Feshbach Resonances in Ultracold Atoms

Isabelle Gunhee Lee

Adviser: Prof. Seth Aubin

Abstract
This thesis presents a simple theoretical model of a Feshbach resonance between two col-

liding ultracold atoms and experimental progress towards the observation of such a resonance

in 85 Rb. On the theoretical side, the inter-atomic scattering potential is approximated by a

spherical square well potential. The potential supports a number of bounds states whose energy

depends on the width and depth of the well. In the low energy limit, only s-wave scattering

is present, and the associated scattering length for the process diverges if there is a bound

state at the dissociation limit: this is the hallmark of a Feshbach resonance. The model also

shows that the Feshbach resonance can be shifted to higher scattering energies by adjusting

the well depth, indicating the presence of a quasi-bound state above the dissociation limit. On

the the experimental side, a crossed dipole trap has been developed and used to trap ultra-

cold rubidium atoms. The performance of the trap is characterized, and the thesis describes

work towards upgrading it for an experiment to observe a Feshbach resonance in the F = 2,

mF = −2 hyperfine ground state of 85 Rb.

1. Introduction

The objective of this project is to develop a theoretical understanding of Feshbach

resonances and observe them experimentally in a gas of ultracold atoms. Feshbach
resonances in ultracold atomic gases are scattering resonances in which the colli-
sion cross-section for two colliding atoms diverges for specific adjustments to the
interatomic potential, generally with a magnetic field. Feshbach resonances allow

1
 enormous control over the attractive and repulsive interactions of atoms, specifically
by controlling scattering cross sections. By tuning the magnetic field, the atoms can
associate into strongly interacting molecular states or dissociate as free unbound
atoms. Feshbach resonances are essential for reducing interactions in a BEC in-
terferometer and increasing those of strongly interacting superfluids [1]. Moreover,
Feshbach resonances may yield advancements in more practical experiments, such
as suppressing the frequency shifts of atomic clocks [2].

Figure 1: Cross section (a), scattering length (b), and molecular binding energy (c) of a generic
Feshbach resonance. Arrow in Figure 1 (c) indicates the direction of magnetic field sweeps
when creating cold molecules (adapted from reference [3]).

By definition, Feshbach resonances rely on scattering theory. A Feshbach reso-

nance is a strong enhancement in the collision cross section and scattering length
at a particular magnetic field value. As shown in Figure 1, colliding ultracold atoms
experience a dramatic enhancement of their collision properties at a Feshbach reso-
nance located at magnetic field B0 : the scattering length and cross-section diverge.

2
In this case, the scattering length, as , shown in red, is given by:

∆
as (B) = abg (1 − ) (1)
B − B0

where as is the s-wave scattering length, abg is the background scattering length,
∆ is the width of resonance, and B0 is the magnetic field value of resonance. At
B = B0 the scattering length diverges, leading to very large elastic and inelastic
collision cross sections. In particular, the inelastic collisions between atoms increase
at a Feshbach resonance, which leads to a trap loss, an experimental signature of
a Feshbach resonance. The elastic cross section for s-wave two-body collisions is
calculated using the equation (2):

σ = 4πas 2 α (2)

where σ is the cross section, as is the scattering length, α = 1 for non-identical

atoms, α = 2 for identical bosons, and α = 0 for identical fermions. Despite the
increase in inelastic collisions at the Feshbach resonance, careful ramping of the
magnetic field through the resonance (from the as < 0 side to the as > 0 side of
the resonance) can produce a weak molecular bond between the colliding atoms, as
shown in Figure 1 (c) [3].

Feshbach resonances provide an efficient method to producing ultracold molecules.

Although Feshbach molecules themselves are extremely “fragile,” Feshbach reso-
nance association is a potentially coherent method that creates molecules in a sin-
gle molecular state. Therefore, Feshbach molecules can be used as an intermediary
product to producing stable molecules in a single state. Moreover, understand-
ing Feshbach resonances may lead to advancements in studying dipolar quantum
gas, ultracold chemistry, molecular interferometry, and precision measurements of
fundamental symmetries [1]. The rest of the thesis is structured in the following

3
 manner: Section 2 introduces basic concepts of scattering theory, section 3 describes
the basic physics of Feshbach resonances using a toy model based on a spherical
square well inter-atomic potential, and section 4 discusses the experimental work
85
to set-up an experiment to observe a Feshbach resonance in ultracold Rb. In
conclude the thesis in section 5 by discussing prospects for future theoretical and
experimental work.

2. Development of Feshbach Resonance Theory

2.1 Scattering Theory

Figure 2: Basic setup of a system of two scattering atoms. The polar angle θ and azimuthal
angle φ are defined according to the cartesian coordinate system in the diagram.

Scattering theory is essential to understanding and modeling Feshbach resonances,

since Feshbach resonances are defined in terms of collisional cross sections and
scattering lengths. A basic setup of scattering systems is shown in Figure 2. The
setup consists of two particles colliding. If we move to the center of mass frame, a
system of two colliding atoms is equivalent to a reduced incoming wave, represented
as a red plane wave in Figure 2, interacting with a scattering potential and an
outgoing wave, represented as a blue spherical wave. In this setup, a couple of
measurements can be taken as below:

4
Fi = the flux of particles in the incident waves over time

dΩ = solid angle

dn =number of scattered particles per unit time

Then, dn should obviously be proportional to Fi and dΩ:

dn = Fi σ(θ, φ)dΩ (3)

where σ = coefficient of proportionality. Since the potential is assumed to be

dσ
spherically symmetric, the φ component may be omitted. Then, dΩ
is defined as
the differential scattering cross section. More variables of interest can be defined as
we proceed, such as the scattering amplitude and the scattering length. However, a
simple relation between scattering amplitude and the scattering cross section may
be stated here. Scattering amplitude, denoted f (θ), in terms of the differential
cross section is:
dσ
= |f (θ)|2 (4)
dΩ

With these definitions at our disposal, the basic structure of scattering systems can
be visualized. In order to understand the overarching structure, the wave equations
at asymptotic limits are investigated. Assuming that the potential has essentially
no effect in the asymptotic limit, the Schrodinger equation becomes:
 2 
~
− ∇2 ψ = Eψ (outside the influence of potential) (5)
2µ

where µ is the reduced mass of the potential and incoming particles, and ψ is the
wave function of inter-particle separation. Then, two forms of wave equations satisfy
this Schrodinger equation: an incoming plane wave and an outgoing spherical wave.
Therefore, the wave equation of the system at asymptotic limits should have the

5
form:
eikr
ψ(r) ∼ eikz + f (θ) (6)
r

where f (θ) is defined as the scattering amplitude. We use the partial wave analysis
and the phase shift of the scattered wave to determine the scattering amplitude
f (θ).

2.1.1 Partial Wave Analysis

The partial wave analysis constructs solutions of the scattering systems using de-
fined partial waves as a basis set. The partial waves are the solutions that satisfy
the radial Schrodinger equation near the potential. In this case, the partial waves
are spherical hankel functions of the first kind:

(1)
hl (x) ≡ jl (x) + inl (x) (7)

Hankel functions are a valid basis set, since in the x → +∞ asymptotic limit, they
behave like spherical waves:

eikr
lim h → (−i)l+1 (8)
x→∞ kr

Since the potential is spherically symmetrical, the resulting solution is:

( ∞
)
X (1)
ψ(r, θ) = A eikz + k i(l+1) (2l + 1)al hl (kr)Pl (cos θ) (9)
l=0

where A is a general constant, and al is the amplitude of the spherical partial wave
with angular momentum l. Then, from (8) and (9), the scattering amplitude can
be found:

∞
X
f (θ) = (2l + 1)al Pl (cos θ) (10)
l=0

6
2.1.2 Phase Shift

As shown in the previous section, the amplitudes of the incoming waves and out-
going waves are different. However, in order to maintain conservation of particle
number in each angular momentum channel l (elastic scattering), the norms of in-
coming and outgoing waves must be the same, so this difference in amplitudes can
be represented as a difference in phase instead: the outgoing wave is out of phase
with the incoming wave by a factor of 2δl . Then, the scattered wave can be written
solely in terms of phase 2δl . First, the incoming plane waves are written in terms
of spherical waves, using Rayleigh’s formula:

∞
X
ikz
ψincident = e = il (2l + 1)jl (kr)Pl (cos θ) (11)
l=0

Multiplying the outgoing spherical component of (11) by phase component (e2iδ ),

the equation becomes:

X 2l + 1  i(kr+2δ)
− (−1)l e−ikr Pl (cos θ)

ψincident = A e (12)
l
2ikr

Using (10), the outgoing scattered wave can be represented as:

∞
X eikr
ψscattered = (2l + 1)al Pl (cos θ) (13)
l=0
r

As mentioned before, since phase shift is another way of representing the differ-
ence in amplitudes, the scattering amplitude can be written in terms of phases by
matching equation (12) and (13):

∞ ∞
X 2l + 1 2iδl
X
A Pl (cos θ)(e − 1) = A (2l + 1)al Pl (cos θ) (14)
l=0
2ik l=0

and simplifying (14),

7
 ∞ ∞
X X 1
al = eiδl sin δl (15)
l=0 l=0
k

Plugging (15) into (10), the scattering amplitude becomes:

∞
1X
f (θ) = (2l + 1)eiδl sin δl Pl (cos θ) (16)
k l=0

2.1.3 Scattering Length

The phase shift approach is particularly convenient, because it is directly connected

to the scattering length, usually denoted as . The scattering length is defined as the
s-wave limit of the equation (15). In the s-wave limit, the angular term l is 0, and k
approaches 0. The s-wave limit is highly relevant to this project, since the collision
of ultracold atoms only involves l = 0 terms. Then, equation (15) becomes:
 
1
as = − lim eiδ0 sin δ0 (17)
k→0 k

Moreover, another useful variable can be defined in terms of the scattering length,
dσ
which is the scattering cross section. Since dΩ
= |f (θ)|2 , the total scattering cross
section σ becomes:

Z Z Z
dσ 2 1 2 4π
σ= dΩ = |f (θ)| dΩ = sin δ 0 dΩ = sin2 δ0 (18)
dΩ k2 k2

Taking the limit as k approaches 0,

σ = 4πa2s (19)

Example: Hard Sphere Scattering

An example relevant to Feshbach resonances is a hard sphere scattering and its

scattering length. Hard sphere scattering is defined by the potential:

8
 
0,

if r > R.
V = (20)
∞, if r < R.


The radial schrodinger equation of this potential is clearly u(r) = A sin(kr + δ0 ).

The boundary condition of this potential becomes u(R) = 0, for the radial wave
equation u(r). In order for the wave equation to satisfy the boundary condition,
δ0 = −kR. Then, in the case of hard sphere scattering, the scattering length
becomes:  
1 −ikR 1
as = − lim e sin(−kR) ' (1 + ikR)kR ' R (21)
k→0 k k

Then, the scattering cross section follows from (19):

σ = 4πR2 (22)

3. The Feshbach Resonance Toy Model

One of the major theoretical objectives is to make a toy model of Feshbach reso-
nances that explains the basic physics. Feshbach resonances occur when the atoms
encounter a bound state energy level that is almost unbounded. In order to inves-
tigate further, we need to understand bound states better. However, the molecular
potential energy is complicated to model, so we replace it with a simpler poten-
tial that still retains the basic physics we are interested in: we approximate the
internuclear molecular potential with a square well potential for studying collisions
between two identical ground state atoms. In particular, we focus on the scattering
behavior of the atoms when the well barely spports a bound state, which is the
reported condition for a Feshbach resonance. In this section, we first study the
bound states on the one-dimensional square well potential, which we then apply
to the determination of the s-wave bound states of the three-dimensional spherical
square well potential. Second, we investigate s-wave scattering from this spherical

9
well potential and determine the dependence of the scattering phase shift δ0 and
scattering length as on the well parameters. Finally, we investigate the Feshbach
resonances of this spherical well potential.

3.1 1-D Square Well Potential

In order to investigate bound states, bound states of a square well are studied. The
energy levels of the system are calculated, as the well depth varies from shallow to
deep. Figure 3 plots the potential of the system as a function of position.

Figure 3: Potential of an Attractive Well

Dividing the system as shown in Figure 3, the Schrodinger equations of respective

regions are as follows:

~2 ∂ 2
− ψ = Eψ (Region 1 and 3) (23)
2m ∂x2

~2 ∂ 2
 
− − V0 ψ = Eψ (Region 2) (24)
2m ∂x2

Moreover, we define r
2m(E + V0 )
k= (25)
~2

10
 r
2mV0
k0 = (26)
~2

From equations above, the boundary conditions are derived:

   
ka k ka
sin = for tan <0 (27)
2 k0 2
   
ka k ka
cos = for tan >0 (28)
2 k0 2

Equations (27) and (28) cannot be solved analytically, so these are solved graphi-
cally as shown in Figure 4. y1 and y2 in the figures are defined as:


ka ka



 sin

, if tan < 0.
2 2
y1 = (29)
ka ka

 cos


2
, if tan 2
> 0.

k
y2 = (30)
k0

As V0 increases, or as the well deepens, the number of solutions, represented by the

intersection points, increases. This signifies that the number of available energy
levels, or bound states of the system, increases as well.

In order to investigate the behavior of the bound state energies further, the energies
are plotted as the potential is scanned from V0 = 1 to V0 = 120. First, the solutions
are found computationally, using intersections. As shown in Figure 4 the solutions
are given as k values, and they are converted to energies using the equation below:

~2 2
E= (k − k02 ) (31)
2m

11
Figure 4: Plots of the boundary conditions (29) and (30) as the well depth (V0 ) increases. The
intersections are the bound state solutions in terms of k. The plots are for (a) V0 = 2, (b)
V0 = 20, (c) V0 = 100, and (d) V0 = 200. For this plot, ~ is set to 1, m is set to 1, and a is
set to 2. y1 is in blue, and y2 is in green. The circles are the intersections of the two boundary
conditions. So, they represent at which k the boundary conditions are satisfied.

12
Figure 5: Bound state energies versus the depth of the well. For this plot, ~ is set to 1, m is
set to 1, and a is set to 2.

Figure 6: Close-up of the boxed portion of Figure 5

13
Then, these energies are plotted against the varying potential, and Figure 5 is the
resulting plot. The energy levels exhibit discrete behavior: new energy levels appear
at distinctive well depths, as shown in Figures 5 and 6.

The square well potential is highly relevant to our toy model for Feshbach reso-
nances, since varying well depth is analogous to the varying potential created by
sweeping magnetic fields in Feshbach resonances. Thus, understanding and model-
ing attractive well energy levels is a first look at modeling bound states of Feshbach
resonances. The plots are generated by Matlab codes in Appendix A.1.

3.2 Spherical Well Potential

In order to study the phase shifts, a spherical well potential is studied. The spherical
well potential is given by the equation below:


−V0 , if r < r0 .

V (r) = (32)

0, if r > r0 .

Then, the Schrodinger equations of the system becomes:

~2 2
 
− ∇ − V0 ψ = Eψ if r < r0 (33)
2m

~2 2
− ∇ ψ = Eψ if r > r0 (34)
2m

As particles encounter and interact with this potential and scatter they gain a phase.
Then, the system of wave functions that satisfy the Schrodinger equations (33) and
(34) will be the linear combination of the initial wave and the phase shifted scattered
wave. Denoting phase as δ, the radial compotnents of the solutions become, for
angular momentum l = 0:

14
 
A sin(k 0 r)

if r < r0 .
u(r) = (35)

B sin(kr + δ) if r > r .
0

where √
2mE
k= (36)
~

and p
2m(E + V0 )
k0 = . (37)
~

Combining equations (36) and (37),

k 02 k2
− V0 = (38)
2m 2m

3.2.1 Spherical Well Bound states

From the continuity of the wave functions and their derivatives at r = r0 , the
boundary condition becomes:

k
cot(k 0 r0 ) = − (39)
k0

Then, we may define two functions, y1 and y2 :

y1 = cot(k 0 r0 ) (40)

k
y2 = − (41)
k0

Using the same methods as with square well potential, the intersections of y1 and
y2 are found, as V0 is scanned. The resulting plots are shown in Figure 7:

15
Figure 7: Intersections of y1 and y2 at varying values of V0 . y1 is in blue, and y2 is in red.
V0 = 2 for (a), V0 = 20 for (b), V0 = 200 for (c), and V0 = 300 for (d). The circles represent the
intersections of y1 and y2 . So, they represent at what k the boundary conditions are satisfied.
For this plot, ~, m, and r0 are set to 1.

Then, converting the resulting k values of the intersections to energy, the E vs. V0
plot is created as below.

16
Figure 8: E vs. V0 . Bound state Energies as a function of the Well Depth. For this plot, ~, m,
and r0 are set to 1.

Figure 9: Close-up of the boxed portion of Figure 8. For this plot, ~, m, and r0 are set to 1.

Similar to the square well potential, the bound state energies occur at discrete

17
values of the well depth V0 , as shown in Figure 8. The plots are generated by
Matlab codes in Appendix A.2.

3.2.2 Spherical Well Potential Scattering

In order to determine the scattering amplitude for k → 0 and l = 0, another bound-

ary condition for the system can be derived, using phase shifts. From equation (38)
and the continuity of the wave functions and their derivatives at r = r0 requires
that

k 0 cot(k 0 r0 ) = k cot(kr0 + δ) (42)

Solving above for δ yields an implicit expression:

k sin k 0 r0 cos kr0 − k 0 sin kr0 cos k 0 r0

tan δ = (43)
k 0 cos k 0 r0 cos kr0 + k sin kr0 sin k 0 r0

Since the above equation cannot be solved explicitly, it is solved using numerical
methods. The phase δ is plotted against k, as shown in Figure 10.

18
Figure 10: Phase δ vs. k at varying values of V0 . Since V0 = q 2 /2m, V0 is varied by scanning
q. Each line represents a solution for a different value of q, starting at q = 0 and increasing in
steps of ∆q = 0.1 This result matches the phase plot of [8].

Each line of the plot represents phase as a function of k, δ(k) at varying V0 . For
simplicity, we define V0 = q 2 /2m where q is a parameter with units of momentum.
Then, q is numerically scanned to plot each line of Figure 10. Figure 10 suggests
that the phase displays a discrete behavior, similar to energy. The plot is generated
by Matlab codes in Appendix A.3.

3.3 Scattering Length Plot

The definition of scattering length is given by (17). Equation (17) can be rearranged
to yield an equivalent definition of the scattering length:

19
 tan δ
as = − lim (44)
k→0 k

The numerical solution of (43) illustrated in Figure 10 provides function of phase in

terms of k, δ(k). Then, δ(k) of Figure 10 can be used to numerically solve for the
scattering length. By setting k = 0.001 at varying values of V0 in Figure 10, solving
for tan δ/k, and plotting the resulting points, we attain the points in Figure 11.
Then, the points are fitted with the function:

1
as = − (45)
V0 − 1.2324

This function is very similar to equation (1) that characterizes the behavior of atoms
at Feshbach resonances. Similar to (1), (45) has a basic structure of (constant)/(x
- constant). This result indicates that a Feshbach resonance can be modeled using
the spherical well potential, thus validating our choice of toy model.

20
Figure 11: Scattering length vs. well depth (as vs. V0 ). The momentum k is set to 0.001. For
this plot, ~, m, and r0 are set to 1.

This plot is generated by Matlab codes in Appendix A.4.

3.3.1 Analytical Solution of the Scattering Length

While it is numerically confirmed that Feshbach resonances can be modeled using

the spherical well potential, the equation for scattering length of (1) can be also
analytically derived from the implicit expression for δ in terms of k in (43). Dividing
equation (43) by k, we obtain

tan δ k sin k 0 r0 cos kr0 − k 0 sin kr0 cos k 0 r0

= 0 (46)
k k k cos k 0 r0 cos kr0 + k 2 sin kr0 sin k 0 r0

Then, using (44) and (46), the scattering length becomes:

tan δ k sin k 0 r0 cos kr0 − k 0 sin kr0 cosk 0 r0

as = − lim = − lim 0 (47)
k→0 k k→0 k k cos k 0 r0 cos kr0 + k 2 sin kr0 sin k 0 r0

21
 sin k 0 r0 sin kr0 r0 k 0 cos k 0 r0
=− + lim · · lim (48)
k 0 cos k 0 r0 k→0 k r0 k→0 k 0 cos k 0 r0 cos kr0 + k sin kr0 sin k 0 r0

tan k 0 r0
=− + r0 (49)
k0

Using the definition of k 0 , the limit of k 0 as a function of V0 becomes as follows:

p p
lim k 0 = lim k 2 + 2mV0 = 2mV0 (50)
k→0 k→0

Then, substituting (50) into (49),

√
r0 tan r0 2mV0
as = − √ + r0 (51)
r0 2mV0

√
Defining φ = r0 2mV0 ,
r0 tan φ
as = − + r0 (52)
φ

Feshbach resonances are defined by dramatic increases or divergences of scattering

lengths at particular magnetic field values. Since tan φ has divergences at (n+1)π/2,
(52) is series expanded at φ = π/2 as below:
 
2 4 n
as = r 0 − + + O(x − π/2) + r0 (53)
π(φ − π/2) π 2

Ignoring terms other than the Laurent term of the tangent expansion, the scattering
length becomes:    
2 4
as = r0 − + + 1 r0 (54)
π(φ − π/2) π2

We may ignore other terms, because at divergences, the Laurent terms will domi-
√
nate. Substituting φ = r0 2mV0 , the scattering amplitude is finally expressed in

22
terms of the well depth:

   
2 4
as = r0 − √ + + 1 r0 (55)
π(r0 2mV0 − π/2) π2

Equation (55) has the basic components of equation (1). Since the series expansion
only holds near the divergence at π/2, (55) is a local descriptor of the scattering
length, as is (1). This analytically confirms that the Feshbach resonances can be
modeled from scattering of spherical well potential. This can be explained, because
the applied magnetic field used to trapped atoms creates potential that resembles
the spherical well potential.

Moreover, (55) can be generalized for all divergences, since tangent is divergent at
(2n + 1)π/2.
   
2 4
as = r 0 − √ + + 1 r0 (56)
(2n + 1)π(r0 2mV0 − (2n + 1)π/2) π2

3.3.2 Feshbach Resonance Dependence on Well Depth

(2n+1)π
Since the divergences occur at φ = 2
, this implies:

p (2n + 1)π
φ = r0 2mV0 = (57)
2

Thus, the divergences occur for well depths at:

(2n + 1)2 π 2
V0 = (58)
8mr02

This equation (58) is verified numerically. From the numerical model of Figure 11,
the well depths at which Feshbach resonances occur are identified and plotted
against n, as shown in Figure 12. Then, (58) is used as a fitting function for
the points. The resulting plot below verifies that (58) holds.

23
Figure 12: Well depths at which Feshbach resonances occur for discrete n. The analytical
solution (58) is in green. For this plot, ~, m, and r0 are set to 1.

3.4 Feshbach Resonance at Finite Momentum

From 2.1.2, the scattering length is defined in terms of the limit below:

 
1
as = − lim eiδ0 sin δ0 (59)
k→0 k

In section 3.3, the numerical plot of the scattering length and the well depth,
Figure 11, is obtained by scanning V0 and plotting the equation below at k = 0.001:

1
as (k) = − eiδ0 sin δ0 (60)
k

In this section, the scattering lengths at finite momenta are investigated. The cases
at k = 0.499, k = 0.99 are investigated in Figure 13 (b) and (c).

24
Figure 13: Shifts in scattering lengths for finite momentum. For (a), k = 0.001, for (b),
k = 0.4990, and for (c), 0.999. For this plot, ~, m, and r0 are set to 1.

The fitting function for Figure 13 (a) is:

1
as = − (61)
V0 − 1.2324

The fitting function for Figure 13 (b) is:

1
as = − (62)
V0 − 1.3660

The fitting function for Figure 13 (c) is:

1
as = − (63)
V0 − 1.8935

Figure 13 illustrates that Feshbach resonances shift to the right as momenta increase

25
 from near 0. The plot is generated by Matlab codes in Appendix A.4.

4. Experiment

The long term experimental objective of this project is the observation of a magnetic
Feshbach resonance. The basic experimental approach is to load µK-level ultracold
atoms into an optical dipole trap, and the subject them to a magnetic field. The
Feshbach resonance is identified by observing the trap lifetime, or loss rate, at
different magnetic field strengths. An unusually large loss rate, or short lifetime
indicates a magnetic Feshbach resonance.

4.1 Candidates
Table 1. Candidates and Feshbach Resonances

Type B (G) ∆ (G) States Sources

85
Rb (boson-boson) 155 11.65 |2, −2 > +|2, −2 > [4]
39
K-87 Rb (boson-boson) 117.6 -1.3 |1, −1 > +|1, −1 > [5]
39
K-87 Rb (boson-boson) 247.9 0.28 |1, 1 > +|1, 1 > [5]
39
K-87 Rb (boson-boson) 318 7.6 |1, 1 > +|1, 1 > [5]
41
K-87 Rb (boson-boson) 49.5 -6.5 |2, 2 > +|2, 1 > [6]
41
K-87 Rb (boson-boson) 53.5 -39 |1, 1 > +|1, 1 > [6]
41
K-87 Rb (boson-boson) 87.5 -1.3 |1, 1 > +|1, 1 > [6]
40
K-87 Rb (fermion-boson) 545.4 1.2 [7]
87 85 39
Originally, Rb, Rb, and K atoms were considered for the experiment. More
specific groupings of these atoms are identified from the literature, as shown in
Table 1. B is the magnetic field at which Feshbach resonance is observed for the
corresponding atom pair, and ∆ is the width of resonance. In order to ensure the
observation of Feshbach resonances, pairs with wider width of resonances are se-
lected. Pairs with relatively low magnetic fields, around 150G to 200G, are also
preferred. Based on the available isotope and magnets, the experiment will be done
85
using Rb.

26
 4.2 Basic Experimental Setup

The basic experimental setup is described in Figure 14. The MOT, or the magneto-
optical trap, laser cools the atoms to 10 to 100 µK range. Then, the dipole trap
is constructed, because the dipole trap is not sensitive to varying magnetic fields.
The dipole trap is an intense, far off-resonance, conservative laser trap that can
operate at arbitrary magnetic field. The trap potential energy is proportional to
the laser intensity. The experimental setup is also shown in Figure 14. By sweeping
the magnetic field near a Feshbach resonance, which is represented as Bfeshbach in
Figure 14, there would be a loss of atoms. The atom losses most likely are due to
an increase in inelastic collisions between atoms at the resonances, an experimental
signature of Feshbach resonances.

4.3 Initial Dipole Trapping Result

87
The dipole trap is first constructed using Rb as shown in Figure 14.

MOT
Bfeshbach

Laser

Figure 14: Constructed MOT and Dipole Trap using 87 Rb, and the Experimental Setup

85
4.4 Trapping of Rb
87
Figure 14 depicts the trapping of Rb atoms. As decided in 4.1, the experiment

27
 85 87
is conducted using Rb atoms. However, since the lab is most familiar with Rb,
87 85
Rb atoms are used first to create a dipole trap to test the apparatus. Then, Rb
atoms are trapped.

Figure 15: (a) 85 Rb MOT, (b) 85 Rb Dipole Trap. The red ellipse indicates the location of the
very faint dipole trap.

85
First, the MOT of Rb is constructed as shown in Figure 15 (a). Then, using
85
lasers, the dipole trap is created with Rb as shown in Figure 15 (b). However,
this original dipole trap is not dense enough to observe Feshbach resonances. Thus,
the density of the dipole trap is increased using various methods, as shown in
Figure 16.

28
Figure 16: Evolution of the 85 Rb Dipole Trap. (a) the original dipole trap. (b) dipole trap with
compressed MOT and better alignment. (c) retro-reflected cross dipole trap. (d) retro-reflected
cross dipole trap with better alignment.

First, the MOT is compressed by increasing the applied magnetic field, and the
laser is aligned better. The resulting is shown in Figure 16 (b). Then, the laser
is retro-reflected to create a cross dipole trap. The experimental setup is shown
in Figure 18 (b). The result is shown in Figure 16 (c). In this figure, the dipole
trap became more compressed at the center. Then, the laser is better aligned so
that the retro-reflected beam had maximal impact on the cross dipole trap. The

29
 result is shown in Figure 16 (d). Moreover, the temperature measurement of the
dipole trap is taken by measuring the width of the thermal cloud. The temperature
measurement yields that the dipole trap is at around T = 8 µK.

500

450

400
Width of Cloud (microns)

350

300

slope = 31.521
250
slope = 26
200

150

100

50

0
0 2 4 6 8 10 12 14 16

Time of Flight (ms)

Figure 17: Temperature Measurement of the Dipole Trap in Figure 16 (d). The slope of the
green fit yields the average velocity of the atoms in the thermal cloud. Then, this average
velocity is used to calculate temperature, since 1/2mv 2 = 1/2kT . The slope of the fitting
function in green gives the average velocity of expansion, v = 31.5 µm/ms. The red fit is used
to calculate the error bar. The calculated error is 8.76 %.

Through these processes, we were able to create a dense dipole trap, but the obser-
vation of Feshbach resonances would require all of the atoms to participate. At this
stage, only 20 percent of the atoms would contribute. In order to make all atoms
contribute, the atoms all have to be in the same desired state. The ideal atomic
state for 85 Rb atoms for Feshbach resonance observation is F = 2 and mF = −2. In

30
 order to obtain the atoms in a single quantum state, optical pumping is required.
Furthermore, the platform for a second dipole trap laser is built and shown in
Figure 18 (b).

(a) (b)

Figure 18: Experimental trapping laser setup. (a) Retro-reflected dipole trap scheme. (b)
Planned path of the 2nd dipole trap laser for further compression of atoms. The aluminum
platform features a hole for easy passage of the dipole trap laser beam

5. Conclusion

The objective of this project is to study and observe Feshbach resonances. There are
two theoretical cornerstones of Feshbach resonances: the scattering of cold atoms

31
and the bound states. Thus, in order to develop a theoretical understanding of
Feshbach resonances, scattering theory and bound states are studied. The phase
shifts of a spherical well is both numerically and analytically modelled. The bound
states of the well potential are studied. On the experimental side, atom pairs with
favorable resonance widths and Feshbach magnetic field values were identified. The
85 87
MOT and the dipole trap using Rb and Rb are constructed.

This study will facilitate the future observation of Feshbach resonance. Once the
Feshbach resonances are observed, the association of atoms into molecular states
can be investigated. On the theory side, the theory of Feshbach molecules may be
investigated. The control of the resonance width in Feshbach resonances can also
be a theoretical endeavour.

32
6. References

[1] Quemener, Goulven, and Paul S. Julienne. “Ultracold Molecules under Con-
trol!.” Chemical Reviews 112.9 (2012): 4949-5011.

[2] Carr, Lincoln D., et al. “Cold and ultracold molecules: science, technology and
applications.” New Journal of Physics 11.5 (2009): 055049.

[3] C. Chin, R. Grimm, P. Julienne, E. Tiesinga, Feshbach resonances in ultracold

gases. Rev. Mod. Phys. 82, 1225 (2010).

[4] Roberts, J. L., N. R. Claussen, S. L. Cornish, and C. E. Wie- man, 2000, Mag-
netic field dependence of ultracold inelas- tic collisions near a Feshbach resonance,
Phys. Rev. Lett. 85, 728731.

[5] Simoni, A., M. Zaccanti, C. DErrico, M. Fattori, G. Roati, M. Inguscio, and G.

Modugno, 2008, Near-threshold model for ultracold KRb dimers from interisotope
Feshbach spectroscopy, Phys. Rev. A 77, 052705.

[6] Simoni, A., F. Ferlaino, G. Roati, G. Modugno, and M. Ingus- cio, 2003, Mag-
netic control of the interaction in ultracold K-Rb mixtures, Phys. Rev. Lett. 90,
163202.

[7] Ferlaino, F., C. DErrico, G. Roati, M. Zaccanti, M. Inguscio, G. Modugno, and

A. Simoni, 2006, Feshbach spectroscopy of a K-Rb atomic mixture, Phys. Rev. A
73, 040702(R) , ibid 74, 039903(E).

33
[8] Mueller, Erich. ”Scattering Tutorial.” Mueller Research Group. Erich Mueller,
12 Aug. 2002. Web. 10 Apr. 2013.

34
A. Appendix- Source Codes

7.1 Matlab Codes for 1-D Attractive Well

function [E,rtx,rty,y1,y2,k] = attr_well(V_0)

%from the boundary condition, this function solves for energies, and
%intersections at varying well depths

m=1;%mass=1
a=2;%radius of well=2
start=0;
finish=10*pi/a;
step=.001;
k=start:step:finish;
%momentum scanned
k_0=sqrt(2*m*V_0);
%hbar set to 1

A=abs(cos(k.*a./2));
B=abs(sin(k.*a./2));
C=sign(tan(k.*a./2));

y1=A.*((1+C)./2)+B.*((1-C)./2);
%boundary condition

y2=k/k_0;
%another boundary condition

[rtx,rty]=rootfinder(k,y1,y2,step);
%y1 and y2 intersections, or where the boundary conditions are satisfied

%scan V_0 from 0 to 200

E=solve_for_E(m,k_0,rtx);
E=rid_the_doubles(E);

end

35
function [rootx, rooty] = rootfinder(k,f1,f2,step)
%make sure f1, f2, indept_var all have the same size!!!
%this function finds "roots," or more accurately, intersections of two given
%functions

[t,s]=size(f1);
if s<t
s=t;
end
%so it doesnt matter if i pass an array or a vector

rootx=zeros(1,20);
rooty=zeros(1,20);
index=1;

for i=1:s
if abs(f1(i)-f2(i))<(step)
rootx(index)=k(i);
rooty(index)=f2(i);
index=index+1;
end
end

rootx=rootx(1:index-1);
rooty=rooty(1:index-1);

end

36
function [E] = solve_for_E(m,k_0,rtx)
% this function solves the resulting momentum intersections to energies.
[s,t]=size(rtx);
if t<s
t=s;
end
%so it doesnt matter if i pass an array or a vector

E=zeros(1,t);

for i=1:t
E(i)=-((rtx(i))^2-(k_0)^2)/2*m;
end

end

37
function [E] = rid_the_doubles(E)
% this function gets rid of double counted intersections.
%if the spacing between two intersections is too small,
% then it only counts one intersection.

[x,y]=size(E);
if y<x
y=x;
end
%so it doesnt matter if i pass an array or a vector

for i=1:(y-1)
if abs(E(i)-E(i+1))<.6
E(i+1)=E(i);
end
end

E=unique(E);

end

38
The function below generates the Figure 4.

function [] = test_attr_well()
%this function plots the boundary conditions and intersections
%at different values of V_0

[E,rtx,rty,y1,y2,k]=attr_well(2);
plot_intersections(k,y1,y2,rtx,rty,1)
disp(’(on figure 1) E when V_0=2 is: ’)
disp(E)

[E,rtx,rty,y1,y2,k]=attr_well(20);
plot_intersections(k,y1,y2,rtx,rty,2)
disp(’(on figure 2) E when V_0=20 is: ’)
disp(E)

[E,rtx,rty,y1,y2,k]=attr_well(100);
plot_intersections(k,y1,y2,rtx,rty,3)
disp(’(on figure 3) E when V_0=100 is: ’)
disp(E)

[E,rtx,rty,y1,y2,k]=attr_well(200);
plot_intersections(k,y1,y2,rtx,rty,4)
disp(’(on figure 4) E when V_0=200 is: ’)
disp(E)

[E,rtx,rty,y1,y2,k]=attr_well(300);
plot_intersections(k,y1,y2,rtx,rty,5)
disp(’(on figure 5) E when V_0=300 is: ’)
disp(E)

end

39
function [] = plot_intersections(k,y1,y2,rtx,rty,number)
%this function plots two bdy conditions, limit, roots.

figure(number)
plot(k,y1)
hold on
plot(k,y2,’g’)
hold off
hold on
plot(k,1,’r’)
hold off
hold on
scatter(rtx,rty,’o’,’m’)
hold off

end

40
The function below generates the Figures 5 and 6.

function [] = E_vs_V(number)
%this function plots the bound state energies vs the well depth, V_0.
step=.01;
V_0=2:step:70;
[~,t]=size(V_0);
V=zeros(1,10*t); %scans V_0
E=zeros(1,10*t);
%as we solve for the energy levels, we will be filling this vector.
n=1;
m=1;

for i=1:t
[E_0,~,~,~,~,~]=attr_well(V_0(i));
[~,y]= size(E_0);
a=1;

for l=m:(m+y)
V(l)=V_0(i);
end
m=m+y;
for j=n:(n+y)
if a<=y
E(j)=E_0(a);
a=a+1;
end
end
n=n+y;
end

[~,d]=size(E);
V=V(1:d);

figure(number)
scatter(V,E,5,’.’)
hold on
E1=0;
plot(V,E1)
hold off

end

41
7.2 Matlab Codes for 3-D Attractive Well Potential

function [E,rtx,rty,y1,y2,k] = attr_well(V_0)

%this function solves for intersections, or at what
%values of momentum the boundary conditions are satisfied.
m=1;
%mass=1
a=2;
%radius of well=1
start=0;
finish=10*pi/a;
step=.001;
k=start:step:finish;
k_p_squared=V_0-k.^2;
k_p=sqrt(k_p_squared);
k_p=real(k_p);

%the boundary conditions, y1 and y2

y1=cot(k_p.*a);
y2=-k./k_p;

%intersections
[rtx,rty]=rootfinder(k,y1,y2,step*30);

%scan V_0 from 0 to 200

E=solve_for_E(rtx);
E=rid_the_doubles(E);

end

This function uses the same functions solve for E, rid the doubles, and rootfinder
from Appendix A.1. Using the same function test attr well generates Figure 7.

42
The function below plots Figures 8 and 9.

function [] = E_vs_V(number)
%this function plots the bound state energies at different values of V_0.
step=.01;
V_0=0.6:step:60;
%scans V_0
[~,t]=size(V_0);
V=zeros(1,10*t);
E=zeros(1,10*t);
n=1;
m=1;

for i=1:t
[E_0,~,~,~,~,~]=attr_well(V_0(i));
[~,y]= size(E_0);
a=1;

for l=m:(m+y)
V(l)=V_0(i);
end
m=m+y;
for j=n:(n+y)
if a<=y
E(j)=E_0(a);
a=a+1;
end
end
n=n+y;
end

[~,d]=size(E);
V=V(1:d);

V1=0:step:60;

figure(number)
scatter(V,E,5,’.’)
hold on
E1=0;
plot(V1,E1)
hold off
axis([0 60 -1 30])

43
figure(number+1)
scatter(V,E,5,’.’)
axis([0 5 -1 2])

hold on
E1=0;
plot(V1,E1)
hold off

end

44
7.3 Matlab Codes for Phase Plot
The Matlab function “phase” generates Figure 10.

function [] = phase()
%This function plots the phases vs momentum at varying well depths.

m=1;%mass set to 1
a=1;%radius of the well set to 1

start=0;
finish=5*pi/a;
step=.001;
k=start:step:finish;

q=0:0.1*pi:5*pi;

n=(q-rem(q,pi))./pi;

[~,c]=size(n);

l=round(q./pi);

figure;

for i=1:c
done=0;
V_0=(q(i).^2)./(2*m);

[delta]=sph_attr_well2(V_0);
while done==0
[delta,done]=scale(delta,1);
end

if i>1
if l(i)-l(i-1)==1
delta=delta+(2*l(i)-1)*pi/2;
else
delta=delta+l(i)*pi;
end
end

hold on
plot(k,delta)

45
 hold off

end

end

46
function [y3] = sph_attr_well2(V_0)
%this function determines the boundary condition using phase shifts.
a=1;%radius=1
start=0;
finish=5*pi/a;
step=.001;

k=start:step:finish;

k_p=sqrt(k.^2+2*V_0);

S_p=sin(k_p.*a);
C=cos(k.*a);
S=sin(k.*a);
C_p=cos(k_p.*a);

A=k.*S_p.*C;
B=k_p.*S.*C_p;
D=k_p.*C_p.*C;
E=k.*S.*S_p;
%boundary conditions
y1=(A-B)./(D+E);

y3=atan(y1);

end

47
function [A,done] = scale(A,step)
%this function scales the phase by getting rid of discontinuities
%that arise from taking arctangent
[~,b]=size(A);
done=0;
c=0;

for i=1:b

if (i+1)>b
done=1;
c=-1;
return;
end
if abs(A(i)-A(i+1))>step
change=abs(A(i)-A(i+1));
c=i;
break;
end
end

if (c>0)&&(c+1<b)
for i=c+1:b
A(i)=A(i)-change;
end
end

end

48
7.4 Matlab Codes for Scattering Length Plot

function [V,A] = iterate_pi_slength(number,n)

%this function solves for the scattering length from phase
a=1;%radius=1
m=1;%mass=1

start=0;
finish=5*pi/a;
step=.001;
k=start:step:finish;%momentum scanned

start=number*pi-2;
finish=number*pi+2;
step=.0003*pi/a;
q=start:step:finish;%q is used to scan V_0

[~,c]=size(q);

V=zeros(1,c);
A=zeros(1,c);
j=1;

for i=1:c
done=0;
V_0=(q(i).^2)./(2*m);

[delta]=sph_attr_well2(V_0);
delta=delta+number*pi;

y=tan(delta)./k;

V(j)=V_0;
A(j)=y(n);
j=j+1;

end
end

49
function [j,x] = diverge_at(V,A)
%this function figures out where the input function diverges.
[~,b]=size(A);
%j=index
j=0;
for i=1:b
if (sign(A(i)) ~= sign(A(i+1)))&&(abs(A(i)-A(i+1))>100)
j=i;
break;
end
end

x=(V(j)+V(j+1))/2;

end

50
The function below generates the Figures 11 and 13.

function [] = A_vs_V()

[V,A]=iterate_pi_slength(1,2);
figure;
scatter(V,A,’.’,’b’);
[~,x_0]=diverge_at(V,A);
hold on
x=0:.001:200;
plot(x, -1./(x-x_0),’b’)
x_0
hold off

[V,A]=iterate_pi_slength(2,2);
figure;
scatter(V,A,’.’,’b’);
[~,x_0]=diverge_at(V,A);
hold on
x=0:.001:200;
plot(x, -1./(x-x_0),’b’)
x_0
hold off

[V,A]=iterate_pi_slength(3,2);
figure;
scatter(V,A,’.’,’b’);
[~,x_0]=diverge_at(V,A);
hold on
x=0:.001:200;
plot(x, -1./(x-x_0),’b’)
x_0
hold off

[V,A]=iterate_pi_slength(4,2);
figure;
scatter(V,A,’.’,’b’);
[~,x_0]=diverge_at(V,A);
hold on
x=0:.001:200;
plot(x, -1./(x-x_0),’b’)
x_0
hold off

51
[V,A]=iterate_pi_slength(5,2);
figure;
scatter(V,A,’.’,’b’);
[~,x_0]=diverge_at(V,A);
hold on
x=0:.001:200;
plot(x, -1./(x-x_0),’b’)
x_0
hold off

end

52