HutsonAMVCD (Rev) 1990

Published in Advances in Molecular Vibrations and Collision Dynamics 1A, 1 (1990).
However, the published version had many typographical errors. This is the version that
was submitted to the journal, and is much more accurate.
An Introduction to the Dynamics of Van der Waals Molecules

Jeremy M. Hutson
Department of Chemistry, University of Durham, South Road, Durham, DH1 3LE,
England
OUTLINE
1. Introduction
2. Atom–atom Van der Waals Molecules
2.1 Numerical Methods
2.2 Spectroscopic Constants
3. Atom–diatom Van der Waals Molecules
3.1 Representation of the Intermolecular Potential
3.2 The Coupled Equations
4. Approximate Methods
4.1 Angular Momentum Coupling Cases
4.2 Centrifugal Decoupling or Helicity Decoupling
4.3 Distortion Approximations
4.4 Perturbation Theory
4.5 Adiabatic Approximations
5. Dipole Moments and Spectroscopic Intensities
5.1 Space-fixed
5.2 Body-fixed
6. Larger Systems
6.1 Atom–polyatom Systems
6.2 Molecule–molecule Systems
6.3 Trimeric Systems
Acknowledgments
References
1
1. Introduction
Van der Waals molecules are complexes formed from pairs of chemically stable neu-
tral molecules. Studies of their structure and dynamics can provide detailed information
on anisotropic intermolecular potentials, and they provide valuable prototypes for the
large-amplitude motions found in reacting systems. They are readily formed in molecu-
lar beams, and an enormous variety of spectroscopic experiments has been carried out
on them. There has also been a great deal of theoretical work, aimed at interpreting ex-
perimental spectra and understanding the dynamics. However, the theoretical literature
in this area draws heavily on scattering theory, and there is no introductory treatment
of the subject available. The purpose of this article is to describe the basic theory of
Van der Waals molecules, without assuming a knowledge of the scattering literature,
and to describe some of the most important results. The emphasis will be on describing
the theoretical methods used to calculate vibration-rotation states of Van der Waals
molecules, and the coupling schemes and sets of quantum numbers that should be used
to describe the resulting states.
Earlier reviews in this area include those of Howard [1], Ewing [2] and Le Roy and
Carley [3]; however, all these are now rather dated. Experimental studies of Van der
Waals spectra [4,5], their use to determine intermolecular forces [6,7] and the photodis-
sociation of Van der Waals molecules [8] are outside the scope of this article.
The structure of this article is as follows: Section 2 describes atom–atom Van der
Waals molecules and the computational methods used for calculating their vibration-
rotation energy levels. Sections 3–5 discuss the theory of atom–diatom complexes: sec-
tion 3 derives the (exact) coupled equations in both space-fixed and body-fixed coor-
dinates; section 4 describes approximate methods, and discusses the different angular
momentum coupling cases that can occur; and section 5 describes the factors influenc-
ing the intensities of spectroscopic lines. Finally, section 6 gives a brief discussion of
larger Van der Waals complexes, and indicates how the methods used for atom–diatom
complexes can be generalised.
2. Atom–atom Van der Waals molecules

The simplest Van der Waals molecules are those formed between two closed-shell
atoms, A and B, with nuclear masses mA and mB . Within the electronic Born-
Oppenheimer approximation, the potential energy for a system of this type depends
only on the internuclear separation R, and the Hamiltonian for nuclear motion, after
separating out the motion of the centre of mass, is
h̄2 2
H = − ∇ + V (R), (1)
2µ
where µ is the reduced mass, µ = mA mB /(mA + mB ), and V (R) is the intermolecular
potential energy. In spherical polar coordinates the Laplacian is
ˆl2
−1 ∂2
∇ =R
2
R− 2 (2)
∂R2 R
2
where
ˆl2 = − 1 ∂ ∂ ∂ 2
sin β sin β + (3)
sin2 β ∂β ∂β ∂α2
is the angular momentum operator for end-over-end rotation of the nuclei. The two Euler
angles α and β are the spherical polar coordinates of the internuclear vector R relative
to a Cartesian axis system fixed in space [(β, α) ≡ (θ, φ) in the usual notation]. The
Schrödinger equation involving the diatomic molecule Hamiltonian is thus separable, and
its solutions may be written
ψnlml = R−1 χnl (R)Ylml (β, α). (4)
In this equation, n is a stretching quantum number for the Van der Waals bond, l
is the angular momentum quantum number for the rotation of the nuclei, and ml is
the projection of l onto the space-fixed Z axis. The functions Ylml (β, α) are spherical
harmonics [9] satisfying the equation
ˆl2 Ylm (β, α) = l(l + 1)Ylm (β, α). (5)

l l
Note that the definition of ˆ

l2 used here does not include a factor of h̄2 , so that the eigen-
values of angular momentum operators are pure numbers. This convention is commonly
used in the published literature, but is not universal.
Substitution of equations (1) – (5) into the total Schrödinger equation gives a one-
dimensional equation for the function χnl (R),

h̄2 d2 h̄2 l(l + 1)
− + V (R) + − Enl χnl (R) = 0. (6)
2µ dR2 2µR2
Since ψnlml must be finite at R = 0 and become zero as R → ∞ for bound states, the
boundary conditions on χnl (R) are
χnl (R) = 0 at R = 0
(7)
χnl (R) = finite as R → ∞.
In practice, χnl (R) decreases exponentially with R as R → ∞ for bound states because
of the form of equation (6). The eigenvalues Enl obtained by solving this equation are
also the eigenvalues of the full Hamiltonian of equation (1).
2.1 Numerical methods

There are several methods available for solving one-dimensional Schrödinger equa-
tions such as (6). For a few simple model problems, such as the harmonic oscillator and
the Morse oscillator, the rotationless problem (l = 0) can be solved analytically, but
for most potentials of practical interest this is not possible and it is necessary to use
numerical methods.
3
An obvious approach is to use a basis set expansion for χnl (R). However, because
of the extreme anharmonicity of Van der Waals interaction potentials, a basis set expan-
sion in harmonic oscillator functions is very slowly convergent. In particular, the actual
wavefunction often dies off very slowly at large R, and most basis set methods expend a
lot of effort in trying to represent this long-range tail in terms of oscillatory basis func-
tions. Various ways around this problem have been proposed: Tennyson and Sutcliffe [10]
favour a basis set of Morse-oscillator-like functions, including continuum contributions,
and various other workers have used a nonorthogonal basis set of Gaussians functions
with centres distributed along the R axis [11].
The most common method of solving equation (6) for Van der Waals complexes
is to integrate the differential equation numerically. Solutions satisfying the boundary
conditions are propagated from long and short range to a matching distance in the
classically allowed region; the two wavefunctions and their derivatives can be made to
match only if the energy chosen is an eigenvalue. The strategy used is to integrate the
equation first using a guessed eigenvalue, and then to use the extent of the mismatch
to make an improved estimate of the eigenvalue; this is known as the shooting method
[12]. This procedure can then be iterated until the eigenvalue is known to any desired
accuracy.
A wide variety of methods is available for propagating ordinary differential equations
such as (6). Cooley [13] proposed the use of Numerov integration, which takes advantage
of the absence of first-derivative terms in the one-dimensional Schrödinger equation to
obtain an integration formula accurate to the fourth power of the step size. Cooley also
gave an energy correction formula that converges quadratically to the true eigenvalue.
Various other propagators have also been used for the one-dimensional problem [14],
but the Cooley algorithm suffices in most cases. However, it is worth mentioning the
log-derivative propagators, which are important in many-channel problems, as will be
seen below. The log-derivative methods rely on the fact that it is not actually necessary
to propagate both the wavefunction and its derivative, since the normalisation of the
wavefunction is arbitrary; it is adequate instead to propagate the logarithmic derivative
Y (R) of the wavefunction [14], defined by
dχ d ln χ
Y (R) = [χ(R)]−1 = . (8)
dR dR
The condition for a trial energy to be an eigenvalue is then that the incoming and
outgoing log-derivatives should be the same at the matching point. The log-derivative
is a discontinuous function, with poles wherever χ(R) has nodes, so that some care is
necessary is propagating it. Nevertheless, numerical methods are available [14,15,16] and
are actually more stable than those for the wavefunction itself. Log-derivative methods
are not widely used for the one-dimensional problem, but their greater numerical stability
makes them the method of choice for many-channel problems.
2.2 Spectroscopic constants

The one-dimensional Schrödinger equation for an atom–atom Van der Waals molec-
4
ule has exactly the same form as for a normal chemically bound diatomic molecule.
However, a Van der Waals molecule is typically much more weakly bound, and the in-
termolecular potential is capable of supporting only a few vibrational levels. The effects
of anharmonicity and centrifugal distortion are much more pronounced than for most
chemically bound species, and the representations necessary to describe the energy lev-
els are somewhat different. In particular, the conventional expansion of diatomic energy
levels in terms of Dunham coefficients Yij ,

Enl = Yij (n + 12 )i [l(l + 1)]j (9)
ij
is only slowly convergent for Van der Waals complexes. It is usually necessary to abandon
the power series in (n + 12 ), and to define a separate rotational expansion for each
vibrational level,
Enl = En0 + Bn l(l + 1) − Dn [l(l + 1)]2 + Hn [l(l + 1)]3 + . . . (10)
where Bn and Dn , Hn etc. are rotational and centrifugal distortion constants for vi-
brational level n. Efficient methods of calculating rotational and centrifugal distortion
constants from the potential energy curve V (R) are available [17]. If a very long progres-
sion in the rotational quantum number l is observed, the centrifugal distortion effects
may be so large that even equation (10) is inadequate. Under these circumstances, it is
necessary to solve equation (6) separately for each value of l.
For chemically bound diatomic molecules, it is usual to obtain the potential curve
V (R) by semiclassical (RKR) inversion [18,19] of (continuous) functions G(n) and B(n),
which characterise the dependence of the vibrational energy and rotational constant on
the vibrational quantum number n. This remains possible for Van der Waals complexes,
although the procedure is hampered because there are usually relatively few vibrational
levels, and this causes difficulty in interpolating G(n) and B(n). However, the shallow
potential wells of Van der Waals complexes support only a limited number of rotational
levels, and it is sometimes possible to follow Enl for a particular vibrational level all the
way to (centrifugal) dissociation. Child and Nesbitt [20] have proposed a variant of the
RKR inversion procedure that allows a potential curve to be extracted from such data.
3. Atom–diatom Van der Waals molecules

When one of the atoms in an atom–atom Van der Waals molecule is replaced by a
diatomic molecule, the dynamical problem is complicated in two ways. First, the Van
der Waals complex may be formed from a diatom in any one of its vibration-rotation
states, and each of these monomer states will give rise to a manifold of states of the
complex. Secondly, the intermolecular potential for an atom–diatom system depends on
the length of the diatom bond and on the relative orientation of the interacting species
as well as on their separation. For complexes formed from diatomic molecules with
electronic angular momentum the situation is even more complicated, but for simplicity
5
Figure 1. Coordinate system for atom–diatom Van der Waals complexes.
the treatment given below will be confined to Van der Waals molecules formed from
closed-shell atoms and molecules.
An important feature of Van der Waals complexes is that the monomers involved
retain their identity in the complex. It is thus desirable to choose a coordinate system
and a form of the Hamiltonian that reflect this. Conventional normal mode Hamiltonians
are generally not useful for Van der Waals complexes, because of the very wide amplitude
and anharmonic character of the vibrational motions.
The coordinate system most commonly used for an atom–diatom Van der Waals
complex is shown in Figure 1. The vector from the centre of mass of the diatom BC to
the atom A is denoted R, and has length R. The vector between the atoms B and C is
r , and is usually taken to originate on the heavier of atoms B and C. The length of r is
r, and the angle between R and r is θ. This is sometimes referred to as a Jacobi axis
system. The unit vectors corresponding to R and r are denoted R̂ and r̂ , and their
orientations in a space-fixed axis system are described by the Euler angles (αR , βR ) and
(αr , βr ). The use of these unit vectors as function arguments will be used below as a
shorthand for the angles defining their orientations: thus Ylm (R̂) ≡ Ylm (βR , αR ) etc.
The Hamiltonian for an atom–diatom complex is a simple generalisation of the
atom–atom Hamiltonian

h̄2 −1 ∂2 h̄2 ˆl2
H=− R R+ + V (R, r, θ) + Hmon , (11)
2µ ∂R2 2µR2
where µ is now mA mBC /(mA + mBC ) and Hmon is the Hamiltonian for the isolated
diatomic molecule BC. Unfortunately, the θ - and r -dependence of the intermolecular
6
potential destroys the separability of the Schrödinger equation, so that the dynamics of
triatomic Van der Waals complexes are much more complicated than those of diatomic
species.
3.1 Representation of the intermolecular potential

The intermolecular potential between an atom and a closed-shell diatomic molecule
is a function of the three coordinates R, r and θ . When performing dynamical calcu-
lations, matrix elements of the potential are often required between angular-momentum
eigenfunctions, and these matrix elements are greatly simplified if the potential is ex-
panded in terms of Legendre polynomials

V (R, r, θ) = Vλ (R, r)Pλ (cos θ) (12)
λ
where the functions Vλ (R, r) are known as radial strength functions. This is a com-
pletely general representation, in that any potential function can be represented in this
way if enough terms are included in the expansion.
Although equation (12) is completely general, it does not necessarily provide a
compact parameterisation of the intermolecular potential, and other functional forms
are often used instead. However, the Legendre expansion is readily obtained from any
other representation by numerical quadrature. Because of the orthogonality property of
the Legendre polynomials [21], equation (12) may be inverted to give
π
Vλ (R, r) = (λ + 2)
1
V (R, r, θ)Pλ (cos θ) sin θ dθ. (13)
0
This integral may be simply evaluated numerically to any desired accuracy using Gaus-
sian quadrature
N

Vλ (R, r) ≈ (λ + 2)
1
V (R, r, arccos xi )Pλ (xi )wi , (14)
i=1
where the quantities xi and wi are points and weights for N -point Gauss-Legendre
quadrature [21]. A minimum number of (λ + 1) quadrature points is required to extract
Legendre components up to Vλ (R, r).
For Van der Waals molecules involving homonuclear diatomic molecules, the poten-
tial is symmetric about θ = π/2,
V (R, r, θ) = V (R, r, π − θ). (15)
This causes the integral of equation (13) to vanish for odd values of λ, so that for
homonuclear diatoms the sum over λ in equation (12) is restricted to even values.
7
A further point of interest is the relationship between the intermolecular potentials
for two different isotopic species of a Van der Waals molecule [22,23]. Within the elec-
tronic Born-Oppenheimer approximation, the intermolecular potential for a particular
choice of the internuclear distances is independent of the nuclear masses. However, the
coordinates R and θ are referred to the centre of mass of the diatomic molecule, and the
position of this does depend on the nuclear masses involved. This affects the coefficients
of the Legendre expansion of the intermolecular potential: for example, although the
the potential for H2 –Ar is symmetric about θ = π/2 and thus includes only even-order
Legendre terms, that for HD–Ar includes odd-order terms when expanded about the
centre of mass of HD.
If the centre of mass shifts by a distance δ in the direction of r , the new coordinates
(R, θ) corresponding to a particular nuclear geometry are related to the old coordinates
(R , θ ) by
1
R = R(1 + t2 + 2t cos θ) 2 , (16)
1
cos θ = (cos θ + t)/(1 + t2 + 2t cos θ) 2 , (17)
where t = δ/R. Applying equation (13), the radial strength functions in the transformed
coordinate system are
π
Vλ (R, r) = (λ + 2)
1
V (R , r, θ )Pλ (cos θ) sin θ dθ. (18)
0
This integral may again be evaluated efficiently by Gaussian quadrature. It provides a

method for obtaining the Legendre expansion of the potential energy surface for one Van
der Waals molecule from the potential of an isotopically related system.
3.2 The coupled equations

A characteristic feature of Van der Waals molecules is that the intermolecular po-
tential is sufficiently weak that the constituent molecules remain identifiable. It is thus
natural to expand the wavefunction of the complex in terms of the vibration-rotation
wavefunctions of the separated monomers, and a convenient expansion for atom–diatom
systems is
ψα (R, r) = r−1 R−1 φvj (r)Φa (R̂, r̂)χα
va (R). (19)
va
Here α labels a particular quantum state of the complex, φvj (r) is the stretching wave-
function of the free diatom with vibrational quantum number v and angular momentum
j , and the functions Φa (R̂, r̂) are a complete orthonormal set of channel functions
spanning the space of the angular coordinates αR , βR , αr and βr . The index a collec-
tively labels the set of angular quantum numbers, including j . There are several possible
choices for the channel functions, which will be discussed in detail below.
When this representation of the wavefunction is substituted into the total Schrö-
dinger equation of the complex, using the Hamiltonian of equation (11), the equation
8
obtained is
h̄2 −1

∂2

h̄2 ˆl2

− R R+ + V (R, r, θ) + Evj
mon
−E
va
2µ ∂R2 2µR 2
(20)
r−1 R−1 φvj (r)Φa (R̂, r̂)χα
va (R) = 0.
∗
−1 −1
Multiplying this equation from the left by r R φv j (r)Φa (R̂, r̂) and integrating
over all coordinates except R yields the coupled equations for the system

h̄2 d2 h̄2 ˆl2
− + (va|V |va) + (a| |a) + Evj − E χα
mon
va (R)
2µ dR 2 2µR 2
(21)
h̄2 ˆl2

=− (va|V |v a ) + (a| |a )δvv χα
v a (R).

2µR 2
v a
This is a set of differential equations, one for each channel (labelled by va) included in
the basis set. Terms off-diagonal in va, which couple the different equations, have been

taken to the right hand side. The symbol indicates summation over all v a = va.
The round bracket notation ( | | ) has been adopted to indicate integration over all
dynamical variables for which the associated quantum numbers are given; thus (a|V |a )
implies integration over the angular variables only, while (va|V |v a ) implies an addi-
tional integral over the diatom stretching coordinate r .
The coupled equations are the most fundamental form of the Schrödinger equation
for Van der Waals molecules, and most of the approximate methods described below can
be derived by making simplifying approximations to the coupled equations. However,
before proceeding further it is desirable to investigate the properties of the various chan-
nel basis sets which can be used in equation (19), and to describe the evaluation of the
matrix elements involved in equation (21).
3.2.1 The space-fixed representation

There are two sources of angular momentum in an atom–diatom Van der Waals
molecule: the end-over-end angular momentum of the complex l and the internal angular
momentum of the diatom j . It would thus be possible to use as channel functions the
simple products
Φa (R̂, r̂) = Yjmj (r̂)Ylml (R̂). (22)
However, j and l couple together to form the total angular momentum J , which (in the
absence of nuclear spin) is a rigorously good quantum number because of the isotropic
nature of space. A more convenient choice of angular basis set is thus the set of simul-
9
taneous eigenfunctions of ĵ2 , ˆ
l2 , Jˆ2 and JˆZ
JM
Φa (R̂, r̂) ≡Yjl (R̂, r̂)

= jlmj ml |JM Yjmj (r̂)Ylml (R̂)
mj ml (23)

1 j l J
= (−)j−l+M (2J + 1) 2 Yjmj (r̂)Ylml (R̂),
mj ml −M
mj ml

. . .
where jlmj ml |JM is a Clebsch-Gordan coefficient and is a Wigner 3j -
. . .
symbol [9]. Equation (19) with this choice of the Φa (R̂, r̂) is known as the space-fixed
representation of the wavefunction. For given values of j and l, J may take values from
|j − l| to j + l in unit steps.
In order to calculate matrix elements of the total Hamiltonian between these angular
functions, we need to know the effects of the operators Hmon and V (R, r, θ). The space-
fixed channel functions are eigenfunctions of Hmon when combined with the diatom
stretching functions φvj (r)

−1 −1
Hmon r φvj (r)Φa (R̂, r̂) = Evj r φvj (r)Φa (R̂, r̂) ,
mon
(24)
mon
where Evj is the energy of the isolated diatom in internal state (v, j). Hmon thus has
only diagonal matrix elements between the channel functions.
If the intermolecular potential is expanded in the form (12), its matrix elements
may be expressed in terms of matrix elements of Legendre polynomials,

(vjlJ|V |v j l J) = fλ (jl; j l ; J) φvj (r)Vλ (R, r)φv j (r)dr, (25)
λ
where the Percival-Seaton coefficients fλ (jl; j l ; J) are
fλ (jl; j l ; J) = (jlJM |Pλ (cos θ)|j l JM )

1
= (−)l+l +J [(2j + 1)(2j + 1)(2l + 1)(2l + 1)] 2
(26)
l λ l j λ j l λ l
× ,
0 0 0 0 0 0 j J j
. . .
and is a Wigner 6j -symbol [9]. The potential matrix elements off-diagonal in
. . .
v have only a very small effect on bound state energy levels, but are crucial in calculations
of vibrational predissociation or vibrational relaxation rates. Because of the presence
of the two 3j -symbols with vanishing projection quantum numbers, the Percival-Seaton
coefficients are zero unless
10
1) Triangle relationships are satisfied by (j, λ, j ) and (l, λ, l )
2) Both (j + λ + j ) and (l + λ + l ) are even.
These restrictions have important consequences for the diagonal matrix elements of the
intermolecular potential. In particular,
1) Only Legendre components with λ ≤ 2j can contribute to the diagonal potential
for a given channel.
2) Odd-order Legendre terms have no matrix elements diagonal in j .
The Percival-Seaton coefficients are diagonal in both J and the parity p = (−)j+l .
The intermolecular potential cannot mix states of different J or p , and these are rigor-
ously good quantum numbers for a Van der Waals molecule. *
In the space-fixed representation, the coupled equations are thus

h̄2 d2 h̄2 l(l + 1)
− + (vjlJ|V |vjlJ) + + Evj − E χJjlJ (R)
mon
2µ dR 2 2µR 2
(27)
=− (vjlJ|V |v j l J)χJv j l (R).
v j l
3.2.2 The body-fixed representation

The space-fixed basis set described in the previous section provides a representation
in which the end-over-end angular momentum operator ˆ l2 is diagonal, and channels
of different l and j are coupled by the intermolecular potential. This is appropriate
when the anisotropy of the intermolecular potential is weak, but for strongly anisotropic
systems it is often more convenient to use a body-fixed (or molecule-fixed) basis set, in
which the intermolecular potential is block-diagonal, and most of the coupling between
channels comes from the ˆ l2 operator.
In order to obtain the body-fixed representation, the Hamiltonian is first trans-
formed to a coordinate system that rotates with the complex, rather than being fixed in
space. The orientation of the body-fixed axes (x, y, z) relative to the space-fixed axes
(X, Y, Z) is described by the Euler angles (α, β, 0), where (β, α) ≡ (βR , αR ) are the
polar coordinates of the R vector in the space-fixed axis system. The polar and az-
imuthal angles of the r vector in the body-fixed axis system are denoted θ and φ. This
coordinate system has been termed the “two-thirds body-fixed” system [24], since only
two of the three Euler angles are used in specifying the orientation of the body-fixed
axes.
* A slightly different quantity, p = (−)j+l+J , is used in much of the scattering

literature. This too is sometimes referred to as parity, and is a rigorously good quantum
number, but is sometimes less convenient than p when formulating spectroscopic selec-
tion rules. In spectroscopic terms, states with p = +1 and −1 are often labelled e and
f respectively.
11
The Hamiltonian obtained directly on transforming to the body-fixed system is
inconvenient to use, since the commutation properties of the angular momenta involved
are very complicated [24]. However, it is possible to define an isomorphic Hamiltonian
where the commutation properties are simpler [25,26]. The isomorphic Hamiltonian is
related to the true Hamiltonian by
Hiso = U HU −1 , (28)
where
U = exp(iφ ĵz ) (29)
and φ is an artificial independent variable not present in the true Hamiltonian. The
isomorphic Hamiltonian has the same eigenvalues as the true Hamiltonian [26], and may
be written in the form

h̄2 −1 ∂2 h̄2 (Jˆ − ĵ)2
Hiso =− R R+ + V (R, r, θ) + Hmon , (30)
2µ ∂R2 2µR2
Here, Jˆ is the operator for the total angular momentum, evaluated in the space-fixed
axis system and projected onto the body-fixed frame defined by the angles (α, β, φ ),
and ĵ is the body-fixed angular momentum operator for the diatomic molecule. Since the
isomorphic Hamiltonian involves one more angular coordinate than the true Hamiltonian,
its eigenfunctions are characterised by an extra body-fixed projection quantum number.
The eigenfunctions may be expanded in the angular basis set
12
2J + 1 J∗
DM K (α, β, φ )Yjk (θ, φ), (31)
8π 2
J
where DM K (α, β, φ ) is a rotation matrix element with the phase convention of Brink
and Satchler [9] and Yjk (θ, φ) is a spherical harmonic involving the angular coordinates
of the diatomic molecule in the body-fixed axis system.
Physically, the end-over-end angular momentum of the complex cannot have any
body-fixed projection along R, so that the projection of J onto the R vector must be the
same as the projection of j . The physically significant eigenfunctions of the isomorphic
Hamiltonian are thus those with K = k , and the basis set appropriate for expanding
the eigenfunctions of the true Hamiltonian is
12
1 2J + 1
ΦJM
jK (R̂, r̂) = (2π) U
−1 2 J∗
DM
K (α, β, φ )YjK (θ, φ)
8π 2
12 (32)
2J + 1 J∗
= DM K (α, β, 0)YjK (θ, φ).
4π
The artificial angle has thus disappeared in the basis functions for the true Hamiltonian,
as required.
12
The matrix elements of the intermolecular potential are particularly simple in the
body-fixed representation, and are diagonal in both J and K . They are given by
(jKJM |Pλ (cos θ)|j K JM ) = δKK gλ (jj K), (33)
where

K 1 j λ j j λ j
gλ (jj K) = (−) [(2j + 1)(2j + 1)] 2 . (34)
0 0 0 −K 0 K
The potential matrix elements are thus independent of J , unlike those of the space-fixed
representation. The presence of the first 3j -symbol in this equation again ensures that
the matrix elements vanish unless (j + λ + j ) is even and (j, λ, j ) satisfy a triangle
relationship.
The matrix elements of the operator (Jˆ − ĵ)2 may be obtained by expanding it as
follows
(Jˆ − ĵ)2 = Jˆ2 + ĵ2 − 2ĵ · Jˆ
= Jˆ2 + ĵ2 − 2ĵz Jˆz − 2ĵx Jˆx − 2ĵy Jˆy (35)
= Jˆ2 + ĵ2 − 2Jˆz2 − ĵ− Jˆ− − ĵ+ Jˆ+
where
Jˆ± = Jˆx ∓ iJˆy (36)
and
ĵ± = ĵx ± iĵy (37)
are raising and lowering operators in the body-fixed axis system, and the equivalence of
the operators Jˆz and ĵz has been used. Note the inverted sign in the definition of Jˆ± ,
which arises because the components of Jˆ referred to body-fixed axes obey “anomalous”
commutation relationships [27].
The matrix elements of (Jˆ − ĵ)2 are thus
jKJM |(Jˆ − ĵ)2 |jKJM = J(J + 1) + j(j + 1) − 2K 2

jKJM |(Jˆ − ĵ)2 |jK ± 1JM = [J(J + 1) − K(K ± 1)] 2 [j(j + 1) − K(K ± 1)] 2
1 1
= c(j; KK ± 1; J),
(38)
with all other matrix elements zero.
13
In the body-fixed representation, the coupled equations are thus

h̄2 d2
− + (vjKJ|V |vjKJ)
2µ dR2

h̄2
+ J(J + 1) + j(j + 1) − 2K 2 + Evj mon
− E χJvjK (R)
2µR 2
(39)
=− (vjKJ|V |v j KJ)χJv j K (R)
v j
h̄2
+ c(j; KK ; J)χJvjK (R),
2µR 2
K =K±1
where

(vjKJ|V |v j KJ) = gλ (jj K) φvj (r)Vλ (R, r)φv j (r)dr. (40)
λ
The primitive body-fixed basis functions as described above do not have definite
parity, except for K = 0. However, since parity is a rigorously good quantum number,
it is usually advantageous to choose basis functions that do have definite parity, and it
is straightforward to define linear combinations of the primitive functions for which this
is the case. Adopting the notation Ω ≡ |K|, these are
J∗
ΦJM
jΩ
±
( R̂, r̂) = N D MΩ (α, β, 0)YjΩ (θ, φ) ± (−)J J∗
D M −Ω (α, β, 0)Yj−Ω (θ, φ) ,
(41)
1 1
where the normalising factor N is [(2J + 1)/16π] 2 for Ω = 0 and [(2J + 1)/8π] 2 for
Ω > 0. These basis functions are referred to as the parity-adapted body-fixed basis
set; the matrix elements of the Van der Waals Hamiltonian between these functions are
readily constructed from the matrix elements between the primitive body-fixed functions.
3.2.3 Transformation between space-fixed and body-fixed representations

The space-fixed and body-fixed basis sets both span the space of the angular coor-
dinates αR , βR , αr and βr , and are related by the expressions

JM J∗
Yjl (R̂, r̂) = DM K (α, β, 0)jlK0|JKYjK (θ, φ)Yl0 (0, 0)
K
(42)
1
j−l−K j l J
= (2l + 1) 2 (−) ΦJM
jK (R̂, r̂)
K 0 −K
K
and conversely

1 j l J
ΦJM
jK (R̂, r̂) = (−)j−l−K (2l + 1) 2 JM
Yjl (R̂, r̂). (43)
K 0 −K
l
14
These equations have several important consequences:
1) The j quantum number is unaffected by the transformation. Thus, a space-fixed
basis set which includes all l values corresponding to a particular J and j spans
exactly the same space as a body-fixed basis set which contains all values of K for
that J and j . If the coupled equations are solved exactly, there is no difference
between the solutions obtained in the space-fixed and body-fixed representations.
2) If J = 0 or j = 0, the summations on the right hand sides of equations (42) and
(43) collapse to a single term. Under these circumstances, there is no difference
between the space-fixed and body-fixed representations. This is also true for the
even parity (p = +1, f symmetry) case with J = 1.
3.2.4 Solving the coupled equations

The coupled equations (27) and (39) take the general matrix form
d2 χ
= [W (R) − ]χ(R). (44)
dR2
There are in principle an infinite number of channels (basis functions); it is usual to
truncate the set to include only those channels which lie reasonably close in energy
to the state(s) of interest. This is known as the close-coupling approximation, and
calculations which make no other dynamical approximation are known as close-coupling
calculations to distinguish them from the various decoupling approximations discussed
below. If N channels are included in the expansion, W (R) is an N × N matrix and =
(2µE/h̄2 )I is a constant times the unit matrix. The physically significant bound state
wavefunction χ(R), satisfying the boundary conditions at both R = 0 and R = ∞, is
represented as a column vector with N components. However, if the boundary conditions
are neglected, there are N linearly independent solution vectors at each energy, so that
until the boundary conditions are applied it is actually necessary to propagate an N ×N
wavefunction matrix.
These equations are exactly the same as the coupled equations of molecular scatter-
ing theory, except that the boundary conditions are different for the bound state case.
There are solutions of the coupled equations satisfying scattering boundary conditions
for any energy greater than the dissociation energy of the complex, so that the scattering
problem reduces to propagating solutions of the coupled equations from one value of R
to another for a specified energy E . Many methods of doing this have been developed
[15,16,28-33], but they are adequately treated in the scattering literature and are outside
the scope of this article.
The additional problem present in the bound state case, at energies below the
dissociation energy of the complex, is that of locating energies which are eigenvalues
of the coupled equations, where a solution may be found that satisfies bound state
boundary conditions: for bound states, each component of χ must satisfy equation (7).
There are several procedures available for doing this [34-37]. The earliest, due to Dunker
and Gordon [34], is a straightforward extension of the shooting procedure used in the
15
one-dimensional case: incoming and outgoing solutions of the coupled equations are
started in the short-range and long-range classically forbidden regions, and matched in
the classically allowed region, allowing an initial estimate of the eigenvalue to be refined
iteratively. Dunker and Gordon have given an explicit generalisation of the matching
criterion, involving both the wavefunction matrix and its derivative. However, as is well
known in scattering theory, wavefunction propagation methods are subject to a classic
numerical instability, which is particularly serious for bound-state problems because it is
usually necessary to include many closed channels in the calculations. The wavefunction
component χi (R) for a locally open channel i (with Wii (R) < ) is an oscillatory
function of R, whereas that for a locally closed channel (with Wii (R) > ) is made up
of exponentially increasing and decreasing components. If there are both locally open
and locally closed channels over any range of R, there is a tendency for the closed channel
components to grow so quickly that (because of numerical rounding errors) the linear
independence of the different solutions is lost.
This problem is neatly circumvented by log-derivative methods. In the many-
channel case, the log-derivative matrix Y (R) is defined by [15]
Y (R) = χ (R)[χ(R)]−1 , (45)
where χ(R) is the N × N wavefunction matrix and the prime indicates radial differen-
tiation. The diagonal elements of the log-derivative matrix become constant when χ(R)
is exponentially increasing or decreasing, so that loss of linear independence does not
occur. In addition, as foreshadowed in the discussion of the single-channel case above,
the log-derivative matrix contains exactly the information needed to locate eigenvalues.
As before, incoming and outgoing solutions are propagated from the two classically for-
bidden regions to a matching point in the classically allowed region: at an eigenvalue,
the determinant of the difference between the two solutions is zero,
|Y in (Rmid ) − Y out (Rmid )| = 0. (46)
The strategy to be adopted in searching for the zeroes of the matching determinant has
been discussed in detail by Johnson [35] and Manolopoulos [16].
One apparent disadvantage of the log-derivative methods is that they do not di-
rectly give explicit wavefunctions, which are needed to calculate molecular properties
(via expectation values) and spectroscopic intensities (via off-diagonal matrix elements).
However, the restriction is not as serious as it might appear: a finite-difference approach
for extracting expectation values from coupled channel calculations has been described
by Hutson [38].
3.2.5 Matrix methods

An alternative approach to solving the coupled equations is to use a basis set expan-
sion for the R coordinate as well as for r and the angular variables. The angular basis
sets used in such calculations are generally the same as in coupled channel calculations.
16
This approach was pioneered for Van der Waals molecules by Le Roy and Van Kranen-
donk [39], who used numerical basis sets for the radial (R) functions. Such basis sets
are adequate for the rare gas–H2 systems, but converge very poorly for more strongly
anisotropic systems. An alternative basis set, based on Morse-oscillator-like functions,
has been used extensively by Tennyson and coworkers [10,40].
A recent development in this area has been the use of non-orthogonal basis sets
of Gaussian functions (“distributed Gaussians”) [11]. These circumvent the problem of
representing non-oscillatory regions of the wavefunction in terms of oscillatory functions,
which is the major source of poor convergence in other types of basis-set calculation.
They have been applied to simulating the spectra of a range of rare gas – hydrogen
halide Van der Waals complexes by Clary and Nesbitt [41]. A particularly promising
approach is the combination of distributed Gaussian basis sets (DGB) for the R motion
with a discrete variable representation (DVR) for the angular motion [42].
4. Approximate methods
Although it is possible to solve the Schrödinger equation exactly for a particular
potential surface, as described in the previous section, the computer time required is
often substantial, and physical insight may be lost because of the complexity of the
calculation. It is often more helpful to attempt an approximate factorisation of the
wavefunction into components recognizable as vibrations or rotations of a particular
part of the molecular framework. Calculations based on such approximations can offer
considerable savings in computer time, and also preserve the physics of the problem in
a more transparent form.
There is no single factorisation scheme that works well for all Van der Waals molec-
ules. However, most of the commonly used schemes make a basic separation into three
types of motion
(1) Vibrations of individual monomers
(2) The stretching vibration of the Van der Waals bond
(3) Angular motions, including internal rotations, bending vibrations of the Van der
Waals bond, and overall rotation of the complex.
The total energy is thus given approximately by
E tot = Evj
mon
+ Enstretch + Eaang . (47)
The vibrational motions (1) and (2) are easily visualised and the energy contributions
mon
from them are simple. To a first approximation, Evj is just the energy that the
monomers would have if isolated, and Enstretch is an eigenvalue of an effective one-
dimensional potential, as discussed in section 4.3 below. The angular motions, however,
are much more complex, and further factorisations are usually attempted; the following
section will be devoted to the various separations which can be used. It is very important
to choose the coupling scheme appropriate to the particular molecule of interest when
performing approximate calculations.
17
4.1 Angular momentum coupling cases
There are several ways in which the angular momenta in a Van der Waals molecule
may couple together, as pointed out by Bratoz and Martin [43], and the different cou-
pling schemes give rise to qualitatively different patterns of energy levels. The coupling
cases are most readily appreciated by considering a simplified model problem in which
coupling between the angular (αR , βR , αr , βr ) and radial (R, r) motions is neglected.
The Schrödinger equation for the angular motion is then

2 ˆ
bĵ + B l + V (θ) − E ΦJα (R̂, r̂) = 0,
2
(48)
where B and b are the rotational constants of the complex and of the diatomic molecule,
h̄2 −2
B= R ,
2µ
(49)
h̄2
b= r−2 ,
2µmon
and V (θ) is the expectation value of V (R, r, θ) over the radial motions
V (θ) = V (R, r, θ)

= Vλ Pλ (cos θ). (50)
λ
The isotropic potential term, V0 , merely shifts all the levels of the complex by a constant
amount relative to the levels of the free diatom; the anisotropic terms V1 , V2 etc. cause
additional shifts and splittings of the observed levels.
It is usually true for Van der Waals molecules that B b, since µ > µmon
and R > r . Denoting the dominant anisotropic term by Vaniso , three major coupling
schemes can occur, depending on the relative magnitudes of B , b and Vaniso . These will
be referred to here as coupling cases 1, 2 and 3 in order of increasing anisotropy; the
coupling cases followed by various Van der Waals molecules in their lower energy states
are given in Table 1, although it should be appreciated that the coupling case observed
can be different for different internal states of the same Van der Waals molecule. *
4.1.1 Case 1
For small anisotropies, even-order Legendre terms in the potential are much more
important than odd-order terms because only the former have diagonal matrix elements.
For very small anisotropies, V2 < B (and consequently V2 b), the space-fixed repre-
sentation of section 3.2.1 is appropriate. Both j and l are nearly good quantum numbers,
* The coupling cases referred to here as cases 1, 2 and 3 correspond to Bratoz and
Martin’s cases a, b and c. Their notation has been modified here to avoid confusion with
Hund’s coupling cases for diatomic molecules.
18
Table 1. Van der Waals complexes exhibiting different angular momentum coupling
cases.
Case 1 Case 2∗ Case 3
Rg-H2 Rg-HF Rg-I2

Rg-HD Rg-HCl Rg-N2
Rg-D2 Rg-HBr Rg-CO
Rg-OCS
* These complexes exhibit case 2 coupling in states of low angular momentum.

For higher angular momenta, a transition to case 1 is expected.
and couple together to form J . In a vector model, j and l may be considered to be pre-
cessing around the direction of J , as shown in Figure 2a, so that mj and ml are not
individually conserved.
Pure case 1 coupling occurs only in the limit of zero anisotropy. For small
anisotropies, however, the energy levels of the complex may be obtained from perturba-
tion theory in the space-fixed basis set. To first order,

J
Ejln = Evj
mon
+ Enstretch + Bl(l + 1) + Vλ jlJM |Pλ (cos θ)|jlJM . (51)
λ
The potential matrix elements are simply Percival-Seaton coefficients, equation (26),
so that the angular factors in the diagonal matrix elements are zero unless λ is even.
Neglecting V4 and higher order terms, which are usually much smaller than V2 , and
mon
substituting the diatomic rigid-rotor value for Evj , the energy expression becomes
J
Ejln = bj(j + 1) + Enstretch + Bl(l + 1) + V0 + V2 f2 (jl; jl; J). (52)
Thus for atom–diatom complexes exhibiting case 1 coupling, there are groups of energy
levels with a particular j , l and n, split by the V2 anisotropy into components of different
J . For each j and l, J can take all integer values from |j − l| to j + l.
The classic systems exhibiting case 1 coupling are the rare gas–H2 complexes, for
which the anisotropy is quite small and B is relatively large. However, even for these
systems the first-order treatment is only qualitatively valid, and it is necessary to consider
off-diagonal potential terms. Second-order perturbation theory is reasonably accurate
for these systems [39] although more sophisticated calculations are usually used [44].
Case 1 coupling begins to break down when perturbation theory in a space-fixed
basis set is no longer adequate to treat the effects of anisotropy. The nearest level that
can be coupled to |jlJM by an anisotropic potential is |jl − 2JM ; the zeroth order
energy separation between these levels is (4l − 2)B . Case 1 coupling is thus appropriate
if
V2 f2 (jl; jl − 2; J) 4Bl. (53)
19
Figure 2. Angular momentum coupling cases.
This is a quantitative criterion for the applicability of case 1 coupling; it has the inter-
esting consequence that case 1 coupling should be valid for all Van der Waals molecules
at sufficiently high values of l (and J ).
4.1.2 Case 2
Case 2 coupling occurs when the anisotropic potential terms are large compared to
the separation of levels of different l for a particular j and J , but are still small compared
to the separation between levels of different j . The end-over-end angular momentum l
is no longer even nearly conserved, but j is still nearly a good quantum number. In a
vector model, j is strongly coupled to the intermolecular axis R, with a projection K . K
then couples to the end-over-end rotation of the complex to form J , as shown in Figure
2b.
Under case 2 conditions, the energy levels may be obtained approximately from a
first-order perturbation treatment in the body-fixed basis set of section 3.2.2,
J
EjKn = Evj
mon
+ Enstretch + B[J(J + 1) + j(j + 1) − 2K 2 ]
(54)
+ Vλ jKJM |Pλ (cos θ)|jKJM .
λ
Once again there are no diagonal matrix elements of odd-order Legendre terms, and the
first-order energy level expression neglecting high-order anisotropies is
J
EjKn = bj(j + 1) + Enstretch + B[J(J + 1) + j(j + 1) − 2K 2 ] + V0 + V2 g2 (jjK), (55)
where g2 (jjK) is defined by equation (34) above. However, since V2 must be substantial
for case 2 coupling to occur at all, it is nearly always necessary in accurate calculations
to include higher-order effects due to potential terms off-diagonal in j (but diagonal in
K ).
20
In case 2 coupling, therefore, each monomer rotational level is split into groups
of levels with K = 0, ±1, ±2, . . . ± j . Within each (j, K) manifold, J can take any
value J ≥ |K|, and the resulting energy levels follow a diatomic-molecule-like energy
expression
J (0)
EjKn = EjKn + BJ(J + 1). (56)
(0)
However, it may be noted that the pattern of energy levels EjKn for different values of
K is not that appropriate to a symmetric top. The splittings between the different K
levels contain important contributions from the potential energy terms V2 g2 (jjK), in
addition to those arising from the rotational kinetic energy.
For real Van der Waals molecules, it is necessary to go beyond the case 2 limit and
consider the neglected Coriolis terms arising from B(Jˆ− ĵ)2 in second order. In the case
2 basis set, the matrix elements of (Jˆ − ĵ)2 are given by equation (38). They connect
states of the same J , j and parity, with Ω differing by ±1. The matrix elements are the
same for (+) and (−) parities, except that only (−)J (e) parity states exist for Ω = 0.
Their effect is to resolve the degeneracy between (+) and (−) parity states for Ω > 0;
the energy splitting is greatest for Ω = 1 states, because the (−)J (e) parity levels are
shifted by coupling to the corresponding Ω = 0 state, and the (−)J+1 (f ) parity states
are not. The effect of Coriolis coupling on an Ω-state first appears in the 2Ωth order of
perturbation theory.
The Coriolis matrix elements increase substantially with increasing j and J . For
small Coriolis matrix elements and well-separated case 2 states, it may be adequate to
treat the Coriolis coupling by second-order perturbation theory. Under these circum-
stances, their effect on the spectrum is simply to introduce different rotational constants
for e and f levels. The difference is usually only important for Π (Ω = 1) states, and is
characterised by an l-type doubling constant ql , such that
1 h̄2 −2
Bf = B + ql = R
2 2µ
1 h̄2 −2 (57)
Be = B − ql = R − ql
2 2µ
.
However, for sufficiently large J and j the perturbation treatment always breaks down.
Under these circumstances it is necessary to construct and diagonalise a (small) Hamil-
tonian matrix directly, as described by Lovejoy and Nesbitt [45] in the case of Ar–HCl.
The range of validity of case 2 coupling is limited by the requirement that the
off-diagonal matrix elements be small compared to the separation of the energy levels
involved. Since the Coriolis matrix elements couple levels with ∆K = ±1, this require-
ment becomes
1 1
B[j(j +1)−K(K ±1)] 2 [J(J +1)−K(K ±1)] 2 V2 [g2 (jjK ±1)−g2 (jjK)]. (58)
21
This condition is complementary to equation (53); case 2 coupling takes over from case 1
coupling as the V2 anisotropy is increased or the angular momentum quantum numbers
decrease. This changeover will be described in more detail below, but it is important
to note here that odd-order anisotropies (in particular, V1 ) play very little role in the
changeover from case 1 to case 2.
Case 2 coupling may also break down if the anisotropy is high enough to cause
significant mixing of states of different j . This will occur if
V1 g1 (j, j − 1, K) ≈ 2bj
(59)
or V2 g2 (j, j − 2, K) ≈ 4bj etc.
and under these circumstances case 3 coupling takes over as discussed below. Since
the potential matrix elements involved here are no longer diagonal in j , odd-order
anisotropies can be effective in causing the changeover from case 2 to case 3 coupling.
In contrast to the 1–2 changeover, the 2–3 changeover is not affected by the degree of
rotational (J ) excitation of the complex, since neither the potential matrix elements nor
the separation between levels of different j depend on the total angular momentum J .
However, since j is no longer a good quantum number in case 3, there can be Coriolis
matrix elements between states that are nominally labelled by different j values.
4.1.3 Case 3
Case 3 coupling occurs when the potential anisotropy is large compared to the
rotational constant of the diatom b, so that the complex is a nearly rigid molecule
executing torsional oscillations about its equilibrium geometry. Under these conditions,
neither j nor l is a good quantum number, but K is still nearly conserved. The vector
model for case 3 is similar to that for case 2, except that only the projection of j onto
R is well-defined, not j itself.
The choice of a basis set for a case 3 Van der Waals molecule is more difficult than
for cases 1 and 2. It is still often feasible to expand the wavefunction in a body-fixed (case
2) basis set, although it may be necessary to include many j levels to obtain convergence.
Alternatively, and more appropriately, a basis set of vibrational functions centred on the
equilibrium geometry may be used. The basis sets that are appropriate are different for
linear and non-linear equilibrium geometries.
For a linear equilibrium geometry, the angular Hamiltonian may be approximated
around θ = 0,
H = bĵ2 + V (θ)

b ∂ ∂ ∂2
=− 2 sin θ sin θ + + V (θ)
sin θ ∂θ ∂θ ∂α2 (60)

1 ∂ ∂2 1 ∂2
≈ −b + + 2 + V (θ).
θ ∂θ ∂θ2 θ ∂α2
22
The quantity in brackets in the last line is just the two-dimensional Laplacian in polar
coordinates; transforming to Cartesian coordinates defined by
x = sin θ cos α
(61)
y = sin θ sin α,
the Hamiltonian becomes

∂2 ∂2
H = −b + 2 + V (θ). (62)
∂x2 ∂y
Since the potential has cylindrical symmetry, it may be written as a function of ρ, where
ρ2 = x2 + y 2 = sin2 θ
(63)
V (θ) = V (ρ) = V0 + V2 ρ2 + V4 ρ4 + . . .
Note that the coefficient V2 does not have the same meaning here as in the earlier
discussion. The eigenvalues of a linear case 3 complex are thus those of a symmetric two-
dimensional oscillator. If the complex is sufficiently anisotropic that only the quadratic
potential term in equation (63) is significant in the classically allowed region, the levels
are those of a two-dimensional harmonic oscillator
ang

EvK = V0 + 2 V2 b(v + 12 ). (64)
In this limit, the bending vibrational levels are equally spaced, with the v th level having
a degeneracy of (v + 1) corresponding to states with vibrational angular momentum
K = −v, −v + 2, . . . v − 2, v . Levels of the same v but different |K| are in fact not
quite degenerate. However, it should be appreciated that the anisotropy required to reach
this “rigid” limit is very great, so that equation (64) will seldom be a good representation
of the energy levels of real Van der Waals molecules.
For a case 3 complex with a non-linear equilibrium geometry, the situation is rather
different. For an equilibrium angle θ = π/2, the angular Hamiltonian may again be
simplified
H = bĵ2 + V (θ)
2
∂ ∂2 (65)
≈ −b + + V (θ).
∂θ2 ∂α2
Since the intermolecular potential is independent of the angle α, the angular Schrödinger
equation is separable, with energy levels
ang
EvK = Ev + bK 2 . (66)
The energy levels Ev are solutions of the one-dimensional equation

d2
−b 2 + V (θ) − Ev Φv (θ) = 0. (67)
dθ
23
In the limit of large anisotropy this may again be approximated by a harmonic oscillator,
giving energy levels
ang

EvK = V0 + 2 V2 b(v + 12 ) + bK 2 . (68)
where V2 is now the quadratic potential term in an expansion about θ = π/2. For
smaller anisotropies the solutions may be obtained by the Cooley method (section 2.1),
although cyclic boundary conditions are required in the present case. The quantum
number v describes the torsional oscillations of the diatom in the plane of the complex,
while K is now a rotational angular momentum describing the free rotation of the diatom
about the molecular axis.
For a rigid atom–diatom complex with equilibrium angle θ , the rotational part of
the Hamiltonian may be written in the form [46]
h̄2 −1
Hrot = [I ]αβ Jˆα Jˆβ , (69)
2
αβ
where I is the inertial tensor of the complex. Equation (69) may be rewritten in a form
similar to that for an asymmetric top,
Hrot = AJâ2 + B Jˆb2 + C Jˆc2 + dab (Jâ Jˆb + Jˆb Jâ ), (70)
where
A ≈ h̄2 /(2µmon r2 sin2 θ) + h̄2 /(2µR2 tan2 θ) (71)

B ≈ h̄2 /2µR2 (72)

C ≈ h̄2 /(2µR2 + 2µmon r2 ) (73)

dab = h̄2 /2µR2 tan θ , (74)
For an equilibrium geometry with θ = 90◦ , dab is identically zero, and the expectation
values of equations (71) – (73) may be identified with the experimental rotational con-
stants A, B and C . It may be noted that the equations for A and B are not the same
as would be obtained by simply inverting Iaa and Ibb .
4.1.4 Transition between case 1 and case 2 coupling

The transition between case 1 and case 2 coupling is illustrated in Figure 3a, where
the energy levels correlating with a diatom in its j = 2 state are plotted as a function of
the V2 anisotropy for 0 ≤ J ≤ 4. The centre of the figure (V2 = 0) corresponds to pure
case 1 coupling, while the outer edges are near the case 2 limit. Several features may be
noted:
1) The even and odd parity levels for a particular value of Ω become more nearly
degenerate with increasing anisotropy, as the case 2 limit is approached. However,
this happens much more quickly for the higher values of Ω. This is because the
24
Figure 3a. The transition between coupling cases 1 and 2, caused by a V2 P2 (cos θ)
anisotropy. The diagram shows the energy levels for j = 2, 0 ≤ J ≤ 4 as a function of
V2 .
Figure 3b. The transition between coupling cases 1 and 2, caused by a V2 P2 (cos θ)
anisotropy. The diagram shows the energy levels for j = 2, J = 4 only as a function of
V2 .
25
essential difference between the even and odd parity manifolds is the existence of
the Ω = 0 state for (−)J (e) parity only. The other levels of this parity are shifted
by coupling to this state via the Coriolis terms in the Hamiltonian, and the low Ω
levels are shifted most because they are more directly coupled to Ω = 0.
2) The energy level pattern may be regarded as originating from avoided crossings
between pure case 2 states, which are themselves linearly dependent on the V2
anisotropy coefficient. This is more clearly seen in Figure 3b, where only the J = 4
levels are shown; the dotted straight lines show the unperturbed positions of the
pure case 2 states. The Ω = 0 and 1+ states interact strongly via Coriolis coupling,
and are shifted furthest from the unperturbed positions.
4.1.5 Transition between case 2 and case 3 coupling

The transition between case 2 and case 3 coupling may be achieved by a V1 or V2
anisotropy (or any linear combination of the two). The correlation diagrams for pure V1
and V2 anisotropies are shown in Figures 4a and 4b. The two diagrams are qualitatively
different:
1) For a V1 anisotropy of either sign, the equilibrium geometry is linear. The high-
anisotropy levels are thus those of a two-dimensional harmonic oscillator, as dis-
cussed above. The j = 0 level correlates with v = 0, while the Ω = 0 and 1
components of j = 1 correlate with v = 2, Ω = 0 and v = 1, Ω = 1 respec-
tively. The different Ω levels for v ≥ 2 remain non-degenerate even at very high
anisotropy, and for v > 4 it is quite difficult to assign the quantum numbers of the
near-rigid limit. This illustrates the general point that case 2 coupling persists to
much higher anisotropies for excited internal rotor (j ) states than for the ground
state.
2) For a V2 anisotropy, the equilibrium structure depends on the sign of V2 : it is
linear for V2 < 0 and T-shaped for V2 > 0. The correlation diagram is thus
unsymmetrical, with a level structure resembling that for V1 on the left but a quite
different structure on the right. However, since a negative V2 gives a symmetric
double well, all the levels of the V1 case are doubled, separated by a tunnelling
splitting which decreases as the anisotropy (which provides the barrier to tunnelling)
increases. The tunnelling doubling is quenched if an additional V1 anisotropy is
introduced, making the two wells non-equivalent. The levels on the right of the
diagram, by contrast, show the energy level pattern of equation (68): there is a set
of approximately equally spaced vibrational levels, each of which has built upon
it a stack of Ω (K ) levels with a separation determined by the diatom rotational
constant b.
4.1.6 Ar–HCl as a prototype system

The Van der Waals complex whose excited states have been studied in most detail
is Ar–HCl, and it is interesting to consider it in some detail. The potential energy
surface for this system has been through several cycles of refinement as better and better
26
Figure 4a. The transition between coupling cases 2 and 3, caused by a V1 P1 (cos θ)
anisotropy. The diagram shows the energy levels correlating with diatom free-rotor
states with j ≤ 4, calculated using a basis set including j values up to 12.
Figure 4b. The transition between coupling cases 2 and 3, caused by a V2 P2 (cos θ)
anisotropy. The diagram shows the energy levels correlating with diatom free-rotor
states with j ≤ 4, calculated using a basis set including j values up to 12.
27
experimental information became available. At each stage, the predictions obtained from
the best-fit potential have been useful in guiding further experiments.
The first reasonably accurate potential was that of Holmgren et al. [47], who de-
termined an anisotropic potential by least-squares fitting of a parameterised form to
molecular beam microwave and radiofrequency spectra of the Ar–HCl complex in its
ground vibrational state. The resulting potential (potential IIb of ref. 47, designated the
HWK potential here) had an absolute well depth of around 180 cm−1 , with a linear Ar–
H-Cl equilibrium geometry. However, this potential was in marked disagreement with
potentials determined from pressure broadening of HCl rotational lines by Ar [48,49],
and was itself unable to reproduce the pressure broadening results. Accordingly, Hutson
and Howard [50,51] obtained new potentials (the M3 and M5 potentials) by simulta-
neous fitting to molecular beam spectra, pressure broadening cross sections and second
virial coefficients. The M3 and M5 potentials were quite similar to the HWK poten-
tial in the region of the absolute minimum, but had a much more anisotropic repulsive
wall, which allowed them to model the pressure broadening results. However, the M3
and M5 potentials were quite different from one another in the region of the alternative
linear geometry, Ar–Cl-H; the potential in this region had little effect on the calculated
microwave and radiofrequency spectra, because the ground state bending wavefunction
did not extend to high enough angles. This region of the potential was therefore not
reliably determined by the data then available. Hutson and Howard [51] suggested that
far-infrared spectroscopy would provide the most satisfactory diagnostic of the presence
or absence of the secondary minimum, and gave calculated spectroscopic frequencies for
the different potentials.
In the last few years, high-resolution far-infrared spectroscopy of Van der Waals
complexes such as Ar–HCl has at last become experimentally feasible [52-58], using
either laser Stark resonance or tunable far-infrared lasers. Near-infrared spectra of Ar–
HCl have also been measured [59,45]; the vibrational frequencies for states of the complex
correlating with HCl (v = 1) are very similar to those for states correlating with HCl
(v = 0). The far-infrared spectra have been used to determine a new intermolecular
potential (the H6(3) potential [60]) which is reliable over the whole range of angles.
The lowest few bending levels of Ar–HCl, calculated from the H6(3) potential using
close-coupling calculations [60], are shown in Figure 5. The observed pattern may be
compared with that expected for free rotation of the HCl (case 1 or 2) and for a near-
rigid linear molecule (case 3). It may be seen immediately that the free-rotor picture is
much closer to reality. Since the Ω quantum number is well-conserved, at least for low
J , Ar–HCl is best viewed as a case 2 complex, with quantum numbers j , Ω, J and p .
The most striking feature of Figure 5 is that the first excited Σ (Ω = 0) state
actually lies below the lowest Π (Ω = 1) state. This is not at all the behaviour expected
for a near-rigid molecule: in the language usually applied to linear triatomics, the Π
state is the fundamental bending vibration, labelled 011 0, and the Σ state is its overtone,
labelled 020 0. However, in the free-rotor (case 2) picture, both these states correlate
with j = 1, and are expected to be degenerate except for the potential term V2 g2 (jjK)
28
Figure 5. Bending energy levels of Ar–HCl, calculated using the H6(3) potential, and
the patterns expected for a free internal rotor and for a semirigid linear molecule.
in equation (55). Since V2 is negative in Ar–HCl, the Σ state lies below the Π state.
Ar–HCl also illustrates the changeover between case 2 and case 1 coupling. The
j = 1, Ω = 0 and 1 states discussed above are reasonably well separated, so that
the Coriolis matrix elements between them can be treated by perturbation theory, at
least at low J , although even here Lovejoy and Nesbitt [45] have found it necessary
to go beyond perturbation theory at high J . In addition, because j is not in fact a
good quantum number, there is a significant Coriolis matrix element between the first
excited stretching state (nominally j = 0) and the Π bending state (nominally j = 1)
which would not be present for pure case 2 coupling. Since these two states are only 1.5
cm−1 apart, the Coriolis mixing is quite substantial and results in the stretching band
acquiring significant spectroscopic intensity.
Another consequence of the breakdown of the j quantum number is that the ∆j =

±1 selection rule, which applies rigorously in free HCl, is only approximate in Ar–HCl.
Thus the fundamental band in the near-infrared spectrum, which is nominally j = 0 ←
0, is of comparable intensity to the j = 1 ← 0 combination bands in the complex [45].
Similarly, bands of the complex corresponding to j = 2 ← j = 0 transitions are weakly
allowed [60]. The j = 2 states also illustrate incipient case 1 coupling, because the j = 2,
Ω = 0 and 1 states are very closer together (though the Ω = 2 state is much further
away). The Ω = 0 and 1 states are thus thoroughly mixed by Coriolis coupling, and
exhibit coupling intermediate between case 1 and case 2. A systematic theoretical study
of the spectra of the rare gas – HCl complexes, using simulations of spectra to illustrate
the effects of anisotropy on band positions and intensities, has been carried out by Clary
and Nesbitt [41].
29
4.2 Centrifugal decoupling or helicity decoupling
One of the most useful approximations to the close-coupled equations is the centrifu-
gal decoupling (CD) or helicity decoupling (HD) approximation [61,62]. This consists of
neglecting the off-diagonal Coriolis terms in the body-fixed coupled equations, so that
the coupled equations become diagonal in K . This is analogous to the coupled states
(or centrifugal sudden) approximation of molecular scattering theory [63,64], although
for bound states it is usual to include the diagonal Coriolis terms exactly, rather than
approximating them as in coupled states calculations. The CD approximation factorises
the large set of coupled equations for each J into a series of smaller sets for each allowed
value of K . For high J and a given value of jmax , it reduces the number of channels N
from (jmax + 1)2 to jmax + 1 − K . Since the computational effort involved in solving
a set of coupled equations is proportional to N 3 , the CD approximation often provides
considerable savings in computer time.
The CD approximation is appropriate for Van der Waals molecules exhibiting case
2 or case 3 coupling, where K is a good quantum number, but its results are significantly
in error for case 1 complexes. Since all Van der Waals molecules exhibit case 1 coupling
for sufficiently high rotational states, caution must be exercised in using the CD approx-
imation for high values of J . An estimate of the validity of the CD approximation may
be obtained by considering whether the condition (58) holds for the state of interest,
using appropriate average values for B and V2 .
Since the different K blocks are decoupled from one another in the CD approxi-
mation, and the angular coefficients g2 (jj K) are independent of the sign of K , states
that differ only in the sign of K are degenerate in this approximation. Consequently, the
parity-adapted linear combinations with Ω = |K| and p = ±1 (equation (41)) are also
degenerate. This degeneracy is an artefact of the CD approximation, and does not hold
for the exact solutions, although it is a good approximation for complexes exhibiting case
2 or case 3 coupling. The Coriolis terms may be included in a subsequent calculation
step if necessary [65].
4.3 Distortion approximations

The CD approximation simplifies the coupled equations, but does not completely
decouple them. The distortion approximations, on the other hand, neglect all the off-
diagonal terms in the coupled equations, and thus result in uncoupled differential equa-
tions of the form

h̄2 d2
− + Vva (R) − Evan χvan (R) = 0. (75)
2µ dR2
Distortion approximations may be obtained in either the space-fixed or body-fixed rep-
resentations. The space-fixed distortion (SFD) approximation consists of neglecting all
the off-diagonal potential terms in the space-fixed coupled equations (27). This gives
the effective potential
h̄2 l(l + 1)
SFD
VvjlJ (R) = (vjlJ|V |vjlJ) + + Evj
mon
. (76)
2µR 2
30
Conversely, the body-fixed distortion (BFD) approximation is obtained by neglecting
all off-diagonal potential and Coriolis terms in the body-fixed coupled equations (39),
giving
h̄2
BFD
VvjKJ (R) = (vjKJ|V |vjKJ) + J(J + 1) + j(j + 1) − 2K 2
+ Evj
mon
. (77)
2µR 2
Equations (76) and (77) are simple one-dimensional Schrödinger equations identi-
cal in form to equation (6), and may be solved using the methods of section 2.1. All
bound stretching (n) states for a particular (j, l, J) or (j, K, J) channel are taken to be
supported by the same effective potential curve, which includes a centrifugal term and
the diagonal matrix elements of the isotropic and anisotropic potentials. However, since
only even order Legendre polynomials can have non-zero diagonal matrix elements, the
distortion approximations completely neglect the effects of odd-order potential terms.
The distortion approximations are very simple to apply, and are valuable for pro-
viding preliminary estimates of the energy levels for a proposed potential surface, and
for obtaining physical insight into the results of more accurate calculations. Space-fixed
distortion calculations are reasonably accurate for case 1 complexes such as the rare
gas–H2 complexes, and body-fixed distortion calculations give a useful first estimate of
the energy levels of case 2 complexes such as Ne–HCl. They are not usually adequate
for quantitative work, and are not nowadays used much for atom–diatom systems, since
coupled channel calculations are fairly cheap. However, analogous methods are some-
times used for larger systems, where coupled-channel calculations may be prohibitively
expensive.
4.4 Perturbation theory

The distortion approximations neglect all the off-diagonal matrix elements in either
the space-fixed or the body-fixed coupled equations. However, these neglected terms
may be reintroduced by perturbation theory, using the wavefunctions of the distortion
approximations as the zeroth order states. Representing the distortion eigenfunctions by
the kets |van, where a collectively represents the angular quantum numbers (j, Ω, J, p)
or (j, l, J, p), we have
van|H |v a n (0)
ψvan = ψvan
(0)
+ (0) (0)
ψv a n , (78)
v a n Evan − Ev a n
|van|H |v a n |2
Evan = (0)
Evan + (0) (0)
. (79)
v a n Evan − Ev a n
The perturbation Hamiltonian H contains all the off-diagonal terms in the coupled
equations. Note that there is no first-order correction to the energy in either the SFD or
the BFD case, since all the diagonal terms are included in zeroth order. These equations
31
are readily generalised to higher order, but it is usually found that the results converge
slowly, if at all, beyond second order.
A special problem for Van der Waals complexes is that, since there are relatively few
stretching states n, the set of zeroth-order bound states |van may not be sufficiently
complete for the sum to converge. In physical terms, this means that the perturbation
correction to the wavefunction has significant contributions from continuum states. A
method for circumventing this problem, based on solving inhomogeneous differential
equations instead of using the summation over the zeroth-order solutions, has been
described by Hutson and Howard [66].
4.5 Adiabatic approximations

It is also possible to decouple the problem by performing an adiabatic or Born-
Oppenheimer separation with respect to one coordinate, usually R. In this approxi-
mation, the wavefunction is written in the single product form
BO
ψan (R, r) = r−1 R−1 Φa (R̂, r; R)χBO
an (R), (80)
where the functions Φa (R̂, r; R) are eigenfunctions of the fixed-R Hamiltonian with
eigenvalues Ua (R),
h̄2 ˆl2
Hmon + + V (R, r, θ) − Ua (R) Φa (R̂, r; R) = 0. (81)
2µR2
If the effect of the radial kinetic energy operator on the functions Φa (R̂, r; R) is ne-
glected, the wavefunctions and energies are the solutions of the single-channel equation

h̄2 d2
− + Ua (R) − Ean χBO
BO
an (R) = 0. (82)
2µ dR 2
The eigenvalues of the fixed-R Hamiltonian thus provide a set of effective potentials
(adiabats) Ua (R) for the stretching motion. The method used is to solve the fixed-R
equation (81) on a grid of R values, and then to solve the resulting one-dimensional
equations for χBOan (R) and Ean . This approach was first applied to Van der Waals
BO
complexes by Levine et al. [67], using a space-fixed basis set, and subsequently rederived
by Holmgren et al. [68] in a body-fixed basis set. The body-fixed formulation has the
major advantage that the Coriolis coupling terms can be neglected in equation (81),
giving adiabatic curves with fewer avoided crossings. The criterion for validity of the
adiabatic approximation is that the angular functions Φa (R̂, r; R) should be slowly-
varying functions of R: this does not necessarily require that the angular frequencies
should be much faster than the stretching frequency. Adiabatic approximations are often
quite accurate for the ground state of Van der Waals complexes, but become progressively
less accurate for excited states, because of the presence of avoided crossings between
different adiabatic curves.
32
It is not necessary to ignore the effect of the radial kinetic energy operator on the
angular functions completely: both diagonal [68] and off-diagonal [69] corrections may be
included if desired. A distinction which is sometimes made is that calculations in which
no correction is included are termed Born-Oppenheimer calculations, whereas those in
which the diagonal correction is included are termed adiabatic calculations.
An alternative adiabatic approximation is to carry out the separation with respect
to the internal angular coordinate θ instead of R [51]; this has been termed the reversed
Born-Oppenheimer approximation. In this approach, a one-dimensional stretching equa-
tion is first solved on a grid of θ values, and the eigenvalues of the fixed-θ equations
provide an effective potential for the bending motion. The reversed Born-Oppenheimer
approximation is particularly useful in obtaining qualitative insight into the nature of
Van der Waals bending states, since it allows them to be considered independently of
the stretching motions.
5. Dipole moments and spectroscopic intensities

Both the molecular Stark effect and the intensities of spectroscopic lines depend
on matrix elements of the dipole moment operator between molecular eigenfunctions.
The radial (R and r ) parts of these integrals are straightforward to calculate once the
wavefunctions are known, but the angular factors involved are quite complicated. This
section will discuss the evaluation of the angular factors appearing in dipole moment
matrix elements.
The dipole moment of a Van der Waals molecule is a vector function of the nuclear
coordinates; thus for an atom–diatom Van der Waals molecule the dipole moment is
denoted µ(R, r). There may be contributions to µ from 3 distinct sources:
1) The permanent dipole moments of the monomers
2) Dipole moments induced in one monomer by the electric field due to the non-
spherical charge distribution of the other. *
3) Distortion of the electronic structure of the monomers due to “chemical” interac-
tions such as electron overlap or charge transfer
For most Van der Waals molecules in their ground electronic states the dipole moment
function is dominated by the first and second effects, although the third may become
important at short range.
In order to investigate the properties of the dipole moment operator, it is convenient
to expand the dipole moment function in terms of angular functions with well-defined
angular momentum properties. This may be done in either space-fixed or body-fixed
representations, as described in the following sections.
5.1 Space-fixed
The dipole moment vector of a Van der Waals molecule may be resolved into com-
ponents along the space-fixed X , Y and Z directions. However, it is more convenient to
* Dispersion contributions to dipole moments are usually very small
33
Table 2. Symmetries of multipole moment contributions in atom–diatom Van der
Waals molecules.
Origin j l J
Direct dipole moment of diatom 1 0 1

Charge-induced dipole 0 1 1
Dipole-induced dipole 1 2 1
Quadrupole-induced dipole 2 3 1
Direct quadrupole moment of diatom 2 0 2
use linear combinations of these components
µ0 = µZ
1 (83)
µ±1 = ∓ √ (µX ± iµY ).
2
JM
The Yjl (R̂, r̂) functions of equation (23) form a complete basis set for the angular
coordinates in an atom–diatom system, so that the components of the µ vector may be
expanded in terms of them,

µM (R, r) = jl (R, r)Yjl (R̂, r̂).
µSF 1M
(84)
jl
The summation may be restricted to terms with J = 1 because µ must transform as

a vector with respect to rotations of the space-fixed axes, and to terms with (j + l)
odd since the dipole moment function has odd parity. Permanent and induced dipole
moments in atom–diatom Van der Waals molecules are very simply expressed in the
space-fixed representation, and the symmetries of the most commonly encountered terms
are given in Table 2.
The matrix elements of these expansion functions between space-fixed basis func-
tions may be obtained from the Wigner-Eckart theorem [9,70]
(ji li Ji Mi |Yjl
1M
|jf lf Jf Mf )

Ji 1 Jf
= (−)ji +li +Ji +Mi
−Mi M Mf
+ 1)(2li + 1)(2lf + 1)] 2 (85)
1
× [3(2Ji + 1)(2Jf + 1)(2ji + 1)(2jf
⎛ ⎞
Ji Jf 1
ji j jf li l l f ⎝
× ji jf j ⎠ .
0 0 0 0 0 0
li lf l
34
Since the dipole moment function has odd parity, it has no non-zero diagonal matrix
elements. In addition, the 3j -symbols with zero projections ensure that the off-diagonal
matrix elements are zero unless
1) Triangle relationships are satisfied by (ji , j, jf ) and (li , l, lf ).
2) (ji + j + jf ) and (li + l + lf ) are both even.
The matrix element is also zero if |Ji − Jf | > 1, because of the triangle relationship for
the 3j -symbol involving the total angular momentum.
Equation (85) immediately gives the spectroscopic selection rules for atom–diatom
Van der Waals molecules exhibiting case 1 coupling. For the Ar-H2 complex, for exam-
ple, the transition moment is dominated by the dipole moment induced on Ar by the
quadrupole moment of H2 . The quadrupole-induced dipole transforms as Y23 1
(Table 2),
so that the selection rules are
∆J = 0, ±1
∆j = 0, ±2 (86)
∆l = ±1, ±3.
These selection rules rely on two assumptions:

1) The contributions from dipole moments due to other sources must be small.
2) The j and l quantum numbers must provide a good description of the wavefunctions.
Both these assumptions are valid for Ar-H2 , which is a weakly anisotropic system near
the case 1 limit. However, it is important to realise that space-fixed selection rules such
as these are not applicable to Van der Waals molecules exhibiting coupling cases 2 or 3.
5.2 Body-fixed
For Van der Waals molecules following case 2 or case 3 quantisation, it is more
convenient to represent the dipole moment function in terms of the body-fixed functions
of equation (32). The dipole moment is expanded

µM (R, r) = jK (R, r)ΦjK (R̂, r̂).
µBF 1M
(87)
jK
Once again, the summation includes only functions with J = 1 and odd parity because
of the requirement that µ transform as a vector in the space-fixed axis system.
The space-fixed and body-fixed basis functions are related by equations (42) and
(43). These may readily be inverted to obtain the relationship between the body-fixed
and space-fixed coefficients

1 j l 1
jK (R, r) =
µBF (−)j−l−K (2l + 1) 2
jl (R, r).
µSF (88)
K 0 −K
l
35
The matrix elements of the Φ1M
jK (R̂, r̂) functions in the body-fixed representation
are
(ji Ki Ji Mi |Φ1M
jK |jf Kf Jf Mf )
1
= (4π)−1 (−)Mi −Ki [3(2Ji + 1)(2Jf + 1)(2ji + 1)(2jf + 1)(2j + 1)] 2

Ji 1 Jf Ji 1 Jf
×
−Ki K Kf −Mi M Mf

ji j jf ji j jf
× .
−Ki K Kf 0 0 0
(89)
For complexes containing dipolar monomers, the principal contribution to the transition
dipole is just the monomer dipole moment, which has components with j = 1, K =
0, ±1. If j is a good quantum number for the complex (pure case 2 coupling), the
resulting selection rules are
∆J = 0, ±1
∆j = ±1 (90)
∆K = 0, ±1.
In addition, the total parity of the wavefunction must change in a transition, so that only
e ↔ e and f ↔ f transitions are allowed for ∆J = ±1, and only e ↔ f transitions for
∆J = 0. The band structures are thus as expected for bending transitions in a linear
molecule: Σ ↔ Σ bands have P and R branches, while Σ ↔ Π bands have P, Q and
R branches; the P and R branches involve e levels of the Π state, while the Q branch
involves f levels, which have a different rotational constant.
For real Van der Waals molecules, j is not a good quantum number, so that the
selection rules above start to break down. In particular, it should be noted that they
predict that the “fundamental” band, correlating with a pure vibrational transition of the
monomer (i.e. v, j = 1, 0 ← 0, 0) should be forbidden (as it is for the free monomer). It is
indeed true that, for weakly anisotropic systems such as Ne–HCl [71], the fundamental
band is considerably weaker than the 1, 1 ← 0, 0 bands. However, one effect of the
anisotropy is to mix the j levels, and the fundamental band becomes allowed: for Ar–
HCl, its intensity is comparable to that of the bending bands. For the same reason,
bands such as 1, 2 ← 0, 0 are also weakly allowed. There are also significant induced
dipole terms in the transition moment, which can also cause the selection rules (90) to
be relaxed.
Van der Waals stretching bands are also of importance. In principle, if the bending
and stretching motions were independent, the stretching bands would acquire intensity
only because the anisotropy of the intermolecular potential varies with R, so that the
dipole moment of the complex is also a function of R. However, there is often quite
strong mixing between bending and stretching states, and the stretching bands may
borrow intensity from the bends. In Ar–HCl, for example, the lowest stretching state
lies at 32.4 cm−1 , and is strongly mixed by Coriolis coupling to the Π bending state at
36
34.0 cm−1 . As a result, the stretching band has significant intensity. A similar effect
occurs in Ar–HF [72], where the overtone of the Van der Waals stretch, n = 2, lies close
to the Π bend and acquires intensity from it.
6. Larger systems
Although atom–diatom systems have been studied in most detail, there has also
been some work on more complex systems. The principles governing the dynamics of
such systems are similar to those for atom–diatom systems, but the angular momentum
algebra is considerably more complicated. In addition, the number of possible coupling
schemes is considerably greater. Again, most of the techniques used for Van der Waals
complexes have direct analogues in the scattering literature. Three different classes of
complex will be considered briefly here: atom–polyatom complexes, molecule–molecule
complexes, and trimeric systems. A fuller account of the dynamics in such systems will
be published separately [73].
A general feature of Van der Waals complexes is that they exhibit wide-amplitude
motion, and are seldom well-described by near-rigid models. For larger systems, there are
frequently a number of equivalent geometries which can be interchanged by vibrational
motions. The resulting symmetries give rise to tunnelling splittings in the observed
spectra. A major advantage of computational methods based on an expansion in free-
rotor functions for the monomers is that these symmetries are taken into account in a
natural way. However, even when the potential anisotropy is too strong for calculations
based on monomer rotational functions to be feasible, the symmetries still exist and
must be taken into account. Conventional group-theoretical methods, based solely on
the point-group symmetry of the equilibrium geometry, are not adequate under such
circumstances, and it is necessary to consider instead the complete molecular symmetry
group arising from the permutation-inversion symmetry of the complex [74,75].
6.1 Atom–polyatom systems

Complexes formed from atoms and linear polyatomic molecules are very similar
to atom–diatom systems: the coupled equations are identical, and the same angular
momentum coupling schemes apply. The only added degree of complexity is that per-
pendicular vibrations of the polyatomic monomer are possible, and these introduce an
extra quantum number (l or k ) for the monomer vibrational angular momentum. Such
states are analogous to those arising from k > 0 states of a symmetric top monomer, as
discussed below.
Atom–nonlinear molecule complexes are of two basic types: atom–symmetric top
and atom–asymmetric top. Several such complexes have been studied through their
pure rotational spectra, but high-resolution infrared spectra, involving excitation of Van
der Waals bending and stretching modes, are only just starting to become available [76].
There has also been a fair amount of theoretical work on the photodissociation spectra of
such systems [77], but this has concentrated on the rates of photodissociation processes
rather than on the energy level patterns.
37
As for atom–diatom complexes, three different angular momentum coupling schemes
may be envisaged, and different sets of quantum numbers are appropriate in each case.
The only system that has been studied in any detail is Ar–H2 O, which is close to the case
2 limit [78]: the anisotropy of the potential splits and shifts the H2 O free-rotor levels,
but the free-rotor quantum numbers are still approximately conserved. Since H2 O is an
asymmetric top, the energy level pattern is considerably more complicated than for an
atom–diatom system, but can nevertheless be understood in terms of a similar physical
picture.
Polyatomic hydrides are a special case in that they have quite large rotational
constants. It is to be expected that most Van der Waals complexes involving non-
hydride monomers will be much closer to the case 3 limit, and that their spectra will be
best interpreted in terms of conventional near-rigid quantum numbers.
6.2 Molecule–molecule systems

The general molecule–molecule case, with two polyatomic fragments, has been very
little studied. Brocks et al. [24] have derived the Hamiltonian for such a system in body-
fixed coordinates, but have not performed actual calculations. This is a very important
area, and high-resolution spectra are available for systems such as the H2 O dimer [79].
There has been a good deal of work on understanding these spectra, especially with
regard to the symmetries of the states involved [80]. Coker and Watts [81] have simulated
vibrational spectra of the water dimer using a semiempirical potential [82] obtained from
a variety of experimental data. However, understanding the detailed spectroscopy and
angular momentum coupling in such systems remains a research topic for the future.
There has been rather more work on the specific case of diatom–diatom systems,
mostly aimed at understanding the spectrum of the HF dimer. For a diatom–diatom
complex, there are three sources of angular momentum: the rotation of each of the
monomers, characterised by quantum numbers j1 and j2 , and the end-over-end rotation
of the complex as a whole. There are therefore many possible coupling schemes. The
simplest is the fully space-fixed coupling scheme, in which j1 and j2 first couple to
give a resultant j12 , which then couples with the end-over-end angular momentum l to
give a resultant J , the total angular momentum. The Hamiltonian corresponding to this
scheme is a straightforward generalisation of equation (11) [83,84] and this is the simplest
representation in which to set up the coupled equations. Unfortunately, the space-fixed
quantum numbers (j1 , j2 , j12 , l, J ) bear little relationship to reality, so calculations in
this representation are only meaningful if all the l channels for each (j1 , j2 , j12 , J ) are
included, and the number of such channels required for convergence is enormous.
There are at least two ways of formulating the diatom–diatom problem in body-
fixed coordinates. In the first, j1 and j2 are each quantised along R, with projections
k1 and k2 . The total projection quantum number, K = k1 + k2 , is nearly conserved,
but the potential mixes basis functions with different k1 and k2 . This approach has
been used by Barton and Howard [85], who performed adiabatic calculations on HF
dimer and obtained an anisotropic intermolecular potential from microwave spectroscopic
38
data. Alternatively, the coupling of j1 and j2 to form j12 may be retained, but j12 is
then quantised in the body-fixed frame with projection K along R. This approach has
been used by Danby [37] in coupled-channel calculations on the H2 dimer. There is no
clear reason for preferring either of these formulations over the other: both provide a
K quantum number, which is indeed a nearly good quantum number for (HF)2 , but
unfortunately neither k1 and k2 nor j12 is particularly useful in analysing the spectrum.
6.3 Trimeric systems

The study of the dynamics of trimeric Van der Waals systems is in its infancy,
although there has been a lot of work on H+ 3 and its isotopic variants [86], which have
some similarities with Van der Waals clusters, at least for their higher vibrational levels.
Once again, for sufficiently strongly bound trimers, conventional near-rigid theory may
be employed; the systems that cause difficulty are those for which internal rotations or
exchange of identical monomers are feasible, so that conventional Hamiltonians involving
an expansion about the equilibrium geometry are inappropriate.
The methods used for atom–diatom systems are not generally suitable for trimeric
systems. For a system such as Ne3 , there is little reason to treat one pair of Ne atoms dif-
ferently from the third, and an expansion in Ne–Ne vibration-rotation functions φvj (r)
as in equation (19) would not be expected to be quickly convergent. However, sym-
metrised normal coordinates are not suitable either, because such coordinate systems
are usually not single-valued for all possible geometries. One possible way around this
problem is to use hyperspherical coordinates [87], in which the three internal coordi-
nates are chosen to be two angles and a hyperradius ρ rather than one angle and two
distances. This allows the three-particle Schrödinger equation to be cast as a set of cou-
pled differential equations in the hyperradius, and the same numerical techniques may
be used as in the atom–diatom case. Frey and Howard [88-90] have attempted to solve
the coupled equations by making an adiabatic separation with respect to ρ, and have
included non-adiabatic corrections to the ground-state energy by perturbation theory.
Hutson and Jain [91] have solved the coupled equations directly, using the log-derivative
propagator. These methods are promising, but very demanding of computer time, and
it has not yet been possible to include overall rotation of the complex.
Systems such as Ar2 HCl and Ar2 HF are of great interest, because they offer the
hope of a spectroscopic determination of non-additive contributions to intermolecular
potentials. Microwave spectra of these systems have been observed [92,93] and it should
be possible to obtain infrared spectra. A complete solution of the dynamical problem
for such systems is beyond our capabilities at present, but Hutson et al. [94] have carried
out a preliminary study, investigating the hindered rotation of an HCl molecule under
the influence of a (fixed) pair of Ar atoms. They concluded that infrared spectra of
these systems would probably contain enough information to determine the effects of
three-body forces. This is likely to be an important research topic in the future.
Acknowledgment
This work is supported by the Science and Engineering Research Council.
39
References
[1] B. J. Howard, MTP International Review of Science, Series 2, Physical Chemistry
2 (1975) 93.
[2] G. E. Ewing, Can. J. Phys. 54 (1976) 487.
[3] R. J. Le Roy and J. S. Carley, Adv. Chem. Phys. 42 (1980) 353.
[4] A. C. Legon and D. J. Millen, Chem. Rev. 86 (1986) 635.
[5] D. J. Nesbitt, Chem. Rev. 88 (1988) 843.
[6] A. D. Buckingham, P. W. Fowler and J. M. Hutson, Chem. Rev. 88 (1988) 963.
[7] J. M. Hutson, Ann. Rev. Phys. Chem. 41 (1990) 123.
[8] K. C. Janda, Adv. Chem. Phys. 60 (1985) 201.
[9] D. M. Brink and G. R. Satchler, Angular Momentum: 2nd ed., Clarendon Press,
Oxford (1968).
[10] J. Tennyson and B. T. Sutcliffe, J. Chem. Phys. 77 (1982) 4061.
[11] I. P. Hamilton and J. C. Light, J. Chem. Phys. 84 (1986) 306.
[12] W. H. Press, B. P. Flannery, S. A. Teukolsky and W. T. Vetterling, Numerical
Recipes: Cambridge University Press, Cambridge (1986).
[13] J. W. Cooley, Math. Comput. 15 (1961) 363.
[14] B. R. Johnson, J. Chem. Phys. 67 (1977) 4086.
[15] B. R. Johnson, J. Comp. Phys. 13 (1973) 445.
[16] D. E. Manolopoulos, J. Chem. Phys. 85 (1986) 6425; D. E. Manolopoulos, Ph. D.
thesis, Cambridge University (1988).
[17] J. M. Hutson, J. Phys. B 14 (1981) 851.
[18] R. Rydberg, Z. Phys. 73 (1931) 326; Z. Phys. 80 (1933) 514; O. Klein, Z. Phys. 76
(1932) 226; A. L. G. Rees, Proc. Roy. Soc. (London) 59 (1947) 998.
[19] J. Tellinghuisen, Comp. Phys. Commun. 6 (1974) 221.
[20] M. S. Child and D. J. Nesbitt, Chem. Phys. Lett. 149 (1988) 404.
[21] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions: Na-
tional Bureau of Standards (1965).
[22] H. Kreek and R. J. Le Roy, J. Chem. Phys. 63 (1975) 338.
[23] W.-K. Liu, J. E. Grabenstetter, R. J. Le Roy and F. R. McCourt, J. Chem. Phys.
68 (1978) 5028.
[24] G. Brocks, A. van der Avoird, B. T. Sutcliffe and J. Tennyson, Mol. Phys. 50
(1983) 1025.
[25] J. T. Hougen, J. Chem. Phys. 36 (1962) 519.
[26] J. K. G. Watson, Mol. Phys. 19 (1970) 465.
[27] J. H. van Vleck, Rev. Mod. Phys. 23 (1951) 213.
[28] R. G. Gordon, J. Chem. Phys. 51 (1969) 14.
[29] W. N. Sams and D. J. Kouri, J. Chem. Phys. 51 (1969) 4809; J. Chem. Phys. 51
(1969) 4815.
[30] J. C. Light and R. B. Walker, J. Chem. Phys. 65 (1976) 4272; E. B. Stechel, R. B.
Walker and J. C. Light, J. Chem. Phys. 69 (1978) 3518.
[31] G. A. Parker, T. G. Schmalz and J. C. Light, J. Chem. Phys. 73 (1980) 1757.
40
[32] M. H. Alexander, J. Chem. Phys. 81 (1984) 4510.
[33] M. H. Alexander and D. E. Manolopoulos, J. Chem. Phys. 86 (1987) 2044.
[34] A. M. Dunker and R. G. Gordon, J. Chem. Phys. 64 (1976) 4984.
[36] M. Shapiro and G. G. Balint-Kurti, J. Chem. Phys. 71 (1979) 1461.
[37] G. Danby, J. Phys. B 16 (1983) 3393.
[38] J. M. Hutson, Chem. Phys. Lett. 151 (1988) 565.
[39] R. J. Le Roy and J. van Kranendonk, J. Chem. Phys. 61 (1974) 4750.
[40] J. Tennyson, Comp. Phys. Reports 4 (1986) 1.
[41] D. C. Clary and D. J. Nesbitt, J. Chem. Phys. 90 (1989) 7000.
[42] Z. Bačić and J. C. Light, J. Chem. Phys. 85 (1986) 4594; J. Chem. Phys. 86 (1987)
3065.
[43] S. Bratoz and M. L. Martin, J. Chem. Phys. 42 (1965) 1051.
[44] G. C. Corey, R. J. Le Roy and J. M. Hutson, Faraday Discuss. Chem. Soc. 73
(1982) 339; J. M. Hutson and R. J. Le Roy, J. Chem. Phys. 83 (1985) 1107.
[45] C. M. Lovejoy and D. J. Nesbitt, Chem. Phys. Lett. 146 (1988) 582.
[46] J. A. Beswick, N. Halberstadt and J. M. Hutson, unpublished work (1988).
[47] S.L. Holmgren, M. Waldman and W. Klemperer, J. Chem. Phys. 69 (1978) 1661.
[48] A. M. Dunker and R.G. Gordon, J. Chem. Phys. 64 (1976) 354.
[49] J. G. Kircz, G. J. Q. van der Peyl, J. van der Elsken and D. Frenkel, J. Chem.
Phys. 69 (1978) 4606.
[50] J. M. Hutson and B. J. Howard, Mol. Phys. 43 (1981) 493.
[52] M. D. Marshall, A. Charo, H. O. Leung and W. Klemperer, J. Chem. Phys. 83
(1985) 4924.
[53] D. Ray, R. L. Robinson, D.-H. Gwo and R. J. Saykally, J. Chem. Phys. 84 (1986)
1171.
[54] R. L. Robinson, D.-H. Gwo, D. Ray and R. J. Saykally, J. Chem. Phys. 86 (1987)
5211.
[55] R. L. Robinson, D.-H. Gwo and R. J. Saykally, J. Chem. Phys. 87 (1987) 5156.
[56] R. L. Robinson, D.-H. Gwo and R. J. Saykally, J. Chem. Phys. 87 (1987) 5149.
[57] R. L. Robinson, D.-H. Gwo and R. J. Saykally, Mol. Phys. 63 (1988) 1021.
[58] K. L. Busarow, G. A. Blake, K. B. Laughlin, R. C. Cohen, Y. T. Lee and R. J.
Saykally, Chem. Phys. Lett. 141 (1987) 289.
[59] B. J. Howard and A. S. Pine, Chem. Phys. Lett. 122 (1985) 1.
[60] J. M. Hutson, J. Chem. Phys. 89 (1988) 4550.
[61] H. Klar, Nuovo Cimento 4A (1971) 529.
[62] M. Tamir and M. Shapiro, Chem. Phys. Lett. 31 (1975) 166.
[63] R. T Pack, J. Chem. Phys. 60 (1974) 633.
[64] P. McGuire and D. J. Kouri, J. Chem. Phys. 60 (1974) 2488.
[65] J. Tennyson and B. T. Sutcliffe, Mol. Phys. 58 (1986) 1067.
41
[67] R. D. Levine, J. Chem. Phys. 46 (1967) 331; J. Chem. Phys. 49 (1968) 51; R. D.
Levine, B. R. Johnson, J. T. Muckerman and R. B. Bernstein, J. Chem. Phys. 49
(1968) 56.
[68] S.L. Holmgren, M. Waldman and W. Klemperer, J. Chem. Phys. 67 (1977) 4414.
[70] A. M. Dunker and R. G. Gordon, J. Chem. Phys. 68 (1978) 700.
[71] C. M. Lovejoy and D. J. Nesbitt, Chem. Phys. Lett. 147 (1988) 490.
[72] C. M. Lovejoy, M. D. Schuder and D. J. Nesbitt, J. Chem. Phys. 85 (1986) 4890.
[73] J. M. Hutson, in Dynamics of Polyatomic Van der Waals Complexes, ed. N.
Halberstadt and K. C. Janda: Plenum, New York (1990).
[74] H. C. Longuet-Higgins, Mol. Phys. 6 (1963) 445.
[75] P. R. Bunker, Molecular Symmetry and Spectroscopy: Academic Press, New
York (1979).
[76] R. C. Cohen, K. L. Busarow, K. B. Laughlin, G. A. Blake, M. Havenith, Y. T. Lee
and R. J. Saykally, J. Chem. Phys. 89 (1988) 4494.
[77] J. M. Hutson, D. C. Clary and J. A. Beswick, J. Chem. Phys. 81 (1984) 4474; A.
C. Peet, D. C. Clary and J. M. Hutson, J. Chem. Soc., Faraday Trans. II 83 (1987)
1719.
[78] J. M. Hutson, J. Chem. Phys. 92 (1990) 157.
[79] G. T. Fraser, R. D. Suenram and L. H. Coudert J. Chem. Phys. 90 (1989) 6077
and references therein.
[80] L. H. Coudert and J. T. Hougen, J. Mol. Spectrosc. 130 (1988) 86.
[81] D. F. Coker and R. O. Watts, J. Phys. Chem. 91 (1987) 2513.
[82] J. R. Reimers, R. O. Watts and M. L. Klein, Chem. Phys. 64 (1982) 95.
[83] H. Klar, Z. Phys. 228 (1969) 59.
[84] M. H. Alexander and A. E. DePristo, J. Chem. Phys. 66 (1977) 2166.
[85] A. E. Barton and B. J. Howard, Faraday Discuss. Chem. Soc. 73 (1982) 45.
[86] J. Tennyson and B. T. Sutcliffe, Mol. Phys. 51 (1984) 887; Mol. Phys. 54 (1985)
141; Mol. Phys. 56 (1985) 1175; Mol. Phys. 58 (1986) 1067.
[88] J. G. Frey and B. J. Howard, Chem. Phys. 99 (1985) 415.
[89] J. G. Frey and B. J. Howard, Chem. Phys. 99 (1985) 427.
[90] J. G. Frey, Mol. Phys. 65 (1988) 1313.
[91] J. M. Hutson and S. Jain, J. Chem. Phys. 91 (1989) 4197.
[92] H. S. Gutowsky, T. D. Klots, C. Chuang, C. A. Schmuttenmaer and T. Emilsson,
J. Chem. Phys. 86 (1987) 569.
[93] T. D. Klots, C. Chuang, R. S. Ruoff, T. Emilsson and H. S. Gutowsky, J. Chem.
Phys. 86 (1987) 5315.
[94] J. M. Hutson, J. A. Beswick and N. Halberstadt, J. Chem. Phys. 90 (1989) 1337.
42

HutsonAMVCD (Rev) 1990

Uploaded by

Copyright:

Available Formats

HutsonAMVCD (Rev) 1990

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HutsonAMVCD (Rev) 1990

Uploaded by

Copyright:

Available Formats

Published in Advances in Molecular Vibrations and Collision Dynamics 1A, 1 (1990).

An Introduction to the Dynamics of Van der Waals Molecules

2. Atom–atom Van der Waals molecules

ψnlml = R−1 χnl (R)Ylml (β, α). (4)

ˆl2 Ylm (β, α) = l(l + 1)Ylm (β, α). (5)

Note that the deﬁnition of ˆ

2.1 Numerical methods

2.2 Spectroscopic constants

Enl = En0 + Bn l(l + 1) − Dn [l(l + 1)]2 + Hn [l(l + 1)]3 + . . . (10)

3. Atom–diatom Van der Waals molecules

3.1 Representation of the intermolecular potential

V (R, r, θ) = V (R, r, π − θ). (15)

This integral may again be evaluated eﬃciently by Gaussian quadrature. It provides a

3.2 The coupled equations

3.2.1 The space-fixed representation

where the Percival-Seaton coeﬃcients fλ (jl; j l ; J) are

fλ (jl; j l ; J) = (jlJM |Pλ (cos θ)|j l JM )

3.2.2 The body-fixed representation

* A slightly diﬀerent quantity, p = (−)j+l+J , is used in much of the scattering

(jKJM |Pλ (cos θ)|j K JM ) = δKK gλ (jj K), (33)

Jˆ± = Jˆx ∓ iJˆy (36)

ĵ± = ĵx ± iĵy (37)

jKJM |(Jˆ − ĵ)2 |jKJM  = J(J + 1) + j(j + 1) − 2K 2

3.2.3 Transformation between space-fixed and body-fixed representations

3.2.4 Solving the coupled equations

Y (R) = χ (R)[χ(R)]−1 , (45)

|Y in (Rmid ) − Y out (Rmid )| = 0. (46)

3.2.5 Matrix methods

V (θ) = V (R, r, θ)

Rg-H2 Rg-HF Rg-I2

* These complexes exhibit case 2 coupling in states of low angular momentum.

The energy levels Ev are solutions of the one-dimensional equation

4.1.4 Transition between case 1 and case 2 coupling

4.1.5 Transition between case 2 and case 3 coupling

4.1.6 Ar–HCl as a prototype system

Another consequence of the breakdown of the j quantum number is that the ∆j =

4.3 Distortion approximations

4.4 Perturbation theory

4.5 Adiabatic approximations

5. Dipole moments and spectroscopic intensities

* Dispersion contributions to dipole moments are usually very small

Direct dipole moment of diatom 1 0 1

use linear combinations of these components

The summation may be restricted to terms with J = 1 because µ must transform as

These selection rules rely on two assumptions:

6.1 Atom–polyatom systems

6.2 Molecule–molecule systems

6.3 Trimeric systems

You might also like

jKJM |(Jˆ − ĵ)2 |jKJM = J(J + 1) + j(j + 1) − 2K 2

V (θ) = V (R, r, θ)