First Order Polarizability

Shoresh Shafei∗and Mark G. Kuzyk

Department of Physics and Astronomy
Washington State University
Pullman, WA 99164-2814

The dipole moment, p, of an atom or molecule in a weak applied electric

field E can be approximated by

p = αE, (1)

where α is the polarizability tensor. Being a tensor, it allows for the fact that
the induced dipole moment is not necessarily in the direction of the applied
field. For simplicity, we treat the polarizability as a scalar so that the induced
dipole moment and electric field are parallel and choose the coordinate system
so that the z-axis is along the field. This reduces the problem to one dimension.
All models of the polarizability begin by calculating the dipole moment as
a function of the electric field. The resultant model function for the dipole mo-
ment, p(E), is expanded in a power series of the electric field, and the coefficient
of the linear term is the polarizability.
Here, we are interested in modeling the polarizability using time dependent
perturbation theory, where the electric coupling between the applied field and
the electrons is weak compared with the binding energy of the electrons in the
molecule. The time-dependent perturbation potential is of the form

V (t) = −p · E,
= −pE0 cos ωt, (2)

where ω is the applied field’s frequency. For convenience, we express cos(ωt) as

eiωt + e−iωt
cos(ωt) = , (3)
2
The full Hamiltonian is H = H0 + V (t), where H0 is Hamiltonian of the
system in the absence of perturbation with

H0 |ni = En |ni , (4)

where En is the eigenvalue and |ni an eigenvector of H0 .

∗ email: [email protected]

1
NLOsource.com 2

Using standard perturbation theory, the eigenvectors of the full Hamiltonian,

H = H0 + V (t), are expanded in terms of the unperturbed eigenvectors,
X
|ui ≡ |n; ti = cn (t)e−iΩn t |ni , (5)
n

where Ωn = En /~ and the time-varying coefficients, cn (t), are related to the

time-dependent perturbation V (t). cn (t) can be written as series of coefficients,
∞
X
cn (t) = c(0) (1) (2)
n (t) + cn (t) + cn (t) + . . . = cin (t), (6)
i=0

that are given by[1]

Z
−i X t
(N −1) 0
c(N )
n (t) = Vnl (t0 ) cl (t0 ) eiΩnl t , (7)
~ −∞
l

where Vnl (t0 ) = hn| V (t0 ) |li and

ωnl = ωn − ωl (8)

is internal frequency difference between state n and l.

Substituting Equation (2) into Equation (6) and assuming that the system
is in ground state at t = 0 (subscript g denotes the ground state), we get

c(0)
n = δng (9)

and
Z t
−i 0
c(1)
n (t) = eiΩng t Vng (t0 )dt0
~ −∞
µ ¶
qE0 ei(Ωng +ω)t ei(Ωng −ω)t
= xng + , (10)
2 Ωng + ω Ωng − ω

where xnm ≡ hn| x |mi. Note that Equation (10) assumes that the perturbation
(1)
is turned on at a time far in the past so that cn (t → −∞) → 0.
The dipole moment, p = qx, is given by the expectation of the dipole oper-
ator using the perturbed state

hu| p |ui
hpi (t) = = q hu| x |ui , (11)
hu|ui

where we keep only terms that are first order in the applied field. As such,
hu|ui = 1.
NLOsource.com 3

To first order in the external field E, the dipole moment is given by

hpi (t) = q hu|x|ui
XD ³ ´ ¯ ¯ ³ ´ E
−iΩn t ¯ ¯ −iΩm t
= q n c(0)
n + c(1)
n e ¯ x ¯e c(0) (1)
m + cm m
m,n
X h³ ∗
´³ ´i
= q e−iΩmn t δn0 + c(1)
n (t) δm0 + c(1)
m (t) xnm
n,m
( )
X ∗ X
= q xgg + c(1)
n (t)e
−iΩgn t
xng + c(1)
m (t)e
−iΩmg t
xgm , (12)
n m

Substituting Equation (10) into (12), we find

·µ ¶
qE0 X 2 eiωt e−iωt
hpi (t) = p0 + q |xgn | +
2~ n Ωng + ω Ωng − ω
µ −iωt ¶¸
e eiωt
+ + . (13)
Ωng + ω Ωng − ω
Decay in excited state population is taken account by including a damping
term that leads to a complex internal frequency of the form
Ωn = ωn − iΓn /2, (14)
where ωn and Γ are real. The negative sign in Equation (14) arises from the
requirement that the time evolution of the wavefunction, given by exp (−iΩn t),
must yields a loss in excited state population for Γn > 0. Furthermore, since
the ground state is stable (spontaneous transitions from the ground state do
not occur in vacuum), we assume that the ground state energy is real.[2] Thus
Γg = 0. (15)
With this generalization to include loss, Equation (13) becomes
q 2 E0 X
hpi (t) = p0 + |xgn |2
2~ n
·µ ¶ µ ¶¸
eiωt e−iωt e−iωt eiωt
+ + + . (16)
Ωng + ω Ωng − ω Ω∗ng + ω Ω∗ng − ω
Note that since the induced dipole moment is real, the term in the second set
of parentheses must be the complex conjugate of the term in the first set of
parenthesis. This requirement determines which frequencies require complex
conjugation. By combining terms, Equation (16) can be expressed as
q 2 E0 X
hpi (t) = p0 + 2 |xgn |2 ωng
~ n
³ ´
2 Γ2
ωng − ω 2 + 4ng cos ωt + ωΓng sin ωt
× ³ 2
´2 , (17)
2 − ω 2 + Γng
ωng + ω 2 Γ2
4 ng
NLOsource.com 4

where we have used exp (iφ) = cos φ + i sin φ.

According to Equation (17), the induced dipole is not in phase with the
applied electric unless Γng = 0. The induced dipole moment can be expressed
as
hpi (t) = α0 (−ω; ω) cos ωt + απ/2 (−ω; ω) sin ωt, (18)
which can be inverted by fourier transform to get the polarizabilities,
³ ´
2 2 Γ2ng
q 2 X ω ng ωng − ω + 4
α0 (−ω; ω) = 2 |xgn |2 ³ ´2 , (19)
~ n 2 − ω 2 + ngΓ 2
ωng 4 + ω 2 Γ2ng
q2 X (ωωng Γng )
απ/2 (−ω; ω) = 2 |xgn |2 ³ ´2 . (20)
~ n 2
2 − ω 2 + Γng 2 Γ2
ωng 4 + ω ng

Noting that cos(ωt−φ) = cos ωt cos φ+sin ωt sin φ, the phase of the polarizability
is given by µ ¶
−1 απ/2
φ = tan . (21)
α0
If only the excited state n contributes, as is approximately the case when the
photon energy is in resonance with the nth state, the phase is given by,
µ ¶ Ã !
−1 απ/2 −1 2ωΓng
φ = tan = tan Γ2
. (22)
α0 ω 2 − ω 2 + ng
ng 2

A more elegant approach to the problem of calculating the polarizability is

to use complex basis exp (±iωt) rather than the real basis (sin ωt, cos ωt). In
that case, the induced dipole moment in Equation (16) can be written in the
form
1
hpi (t) = p(ω)e−iωt + c.c.
2 µ ¶
q 2 E0 e−iωt X 2 1 1
= |xgn | ∗ +ω
+ + c.c., (23)
2~ n
ωng ωng − ω

were c.c. represents the complex conjugate of the previous term. Using orthog-
onality of exp (−iωt) and exp (+iωt), and the definition given in Equation (1),
we get the complex polarizability
µ ¶
q2 X 2 1 1
α(−ω; ω) = |xgn | ∗ +ω
+ . (24)
~ n ωng ωng − ω

The real an imaginary parts of the polarizability, α(−ω; ω), given by Equation
(24) are just α0 and απ/2 , respectively.
One additional comment needs to be made about the damping factor in
Equation (14). The astute reader may notice that the introduction of complex
NLOsource.com 5

internal frequencies prior to applying perturbation theory leads to divergences

that invalidate the use of perturbation theory. Furthermore, even without the
damping factors, perturbation theory does not - in principle - hold on resonance
(when the internal frequency and the electric field frequency are the same)
(1)
because the coefficients cn can become larger than unity, leading to probability
non-conservation. However, the ad hoc addition of damping in Equation (16)
leads to a theory that accurately fits experiment. Methods for dealing with
such divergence are described in the paper by Orr and Ward,[2] and references
therein.

References
[1] R. W. Boyd, Nonlinear Optics Academic, Press, New York (1992).
[2] B. J. Orr and J. F. Ward, Mol. Phys. 20, 513, (1971).