Saturated Absorption Spectroscopy

Experiment SAS

University of Florida — Department of Physics

PHY4803L — Advanced Physics Laboratory

Overview saturated absorption spectrometer for Cs

and Rb, Amer. J. of Phys. 60, 1098-1111
You will use a tunable diode laser to carry out
(1992).
spectroscopic studies of the rubidium atom.
You will measure the Doppler-broadened ab- 3. John C. Slater, Quantum Theory of
sorption profiles of the D2 transitions at Atomic Structure, Vol. I, (McGraw-Hill,
780 nm and then use the technique of satu- 1960)
rated absorption spectroscopy to improve the
resolution beyond the Doppler limit and mea-
Theory
sure the nuclear hyperfine splittings, which are
less than 1 ppm of the wavelength. A Fabry- The purpose of this section is to outline the
Perot optical resonator is used to calibrate the basic features observed in saturated absorp-
frequency scale for the measurements. tion spectroscopy and relate them to simple
Saturated absorption experiments were atomic and laser physics principles.
cited in the 1981 Nobel prize in physics and re-
lated techniques have been used in laser cool- Laser interactions — two-level atom
ing and trapping experiments cited in the 1997
Nobel prize as well as Bose-Einstein condensa-We begin with the interactions between a laser
field and a sample of stationary atoms having
tion experiments cited in the 2001 Nobel prize.
Although the basic principles are straightfor-only two possible energy levels. Aspects of
thermal motion and multilevel atoms will be
ward, you will only be able to unleash the full
power of saturated absorption spectroscopy by treated subsequently.
carefully attending to many details. The difference ∆E = E1 − E0 between the
excited state energy E1 and the ground state
energy E0 is used with Planck’s law to deter-
References mine the photon frequency ν associated with
1. Daryl W. Preston Doppler-free saturated transitions between the two states:
absorption: Laser spectroscopy, Amer. J. ∆E = hν0 (1)
of Phys. 64, 1432-1436 (1996).
This energy-frequency proportionality is why
2. K. B. MacAdam, A. Steinbach, and C. energies are often given directly in frequency
Wieman A narrow band tunable diode units. For example, MHz is a common energy
laser system with grating feedback, and a unit for high precision laser experiments.

SAS 1
SAS 2 Advanced Physics Laboratory

There are three transition processes involv-

ing atoms and laser fields:

Stimulated absorption in which the atom

starts in the ground state, absorbs a pho-
ton from the laser field, and then ends up
in the excited state.

Stimulated emission in which the atom starts

in the excited state, emits a photon with
the same direction, frequency, and polar-
ization as those in the laser field, and then
Figure 1: The Lorentzian line shape profile for
ends up in the ground state.
resonance absorption.
Spontaneous emission in which the atom
−8
starts in the excited state, emits a pho- For our rubidium levels, ∆E ≈ 2.5 × 10 eV
ton in an arbitrary direction unrelated to or Γ ≈ 6 MHz.
1

the laser photons, and then ends up in the The stimulated emission and absorption
ground state. processes are also described by a transition
rate — a single rate giving the probability per
Stimulated emission and absorption are as- unit time for a ground state atom to absorb
sociated with external electromagnetic fields a laser photon or for an excited state atom to
such as from a laser or thermal (blackbody) emit a laser photon. The stimulated transi-
radiation. We consider spontaneous emission tion rate is proportional to the laser intensity
first — a process characterized by a transition I (SI units of W/m2 ) and is only significantly
rate or probability per unit time for an atom different from zero when the laser frequency ν
in the excited state to decay to the ground is near the resonance frequency ν0 . This tran-
state. This transition rate will be denoted γ sition rate will be denoted αI, where
and is about 3.6 × 107 /s (or 36 MHz) for the
rubidium levels studied here. α = α0 L(ν, ν0 ) (3)
In the absence of an external field, any ini- and
tial population of excited state atoms would 1
L(ν, ν0 ) = (4)
decay exponentially to the ground state with 1 + 4(ν − ν0 )2 /Γ2
a mean lifetime ∆t = 1/γ ≈ 28 ns. In the rest gives the Lorentzian (or natural resonance)
frame of the atom, spontaneous photons are frequency dependence as shown in Fig. 1.
emitted in all directions with an energy spec-
1
trum having a mean E = hν0 and a full width The natural linewidth Γ normally represents the
sharpest observable energy distributions, but most
at half maximum (FWHM) ∆E given by the attempts to measure it in gases are confounded by
Heisenberg uncertainty principle ∆E∆t = h̄ Doppler shifts associated with the random thermal
or ∆E = γh̄. Expressed in frequency units, motion of the atoms, which broaden the emission or
the FWHM is called the natural linewidth and absorption spectrum by an order of magnitude or
given the symbol Γ. Thus more. Saturated absorption spectroscopy specifically
overcomes the Doppler-broadening limit by providing
γ for a two-photon interaction which only occurs for
Γ= (2) atoms with a lab frame velocity very near zero.
2π

January 30, 2020

Saturated Absorption Spectroscopy SAS 3

laser beam and affect the density of atoms (num-

beam ber per unit volume) in the ground and ex-
cited states. Moreover, Doppler shifts associ-
vapor cell photodiode ated with the random thermal motion of the
absorbing atoms must also be taken into ac-
count. There is an interplay among these ef-
Figure 2: Basic arrangement for ordinary laser fects which is critical to understanding satu-
absorption spectroscopy. rated absorption spectroscopy. We begin with
the basic equation describing how the laser in-
L(ν, ν0 ) also describes the spectrum of radia- tensity changes as it propagates through the
tion from spontaneous emission and the width sample and then continue with the effects of
Γ is the same for both cases. The maximum Doppler shifts and population changes.
transition rate α0 I occurs right on resonance
(ν = ν0 ) and for the rubidium transitions Laser absorption
studied here α0 ≈ 2 × 106 m2 /J.
The value of γ/α0 defines a parameter of the Because of stimulated emission and absorp-
atoms called the saturation intensity Isat tion, the laser intensity I(x) varies as it prop-
agates from x to x + dx in the medium.
γ
Isat = (5)
α0 Exercise 1 Show that
and is about 1.6 mW/cm2 for our rubidium
I(x + dx) − I(x) = −hναI(x)(P0 − P1 )n0 dx
transitions.2 Its significance is that when the
(6)
laser intensity is equal to the saturation in-
Hints: Multiply both sides by the laser beam
tensity, excited state atoms are equally likely
cross sectional area A and use conservation of
to decay by stimulated emission or by sponta-
energy. Keeping in mind that I(x) is the laser
neous emission.
beam intensity at some position x inside the
vapor cell, explain what the left side would rep-
Basic laser absorption spectroscopy resent. With n0 representing the overall den-
The basic arrangement for ordinary laser sity of atoms (number per unit volume) in the
absorption (not saturated absorption) spec- vapor cell, what does n0 Adx represent? P0 and
troscopy through a gaseous sample is shown P1 represent the probabilities that the atoms
in Fig. 2. A laser beam passes through the are in the ground and excited states, respec-
vapor cell and its intensity is measured by a tively, or the fraction of atoms in each state.
photodiode detector as the laser frequency ν They can be assumed to be given. What then
is scanned through the natural resonance fre- is the rate at which atoms undergo stimulated
quency. absorption and stimulated emission? Finally,
When a laser beam propagates through a recall that hν is the energy of each photon. So
gaseous sample, the two stimulated transi- what would the right side represent?
tion processes change the intensity of the laser
Equation 6 leads to
2
γ, Γ, α0 , Isat vary somewhat among the vari-
ous rubidium D2 transitions studied here. The values dI
given are only representative. = −κI (7)
dx

January 30, 2020

SAS 4 Advanced Physics Laboratory

where the absorption coefficient (fractional ab-

sorption per unit length)

κ = hνn0 α(P0 − P1 ) (8)

The previous exercise demonstrates that the

proportionality to P0 − P1 arises from the
competition between stimulated emission and
absorption and it is important to appreciate
the consequences. If there are equal num-
bers of atoms in the ground and excited state
(P0 − P1 = 0), laser photons are as likely to
be emitted by an atom in the excited state Figure 3: Maxwell-Boltzmann velocity distri-
as they are to be absorbed by an atom in the bution. The density of atoms vs. their velocity
ground state and there will be no attenuation component in one direction for room temper-
of the incident beam. The attenuation maxi- ature rubidium atoms.
mizes when all atoms are in the ground state
(P0 − P1 = 1) because only upward transitions direction will be important when taking into
would be possible. And the attenuation can account Doppler shifts and it is this compo-
even reverse sign (become an amplification as nent we refer to with the symbol v. The den-
it does in laser gain media) if there are more sity of atoms dn in the velocity group between
atoms in the excited state (P1 > P0 ). v and v+dv is given by the Boltzmann velocity
In the absence of a laser field, the ratio distribution:
of the atomic populations in the two energy √
m −mv2 /2kT
states will thermally equilibrate at the Boltz- dn = n0 e dv (9)
2πkT
mann factor P1 /P0 = e−∆E/kT = e−hν0 /kT . At
room temperature, kT (≈ 1/40 eV) is much With a standard deviation (proportional to
smaller than the hν0 (≈ 1.6 eV) energy dif- the width of the distribution) given by:
ference for the levels involved in this exper- √
iment and nearly all atoms will be in the σv = kT /m (10)
ground state, i.e., P0 − P1 = 1. While you
will see shortly how the presence of a strong this is just a standard Gaussian distribution
laser field can significantly perturb these ther-
mal equilibrium probabilities, for now we will 1
dn = n0 √ e−v /2σv dv
2 2
(11)
only treat the case where the laser field is weak 2π σv
enough that P0 − P1 = 1 remains a good ap-
proximation throughout the absorption cell. with a mean of zero — indicating the atoms
are equally likely to be going in either direc-
tion. It is properly normalized so that the in-
Doppler shifts
tegral over all velocities (−∞ → ∞) is n0 , the
Atoms in a vapor cell move randomly in all di- overall atom density. Note that the distribu-
rections with each component of velocity hav- tion’s variance σv2 increases linearly with tem-
ing a distribution of values. Only the com- perature and decreases inversely with atomic
ponent of velocity parallel to the laser beam mass.

January 30, 2020

Saturated Absorption Spectroscopy SAS 5

Exercise 2 Determine the width parameter

σv of the Maxwell-Boltzmann distribution for
room temperature rubidium atoms and com-
pare with the plot of Fig. 3.

Atoms moving with a velocity v see the laser

beam Doppler shifted by the amount ν (v/c).
We will take an equivalent, alternate view that
atoms moving with a velocity v have a Doppler
shifted resonance frequency
( )
v
ν0′ = ν0 1 + (12) Figure 4: Doppler profile. The absorption co-
c
efficient vs. the laser frequency offset from res-
in the lab frame. The sign has been chosen to onance has a Gaussian lineshape.
be correct for a laser beam propagating in the
positive direction so that the resonance fre-
(b) Determine the rough size of this range
quency is blue shifted to higher frequencies if
assuming Γ = 6 MHz. (ν0 = c/λ where
v is positive and red shifted if v is negative.
λ = 780 nm.) (c) Show that over this
The absorption coefficient dκ from a veloc-
range, e−mv /2kT remains relatively constant,
2

ity group dn at a laser frequency ν is then

i.e., compare the range with σv (Eq. 10).
obtained from Eq. 8 by substituting dn for n0
Consequently, the integral of Eq. 14 can be
(keeping in mind its dependence on v through
accurately determined as the integral of the
Eq. 9) and by adjusting the Lorentzian de-
Lorentzian times the value of the exponen-
pendence of α so that it is centered on the
tial at v =∫ vprobe . The integration of the
Doppler shifted resonance frequency ν0′ (keep-
Lorentzian L(ν, ν0′ )dv is easier to look up af-
ing in mind its dependence on v through
ter using Eq. 12 to convert dv = (c/ν0 )dν0′ .
Eq. 12).
(d) Show∫ that integrating the absorption co-
dκ = hνα0 (P0 − P1 )L(ν, ν0′ )dn (13) efficient dκ over velocity (from −∞ to ∞)
then gives a Gaussian dependence on ν cen-
The absorption coefficient from all atoms tered around the resonance frequency ν0
is then found by integrating over all velocity
κ = κ0 e−(ν−ν0 ) /2σν
2 2
(16)
groups. We treat the weak-laser case first set-
ting P0 − P1 = 1 so that with the width parameter given by
√ √
m kT
L(ν, ν0′ )e−mv /2kT dv
2
dκ = n0 hνα0 σ ν = ν0 (17)
2πkT mc2
(14)
and √
m c πΓ
Exercise 3 (a) Show that the Lorentzian κ0 = n0 hνα0 (18)
function L(ν, ν0′ ) at a particular ν is signifi- 2πkT ν0 2
cantly different from zero only for atoms with (e) Evaluate the Doppler-broadened width pa-
velocities within a very narrow range around rameter σν for room temperature rubidium
atoms and compare your results with the pro-
vprobe = c(ν/ν0 − 1) (15) file illustrated in Fig. 4.

January 30, 2020

SAS 6 Advanced Physics Laboratory

Populations [γ +2αI]−1 , which is less than 28 ns for our ru-

bidium transitions. Thus as atoms, mostly in
Now we would like to take into account the
the ground state, wander into the laser beam
changes to the ground and excited state pop-
at thermal velocities, they would only travel a
ulations arising from a laser beam propagat-
few microns before reaching equilibrium prob-
ing through the cell. The rate equations for
abilities.
the ground and excited state probabilities or
fractions become: C.Q. 1 Take into account velocity groups and
their corresponding Doppler shifts for the case
dP0
= γP1 − αI(P0 − P1 ) P0 − P1 ̸= 1. The calculation is performed
dt
dP1 as in Exercise 3. Show that the result after
= −γP1 + αI(P0 − P1 ) (19) integrating over all velocity groups is:
dt
κ = κ′0 e−(ν−ν0 ) /2σν
2 2
(24)
where the first term on the right in each equa-
tion arises from spontaneous emission and the where the width parameter, σν is the same as
second term arises from stimulated absorption before (Eq. 17), but compared to the weak-field
and emission. absorption coefficient (Eq. 18), the strong-field
coefficient decreases to
Exercise 4 (a) Show that steady state condi-
κ0
tions (i.e., where dP0 /dt = dP1 /dt = 0) lead κ′0 = √ (25)
to probabilities satisfying 1 + 2I/Isat

1
P 0 − P1 = (20) Laser absorption through a cell
1 + 2αI/γ
In the weak-field case κ at any frequency is
Hint: You will also need to use P0 + P1 = 1. given by Eq. 16 (with Eq. 18) and is indepen-
(b) Show that when this result is used in Eq. 8 dent of the laser intensity. In this case, Eq. 7
(for stationary atoms) and the frequency de- is satisfied by Beer’s law which says that the
pendence of α as given by Eqs. 3-4 is also in- intensity decays exponentially with distance
cluded, the absorption coefficient κ again takes traveled through the sample.
the form of a Lorentzian
I(x) = I0 e−κx (26)
hνn0 α0
κ= L′ (ν, ν0 ) (21) In the general case, κ is given by Eq. 24
1 + 2I/Isat
(with Eq. 25)
√ and at any frequency is propor-
′
where L is a standard Lorentzian tional to 1/ 1 + 2I/Isat . The general solution
to Eq. 7 for how the laser intensity varies with
′ 1
L (ν, ν0 ) = (22) the distance x into the cell is then rather more
1 + 4(ν − ν0 )2 /Γ′ 2 complicated than Beer’s law and is given in
with a power-broadened width parameter an appendix that can be found on the lab web
site. However, the strong-field I >> Isat be-
√
′
Γ = Γ 1 + 2I/Isat (23) havior is easily determined by neglecting the
1 compared to I/Isat so that Eq. 7 becomes
The approach to the steady state probabili- dI √
ties is exponential with a time constant around = −k I (27)
dx

January 30, 2020

Saturated Absorption Spectroscopy SAS 7

pump As mentioned, the stimulated emission and

probe beam
beam absorption rates are nonzero only when the
laser is near the resonance frequency. Thus,
vapor cell photodiode we will obtain P0 −P1 from Eq. 20 by giving α a
Lorentzian dependence on the Doppler shifted
resonance frequency:
Figure 5: Basic arrangement for saturated ab-
sorption spectroscopy. α = α0 L(ν, ν0′′ ) (29)

with the important feature that for the pump

where beam, the resonance frequency for atoms mov-
√ ing with a velocity v is
k = κ0 Isat /2e−(ν−ν0 )
2 /2σ 2
ν (28)
( )
v
ν0′′ = ν0 1− (30)
Saturated absorption — simple model c

Up to now, we have considered only a single This frequency is Doppler shifted in the direc-
laser beam propagating through the cell. Now tion opposite that of the probe beam because
we would like to understand what happens the pump beam propagates through the vapor
when a second laser propagates through the cell in the negative direction. That is, the res-
cell in the opposite direction. This is the ba- onant frequency for an atom moving with a
sic arrangement for saturated absorption spec- signed velocity v is ν0 (1 + v/c) for the probe
troscopy shown in Fig. 5. The laser beam beam and ν0 (1 − v/c) for the pump beam.
traveling to the right — the one we have been
considering up to now and whose absorption Exercise 5 (a) Plot Eq. 20 versus the detun-
is measured — is now called the probe beam. ing parameter
The second overlapping laser beam propagat-
ing in the opposite direction is called the pump δ = ν − ν0′′ (31)
beam. Both beams will be from the same laser
and so they will have the same frequency, even for values of I/Isat from 0.1 to 1000. (Use Γ =
as that frequency is scanned through the res- 6 MHz.) (b) Use your graphs to determine how
onance. the FWHM of the dips in P0 − P1 change with
With only a single weak laser propagat- laser power.
ing through the sample, P0 − P = 1 would
be a good approximation throughout the cell. Exercise 5 should have demonstrated that
With the two beams propagating through the for large detunings (δ ≫ Γ), P0 − P1 = 1 im-
cell, the probe beam will still be kept weak — plying that in this case the atoms are in the
weak enough to neglect its affect on the pop- ground state — the same as when there is no
ulations. However, the pump beam will be pump beam. On resonance, i.e., at δ = 0,
made strong — strong enough to significantly P0 − P1 = 1/(1 + 2I/Isat ), which approaches
affect the populations and thus change the zero for large values of I. This implies that
measured absorption of the probe beam. To atoms in resonance with a strong pump beam
understand how this comes about, we will will have equal populations in the ground and
again have to consider Doppler shifts. excited states (P0 −P1 = 0). That is, a strong,

January 30, 2020

SAS 8 Advanced Physics Laboratory

ues of P0 − P1 perturbed from the “pump off”

value of unity. Of course, whether there are
a significant number of atoms in either level
near v = vpump depends on the width of the
Boltzmann velocity distribution and the rela-
tive values of ν and ν0 . But if vpump is within
the distribution, the atoms near that velocity
will have perturbed populations.
The density of atoms in the ground state
dn1 = P1 dn plotted as a function of v will fol-
low the Maxwell-Boltzmann distribution ex-
cept very near v = vpump where it will drop
off significantly as atoms are promoted to the
excited state. This is called “hole burning”
as there is a hole (a decrease) in the density
of atoms in the ground state near v = vpump
(and a corresponding increase in the density of
atoms in the excited state) as demonstrated in
Fig. 6.
How does hole burning affect the probe
beam absorption? In Exercise 3 you learned
that the absorption at any frequency ν arises
only from those atoms moving with veloc-
ities near vprobe = c(ν/ν0 − 1). Also re-
Figure 6: Hole burning by the pump beam. call, the probe beam absorption is propor-
The density of ground state atoms is plotted tional to P0 − P1 , which we have just seen
vs. their velocity and becomes depleted near remains constant (≈ 1) except for atoms hav-
the velocity that Doppler shifts the laser fre- ing nearly the exact opposite velocities near
quency into resonance with ν0 . vpump = c(1 − ν/ν0 ). Therefore, when the
laser frequency is far from the natural reso-
nance (|ν − ν0 | ≫ Γ), the probe absorption
resonant laser field causes such rapid transi-
arises from atoms moving with a particular ve-
tions in both directions that the two popula-
locity in one direction while the pump beam is
tions equilibrate. The laser is said to “satu-
burning a hole for a completely different set of
rate” the transition, which is the origin of the
atoms with the opposite velocity. In this case,
name of the technique.
the presence of the pump beam will not affect
According to Eq. 30, the resonance condi-
the probe beam absorption which would follow
tion (δ = 0) translates to v = vpump where
the standard Doppler-broadened profile.
vpump = c(1 − ν/ν0 ) (32) Only when the laser frequency is very near
the resonance frequency (ν = ν0 , vprobe =
Consequently, for any frequency ν within the vpump = 0) will the pump beam burn a hole for
Doppler profile, only atoms near this veloc- atoms with velocities near zero which would
ity will be at zero detuning and will have val- then be the same atoms involved with the

January 30, 2020

Saturated Absorption Spectroscopy SAS 9

E
V-system Λ-system
2 2
1
hν2 hν0
hν1 hν1
1
0 0

Figure 8: Energy levels for two possible three-

Figure 7: Absorption coefficient vs. the laser level systems. The V-configuration with two
frequency offset from resonance for a two-level upper levels and the) Λ-system with two lower
atom at values of I/Isat = 0.1, 1, 10, 100, 1000 levels.
(from smallest dip to largest). It shows a
Gaussian profile with the saturated absorption
tional narrow absorption dips arising because
dip at ν = ν0 .
several upper or lower levels are close enough
in energy that their Doppler-broadened pro-
probe beam absorption at this frequency. The files overlap. Optical pumping occurs when
absorption coefficient would be obtained by the excited level can spontaneously decay to
taking P1 −P0 as given by Eq. 20 (with Eqs. 29 more than one lower level. It can significantly
and 30) to be a function of the laser frequency deplete certain ground state populations fur-
ν and the velocity v, using it in Eq. 13 (with ther enhancing or weakening the saturated ab-
Eqs. 9 and 12), and then integrating over ve- sorption dips.
locity. The result is a Doppler-broadened pro- The basics of crossover resonances can be
file with what’s called a saturated absorption understood within the three-level atom in ei-
dip (or Lamb dip) right at ν = ν0 . Numeri- ther the Λ- or V-configurations shown in Fig. 8
cal integration was used to create the profiles where the arrows represent allowed sponta-
shown in Fig. 7 for several values of I/Isat . neous and stimulated transitions. In this ex-
periment, crossover resonances arise from mul-
tiple upper levels and so we will illustrate with
Multilevel effects
the V-system. Having two excited energy lev-
Real atoms have multiple upper and lower els 1 and 2, the resonance frequencies to the
energy levels which add complexities to the ground state 0 are ν1 and ν2 , which are as-
simple two-level model presented so far. For sumed to be spaced less than a Doppler width
this experiment, transitions between two lower apart.
levels and four upper levels can all be Without the pump beam, each excited state
reached with our laser and add features called would absorb with a Doppler-broadened pro-
crossover resonances and a process called op- file and the net absorption would be the sum of
tical pumping. Crossover resonance are addi- two Gaussian profiles, one centered at ν1 and

January 30, 2020

SAS 10 Advanced Physics Laboratory

saturated absorption dip at ν = ν1 . A similar

argument predicts a saturated absorption dip
at ν = ν2 . Thus at ν = ν1 and at ν = ν2
there will be saturated absorption dips similar
to those occurring in two-level atoms.
A third dip, the crossover resonance, arises
at a frequency midway between — at ν12 =
(ν1 + ν2 )/2. All three dips are illustrated in
Fig. 9 assuming I/Isat = 10 and assuming
that the excited levels have the same spon-
taneous transitions rate γ and the same stim-
ulated rate constant α0 .
At the crossover frequency, the pump and
Figure 9: Absorption coefficient vs. the laser
probe beams are resonant with the same
frequency for a V-type three-level atom. The
two opposite velocity groups: v ≈ ±c (ν2 −
three Bell shaped curves are with the pump
ν1 )/2ν12 . Atoms at one of these two velocities
off and give the Doppler-broadened absorp-
will be resonant with one excited state and
tion from the individual resonances at ν1 and
atoms at the opposite velocity will be resonant
ν2 and their sum. The pump on curve shows
with the other excited state. The pump beam
normal saturated absorption dips at ν = ν1
burns a hole in the ground state populations
and ν = ν2 and a crossover resonance midway
at both velocities and these holes affect the
between.
absorption of the probe beam, which is simul-
taneously also arising from atoms with these
one centered at ν2 . If the separation |ν1 − ν2 |two velocities.
is small compared to the Doppler width, they
C.Q. 2 Explain why the crossover resonance
would appear as a single broadened absorption
dip disappears as the laser frequency changes
profile.
by a few linewidths from the mid-frequency
When the pump beam is turned on, two where the dip occurs. Remember, in this region
holes are burned in the ground state velocity each beam will still be interacting with two ve-
distribution at velocities that put the atoms in locity groups: one group for one excited state
resonance with ν1 and ν2 . These two velocities and another for the other excited state. For ex-
would depend on the laser frequency ν. For ex- ample, consider the case where the laser goes
ample, at ν = ν1 , the probe absorption involv- above the mid-frequency by a small amount.
ing upper state 1 is from v ≈ 0 atoms while the How would the velocities change for the two
probe absorption involving the higher-energy groups of atoms involved in probe beam absorp-
state 2 arises from some nonzero, negative- tion? How would the velocities change for the
velocity atoms. At this frequency, the pump two groups of atoms involved in pump beam
beam burns one hole in the ground state for absorption?
v ≈ 0 atoms due to upper state 1 and another
hole for some nonzero, positive-velocity atoms Optical pumping in rubidium occurs where
due to upper state 2. As with the two-level one excited level can decay to two different
system, the hole at v ≈ 0 leads to a decreased lower levels. It can be modeled in terms of the
absorption to upper state 1 and produces a Λ-type three-level atom, where the two lower

January 30, 2020

Saturated Absorption Spectroscopy SAS 11

levels are separated in energy by much more Energy levels in rubidium

than a Doppler width.
One common application of saturated absorp-
Assume the laser beam is resonant with tion spectroscopy is in measuring the hyper-
atoms in only one of the lower levels, but fine splittings of atomic spectral lines. They
atoms in the excited level spontaneously de- are so small that Doppler broadening normally
cay to either lower level more or less equally. makes it impossible to resolve them. You
Then, each time an atom in the “resonant” will use saturated absorption spectroscopy to
lower level is promoted by the laser to the ex- study the hyperfine splittings in the rubidium
cited level, some fraction of the time it will atom and thus will need to know a little about
decay to the “non-resonant” lower level. Once its energy level structure.
in the non-resonant lower level, the atom no In principle, quantum mechanical calcula-
longer interacts with the laser field. It be- tions can accurately predict the energy levels
comes “shelved” and unable to contribute to and electronic wavefunctions of multielectron
the absorption. Analysis of the level popula- atoms. In practice, the calculations are diffi-
tions then requires a model where atoms out- cult and this section will only present enough
side the laser beam interaction volume ran- of the results to appreciate the basic structure
domly diffuse back into it thereby replenishing of rubidium’s energy levels. You should con-
the resonant lower level populations. Depend- sult the references for a more in-depth treat-
ing on the laser intensity and beam geometry, ment of the topic.
the lower level populations can be significantly The crudest treatment of the energy lev-
altered by optical pumping. els in multielectron atoms is called the central
field approximation (CFA). In this approxima-
As with the population variations arising tion the nuclear and electron magnetic mo-
from stimulated emission and absorption, op- ments are ignored and the atomic electrons
tical pumping also depends on the laser fre- are assumed to interact, not with each other,
quency and atom velocities. Optical pumping but with an effective radial electric field aris-
due to the pump beam can drastically deplete ing from the average charge distribution from
resonant ground state atoms and can signif- the nucleus and all the other electrons in the
icantly increase the size of the saturated ab- atom.
sorption dips. Optical pumping due to the Solving for the energy levels in the CFA
probe beam can also affect saturated absorp- leads to an atomic configuration in which each
tion measurements, particularly if the probe electron is described by the following quantum
laser intensity is high. In this case, the probe numbers:
beam shelves the atoms involved in the ab-
sorption at all frequencies, not just at the sat- 1. The principal quantum number n (al-
urated absorption dips. lowed integer values greater than zero)
characterizes the radial dependence of the
Surprisingly, optical pumping does not sig- wavefunction.
nificantly change the predicted form for the
absorption signals. Largely, its effect is to 2. The orbital angular momentum quantum
change the widths (Γ’s), the saturation inten- number ℓ (allowed values from 0 to n −
sities (Isat ’s) and the strengths (κ’s) appearing 1) characterizes the angular dependence
in the formulae. of the wavefunction and the magnitude

January 30, 2020

SAS 12 Advanced Physics Laboratory

of the orbital angular momentum ℓ of an Fine structure levels

individual electron.3
Within a configuration, there can be several
fine structure energy levels differing in the en-
3. The magnetic quantum number mℓ (al-
ergy associated with the coulomb and spin-
lowed values from −ℓ to ℓ) further char-
orbit interactions. The coulomb interaction is
acterizes the angular dependence of the
associated with the normal electrostatic po-
wavefunction and the projection of ℓ on a
tential energy kq1 q2 /r12 between each pair of
chosen quantization axis.4
electrons and between each electron and the
nucleus. (Most, but not all of the coulomb in-
4. The electron spin quantum number s
teraction energy is included in the configura-
(only allowed value equal to 1/2) charac-
tion energy.) The spin-orbit interaction is as-
terizes the magnitude of the intrinsic or
sociated with the orientation energy −µ · B of
spin angular momentum s of an individ-
the magnetic dipole moment µ of each electron
ual electron.
in the internal magnetic field B of the atom.
The form and strength of these two interac-
5. The spin projection quantum number ms
tions in rubidium are such that the energy
(allowed values ±1/2) characterizes the
levels are most accurately described in the L-
projection of s on a chosen quantization
S or Russell-Saunders coupling scheme. L-S
axis.
coupling introduces new angular momentum
quantum numbers L, S, and J as described
The rubidium atom (Rb) has atomic num- next.
ber 37. In its lowest (ground state) config-
uration it has one electron outside an inert
gas (argon) core and is described with the no- 1. L is the quantum number describing the
tation 1s2 2s2 2p6 3s2 3p6 3d10 4s2 4p6 5s. The inte- magnitude of the total orbital angular
gers 1 through 5 above specify principal quan- momentum L, which is the sum of the
tum numbers n. The letters s, p, and d specify orbital angular momentum of each elec-
tron:
orbital angular momentum quantum numbers ∑
ℓ as 0, 1, and 2, respectively. The superscripts L= ℓi (33)
indicate the number of electrons with those
values of n and ℓ. 2. S is the quantum number describing the
The Rb ground state configuration is said magnitude of the total electronic spin an-
to have filled shells to the 4p orbitals and a gular momentum S, which is the sum of
single valence (or optical) electron in a 5s or- the spin angular momentum of each elec-
bital. The next higher energy configuration tron:
∑
has the 5s valence electron promoted to a 5p S= si (34)
orbital with no change to the description of
the remaining 36 electrons. 3. J is the quantum number describing the
magnitude of the total electronic angular
3
Angular momentum operators such as ℓ satisfy an momentum J, which is the sum of the to-
eigenvalue equation of the form ℓ2 ψ = ℓ(ℓ + 1)h̄2 ψ. tal orbital and total spin angular momen-
4
Angular momentum projection operators such as
ℓz satisfy an eigenvalue equation of the form ℓz ψ = tum:
mℓ h̄ψ. J=L+S (35)

January 30, 2020

Saturated Absorption Spectroscopy SAS 13

The values for L and S and J are specified than the electron magnetic moment and this is
in a notation (2S+1) LJ invented by early spec- why the hyperfine splittings are so small. The
troscopists. The letters S, P, and D (as with nuclear magnetic moment is proportional to
the letters s, p, and d for individual electrons) the spin angular momentum I of the nucleus,
are used for L and correspond to L = 0, 1, whose magnitude is described by the quantum
and 2, respectively. The value of (2S + 1) is number I. Allowed values for I depend on nu-
called the multiplicity and is thus 1 for S = 0 clear structure and vary with the isotope.
and called a singlet, 2 for S = 1/2 (doublet), The hyperfine energy levels depend on the
3 for S = 1 (triplet), etc. The value of J (with the total angular momentum F of the atom:
allowed values from |L − S| to L + S) is anno- the sum of the total electron angular momen-
tated as a subscript to the value of L.5 tum J and the nuclear spin angular momen-
The sum of ℓi or si over all electrons in tum I:
any filled orbital is always zero. Thus for Rb F=J+I (36)
configurations with only one valence electron,
The magnitude of F is characterized by the
there is only one allowed value for L and S:
quantum number F with allowed values from
just the value of ℓi and si for that electron.
|J − I| to J + I. Each state with a different
In its ground state (5s) configuration, Rb is
value of F will have a slightly different energy
described by L = 0 and S = 1/2. The only
due to the interaction of the nuclear magnetic
possible value for J is then 1/2 and the fine
moment and the internal field of the atom.
structure state would be labeled 2 S1/2 . Its
There is no special notation for the labeling
next higher (5p) configuration is described by
of hyperfine states and F is usually labeled
L = 1 and S = 1/2. In this configuration
explicitly in energy level diagrams.
there are two allowed values of J: 1/2 and 3/2
There are two naturally occurring isotopes
and these two fine structure states are labeled
2
of Rb: 72% abundant 87 Rb with I = 3/2
P1/2 and 2 P3/2 .
and 28% abundant 85 Rb with I = 5/2. For
both isotopes, this leads to two hyperfine lev-
Hyperfine levels els within the 2 S1/2 and 2 P1/2 fine structure
Within each fine structure level there can be levels (F = I − 1/2 and F = I + 1/2) and four
an even finer set of hyperfine levels differing in hyperfine levels within the 2 P3/2 fine structure
the orientation energy (again, a −µ · B type level (F = I − 3/2, I − 1/2, I + 1/2, I + 3/2).
energy) associated with the nuclear magnetic The energies of the hyperfine levels can be
moment in the magnetic field of the atom. expressed (relative to a “mean” energy EJ for
The nuclear magnetic moment is much smaller the fine structure state) in terms of two hyper-
fine constants A and B by the Casimir formula
5
The terms singlet, doublet, triplet etc. are associ-
ated with the number of allowed values of J typically
possible with a given L and S (if L ≥ S). Historically, κ
EF = EJ + A (37)
the terms arose from the number of closely-spaced 2
spectral lines typically (but not always) observed in 3κ(κ + 1)/4 − I(I + 1)J(J + 1)
the decay of these levels. For example, the sodium + B
doublet at 589.0 and 589.6 nm occurs in the decay of
2I(2I − 1)J(2J − 1)
the 2 P1/2 and 2 P3/2 fine structure levels to the 2 S1/2
ground state. While the 2 P1/2 and 2 P3/2 are truly a
where κ = F (F + 1) − J(J + 1) − I(I + 1). (If
doublet of closely-spaced energy levels, the 2 S1/2 state either I = 1/2 or J = 1/2, the term containing
has only one allowed value of J. B must be omitted.)

January 30, 2020

SAS 14 Advanced Physics Laboratory

Iso fss A B hyperfine levels are less than 0.00003 eV.

85 2
Rb S1/2 1011.91 —
2
P1/2 120.72 —
2
Transitions
P3/2 25.01 25.88
87 2
Rb S1/2 3417.34 — The 2 S1/2 to 2 P1/2 transitions are all around
2
P1/2 406.20 — 795 nm while the 2 S1/2 to 2 P3/2 transitions are
2
P3/2 84.80 12.52 all around 780 nm. We will only discuss the
780 nm transitions that can be reached with
the laser used in this experiment. Dipole tran-
Table 1: Hyperfine constants A and B (in
sitions follow the selection rule ∆F = 0, ±1.
MHz) for the lowest three fine structure states
Thus, in each isotope, the allowed transitions
in the two naturally-occurring Rb isotopes
from the 2 S1/2 to 2 P3/2 fall into two groups of
three, with each group labeled by the F of the
2
Figure 10 shows the configuration-fine S1/2 state. Because the hyperfine splitting be-
structure-hyperfine energy structure of the Rb tween the two 2 S1/2 levels is large compared to
atom. The scaling is grossly inaccurate. The the hyperfine splittings among the four 2 P3/2
energy difference between the 5s and 5p con- levels, the groups will be well separated from
figurations is around 1.6 eV, the difference be- each other. Within each group, the three pos-
tween the 2 P1/2 and 2 P3/2 fine structure levels sible transitions can be labeled by the F ′ of
is around 0.03 eV, and the differences between the 2 P3/2 state. These three transitions will
be more closely spaced in energy.

C.Q. 3 The energy level diagrams of Fig. 11

energy 2S+1 are based on the information in Table 1 with
LJ
nl coulomb and
and Eq. 37. Use the same information to cre-
hyperfine F=
configuration
spin-orbit 4 ate a table and scaled stick spectrum for all
energy 2 3
P 3/2 2
1 twenty-four 780 nm resonances (including the
780.0 nm
5p 3 crossover resonances) occurring in both iso-
2 2
P 1/2 topes. The horizontal axis should be at least
794.8 nm
25 cm long and labeled in MHz. For the fig-
5s
3 ure, draw a one- or two-centimeter vertical
2S 2
1/2 line along the horizontal frequency axis at the
position of each resonance. Label the isotopes
and the F of the 2 S1/2 level for each of the four
Figure 10: Energy structure of the lowest lev- groups and within each group label the value of
els in 85 Rb, (I = 5/2). Energy increases to- F ′ in the 2 P3/2 level for each of the three nor-
ward the top, but the levels are not shown mal resonances and the two values of F ′ for
to scale; the separation between the 5s and 5p each of the three crossover resonances. Also
configuration energies is about 50 times bigger label their transition energies (in MHz) rela-
than the spacing between the 2 P1/2 and 2 P3/2 tive to the overall 385 THz average transition
fine structure levels, and it is about 105 times frequency. Hint: you should get the lowest two
bigger than the spacing within any group of resonances in 87 Rb — a normal resonance be-
hyperfine levels. tween the F = 2 and F ′ = 1 levels 2793.1

January 30, 2020

Saturated Absorption Spectroscopy SAS 15

Figure 11: Energy levels for 85 Rb and 87 Rb (not to scale). Note that the hyperfine splittings
are about an order of magnitude larger in the ground 2 S1/2 levels compared to the excited
2
P3/2 levels.

MHz below the average and a crossover reso- 4. Take care when changing optics in your
nance between the F = 2 and F ′ = 1, 2 levels experiment so as not to inadvertently
2714.5 MHz below the average. The highest place a highly reflective object (glass, mir-
energy resonance is again in 87 Rb; a normal ror, post) into the beam.
resonance from F = 1 to F ′ = 2 4198.7 MHz
The layout of the saturated absorption spec-
above the average.
trometer is shown in Fig. 12. Two weak beams
are reflected off the front and rear surfaces of a
Apparatus thick beam splitter (about 4% each). One (the
probe beam) is directed through the rubidium
Laser Safety
cell from left to right and then onto a photo-
The diode laser used in this experiment has diode detector. The other is directed through
sufficient intensity to permanently damage a Fabry-Perot interferometer and onto a sec-
your eye. In addition, its wavelength (around ond photodiode detector for calibrating the
780 nm) is nearly invisible. Safety goggles frequency scan.
must be worn when working on this ex- The strong beam is transmitted through the
periment. You should also follow standard thick beam splitter. The two lenses after the
laser safety procedures as well. beam splitter should be left out at first and
only used if needed to expand the laser beam.
1. Remove all reflective objects from your This is the pump beam passing through the
person (e.g., watches, shiny jewelry). cell from right to left, overlapping the probe
2. Make sure that you block all stray reflec- beam and propagating in the opposite direc-
tions coming from your experiment. tion. (The intersection angle is exaggerated in
the figure.)
3. Never be at eye level with the laser You should not adjust the components from
beam (e.g., by leaning down). the thick beam splitter back to the laser.

January 30, 2020

SAS 16 Advanced Physics Laboratory

M
FPM Fabry-Perot FPM PD2

M
NBS PD1
Rb cell
M probe beam
laser

λ/4
A1
pump beam
M M
APP PBS TBS λ/4 L1 L2

Figure 12: Optical layout of the saturated absorption spectrometer. Key: M mirror; A aper-
ture; APP anamorphic prism pair; PBS polarizing beam splitter; TBS thick beam splitter;
λ/4 quarter wave plate; L lens; NBS non-polarizing beam splitter; FPM Fabry-Perot mirror
PD photodiode detector.

The diode laser voltage (0-150 V) from the PZT controller (de-
scribed later). The voltage can be adjusted
A schematic of our diode laser is shown in
manually or under computer-control.
Fig. 13. The American Journal of Physics ar-
ticle on which its construction is based as well As the laser cavity length varies, because
as manufacturer literature on the laser diode the number of waves in the cavity will stay
used in the laser can be found in the auxiliary constant (for a while), the wavelength and
material. therefore the laser frequency will vary as well.
The frequency of the laser depends on three But if the cavity length change goes too far,
parameters: the number of half-wavelengths inside the cav-
ity will jump up or down (by one or two) so
• The laser temperature
that the laser frequency will jump back into
• The laser current the range set by the temperature and current.
These jumps are called mode hops. Continu-
• The position and orientation of the grat- ous frequency tuning over the four hyperfine
ing groups (≈ 8 GHz) without a mode hop can
only be achieved if the settings for the tem-
The laser temperature and current set the perature and current are carefully optimized.
frequency range over which the diode laser
will operate (coarse tuning). Within this The beam profile (intensity pattern in a
range, the laser frequency can be continuously cross-section of the laser beam) from the laser
scanned using the position and orientation of head is approximately a 1.5 × 3 mm elliptical
the grating (fine tuning). Fine tuning of the spot. An anamorphic prism pair is used to
laser frequency is accomplished by a piezoelec- transform this to a circular pattern. In addi-
tric transducer (PZT) located in the grating tion, lenses may be used to change the beam
mount. The PZT expands in response to a diameter.

January 30, 2020

Saturated Absorption Spectroscopy SAS 17

particularly from the polarizing and non-

EHDP polarizing beam splitters, and use beam blocks
for any that are directed toward areas where
people have access.
ODVHUGLRGH

7 Photodetectors
IRFXV OHQV
7
Two photodetectors are used to monitor the
JUDWLQJ laser beam intensities. A bias voltage from a
3=7 battery (≈ 22 V) is applied to the photode-
tector which then acts as a current source.
The current is proportional to the laser power
impinging on the detector and is converted to
FRDUVH YHUW a voltage (for measurement) either by send-
IUHT WLOW ing it through a low noise current amplifier or
EDVHSODWH by letting it flow through a resistor to ground.
HQFORVXUH With the latter method, the voltage developed
across the resistor subtracts from the battery
voltage and the resistance should be adjusted
to keep it below one volt.
Figure 13: Schematic of our diode laser. Ex-
cept for the thermistors (T), the temperature
control elements and electrical connections are Fabry-Perot Interferometer
not shown. Once the laser temperature has stabilized and
a laser current has been set, a voltage ramp
Beam Paths will be applied to the PZT causing the laser
frequency to smoothly sweep over several res-
Several beams are created from the single onance frequencies of the rubidium atoms in
beam coming from the laser. The polarizing the cell. The laser frequency during the sweep
beam splitter and the quarter-wave plate help is monitored using a confocal Fabry-Perot in-
prevent the pump beam from feeding back into terferometer. The Fabry-Perot consists of two
the laser cavity thereby affecting the laser out- partially transmitting, identically-curved mir-
put. rors separated by their radius of curvature as
The thick beam splitter creates the three shown in Fig. 14.
beams used in the experiment. Two beams — With the laser well aligned going into the
one reflected off the front surface and one re- Fabry-Perot cavity, Fig. 15 shows how much
flected off the back surface — create the probe is transmitted through the cavity as a func-
beam and the beam sent through the Fabry- tion of the laser frequency. The Fabry-Perot
Perot interferometer. Each of these beams only transmits the laser beam at a set of dis-
contains about 4% of the laser output. About crete frequencies separated by the free spectral
92% passes through this beam splitter to be- range (FSR). For a confocal Fabry-Perot it is
come the pump beam. given by:
c
You should also check for additional beams, FSR = (38)
4L

January 30, 2020

SAS 18 Advanced Physics Laboratory

Rb vapor cell
L The Rb vapor cell is a glass cell filled with nat-
ural rubidium having two isotopes: 85 Rb and
87
Rb. The vapor pressure of rubidium inside
the cell is determined by the cell temperature
and is about 4×10−5 Pa at room temperature.

Low-noise current amplifier

The low noise current amplifier can be used
Figure 14: The optical resonator or Fabry- to amplify and filter the photodetector signal.
Perot interferometer consists of two partially This signal will change at a rate which de-
transmitting curved mirrors facing each other. pends on the rate at which you scan the laser
frequency. For a very slow scan, the interest-
ing parts of the signal will change very slowly
where L is the distance between the mirrors. and you can use the low pass filter function of
Consequently, as the frequency of the laser the amplifier to reduce the higher frequency
beam going into the cavity is swept, the trans- noise like 60Hz pick-up. The low pass filter
mission (as monitored by a photodetector on should not change your signal if the cutoff fre-
the other side of the cavity) will show a series quency is set properly. Another possibility is
of peaks separated by the FSR. These peaks to scan rather fast and use the bandpass filter
(or fringes) will be monitored simultaneously function to reduce the low and high frequency
with absorption through the rubidium and will noise. You should play with these options a
serve as frequency markers for how the laser bit and try to find the optimum setting for
frequency is changing during the sweep. your scan rate and detector noise.

Alignment procedure
Most students have difficulties aligning the
position and propagation direction of a laser
beam simultaneously — usually due to a mix-
ture of inexperience and impatience. In the
FWHM following we describe the standard alignment
procedure. As usual, standard procedures are
FSR
not applicable to all cases but they should give
you an idea how do do it.
Task: Align a laser beam to propagate
along a particular axis in the apparatus, e.g.,
the axis of our Fabry-Perot resonator. This
alignment procedure is called beam walking
and requires two mirrors in adjustable mirror
Figure 15: The transmitted intensity for a
mounts.
well aligned Fabry-Perot interferometer shows
sharp peaks equally spaced in frequency. • Coarse alignment:

January 30, 2020

Saturated Absorption Spectroscopy SAS 19

2. Optimize the signal with the first

mirror in the horizontal and vertical
mirror b a directions.
3. Optimize the signal with the second
b
mirror in the horizontal and vertical
directions.
4. Use the first mirror to de-optimize
laser beam
the signal very slightly in the hori-
mirror a
zontal direction. Remember the di-
rection you moved the mirror.
5. Use the second mirror to re-optimize
the signal again.
Figure 16: Procedure to align the laser beam
(dashed line) along a prescribed line (solid line 6. Compare this signal with the signal
ab). Pick two reference points a and b and use you had before step 4. If the sig-
the first mirror to align the laser to point a and nal improved repeat step 4 moving
the second mirror to align the laser to point b. in the same direction. If the signal
Iterate until laser beam is perfectly aligned. got worse, repeat step 4 moving in
the opposite direction.
1. Pick two reference points along the 7. Iterate until you can’t further im-
target axis. These points should prove the signal.
be reasonably far apart from each 8. Do the same for the vertical direc-
other. tion until you can’t improve the sig-
2. Use the first alignment mirror to nal anymore.
align the laser on the first reference 9. Iterate steps 4 to 8 until you can’t
point. improve the signal furthermore.
3. Use the second alignment mirror to This procedure requires some patience and can
align the laser on the second refer- not be done in five minutes.
ence point.
4. Iterate points 2 and 3 until the laser The Data Acquisition Program
beam is aligned as well as possible
with respect to the optical axis. YouRead the reference material for detailed in-
might overshoot in step 2 or step 3 formation on our PCI-MIO-E4 multifunction
occasionally to accelerate the proce-data acquisition (DAQ) board located inside
dure. the computer. Here we will only present what
is needed for this experiment. The func-
• Fine alignment or optimizing the signal: tionality used here is the DAQ board’s abil-
ity to measure analog voltages using analog-
1. Display the signal you wish to opti- to-digital converters (ADC’s) and to produce
mize. For example, the photodetec- analog voltages using digital-to-analog con-
tor signal after the Fabry-Perot res- verters (DAC’s). The input and output volt-
onator. ages from the ADC’s and DAC’s are connected

January 30, 2020

SAS 20 Advanced Physics Laboratory

to and from the experiment through the BNC- for the DAC1 voltage (vref ) (labeled ramp ref
2090 interface box. (V)) and the value of Nmax (labeled ramp size).
The DAQ board will be programmed to gen- The speed of the sweep is determined by the
erate a (discrete) voltage sweep which is sent front panel setting of the sampling rate control.
to the PZT controller to sweep the laser fre- It is specified in samples per second and gives
quency. While the voltage sweep is occurring, the rate at which the DAC0 output changes
the DAQ board is programmed to simultane- during a sweep. Thus, for example, with the
ously read any photodetectors monitoring the ramp size set for 2048, the DAC1OUT set for
probe, pump, and/or the Fabry-Perot laser 4 V, and the Sweep rate set for 10000/s, the
beam intensities. ramp will take about 204.8 ms to go from 0
There are two, 12-bit DAC’s (DAC0 and to 2 V (2048 · 4V/4096) and another 204.8 ms
DAC1) on our DAQ board and both are used to come back down to zero, after which it re-
to generate the sweep. They are set to run in peats.
unipolar mode so that when they are sent inte- The triangle ramp should be connected to
gers N from 0 to 4095, they produce propor- EXT INPUT on the MDT691 Piezo Driver.
tional voltages vout from 0 to some reference This module amplifies the ramp by a factor of
voltage vref . 15 and adds an offset controlled by an OUT-
N PUT ADJ knob on the front of the module to
vout = vref (39) produce the actual voltage ramp at its OUT-
4095
PUT. This output drives the PZT in the laser
The voltages appear at the DAC0OUT and head which in turn changes the laser cavity
DAC1OUT outputs on the DAQ interface box. length and thus the laser frequency. To change
The reference voltage vref for either DAC can the starting point of the ramp sent to the PZT
be chosen to be an on-board 10 V source or in the laser head, you should simply offset it
any value from 0 to 11 V as applied to the using the OUTPUT ADJ knob. Alternatively,
EXTREF input on the DAQ interface box. the front panel control ramp start can be set
Our data acquisition program runs DAC1 in at a non-zero value to have the DAC0 pro-
the 10 V internal reference mode to create a gramming integers begin at that value. For
fixed voltage at DAC1OUT, which is then con- example, setting the ramp start to 1024 and
nected to the EXTREF input and becomes the the ramp size to 2048 makes the programming
voltage reference vref for DAC0. The DAC0 integers sent to DAC0 to run from 1024 to
programming integers are a repeating triangu- 3072 (and back down again). Then, for exam-
lar ramp starting N=0, 1, 2 ..., Nmax and then ple, if vref were set for 8 V, the ramp would go
going backward from Nmax back to 0 (where from 2 V to 6 V.
Nmax is chosen by the user). This program- Voltages from the photodetector outputs
ming thereby generates a triangular voltage should be connected to any of the inputs to
waveform from zero to vmax and then back to the analog-to-digital converter (ADC) on the
zero, where interface box. These inputs are connected to
Nmax a multiplexer, which is a fast, computer con-
vmax = vref (40) trolled switch to select which input is passed
4095
on to be measured by the ADC. Before be-
The properties of the triangle ramp are ing measured by the ADC, the signal is ampli-
specified on the program’s front panel settings fied by a programmable gain amplifier (PGA).

January 30, 2020

Saturated Absorption Spectroscopy SAS 21

The gain of this amplifier and the configura- quency depends strongly on temperature
tion (unipolar or bipolar) is determined by the so wait for the temperature to stabilize
Range control on the front panel. You should before making other adjustments.
generally use the most sensitive range possible
for the signal you wish to measure so that the 3. Check that the laser beam paths are all
12-bit precision of the ADC is most fully uti- clear of reflective objects.
lized. When measuring more than one analog
input, it is also preferable if they all use the 4. Turn on the LFI-4505 laser diode current
same range so that the PGA does not have to controller. When you push the OUTPUT
change settings as it reads each input. ON button, the laser current comes on
and the diode laser starts lasing. The
laser current needed is typically some-
Measurements where in the range from 60-90 mA. For
Turning the Diode Laser On and Off now, look at the laser output beam with
an infrared viewing card and watch the
1. Ask the instructor to show you how to laser beam come on as you increase the
turn on the laser electronics. The LFI- laser current past the 40 mA or so thresh-
3526 temperature controller will maintain old current needed for lasing. Watch the
the aluminum baseplate at a temperature beam intensify as you bring the current
about 10 degrees below room temperature up to around 70 mA.
using a thermoelectric cooler. The home-
made interface box (with a white panel 5. Later, when you need to turn the laser
on the front and a power switch on the off, first turn down the laser current to
back) contains a second temperature con- zero, then hit the OUTPUT ON button
trol stage which heats the diode mount- again, then push the power button. Af-
ing block about 2 degrees with a resis- ter the laser is powered off, the rest of
tive heater. Check that the fan under the the laser electronics can turned off by the
laser head comes on and is suitably po- switch on the outlet box. The computer
sitioned to blow on the laser head heat and detector electronics should be pow-
sink. Watch out that you don’t skin your ered from another outlet box and should
fingers moving the fan as the blade is un- be turned off separately. Also remember
shielded. The fan sits on a thin foam base to turn off the battery-powered photodi-
to minimize vibrations. ode detectors.

2. Do not change the temperature controller Tuning the laser

set point, but learn how to read it and
write its present value in your lab note- 6. Align the beam through the center of the
book. The setting is the voltage across F-P and onto its detector. Align the
a 10 kΩ thermistor mounted to the laser probe beam through the center of the Rb
head’s aluminum baseplate and used for cell and onto its detector. Align the pump
feedback to stabilize the temperature. It beam to be collinear with the probe beam
should be set to produce the correct tem- but propagating in the opposite direction.
perature to allow the laser to operate at Turn on the detectors and the PAR cur-
a wavelength of 780 nm. The laser fre- rent preamp.

January 30, 2020

SAS 22 Advanced Physics Laboratory

(a) (b)

Figure 17: Single-beam Rb absorption spectra. (a) Shows a laser frequency sweep of about
8 GHz without a mode hop over the appropriate region of interest for the Rb hyperfine levels
near 780 nm. (b) Shows a sweep with a mode hop around 0.2 s.

7. Focus the video camera on the Rb cell. to 70 mA.

Block the probe beam.
10. Start the Analog IO data acquisition pro-
8. Wait for the temperature to stabilize and gram. Set the ramp ref (V) control to 3 V.
then adjust the laser current in the range Set the ramp size to 2000 channels. Leave
from 60-100 mA while monitoring the ru- the ramp start control at zero and the sam-
bidium cell with the video camera. When pling rate at 20,000 samples per second.
the current is set properly, the moni- This makes about a 1.5 V sweep — up
tor should show strong fluorescence all and back down — run at about 5/s. The
along the pump beam path inside the cell. 1.5 V ramp would normally be sufficient
There can be up to four distinct reso- to produce a laser frequency sweep of 6-9
nances. They are all very sharp and the GHz or more. However, the laser typi-
laser frequency may drift so you may need cally mode hops during the ramp.
to repeatedly scan the laser current back
and forth over a small range to continue 11. In this and the next few steps, you will
observing the fluorescence. Ask the in- look for evidence of mode hops in your
structor for help if no current in this range absorption spectrum and/or in the F-P
shows strong Rb fluorescence. markers and you will adjust the temper-
ature, current, and PZT offset to get a
9. With the strong pump fluorescence ob- long sweep that covers all four Rb reso-
served on the monitor, block the pump nances without a mode hop. All adjust-
beam and unblock the probe beam to see ments should be performed while running
the path of the probe beam and its weaker the sweep described above and monitor-
fluorescence on the monitor. The reso- ing the computer display of the absorp-
nances may be observed at two laser cur- tion spectrum and F-P markers, as well
rents separated by twenty milliamps or as the video display of the fluorescence.
more. If so, set it for the current closest At this point, the back and forth sweep

January 30, 2020

Saturated Absorption Spectroscopy SAS 23

should produce a blinking fluorescence on tering the beam on the detectors, mirrors,
the camera video. Turn off all preamp or splitters.
filtering. You may need to adjust the
gain settings on the DAQ board’s ADC 13. Adjust the PZT offset knob to get the
so that the measured signals are not satu- longest possible sweep without a mode
rating the ADC. Try to adjust the current hop while also trying to cover the fre-
preamplifier gain so that the absorption quency range for the four absorption
spectrum has just under 0.5 V baseline peaks. Adjust the laser current 1 or 2 mA
(the signal level away from the peaks) and one way or the other and then go back to
then be sure to adjust the ADC gain for the PZT knob to see if the sweep is im-
0-0.5V signals. Also adjust the F-P detec- proving in length or frequency coverage.
tor resistor so that the F-P peaks are also Repeat until you have a feel for how this
just under a 0.5 V and adjust the ADC works and have roughly found the best
gain for 0-0.5 V signals. possible sweep.

14. If no combination of laser current and

12. A good absorption spectrum with F-P PZT offset gives a frequency sweep that
markers (for only the half-sweep up of shows all four Rb hyperfine groups, the
the up and down triangular waveform) is laser temperature will need to be ad-
shown in Fig. 17(a). Fig. 17(b) shows justed. Save the best spectrum you cur-
what these two signals look like with a rently have for later comparisons. The
single mode hop about halfway into the temperature should be changed in small
sweep. Notice the break in the F-P mark- steps corresponding to about 0.05-0.10 kΩ
ers and how the section of the absorption on the 10 kΩ thermistor used to control it.
spectrum repeats after the 0.2 s mark as Start by increasing the setting by 0.05 kΩ
the laser frequency continues scanning af- (lowering the temperature). After chang-
ter the abrupt change in frequency caused ing it, go back to the previous step. If
by the mode hop. Even without any Rb the frequency sweep improved, but is still
absorption, the lasers intensity may not not long enough, change the tempera-
be very constant over the entire sweep. ture controller by another 0.05 kΩ in the
If this is the case, the photodetector cur- same direction. Otherwise, change it in
rent between absorption peaks from one the opposite direction. Continue iterat-
side of the sweep to the other will show ing these two steps until you have the de-
a significant slope. This is partly because sired sweep. Ask the instructor for help if
the beam from the laser changes direction this step is taking more than 60 minutes
a small amount as the piezo scans and the or if the temperature controller has been
beam may then move partially on and off changed by 0.25 kΩ without getting the
the detector by a small amount. However, needed sweep. Keep in mind, however,
if the detected laser intensity varies more that on some days it seems the best the
than about 10%, the beam may be hitting laser can do is a sweep that only covers
very near the edge of the detector or one three out of the four peaks. In this case,
of the mirrors or beam splitters thereby the four peaks can usually be observed
making the effect bigger than it could be. three at a time in two separate sweeps
Fix this problem if it occurs by better cen- with slightly different settings on the PZT

January 30, 2020

SAS 24 Advanced Physics Laboratory

or laser current controller. differences so this is not a problem. Then

the frequency at the markers can be fit to a
Preliminary Measurements quadratic or cubic polynomial in the time or
channel number where the F-P peak occurred
15. Measure the width (FWHM) and spac- and the polynomial coefficients can then be
ing between the F-P fringe peaks and de- used to convert the time or channel number of
termine the finesse, which is the ratio of spectral features to frequency units.
the spacing to the FWHM. If the finesse A second, slightly simpler technique would
is below three, you may want to adjust be to count whole marker spacings between
the beam steering to better align it to features and interpolate the location of fea-
the Fabry-Perot. Note the spacing be- tures as fractions of the spacing relative to the
tween the fringes. Measure the F-P mir- markers located near them. The whole and
ror spacing and calculate the free spectral fractional parts multiplied by the FSR would
range. Measure the distance (in channels) then give the frequency spacing between the
between the F-P frequency markers at features.
various places within the sweep and note
the nonlinearity of the sweep. The non-
Doppler broadening and Laser Absorp-
linearity will complicate the analysis be-
tion
cause the frequency spacing between data
points will not be constant. 17. Obtain a spectrum of the four absorption
groups (with the F-P frequency markers)
16. Change the sweep to 4000 points with and determine the spacing of the peaks
vref = 1.5 V to produce about the same in frequency. Use this information with
size sweep and explain how the F-P peaks your stick spectrum to identify the four
are affected. make sure all absorption groups. Does the frequency increase or
spectra have the F-P frequency markers decrease as the PZT ramp voltage goes
along with them. They will be needed to up?
convert the spacing between Rb absorp-
tion features (such as peak positions and 18. For each group, measure the Doppler-
widths) to frequency units. broadened FWHM and compare with pre-
dictions. Do you see any effects in the
There are several ways the F-P markers can FWHM due to the fact that each group
be used to perform the conversion. The ab- includes three transitions at different fre-
sorption spectra are normally plotted vs. time quencies?
or channel number. In one method, the lo-
cation of the F-P markers in time or chan- 19. Measure the absorption as a function of
nels would be determined using cursors. Then incident laser intensity for at least one
since the spacing between frequency markers of the four Doppler-broadened groups.
is constant and is given by the FSR (let’s call That is, use one graph cursor to mea-
it ∆ν), the frequency at consecutive markers sure the probe detector voltage (and con-
would be taken as ν0 , ν0 +∆ν, ν0 +2∆ν, ... The vert to a current using the preamp gain)
frequency ν0 is about 3.85 × 1014 Hz and not at maximum absorption. This reading
really determined with our apparatus. How- would correspond to the transmitted in-
ever, we will only be interested in frequency tensity at resonance. Set the other cur-

January 30, 2020

Saturated Absorption Spectroscopy SAS 25

sor to“eyeball” a background reading at use is the beam area. If the beam is larger
the same place — what the voltage read- than the detector, then the detector area (13
ing would be if the absorption peak were mm2 for the DET110) should be used. Keep
not there. This reading, again converted in mind that the beam probably has a roughly
to a photocurrent) would correspond to circular intensity profile with the central area
the incident intensity. fairly uniform and containing around 90% of
So that you can get the highest laser the power. The diameter of this central area is
power, move the Rb cell into the pump probably around 1/2 to 1/3 that of the beam
beam and move the detector behind the as measured on the IR viewing card.
cell so it will measure the pump beam ab- Make sure to get to low enough laser inten-
sorption. Do not change the laser path; sities that you see the fractional transmission
only adjust the Rb cell and photodetector become constant (Beer’s law region). Plot the
positions. data in such a way that you can determine
where the Beer’s law region ends and the frac-
Use neutral density filters to lower the in- tional absorption starts to decrease. This hap-
tensity of the beam. pens near Isat , so see if you can use your data
to get a rough estimate of this quantity.
Be sure to keep track of the gain setting
Plot your data in such a way that it shows
on the current preamp so that you will be
the prediction for the absorption or transmis-
able to convert the photodetector voltages to
sion for the strong-field case. There are several
photocurrents. Also be sure to check the
ways to do this. Explain your plot and how it
background photocurrent, i.e., with the laser
demonstrates the prediction.
blocked, and subtract this from any photocur-
rents measured with the laser on. This back- C.Q. 4 Use the measured weak-field trans-
ground subtraction is particularly important mission fraction and Beer’s law (Eq. 26) with
at low laser power where the background pho- the cell length to determine κ0 . Use this κ0
tocurrent can be a sizable fraction of the total. with Eq. 18 to determine α0 and then use Eq. 5
In the following discussion, photocurrent will to determine Isat . Compare this Isat with the
refer to background corrected photocurrent. rough value determined by the graph — where
The photocurrent is proportional to the Beer’s law starts to be violated and with the ap-
power incident on the detector and for the proximate value of 1.6 mW/cm2 given in the
Thorlabs DET110 detector, the responsivity text.
(conversion factor from incident laser power
to current) is around 0.5 A/W at 780 nm.
Rb spectra
You can also use the Industrial Fiber Optics
Photometer at a few of the laser intensities 20. Return the Rb cell to its position in
for comparison purposes. It reads directly in the probe beam. Move the photodetec-
watts, but is calibrated for a HeNe laser. The tor back to its original position monitor-
reading needs to be scaled down by a factor ing the probe beam absorption. If nec-
of about 0.8 for the 780 nm light used here. essary, realign the probe beam through
To get intensity, you will also have to esti- the splitter and onto the photodetector.
mate the laser beam diameter and determine Realign the pump beam to be collinear
how well you are hitting the detector. If the with the probe beam. Observe the sat-
beam is smaller than the detector, the area to urated absorption dips from the probe

January 30, 2020

SAS 26 Advanced Physics Laboratory

beam. Readjust the pump beam overlap mirror spacing should be measured and used
to maximize the depth of the dips. Note to determine the free spectral range, which
how the dips disappear when you block should be used with the overlayed fringes
the pump beam. Measure the saturated to determine (in MHz) the following spec-
absorption dips for each of the four hyper- tral features: (a) The separation between the
fine groups and explain your results. Use Doppler-broadened peaks from the same iso-
short sweeps that cover only one group at tope of the “no pump” spectrum. Compare
a time. Experiment with the sweep speed, with expectations based on the energy level di-
the number of points in the sweep and the agrams in Fig. 11. (b) The FWHM of the each
preamp gain and filter settings to see how Doppler-broadened peak. Compare with the
they affect your measurements. predictions based on Eq. 17. (c) The FWHM
of the narrowest saturated absorption dip.
21. Measure the FWHM of one or more dips
as a function of the pump and probe
beam intensities and explain your results.
How do you get the narrowest resonances?
Why can’t you get a FWHM as low as
the predicted natural resonance width
of 6 MHz? Hints: Is the laser truly
monochromatic?

22. If the displayed spectrum is just about

perfectly stable from sweep to sweep, you
might try turning on the program’s aver-
aging feature to improve the quality of the
spectra. However, if the spectrum moves
around on the display — even a little —
averaging will blur the spectrum and de-
crease its quality.

23. Measure the relative splittings within

each group and use these results to iden-
tify the direct and crossover resonances.

CHECKPOINT: You should have obtained

the “no pump” spectrum (Doppler-broadened
without saturated absorption dips) and the
saturated absorption spectrum (with the
pump beam and showing the saturated ab-
sorption dips). Both spectra should cover at
least three of the four possible groups of hyper-
fine transitions (with the same lower 2 S1/2 hy-
perfine level) and they should include the over-
layed Fabry-Perot peaks. The Fabry-Perot

January 30, 2020