Diploma Thomas Uehlinger
Diploma Thomas Uehlinger
Diploma Thomas Uehlinger
A 2D Magneto-Optical Trap as a
High-Flux Source of Cold
Potassium Atoms
Thomas Uehlinger
A 2D Magneto-Optical Trap as a
High-Flux Source of Cold
Potassium Atoms
Submitted by
Thomas Uehlinger
Citizen of Zurich and Neunkirch SH,
Switzerland
iii
iv
Contents
1 Introduction 1
2 Theory 3
2.1 Potassium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Doppler-free saturated absorption spectroscopy . . . . . . . . . . . . . . . . . . . . 4
2.3 Light forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 Doppler cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5 Magneto-optical traps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5.1 Magneto-optical traps in a multi-level atom . . . . . . . . . . . . . . . . . . 9
2.5.2 Atom number dynamics in a MOT . . . . . . . . . . . . . . . . . . . . . . . 10
2.5.3 Loading of MOTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.6 2D magneto-optical traps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Experimental set-up 15
3.1 2D MOT vacuum chamber and optics . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.1 Magnetic coils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.2 Primary atom source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.3 Residual gas analyzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.4 Differential pumping tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.5 MOT optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 3D MOT vacuum chamber and optics . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Laser system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3.2 Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.3 Frequency stabilization of the lasers . . . . . . . . . . . . . . . . . . . . . . 25
3.3.4 Electro-optical modulators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.4 Measurement set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4.1 Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4.2 Atomic beam fluorescence in the 2D MOT chamber . . . . . . . . . . . . . 35
3.4.3 3D MOT size and loading parameters . . . . . . . . . . . . . . . . . . . . . 36
4 Measurements 37
4.1 Calibration of the laser frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Atomic beam fluorescence in the 2D MOT chamber . . . . . . . . . . . . . . . . . . 37
4.3 3D MOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3.2 The 3D MOT as a detection device for the atomic beam . . . . . . . . . . 40
4.4 2D MOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4.1 Laser detuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4.2 Laser power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4.3 Repumping light power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4.4 Magnetic field gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4.5 Beam alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4.6 Cooling volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.4.7 2D MOT position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.4.8 39 K Partial pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.4.9 Influence of UV light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.4.10 Best parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
v
Contents
List of Figures 61
List of Tables 63
Bibliography 65
vi
1 Introduction
In recent years, ultracold atoms have emerged as a promising system to study quantum many-
body physics. This development was sparked by the development of laser cooling methods [1],
which enabled cooling of a gas of cold atoms to micro-Kelvin temperatures. These advances in
cooling culminated in the observation of Bose-Einstein condensation in a dilute atomic vapor [2]
in 1995 and of quantum degenerate fermions in 1999 [3].
By loading these ultracold samples of „quantum matter“ into periodic light potentials - called
optical lattices - created by one or more standing light waves, crystal like structures can be
formed. Quantum degenerate gases in optical lattices represent an ideal starting point to test
models from condensed matter physics [4], while also finding applications in quantum optics
and quantum information processing and in understanding atomic and molecular physics [5]. A
unique feature that sets these „optical crystals“ apart from their solid counterparts is that the
ratio of the interaction energy to the kinetic energy can be easily controlled by changing the lattice
depth. Additionally the interactions between single particles can be tuned from the attractive to
the repulsive regime by accessing so called Feshbach resonances. Using these techniques, different
quantum many-body phases can be investigated. A prominent example showing a transition from
a weakly interacting quantum system to a strongly correlated quantum many-body system is the
transition from the superfluid to the Mott insulating regime for bosons [6].
Fermionic atoms in optical lattices mimic the behavior of electrons in a solid (see figure 1.1 for
an example). The exploration of these novel systems is driven by the ambition to get deeper
insight into long-standing problems of quantum many-body physics, such as high-temperature
superconductivity. Studying these systems, researchers could already investigate the transport
properties of fermions in periodic potentials for varying interaction strengths [8, 9]. The recent
observation of a Mott insulator of fermionic atoms [10] constitutes a further step towards the
simulation of many-body physics with ultracold atoms, which might ultimately throw light on
the nature of high-temperature superconductivity.
From an experimental point of view, one of the main challenges to study fermionic many-body
physics in optical lattices is achieving sufficiently low temperatures. A high initial atom number
in the magneto-optical trap (MOT) is crucial for the efficiency of the evaporative cooling process.
Evaporatively cooling fermions gets increasingly more difficult at low temperatures due to Pauli
blocking [11], which reduces the effective collision cross section and consequently the thermaliza-
tion efficiency in the quantum degenerate regime. To further reduce the relative temperature TTF ,
Figure 1.1: Observing Fermi surfaces: Momentum distribution of an ultracold Fermi gas of 40
K
atoms in an optical lattice (from [7])
1
1 Introduction
ion pump
Rb
ion pump
Ti
pump
MOT
chamber atom
transport
standard
2D MOT as a science glass cell
potassium
replacement K
source
characterizing the quantum degeneracy, increasingly more hot atoms have to be removed during
the cooling process and therefore higher initial atoms numbers are needed.
We plan to increase the initial atom number trapped in the 3D MOT of the optical lattice
experiment shown in figure 1.2 by implementing a 2D MOT as a high atomic flux source for
40
K (also shown in the figure). This 2D MOT generates an atomic beam which will be used to
load the 3D MOT, leading to higher atom numbers and ultimately to lower temperatures of the
degenerate Fermi gas. Previously, 2D MOTs have mainly been implemented for rubidium, while
to our knowledge only very few have been realized for fermionic [12] and bosonic [13] potassium.
This thesis reports on the experimental realization of a 2D MOT for 39 K which will also be
suitable for 40 K after small modifications. Trapping and cooling in two dimensions is achieved
by using two perpendicular retro-reflected laser beams and two pairs of magnetic coils. A push
laser can then be used to generate a cold atomic beam which can be directed into the 3D MOT
via a differential pumping tube. The dependency of the atomic beam on different factors such as
laser detunings, laser powers and push beam parameters is investigated in detail.
2
2 Theory
This chapter will guide through the theory that one needs to understand how magneto-optical
traps work. In section 2.1, important facts about potassium as well as its level scheme are
presented. The scheme of Doppler-free saturated absorption spectroscopy which is used in our
set-up is briefly introduced in section 2.2. Starting from section 2.3, the concepts from atomic
physics needed to explain magneto-optical trapping in section 2.5 are introduced. Finally, in
section 2.6, we will familiarize the reader with the different types of two-dimensional magneto-
optical traps.
2.1 Potassium
Potassium is a soft silvery-white alkali metal that occurs naturally bound to other elements in for
example seawater and many minerals. It oxidizes rapidly in air and is very reactive with water.
Potassium is the seventh abundant element in the earth’s crust.
Table 2.1 gives an overview of the different potassium isotopes. Note that 40 K is besides 6 Li
the only fermionic isotope of an alkali metal and is therefore one of the main candidates for
experiments with ultracold fermions. Because of its very low natural abundance, enriched samples
(containing up to 14% of 40 K) are available, but they are relatively expensive.
Having only one valence electron, potassium is very similar to hydrogen. Therefore the same
Hamiltonian can be used to deduce the level scheme [14] and only the effective nuclear charge has
to be adapted from the hydrogen atom calculation. The Hamiltonian for this problem is given
by:
~2 Ze2 µ0 Z 2 e2
H=− 4+ + L·S (2.1)
2m0 4π0 r 8πm0 r3
with m0 being the electron mass, Z the number of protons in the nucleus, r the distance of the
electron from the nucleus and e the elementary charge. The last term describes the spin-orbit
coupling. Since neither L nor S commutate with the Hamiltonian and are therefore not good
quantum numbers, the total electron angular momentum J = L + S is introduced. As a result
3
2 Theory
of the spin-orbit coupling, the energy levels are split up in what is called the fine structure. The
eigenvalues of the new operator can be calculated using
1 2
L·S= (J − L2 − S2 ) (2.2)
2
to be
1
EFS ∝ [j(j + 1) − l(l + 1) − s(s + 1)] (2.3)
2
Here l and s are the orbital angular momentum and spin quantum numbers. As one can easily
see, only energy levels with l 6= 0 are split up.
Taking also the so called hyperfine splitting into account, that is the coupling of the total angular
momentum of the electrons J to the nuclear spin I, the new eigenvalues can be calculated in a
similar way by introducing the new total angular momentum F = J + I :
Now levels with l = 0 are also split up and one gets for example for the ground state of 39
K two
levels with
4
2.3 Light forces
D2 766.7017 nm
Γ = 6.1 MHz
F = 2 (173.1 MHz)
2 125.6 MHz 235.3 MHz
S1/ 2 2
461.7 MHz S1/ 2 1285.8 MHz
2 F = 2 (-140.0 MHz)
S1/ 2
F = 1 (-288.6 MHz) 254.0 MHz
F = 1 (-394.0 MHz)
F = 9/2 (-697.1 MHz)
• The laser is resonant with an optical transition for atoms in the velocity class v = 0. In this
case both the incoming and the reflected beam can drive the transition of atoms of that
velocity class. The incoming beam saturates the transition and thus the reflected beam will
not get absorbed anymore. Therefore only half the light will be absorbed compared to the
above case.
• The laser is blue detuned to one transition and red detuned with the same detuning to
another transition having a higher frequency than the first. The incoming beam saturates
the first transition for atoms with velocity v and at the same time the other transition of
atoms of the velocity class −v. The reflected beam will once again go through the vapor cell
without absorption. This is called a „crossover“ because (at least) two different transitions
are involved.
As will be explained in chapter 3.3.3 on page 25 this technique can be used as part of a frequency
stabilization scheme for lasers.
An atom in a light field will either scatter photons or get polarized by the light and therefore
experience a force. One can easily derive these forces acting on an atom when considering the
interaction of electro-magnetic radiation with a two-level atom [17].
Take an atom with its center at position R and a single electron with a spatial coordinate r
relative to the center of the atom. Consider an incoming plane wave of the form
5
2 Theory
where e is the polarization vector, E0 (R) the amplitude, ωL the angular frequency of the light
and φ(R) its phase.
The interaction of the two-level atom with the electro-magnetic radiation is described by a Hamil-
tonian of the form
Here, the energy of the ground state |gi has been set to zero and thus the energy difference
between the ground state and the excited state |ei where the atom has absorbed a photon is
E = ~ω0 . In Hint r stands for the position operator of the electron and E(R, t) is the electric
field at the position R of the atom.
The difference between the frequency of the laser light and the atomic resonance δ = ωL − ω0 is
called the detuning. A positive detuning means that the laser frequency is higher than that of
the resonance and is called a blue detuning, while a negative detuning is called a red detuning.
By writing the eigenfunctions |ψi i of the Hamiltonian in terms of density matrices %ij = |ψi i hψj |
and applying the rotating wave approximation (RWA), in which non-resonant terms are neglected,
the optical Bloch equations are obtained. Out of these, the forces acting on an atom in an electro-
magnetic field can be deduced. On one hand there is the conservative dipole force which only
comes into play for far-detuned light sources with spatially dependent amplitudes. On the other
hand there is the dissipative spontaneous light force, which can be easily understood within the
following picture:
When an atom absorbs a photon, its momentum and energy are increased by p = ~k and E = ~ωL ,
respectively. On average it will remain in the excited state for a time τ which corresponds to the
linewidth of the transition Γ = τ1 (note that Γ is an angular frequency 2π · Hz) . The photon will
then be re-emitted in a random direction and the total momentum is conserved. Therefore the
maximum force exerted on a two-level atom is
~kΓ
Fmax = (2.9)
2
The factor of 2 in the denominator comes from the fact that the atom is assumed to be in the
ground state for half of the excited state lifetime ( τ2 ) and in the excited state for the other half.
The actual spontaneous force for an arbitrary transition and laser detuning δ is given by
Ω21
Γ
F = ~kL 2
, (2.10)
2 δ2 + Γ2
+
Ω21
4 2
d·E
with Ω1 = being the so called Rabi frequency and d = er the dipole operator.
~
The scattering rate on the other hand is given by
Γ s F
γ= · = (2.11)
2 1+s+ 2δ 2 ~kL
Γ
6
2.4 Doppler cooling
Here s = I
IS is the saturation parameter with
~ω03 Γ
IS = (2.12)
12πc2
being the saturation intensity. When approaching the saturation intensity, additional light in-
tensity will not significantly increase the spontaneous light force any more because the transition
cannot be driven any faster than at the maximum rate given by the finite lifetime τ of the excited
state.
For potassium, the saturation intensity is IS = 1.77 mW/cm2 .
v
ν 0 = ν(1 + ) (2.13)
c
δD = δ − ωD = ωL − ω0 − kL · v0 (2.14)
F = ~kL 2
(2.15)
2 (δ − ωD )2 + Γ2
+
Ω21
4 2
If we now consider two counterpropagating laser beams with k1 = −k2 , the total force - called
the Doppler force - will be the sum of the forces of the two light beams:
!
~kL ΓΩ21 1 1
F = F+ + F− = − (2.16)
4 2
(δ − ωD )2 + Γ + 1
Ω2 2 Ω2
(δ + ωD )2 + Γ + 1
4 2 4 2
If these two laser beams are both red detuned with respect to the atomic transition they cancel
for an atom a rest. If an atom moves towards a laser beam its effective detuning with respect to
this laser beam will get smaller and thus the force exerted upon the atom from this beam will
get bigger. As a result the atom will slow down. Therefore the Doppler force can formally be
considered a friction force.
Note that using Doppler cooling atoms cannot be cooled to complete rest, because there is always
a heating process counteracting the cooling. This heating results from the discrete momentum
transfer when the atom emits a photon. In momentum space the movement of atoms at low
temperatures is a random walk with step size ~k. An equilibrium is reached when the effect of
7
2 Theory
{
E
mJ’ = +1
J’ = 1 mJ’ = 0
mJ’ = –1
J=0
σ+ σ–
mJ = 0
z=0
cooling is canceled by that of the heating. The lowest temperature that can be achieved using
Doppler cooling is given by the Doppler limit:
~Γ
TDoppler = (2.17)
2kB
Such a four level atom is depicted in figure 2.2. We can see that for the ground state J = 0 the
energy remains undisturbed because it has no angular momentum (l = 0) while for the excited
state J0 = 1 (l = 1) the degeneracy is lifted for z 6= 0 according to the energy term mentioned
above with mJ0 = +1, 0, −1.
To drive the transitions J = 0 → J0 = 1, mJ0 = +1 and J = 0 → J0 = 1, mJ0 = −1 right-circular
polarized (σ + ) light is needed for the first transition or left-circular polarized (σ − ) light for the
latter, respectively, as a direct consequence of the conservation of angular momentum.
Now consider an atom at a position z > 0 and two laser beams, one propagating in positive z
direction with σ + polarization and the other one propagating in −z direction with σ − polarization.
Both beams are red detuned to the atomic transition by the same frequency. In this case, more
σ − than σ + photons will be absorbed since the transition to the mJ0 = −1 state is much closer
to resonance than that to the mJ = +1 state. Therefore, the atom will experience a net force
8
2.5 Magneto-optical traps
Figure 2.3: Numerical calculation of the MOT forces on an atom at rest (from [16])
-10Γ
-2.5Γ
1
0Γ
+0.5Γ
N)
-20 0
F / (10
–1
–20 –10 0 10 20
z / cm
pushing it towards the center. The same argument holds for atoms at a position z < 0, while
those at position z = 0 will experience no net force at all.
The total force experienced by an atom has the following form:
!
~kΓΩ21 1 1
F= − (2.19)
4 (δ − ωD − Γ2 Ω21 Γ2 Ω21
~ ) + + (δ + ωD + ~ ) + +
µB b z 2 µB b z 2
4 2 4 2
If we numerically calculate this force for a 1D 39 K-MOT with a magnetic field gradient b = 4 G/cm
and a saturation parameter of s = 8 and plot it versus the position of the atom for different
detunings (figure 2.3), we can see that the shape is in fact highly dependent on the detuning.
While with a large negative detuning the capture range and also the magnitude of the force
increases, the localization of the MOT around the trap center decreases at the same time. With
smaller detunings, the capture range decreases but the density of the MOT gets bigger. For zero
detuning obviously no atoms can be captured while for positive detunings they are even repelled
from the center.
Finally it should be stated that while a 1D scheme was used to simplify the argument, it can
easily be extended to two or three dimensions by using four or six laser beams and a magnetic
field that grows linearly along two or three spatial axes.
In the model of a MOT discussed in the last chapter we assumed that the atom is a pure four-level
system, which means that the transition is closed and no atoms decay into other states during
the cooling process. However, this is not the case for most real atoms.
Most elements are not suitable for laser cooling due to their atomic structure, for example because
their optical transitions are below 300 nm where it is extremely difficult to generate the amounts
of laser power needed. Even if an element has optical transitions at an accessible wavelength
there may be possibilities due to an element’s hyperfine structure for it to emit a photon from its
excited state and not return to its original state. This third state is not accessible by the cooling
light and therefore the atom is lost from the trap.
The cooling scheme explained above not only works for J = 0 → J0 = 1 transitions but for any
J → J0 = J + 1 transition. In alkali metals, cooling and trapping can be achieved by using the
9
2 Theory
Potassium. For 39 K the 2 S1/2 , F = 2 → 2 P3/2 , F0 = 3 transition will be used as the cooler and
0
the 2 S1/2 , F∗ = 1 → 2 P3/2 , F∗ = 2 transition as the repumper.
An experimental and theoretical investigation of the cooling mechanisms in potassium 3D MOTs
can be found in [19].
ˆ
dN
= αφB − (χφB + γ)N − β n2 dV (2.20)
dt
Here N is the atom number in the trap, n the atom density, φB the flux of the atomic beam
loading the trap, α the probability that an atom from the atomic beam is captured by the MOT, χ
the probability that an atom in the MOT is knocked out of it by a particle from the atomic beam,
γ a constant that accounts for trap losses due to collisions of trapped atoms with background
gas and β a constant that accounts for trap losses by collisions of two trapped atoms. We can
simplify this model by assuming that the shape of the atom density distribution is Gaussian and
independent of time, allowing us to write the equation with V being the trap volume as follows:
dN N2
= αφB − (χφB + γ)N − β (2.21)
dt V
One can now study two separate cases:
To get analytical solutions, we have to set β = 0, which can be justified by the fact that losses
due to collisions of two atoms in the trap only play a role for very dense traps. With the initial
condition N (t = 0) = 0 we obtain the following solution:
N (t) = N0 1 − e−(χφB +γ)t with (2.22)
αφB
N0 = being the maximum atom number (2.23)
χφB + γ
10
2.5 Magneto-optical traps
. 1
T = (2.24)
χφB + γ
dN .
= αφB = d (2.25)
dt t=0
To study the decay of the trap the atomic beam loading the trap is turned off and we set φB = 0
to get the following solution:
N0 e−γt
N (t) = (2.26)
1 + NV0γβ (1 − e−γt )
. 1
τ= (2.27)
γ
This is the simplest way to load a MOT. A dispenser or metallic source is attached to the MOT
chamber and by letting a current flow through the dispenser or heating the intended background
pressure is established. The loading rate of the MOT in this scheme is highly dependent on the
gas pressure. But while a higher pressure enhances the loading rate, it reduces the MOT lifetime
due to increasing collisions with the background gas.
Double-MOT systems
This scheme consists of two 3D MOTs connected by a differential pumping tube [22]. First, a
MOT is loaded in one chamber from the background gas at a relatively high pressure. Then the
atoms are transferred by a laser pulse to the other chamber, which is under UHV conditions. Up
to 80% of the originally trapped atoms can be recaptured in the second chamber by this method.
By repeating this step, even more atoms can be trapped in this second chamber, where due to
the low pressure the MOT lifetime is much higher than in the first chamber.
In place of the first MOT one can also use a Low-Velocity Intense Source (LVIS) [23]. An LVIS
is essentially a normal MOT loaded from the background gas. The only difference is that one of
the mirrors of the LVIS has a hole so that the atoms do not feel any light force coming from that
place. This way an atomic beam of cold atoms escapes through the hole. This beam is then fed
into the second MOT.
11
2 Theory
There are two common techniques to decelerate a thermal atomic beam created in an oven by
monochromatic laser radiation directed towards the atomic beam. Since the scattering rate of
the photons is velocity dependent, they would be tuned out of resonance during the deceleration.
Therefore either the magnetic field has to be varied along the deceleration path, changing the Zee-
man splitting (Zeeman-Slower) or the laser frequency needs to be changed during the deceleration
(Chirp-Slower). The decelerated beam is then used to load the 3D MOT.
2D MOT
This is another possibility to efficiently create a high-flux beam of cold atoms, first implemented by
Nellessen et al. [24]. In this configuration, which will be explained in detail in chapter 2.6, atoms
are trapped in two dimensions while in the third they can, depending on the exact configuration,
move more or less freely creating the atomic beam.
Comparison
While loading from the background gas is certainly the simplest method, it has disadvantages:
the loading can take relatively long and the MOT chamber cannot be kept under UHV conditions.
With a double-MOT scheme it is possible to separate the high-pressure region for loading from
the main experiment, but loading can take up to 60 seconds, considerably slowing down the
experimental cycles. Both capturing from a decelerated atomic beam or using a 2D MOT require
much shorter loading times (only a few seconds). While a Zeeman slower has the advantage that
it can decelerate atoms having a very low vapor pressure at room temperature it takes up a lot of
space on an experimental table, is not isotope-selective and also requires relatively high amounts
of source material. A 2D MOT on the other hand does not have any of these drawbacks, being
therefore at least in our case the optimal choice.
Pure 2D MOT The basic scheme explained above using just two pairs of counterpropagating
laser beams is commonly called a pure 2D MOT.
While 87 Rb fluxes of up to 6 × 1010 atoms/s have been reported by Schoser et al. [25], Catani et
al. [13] reported that there was no observable atomic flux for 39 K at all when using only a pure
2D MOT scheme. This, however, will be refuted in this work as a considerable flux has been
observed.
12
2.6 2D magneto-optical traps
σ+ MOT beams
Atomic beam σ−
σ+ Pus
h bea
m
MOT coils
σ−
One important feature of pure two-dimensional traps is that atoms which exit the MOT chamber
through a small aperture are velocity-filtered by the geometry of the trap [25]. To get cooled in
the MOT and to pass through the aperture the atoms have to fulfil two conditions:
1. Their radial velocity needs to be smaller than the capture velocity of the trap which is
mainly given by the transversal extent of the MOT beams.
2. Their axial velocity must be small enough to allow them to stay within the trap until they
are cooled to low radial velocities. This is because their divergence (which is determined
by their radial velocity) must be small enough for the atoms to be able to pass through the
aperture upon exiting from the trap.
As a result, the atomic beam shows a very narrow velocity profile peaking at a velocity which
is well capturable by a 3D MOT (Catani et al. [13] report a value of 33 ms/s). Note that the
number of background atoms leaving through the aperture is negligible due to its small size.
2D MOT with push beam One problem of the pure 2D MOT scheme is that an atomic beam
exits on both sides while normally only the beam on one side is used. To circumvent this problem
a fifth laser beam, called the push beam, can be applied as shown in figure 2.4. This beam will
either act as a „plug“ which just prevents the atoms from exiting from one end of the cloud or
it can even additionally accelerate the atoms in the cloud or in the atomic beam to push them
towards the exit of the chamber.
When using a 3D MOT to capture the atomic beam and a push beam for the 2D MOT, however,
one has to take into consideration that the push beam may disturb the 3D MOT and aggravate
the capturing process.
2D+ MOT A 2D+ MOT is an extension of the „2D MOT with push beam“ scheme explained
above. Here, not only a push beam but another MOT beam counterpropagating with the push
beam is used. This beam usually enters the vacuum from the side and is reflected along the
MOT axis by a 45° mirror, in the center of which a small exit hole for the atomic beam is drilled.
Therefore the counterpropagating beam contains a shadow along the axis so that atoms close to
the axis are strongly accelerated out of the capture region by the push beam.
13
2 Theory
This is reportedly the most effective configuration for rubidium [25] while Catani et al. [13] report
for 39 K no significant improvement for the 2D+ MOT configuration over a 2D MOT with push
beam. According to the findings of Catani et al. it was decided to build a 2D MOT with push
beam thereby avoiding the technical difficulties associated with a 2D+ MOT configuration.
Schoser et al. [25] and Catani et al. [13] both suggested theoretical models for the dynamics in
2D MOTs for 87 Rb and 39 K, respectively.
14
3 Experimental set-up
In this chapter, the experimental set-up is explained in detail. The vacuum chamber and the
surrounding optics were built by Roger Gehr [26] and completed by Dominik Leitz [27] during
two past diploma theses and are explained in sections 3.1 (2D MOT) and 3.2 (3D MOT). The
laser set-up to generate the MOT light is discussed in section 3.3. During the course of this
thesis, detection set-ups for both the atomic beam in the 2D MOT chamber and for the 3D MOT
fluorescence have been built. These as well as the implemented computer control system are
discussed in section 3.4.
Figure 3.1 gives an overview of the core part of the experimental set-up: the vacuum chambers and
the measurement devices. The whole vacuum chamber including the optical set-up for generating
the laser light needed to operate the system is built on a single laminar-flow-isolated optical table.
To ease the planned inclusion of the 2D MOT into the optical lattice experiment the optics to
operate the 2D MOT is already set-up on a separate breadboard which can be easily detached
and moved.
The vacuum chamber consists of two main parts, the 2D MOT on the right where the cold atomic
beam is generated and on the other hand a 3D MOT as a detection device to the left. To keep
the chamber under ultra high vacuum (UHV) conditions, a 80 l/s ion pump is attached to the
3D MOT. The two MOT chambers are connected by a differential pumping tube (see 3.1.4).
For the detection of the MOTs there are two CCD finger cams and two photodiodes (one for the
detection of the 3D MOT fluorescence and another one in a detachable set-up for the measurement
of fluorescence from the atomic beam induced by a probe beam in the 2D MOT chamber).
80 l/s ion
vacuum pump
RGA
p=8· 10-10 mbar
coils coils
photo
diode
camera
and photo diode atomic beam
potassium vial
differential pumping tube
low pressure high pressure
15
3 Experimental set-up
2D MOT
optics
ion pump
2D MOT
coils
2D MOT
3D MOT chamber
atomic
beam
3D MOT detection
camera
Additionally there is a residual gas analyzer (RGA) mounted on the 2D MOT chamber and the
vacuum pump pressure sensor.
A photograph of the vacuum set-up and the surrounding equipment can be seen in figure 3.2.
When designing a 2D MOT several different aspects have to be taken into account:
• Both the axial (i.e. the MOT length) as well as the transversal cooling beam size and the
aperture diameter need to be chosen in such a way that most of the atoms in the atomic
beam are in a velocity class that can be captured by the 3D MOT.
• The mean free path in the chamber which is pressure-dependent needs to be at least on the
order of the length of the MOT if one wants to keep losses by collisions with the residual
gas low [26].
• Depending on the targeted pressure in the 3D MOT, one may have to use a differential
pumping tube to connect the 2D MOT and the 3D MOT to allow for higher pressures in
the 2D MOT while keeping the pressure in the 3D MOT low.
• The shape of the differential pumping tube needs to take the expected atomic beam diver-
gence into account to avoid losing too many atoms by collisions with the tube.
Additionally, in our case, the whole set-up needed to be as compact as possible such that it fits
into the remaining space in the optical lattice experiment.
Figure 3.3 shows a picture of the bare stainless steel 2D MOT chamber. The whole chamber has
a size of 210 mm × 70 mm × 70 mm. On the left one can see the insets for the four large MOT
windows with a size of 90 mm × 40 mm. The windows are anti-reflection (AR) coated on one side
to reduce losses. On the right, there are four CF40 flanges of which currently only the one on
16
3.1 2D MOT vacuum chamber and optics
CF40 flange
push window
Figure 3.4: Magnetic field gradient for different coil currents (from [26])
magnetic field gradient / (G / cm)
20
10
measurement
0 fit ∂B/∂x = 4.33 (G/(cm⋅A) ⋅ I
the back side and the one below are used to connect a residual gas analyzer and the bellow that
holds the potassium source, respectively. The other two flanges are equipped with valves.
The back window with a diameter of 40 mm allows for optical access for a push beam and a
camera or a photo diode. All windows are sealed onto the chamber body with indium wires of a
purity of 99.999% which results in leak rates of less than 2 × 10−7 mbar l/s [28] (for details see
[26]).
Two magnetic coils in anti-Helmholtz configuration are used to generate a quadrupole field inside
of the chamber leading to a spatially dependent Zeeman splitting of the energy levels of the
atoms. In a 2D MOT one only wants to have magnetic gradients in the axial plane while along
the MOT axis the magnetic field should be constant to let the atomic beam easily escape towards
the aperture. To create a more homogeneous field, two pairs of rectangular coils are used in our
set-up. Each coil has a size of 145 mm × 85 mm and they have a distance of 90 mm. The wire
cross section is 3 mm2 .
The four coils are wired in pairs of two facing coils for which the currents can be independently
controlled. For all measurements the two pairs of coils were connected in series, though. Addi-
17
3 Experimental set-up
tionally, there are two coils interleaved with the main coils to apply an offset field, one in the
back horizontal coil and another one in the upper vertical coil.
The magnetic field gradient produced for different currents is shown in figure 3.4.
A metallic source was used as the primary atom source. Such a metallic source is just a glass
ampul containing a few grams of potassium in its metallic form. This ampul is placed in a bellow
flanged to the bottom of the chamber and is broken inside of the chamber by bending the flexible
bellow after the vacuum has been established. Thereupon, the source emits atoms from its surface
at a rate corresponding to its vapor pressure and the emittance can therefore only be regulated
by manually heating the bellow from the outside.
The metallic source currently in use contains the different potassium isotopes in amounts accord-
ing to their natural abundance (i.e. it contains no enriched fermionic 40 K). Therefore the most
abundant isotope is the bosonic 39 K.
A residual gas analyzer (RGA) from MKS Instruments (e-Vision+ ) is attached to the 2D MOT
to measure the partial pressures of residual gas elements in the MOT chamber.
Such a RGA works according to the following principle: At the inlet the residual gas elements
are ionized by electron ionization (E + e− → E+ + 2e− ). Then their masses are detected by mass
spectrometry: the ions are first accelerated in an electric field from where they reach an area with
a homogeneous magnetic field in which they are differently deflected according to their m q ratio.
The mass-separated ions (q can be assumed to be 1 for almost all of the produced ions) can then
be detected by ion counters in different spatial positions. In the end, partial pressures for the
elements are calculated from the count rates.
A typical spectrum from the RGA is shown in figure 3.5 on the next page. It is important to
be aware that using the ionizer filament leads to heating of the vacuum chamber and thus to
the ablation of elements which were deposited on the chamber walls. Therefore measurements
should only be made after the RGA filament ran for more than an hour. During the course
of our measurements we also found that the RGA filament constantly enhances the amount of
potassium available for the 2D MOT and thus acts as nice potassium dispenser. To always
operate the system under the same conditions we therefore kept the filament running over the
whole measurement period.
One of the major advantages of using a 2D MOT to load a 3D MOT is that one can separate
the two chambers by a differential pumping tube in order to keep a relatively low pressure in the
3D MOT. This allows for long lifetimes (by reducing collisions with the background gas) while
maintaining a higher pressure in the 2D MOT for efficient loading.
The targeted pressure in the 3D MOT was defined to be about 10−9 mbar. Gases at these pres-
sures at room temperature are in the molecular flow regime, in which gas flow is well understood
(for an overview see [29]). The pressure ratio of the two chambers separated by a differential
pumping tube and a pump connected to the 3D MOT chamber is given by (calculation from [26])
P3D C
= (3.1)
P2D SUHV
18
3.1 2D MOT vacuum chamber and optics
Figure 3.5: Typical measurement from the residual gas analyzer (RGA). The spectrum was
measured after the chamber had first been heated to 85 °C, but then let cool down and stabilize
at room temperature over the course of several days. The pressure in the 3D MOT chamber was
8 × 10−8 mbar. Besides 39 K only H2 is available in significant amounts in the chamber. As can
be expected from their natural abundance, the other potassium isotopes can be seen as well: 41 K
in an amount about 10 times lower than that of 39 K, and only a few traces of 40 K.
10–7
partial pressure / mbar
10–8
10–9
10–10
10–11
1 5 10 15 20 25 30 35 40 45 50
atomic mass / amu
with SUHV being the pump rate in m3 /s and C being the conductivity of the differential pumping
tube:
d3
C = 121 m/s · (3.2)
l
The tube used in our setup has a length of l = 90 mm and an inner diameter of d = 16 mm. When
operating the pump at a rate of 80 l/s one gets a pressure ratio of more than 10−5 which would
theoretically allow for operation of the 2D MOT at a maximum pressure of about 10−4 mbar,
although pressures above 10−6 mbar lead to losses by collisions with background atoms which are
too high to operate a MOT.
To get a good velocity selection (see 2.6 on page 12) while reducing collisions of the atomic beam
with the tube walls, the pumping tube is fitted with an additional tube with a conical hole. The
hole has a divergence of 50 mrad which corresponds to the beam divergence measured in [13]. Its
diameter is increasing from 2 mm to 10 mm. We assume that the results obtained in the above
calculation are still applicable for a tube with such a conical shape.
The preparation of the MOT light is done in two stages which are connected by optical fibers. In
the first stage, the laser light is generated, frequency stabilized, amplified and additional frequency
components are modulated onto the light. Details on this first stage can be found in chapter 3.3
on page 21. Then the light is coupled into a fiber which cleans the beam profile and allows to
guide the light to where the MOT is set up. In a second stage the light is split into the different
MOT beams, their polarization is adjusted and the beams are enlarged before they are fed into
the MOT chamber.
In detail, the set-up is as follows (see also figure 3.6):
19
3 Experimental set-up
λ/4
Aspherical
lenses
First
cylindrical
lenses
Second
cylindrical
lenses
2D MOT
chamber
λ/4
Mirror
Fibercoupler. An adjustable fibercoupler from Schäfter + Kirchhoff is used. The beam has a
diameter of 1.43 mm upon exiting the fibercoupler.
Half-wave plate and polarizing beam splitter. These two components split the main beam into
the two MOT beams. The ratio of the intensities in the two beams is adjustable by turning
the half-wave plate.
Quarter-wave plates. They turn the linearly polarized light into circularly polarized light.
Aspherical and cylindrical lenses. Together they form a Kepler type telescope consisting of a
small focal length (11 mm) objective lens and two cylindrical lenses with focal lengths of
75 mm and 150 mm for the vertical beam and 100 mm and 150 mm for the horizontal beam,
respectively, forming the ocular. This telescope enlarges the beams to their final elliptical
shapes with a size of 80 mm × 40 mm. The telescope is adjusted in such a way that the
beam is slightly focused to correct for the intensity losses caused by the windows (they are
estimated to be around 12%). This way radiation pressure equilibrium of the incoming and
reflected MOT beam can be restored despite of the losses.
Quarter-wave plates. Since the light passes two times through the quarter-wave plates, it keeps
its handedness despite of being reflected from the mirror.
Retro-reflecting mirrors. They retro-reflect the MOT beams back into the chamber.
An initial adjustment of the MOT optics was done using a power meter for the light balance and
calculated values for the loss correction of the MOT beams and by ensuring that the incoming
and reflected MOT beams were perfectly overlapping.
20
3.2 3D MOT vacuum chamber and optics
3.3.1 Overview
Three separate laser sources at a wavelength of 766.7017 nmare used in this experiment, one for
the operation of the 3D MOT, one for the 2D MOT and a third for a push beam. Figure 3.7
gives an overview on how the light for these three components is created.
The 3D MOT light is produced in a tapered laser (TL) which combines both a diode laser and
a tapered amplifier. A small fraction of the output of 450 mW is sent through a double-pass
125 MHz AOM which induces a blue-shift of 250 MHz (adjustable) to that part of the light.
This light is then used for a RF lock set-up to frequency stabilize the TL (see chapter 3.3.3 on
page 25 for details on laser frequency stabilization). Therefore the main part of the TL light
is red-detuned by the AOM frequency shift with respect to the lock point. The lock point of
the TL is chosen in such a way that taking this frequency shift into account the main part of
the TL light is red detuned by about 30 MHz to the cooler transition F = 2 → F0 = 3 of 39 K.
An electro-optical modulator (EOM) modulate sidebands with a distance of 460 MHz (slightly
adjustable) onto that light. The blue sideband will be used to drive the repumper transition
0
F∗ = 1 → F∗ = 2 of 39 K. After the fiber, which acts as cleaning stage, only 40 mW of laser power
remain because the beam profile produced by the tapered laser is very bad leading to a low fiber
coupling efficiency.
Two standard diode lasers [30] with an output of 20 mW and 35 mW generate light to seed the
2D MOT tapered amplifier and for the push beam respectively. Both are frequency stabilized
using offset locks to the light of the tapered laser after the AOM stage . The offset locks are set
near 250 MHz which again results in a net detuning to the cooler transition of about 30 MHz.
An EOM is used to modulate sidebands with a distance of 460 MHz onto the carrier to drive
the repumper transition in the same way as for the 3D MOT light. The tapered amplifier (TA)
input is about 10 mW resulting in an output of 700 mW. 200 mW of the generated output can
be coupled into the fiber leading to the 2D MOT.
In contrast to the two other lasers, the push laser is coupled unamplified into a fiber and contains
light at only one frequency. By adjusting the offset lock, the push beam can either operate near
21
3 Experimental set-up
Tapered
Amplifier
700 mW
the cooler or repumper transition, though. The laser power of the push beam behind the fiber is
about 5 mW.
Figure 3.8 shows the optical set-up in detail. On the schematics one can also see two additional
vapor cells to perform FM spectroscopy of the seed and the push laser and three removable
mirrors to guide the light into cavities for diagnostics. There is also a shutter to turn the push
beam off when not needed.
3.3.2 Lasers
To operate a MOT a narrow-band (< 1 MHz) light source with a frequency stabilized with respect
to an atomic transition and a power of several 100 mW is needed. Tunable diode lasers are widely
used for such scientific applications because they are reliable and easily available at a moderate
price. However, the laser output is typically some tens of MHz wide, can only be continuously
tuned over very limited regions and single-mode operation cannot be enforced. By using external
optical feedback to control the frequency, these characteristics can be greatly improved [31]. By
combining a laser diode (LD) and a tilted optical grating an external cavity is created which
stabilizes the laser frequency and allows for fine control over it. Figure 3.9 shows one of the used
lasers in this so called Littrow configuration. The design is based on the one described in [30].
The light coming from the diode which is mounted on a temperature-controlled heat-sink gets
diffracted from a grating with a line spacing of 1800 lines/mm mounted on a adjustable flexture.
The first diffraction order gets reflected back into the laser diode forming the external cavity,
selecting the longitudinal mode of the cavity to lase which is closest to the gain profile maximum
of the laser diode. The zeroth order is coupled out of the cavity. Thus such a laser diode with an
external cavity with a length L of about 1 cm has a mode separation of
c
∆ν = ≈ 10 GHz (3.3)
nL
while the laser diode itself (cavity length 1 mm) has a mode separation of 100 GHz. The band-
width of the laser diode can be reduced to several kHz in the external cavity set-up [30]. By either
22
3.3 Laser system
Vapor
Cell
Push LD Seed LD
λ/4
PBS PD
TL
flexture
vertical adjustment
laser diode λ/2 screw
horizontal
adjustment
safety screw
circuitry
piezo-
temperature electrical
sensor element
23
3 Experimental set-up
adjusting the vertical adjustment screw or by changing the voltage on a piezo-electrical element
one can change both the cavity length and the angle of the grating. Note that the change in the
Bragg condition resulting from the different inclination of the grating can be neglected compared
to the change in length of the cavity and thus one changes de facto only the cavity length by
using either of these two methods.
Since the laser is very susceptible to small changes in temperature and diode current, it is mounted
on a big thermal mass which is temperature controlled by a peltier element to an accuracy of
about 10 mK and is driven by a stabilized laser current supply. Additional elements which can
be seen in the figure are: A collimation lens which is adjusted when building the laser in such a
way that the laser beam stays collimated and keeps a nice beam for several meters. Optionally
a half-wave plate can be placed into the cavity to change the polarization inside the cavity. The
horizontal adjustment screw is used to adjust the vertical angle under which the light gets back-
reflected into the laser diode. The safety circuitry includes diodes which protect the laser diode
from being connected to a current source with the wrong polarity and also capacitors which
suppress high-frequency noise from the laser current supply.
Both the seed and the push laser use the set-up described above, whereas the tapered laser (TL)
for the 3D MOT light uses a different one. Details on the three lasers are explained below.
Seed laser
The seed laser features a standard laser diode in Littrow configuration and is operated with a
current of 50 mA at a temperature of 16.5 °C resulting in an output of 20 mW.
Since this laser power is much too low to operate a MOT and especially a 2D MOT with its
large beam profiles, the seed beam is fed into a „Tapered Amplifier“ (TA) after going through an
optical isolator and the EOM. A TA chip is basically a laser diode with a tapered shape in which
laser light is amplified several times.
From the seed beam 10.3 mW of laser power finally reach the TA which is operated with a
current of 2 A at 20 °C and produces at a maximum 730 mW of output of which 200 mW can be
successfully coupled into the fiber leading to the 2D MOT optics.
Push laser
The push laser features a standard laser diode in Littrow configuration. It is operated with a
current of 78 mA at 18.0 °C and produces 37 mW of output. The difference in output power
compared to the seed laser can be explained by the fact that here an AR-coated diode is used.
The push beam needs only to operate with a much smaller beam diameter than the actual 2D
MOT beams and therefore does not need a tapered amplifier.
A tapered laser (TL) combines both a conventional laser diode in Littrow configuration with a
tapered amplifier and therefore directly produces a small-bandwidth frequency-stabilized laser
beam with a power which can exceed 1 W. A schematic drawing of a tapered laser is shown in
figure 3.10. In contrast to a normal laser diode for external cavity set-ups the diode used for a
tapered laser has a tapered shape, with one diode facet being 3 µm in size and the other 200 µm.
The larger facet of the diode together with a grating builds an external cavity because a small
part of the spontaneously produced light is able to exit the diode through the small facet towards
the left. But here, while the first diffraction order of the grating is reflected back into the laser
diode, the zeroth order is dumped onto the laser diode box. When the light in the external cavity
passes through the tapered diode it is highly amplified and as the large facet is reflecting only a
small part of the light most of it will be coupled out of the diode (and thus the cavity) via the
right facet.
24
3.3 Laser system
50.5 O
79 O
External Cavity
TL diode
grating
piezo
crystal
cylindrical lens
For illustration, a picture of the tapered laser used in our experiment can be seen in figure 3.11.
In contrast to a standard diode laser the TL features two more lenses for collimation of the beam
outside the cavity. The adjustment and the tuning of the frequency work the same way as for the
conventional external cavity diode lasers. However, a TL is much more fragile with respect to
vibrations and acoustical noise and is also much more sensitive to small drifts in the laser diode
current. As a consequence, it was very hard to adjust the laser for stable operation while at the
same time optimizing it for high output power. A more or less stable operating point was found
at a current of 1.93 A and a temperature of 17.9 °C resulting in an output of up to 450 mW of
which 40 mW could be coupled into the fiber.
To compensate for fluctuations in the laser diode current and in the diode temperature but also
for mechanical vibrations, the laser frequency needs to be actively stabilized. This is achieved
through an electronic feedback circuit, namely a proportional-integral feedback loop. First a so-
called error signal needs to be measured, i.e. the deviation of the laser frequency from a set point.
In our experiment this error signal is deduced from either a FM spectroscopy or an offset-lock
set-up. By continuously ramping the voltage of the piezo-electrical transducer of the laser up and
down, the spectrum can be displayed on an oscilloscope. The set point of the feedback circuit
which is integrated in a „Lock-Box“ can be set by zooming in and locking (therefore also the
name of the box) the circuit to an edge of the abovementioned spectrum.
In the next two sections, the two laser locking techniques are explained in more detail.
25
3 Experimental set-up
Mirror
LD
ωL
Vapor Cell
VCO
ωm
PD
Mixer
DC Signal
Figure 3.12 shows a schematic overview of an FM spectroscopy set-up. The goal is to generate
a well resolved absorption spectrum of the atomic element in a vapor cell. Because of the large
Doppler broadening at room temperature, the technique of Doppler-free saturated absorption
spectroscopy needs be utilized to resolve transitions which are close together (see chapter 2.2 on
page 4). However, absorption spectra cannot directly be used as error signals. Since the lock
points one wants to use in such a spectrum are generally peaks which are either local maxima
or minima, deviations from the extrema will result in a change of the error signal with the same
sign for either direction of the deviation. To circumvent this problem it is desirable to generate
the derivative of such an absorption spectrum.
This is exactly what a FM spectroscopy set-up does [32, 33]: A local oscillator (VCO1 ) generates
a sinusoidal signal with a certain frequency (64 MHz in our case) which is brought onto the laser
light by either modulating the laser current with this signal or by letting the light pass through
an EOM which is driven with that frequency (which is the way it is done in our set-up). Doppler-
free saturated absorption spectroscopy is then performed in a vapor cell and the resulting signal
is measured with a fast AC coupled photodiode (PD). This signal is then demodulated with
the original signal from the local oscillator and the resulting signal is in first approximation the
derivative of the absorption signal.
This can easily be seen from the following short calculation [26]: Let the laser signal be modulated
with a frequency ωm and an amplitude M from the local oscillator. This leads together with the
laser carrier frequency ωL to an electric field of the following form:
For small modulations M one can only consider the carrier and the two first sidebands:
M M
E(t) ≈ E0 − exp(i(ωL − ωm )t) + exp(iωL t) + exp(i(ωL + ωm )t) (3.5)
2 2
When going through the vapor cell, the different frequency components undergo different phase
shifts (φ0,+,− ) and absorptions (δ0,+,− ) and the electric field becomes
1 voltage controlled oscillator
26
3.3 Laser system
lock point
CO
F=2 → F' = 1,2,3
signal / a.u.
F=1 → F' = 0,1,2
F=2 → F'=1,2,3
F=1 → F'=0,1,2
E (t) ≈ E0 exp [−δ0 + i(ωL + φ0 )]
0
(3.6)
M M
− exp [−δ− + i((ωL − ωm )t + φ− )] + exp [−δ+ + i(ωL + ωm )]
2 2
Optical frequencies of the order of ωL are not seen by the photodiode. If we now additionally
neglect terms in M 2 , the intensity detected by the photodiode can be expressed in the following
way:
In this equation, the cosine term is proportional to the absorption difference between the two
sidebands (that is the first derivative of the absorption in first approximation) and the sine term
on the other hand is proportional to the difference in phase shifts experienced by the carrier and
the average of the sideband phaseshifts (this is the second derivative of the dispersion in first
approximation). Depending on the phase of the original signal at the demodulating mixer, either
the amplitude of the absorption-dependent (cosine) or dispersion-dependent (sine) term of above
equation is obtained.
In figure 3.13 one can see a typical FM spectrum obtained from our tapered laser. The locations
of the zero-crossings resulting from a crossover peak (in the middle, „CO“) and two normal
absorption peaks for 39 K as well as the lock point used for all measurements are marked in the
plot.
Offset lock
FM spectroscopy allows frequency stabilization to fixed atomic transitions. However, one gen-
erally wants to be able to stabilize a laser to arbitrary detunings from a fixed frequency. To
achieve this, one uses a reference laser locked to an atomic transition and locks all other lasers
with respect to this reference laser by an offset locking scheme. The main idea here is, as depicted
in figure 3.14, to heterodyne the reference laser with the one that is to be locked and to measure
the beat signal between these two lasers with an AC coupled photodiode.
The detailed scheme works as follows: The beat signal ∆ν = ν1 − ν2 captured by the photodiode
will be mixed with the signal νVCO from the voltage controlled oscillator (VCO) resulting in
27
3 Experimental set-up
Ref.
LD LD
ν1 ν2
PD
VCO
νVCO Mixer
Delay line 1m
a signal with frequency ∆ν − νVCO . The oscillation frequency of the VCO can be tuned by
changing its input voltage via a potentiometer or an external signal. The signal is then split into
two with one part going straight into a mixer and the other going through a 1 m long coaxial
cable before entering the mixer. In the cable the signal acquires a frequency dependent phase
shift of φ = 2π(∆ν − νVCO )τ with τ ≈ 5 ns being the signal delay for a cable with a length of
1 m. This second mixer will multiply the two sinusoidal signals with their different phases:
1
cos(ωt + φ) cos(ωt) = {cos φ + cos(2ωt + φ)} (3.8)
2
Low-pass filtering this signal yields a signal which is only dependent on cos φ which in turn is
related to the frequency difference of the VCO and the beat signal. So an offset-lock signal is
therefore nothing more than the cosine of the frequency difference of the beat signal and the VCO
frequency. It is in principle possible to lock to any of the zero-crossings of the signal which are
spaced by about 1/τ ≈ 200 MHz. The biggest advantage of this set-up though is the possibility
to shift these zero-crossings along the frequency axis by changing νVCO which means one can
continuously change the detuning of a laser over a broad range of up to 100 MHz while keeping
it locked to the same zero-crossing.
Figure 3.15 displays a typical offset lock spectrum. The sinusoidal modulations of the signal
can easily be seen. The fact that the amplitude of the modulation decays towards the middle
of the spectrum and also far outside can be explained by the finite bandwidth of the offset lock
electronics. The lock points used to either stabilize the laser on the cooler or the repumper
transition of 39 K are marked in the plot.
Another convenient feature of an offset lock set-up is that one can simply feed the beat signal
captured by the photodiode into a frequency meter which then directly displays the offset to
the reference laser in MHz. Both our seed and push laser offset locks where equipped with such
digital frequency meters.
28
3.3 Laser system
lock point
lock point on repumper
on cooler
signal / a.u.
to be overlapped or the additional frequency needs to be modulated onto light. For this, either
an acousto-optical modulator (AOM, with the disadvantage that the frequency shifted beam will
also be spatially shifted) or an electro-optical modulator (EOM) can be used.
In our set-up EOMs are used because they offer the most convenient way to generate the repumper
frequency. An EOM uses the electro-optic effect to introduce a phase shift to a part of the
light traversing it. As an electric field is applied to a nonlinear crystal such as lithium niobate
(LiNbO3 ) its index of refraction changes proportionally to the applied field according to the
following formula:
E
∆n = n30 r (3.9)
2
with ∆n being the change in the index of refraction, n0 the index of refraction without an applied
field, r the corresponding element in the electro-optic tensor and E the applied electric field. The
resulting phase shift is:
πn30 rV l
φ= · (3.10)
λ d
Here λ is the wavelength of the light, l the crystal length and d the distance of the electrodes on
which a voltage V is applied.
If we now vary V = V (t) over time and define the complete phase of the signal after the crystal
to be Φ(t) = ωt + φ(t) one can easily see that this results in a frequency modulation:
1 dΦ(t) 1
dφ(t)
f (t) = = ω+ (3.11)
2π dt 2π dt
If one applies a sinusoidal voltage V (t) = cos(Ωt) and writes dφ(t) dφ(t)
dt in the form dt = δ sin(Ωt)
one can see that the EOM produces sidebands on the carrier ω at frequencies ω ± kΩ with k being
an integer > 0 :
29
3 Experimental set-up
while Jn is the n-th Bessel function. From this last formula one can also deduce the energy
distribution for the different sidebands.
Technically, EOMs are implemented as resonators and the frequencies with which they can be
efficiently driven are only in the range of a few MHz around its resonance. The two EOMs used
to create the repumper frequencies for the 2D and 3D MOT have its resonances somewhere below
460 MHz and they were set to 462 MHz during the measurements when not otherwise stated.
One important thing that needs to be considered is that the power an EOM will put into the
sidebands drastically depends on its temperature and thus it may be required to tune the EOM
frequency closer to its resonance to heat it up and then tune it back to its original value to
bring a reasonable amount of power into the sidebands at an off-resonance frequency [27]. Both
EOMs in use were driven by separate VCOs and 5 W amplifiers of the type ZHL-5W-1-SMA from
Minicircuits. Additionally to just changing the frequency by tuning the VCO voltage the EOM
power could also be varied by changing the amplifier supply voltage.
This section describes the control and detection devices that were used for the measurements
presented in chapter 4. Most of them were newly implemented in the course of this work. Figure
3.16 gives an overview of the whole measurement set-up.
In section 3.4.1 the different devices will be explained in detail. Then, in sections 3.4.2 and 3.4.3
the two set-ups to measure the atomic beam fluorescence in the 2D MOT chamber and to measure
the fluorescence of the 3D MOT are presented.
3.4.1 Components
Two simple CCD finger cameras from Conrad Electronics together with an enlarging optical
telescope consisting of two lenses are installed, one looking into the 2D MOT chamber and the
other one into the 3D MOT chamber. The 2D MOT camera is looking into the chamber through
the push beam window and a 50:50 beam splitting plate which allows for a simultaneous operation
of the push beam and the 2D MOT camera (although the 2D MOT cannot be seen when the
push beam is turned on due to stray light from the push beam). In this alignment, the 2D MOT
camera looks axially onto the 2D MOT and thus one can only see its axial projection as a small
dot. It would certainly be desirable to look at the 2D MOT from the side, but this is currently
not feasible since on one hand the space on the side of the 2D MOT is already taken by the MOT
mirrors and on the other hand the MOT fluorescence from the side would be much too small
compared to the stray light from the MOT windows. In fact, the 2D MOT can currently not be
seen at all from the side, even not with the infrared viewer. Also, the 2D MOT camera needs to
be focused on the very front of the 2D MOT chamber (the entry of the differential pumping tube
should be in focus) to be able to see the 2D MOT.
In contrast, the 3D MOT camera was easier to install as there were still two spare windows on
the 3D MOT chamber and because the 3D MOT fluorescence is much bigger. Setting the camera
to look through the lower window of the 3D MOT chamber via a mirror proved to be the easiest
way. The whole camera set-up is movable in one single part so that the focus could be adjusted
on a test set-up to be in the exact center of the 3D MOT.
A USB TV adapter with analog video inputs captures digital images and even movies of the
MOTs, whereas two TV monitors are used for adjustment.
30
3.4 Measurement set-up
Atomic Lock-in
Pump 3D MOT 3D MOT 2D MOT Beam Frequency
Pressure Photodiode Camera Camera Photo- Generator
diode
39
K Source
Temp. Residual
Digital
Gas
Oscilloscope
Analyzer
Ethernet
network
Measurement
n
Programs
gI
Labjack
An
ut
Amplifier
gO
Computer
alo
Digital Out
An
Fixed
Voltages
Defining AOM Freq.
31
3 Experimental set-up
Photodiodes
The CCD finger cameras are both continuously and automatically adjusting their exposure times
according to the amount of incoming light. The cameras proved to be very useful for qualitative
adjustments of e.g. the MOT beam geometry, but on the other hand they cannot be used for
qualitative measurements of the MOT fluorescence due to the auto-exposure feature. Therefore we
additionally installed two DC photodiodes (type OSD15-E) with integrated adjustable amplifiers:
One was used for a beam split off the from the 3D MOT camera optics by a 30:70 beam splitting
plate to measure the MOT fluorescence and the other one was installed in a removable set-up to
measure fluorescence from the atomic beam in the front part of the 2D MOT chamber.
Digital oscilloscope
A digital oscilloscope from Tektronics which can be connected to the Ethernet network is used
to capture signals from various sources directly to a computer.
Pump pressure
The ion pump can detect the pressure in the 3D MOT chamber by measuring the ion current.
Although the pressure in the center of the 3D MOT chamber is expected to be somewhat higher
than the displayed value because the gauge is more than 30 cm away from the trap center, the
value certainly gives a good indication of the pressure in the area of the 3D MOT.
Heating tapes were installed around the rear end of the 2D MOT and the bellow containing the
potassium ampul. When installing the tapes a PT100 resistance temperature detector (RTD)
was positioned near the bellow which allows to measure its temperature with a simple ohmmeter.
While heating the potassium source the RGA mounted close to the rear end of the 2D MOT was
used to measure the 39 K levels and those of other elements. Details on the RGA can be found in
chapter 3.1.3 on page 18.
In order to study the 3D MOT one needs to be able to change the detuning of the TL with respect
to the cooler transition. The simplest way to achieve this would be to choose another lock point
in the FM spectrum of the TL. This however, must be done manually and separately for each
detuning. Another way is to change the frequency of the AOM which shifts the part of the TL
light that is used for the spectroscopy.
While changing the AOM frequency for example to a higher value, the lock electronics will always
try to keep the TL stabilized to the same point on the crossover edge in the spectroscopy signal
(which will be shifted further up in frequency by increasing the AOM frequency) and thus it has
to shift the TL frequency to a correspondingly lower value. Therefore a higher AOM frequency
results in an additional red detuning of the 3D MOT light.
Because the AOM is only built for a certain frequency range and also because the angle of the first
order beam coming out of the AOM changes with the frequency, it can only be tuned by about
5 MHz in either direction, resulting in a net shift of 10 MHz in each direction in the double-pass
configuration.
32
3.4 Measurement set-up
The AOM driver includes a voltage controlled oscillator (VCO) to generate the oscillating signal
that is fed into the AOM. This means the AOM frequency can easily be controlled by an external
voltage connected to the AOM driver.
To perform fast switches (< 3 ms) between different fixed AOM frequencies an AOM preheat
switch is used. This device contains several analog switches that are controllable by TTL signals.
The outputs of the preheat switch are all connected together and fed into the VCO of the AOM
driver while its inputs are connected to one variable voltage source (connected to the output when
all three switches are off) and three other fixed voltage sources each defining a different fixed TL
detuning.
Compared to the TL detuning, the seed and push detunings are much easier to control because
offset locks are used to stabilizes these lasers. As there was no need to switch these as fast as
the TL, the VCO inputs of the two offset locks were simply connected to variable voltage sources
which allowed for continuous tuning of the seed and push detuning over a range of up to 100 MHz.
A shutter is installed to turn the push beam on and off by setting a TTL line to high or low
respectively.
A device containing eight relais controllable from a computer by a RS 232 serial line was installed
to switch high currents. Cable connections to independently invert the currents running through
the 2D and 3D coils were set up. By doing so, σ+ becomes σ− light and vice versa and thus the
MOT will not work anymore. The absolute values of the currents and the magnetic gradients
remain the same and therefore as few parameters as needed are changed while still extinguishing
the MOT.
This feature is used to measure the background light levels on the 3D MOT photodiode as well
as to capture loading curves of the 3D MOT.
To control the abovementioned devices with digital or analog inputs and to measure the 3D
MOT photodiode voltages a simple but versatile analog and digital input/output device with an
USB interface of the type „Labjack U12“ (displayed in figure 3.17) is used. This device has the
following features:
• 8 single-ended, 4 differential 12-bit analog inputs with ±10 volt input range and pro-
grammable gain amplifiers (PGAs) with gains of up to 20
• an analog data acquisition rate of up to 8 kSamples/s
• two 10-bit analog outputs (0-5 V)
• 20 digital I/O channels (TTL) with data rates of up to 50 Hz
Software drivers and an application programming interface was provided by the supplier of the
device but these had to be rewritten from the ground due to their poor quality. The new driver was
written in the scripting language Python [34] using libusb [35] and the corresponding interface
to Python [36].
33
3 Experimental set-up
Lock-in amplifier
To measure the atomic beam fluorescence in the 2D MOT chamber during the first part of this
work, a so called lock-in amplifier was used as a sensitive device to measure signals under severe
noise conditions - that is, where the noise-to-signal ratio can be as high as 130 dB.
A lock-in amplifier is a measurement device which can be understood as a synchronous demodu-
lator or phase-sensitive detector. It is basically a specialized AC voltmeter that uses synchronous
demodulation to measure signal strength or phase and can be used wherever the signal of interest
can be synchronized with or derived from a suitable reference signal. The output of the lock-in
amplifier are the magnitude and the phase of the signal.
Details on how the lock-in amplifier was used can be found in chapter 3.4.2.
Computer software
To control all the devices connected to the computer an extensive library of software components
was written in the scripting language Ruby [37] which in turn has provided the base to implement
different measurement programs on top of them.
A short overview of the program library follows:
labjack_server.py A small TCP/IP server written in Python which offers a simple interface to
the Labjack device for multiple clients.
pythonlabjack_client.rb A TCP/IP client which connects to the Labjack server.
display.rb A class which provides a live data display in large letters.
lockin.rb A class to control all aspects of the Lock-in amplifier.
rga.rb A class to control the residual gas analyzer.
relais.rb A class to control the relais box.
measurement.rb The core class to measure the 3D MOT fluorescence and loading curves while
varying different parameters. Also contains functions to continuously monitor and plot the
MOT fluorescence.
m_manual.rb Provides functions to perform a series of 3D MOT loading curve measurements
while manually varying a parameter such as the 2D MOT length.
loading_fit.rb A module which takes measured MOT loading and decay curves and automatically
fits them with exponential functions.
34
3.4 Measurement set-up
require 'measurement' # load the measurement library and all its dependencies
while d_2d < -3 + delta/2 # loop over the 2D MOT detuning in the range [-7..-3]
$m.detuning_2d = d_2d # set the next 2D MOT detuning
puts d_2d # print the 2D MOT detuning to indicate the measurement progress
d_3d = -7
while d_3d <= -3 + delta/2 # loop over the 3D MOT detuning in the range [-7..-3]
$m.detuning_3d = d_3d # set the next 3D MOT detuning
pnumber.rb This is a class which provides functions to convert photodiode voltages to 3D MOT
atoms numbers.
Using this library, the composition of different measurement programs was a relatively straight-
forward task resulting in clean, easily understandable and well maintainable program code. For
illustration purposes, the code automatically performing a two-dimensional scan over both the
3D MOT and 2D MOT detuning is listed in figure 3.18. The program results in 400 single
measurements and a total measurement time of over half an hour .
35
3 Experimental set-up
Therefore a lock-in technique was used. For this purpose, the TL light was used for the probe
beam and a frequency generator running at about 16 kHz continuously switched the repumper
EOM on and off. This caused the fluorescence of the atomic beam to get bigger (when the EOM is
turned on) and smaller (EOM off). Using both the reference signal from the frequency generator
and the photodiode signal the lock-in amplifier was now able to extract a signal proportional to
the atomic beam fluorescence.
The qualitative results obtained from measurements with this set-up are discussed in chapter 4.2
on the next page.
As soon as the atomic beam emanating from the 2D MOT could be captured in the 3D MOT, a
photodiode measuring its fluorescence was installed additionally to the camera which was already
in place to see the 3D MOT at all.
The amplification of the photodiode was tuned such way that the difference in the photodiode
voltage with and without a loaded 3D MOT was several 100 mV. This voltage difference can be
well resolved by the 12-bit analog-to-digital converter (ADC) of the Labjack. Also a bandpass
filter for 767 nm was mounted in front of the photodiode to eliminate influences from room light.
After that the set-up of the photodiode remained the same for all further measurements and its
signal was used as the main data source for these measurements.
Details on the data processing and the results of the 3D MOT measurements can be found in
chapter 4.3.2 on page 40.
36
4 Measurements
In this chapter, the different measurements that have been performed are presented and their
results are discussed and compared to related work.
First, in section 4.1 the calibration of the laser frequencies is explained, which is a crucial mea-
surement for obtaining correct results when probing different laser detunings. Then the insights
gained from the atomic beam fluorescence set-up are presented in section 4.2.
The parameters of the 3D MOT and the measurement procedures that were established are
discussed in sections 4.3.1 and 4.3.2 of this chapter, while the results from varying 2D MOT and
push beam parameters are presented in chapter 4.4 and 4.5 respectively.
37
4 Measurements
Figure 4.1: Atomic beam fluorescence when continuously inverting the magnetic field
fluorescence / a.u.
0 2 4 6 8 10
time / (coil cycles ≈ 300 ms)
while continuously inverting the magnetic field could be taken as an indication for the flux or size
of the atomic beam.
Using this method we could confirm that the signal was in fact coming from the atomic beam,
because it immediately disappeared when e.g. the 2D MOT detuning was changed to a value
where no MOT could be seen, when one of the MOT beams was blocked or when the probe beam
alignment was changed so that it did not hit the atomic beam anymore.
Additionally we estimate the atomic beam width 5 mm away from the end of the atomic cloud to
be a few millimeters. Also, the push beam used on the cooler frequency seemed to enhance the
flux. The flux was on the other hand not very sensitive to maladjustments of the MOT mirrors
or of the movable telescope lens to compensate for the light force imbalance due to the MOT
windows. The flux was also insensitive to unbalanced light intensities in the horizontal and the
vertical MOT beam pair.
Unfortunately the quality in general and especially the reproducibility of the data obtained from
the temporary atomic beam fluorescence set-up were very bad due to unknown drifts. Therefore
no quantitative results can be given here.
4.3 3D MOT
The 3D MOT in our experimental set-up was only used as a detection device for the atomic beam
produced in the 2D MOT. Therefore we carried out (after we managed to capture the beam
for the first time) only some initial adjustments and characterization measurements to establish
reliable measurement procedures for the atomic beam.
After this initial adjustments, all 3D MOT parameters were kept constant at the values given in
table 4.1 while performing parameter variations on the 2D MOT and the push beam. Loading
of the 3D MOT on its own (without operating the 2D MOT) was never observed due to the
low partial pressure of 39 K in the 3D MOT chamber, not even when applying UV light to the
chamber2 or when heating the 2D MOT source to temperatures as high as 85 °C.
The adjustment of the 3D MOT after the atomic beam could be captured in it for the first time
was mainly based on the images obtained by the 3D MOT CCD camera and the continuous
monitoring of the photodiode signal. For illustration purposes, two images of the 3D MOT are
shown in figure 4.2.
Typical diameters of the measured MOTs ranged from 1 mm to about 7 mm with atom numbers
ranging from about (0.3 ± 0.1) × 108 atoms up to an absolute maximum of (1.0 ± 0.3) × 109 atoms.
The atom numbers were calculated using the method explained in section 4.3.2.
2 Applying UV light is reported to remove alkali metals that are sticking to the chamber walls [38].
38
4.3 3D MOT
Figure 4.2: CCD fluorescence images of 3D MOTs. To reduce stray light reaching the CCD
chip, the camera aperture had to be closed and is therefore visible in the obtained images. The
magnification factor of the camera system was determined on a test set-up away from the MOT
chamber. Comparing the two pictures, one can also easily observe the auto-exposure feature of
the camera: even though the second MOT is in fact much brighter than the first (as measured
by the photodiode) it actually does not appear so in the image.
camera aperture
MOT
4 mm 4 mm
(a) a typical 3D MOT: (0.15 ± 0.04) × 109 atoms (b) a big 3D MOT: (0.9 ± 0.3) × 109 atoms
Table 4.1: 3D MOT parameters. Note that for 39 K the frequency of the repumper transition
is (461.7 − (14.3 + 6.7)) MHz = 440.7 MHz higher than that of the cooler transition. So e.g.
a repumper EOM frequency of 462 MHz and a cooler detuning of −4.4Γ results in an effective
detuning of the repumper of (−4.4 · 6.1 + 462 − 440.7) MHz = −5.5 MHz = −0.9Γ.
39
4 Measurements
Figure 4.3: Atom numbers and loading rates for different 3D MOT detunings. For the 2D MOT
the default parameters given in table 4.2 on page 44 were used.
0.1
0.1
0.0 0.0
–6 –5 –4
3D MOT detuning / Γ
4.3.1 Parameters
Laser detuning
The 3D MOT fluorescence and loading curves were measured for different MOT detunings. Using
this data atom numbers and initial loading rates (that is the loading rate just after inverting the
3D MOT magnetic field back to its normal state) have been calculated as discussed in section
4.3.2. The results can be seen in figure 4.3.
The obtained results for the atom numbers are similar to those given in [19], although our data
is much flatter in the region from −5.5Γ to −4.3Γ. Notable is the abrupt drop of the atom
number and loading rate at about −4Γ. This might be due to the fact that at this point the
repumper is on resonance and it becomes blue detuned if the cooler detuning further approaches
0Γ, thereby starting to repel atoms from the trap center instead of pushing them towards it.
Another interesting feature is the small dip in the loading rate at −4.3Γ.
The main goal of this work is to characterize the atomic beam coming from the 2D MOT and
its dependencies on different parameters. The three properties of the atomic beam that are of
interest are the flux, its velocity profile and also its divergence. Although the velocity profile
and the divergence of the beam determine how well the beam can be recaptured and cooled by
a 3D MOT, they were not further investigated because previous work had already shown that
39
K 2D MOTs using a push beam produce a narrow peak (FWHM ≈ 4.5 m/s) centered around
33 m/s in the velocity profile and a beam divergence of 50 mrad which is well suited for a fast
and reliable loading of a 3D MOT [13]. Therefore the main focus was put on the measurement
of the atomic flux.
Using a photodiode, two parameters can be easily measured for a specific 2D MOT configuration:
the final 3D MOT size and its initial loading rate. Both of them could act as measures for the
atomic flux. If we go back and look at the calculations on MOT dynamics in chapter 2.5.2 on
page 10, we can see that while the initial loading rate is directly proportional to the atomic flux,
the total atom number is also dependent on the parameter γ which is the inverse of the 3D MOT
lifetime. But the lifetime of the 3D MOT depends critically on its detuning, power and the
background pressure. Therefore we decided to take the initial loading rate as the main indicator
40
4.3 3D MOT
for the atomic flux while always measuring the total atom number as well. In some cases it was
not possible to measure loading curves due to the longer measurement times required and as a
consequence only the atom number was measured.
In order to extract atom numbers and fluxes from the fluorescence signal of the 3D MOT one has
to first calculate the scattering rate γ 0 of a single atom in the MOT. By dividing the measured
fluorescence light power Pphotodiode by the scattering rate and the energy Eγ = ~ω0 of the scattered
photons one can obtain the atom number.
In detail, the used formula looks as follows:
Pphotodiode
N= (4.1)
γ 0 · Eγ · σ · t
with
Γ
γ 0 = C12 · s · 2
(4.2)
2δ 2
1 + C22 · s +
Γ
being a modified scattering rate [39]. δ is the detuning of the probe laser (note that both δ and Γ
are angular frequencies 2π · Hz in such calculations) and t = 0.9 the transmission of the windows
and optics leading to the photodiode.
A
σ= = 0.0025 (4.3)
4πR2
is the fraction of the solid angle captured by the photodiode. It is calculated by dividing the area
A of the lens that collects the fluorescence by the surface area of a sphere. R is the distance of
the collecting element from the MOT center. In our case A = 2.01 × 10−4 m2 and R = 0.08 m.
The coefficients C12 = C22 = 0.73 are averaged Clebsch-Gordan coefficients correcting the scatter-
ing rate γ for the fact that not all six MOT beams can drive each atom’s transition at the same
time at the full efficiency due to selection rules. They are chosen according to ref. [39]. Although
these values were measured for Cs MOTs, we assume that in first order they should be the same
for 39 K.
I
The intensity used to deduce the saturation parameter s = is calculated the following way:
Isat
2 Ptot
I= · (4.4)
πwx wy 2
Ptot is the total power of all six MOT beams. For potassium, we then have to take Ptot 2 as the
effective power of the MOT light, because only either the cooler or the repumper transition of an
atom can be driven at any given time. wx and wy are the waists of the MOT beams. Remember
2P
that for a Gaussian beam the peak intensity I is given by: I = .
πwx wy
As the amplified photodiode output is just a voltage rather than the power of the incoming
light, the photodiode voltage had to be calibrated with a power meter prior to its first use.
The calibration with MOTs with different sizes being turned on and off to compensate for stray
light led to the result that a voltage difference of one volt on the photodiode corresponds to
(307 ± 5) nW of fluorescence power.
41
4 Measurements
Figure 4.4: MOT fluorescence during a fast MOT laser detuning switch to δ 0 = −3.3Γ. The switch
is performed at the time t = 0 ms. For t < 0 ms one can see the original MOT fluorescence for
the different detunings. Note that for a detuning of −6.5Γ there is no MOT at all. One also sees
that switching the laser frequency takes up to 2 ms depending on the size of the jump. When
going to smaller detunings the fluorescence initially increases due to the enhanced scattering rate
before it decreases because the atoms get lost due to the fact that they are not trapped anymore
at −3.3Γ.
error signals
laser lock
fluorescence
loaded at -4.5Γ
loaded at -5.5Γ
MOT
loaded at -6.5Γ
0 2 4 6 8
time / ms
To test the accuracy of our atom number calculations using the abovementioned method, we did
the following: after loading the 3D MOT we instantly switched the laser of the 3D MOT to a
fixed detuning δ 0 to measure the fluorescence. This would effectively remove the dependence from
the laser detuning from the atom number calculation making it just a matter of multiplying the
photodiode signal with a constant.
To get a good fluorescence signal choosing a detuning of δ 0 = −3.3Γ close to resonance seemed
a sensible choice. However, at this detuning the atoms are not trapped anymore. Therefore
the fluorescence imaging has to be done immediately after the laser frequency switch. The laser
frequency switch, however, cannot be performed instantaneously either. To test the feasibility of
the method, MOT loadings at different detunings followed by this imaging were performed while
measuring the MOT fluorescence and the laser lock error signal3 . The results are shown in figure
4.4 (including explanations in the caption).
From this measurement we concluded that while the atoms almost immediately start to escape
from the imaged region as soon as (or even before) the laser reaches the new detuning δ 0 , the
proposed method should still work if we capture the fluorescence curves with a computer and
extract the height of the fluorescence peak.
Following this first measurement, the scheme was implemented as described above and measure-
ments with different fixed imaging detunings as well as without any imaging were made and the
obtained atom numbers are shown in figure 4.5.
First of all, comparing the shapes of the atom number curves to the photodiode signal given in
arbitrary units, one can see that the brightest MOT does not necessarily need to be the biggest
one in terms of atom numbers. Interestingly, the curve with an imaging at −4.5Γ is in very good
agreement with the curve obtained without any imaging (fluorescence measured at the detuning
where to MOT is loaded), while for the two other imaging detunings the obtained atom numbers
are significantly lower and their maxima differ in position. We assume that these lower atom
numbers are partly due to atoms already escaping from the trap during the switch to the imaging
3 that is the deviation of the actual laser frequency from the set point
42
4.3 3D MOT
imaging at -6Γ
0.3
imaging at -4.5Γ
atom number / 109
imaging at MOT
0.1 detuning
photodiode signal at MOT
detuning for comparison (a.u.)
0.0
–6 –5 –4
3D MOT detuning / Γ
Figure 4.6: Typical loading curves of the 3D MOT including fits both with and without push
beam
0.9
with push
atom number / 109
frequency. We conclude that just taking the MOT fluorescence at the detuning used for loading
the trap to calculate the atom number is actually the most reliable way. All atom numbers
given in this work are therefore calculated this way. To be on the safe side, we still attribute
an estimated error of ±30% to all atom number calculations which makes the atom numbers
obtained from the imaging at −6Γ and −3Γ lie within the error bounds.
Loading curves for the 3D MOT were obtained by first inverting the 3D MOT coil current to
extinguish the MOT, then setting it back to its normal state and waiting for several seconds for
the MOT to load while continuously measuring the MOT fluorescence. The MOT lifetime was
measured by letting the MOT load and then turning off the 2D MOT by inverting its coils and
waiting until the 3D MOT was completely emptied.
Figure 4.6 shows two typical loading curves of the 3D MOT, one without using a push beam and
the other one with a push beam tuned to the cooler resonance (push beam power: 0.75 mW).
According to the calculations in chapter 2.5.2 on page 10 fits to the loading and decay curves
43
4 Measurements
N (t) = N0 1 − e−(χφB +γ)t (4.5)
1
for the decay curve. The results of the fits for the curves shown are given in the plot. T =
χφB + γ
1
is the loading time, d = N0 (χφB + γ) the initial loading rate, and τ = the lifetime of the MOT.
γ
Fitting worked usually quite well and there was no need to include quadratic terms in N in the
MOT rate equation. Only the loading curves of very large MOTs (> 0.8×109 atoms) with loading
times of a few 100 ms or less could not be reliably fitted with exponential functions. Instead,
linear fits were applied to the first 5 to 10 points4 of the loading curves to obtain the initial
loading rate.
The errors of the fitted parameters are estimated to be less than 10%.
4.4 2D MOT
If not otherwise stated, the default parameters given in table 4.2 are used for the following
measurements on the 2D MOT. The default parameters for the push beam can be found in
table 4.4 on page 53.
As a reference, a CCD fluorescence image of the 2D MOT in axial view is given in figure 4.7.
44
4.4 2D MOT
Figure 4.7: CCD fluorescence image of a 2D MOT. In the center the entry of the differential
pumping tube with a diameter of 2 mm can be seen while the outer circle is the diameter of
the differential pumping tube inset piece. The picture was taken with a 39 K partial pressure
of 1.5 × 10−7 mbar after the temperature dependence measurement discussed in section 4.4.8.
Therefore the MOT fluorescence is maximized. Additionally, because the EOM frequency had
been shifted to 450 MHz, the MOT was also visible for the chosen detuning of only −3Γ. To
make the MOT clearly distinguishable, the offset coils centering the MOT in front the tube entry
have been turned off. The total MOT light power is 180 mW.
differential pumping
tube entry
2D MOT
2 mm
Figure 4.8: Atom numbers and loading rates for different 2D MOT detunings
0.5
without push beam
without push beam
0.25
with push beam
0.4
0.20
atom number / 109
0.3
0.15
0.2
0.10
0.1
0.05
0.00 0.0
–7.0 –6.5 –6.0 –5.5 –5.0 –4.5 –4.0 –3.5 –3.0 –7.0 –6.5 –6.0 –5.5 –5.0 –4.5 –4.0 –3.5 –3.0
2D MOT detuning / Γ 2D MOT detuning / Γ
(a) atom numbers (b) loading rates
45
4 Measurements
–3 0.04 –3
0.3
0.03 –4
0.2
0.02
–5 –5
0.1
0.01
–6 –6
0.00 0.0
–7 –7
–7 –6 –5 –4 –7 –6 –5 –4
3D MOT detuning / Γ 3D MOT detuning / Γ
(a) without push beam (b) with push beam
0Γ). The results can be seen in figure 4.8. All four curves peak at about −5Γ (with the peak
without push beam being slightly closer to resonance), but compared to the 3D MOT detuning
scan, the slopes of the red and blue edge are about the same, i.e. they show not such a sharp
edge at the blue side as the 3D MOT scan did. But according to the explanation given for this
sharp edge in chapter 4.3.1, it was to be expected to appear at about −4Γ where there is almost
no capturable atomic beam anyway. Another important fact to notice is, that while the push
beam enhances the atom number by about a factor of 4, the loading rate is enhanced much more,
namely by a factor of almost 10.
A two dimensional scan over both the 3D and 2D MOT detuning was performed to see whether
one needs different 3D MOT detunings to most efficiently capture the atomic beam produced at
different 2D MOT detunings. The results are shown in figure 4.9 displaying one scan without
and another one with the push beam.
Most importantly, the capture efficiency does not seem to depend on the 2D MOT detuning.
This can be deduced from the shape of the contours; they are apparently just the convolution
of the one-dimensional atom number curves for the 3D and the 2D MOT measured before. The
contours show no diagonally running features which indicates that the 2D MOT does only work
efficiently in one single range around a detuning of about −5Γ.
Again one can see that the 2D MOT works best at a slightly bluer (−4.6Γ) detuning without
push beam than with it (−5Γ). Also, it is easily observed that the push beam makes the range
of detunings where the 2D MOT works efficiently much broader. The sharp edge in the atom
number when going to 3D MOT detunings of about −4.5 to − 4Γ can be seen as well.
Since the EOM modulates both blue and red sidebands onto the cooler carrier, we should be able
to see a MOT while locking the seed laser to the repumper so that the carrier of the laser will
be the repumper and the red sideband the cooler. That this inverted scheme indeed works can
be seen in figure 4.10 on the facing page (the normal scheme is on the left and the inverted on
the right, both using the push beam). Additionally we also scanned the 2D MOT laser over a
much broader range of detunings and to our great amazement we discovered a small atomic beam
generated by a blue-detuned 2D MOT. This beam could be observed in both the normal and the
inverted scheme.
46
4.4 2D MOT
0.2 0.2
0.1 0.1
0.0 0.0
–7 –6 –5 –4 –3 –2 –1 0 1 2 –3 –2 –1 0 1 2 3 4 5 6
2D MOT detuning / Γ 2D MOT detuning / Γ
(a) carrier on the cooling transition (b) carrier on the repumper transition
0.3
initial loading rate 2.5
loading time / s
loading time
0.2 2.0
1.5
0.1
1.0
0.0
0 20 40 60 80 100 120 140 160 180
total 2D MOT power / mW
Note that to observe an atomic beam in the inverted scheme, the laser had to be tuned very
closely to resonance so that the cooler which is in this case 462 MHz red-shifted has the same
effective detuning of about −5Γ. Additionally, the atomic beam is much smaller in the inverted
scheme because in this case the repumper : cooler ratio is also inverted (therefore being 3:2) and
as a consequence less laser power is available for the cooler while some of the laser power is wasted
for repumping at a rate which is too high. Please refer to section 4.4.3 for details on this.
In figure 4.11 the initial loading rate and loading times are shown for increasing total power in all
2D MOT beams. The loading rate increases almost linearly and while at some point a saturation
is to be expected, this cannot be observed at the available powers.
47
4 Measurements
Figure 4.12: Initial 3D MOT loading rate for varying 2D MOT repumper : cooler ratios
0.1
0.0
0.0 0.5 1.0 1.5
ratio repumper to cooling power
48
4.4 2D MOT
Figure 4.13: Magnetic field gradient dependence with a 2D MOT detuning of −4.7Γ
loading time / s
0.2
initial loading rate 1.5
1.0
0.0
0 2 4 6 8 10 12 14
2D magnetic field gradient / (G/cm)
The dependence of the loading rate on the length of the 2D MOT was studied and the correspond-
ing measurements are presented in figure 4.14. To change the cooling volume, black cardboard
was used to cover parts of the MOT mirrors starting from both sides and going towards the
center. While for the case without push beam the increase with the cooling volume length is only
very moderate and linear, with the push beam it is almost quadratic.
Additionally the effect of different cooling volume lengths for the horizontal and the vertical MOT
beams was investigated. Small differences of up to about 5 millimeters had no measurable effect
while bigger ones reduced the atomic beam flux rather fast.
0.25
with push beam
without push beam
0.20
initial loading rate / (109 atoms / s)
0.15
0.10
0.05
0.00
0 4 8
cooling volume length / cm
49
4 Measurements
1.5
1.0
0.5
0.0
39
4.4.8 K Partial pressure
After all other measurements had been performed we started to continuously heat the area around
the rear end of the 2D MOT chamber and especially the bellow containing the potassium ampul.
During the heating process we monitored the following parameters: the temperature of the bellow,
the pressure in the 3D MOT, the partial pressure of 39 K in the 2D MOT, as well as the 3D MOT
loading rate and lifetime. The results can be seen in figures 4.16 to 4.18. In contrast to all other
measurements, these have been performed with different 2D MOT parameters: the repumper
EOM was set to 450 MHz and the 2D MOT detuning to −2.6Γ.
The bellow temperature was increased from 40 °C to a maximum of 85 °C within ten hours. As a
reference, the partial pressure of 39 K and the 3D MOT pressure for these temperatures are given
in figure 4.16. As expected, both pressures increased continuously with the exception of the last
3D MOT pressure point which was measured three days after the temperature of 85 °C had been
established.
In figure 4.17 the loading rate is plotted against the increasing 39 K partial pressure. Between
1.5 × 10−7 mbar and 1.0 × 10−6 mbar data was taken, but it was not usable because some drift of
50
4.4 2D MOT
3.10–9
8.10–7
pressure / mbar
2.10–9
1.10–9
0
40 °C 50 °C 60 °C 70 °C 80 °C 90 °C 40 °C 50 °C 60 °C 70 °C 80 °C 90 °C
potassium source temperature potassium source temperature
(a) partial pressure of 39 K at the rear end of the 2D (b) pressure at the 3D MOT pump
MOT
0
≈
51
4 Measurements
3D MOT lifetime / s
without push beam
1.6
1.2
0.8
5.0.10–10 1.5.10–9 2.5.10–9
pressure at the 3D MOT pump / mbar
the offset coils had occurred during this time. Therefore we can estimate that there is a maximal
loading rate for a partial pressure somewhere between 1.5 × 10−7 mbar and 1.0 × 10−6 mbar,
which is consistent with Catani et al. who measured the maximum flux at a total pressure of
2.1 × 10−7 mbar. According to our RGA measurements, we can assume that besides 39 K only H2
contributes to the total pressure to a significant amount (with about the same partial pressure
as 39 K), and therefore we can estimate that our flux is maximized for a total pressure between
3 × 10−7 mbar and 2 × 10−6 mbar. The decrease in flux after reaching the maximum can be
explained by the fact that the atomic flux starts to be depleted by collisional effects.
The effect of the increasing pressure in the 3D MOT on the MOT lifetime can be nicely observed
in figure 4.18. Interestingly, for low pressures, the MOT lifetime seems to depend on whether the
push beam was used to load it. But we think that this is rather due to a large difference in the
MOT size while for higher pressures the difference is not so pronounced.
52
4.5 Push beam
respective 3D MOT atom numbers were (1.0 ± 0.3) × 109 and (0.8 ± 0.3) × 109 .
From the initial loading rate d we can infer a lower bound for the total atomic flux φB , since (as
deduced in section 2.5.2) d = αφB with α ≤ 1 being the probability that an atom from the atomic
beam is captured by the MOT. Catani et al. [13], who measured the total atomic flux directly
by its fluorescence, give a maximum flux of 1.0 × 1011 atoms/s at a pressure of 2.1 × 10−7 mbar,
which is 15 times higher than our loading rates. Unfortunately they do not give any loading rates
and therefore a direct comparison is not possible since we expect that they can only capture a
fraction of the atoms from their atomic beam.
A two-dimensional scan of both the 2D MOT detuning and the push beam detuning was performed
and the results are shown in figure 4.19. One can see that both on the cooler and on the repumper
transition, there are two or even three push detuning ranges where the push beam indeed increases
the atom number. But to our amazement, in the areas in between, the push beam not only had
53
4 Measurements
Figure 4.19: Varying push beam and 2D MOT detunings with a push beam power of 0.75 mW
8 0.30 8 0.30
4 4
2 0.10 2 0.10
0 0
–2 0.00 –2 0.00
–4 –4
no effect but completely or almost extinguished the atomic flux. Additionally, on the cooler
transition for negative push detunings the effect of the push beam depends strongly on the 2D
MOT detuning, while this is not the case on the repumper transition.
To explain the observed behavior, remember the interaction of laser light with an atom: if it is
red detuned, it slows down atoms moving towards it, if it is on resonance, it accelerates atoms at
rest in the direction of the beam, and if it is blue detuned it further accelerates atoms which are
already moving in the direction of the beam. The push beam can theoretically enhance the flux
in all three cases: for a red detuning it could act as a plug preventing the atoms from escaping
from the atomic cloud on the rear side. If the beam is on resonance, it could slightly accelerate
atoms at rest out of the cloud. If the beam is blue detuned it might accelerate the atoms so that
they reach the 3D MOT without hitting the aperture or the pumping tube due to their inherent
divergence.
Currently it cannot be explained why these explanations only seem to be applicable for some
detunings and also why this depends on the 2D MOT detuning. We expect the behavior to
be highly dependent on the velocity profile produced by the 2D MOT and also on the capture
efficiency for these velocities in the 3D MOT.
Also note that due to the fact that our push beam contains either light tuned to the cooler or the
repumper transition, it will only have an effect on the atoms while they still see the main MOT
light; outside of it they will be quickly pumped to a dark state.
Comparing our results to those obtained by Catani et al. [13], who only give results for negative
detunings of the push beam on the repumper transition with much higher powers (6 mW vs.
0.75 mW), one can see that these are qualitatively not so different. Both data show a flux curve
peaking at about −5Γ (Catani et al.: −5.2Γ). Nevertheless, at a power of 0.75 mW, pushing with
a detuning of 0Γ was much more efficient for us.
54
4.5 Push beam
the atomic beam is extinguished gets bigger). For powers above 1 mW one can also observe on
the 3D MOT camera that for detunings around 5Γ on the cooler not only the atomic beam is
extinguished but the 3D MOT is immediately extinguished as well by the push beam itself. It is
also interesting to see that while the qualitative curve shapes are the same for the two 2D MOT
detunings, their relative heights can differ by more than a factor of two.
Compared with the laser power scan at −5.2Γ on the repumper transition by Catani et al. [13],
we find again that their and our results are qualitatively the same: the atomic flux increases with
higher push beam powers in both measurements.
4.5.3 Polarization
All our measurements were performed with circularly polarized push beam light. When going
to linearly or elliptically polarized light, the push beam efficiency decreased no more than 10%,
while just changing the handedness of the circularly polarized light did not have any effect. This
is expected because the magnetic field strength is almost zero along the MOT axis and therefore
its direction is not well defined there.
To achieve a maximal atomic flux, the push beam had to be exactly on the MOT axis and also
had to go centered through the differential pumping tube entry. Here only the direction, i.e.
whether the beam was really going straight through the MOT and was hitting the tube entry in
the center, was crucial. Slightly changing the overlap of the push beam with the atomic cloud
did not have such a big influence.
Using a movable lens we could also change the beam diameter and focus. The default setting here
was to focus the beam slightly towards the tube entry. However, this barely had a measurable
influence on the atomic flux.
55
4 Measurements
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
push detuning / Γ push detuning / Γ
56
4.5 Push beam
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
0.1 0.1
0.0 0.0
–10 –5 0 5 10 15 –10 –5 0 5 10 15
push detuning / Γ push detuning / Γ
57
4 Measurements
58
5 Conclusions and Outlook
In the course of this diploma work, an already existing test set-up for a 2D magneto-optical trap
for potassium was taken into operation. All the lasers and the optics were readjusted and an
atomic cloud was observed, as already reported in [27]. In a next step, dedicated optics were
set-up for the detection of the atomic beam emerging from the cloud. To sensitively detect the
beam, a lock-in amplifier was used to filter the photodiode signal. Using this technique, we were
able to clearly identify an atomic beam and perform an initial characterization.
For quantitative measurements the attached 3D MOT was used as a detector for the atomic flux,
requiring a complete readjustment of the 3D MOT optics. The stability of the laser locks and the
MOT beam polarization were significantly improved. We were then able to capture the atomic
beam in the 3D MOT for the first time in this set-up.
Subsequently, both hard- and software to control the laser detunings, shutters and magnetic coils
was implemented and set up. This allowed for sophisticated computer controlled measurements
of the MOT fluorescence including scans over multi-dimensional parameter spaces and automatic
fitting of the results. The obtained MOT loading rates were then used to characterize the atomic
beam.
It was found that already without a push beam, atomic fluxes of more than 109 atoms/s can
be achieved by choosing an appropriate 2D MOT detuning and radial position in front of the
differential pumping tube and above all by establishing a relatively high 39 K partial pressure of
around 10−7 mbar. Due to the differential pumping tube the 3D MOT can at the same time still
be operated at a low pressure of 10−9 mbar.
When applying a push beam with a moderately low power (0.75 mW) the atomic flux increases
by a factor of ten under certain circumstances, leading to a maximum flux of more than 6 ×
109 atoms/s. In contrast to what was found in other work [13], the push beam works almost
equally well when tuned to either the cooler or repumper atomic transition. Using this well
optimized atomic beam, we were able to load our 3D MOT to a final atom number exceeding
109 atoms in less than 100 ms. It could be shown that the effect of the push beam for different
detunings changes drastically when going from low (< 1 mW) to high powers (5 mW). While for
low powers it only works on resonance, for higher powers the push beam also has a significant
effect for a blue or red detuning of −5Γ or 10Γ, respectively. The 2D MOT on the other hand
operates best at a detuning of −5Γ for a repumper frequency shift of 462 MHz1 . Surprisingly, we
also found evidence of another small 2D MOT blue-detuned to the atomic resonance.
In a next step, the 2D MOT needs to be switched to the fermionic 40 K which is the ultimately
targeted isotope - 39 K was used for test purposes due to its easier availability. Therefore a metallic
source containing enriched 40 K needs to be brought into the vacuum and two EOMs and an AOM
are to be exchanged in the laser set-up. If the 2D MOT works as well for 40 K as for 39 K one can
then move the 2D MOT over to the optical lattice experiment and connect it to its 3D MOT.
The increase in atom number due to the use of the 2D MOT will allow to go to lower relative
temperatures TTF , hopefully leading to the observation of novel phenomena in future experiments
such as antiferromagnetic ordering of a repulsive spin mixture of fermions in an optical lattice.
1 The absolute maximum value given above was found at a frequency of 450 MHz and a detuning of −2.6Γ, but
no further measurements were made at this repumper frequency shift.
59
5 Conclusions and Outlook
60
List of Figures
4.1 Atomic beam fluorescence when continuously inverting the magnetic field . . . . . 38
4.2 CCD fluorescence images of 3D MOTs. To reduce stray light reaching the CCD
chip, the camera aperture had to be closed and is therefore visible in the obtained
images. The magnification factor of the camera system was determined on a test
set-up away from the MOT chamber. Comparing the two pictures, one can also
easily observe the auto-exposure feature of the camera: even though the second
MOT is in fact much brighter than the first (as measured by the photodiode) it
actually does not appear so in the image. . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Atom numbers and loading rates for different 3D MOT detunings. For the 2D
MOT the default parameters given in table 4.2 on page 44 were used. . . . . . . . 40
61
List of Figures
4.4 MOT fluorescence during a fast MOT laser detuning switch to δ 0 = −3.3Γ. The
switch is performed at the time t = 0 ms. For t < 0 ms one can see the original
MOT fluorescence for the different detunings. Note that for a detuning of −6.5Γ
there is no MOT at all. One also sees that switching the laser frequency takes
up to 2 ms depending on the size of the jump. When going to smaller detunings
the fluorescence initially increases due to the enhanced scattering rate before it
decreases because the atoms get lost due to the fact that they are not trapped
anymore at −3.3Γ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.5 Atom number measurements for different imaging detunings . . . . . . . . . . . . . 43
4.6 Typical loading curves of the 3D MOT including fits both with and without push
beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.7 CCD fluorescence image of a 2D MOT. In the center the entry of the differential
pumping tube with a diameter of 2 mm can be seen while the outer circle is the
diameter of the differential pumping tube inset piece. The picture was taken with
a 39 K partial pressure of 1.5 × 10−7 mbar after the temperature dependence mea-
surement discussed in section 4.4.8. Therefore the MOT fluorescence is maximized.
Additionally, because the EOM frequency had been shifted to 450 MHz, the MOT
was also visible for the chosen detuning of only −3Γ. To make the MOT clearly
distinguishable, the offset coils centering the MOT in front the tube entry have
been turned off. The total MOT light power is 180 mW. . . . . . . . . . . . . . . . 45
4.8 Atom numbers and loading rates for different 2D MOT detunings . . . . . . . . . . 45
4.9 Atom numbers for different 3D and 2D MOT detunings . . . . . . . . . . . . . . . 46
4.10 2D MOT when inverting cooler and repumper . . . . . . . . . . . . . . . . . . . . . 47
4.11 Total 2D MOT laser power dependence . . . . . . . . . . . . . . . . . . . . . . . . 47
4.12 Initial 3D MOT loading rate for varying 2D MOT repumper : cooler ratios . . . . 48
4.13 Magnetic field gradient dependence with a 2D MOT detuning of −4.7Γ . . . . . . 49
4.14 Cooling volume dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.15 2D MOT position dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.16 Pressure for different potassium source temperatures . . . . . . . . . . . . . . . . . 51
4.17 Loading rates for different 39 K partial pressures . . . . . . . . . . . . . . . . . . . . 51
4.18 3D MOT lifetime for different pressures . . . . . . . . . . . . . . . . . . . . . . . . 52
4.19 Varying push beam and 2D MOT detunings with a push beam power of 0.75 mW . 54
4.20 Push beam on the cooler transition . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.21 Push beam on the repumper transition . . . . . . . . . . . . . . . . . . . . . . . . . 57
62
List of Tables
4.1 3D MOT parameters. Note that for 39 K the frequency of the repumper transition
is (461.7−(14.3+6.7)) MHz = 440.7 MHz higher than that of the cooler transition.
So e.g. a repumper EOM frequency of 462 MHz and a cooler detuning of −4.4Γ
results in an effective detuning of the repumper of (−4.4 · 6.1 + 462 − 440.7) MHz =
−5.5 MHz = −0.9Γ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 2D MOT parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.3 Best parameters for the 2D MOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.4 Push beam parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
63
List of Tables
64
Bibliography
[1] T. W. Hänsch, A. L. Schawlow. Cooling of gases by laser radiation. Optics Communications
13, 68-69, 1975.
[2] M. H. Anderson, J. R. Ensher, M. R. Matthews, C. E. Wieman, E. A. Cornell. Observation
of Bose-Einstein condensation in a dilute atomic vapor. Science, 269, 198, 1995.
[3] B. DeMarco, D. Jin. Onset of fermi degeneracy in a trapped atomic gas. Science, 285, 1703,
1999.
[4] I. Bloch, J. Dalibard, W. Zwerger. Many-body physics with ultracold gases.
arXiv:0704.3011v2, 2007.
[5] I. Bloch. Ultracold quantum gases in optical lattices. nature physics, vol 1, 23, 2005.
[6] M. Greiner, O. Mandel, T. Esslinger, T. W. Hänsch, I. Bloch. Quantum phase transition
from a superfluid to a Mott insulator in a gas of ultracold atoms. Nature 415, 39, 2002.
[7] M. Köhl, H. Moritz, T. Stöferle, K. Günter, T. Esslinger. Fermionic atoms in a 3D optical
lattice: Observing fermi-surfaces, dynamics and interactions. PRL 94, 080403, 2004.
[8] H. Ott, E. de Mirandes, F. Ferlaino, G. Roati, G. Modugno, M. Inguscio. Collisionally
induced transport in periodic potentials. PRL 92, 160601, 2004.
[9] N. Strohmaier, Y. Takasu, K. Günter, R. Jördens, M. Köhl, H. Moritz, T. Esslinger.
Interaction-controlled transport of an ultracold fermi gas. PRL 99, 220601, 2007.
[10] R. Jördens, N. Strohmaier, K. Günter, H. Moritz, T. Esslinger. A Mott insulator of fermionic
atoms in an optical lattice. arXiv:0804.4009v1, 2008.
[11] B. DeMarco, S. B. Papp, D. S. Jin. Pauli blocking of collisions in a quantum degenerate
atomic fermi gas. PRL 86, 5409, 2001.
[12] C. Ospelkaus, S. Ospelkaus, K. Sengstock, K. Bongs. Interaction-driven dynamics of 40 K −
87
Rb fermion-boson gas mixtures in the large-particle-number limit. PRL 96, 020401-4,
2006.
[13] J. Catani, P. Maioli, L. De Sarlo, F. Minardi, M. Inguscio. Intense slow beams of bosonic
potassium isotopes. PRA 73, 033415, 2006.
[14] H. Haken, H. C. Wolf, W.D. Brewer. The Physics of Atoms and Quanta: Introduction to
Experiments and Theory. Springer-Verlag, seventh edition, 2004.
[15] R. S. Williamson III. Magneto-optical trapping of potassium isotopes. Dissertation, Univer-
sity of Wisconsin-Madison, 1997.
[16] B. Zimmermann. Aufbau einer magneto-optischen Falle für Kalium-Atome. Diploma thesis,
2005.
[17] H. J. Metcalf, P. van der Straaten. Laser cooling and trapping. Springer, 1999.
[18] E. L. Raab, M. Prentiss, Alex Cable, Steven Chu, D. E. Pritchard. Trapping of neutral
sodium atoms with radiation pressure. PRL 59, 2631, 1987.
[19] C. Fort, A. Bambini, L. Cacciapuoti, M. Prevedelli, G.M. Tino, M. Inguscio. Cooling mech-
anisms in potassium magneto-optical traps. Eur. Phys. J. D. 3, 113-118, 1998.
[20] C. C. Bradley, J. J. McClelland, W. R. Anderson, R. J. Celotta. Magneto-optical trapping
of chromium atoms. PRA 61, 053407, 2002.
[21] J. Fuchs. Aufbau und Charakterisierung einer 2D und 3D magneto-optischen Fallenkombi-
nation für 87 Rb. Diploma thesis.
65
Bibliography
66
Acknowledgment
Looking back, it really amazes me how much I could learn and also achieve in the course of this
thesis. A great part of this is certainly due to the inspiring and welcoming atmosphere brought
forth by the people I worked with during this time. Therefore I would like to thank all the
individuals who have contributed in one or another way to the success of this work.
First of all, I am very grateful to Prof. Dr. Tilman Esslinger, who gave me the chance to write
this thesis in his group and with this put the key into my hands to further explore the field of
quantum optics, which after these four months fascinates me more than ever.
My supervisors, Niels Strohmaier and Robert Jördens continuously supported me from adjusting
my first mirror on an optics table ever to tracking down mysterious laser drifts and improving
my laser set-up with ever new gadgets (which were in the end controlled by a peaceful symbiosis
of both Python and Ruby code). Thank you both for your commitment for my thesis and all the
enlightening discussions.
Whenever I thought I was facing an insoluble problem, Henning Moritz was still a sure bet. His
knowledge in experimental physics seems to be inexhaustible and oftentimes he had already solved
the problem before I even completely understood it. Also, he always showed a great interest in my
work and it was a pleasure to discuss with him new approaches how this or another measurement
could still be improved.
Then there was Bruno Zimmermann, who originally set up the 3D MOT. He helped me taming
the tapered laser and the ever-drifting 3D MOT polarization. Apart from that, it was always
a joy discussing with him current political issues or just exchanging stories from our boy scouts
time.
Jakob Meineke, Kristian Baumann and Daniel Greif all started their PhD theses at about the
same time I started my diploma thesis. Jakob Meineke not only proved to be a dependable guide
through Florence where we had the chance to attend the Young Atom Opticians conference, but he
is also a really clever physicist with deep insights in theoretical physics and a very sociable person.
Kristian Baumann was always the right person to try out some new electronic gadgets with or
discuss the newest advancements in computer technology and he also generously contributed the
USB TV adapter which I used to capture digital images of my MOTs. With Daniel Greif I not
only spent an enjoyable day snowboarding, but also many more not less enjoyable days in the lab
and the office.
A special thank goes to Alexander Frank, who always kept his patience with me, even when he
had to do some major modifications before one of the electronic devices I soldered really worked
the way it was supposed to. Another special thank goes to Veronica Bürgisser, who runs the
whole administration of the lab so smoothly, that one barely ever notices her great efforts. I also
want to thank all the other people in the group which I did not mention in person here for all
the hands-on help and the words of advice they gave me.
Last but certainly not least, I want to thank my whole family and especially my parents for
paving me the way so I could make it to were I am now, and for their support and understanding
over all the years. Thank you, Regula, for the time we spent together over the last years and
especially for your support during the course of this thesis, and I am looking forward to many
more years!
67