Vibrational Sudden Approximation

September 13, 2010

1 A simple model
1.1 Time independent picture

Let's consider a very simple two dimensional model: the linear collision of an
atom and a diatomic molecule.

With a simple linear transformation of the coordinate Z (the distance be-

tween the incident atom and the center of mass of the molecule) and of the
coordinate Q (the internal degree of freedom of the molecule) [see ref Bow-
man79], we can dene the dimensionless mass scaled coordinates z and q such
that the Hamiltonian of the system is

1 1
H(z, q) = − ∂z2 − ∂q2 + Vvibr (q) + Vscatt (z, q) (1)
2 2
where Vvibr is the potential for the vibrations of the diatomic molecule and Vscatt
is the atom-molecule interaction.
When the atom is far from the molecule (z → ∞), Vscatt is negligible and a
convenient set of eigenstates for this system is made of the product functions

hzq | npi = χn (q) eıpz n = 0, 1... p ∈ [−∞; +∞]

where ϑp (z) is a plane wave of momentum p and χn (q) is the n-th eigenfunction
of the one dimensional molecular Hamiltonian
 
1 2
− ∂q + Vvibr (q) χn (q) = En χn (q)
2

We label the eigenstates χn with the index n assuming that the potential
admits only bound states. If the potential allows the molecule to dissociate,

1
also continous states are possible. However in such case our discussion can be
generalized in a straightforward manner .
1
Let's consider the scattering states for our model system

|np±i = Ω± |npi

These states represent an atom with momentum p that collide with a molecule
in the n-th vibrational state.
Without any approximation, we can write the scattering states as the prod-
uct
hzq | np±i = χn (q)ϕ±
np (z, q)

Since the scattering event couples dierent vibrational states of the target
molecule, hzq | np±i is not a simple product state, and the function ϕnp still
depends on q.
The Vibrational Sudden Approximation consists in assuming that the func-
tion ϕnp (z, q) is a slowly varying function of the vibrational coordinate q, i.e.

∂2 
χn (q)ϕ± ± 2

np (z, q) ≈ ϕnp (z, q) ∂q χn (q) (2)
∂q 2
The physical meaning of this assumption will be discussed later, from a
time dependent point of view. Now we want to derive an equation to compute
ϕnp (z, q). The scattering states are eigenfunctions of the Hamiltonian operator

p2
  
1 1
− ∂z2 − ∂q2 + Vvibr (q) + Vscatt (z, q) − En + χn (q)ϕ±
np (z, q) = 0
2 2 2

where the eigenvalue can be found applying the intertwining relation.

Applying the VSA approximation we get

p2
  
1
− ∂z2 + En + Vscatt (z, q) − En + ϕ±
np (z, q) = 0
2 2

p2
 
1
− ∂z2 + Vscatt (z, q) − ϕ±
np (z, q) = 0 (3)
2 2
This part of the scattering wavefunction is just a solution of a simpler one
dimensional problem, in which the scattering potential depends parametrically
on q. We dene the VSA Hamiltonian as the the hamiltonian of this lower
dimensional system

1
HVSA (q, z) = − ∂z2 + Vscatt (z, q) (4)
2
1 If the vibrational coordinate is higly excited and the continuum states are populated, the
assumption of Vibrational Sudden Approximation (Eq. (2)) is most unlikely to hold. This
however has nothing to do with the present mathematical discussion but with the specic
physical properties of the system.

2
 The VSA hamiltonian is a family of one dimensional system, in which the
solution is parametrically dependent of the coordinate q. Hence, instead of
dealing with one single two dimensional problem, we now can solve an innite
number of simpler one dimensional problems with potential V q (z) = Vscatt (z, q).
In the following we will use the semicolon to point out that the dependence
on q is parametrical. From the same equation, we can see that the solution is
independent on the vibrational state n. From now on we will drop the n label
from the wavefunction. In conclusion, the scattering state in the VSA is

hzq | np±i = χn (q)ϕ±

p (z; q)

Consequently, the scattering matrix can be written as

S(pn ← p0 m) = hpn− | mp0 +i =

 
= dq dz hpn− | zqi hzq | mp0 +i =
 
?  
= dq dz χn (q)ϕ−
p (z; q) χm (q)ϕ +
p0 (z; q) =
 

?

= dq χ?n (q) dz ϕ−
p (z; q) ϕ+
p0 (z; q) χm (q)

In square parentheses, we can recognize the q dependent S matrix for the scat-
tering problem corresponding to the VSA Hamiltonian

S(pn ← p m) = 0
dq χ?n (q) S VSA (p ← p0 ; q) χm (q)

1.2 Time dependent picture

To understand what are the implications of the VSA from a dynamical point of
view, let's analyze the assumption of Eq. (2) from a time dependent point of
view.
Let's consider an initial wavefunction χn (q)ϕ(z) which evolves in time ac-
cording to the full hamiltonian of Eq. (1)

ψ(q, z, t) = Ut χi (q)ϕ(z)

The VSA consists in assuming that the coupling potential commutes with the
2
vibrational Hamiltonian
∂
Hq = − 21 ∂q 2 + Vvibr (q)

[Hq , Vscatt ] ≈ 0 (5)

so that the vibrational states is not perturbed during the propagation. In fact
under such assumption the vibrational hamiltonian commutes with the VSA
hamiltonian and the evolution operator can be split according to

Ut = exp(−ıHt) ≈ exp(−ıHq t) exp(−ıHVSA t)

3
 Hence the evolving state in the VSA is

ψ(q, z, t) = exp (−ıHq t) χi (q) exp (−ıHVSA t) ϕ(z)

i.e. it can be written as a simple product of two functions, each solution of
a one dimensional problem. The part depending just on q , exp (−ıHq t) χi (q),
is a stationary eigenstate, representing the vibration of the molecule, unper-
turbed by the scattering event. The other factor exp (−ıHV SA t) ϕ(z) is a sim-
ple wavepacket, that evolves according to a potential which just parametrically
depends on the vibrational coordinate q.
Now we can examine the assumption of Eq. (5) with greater detail. The
part of Hq which does not commute with the coupling potential is the kinetic
energy part, since multiplicative operators always commute. Explicitely, the
commutation relation can be written as

∂q2 Vscatt (z, q) ψ(q, z) ≈ Vscatt (z, q) ∂q2 ψ(q, z) (6)

which means that the coupling potential is a slowly varying function of q

with respect to the vibrational eigenstates. By applying the derivation rule, we
nd that the commutator (and hence the error) is proportional to

ε ∼ ∂q2 Vscatt (z, q) ψ(q, z) + 2 (∂q Vscatt (z, q)) (∂q ψ(q, z))

For small displacement of the vibrational coordinate, we can further assume

the coupling potential to be a linear function of q
Vscatt = κq f (z)
The VSA error then becomes

ε ∼ 2κ f (z) (∂q ψ(q, z))

For each (q, z) point, the approximation is satised if

1. the local velocity of the oscillator is small (compared to the scattering

degree of freedom)
∂q ψ(q, z) ≈ 0
2. the coupling coecient is small (compared to the total potential of the
system)
κ f (z) ≈ 0

2 Vibrational Sudden Approximation in molecule

surface scattering
Now let's turn our attention to the scattering of molecules on surface. If we
consider the six degree of freedom of the molecule and one additional vibrational
degree of freedom for the surface, the Hamiltonian can be written as

1 2 1 2 1 2
H=− ∇ − ∇ − ∂ + V6D (R, r) + Vcoupling (R, r, Q) + Vvib (Q)
2M R 2M r 2µ Q

4
 We have split the total 7D potential in three terms: V6D is the interaction of
the molecule with the static equilibrium surface, Vvibr is the vibrational potential
for the surface degree of freedom and Vcoupling is the coupling between the
molecule and the surface degrees of freedom.
We can apply the Vibrational Sudden Approximation to the vibrational
degree of freedom of the surface:

 
1 2
− ∂Q + Vvib , Vcoupling ≈ 0
2µ

The VSA Hamiltonian will be

1 2 1 2
HVSA (R, r; Q) = − ∇R − ∇ + V6D (R, r) + Vcoupling (R, r, Q)
2M 2M r
As in the case of the simple two dimensional problem, this Hamiltonian is a
scattering problem with one less degree of freedom and parametrically dependent
on the vibrational coordinate Q. The VSA scattering matrix will be

S VSA (f ← i; Q)

where the indices i and f labels the initial and nal state of the scattered
molecule
2 . Labelling the initial and nal vibrational state of the surface with
µ and µ0 , the 7D scattering matrix in the VSA is given by

0
S(f µ ← i µ) = dQ χ?µ0 (Q) S VSA (f ← i; Q) χµ (Q)

where χµ (Q) and χµ0 (Q) are two eigenfunctions of the vibrational hamiltonian

1 2
Hvib = − ∂ + Vvib (Q)
2µ Q
The scattering probability, summed over the nal vibrational state of the
surface is
2
X
Pscatter (f ← i µ) = |S(f µ0 ← i µ)|
µ0

In the VSA, this sum can be simplied according to

2
X X
|S(f µ0 ← i µ)| = S ? (f µ0 ← i µ)S(f µ0 ← i µ) =
µ0 µ0

X  ? 
= dQ χ?µ0 (Q) S VSA (f ← i; Q) χµ (Q) dQ0 χ?µ0 (Q0 ) S VSA (f ← i; Q0 ) χµ (Q0 ) =
µ0

2 In detail, i will identify the initial momentum p and the initial rovibrational state of the
molecele ν j mj . The nal index f will specify the quantum numbers of the diraction channel
nX nY and the rovibrational state of scattered molecule ν 0 j 0 m0j . The nal momentum along
Z is xed by energy conservation.

5
 
χµ0 (Q)χ?µ0 (Q0 ) S VSA (f ← i; Q) S VSA (f ← i; Q0 ) χ?µ (Q)χµ (Q0 )
X
?
dQ dQ0
µ0

From the completeness of the vibrational states χµ0

X
χµ0 (Q)χ?µ0 (Q0 ) = δ(Q − Q0 )
µ0

In conclusion, the (molecule) state resolved reaction probability is given by

the average of the VSA scattering probability, with a weight which is given by
the square of the initial vibrational eigenfunction

dQ S VSA (f ← i; Q)χµ (Q)
2
Pscatter (f ← i µ) =

Note that with a knowledge of S VSA (f ← i; Q), the scattering probability

can be computed for any initial state just by changing the vibrational eigenstate
in the integral.
From a practical point of view, we can use this equation by computing the
VSA scattering matrix on a suitable grid {Qn } and by performing the integra-
tion with a quadrature rule

S VSA (f ← i; Qn )χµ (Qn ) wn

X 2
Pscatter (f ← i µ) =
n

where the coecients wn are the weights of the quadrature rule.

From the state resolved scattering probability, the reaction probability can
be computed by
X
Preact (i, µ) = 1 − Pscatter (f ← i µ)
f

The reaction probability then become


S VSA (f ← i; Q)
2
X 2
Preact (i, µ) = 1 − dQ |χµ (Q)| =
f
 
2 V SA

2
V SA
=1− dQ |χµ (Q)| 1 − Preact (i, Q) = dQ |χµ (Q)| Preact (i, Q)

As for the scattering probability, the reaction probability can be computed

simply averaging the reaction probability with weights equal to the square of
the initial vibrational eigenfunction.
In the last formula the reaction probability depends on the vibrational state
of the surface. If we assume the surface to be in thermal equilibrium with
temperature T, we can average the reaction probabily with a Boltzmann distri-
bution

X Eµ X Eµ 2 V SA
Preact (i, T ) = exp( )Preact (i µ) = dQ exp( ) |χµ (Q)| Preact (i, Q)
µ
kT µ
kT