rare-earth ions in
crystal lattices1, 712 or in ultrafast radiationless transitions
and photochemical reactions in organometallic coordination
complexes.1316
The interplay of the splitting of degenerate electronic
states by linear and quadratic JT coupling of electrostatic origin and by spin-orbit (SO) coupling is ubiquitous in solidstate spectroscopy. In the vast theoretical literature on this
field, the JT coupling is described by a Taylor expansion of
the spin-free electronic Hamiltonian up to second order in the
relevant vibrational displacement coordinates. The SO coupling, on the other hand, is approximated in zeroth order, that
is, by its value at the high-symmetry reference geometry. In
most cases, the SO coupling is represented by an effective
single-center atomic SO operator which, as such, is independent of the nuclear coordinates.16 This level of description
(spin-free electronic potentials up to second order, SO coupling in zeroth order) may be termed the standard model of
JT theory.
In the present work, we address the treatment of SOcoupling effects beyond the standard model for cubic and
octahedral systems. The analysis will be based on the microscopic Breit-Pauli SO operator17 and all SO-induced
vibronic-coupling terms will systematically be included up to
first order in the normal-mode displacements. It will be shown
that there exist JT as well as pseudo-JT (PJT) couplings which
arise from the SO operator and thus are of relativistic origin.
Since the symmetry selection rules in the relativistically generalized cubic and octahedral spin double groups (O  , Oh ) are
different from the selection rules in the spin-less case (point
0021-9606/2012/137(11)/114101/11/$30.00

groups O, Oh ), there arise novel SO-induced JT and PJT couplings which partly are complementary to the electrostatic JT
couplings.
Octahedral coordination is particularly common in solids
and in organometallic compounds. The simplest example of
such systems is an octahedrally coordinated transition-metal
ion with a single electron in the d-shell, such as the Ti3+
ion in sapphire (Al2 O3 ),10, 18 which is of particular interest
as a highly versatile lasing material.19, 20 As is well known,
the fivefold degenerate d orbital splits into a threefold degenerate orbital of T2g symmetry and a twofold degenerate orbital of Eg symmetry.21 The JT-active modes are of t2g and
eg symmetry.16 In addition to the JT couplings within the degenerate T2g and Eg electronic manifolds, the T2g and Eg states
can interact via the t2g normal mode (PJT coupling). Inclusion
of spin doubles the minimal electronic basis from the five spatial orbitals of a d level to ten spin-orbitals. The JT and PJT
vibronic Hamiltonians are thus given by 10 10 matrices.
The JT and SO coupling effects in transition-metal and
rare-earth compounds are not only responsible for complex
electronic spectra and ultrafast photophysical dynamics (see
Refs. 1416, 22, 23 for recent experimental studies and
Refs. 2427 for theoretical studies), but also for distortions of
ground-state equilibrium structures from the expected tetrahedral or octahedral shapes (the so-called static JT effect16 ).
In most cases, the electrostatic (spin-free) JT selection rules28
and spin-free electronic-structure calculations were employed
for the analysis of the static JT effect in tetrahedral and octahedral systems.2933 Only relatively recently, due to the availability of broadly applicable relativistic electronic-structure
codes, the effects of strong SO couplings on the equilibrium
geometries of heavy-element compounds have been elucidated for a number of examples.3437 The expected quenching of the static JT distortion38 by large SO splittings was

137, 114101-1

2012 American Institute of Physics

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

114101-2

L. V. Poluyanov and W. Domcke

J. Chem. Phys. 137, 114101 (2012)

observed. It was shown that SO splittings can be of the same

order of magnitude as crystal-field splittings for the hexahalides of the heaviest transition-metal elements.37
In the present work, we develop a systematic theory of
JT, PJT, and SO coupling in octahedral systems which is not
limited to the assumption of weak SO-coupling effects. The
eigenvalues of the 10 10 JT matrix define a tenfold adiabatic potential-energy surface as a function of the t2g and eg
normal coordinates. This model is suitable for a systematic
fitting of ab initio calculated energy data and thus allows the
quantitative extraction of JT couplings and SO couplings, including also SO-induced JT couplings, from ab initio data.
II. ELECTRONIC HAMILTONIAN, ELECTRONIC BASIS
FUNCTIONS, VIBRATIONAL NORMAL MODES AND JT
SELECTION RULES

In this work, we are concerned with XY6 complexes

of octahedral symmetry (point group Oh ), where X is a
transition-metal atom or ion and Y is a main-group atom or
ion. We consider the case of a single electron in the d-shell of
the central atom (or, equivalently, a d9 configuration), while
Y is a closed-shell atom or ion.
For the purpose of symmetry analysis, the electrostatic
(spin-free) Hamiltonian of the single electron in an octahedral
crystal environment can be written as (in atomic units)
H es = 12 2 (r)
(r) =

q0
+
r

6

k=1

(1a)

q1
,
rk

(1b)

r = |r|

(1c)

rk = |r Rk | ,

(1d)

where r is the radius vector of the electron and the Rk , k

= 1. . . 6, denote the radius vectors from the origin to the six
corners of the octahedron. The X atom is at the origin of the
coordinate system.
The Breit-Pauli SO coupling operator for this system
reads17


6

q
1
0
2
H SO = ige e S 3 (r ) + q1
(rk ) , (2a)
r
r3
k=1 k
where
S = 12 (ix + jy + kz ),

(2b)

x , y , z are the Pauli spin matrices, e is the Bohr magneton, ge is the g-factor of the electron, and i, j, k are the Cartesian unit vectors. For the analysis of the symmetry properties,
it is useful to write the Breit-Pauli operator in determinal form

H SO

y
y

x =


,
x

etc.

(3b)

Equation (3a) reveals that the SO coupling operator is completely determined by the electrostatic potential (r). Since
(r) depends on the nuclear coordinates, so does H SO .
The symmetry group of the SO operator (2) is Oh , the
octahedral spin double group. The elements of Oh are of the
form
Zn = Cn Un ,

(4)

where the Cn are the 48 spatial symmetry operations of the

point group Oh and the Un are unitary 2 2 matrices acting
on the spin functions , . To close the group, Zn has to
be included for each Zn of Eq. (4), which doubles the group
order to 96.39 The explicit form of the symmetry operators Zn
is given in Appendix A.
In addition, H SO is time-reversal invariant. The timereversal operator for a single unpaired electron is the antiunitary operator (up to an arbitrary phase factor)39


0 1
=

cc,
(5)
= iy cc
1 0
denotes the operation of complex conjugation. The
where cc
full symmetry group of H SO of Eq. (2) is thus
G = Oh (E, ),

(6)


where q0 and q1 are effective charges of the atomic centers X

and Y, respectively, and


 x

2 
1
= 2 ige e  x



where


z 
z  ,


z

where E is the identity operator. The order of the group G is

192. The operations of Oh commute with .
As is well known, the fivefold degenerate d-orbital on
the transition-metal atom splits into a threefold degenerate orbital of T2g symmetry and a twofold degenerate orbital of Eg
symmetry.21 The former can be written as

x = yzf (r)

y = xzf (r) T2g ,

(7)

z = xyf (r)
where x, y, z are the coordinates of the electron with respect
to the center of the octahedron and f(r) is an exponential or
Gaussian radial function. The orbitals of Eg symmetry can be
written as

a = 16 (2z2 x 2 y 2 )f (r)
(8)
Eg .
b = 12 (x 2 y 2 )f (r)
Including electron spin, the electronic basis set is given by the
ten spin orbitals


x , y , z , z , y , x , a , b , b , a .
(9)
These basis functions define a ten-dimensional double-valued
reducible representation
(T2g + Eg ) Eg1/2 ,

(3a)

(10a)

which can be decomposed into irreducible representations

of Oh
(T2g + Eg ) Eg1/2 = Eg5/2 + 2Gg3/2 .

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

(10b)

114101-3

L. V. Poluyanov and W. Domcke

J. Chem. Phys. 137, 114101 (2012)

The well-known spin-free JT selection rules for T2g and

Eg electronic states in cubic/octahedral symmetry are28

z
5
2
C

C'
4

FIG. 1. Enumeration of the corners of the octahedron (A, B, C, A , B , C )

and the corners of the enclosing cube (1. . . 8).

Here Eg1/2 and Eg5/2 are two-dimensional double-valued irreducible representations, while Gg3/2 is a four-dimensional
double-valued irreducible representation of Oh .40 One of the
two Gg3/2 manifolds corresponds to the 2 Eg state; the other
Gg3/2 manifold is an irreducible component of the 2 T2g electronic state.
The 15 vibrational normal modes of the centered octahedron are of the species
= a1g + eg + t2g + 2t1u + t2u .

(11)

The definitions of the symmetry coordinates and pictorial

representations can be found, for example, in Table 2.2 and
Fig. 2.3 of Ref. 6, respectively. Designating the central atom
as O and the atoms at the corners of the octahedron as A, B,
C, C , B , A , see Fig. 1, symmetry-adapted linear combinations of t2g and eg symmetry are

(rBC + rB C rBC rB C )

1
Sy = 2 (rAC + rA C rAC rA C ) t2g ,

Sz = 12 (rBA + rB A rBA rB A )

sa =

1
2

(2rOC
2 3

(12)

Gg3/2 Eg5/2 = Eg + T1g + T2g .

eg ,

(13)
where rAB , etc., are displacements of interatomic distances
from the octahedral reference geometry. Normal coordinates
Qx , Qy , Qz of t2g symmetry and qa , qb of eg symmetry are
obtained by multiplication of the symmetry coordinates with
appropriate mass-dependent conversion factors.41

(16b)

Here { } denotes the antisymmetrized square of the irreducible representation .39 According to Eq. (16a), the fourfold degeneracy of a Gg3/2 level is lifted by t2g and eg modes,
and according to Eq. (16b), Gg3/2 and Eg5/2 levels interact via
the t2g and eg modes. JT splitting of the twofold degenerate
Eg5/2 level is excluded by time-reversal symmetry, which requires at least twofold degeneracy of levels for an odd number
of electrons (Kramers degeneracy).
It should be noted that in the spin-free case only the eg
mode is JT-active in the Eg electronic state (Eq. (14a)), while
both eg and t2g modes are active in the 2 E (Gg3/2 ) electronic
state (Eq. (16a)). This implies that the JT-activity of the t2g
mode must arise from the SO operator. The same applies in
tetrahedral symmetry.43, 44
III. (2 T2g + 2 Eg ) (t2g + eg ) JT AND PJT HAMILTONIAN
A. Taylor expansion of the electronic Hamiltonian and
calculation of matrix elements

In accordance with the standard model of JT coupling,

we expand the electrostatic Hamiltonian up to second order
in the t2g and eg normal-mode displacements. Assuming that
SO coupling is somewhat weaker than the electrostatic interactions, we terminate the Taylor expansion of the SO operator
after the first order. The Hamiltonian thus has the structure
H = H es + H SO ,

+ 2rOC rOA rOA rOB rOB )

sb = 12 (rOA + rOA rOB rOB )

(15)

The t2g mode is thus PJT-active in first order.

The JT selection rules relevant for the d 1 configuration
in the spin double group Oh are42
2 
Gg3/2 = A1g + Eg + T2g ,
(16a)

Sx =

(14b)

Eg T2g = T1g + T2g .

[T2g ]2 = A1g + Eg + T2g ,

where [ ] denotes the symmetrized square of the irreducible

representation .39 In electronic states of Eg symmetry, the
vibrational mode of eg symmetry is JT-active in first order,
while in electronic states of T2g symmetry, the eg mode as
well as the t2g mode are active in first order according to
Eq. (14). The selection rule for Eg T2g PJT coupling is

A'
O

(14a)

1
B'

[Eg ]2 = A1g + Eg ,

(17)

(1)
(2)
(2)
(1)
(2)
H es = H es(0) + H es,Q
+ H es,q
+ H es,Q
+ H es,q
+ H es,Qq
,
(18)

(0)
(1)
(1)
+ H SO,Q
+ H SO,q
.
H SO = H SO

(19)

Here, H es(0) is the electrostatic Hamiltonian of Eq. (1) at the oc(0)

tahedral reference geometry. Likewise, H SO
is the SO operator of Eq. (2) at the reference geometry. The first- and secondorder expansion terms of the electrostatic Hamiltonian can be

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

114101-4

L. V. Poluyanov and W. Domcke

J. Chem. Phys. 137, 114101 (2012)

written as
(1)
H es,Q

= A 1 (t2g )Qx + A 2 (t2g )Qy + A 3 (t2g )Qz ,

(20)

(1)
H es,q
= B 1 (eg )qa + B 2 (eg )qb ,

(2)
H es,Q
=

(21)


 2
Qx + Q2y + Q2z


+ 16 D 1 (eg ) 2Q2z Q2x Q2y


+ 12 D 2 (eg ) Q2x Q2y
1g )
1 C(a
3

+ E 1 (t2g )Qy Qz + E 2 (t2g )Qx Qz + E 3 (t2g )Qx Qy ,

(22)
(2)
H es,q
=

1 F (a1g )
2

(2)
H es,Qq
=


 2
qa + qb2 +

(e )
1 G
2 1 g


 2
qa qb2

2 (eg )qa qb ,
2G

(23)

3
H1 (t1g )Qx q+ + 23 H 2 (t1g )Qy q + H 3 (t1g )Qz qb
2

23 I1 (t1g )Qx q + 23 I2 (t1g )Qy q + + I3 (t1g )Qz qa .

In Eqs. (20)(27), the symmetries of the electronic opera which are the coefficients of the
tors A . . . I and a . . . f, h,
normal-mode expansion, are indicated in parentheses. Since
the Pauli spin matrices have been factored out, the expansion coefficients in Eqs. (26) and (27) transform according
to single-valued irreducible representations of Oh . Given the
transformation properties of the electronic operators, the calculation of matrix elements with the electronic basis functions
of the d-orbital (Eq. (9)) is straightforward. Note that the expansions ((20)(28)) also are useful for the construction of
JT/PJT vibronic matrices for p-orbitals or f-orbitals in octahedral systems.
The labor of calculating the matrix elements of the
Hamiltonian is significantly reduced by the hermiticity of the
Hamiltonian and by time-reversal symmetry. In the basis set
of Eq. (9), the time-reversal operator has the representation
given by the 10 10 matrix T in Appendix A. The requirement that the Hamiltonian matrices commute with T leads to
the vanishing of many of the matrix elements and requires
others to be equal, or equal up to a minus sign.
The matrix elements of the electronic Hamiltonian defined by Eqs. (17)(28) form a 10 10 hermitean matrix,
which we write as
H = Hes + HSO ,

(29a)

(24)
(2)
(2)
(2)
Hes = Hes(0) + Hes(1) + Hes,Q
+ Hes,q
+ Hes,Qq
,

(0)
H SO
is written as
(0)
H SO
= h x (t1g )x + h y (t1g )y + h z (t1g )z .

(25)

The first-order expansion terms of the SO operator take the

form

(0)
(1)
(1)
HSO = HSO
+ HSO,Q
+ HSO,q
,

(29c)

where Hes(1) contains the linear JT and PJT coupling terms of

both t2g and eg modes. For convenience, the totally symmetric
quadratic part of the potentials is included in Hes(0) , that is,


Hes(0) = ET + 12
2T Q2 + 12 T2 q 2 16


(30)
EE + 12
2E Q2 + 12 E2 q 2 14 ,

(1)
2g )(Qx x + Qy y + Qz z )
H SO,Q
= 13 a(a

(e )(2Qx x
1 b
6 1 g

(e )(Qy y
1 b
2 2 g

Qz z )

1 c1 (t1g )(Qz y
2

+ Qy z )

1 c2 (t1g )(Qx z
2

+ Qz x )

1 c3 (t1g )(Qy x
2

+ Qx y )

1 d1 (t2g )(Qz y
2

Qy z )

where ET and EE are the vertical energies of the 2 T2g and 2 Eg

states, 16 and 14 are 6 6 and 4 4 unit matrices, respectively,
T ,
E and T , E are the harmonic vibrational frequencies of the t2g and eg modes in the 2 T2g and 2 Eg states,
and

1 d2 (t2g )(Qx z
2

Qz x )

Q2 = Q2x + Q2y + Q2z ,

1 d3 (t2g )(Qy x
2

Qx y ),

Qy y Qz z )

(31a)

(26)
q 2 = qa2 + qb2 .

(1)
H SO,q

= e1 (t1g )q x + e2 (t1g )q + y + e3 (t1g )qa z

q = 12 ( 3qa qb ),

(28a)

3qb ).

(28b)

q = 12 (qa

(31b)
Hes(1) ,

+ f1 (t2g )q+ x f2 (t2g )q y + f3 (t2g )qb z , (27)

where

(29b)

(2)
Hes,Q
,

(2)
Hes,q
,

and
The electrostatic vibronic matrices
(2)
(0)
Hes,Qq are given in Figs. 25. The vibronic matrices HSO
,
(1)
HSO,Q resulting from the expansion of the SO operator are
given in Figs. 6 and 7. The vibronic matrices in Figs. 27 are
the main results of the present work. We have verified that the
Hamiltonian matrices in Figs. 27 commute with the symmetry operators of Oh as well as with the time-reversal operator
given in Appendix A.

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

114101-5

L. V. Poluyanov and W. Domcke

J. Chem. Phys. 137, 114101 (2012)

cq

aQz

aQy

dQx

aQz

cq

aQx

dQy

3dQx

3dQy

aQy

aQx

cqa

2dQz

cqa

aQx

aQy

aQz

0
2dQz

3dQy
dQy

3dQx dQx

aQx

cq

aQy

aQz

cq

bqa

bqb

bqb

bqa

bqa

bqb

bqb

bqa

dQx
dQy 2dQz

3dQx 3dQy
0

2dQz

0
0

3dQy 3dQx
dQy

dQx

(1)

(1)

FIG. 2. Representation of the first-order electrostatic JT Hamiltonian, Hes,Q + Hes,q , in the spin-orbital basis (9). a, b, c, d are real parameters representing
linear T2g t2g coupling (a), T2g eg coupling (c), Eg eg coupling (b) and T2g Eg PJT coupling by the t2g mode (d).

B. 2 T2g (t2g + eg ) JT Hamiltonian

In this and Subsection III C, we assume that the

crystal-field splitting is large compared to other parameters of the system, such that the 2 T2g and 2 Eg electronic

states can approximately be considered as decoupled. The

JT Hamiltonian of the 2 T2g state is given by the upperleft 6 6 block of the matrices in Figs. 27. Omitting,
for brevity and clarity, the quadratic JT coupling terms, we
have



 
H 2 T2g (t2g + eg ) = ET + 12
2T Q2 + 12 T2 q 2 16

cq

i + aQz

aQy + iQx

+
 + iQz

i + aQz

aQy iQx

 iQz

Q+

cq +

aQx + iQy

i + Qz

aQx iQy

cqa

i Qz

i + Qz

cqa

aQx + iQy

i Qz

aQx iQy

cq +

 iQz

aQy + iQx

i + aQz

where q is defined in Eq. (28a),

Q = Qx iQy ,

(33)

and a, c, , are real parameters. The parameter a is the

linear electrostatic T2g t2g JT coupling constant, c is the
linear electrostatic T2g eg JT coupling constant,  represents the zeroth-order SO splitting of the 2 T2g state and is
the linear relativistic T2g t2g JT coupling constant. It is seen
that the t2g mode is JT-active via electrostatic (parameter a)
as well as relativistic (parameter ) forces. The eg mode, on
the other hand, is JT-active only via the electrostatic forces
(parameter c).
The structure of the zeroth-order SO contribution (that
is, the upper left 6 6 block in Fig. 6) is the same as found
for a 2 T2 state derived from p-type orbitals in tetrahedral

 + iQz

,
aQy iQx

i + aQz

cq
Q+

(32)

symmetry.44 This nondiagonal matrix can be transformed to

diagonal form by a unitary transformation U which defines a
SO-adapted electronic basis. A suitable choice of U has been
found in Ref. 44. In the transformed basis, the zeroth-order
SO matrix of a 2 T2g state takes the form

(0) 2
T2g = diag(, , , , 2, 2), (34)
HSO
in agreement with the group-theoretical result (10b). The 2 T2
(t2 + e) Hamiltonian matrix in the SO-adapted basis is
given in Ref. 44 (Eqs. (33) and (48)).
It should be noted that in tetrahedral systems the 2 T2
e JT coupling has electrostatic as well as relativistic contributions (parameters c and ). In octahedral and cubic systems, on the other hand, the linear relativistic 2 T2g eg coupling parameter is zero by symmetry. The requirement that

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

114101-6
L. V. Poluyanov and W. Domcke

(2)

FIG. 3. Representation of the second-order electrostatic JT Hamiltonian, Hes,Q , of the t2g mode in the spin-orbital basis (9). A, B, C, D are real parameters representing quadratic T2g t2g coupling (A, B), quadratic T2g
eg coupling (C) and quadratic T2g Eg PJT coupling by the t2g mode (D).

(2)

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

J. Chem. Phys. 137, 114101 (2012)

FIG. 4. Representation of the second-order electrostatic JT Hamiltonian, Hes,q , of the eg mode in the spin-orbital basis (9). A and C are real parameters representing quadratic T2g
t2g coupling (A ) and quadratic Eg eg coupling (C ).

114101-7

L. V. Poluyanov and W. Domcke

J. Chem. Phys. 137, 114101 (2012)

the 2 T2g e vibronic matrix must commute with the symmetry operators of Oh (Appendix A) enforces the vanishing
of the coupling parameter . This is another example which
reveals that the symmetry selection rules for the SO operator
differ from those for the electrostatic Hamiltonian in a subtle
manner.

C. 2 Eg (t2g + eg ) JT Hamiltonian

The JT Hamiltonian of the 2 Eg state is given by the lower

right 4 4 block of the matrices in Figs. 27. Omitting, for
brevity and clarity, the quadratic JT coupling terms, we have



 
H 2 E2g (t2g + eg ) = EE + 12
2E Q2 + 12 E2 q 2 14

bqa

bqb iQz
+
iQ
+

bqb + iQz

iQ

bqa

bqa

iQ+

bqb iQz

Note that there is no zeroth-order SO splitting in the 2 Eg

state. In accordance with the JT selection rules discussed in
Sec. II, the eg mode is JT active through electrostatic forces
(coupling constant b), while the t2g mode is JT-active through
relativistic forces (coupling constant ), as in tetrahedral
systems.43
It is straightforward to determine the adiabatic potentialenergy functions as eigenvalues of the matrix (35)
V1 = V2 = EE + 12
2E Q2 + 12 E2 q 2
V3 = V4 = EE + 12
2E Q2 + 12 E2 q 2 +


b 2 q 2 + 2 Q2
(36)


b 2 q 2 + 2 Q2 .

.
bqb + iQz

bqa
iQ

(35)

The energy functions (36) represent an elliptic Mexican hat

in six-dimensional space (the energy as a function of five nuclear coordinates). Since the adiabatic eigenvectors also are
given in analytic form, it is straightforward to calculate the
geometry phases of the adiabatic wave functions as a contour
integral45 as discussed in Ref. 43. The adiabatic wave functions carry nontrivial geometric phases which depend on the
orientation of the plane of integration in the five-dimensional
nuclear coordinate space.
D. (2 T2g + 2 Eg ) (t2g + eg ) PJT coupling

Omitting, for brevity, second-order terms, the 6 4 block

which couples the 6 6 2 T2g matrix with the 4 42 Eg matrix
reads

H T2g ,Eg

dQx + i Qy

dQy i Qx

2dQz
=

2Q+

3 i + Qz

i 3 + Qz

where

= 3,
=

3 .

3dQx + i+ Qy

3dQy + i+ Qx

i + Qz

2i

2Q+

2Q
+ i Qz
i Qz

(38a)

(38b)

It is seen from Eq. (37) that there exists a zeroth-order relativistic coupling of the 2 T2g and 2 Eg states (parameter ), a
first-order electrostatic coupling by the t2g mode (parameter

+ i Qz

2i

3dQy i+ Qx

3dQx i+ Qy

i 3 + + Qz

3 i + Qz

2Q
,

2dQz

dQy + i Qx

dQx i Qy

(37)

d), as well as first-order relativistic couplings by the t2g mode

(parameters , ). The eg mode does not contribute to the 2 T2g
2 Eg PJT coupling.
The nonrelativistic 2 T2g 2 Eg PJT coupling is determined by the single parameter d in Eq. (37). Our result is in
agreement with the linear PJT coupling Hamiltonian given by
Stoneham and Lannoo for cubic and tetrahedral systems.46
The relativistic linear 2 T2g 2 Eg PJT coupling by the t2g
mode (parameters , ) is a new result.

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

114101-8

L. V. Poluyanov and W. Domcke

EQz qa

EQz qa

3
EQy q +
2

J. Chem. Phys. 137, 114101 (2012)

3
EQy q +
2

3
2 EQx q

3
EQx q
2

HQq

HQq

23 EQx q

3
EQy q +
2

(17)

(27)

(17)

(27)

3
EQx q
2

3
EQy q +
2

(18)

(28)

HQq

HQq

F Qz qa GQz qb
0

GQz qb
(28)

EQz qa

HQq

EQz qa

HQq

(18)

HQq

HQq

F Qz qa

(18)
HQq

(28)
HQq

GQz qb

(18)
HQq

(17)

GQz qb

(28)
HQq

F Qz qa

HQq

(27)

HQq

F Qz qa

(27)
HQq

(17)

HQq

0
0

(2)

FIG. 5. Representation of the electrostatic JT Hamiltonian bilinear in the t2g and eg modes, Hes,Qq , in the spin-orbital basis (9). E, F, G are real parameters
(17)

(27)

(18)

representing bilinear T2g (t2g + eg ) JT coupling (E) and bilinear T2g Eg PJT coupling (F, G). See Appendix B for the definition of HQq , HQq , HQq ,
(28)
HQq .

i i 3

2i

2i

i
0

2i

2i

3
i 3

i
0

3 i 3

i 3 i
0

IV. CONCLUSIONS

(0)

FIG. 6. Representation of the zeroth-order SO coupling Hamiltonian, HSO ,

in the spin-orbital basis (9).  and are real parameters.

iQx iQz

iQy

Qz

iQx

iQy

Qz

iQz

Qz

iQy

Q+

We have presented a systematic derivation of the JT, PJT,

and SO coupling Hamiltonian arising from a singly occupied
d-orbital in an octahedral crystal environment. The electrostatic electronic Hamiltonian has been expanded up to second order in t2g and eg normal-mode displacements. The SO
coupling Hamiltonian has been expanded up to first order in
these modes. While the expansion of the electrostatic potential up to second order corresponds to the standard model of
JT theory,16 the expansion of the SO coupling operator up to
first order extends the theory beyond the standard model, in
which SO coupling is considered as an atomic property which
is independent of the nuclear coordinates.16
It has been shown that there exist JT and PJT forces
which are of relativistic origin, arising from the SO operator.
In some cases, e.g., for the 2 T2g t2g JT effect, the electrostatic and relativistic forces act additively, resulting in constructive or destructive interference of electrostatic and relativistic JT couplings. In other cases, e.g., for the 2 E2g (t2g

i+Qy

Q z

+ Qz

Q+ i Qx i+Qx

iQz

i + Qz

2Q+

2Q

2Q

iQz

i Qy

iQx 2Q+
+

Qz

iQy

i Qz

iQz

iQx

+ Qz Qz i+Qy i Qy

i Qy

i Qx

2Q

i + Qz

+ Qz

2Q+

i+Qy i+Qx
Q z
+

Qz

iQz 2Q
+

i Qz

2Q+

iQz Qz

iQz

i+Qx

i Qx

iQz

iQ

iQz

iQ

i+Qx

i+Qy

iQ+

iQz

iQ+

iQz

i Qx i Qy
(1)

FIG. 7. Representation of the first-order SO coupling Hamiltonian of the t2g mode, HSO,Q , in the spin-orbital basis (9). , , , are real parameters representing
relativistic linear T2g t2g coupling (), E2g t2g coupling () and T2g Eg PJT coupling (, ).

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

114101-9

L. V. Poluyanov and W. Domcke

+ eg ) JT effect, the electrostatic and relativistic forces are

complementary. In the 2 Eg state, the eg mode is JT-active
through electrostatic forces, while the t2g mode is JT-active
through relativistic forces. The 2 T2g and 2 Eg states interact in
zeroth order through the SO operator and in first order through
the t2g mode. The PJT coupling via the t2g mode has electrostatic and relativistic origins. The electrostatic JT coupling
terms derived in the present work agree with the JT textbooks
for the terms linear and quadratic in the t2g and eg modes. For
(2)
(Fig. 5), we could not find
the mixed coupling matrix Hes,Qq
literature results. The first-order relativistic JT/PJT coupling
(1)
(Fig. 7) also is new in JT theory.
matrix HSO,Q
While the analysis has been performed for the model of
an electron in a d-orbital at the central atom of an octahedral
complex, the derived JT and PJT matrices should be generally valid for d-orbitals in systems of cubic symmetry (point
groups O, Oh ), such as cubic X8 systems or centered cubic
XY8 systems.
The magnitude of the relativistic JT and PJT coupling parameters scales such as the SO splittings, that is, with Z2 in the
valence shell of heavy elements.47 For the first-row transitionmetal elements, the relativistic JT forces are expected to be
weak compared to the electrostatic forces. For the third-row
transition metals, on the other hand, not only the zeroth-order
SO couplings can be large, but also the SO-induced JT couplings may be of the same order of magnitude as the electrostatic JT couplings.
It has recently been stated that Today, no one has postulated a relativistic JT theorem; therefore, molecular geometry
distortions due to the dynamic JT effect are a consequence
of nonrelativistic treatments.37 This statement is superseded
by our previous results for tetrahedral systems43, 44 and the
present results for cubic and octahedral systems. A JT theory
is now available which systematically includes SO-coupling
effects up to first order in normal-mode displacements
from the reference geometry. If considered necessary, the
expansion of the BP operator (Eq. (19)) could be extended to
include second-order terms.

J. Chem. Phys. 137, 114101 (2012)

To avoid the need of an explicit diabatization of the ab

initio electronic wave functions,48 it is usually convenient to
determine the electrostatic and relativistic JT couplings and
SO-splitting parameters by the fitting of the eigenvalues of the
JT model Hamiltonian to adiabatic ab initio energy data (diabatization by ansatz). For such applications, it is essential
to know which gradients and second derivatives of the electrostatic and SO energies are zero by symmetry and thus are
not adjustable parameters in the fitting procedure.
To the knowledge of the authors, no attempts have been
made so far towards the ab initio determination of geometric
distortions in transition-metal or rare-earth compounds which
arise from purely relativistic forces and are thus not predicted
by the time-honored electrostatic JT selection rules.28 Since
the zeroth-order SO splitting of 2 E states is zero by symmetry
in tetrahedral and octahedral systems, these states are generically unstable with respect to relativistic distortions in normal
modes of t2 symmetry, cf. Eq. (16a). The existence of such
SO-induced 2 E t2 JT distortions has recently been demonstrated by relativistic multi-reference self-consistent-field calculations for the 2 E ground states of the tetrahedral cluster
+
+
cations of the elements of the fifth main group (P+
4 , As4 , Sb4 ,
+ 49, 50
Bi4 ).

APPENDIX A: THE SYMMETRY OPERATORS

OF THE SPIN DOUBLE GROUP Oh

The corners of the cube enclosing the octahedron are

enumerated 1. . . 8, see Fig. 1. Twofold and fourfold rotations
around the coordinate axes are denoted C2, k , C4, k , k = x, y,
z. Threefold rotations around axes through opposite corners
on the enclosing cube are denoted C3, ij , i, j = 1. . . 8. Twofold
rotations around axes through opposite edges of the enclosing
cube are denoted C2, ij, kl , i, j, k, l = 1. . . 8. The unit operation
is denoted as E, the inversion as I. In this notation, the 24 essential symmetry operators of the vibronic Hamiltonian are



1 0
Z1 = E
0 1


0 i
Z2,y = iC2,y
i 0


i
Z3,18 = e C3,18
i


i
Z3,36 = e+ C3,36
i



2
+ 2
Z3,18 = e C3,18
i i



2
2
= e C3,36
Z3,36
i i


i
Z2,12,78 = C2,12,78

Z2,x = iC2,x


Z2,z
Z3,27
Z3,45
2
Z3,27

2
Z3,45

Z2,43,56

0
1

1
0


1 0
= iC2,z
0 1



= e C3,27
i i



= e+ C3,45
i i


i
+ 2
= e C3,27
i


i
2
= e C3,45
i


i
= C2,43,56
i

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

114101-10

L. V. Poluyanov and W. Domcke


Z2,23,76 = C2,23,76

Z2,16,38 = C2,16,38

Z4,x = C4,x

Z4,z = C4,z

3
Z4,y

3
C4,y

J. Chem. Phys. 137, 114101 (2012)

0
e+
i
i

i
e
0


e
0

i
i

i


0
e+



Z2,14,58 = C2,14,58

Z2,25,38 = C2,25,38


3
Z4,z

where
=

1
2

e = ei/4 .
24 additional symmetry operators are given by the product of
the above operators with the inversion operator:
Z24+n = I Zn ,

n = 1 . . . 24.

Z3,27

= e
0

i
i

i
i


To close the group, the operator Zn has to be included for each Zn . This defines the 96 operators of the
group Oh . Note that the point-group symmetry operators act
both on the electronic basis states as well as on the nuclear
coordinates.
In the spin-orbital basis of Eq. (9), the electronic parts of
the above symmetry operators are given by 10 10 matrices.
Z3, 27 , for example, reads

3
2

12

3
2

i 23

2i

The superscript n in the above equation indicates that this

point-group symmetry operator acts on the nuclear degrees of
freedom.
The 23 other essential symmetry operator matrices are
constructed analogously. The JT+PJT Hamiltonian matrix
must commute with all 24 essential symmetry operators and
the inversion operator I. The commutation with the remaining
symmetry operators follows trivially. In addition, the Hamiltonian matrix must commute with the time-reversal operator.

e+
0



i
3
= C4,x
i
 +

0
e
3
= C4,z
0 e

Z4,y = C4,y
3
Z4,x

0
e

3
2

12

1
2

2i

2i

i 3
2

i 3
2

0
(n)
C .
3,27
0

i 3
2
2i

In the basis of Eq. (9), it is given by

0
0
0 0 0 1

0
0
0 0 1 0

0
0
0 1 0 0

0
0 1 0 0 0

0 1 0 0 0 0
T =
1 0
0 0 0 0

0
0
0 0 0 0

0
0
0 0 0 0

0
0
0 0 0 0

0 0

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions

0
cc.

114101-11

L. V. Poluyanov and W. Domcke

J. Chem. Phys. 137, 114101 (2012)

20 P.

APPENDIX B: ABBREVIATIONS IN FIG. 5

The abbreviations in Fig. 5 are defined as follows:




(17)
HQq
= 34 Gq+ + 43 F q Qx ,
(B1)
(27)
HQq
=

1 M.

3
Gq
4

3
F q+
4


Qy ,

(B2)



(18)
HQq
= 43 Gq+ 34 F q Qx ,

(B3)



(28)
HQq
= 43 Gq 34 F q + Qy .

(B4)

D. Sturge, Solid State Phys. 20, 91 (1967).

Englman, The Jahn-Teller Effect (Wiley, New York, 1972).
3 I. B. Bersuker and V. Z. Polinger, Vibronic Interactions in Molecules and
Crystals (Springer, Heidelberg, 1989).
4 M. C. M. OBrien and C. C. Chancey, Am. J. Phys. 61, 688 (1993).
5 I. B. Bersuker, Chem. Rev. 101, 1067 (2001).
6 I. B. Bersuker, The Jahn-Teller Effect (Cambridge University Press, Cambridge, 2006).
7 S. Sugano, Y. Tanabe, and H. Kamimura, Multiplets of Transition Metal
Ions (Academic, New York, 1970).
8 D. Reinen and C. Friebel, Structure and Bonding (Springer, Berlin, 1979),
Vol. 37.
9 D. Reinen and M. Atanasov, Magn. Reson. Rev. 15, 167 (1991).
10 M. Grinberg, A. Mandelis, and K. Fjeldsted, Phys. Rev. B 48, 5922 (1993).
11 F. Rodriguez and F. Aguado, J. Chem. Phys. 118, 10867 (2003).
12 M. N. Sanz-Ortiz and F. Rodriguez, J. Chem. Phys. 131, 124512 (2009).
13 A. Juris, V. Balzani, F. Barigelletti, S. Campagna, P. Belser, and A. von
Zelewsky, Coord. Chem. Rev. 84, 85 (1988).
14 A. Hauser, Top. Curr. Chem. 234, 155 (2004).
15 E. A. Juban, A. L. Smeigh, J. E. Monat, and J. K. McCusker, Coord. Chem.
Rev. 250, 1783 (2006).
16 A. Cannizzo, F. van Mourik, W. Gawelda, G. Zgrablic, C. Bressler, and
M. Chergui, Coord. Chem. Rev. 254, 2677 (2010).
17 H. A. Bethe and E. E. Salpeter, Quantum Mechanics for One- and TwoElectron Atoms (Springer, Berlin, 1957).
18 S. Garcia-Revilla, F. Rodriguez, R. Valiente, and M. Pollnau, J. Phys. Condens. Matter 14, 447 (2002).
19 P. Moulton, Optics News 8, 9 (1982).
2 R.

Albers, E. Stark, and G. Huber, J. Opt. Soc. Am. B 3, 134 (1986).

Moffitt and C. J. Ballhausen, Annu. Rev. Phys. Chem. 7, 107 (1956).
22 T. Hofbeck and H. Yersin, Inorg. Chem. 49, 9290 (2010).
23 J. N. Schrauben, K. L. Dillman, W. F. Beck, and J. K. McCusker, Chem.
Sci. 1, 405 (2010).
24 C. Daniel, Top. Curr. Chem. 241, 119 (2004).
25 R. G. McKinlay, J. M. Zurek, and M. J. Paterson, Adv. Inorg. Chem. 62,
351 (2010).
26 J.-L. Heully, F. Alary, and M. Boggio-Pasqua, J. Chem. Phys. 131, 184308
(2009).
27 A. Y. Freidzon, A. V. Scherbinin, A. A. Bogaturyants, and M. V. Alfimov,
J. Phys. Chem. A 115, 4565 (2011).
28 H. A. Jahn and E. Teller, Proc. R. Soc. London, Ser. A 161, 220 (1937).
29 A. Ceulemans, D. Beyens, and L. G. Vanquickenborne, J. Am. Chem. Soc.
106, 5824 (1984).
30 L. G. Vanquickenborne, A. E. Vinckier, and K. Pierloot, Inorg. Chem. 35,
1305 (1996).
31 R. Bruyndonckx, C. Daul, P. T. Manoharan, and E. Deiss, Inorg. Chem. 36,
4251 (1997).
32 M. Hargittai, Chem. Rev. 100, 2233 (2000).
33 D. Reinen, M. Atanasov, and P. Khler, J. Mol. Struct. 838, 151 (2007).
34 L. Alvarez-Thon, J. David, R. Arriata-Perez, and K. Seppelt, Phys. Rev. A
77, 034502 (2008).
35 J. David, P. Fuentealba, and A. Restrepo, Chem. Phys. Lett. 457, 42
(2008).
36 M. J. Molski and K. Seppelt, Dalton Trans. 2009, 3379.
37 J. David, D. Guerra, and A. Restrepo, Inorg. Chem. 50, 1480 (2011).
38 F. S. Ham, Phys. Rev. A 138, 1727 (1965).
39 M. Hamermesh, Group Theory and Its Application to Physical Problems
(Addison-Wesley, Reading, 1962).
40 We use the symbols E , E , G
1/2
5/2
3/2 of Landau and Lifshitz and
Hamermesh39 for the representations of O . In the solid-state literature,
these representations are denoted as 6 , 7 , 8 .
41 E. B. Wilson, J. C. Decius, and P. C. Cross, Molecular Vibrations
(McGraw-Hill, New York, 1955).
42 H. A. Jahn, Proc. R. Soc. London, Ser. A 164, 117 (1938).
43 L. V. Poluyanov and W. Domcke, J. Chem. Phys. 129, 224102 (2008).
44 L. V. Poluyanov and W. Domcke, Chem. Phys. 374, 86 (2010).
45 M. V. Berry, Proc. R. Soc. London, Ser. A 392, 45 (1984).
46 A. M. Stoneham and M. Lannoo, J. Phys. Chem. Solids 30, 1769 (1969).
47 P. Pyykk, Annu. Rev. Phys. Chem. 63, 45 (2012).
48 H. Kppel, in Conical Intersections: Electronic Structure, Dynamics and
Spectroscopy, edited by W. Domcke, D. R. Yarkony, and H. Kppel (World
Scientific, Singapore, 2004), Chap. 4.
49 D. Opalka, M. Segado, L. V. Poluyanov, and W. Domcke, Phys. Rev. A 81,
042501 (2010).
50 D. Opalka, L. V. Poluyanov, and W. Domcke, J. Chem. Phys. 135, 104108
(2011).
21 W.

Downloaded 20 Sep 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions