Measurement of the Hyperne Structure of Rubidium 85 and 87

Javier M. G. Duarte

and Sara L. Campbell

Massachusetts Institute of Technology, MA 02142

(Dated: March 12, 2009)
We measure the hyperne splitting of the
5
P
3/2
state in rubidium 85 and 87 using Doppler-
free saturated absorption spectroscopy. The Doppler eect causes a broadening of the spectral
lines, which consumes the hyperne peaks. However, this technique, along with the Fabry-Perot
interferometer, allows us to measure the previously shrouded hyperne energy spacings. From
these spacings, we calculate the magnetic dipole and electric quadrupole coupling constants for the
5
P
3/2

5
S
1/2
transition for each isotope: A(
85
Rb)=24.81.6, B(
85
Rb)=17.46.8, A(
87
Rb)=83.8
1.8, and B(
87
Rb)=19.6 4.4.
I. INTRODUCTION AND MOTIVATION
Traditionally, optical spectroscopy was limited by
the Doppler broadening, or the widening of spectral
linewidths due to the Doppler shift from the motion of
the atom.
However, a sophisticated technique known as Doppler-
free saturated absorption spectroscopy is insensitive to
the Doppler eect. This allows us to resolve smaller fre-
quency spacings, which form the hyperne structure of
atomic lines. A. A. Michelson rst observed optical hy-
perne structure in 1891. In order to explain these hy-
perne splittings, in 1924, W. Pauli proposed that the
atomic nucleus possesses intrinsic angular momentum,
which is labelled by the quantum number I. This radical
idea actually allows one to derive the frequency spacings
up to a pair of parameters known as the hyperne cou-
pling constants.
F=3
5P
3/2
1/2
5P
5S
1/2
5P
3/2
1/2
5P
5S
1/2
85
Rb: I=5/2, 72% Rb: I=3/2, 28%
87
267 MHz
780.23 nm
794.76 nm
29 MHz
63 MHz
121 MHz
3.036 GHz
780.23 nm
794.76 nm
157 MHz
72 MHz
6.835 GHz
F=3
F=2
F=3
F=1
F=2
F=4 F=3
F=2
F=1
F=0
F=2
F=1
F=2
F=1 F=2
FIG. 1: The energy level diagram for
85
Rb and
87
Rb.
In this experiment we are concerned with probing the
hyperne structure of the two most abundant isotopes of
rubidium. Specically, we look in the range of 780 nm
light, which corresponds to the D
2
line or the 5
2
S
1/2

5
2
P
3/2
transition.

Electronic address: woodson@mit.edu

Electronic address: campsoup@mit.edu

II. THEORY
Two important components of the theoretical back-
ground for this experiment are the physics that gives rise
to ne and hyperne structure and the Fabry-Perot in-
terference pattern. In the following exposition, we follow
these conventions: J is the total outer electron angular
momentum, that is the sum of the electrons spin and
orbital angular momentum, J = S + L, I is the spin
of the nucleus, and nally F is the total atomic angular
momentum, F = J +I.
A. Fine and Hyperne Splitting
In addition to the nuclear spin I, higher order elec-
tromagnetic multipole moments are attributed to the nu-
cleus [1]. Typically, we consider only the rst and second
orders of interaction: the magnetic dipole and the electric
quadrupole.
The perturbation to the Hamiltonian describing the in-
teraction between the magnetic eld created by the elec-
trons at the nucleus and the nuclear magnetic moment is
given by:
H =
I
B
el
(0) (1)
= AI J (2)
where equation 2 is the result of the IJ coupling ap-
proximation, that is we assume we are dealing with an
isolated level J, or that the hyperne structure is small
compared with the ne structure. Thus, we may assume

I
is directly proportional to I and that B
el
J.
In order to treat the quadrupole interaction, we must
allow the possibility of a nonuniform charge distribution
in the nucleus. In other words, we no longer assume the
nucleus is a point charge. Our goal in the following is to
expand the Coulombic Hamiltonian to split up the contri-
butions from the electric monopole, dipole, quadrupole,
and higher order terms.
First, we consider the interaction between a single nu-
cleon at r
n
and a single electron at r
e
. Later we will sum
over all the particles to achieve the total interaction. We
want to separate these variables in order to proceed with
2
the expansion, so we assume r
e
> r
n
, that is the electrons
do not penetrate the nuclear volume.
H =
e
2
|r
e
r
n
|
(3)
= e
2

k=0
r
k
n
r
k+1
e
P
k
(cos
en
) (4)
where P
k
(cos
en
) are the Legendre polynomials and
en
is the angle between r
n
and r
e
.
The k = 2 term is the quadrupole interaction that we
seek. Next we apply a useful identity from mathematics,
known as the spherical harmonic addition theorem, which
expresses a Legendre polynomial as a sum of products of
two spherical harmonics
P
k
(cos
en
) =
4
2k + 1
k

q=k
(1)
q
Y
q
k
(
n
,
n
)Y
q
k
(
e
,
e
).
(5)
where the Y
q
k
are the spherical harmonics of rank
k and projection q (analogous to the quantum
numbers l and m
l
). Figure 2 displays how the
angles are dened. This allows us to totally sep-
arate the two variables. So now, picking out the
k = 2 term, we are left with a sum of the form
2

q=2

Complicated Expression r
2
n

Complicated r
3
e

Again we assume the IJ coupling approximation is valid,

r
e
z
x
y

e
r
n

e
FIG. 2: Denition of angles for the nucleon and the electron.
which implies the nucleus has a denite spin I. We
then may recast the sum in terms of angular momentum
operators, yielding the quadrupole interaction Hamilto-
nian. Finally, we conveniently dene B as the product
eQ

2
V
e
z
2
, which must be determined experimentally.
Then, we simply add the Hamiltonians from the mag-
netic dipole and electric quadrupole and nd E
1
, the
rst order correction to the energy. From perturbation
theory E
1
=
0
|H

|
0
, which directly yields the
equation weve been seeking [2][3]:

F

J
=
AC
2
+
B[
3C(C+1)
4
I(I + 1)J(J + 1)]
2I(2I 1)J(2J 1)
(6)
where C = F(F + 1) J(J + 1) I(I + 1).
B. Fabry-Perot Interferometer
The purpose of the Fabry-Perot interferometer in our
setup, roughly speaking, is to measure the laser frequency
scan as a function of time. In other words, the oscillo-
scope displays signals as a function of time and we would
like to convert to signals over a frequency range.
Our particular Fabry-Perot interferometer consists of
two parallel mirrors, facing each other, at a distance L
apart. We expect to see constructive interference at the
far mirror when the path dierence is an integer multiple
of the wavelength.
Thus we expect to see a periodic interference pattern
as a function of time. The period is the time it takes for
the laser controller to scan to an adjacent resonant wave-
length. This period corresponds to a frequency spacing
between adjacent resonant modes. This spacing, known
as the free spectral range, is given by [3]:

FSR
=
c
2n
air
L
(7)
Equating the period of our Fabry-Perot interference
pattern and the free spectral range yields the frequency
calibration.
III. EXPERIMENTAL DESIGN
The experimental setup is similar, in essence, to gure
3.
Splitter
Probes Rubidum
Cell
Mirror
Mirror
Beam
Diode Laser
Head
Faraday
Isolator
Fabry!Perot Cavity
Mirror
Prism
Balanced
Optical
Detector
Trigger
Oscilliscope
Pump
Beam
Splitter
Neutral
Density
Filter
Laser
Controller
FIG. 3: The experimental setup: gure created by S. Camp-
bell.
The 780 nm AlGaAs semiconductor diode laser head
and a collimating lens are both mounted on a tempera-
ture controlled stage. The laser uses a diraction grating
in front of the diode as a wavelength selective mirror.
The orientations of this grating and another reective
3
surface allow for ne adjustment of the output frequency
of the laser. When outputing single mode light, the spec-
tral width of the laser is approximately 30 MHz, which
signicantly aects our measurements.
The laser controller has three components, which con-
trol the diodes temperature, the diodes current, and the
linear scan of the current. Adjusting the diodes temper-
ature is the coarse tuning of the output frequency, while
the current knob allows for ner adjustment. The scan
controller in eect allows us to output a repetitive linear
sweep through a set of wavelengths near 780 nm.
A beam splitter sends one half of the laser to the Fabry-
Perot interferometer, in which we use a photodiode be-
hind the far mirror to measure the interference fringes.
The other half of the laser proceeds through a prism to
create an intense pump beam and two parallel probe
beams, each a factor of 0.04 times less intense than the
pump. The pump continues until it is split by a beam
splitter and half of it is reected through the rubidium
cell. Meanwhile, the two probes are incident upon a mir-
ror which reect them through the same cell in the op-
posite direction as the pump.
By design, the pump overlaps one of the probe beams
throughout the entire extent of the cell. Suppose the
pump and the probe in a particular instance are tuned
on the resonance frequency
0
. The velocity group
1
that the beams interact with is the v = 0 groupthere
is no need for the atoms to be moving to compensate,
through the Doppler shift, for o-resonance beams.
Since the pump excites most of the available rubidium
atoms in the v = 0 velocity group , the intersectant probe
nds less v = 0 atoms than in nearby groups, so more
light intensity is transmitted. Conversely, this means the
absorption spectrum A() from the probe beam exhibits
a sharp dip, known as a Lamb dip, at
0
, the resonant
frequency. We detect the dierence in absorption spec-
tra between the two probes (one intersecting the pump)
using a balanced optical detector. Figure 4 shows full
spectrum with Lamb dips.
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
Frequency [MHz] decreasing
R
e
s
c
a
l
e
d

V
o
l
t
a
g
e

[
A
r
b
.

U
n
i
t
s
]
Full Spectrum
FIG. 4: Raw data of the full spectrum with Lamb dips. We
focus on tting the regions highlighted in red.
1
The number of atoms in a particular atomic state (population)
traveling at a velocity v due to random thermal motion.
IV. DATA AND ERROR ANALYSIS
A. Fabry-Perot Calibration
First, as described in section II.B, we calibrated the fre-
quency axis using the Fabry-Perot interference pattern.
We t the pattern to the theoretical function described
in [4]:
I
T
=
I
0
T
2
(1 R)
2
+ 4Rsin
2

(8)
where T is the power transmission coecient, R is the
power reection coecient, and R+T = 1. We can com-
pute the factor M which converts separations in times
to separations in frequency, = Mt. In the example
shown in gure 6, we get M = 132.8 1.5 GHz/s. To
account for the nonlinearity of the laser scan, we use a
dierent local scaling for each t. However, we note that
M varies little, in agreement with the apparent linearity
of the laser scan.
B. Doppler-Broadened Structure
Once we knew how to convert time spacings into
frequency spacings, we t Gaussians to the Doppler-
broadened spectrum. The predicted FWHM is given by
502 MHz [3]. Figure 5 shows that we nd slightly wider
peaks at 562 6 MHz.
0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18
4.4
4.6
4.8
5
5.2
5.4
5.6
5.8
6
6.2
y(x) =A(exp(!(x!)
2
/(2!
2
)))+(lin)x + DC
Time [s]
A
m
p
lit
u
d
e

[
V
]
Doppler!Broadened
! = 0.00228e!06
"
2
/dof = 1.03
Probability =38 %
! = 0.00195e!06
"
2
/dof = 0.553
Probability =1e+02 %
! = 0.0018771e!06
"
2
/dof = 1.02
Probability =44 %
! = 0.0020.005
"
2
/dof = 5.828
Probability =0 %
0.116 0.129
!0.02
0
0.02
Time [s]
A
m
p
lit
u
d
e

[
V
]
Residuals 1
0.145 0.15
!0.02
0
0.02
Time [s]
Residuals 2
0.165 0.17
!0.04
!0.02
0
0.02
0.04
Time [s]
Residuals 3
0.175 0.18
!0.02
0
0.02
Time [s]
Residuals 4
87
Rb 5P
3/2
#
5S
1/2
(F=1)
85
Rb 5P
3/2
#
5S
1/2
(F=2)
85
Rb 5P
3/2
#
5S
1/2
(F=3) 87
Rb 5P
3/2
#
5S
1/2
(F=2)
FIG. 5: The Gaussians ts to the Doppler-broadened signal
(with the pump blocked). Figure by S. Campbell.
C. Hyperne Structure Fitting
To better discriminate the peaks, we subtracted the
background from the signal. We dened this back-
ground to be the dierence spectrum from the balanced
optical detector when the pump was blocked. Ideally,
this signal would be vanish everywhere (as the two probes
should give identical spectra in the absence of the pump).
However, this is not the case due to several contributions.
The most obvious is the fact that one of the probes is
4
transmitted twice through the surface of the so that its
intensity carries an extra factor of (0.96)
2
.
0 50 100 150 200 250
0
0.1
0.2
0.3
0.4
0.5
0.6

12
= 35.67.5

23
= 59.66.3

34
= 1136.4

3
(
3
+
4
)/2

4
Frequency [MHz]
R
e
s
c
a
l
e
d

V
o
l
t
a
g
e

[
A
r
b
.

U
n
i
t
s
]
85
Rb 5
2
S
1/2
, F=3 5
2
P
3/2
FIG. 6:
85
Rb 5
2
S
1/2
, F = 3 5
2
P
3/2
.
We then performed a least-squares t using an ap-
propriate number of bins about each peak. We used a
Lorentzian lineshape due to the resonance nature of the
atomic transition. However, due to background and other
systematic considerations, our t function contained lin-
ear and DC oset terms. The Lorentzian ts to the
85
Rb
data, shown in gure 6 gave reduced chi-square values on
the order of unity ranging from 0.3 to 6.7. Likewise, for
the Lorentzian ts to the
87
Rb data, shown in gure 7,
we recorded reduced chi-square values ranging from 0.2
to 1.8.
150 200 250 300 350 400 450 500 550
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045

01
=62.1 5.9

12
=148 5.7

23
=271 7

0
(
0
+
1
)/2

1
(
0
+
2
)/2
(
1
+
2
)/2

2
(
0
+
3
)/2
(
2
+
3
)/2
Frequency [MHz]
R
e
s
c
a
l
e
d

V
o
l
t
a
g
e

[
A
r
b
.

U
n
i
t
s
]
87
Rb 5
2
S
1/2
, F=1 5
2
P
3/2
FIG. 7:
87
Rb 5
2
S
1/2
,F = 1 5
2
P
3/2
.
Table IVC summarizes our ndings for the hyperne
splittings. Then by using equation 6 on each splitting
level, that is subtracting
F

F1
, we may solve the
system of linear equations for the coupling constants A
and B in terms of the frequency splittings.
Thus we also list our values for A and B for each iso-
tope. The accepted values for the coupling constants
come from [5]
Parameter Measured [Mhz] Accepted [Mhz]

12
(
85
Rb) 35.6 7.5 29

23
(
85
Rb) 59.6 6.3 63

34
(
85
Rb) 113 6.4 121

01
(
87
Rb) 62.1 5.9 72

12
(
87
Rb) 148 5.7 157

23
(
87
Rb) 271 7.0 267
A(
85
Rb) 24.8 1.6 24.988(31)
B(
85
Rb) 17.4 6.8 25.693(31)
A(
87
Rb) 83.8 1.8 84.7185(20)
B(
87
Rb) 19.6 4.4 12.4965(37)
TABLE I: Measurement of the hyperne splitting and cou-
pling constants of
5
P
3/2
compared to accepted values.
V. CONCLUSIONS
We rst looked at the spectrum of rubidium with the
pump blocked, and thus we saw the eect of Doppler-
broadening. Then we intersected the pump with one
of the probes and saw characteristic Lamb dips. We
looked that dierence of the two probes, which yielded
a sharply peaked spectrum. Finally we subtracted the
background (probes dierence with the pump blocked)
in order to better identify the peaks.
These peaks were t to Lorentzians with linear and
DC oset terms, which yielded good reduced chi-square
values. The spacings between peaks corresponded to fre-
quency spacings between dierent F (total atomic angu-
lar momentum) state levels. We determined these fre-
quency spacings as well as the hyperne coupling con-
stants. Our results are in good agreement with the ac-
cepted values and support the theory of hyperne struc-
ture and Doppler-free saturated absorption spectroscopy.
Acknowledgments
The author acknowledges Sara L. Campbells equal
partnership in the preparation and execution of this
experiment. The author also thanks Professor Ner-
gis Mavalvala, Stephen Jaditz, Regina Yopak, and Dr.
Emily Edwards for their help and input.
[1] G. K. Woodgate. Elementary Atomic Structure. Oxford
University Press, 1980.
[2] D. J. Griths. Introduction to Quantum Mechanics. Pren-
tice Hall, 2005.
[3] MIT Department of Physics. Doppler-free saturated ab-
sorption spectroscopy. Lab Guide, 2009.
[4] A. Melissinos et al. Experiments in Modern Physics. El-
sevier Science, second edition, 2003.
[5] G. P. Barwood et al. Frequency measurements on rb at
780 nm and 795 nm. Appl. Phys., B53, 1991.