Optical Faraday rotation

P. R. Berman
Department of Physics and Michigan Center for Theoretical Physics, University of Michigan,
Ann Arbor, Michigan 48109-1040
共Received 27 July 2009; accepted 2 November 2009兲
Three calculations of optical Faraday rotation are presented in which a linearly polarized field is
incident on a medium of harmonic oscillators in the presence of a longitudinal magnetic field. The
rotation of the plane of polarization of the field is evaluated using classical oscillators and the
Lorentz force equation, quantum oscillators and the Heisenberg equations of motion, and quantum
oscillators and a Schrödinger equation approach. It is shown that a simple argument, based on the
assumption that a circularly polarized field drives either ⌬m = 1 or ⌬m = −1 transitions on absorption
共m is the magnetic quantum number兲, leads to an incorrect result for the Verdet constant. © 2010
American Association of Physics Teachers.
关DOI: 10.1119/1.3266970兴

I. INTRODUCTION

Faraday optical rotation refers to the rotation of the plane

V= 冉
␻ n− − n+
2c B0/␮0
= 冊␳ e 3␮ 0 ␻2
2nm20c⑀0 共␻20 − ␻2兲2
共4a兲

of polarization of an optical field as it propagates in a dielec-

tric medium in which a static magnetic field is applied along
the direction of propagation of the field.1 Sommerfeld has
given a simple derivation of the Faraday rotation, modeling e␮0 dn
= ␻ 共4b兲
the atoms in the medium as classical oscillators.2 2m0c d␻
Sommerfeld explained the effect as arising from a differ-
ence in the index of refraction for the left 共n+兲 and right 共n−兲
circularly polarized components of the field.2 He used the is the Verdet constant. It is seen that the Verdet constant is
Lorentz force to calculate the response of the oscillators to proportional to the dispersion of the medium.
the applied fields and then obtained the polarization of the Although the Faraday effect is mentioned in many optics
medium. From the polarization, he extracted the indices of textbooks and studied experimentally in advanced under-
refraction associated with the left and right circularly polar- graduate laboratories,3 discussions of the Faraday effect in
ized components of the optical field and found that the dif- textbooks and laboratory manuals are usually qualitative in
ference in indices is given approximately by nature. Often, there is no detailed derivation of the Faraday
rotation angle. An exception is the book by Rossi,4 which
contains a derivation that is similar to Sommerfeld’s.
␳e3 B 0␻
n− − n+ = , 共1兲 Quantum-mechanical derivations of Faraday rotation are not
nm0⑀0 共␻0 − ␻2兲2
2 2
readily available in standard optics texts. There is a vast
literature on the Faraday effect, but many of these articles are
where ␳ is the oscillator density, ␻0 is the natural frequency, specialized, including applications to specific molecules,5
−e is the charge, m0 is the mass of an oscillator, B0 is the nonlinear effects,6 and the role of optical pumping.7 It is
magnitude of the external magnetic field, ␻ is the frequency useful to present a quantum calculation for a medium com-
of the optical field, ⑀0 is the vacuum permittivity, and posed of harmonic oscillators, when a weak linearly polar-
ized field propagates in the medium in the presence of a

冉 冊
longitudinal magnetic field. The results should mirror that of
␳e2/m0⑀0 1/2
the classical calculation; paradoxically, the predictions based
n= 1+ 共2兲
␻20 − ␻2 on a simple quantum model seem to be at odds with the
classical results.
The quantum calculation can be formulated for the level
is the index of refraction of the medium of oscillators in the
scheme of Fig. 1. Shown are the two lowest lying electronic
absence of the magnetic field.
state manifolds of a three-dimensional harmonic oscillator
For a sample of length L, the difference in indices leads to
with natural frequency ␻0. The ground state has electronic
a rotation angle
quantum number n = 0 and angular momentum L = 0, and the
first excited state has n = 1 and L = 1. If the incident field is
␪=
c
冉
␻ n− − n+
2
L= 冊␳ e 3B 0L ␻2
2nm20c⑀0 共␻20 − ␻2兲2
共3a兲
x-polarized, we can view the field as an equal superposition
of left and right circularly polarized fields or, equivalently, as
an equal superposition of ␴+ and ␴− radiation, where the ␴⫾
radiation induces ⌬m = ⫾ 1 transitions on absorption and m is
the magnetic quantum number. Because the only effect of the
=VLB0/␮0 , 共3b兲 magnetic field is to shift the excited state m = ⫾ 1 sublevels,
it can be deduced from Eq. 共2兲 that the indices of refraction
where ␮0 is the vacuum permeability and associated with the ␴+ and ␴− radiation are

270 Am. J. Phys. 78 共3兲, March 2010 [Link] © 2010 American Association of Physics Teachers 270
 r共t兲 = x+共t兲x̂ + y +共t兲ŷ + c.c., 共11兲
where x+共t兲 and y +共t兲 are positive frequency components of
the displacement, varying as exp共−i␻t兲. The field amplitudes
Ex共Z兲 and Ey共Z兲 are assumed to vary slowly with the optical
wavelength ␭, and it is assumed that the dipole approxima-
tion is valid, r Ⰶ ␭, allowing us to evaluate both the ampli-
tude and phase of the optical field at the equilibrium position
of the oscillator.
Fig. 1. Energy level scheme of an oscillator. The ground state has n = 0 and
The classical equations of motion for x+共t兲 and y +共t兲 of an
L = 0, while the first excited state has n = 1 and L = 1. A longitudinal magnetic
field splits the magnetic sublevels of the L = 1 manifold. oscillator that is located at position Z and subjected to the
Lorentz force 共8兲 are
eEx共Z兲exp关i共nkZ − ␻t兲兴 eB0ẏ +
ẍ+ + 2␥ẋ+ + ␻20x+ = −
冉 冊
− ,
␳e2/m0⑀0 1/2 2m0 m0
n⫾ = 1 + , 共5兲 共12a兲
␻⫾2
− ␻2
where eEy共Z兲exp关i共nkZ − ␻t兲兴 eB0ẋ+
ÿ + + 2␥ẏ + + ␻20y + = − + ,
eB0 2m0 m0
␻⫾ = ␻0 ⫾ 共6兲
2m0 共12b兲
are the frequencies of the ⌬m = ⫾ 1 transitions from the where ␥ is a radiative decay rate. These equations are iden-
ground to excited state, including the Zeeman shift of the tical to those obtained by Sommerfeld,2 except that I have
levels. From Eqs. 共5兲 and 共6兲 it follows that included radiative decay, which allows for a steady-state so-
␳e3 ␻ 0B 0 lution. It is assumed that the magnitude of the electric field
n− − n+ ⬇ , 共7兲 changes negligibly as a function of Z in the medium due to
nm0⑀0 共␻0 − ␻2兲2
2 2
the fact that the field is detuned by an amount ␦ = ␻0 − ␻
which differs significantly from Eq. 共1兲 if ␻0 Ⰷ ␻. Clearly Ⰷ ␥, eB0 / m0. Although the magnitude of the field remains
something is wrong with this approach. constant, the plane of polarization rotates as the field propa-
The purpose of this article is to explain the breakdown in gates in the medium.
the reasoning that led to Eq. 共7兲 and to provide a classical Sommerfeld solved Eqs. 共12兲 by introducing the circular
and two simple quantum calculations of Faraday rotation. I components x+ ⫾ iy +; to obtain the rate of change of the Far-
begin by reviewing the classical calculation based on the aday rotation angle, d␪ / dZ, it is sufficient to calculate the
Lorentz force equation 共essentially equivalent to that given rotation angle d␪ in a distance dZ for which the rotation
by Sommerfeld兲 and then present the analogous quantum angle is much less than unity.8 In this limit we can take
calculation using a Heisenberg operator approach. Finally, I Ex共Z兲 ⬇ Ex共0兲 ⬅ E0 in the interval dZ. Moreover, we can ne-
use a Schrödinger picture approach that helps to isolate the glect the influence of the y component of the field on the
role of the ⌬m = ⫾ 1 transitions. Such an approach allows us dynamics of the x coordinate and take as the equations of
to uncover the fallacy in the reasoning that led to Eq. 共7兲. motion
Finally, I point out an interesting feature related to the choice eE0ei共nkZ−␻t兲
of the atom-field interaction Hamiltonian. ẍ+ + 2␥ẋ+ + ␻20x+ = − , 共13a兲
2m0

eEy共Z兲ei共nkZ−␻t兲 eB0ẋ+
II. CLASSICAL CALCULATION ÿ + + 2␥ẏ + + ␻20y + = − + . 共13b兲
2m0 m0
The medium is assumed to be composed of a uniform
density of harmonic oscillators, each having charge −e. The The steady-state solution of these equations 共that is, the so-
force on an oscillator is given by lution after all transients are finished兲 is

F = − e共E + v ⫻ B兲 − m0␻20r, 共8兲 eE0ei共nkZ−␻t兲 1

x+共Z,t兲 = − , 共14a兲
2m0 ␻20 − ␻2
where E is the electric field vector in the medium,
E共Z,t兲 = 21 关Ex共Z兲x̂ + Ey共Z兲ŷ兴exp关i共nkZ − ␻t兲兴 + c.c., 共9兲 eEy共Z兲ei共nkZ−␻t兲 1
y +共Z,t兲 = −
B is the magnetic induction, 2m0 ␻20 − ␻2
B = B0ẑ, 共10兲 e2B0E0ei共nkZ−␻t兲 共i␻兲
+ , 共14b兲
2m20 共␻20 − ␻2兲2
␻ = kc, n is the index of refraction of the medium given by
Eq. 共2兲, r is the displacement of an oscillator from its equi- where the limit ␥ → 0 has been taken under the assumption
librium position produced by the applied fields, and v = ṙ. that ␦, ␻0, ␻ Ⰷ ␥.
The field incident on the medium is polarized in the x̂ direc- To calculate the Faraday rotation, we must relate the fields
tion but acquires a y component as it propagates through the to the polarization in the medium. The positive frequency
medium. The displacement r共t兲 can be written as component of the polarization is given by

271 Am. J. Phys., Vol. 78, No. 3, March 2010 P. R. Berman 271
 P+共Z,t兲 = − ␳er+共t兲 = − ␳e关x+共Z,t兲x̂ + y +共Z,t兲ŷ兴, 共15兲 frequency components of the oscillator, obtained using ṗ
= 关p , H兴 / 共iប兲 and ṙ = 关r , H兴 / 共iប兲, are
and the wave equation in the medium can be written as
⳵ 2E + 1 ⳵ 2E + 1 ⳵ 2P + eB0 e2B20
− = , 共16兲 ṗx+ = − m0␻20x+ − py+ − x+
⳵ Z2 c2 ⳵ t2 ⑀ 0c 2 ⳵ t 2 2m0 4m0
e2B0Ey共Z兲exp关i共nkZ − ␻t兲兴
where I have set B = ␮0H and D = ⑀0E + P. Using Eqs. 共16兲, − , 共24a兲
共15兲, and 共14兲, we find that the wave equation for the x 4im0␻
component of the field 共9兲 is
␳ e 2E 0 1 eB0 e2B20
共n − 1兲E0 =
2
, 共17兲 ṗy+ = − m0␻20y + + px+ − y+
⑀0m0 ␻20 − ␻2 2m0 4m0
e2B0Ey共Z兲exp关i共nkZ − ␻t兲兴
which reproduces Eq. 共2兲 for the index of refraction in the + , 共24b兲
absence of the applied magnetic field. If we use a similar 4im0␻
approach for the y component of the field and cancel terms
using Eq. 共17兲, we find px+ eEx共Z兲exp关i共nkZ − ␻t兲兴 eB0
ẋ+ = + − y+ , 共24c兲
⳵ Ey共Z兲 i␳e k B0E0 3 2
␻ m0 2im0␻ 2m0
2ink ⬇ , 共18兲
⳵Z m 0⑀ 0 共 ␻ 0 − ␻ 2兲 2
2 2
py+ eEy共Z兲exp关i共nkZ − ␻t兲兴 eB0
where the assumption that Ey共Z兲 varies slowly in an optical ẏ + = + + x+ , 共24d兲
m0 2im0␻ 2m0
wavelength was used, that is, the ⳵2Ey共Z兲 / ⳵Z2 term was ne-
glected. With the boundary condition Ey共0兲 = 0, the solution from which it follows that
of Eq. 共18兲 for Ey共dZ兲 is ṗx+ eEx共Z兲exp关i共nkZ − ␻t兲兴 eB0
ẍ+ = − − ẏ + 共25a兲
␳e3kB0E0 ␻ m0 2m0 2m0
Ey共dZ兲 ⬇ dZ, 共19兲
2nm0⑀0 共␻0 − ␻2兲2
2 2

eEx共Z兲exp关i共nkZ − ␻t兲兴 eB0

which implies a differential change in the rotation angle for =− ␻20x+ − − ẏ + , 共25b兲
the field polarization given by 2m0 m0

d␪ Ey共dZ兲 ␳ e 3B 0 ␻2 ṗy+ eEy共Z兲exp关i共nkZ − ␻t兲兴 eB0

= = . 共20兲 ÿ + = − + ẋ+ 共25c兲
dZ E0dZ 2nm20c⑀0 共␻20 − ␻2兲2 m0 2m0 2m0
For a medium of length L, we find a rotation angle
eEy共Z兲exp关i共nkZ − ␻t兲兴 eB0ẋ+
␳ e 3B 0L ␻2 =− ␻20y + − + . 共25d兲
␪= , 共21兲 2m0 m0
2nm0c⑀0 共␻0 − ␻2兲2
2 2

These equations agree with Eq. 共12兲 in the limit ␥ → 0. If I

in agreement with Eq. 共3兲.
had used Hamilton’s equations for the classical calculation
instead of the Lorentz force equation, the classical and quan-
III. HEISENBERG EQUATIONS OF MOTION tum calculations would have been identical.
The calculation in this section illustrates the power of us-
The Hamiltonian for an oscillator having charge −e in the ing the Heisenberg equations of motion. We obtained the
presence of the applied fields is equations of motion for the relevant operators without hav-
关p + eA共r,Z,t兲兴2 1 ing to resort to a solution of Schrödinger’s equation. Life
H= + m0␻20r2 , 共22兲 was simple here, owing to the fact that I considered the quan-
2m0 2
tum system to be an 共intrinsically linear兲 oscillator. Had I
where the vector potential, tried to calculate the response of an atom to the applied
fields, I would not be able to write closed form differential
A共Z,t兲 = AE共Z,t兲 + AB共r兲, 共23a兲
equations of the type 共25兲 because the atom acts as a nonlin-
ear object.9
AE共Z,t兲

= 再 1
2i␻
冎
关Ex共Z兲x̂ + Ey共Z兲ŷ兴exp关i共nkZ − ␻t兲兴 + c.c. ,
IV. SCHRÖDINGER PICTURE CALCULATION
共23b兲
The two lowest lying electronic state manifolds of the os-
B0 cillator are shown in Fig. 1. The ground state has angular
AB共r兲 = − 关yx̂ − xŷ兴, 共23c兲 momentum L = 0 and the first excited state L = 1. Under the
2
assumption that the external fields are weak, these are the
is written in the Coulomb gauge. Interactions with the only levels that need be considered.10 Moreover, because the
vacuum field have been neglected, and thus the equations to external fields are weak, we can drop terms that vary as E20 or
be derived should be compared with Eq. 共12兲 in the limit that B20 in the Hamiltonian 共22兲. The approximate Hamiltonian
␥ → 0. The Heisenberg equations of motion for the positive for the oscillator in this limit is

272 Am. J. Phys., Vol. 78, No. 3, March 2010 P. R. Berman 272
 H=
p2 1
+ m0␻20r2 + 再
e关pxEx共Z兲 + pyEy共Z兲兴
2im0␻
amplitudes, the expectation values 具x典 and 具y典 for the state
vector 共29兲 are given by

冎
2m0 2

⫻exp关i共nkZ − ␻t兲兴 + adj. −

eB0
关pxy − pyx兴 −
e 2B 0 具x共t兲典 = 兺
m=−1,1
x0mam共t兲aⴱ0共t兲 + c.c., 共30a兲

再 冎
2m0 2m0

⫻
关yEx共Z兲 − xEy共Z兲兴
exp关i共nkZ − ␻t兲兴 + adj. 具y共t兲典 = 兺 y 0mam共t兲aⴱ0共t兲 + c.c., 共30b兲
2im0␻ m=−1,1

共26a兲 where

再 冎
x0m = 具L = 0兩x兩L = 1,m典, y 0m = 具L = 0兩y兩L = 1,m典 共31兲
eEx共Z兲共px − eyB0/2兲
=H0 + exp关i共nkZ − ␻t兲兴 + adj. are matrix elements between the 兩L = 1 , m典 excited state and
2im0␻

再 冎
the ground state.
eEy共Z兲共py + exB0/2兲 To calculate the state amplitudes, we use the time-
+ exp关i共nkZ − ␻t兲兴 + adj. dependent Schrödinger equation with the Hamiltonian 共28b兲
2im0␻
and the fact that the energy levels in the absence of the
eLzB0 optical field are given by
+ , 共26b兲
2m0 EL=0 = ប␻0/2, 共32a兲
where Lz = 共xpy − ypx兲 is the z component of angular momen-
tum, EL=1,m = 3ប␻0/2 + meបB0/2m0 . 共32b兲
If the oscillator is prepared in its ground state at t = 0, then, in
p2 1
H0 = + m0␻20r2 共27兲 lowest order perturbation theory, the ground state amplitude
2m0 2 is given by
is the Hamiltonian in the absence of any applied fields, and a0共t兲 = exp共− i␻0t/2兲, 共33兲
“adj.” is an abbreviation for “adjoint.” Note that I had to
and the time evolution of the excited state amplitudes is gov-
keep terms that vary as ExB0 or EyB0, originating from the
erned by
2AE · AB cross term in the A2 term in the Hamiltonian 关see
Eqs. 共23兲兴. This term contributes to the Faraday rotation in e
lowest order, and its inclusion is the price we must pay for ȧm = − i␻mam − ␥am − 关E0共px − eyB0/2兲
2m0ប␻
using the eA · p / m0 rather than the er · E interaction Hamil-
tonian for the optical field. I will return to this point in Sec. e
+ pyEy共Z兲兴m0exp关i共nkZ − ␻t兲兴a0共t兲 +
V. 2m0ប␻
In this section, I will calculate 具x典 and 具y典 from
Schrödinger’s equation. As in Sec. II, I assume that the field ⫻关Eⴱ0共px − eyB0/2兲 + pyEⴱy 共Z兲兴m0
is x-polarized initially and propagates for a small distance in ⫻exp关− i共nkZ − ␻t兲兴a0共t兲, 共34兲
which the Faraday rotation angle is small, allowing one to
replace Ex共Z兲 by E0 and to neglect the Ey共Z兲B0 term in Eq. where
共26b兲. As such the effective Hamiltonian is EL=1,m 3␻0 emB0
␻m = = + , 共35兲
e关E0共px − eyB0/2兲 + pyEy共Z兲兴 ប 2 2m0
H = H0 + exp关i共nkZ − ␻t兲兴
2im0␻ and I have included radiative damping to allow for a steady-
共28a兲 state solution.
The steady-state solution of Eq. 共34兲, with a0共t兲 given by
Eq. 共33兲, is
e关Eⴱ0共px − eyB0/2兲 + pyEⴱy 共Z兲兴
− exp关− i共nkZ − ␻t兲兴. e关E0共px − eyB0/2兲 + pyEy共Z兲兴m0 exp关i共nkZ − ␻t兲兴
2im0␻ am共t兲 = −
共28b兲 2m0ប␻ ␥ + i共␻m0 − ␻兲
e关Eⴱ0共px − eyB0/2兲 + pyEⴱy 共Z兲兴m0
Because the external fields are weak, a perturbative approach ⫻e−i␻0t/2 +
can be used.10 2m0ប␻
In a perturbation theory limit, we can work with state am- exp关− i共nkZ − ␻t兲兴 −i␻ t/2
plitudes rather than density matrix elements. The state vector ⫻ e 0 , 共36兲
for the oscillator can be written as ␥ + i共␻m0 + ␻兲
where
兩␺共t兲典 = a0共t兲兩L = 0典 + 兺 am共t兲兩L = 1,m典,
m=−1,1
共29兲
emB0
␻m0 = ␻m − ␻0/2 = ␻0 + 共37兲
2m0
where a0共t兲 is the ground state amplitude and am共t兲 共m
= ⫾ 1兲 is the excited state amplitude for sublevel m 共the m is an optical transition frequency. If we keep terms to zeroth
= 0 excited state sublevel is not excited because the optical order in B0 for 具x+典 and first order in B0 for 具y +典 and neglect
electric field has no z component兲. In terms of these state the back action of field component Ey共Z兲 on x+, we can use

273 Am. J. Phys., Vol. 78, No. 3, March 2010 P. R. Berman 273
Eqs. 共30兲, 共36兲, and 共33兲 to show that the positive frequency
components 关those varying as exp共−i␻t兲兴 of the displacement 兺 2
m=−1,1 共␻
␻m0
m0 − ␻ 兲
2 冋
共␦m,1 − ␦m,−1兲 +
eB0
2m0␻0
册
冋 册
are given by
␻10 ␻−10 eB0
⬇ − 2 +
eE0 exp关i共nkZ − ␻t兲兴 ␻10
2
− ␻2 ␻−10 − ␻2 m0共␻20 − ␻2兲
具x+共Z,t兲典 = − 兺

冤冉 冥
2m0ប␻

再 冎
m=−1,1 eB0 eB0
␻0 + ␻0 −
x0m共px兲m0 关x0m共px兲m0兴ⴱ 2m0 2m0 eB0

冊 冉 冊
⫻ − , 共38a兲 = − +
␥ + i共␻m0 − ␻兲 ␥ − i共␻m0 + ␻兲 eB0 2
eB0 2
m0共␻20 − ␻2兲
␻0 + ␻0 −
2m0 2m0
e exp关i共nkZ − ␻t兲兴 2eB0␻2
具y +共Z,t兲典 = − 兺 2m0ប␻
⬇−
m0共␻20 − ␻2兲2
. 共41兲

再
m=−1,1

y 0m关pyEy共Z兲 + pxE0 − eyE0B0/2兴m0 As a consequence, Eq. 共40b兲 reduces to

⫻
␥ + i共␻m0 − ␻兲
兵y 0m关pyEⴱy 共Z兲 pxEⴱ0 eyEⴱ0B0/2兴m0其ⴱ
冎
具y +共Z,t兲典 ⬇ − 再 eEy共Z兲
2m0共␻20 − ␻2兲

冎
+ −
− .
␥ − i共␻m0 + ␻兲 e2B0E0 共i␻兲
+ ei共nkZ−␻t兲 . 共42兲
共38b兲 2m20 共␻20 − ␻2兲2
Equations 共40兲 and 共42兲 agree with Eq. 共14兲.
The matrix elements for the 3D oscillator are
The Schrödinger approach has led to a result that is con-

冑
sistent with the classical result. Moreover, it can help us to
1 ប understand the breakdown of the reasoning given in Sec. I.
x0m = 共␦m,1 − ␦m,−1兲, 共39a兲 To understand what went wrong, consider a ␴+ circularly
2 m 0␻ 0
polarized field with the electric field vector,

ⴱ
y 0m = y m0 =
i
2
冑 ប
m 0␻ 0
共␦m,1 + ␦m,−1兲, 共39b兲
E共Z,t,+ 兲 =
2
冋 册
E0 x̂ + iŷ i共n kZ−␻t兲
冑2 e
+ + c.c., 共43兲

with E0 real, that drives transitions between the L = 0 ground

state and the L = 1 excited state. Using the Hamiltonian 共see
共px兲m0 =
i
冑បm0␻0共␦m,1 − ␦m,−1兲, 共39c兲 Sec. V兲
2
eLzB0
H = H0 + er · E共Z,t,+ 兲 + , 共44兲
2m0
共py兲m0 =
1
冑បm0␻0共␦m,1 + ␦m,−1兲, 共39d兲 with the state vector 共29兲 and a0共t兲 ⬇ exp共−i␻0t / 2兲, I find
2
that the excited state amplitudes evolve as
where ␦i,j is the Kronecker delta. The matrix elements de- E0ei共n+kZ−␻t兲
ȧm = − i␻mam − ␥am − i 关x + iy兴m0
pend on ␻0 and not ␻m0 because they are taken with respect 2ប冑2
to the unperturbed basis. We combine Eqs. 共38兲 and 共39兲 and
find ⫻exp共− i␻0t/2兲 共45a兲

eE0ei共nkZ−␻t兲 eE0ei共nkZ−␻t兲 E0e−i共n+kZ−␻t兲

具x+共Z,t兲典 = − 兺
m=−1,1 4m0共␻m0
2
− ␻ 2兲
⬇ −
2m0共␻20 − ␻2兲
, −i
2 冑2
关x − iy兴m0exp共− i␻0t/2兲, 共45b兲

共40a兲 where I have included radiative decay. The needed matrix

elements are given by Eq. 共39兲, and it follows that

具y +共Z,t兲典 = − 兺
eE 共Z兲ei共nkZ−␻t兲
y

m=−1,1 4m0共␻
2
m0 − ␻ 兲
2
ȧm = − i␻mam − ␥am − ieE0 冑 1
8បm0␻0
⫻关ei共n+kZ−␻t兲␦m,1 − e−i共n+kZ−␻t兲␦m,−1兴exp共− i␻0t/2兲.
ieE0ei共nkZ−␻t兲␻m0
− 兺
m=−1,1 4m0␻共␻m0 2
− ␻ 2兲
共46兲

冋
⫻ 共␦m,1 − ␦m,−1兲 +
eB0
2m0␻0
, 册 共40b兲
From Eq. 共46兲 it is clear that for the Hamiltonian 共44兲, the
only effect of the magnetic field is to shift the energy of the
m = ⫾ 1 excited state sublevels. Magnetic-field-induced
changes in the index of refraction associated with the ␴+ and
where I have taken the ␥ → 0 limit. Using Eq. 共37兲, we can ␴− components of the electric field result solely from these
show that magnetic-field-induced level shifts. It seems that arguments

274 Am. J. Phys., Vol. 78, No. 3, March 2010 P. R. Berman 274
of the type presented in Sec. I lead to an inconsistent value ␳e3 ␻B0
for the Verdet constant. However, this line of reasoning is n− − n+ ⬇ , 共54兲
nm0⑀0 共␻0 − ␻2兲2
2 2
valid only if ␴+ radiation drives only ⌬m = 1 and ␴− radiation
only ⌬m = −1 transitions on absorption. in agreement with Eq. 共1兲. Note that the index of refraction
In the resonance or rotating-wave approximation, the last for each circularly polarized field in Eqs. 共52兲 and 共53兲 con-
term in Eq. 共46兲 can be dropped. In this approximation, it is tain contributions from both ⌬m = ⫾ 1 transitions.
correct to say that a ␴+ circularly polarized excites only
⌬m = 1 transitions on absorption. If the last term in Eq. 共46兲
V. DISCUSSION
is included, this statement no longer holds. Thus the break-
down in the reasoning in Sec. I is that a circularly polarized I have presented three calculations of Faraday rotation, a
field drives both ⌬m = ⫾ 1 transitions if the resonance ap- classical calculation using the Lorentz force equation, a
proximation is not made. Thus, there are contributions to the quantum calculation using a Heisenberg operator approach,
index of refraction of ␴+ radiation from both the ⌬m = ⫾ 1 and a quantum calculation using a Schrödinger wave func-
transitions. tion approach. The three methods yield consistent results. An
Equations 共30兲 and 共46兲 can be used to calculate n+. The oversimplified picture in which one assigns an index of re-
steady-state solution of Eq. 共46兲 is fraction for ␴⫾ circularly polarized radiation resulting only

冑 冋
from ⌬m = ⫾ 1 transitions on absorption breaks down when
1 ei共n+kZ−␻t兲␦m,1 non-rotating-wave approximation terms are included in the
am共Z,t兲 = − ieE0
8បm0␻0 ␥ + i共␻m0 − ␻兲 calculation.

册
Finally, I return to the question as to the correct form of
e−i共n+kZ−␻t兲␦m,−1 the Hamiltonian in weak fields. It might be thought that us-
− exp共− i␻0t/2兲. 共47兲
␥ + i共␻m0 + ␻兲 ing a Hamiltonian of the form

To obtain an expression for n+ in terms of am共Z , t兲, I first ep · AE共Z,t兲 eLzB0

H1 = H0 + + 共55兲
write the polarization as m0 2m0

P共Z,t兲 = 共Px+x̂ + iPy+ŷ兲ei共n+kZ−␻t兲 + c.c. 共48a兲 with neglect of the A2 term would lead to correct results.
However, we have seen that this assumption is not the case.

冋 册
It is essential to include the 2AE · AB cross term to arrive at
x̂ + iŷ x̂ − iŷ the correct expression for the Faraday rotation for both the
= P+共+ 兲 + P+共− 兲 ei共n+kZ−␻t兲 + c.c.,
冑2 冑2 quantum calculation and a classical calculation based on
Hamilton’s equations. In contrast, if we start from the Hamil-
共48b兲 tonian

where the positive frequency component of the circular com- eLzB0

H2 = H0 + er · E共Z,t兲 + , 共56兲
ponents of the polarization is given by 2m0

Px+ ⫿ iPy+ the correct Faraday rotation is obtained without any problem.
具x+典 ⫿ i具y +典
P +共 ⫾ 兲 = = − ␳e . 共49兲 This result is another example of where it is advantageous to
冑2 冑2 use the er · E rather than the ep · A / m0 form of the interaction
potential for interactions of atoms or oscillators with optical
By combining Eqs. 共30兲, 共39兲, 共47兲, and 共49兲, we can show fields in the dipole approximation.11
that P+共−兲 = 0 共the ␴+-polarized incident field induces only a The Hamiltonians 共22兲 and 共56兲 must lead to the same
␴+ component of the polarization兲 and expectation values for any physical observables because they

冋 册
are related by a unitary transformation.11,12 However, the dy-
␳ e 2E 0 1 1 namics of the state evolution for each Hamiltonian can be
P+共+ 兲 = + , 共50兲
4m0␻0 ␻+ − ␻ ␻− + ␻ quite different. For example, it follows from Eq. 共34兲 that for
the Hamiltonian 共26b兲, the magnetic field shifts the excited
where the limit ␥ → 0 has been taken 共recall that ␻⫾ = ␻⫾10 state sublevels, the optical field couples the ground and ex-
= ␻0 ⫾ eB0 / 2m0兲. Finally, because the polarization and field cited states, and the combined action of the optical and mag-
amplitudes in Eqs. 共48兲 and 共43兲 are related by netic fields couples the ground and excited states. In contrast,
for the Hamiltonian 共56兲, the magnetic field shifts the excited
P+共+ 兲 = 共n+2 − 1兲⑀0E0/2, 共51兲 state sublevels and the optical field couples the ground and
excited states, but there is no cross term involving the com-
it follows that bined action of the optical and magnetic fields. As a conse-

冋 册
quence, it is easier to solve for the dynamics and to interpret
␳e2 1 1 the roles of the fields if we use the er · E Hamiltonian in Eq.
n+2 − 1 = + . 共52兲
2m0⑀0␻0 ␻+ − ␻ ␻− + ␻ 共56兲.
The calculation for the oscillators would be identical to
Similarly, for a ␴− circularly polarized field, we find that for atoms with ground state angular momentum L = 0

冋 册
and excited state angular momentum L = 1 as long as the
␳e2 1 1 fields are weak. In contrast to the oscillator problem, how-
n−2 − 1 = + , 共53兲
2m0⑀0␻0 ␻− − ␻ ␻+ + ␻ ever, there are nonlinear effects for atoms that enter as the
optical field strength is increased.6 Moreover, for ground
implying that states in which there is magnetic state degeneracy, fine or

275 Am. J. Phys., Vol. 78, No. 3, March 2010 P. R. Berman 275
 5
hyperfine structure optical pumping can modify the Faraday See, for example, D. M. Bishop and S. M. Cybulski, “Magnetic optical
rotation.7 For atoms, it is not possible to obtain a closed from rotation in H2 and D2,” J. Chem. Phys. 93, 590–599 共1990兲; M.
Krykunov, A. Banerjee, T. Ziegler, and J. Autschbach, “Calculation of
differential equation for the Heisenberg operator r共t兲 as it
Verdet constants with time-dependent density functional theory: Imple-
was for the oscillator. Instead, using either a Heisenberg or mentation and results for small molecules,” ibid. 112, 074105-1–7
Schrödinger approach, we must calculate the expectation 共2005兲.
value of the dipole moment operator of the atoms. 6
See, for example, F. Schuller, M. J. D. MacPherson, and D. N. Stacey,
“Saturation and collisional effects on magnetic optical rotation,” Physica
ACKNOWLEDGMENTS C 147, 321–331 共1987兲; D. Budker, D. J. Orlando, and V. Yashchuk,
“Nonlinear laser spectroscopy and magneto-optics,” Am. J. Phys. 67,
The motivation for this calculation came from a request 584–592 共1999兲.
7
from Carl Akerlof at the University of Michigan to provide a F. Schuller, M. J. D. MacPherson, and D. N. Stacey, “Magneto-optical
theory write-up for an advanced undergraduate laboratory on rotation in an atomic vapor,” Opt. Commun. 71, 61–64 共1989兲; F.
Faraday rotation. The author is also happy to acknowledge Schuller, D. N. Stacey, R. B. Warrington, and K. P. Zetie, “Theory of
1 1
Faraday rotation produced by atoms near a Jg = 2 → Jg = 2 transition,” J.
helpful discussions with Peter Milonni, and the author ben-
Phys. B 28, 3783–3790 共1995兲. This paper also includes collisional and
efitted from helpful comments of the referees. saturation effects.
8
For successive intervals, we can redefine the direction of polarization as
1 the x direction and then calculate the rotation in the next small interval.
See, for example, A. D. Buckingham and P. J. Stephens, “Magnetic op-
tical activity,” Annu. Rev. Phys. Chem. 17, 399–432 共1966兲, and refer- The treatment is strictly valid only for eB0␻ / 2m0 Ⰶ 共␻20 − ␻2兲.
9
ences therein. P. R. Berman, “Two-level approximation in atomic systems,” Am. J.
2
A. Sommerfeld, Optics 共Academic, New York, 1964兲, pp. 101–106. Som- Phys. 42, 992–997 共1974兲.
10
merfeld used n− for left circularly polarized radiation and n+ for right Because the equations of motion for the displacement operator of a har-
circularly polarized radiation. I have interchanged his notation because monic oscillator contain a source term that is linear in the external elec-
n⫾ corresponds to ␴⫾ radiation for a field propagating in the ẑ direction, tric field, a perturbative calculation of the Heisenberg operator r共t兲 yields
where ␴⫾ radiation induced ⌬m = ⫾ 1 transitions on absorption 共m is the 11
a result that is correct to all orders in the electric field amplitude.
magnetic quantum number兲. See, for example, J. R. Ackerhalt and P. W. Milonni, “Interaction Hamil-
3
F. J. Loeffler, “A Faraday rotation experiment for the undergraduate phys- tonian of quantum optics,” J. Opt. Soc. Am. B 1, 116–120 共1984兲.
12
ics laboratory,” Am. J. Phys. 51, 661–663 共1983兲; F. L. Pedrotti and P. The Hamiltonians are related by the unitary transformation H⬘ = UHU†
Bandettini, “Faraday rotation in the undergraduate advanced laboratory,” with U = exp关ier · AE共Z , t兲 / ប兴, where H is given by Eq. 共22兲, AE共Z , t兲 is
ibid. 58, 542–545 共1990兲. given by Eq. 共23b兲, and H⬘ = H2 + e2B20共x2 + y 2兲 / 8m0 is equal to H2 关Eq.
4
B. Rossi, Optics 共Addison-Wesley, Reading, MA, 1957兲, pp. 427–430. 共56兲兴, plus a term corresponding to the quadratic Zeeman shift.

276 Am. J. Phys., Vol. 78, No. 3, March 2010 P. R. Berman 276