Anda di halaman 1dari 62

Magneto-optical trap of 39K: setup and

characterisation
Report of a 36 EC master project

Tom Opdam (5696011)

Supervisors: Prof. dr. J.T.M. Walraven


Dr. V. Lebedev
Dr. S. Lepoutre

project realised during 1-10-2011 until 19-04-2012

Universiteit van Amsterdam


Faculteit der Natuurwetenschappen, Wiskunde en Informatica
van der Waals-Zeeman Instituut
Abstract
During this six months project a three-dimensional magneto-optical trap (MOT)
for the bosonic isotope 39 K has been realised. The three-dimensional MOT is loaded
by a pre-cooled atomic beam originating from a two-dimensional MOT. A frequency-
modulation spectroscopy setup was built to obtain a feedback signal that stabilizes
the frequency of the laser. A characterization of the MOT is made by analysing the
fluorescence images of the atom cloud produced with a CCD camera. Among the
investigated parameters are the lifetime, loading time and the total number of trapped
atoms in the 3D MOT. Based on the results, possible modifications of the setup are
proposed for an improvement of the performance of the MOT.

1
Contents
1 Introduction 4
1.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Theoretical background 5
2.1 Chapter overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Potassium and its energy levels scheme . . . . . . . . . . . . . . . . . . . . . 5
2.2.1 Potassium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Fine and Hyperfine structure . . . . . . . . . . . . . . . . . . . . . . 6
2.2.3 Natural linewidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3.1 Doppler shift and Doppler width . . . . . . . . . . . . . . . . . . . . 10
2.3.2 Doppler-free saturated absorption spectroscopy . . . . . . . . . . . . 12
2.3.3 Frequency modulation spectroscopy . . . . . . . . . . . . . . . . . . 14
2.4 Gaussian optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.1 Gaussian beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.2 Higher order transverse modes . . . . . . . . . . . . . . . . . . . . . 18
2.4.3 The Fabry-Perot cavity . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.4 The confocal cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.5 Mode degeneracy in the confocal cavity . . . . . . . . . . . . . . . . 21
2.5 Laser cooling and trapping . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.2 Scattering force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.3 Magneto-Optical Trap . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.5.4 MOT’s for multilevel atoms . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 MOT dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3 Experimental setup 30
3.1 Chapter overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Vacuum system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Master Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Experimental setup spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5 Laser locking setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.6 Laser system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.7 39 K 2D MOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.8 39 K 3D MOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.9 Fluorescence imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2
4 Results 41
4.1 Chapter overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Spectroscopy results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Laser-locking performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4 Fluorescence measurement with a CCD camera . . . . . . . . . . . . . . . . 45
4.5 MOT parameters as a function of the magnetic field gradient . . . . . . . . 49
4.6 MOT lifetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.7 MOT loading time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5 Conclusions and Outlook 56

6 Update 22-04-2013 57

Bibliography 58

3
1 Introduction
A remarkable aspect of quantum mechanics is that it tells us that all matter around us
falls in either one of two categories, bosons or fermions. There are important differences
between these two fundamental particles. Bosons, with integer spin, are described by Bose-
Einstein statistics. Fermions, with half-integer spin, obey Fermi-Dirac statistics. In 1925,
Einstein predicted that bosons for temperatures near absolute zero, in the order of nano
Kelvin, condens into the ground state of the system, forming a Bose-Einstein condensate
(BEC) [1]. This is in contrast with their counter-particles, the fermions, for which the
Fermi-Dirac statistics state that indentical fermions can not occupy the same quantum
state at any time. Therefore, for temperatures near absolute zero, fermions do not form
a condensate of particles in the same ground state, but they each occupy one state up to
the Fermi energy, forming a Fermi degenerate gas. These quantum mechanical phenomena
that occur near absolute zero temperature present an opportunity to investigate quantum
mechanical effects with macroscopic entities. The experimental realisation of bosonic and
fermionic ensembles at temperatures in the order of nano Kelvins has proved to be a chal-
lenging one. In 1995, seventy years after Einsteins prediction, experimentalists were able
to cool bosons for the first time to degeneracy, realizing the first Bose-Einstein condensate
in a dilute atomic gas [2], [3]. It took another four years to demonstrate the first fermionic
degenerate gas, [4].
In those seventy years from prediction to realisation, probably the most important
invention that contributed to the creation of degenerate gases, was the laser. Since the
interaction of atoms with light forms the most important mechanism for the creation and
manipulation of ultracold atoms, the laser, capable of emitting narrow band light, is at
the basis of many experimental schemes in which atoms are trapped and cooled. A crucial
step towards quantum degeneracy was made in 1986, when trapping of neutral atoms with
a device called the magneto-optical trap (MOT) was made possible [5]. MOT’s make use
of optical forces that are exerted on atoms that experience Zeeman shifts in the presence
of a position dependent magnetic field. The frequency of the laser light that is used is
required to be stabilized with respect to a specified atomic transition, with an accuracy
below the natural linewith of the transition. A MOT provides experimentalists with atoms
at temperatures in the order of micro Kelvins. Not yet cold enough for degeneracy to
occur, but a great starting point for ultracold gas experiments.
In this project, the cooling and trapping of a bosonic potassium isotope, 39 K, is demon-
strated. MOT’s for potassium-39 have already been built in numerous groups, with various
loading techniques. In our setup we have chosen a two-dimensional MOT as a cold atom
source to provide an atomic beam that loads the MOT in a ultra high vacuum chamber.

4
1.1 Outline
In chapter 2, the theoretical background required for the description of the work done in
this project is discussed. Among the covered subjects are the properties of potassium-39,
saturation absorption spectroscopy, frequency-modulation spectroscopy, Gaussian beams,
the Fabry-Perot interferometer and the principles of laser cooling and trapping. In chapter
3, a detailed description is given of the experimental setup used in this project. The results
of the measurements, their interpretation and comparisons with other work are presented
in chapter 4. Finally, conclusions are drawn and an outlook on future experimental steps
is given in chapter 5.

2 Theoretical background
2.1 Chapter overview
In this chapter a brief review of the theory used in this project is given. The chapter is
divided into four parts. As a starting point, in section 2.2, an introduction to the potassium
atom is made and several concepts of atomic physics are given. In section 2.3, the topic of
spectroscopy and its applications to laser locking are addressed. Gaussian beams and their
relation with the Fabry-Perot interferometer are discussed in section 2.4. In the last section
of this chapter, section 2.5, the principles of cooling and trapping of atoms are covered,
leading to a discussion of the main object in this project, the magneto optical trap.

2.2 Potassium and its energy levels scheme


2.2.1 Potassium
Potassium is an alkali metal denoted with the symbol K (after the Latin name Kalium)
and has an atomic number of 19. Like all other alkali metals, it has one electon in the
outermost shell, giving the atom a low ionization energy. Because of this low ionization
energy, potassium is chemically reactive. It oxidizes rapidly with air and is reactive with
water. Therefore, special care has to be taken when assembling glass ampules containing
potassium. An overview of the different naturally occurring potassium isotopes is presented
in table 1. Whereas 39 K and 41 K are bosonic isotopes, 40 K is a fermionic isotope. Together
with 6 Li it is the only stable fermionic alkali isotope, making it a popular atom for ultracold
fermionic experiments [6]. However, in this project the atom of interest is the bosonic
isotope 39 K, since its high natural abundance is practically favourable for trapping a high
number of cold atoms.

5
Isotope A Neutrons N Abundance (%) m(u) I
39 20 93.2581 38.9637 3/2 bosonic
40 21 0.0117 39.9639 4 fermionic
41 22 6.7302 40.96182 3/2 bosonic

Table 1: Several properties of the naturally occurring potassium isotopes. The atomic
number Z of potassium is 19. The given properties are: the mass number A, the number
of neutrons N , the abundance in %, the atomic mass m and the nuclear spin I. Values
taken from [7].

2.2.2 Fine and Hyperfine structure


The fine structure of an alkali atom is the result of the coupling between the orbital angular
momentum L of the valence electron and its spin angular momentum S. The total angular
momentum of the electron is given by

J = L + S, (1)

with the condition that the quantum number J of J lies in the range

|L − S| ≤ J ≤ L + S. (2)
The Hamiltonian describing the fine structure for hydrogen-like atoms is given by

HLS = ξ(r)L · S, (3)


here ξ(r) is the coupling constant. The eigenvalues of this Hamiltonian can be calculated
with the operator identity
1
L · S = (J2 + L2 + S2 ). (4)
2
It follows that the eigenvalues of the fine structure Hamiltonian for hydrogen-like atoms
are proportional to
1
∆Ef s (n2s+1 LJ ) ∝ [j(j + 1) − l(l + 1) − s(s + 1)]. (5)
2

Here, j, l and s are the quantum numbers associated with J, L and S. (For alkali atoms
only the valence electron contributes to the angular momentum of the atom and the con-
vention is that the angular momentum of one electron is indicated in lower case [8].) It
follows from equation (5) that fine splitting only occurs if l 6= 0.
In the case of 39 K, the electronic groundstate is 42 S1/2 , with L = 0 and S = 1/2 and

6
therefore J = 1/2. For the first excited state L = 1 and S = 1/2, thus J = 1/2 or J = 3/2,
corresponding to the states 2 P1/2 and 2 P3/2 . The fine structure lifts the degeneracy of
these first excited states. The transition 42 S1/2 → 2 P1/2 is called the D1 transition and
the transition 42 S1/2 → 2 P3/2 is called the D2 transition.
Besides the spin-orbit coupling, one should als take into account the coupling between
the nuclear spin I and the total electronic angular momentum. The total angular momen-
tum is given by

F = J + I. (6)

where the quantum number F , associated with the operator F, must lie in the range:

|J − I| ≤ F ≤ J + I. (7)

The Hamiltonian for the hyperfine structure is given by

ahf s bhf s 3(I · J)2 + 32 I · J + I2 J2


Hhf s = I · J + . (8)
~2 ~2 2I(2I − 1)J(2J − 1)

where ahf s is the magnetic-dipole constant and bhf s is the electric-quadrupole coupling
constant.
In a similar way as for the fine structure, the dot product I · J can be written as 21 (J2 +
L2 + S2 ) and this leads to an hyperfine energy shift of
3
ahf s K(K + 1) − 2I(I + 1)J(J + 1)
∆Ehf s (n2s+1 LJ ) = K + bhf s 2 . (9)
2 2I(2I − 1)J(2J − 1)
where K = F (F + 1) − I(I + 1) − J(J + 1).
In contrast to the fine structure, the energy levels in the hyperfine structure split also
when L = 0. This is the reason that the ground state of 39 K with J = 1/2 and I = 3/2
is split in two levels, the states 4S1/2 (F = 1) and 4S1/2 (F = 2). In figure 1, the fine
and hyperfine structure of the three natural abundant potassium isotopes are shown. The
strongest spectral lines of the ground state potassium atoms are the D1 line (2 S1/2 → 2 P1/2 )
and the D2 line (2 S1/2 → 2 P3/2 ). The values of the optical transition frequencies in figure
1 are taken from Falke et al., who have published their high precision measurements of the
potassium energy levels in [9]. It should be noted that the hyperfine structure of the 40 K
isotope is inverted. This a result of its negative magnetic-dipole coupling constant. Both
39 K and 41 K have a positive magnetic-dipole coupling constant and thus a non-inverted

hyperfine structure.

7
Figure 1: The fine- and hyperfine structure of 39 K, 40 K and 41 K. The hyperfine structure of
of 40 K is inverted due to its negative magnetic dipole constant, ahf s . Both other isotopes
have a postive magnetic dipole constant and thus a non-inverted hyperfine structure. Figure
taken from [7].

8
Each of the hyperfine energy levels consists of 2F + 1 magnetic sublevels. Without an
external magnetic field these sublevels are degenerate. However, when an external magnetic
field is aplied, the degeneracy is lifted due to the Zeeman effect. The Zeeman Hamiltonian
is given by
µB
HZ = (gJ J − gI I) · B, (10)
~
where µB is the Bohr-magneton, gJ is the Landé factor for an electron and gI is the
nuclear gyromagnetic factor. The total hyperfine interaction in the presence of an external
magnetic field is thus given by the Hamiltonian

H = Hhf s + HZ . (11)

For weak magnetic fields the I-J coupling is still present and the total angular momentum
F precesses around the direction of the magnetic field. In that case the Hamiltonian HZ
perturbs the zero-field eigenstates of Hhf s and the energy levels split linearly according to

∆EZB,low = µB gF mF B , (12)
where gF is given by

F (F + 1) + J(J + 1) − I(I + 1) F (F + 1) + I(I + 1) − J(J + 1)


gF = gJ + gI . (13)
2F (F + 1) 2F (F + 1)

In the case of a high magnetic field the I-J coupling is lifted. Both angular momenta
precess independently around the direction of the magnetic field. The energy shift in this
regime is called the Paschen-Back effect and is approximated by [10]

3(mJ mI )2 + 32 mJ mI − I(I + 1)J(J + 1)


∆EZB,high = µB gJ mJ + ahf s mI mJ + bhf s . (14)
2J(2J − 1)I(2I − 1)

The magnetic crossover field where the energy of the states in the low-field approximation
equals the energy in the high-field approximation is much higher than the magnetic field
used in a magneto optical trap. Therefore, the linear low-field approximation is valid to
describe the energy shifts of cold atoms in the MOT.

2.2.3 Natural linewidth


The difference between two atomic energy levels can be overcome by the absorption or
emission of a photon with the same energy, E = ~ω0 , where ω0 is the transition frequency.

9
However, because of the generally short lifetimes of excited states, the uncertainty relation
∆E∆t > ~2 tells us that there is a significant uncertainty in the measured energy, and
thus in the measured transition frequency. Emission and absorption lines have therefore
finite widths that are called natural linewidths. These are the full widths at half maximum
(FWHM) of the absorption/emission peaks, usually denoted by the symbol Γ. Natural
linewidths have a Lorentzian profile.
In the case of 39 K, the D2-line has a lifetime of τ ≈ 26.37 × 10−9 s. This corresponds
to a natural linewidth Γ/2π ≈ 6.03 MHz [7].

2.3 Spectroscopy
In the field of laser cooling and trapping, spectroscopy plays an important role. For most
experiments it is of great importance that the main parameters; i.e., frequency and inten-
sity, of the laser are stabilized. Moreover, many experiments require control of the laser
frequency to a fraction of the linewidth of atomic transitions, which typically has a width
in the order of MHz. Fluctuations in the laser diode current and temperature result in
frequency fluctuations. Mechanical vibrations also have to be taken into account. To com-
pensate for these fluctuations, lasers have to be actively stabilized. A number of different
spectroscopy techniques exist in order to accomplish this. In this project we chose to use
frequency modulation (FM) spectroscopy (see section 2.3.3) to deduce a so-called error
signal that is needed for laser stabilization.
Typical spectroscopy works as follows. Laser light is sent through an ensemble of
atoms. If the photon energy, E = ~ω, matches the energy difference between two atom
states, the photons are absorbed or emitted. This absorption can be measured with the
use of a photodiode that converts light intensity into current. In this way information is
gathered about the internal structure of atoms. This information can be used to provide
feedback to a laser system and thereby stabilizing its main parameters.
In the following sections several aspects of spectroscopy and the principle of FM-
spectroscopy will be explained.

2.3.1 Doppler shift and Doppler width


In general, the natural linewidth can not be observed without special techniques, because it
is completely concealed by broadening effects that broaden the linewidth of absorption lines
in atomic spectra. Doppler broadening is one of the major contributions to the spectral
linewidth of atomic gases at room temperature [11], due to the thermal motion of the
atoms.
Consider an atom with transition frequency ω0 . If this atom moves with a velocity
v across a plane electro-magnetic wave, the absorption frequency is shifted. The wave
frequency ω in the restframe is seen in the reference frame of the moving atom as

10
ω 0 = ω − k · v, (15)

where k is the wavevector of the radiation with magnitude k = ω/c = 2π/λ, and v is
the velocity of the moving atom. It is the component of the velocity along k that leads
to the Doppler effect. The atom can only absorb a photon if ω 0 is equal to the transition
frequency ω0 . Hence, the frequency at which absorption could occur is

ω = ω0 + k · v. (16)

If we assume for simplicity that the direction of movement of the atoms coincides with the
light propagation (k·v = kv), it means that atoms moving with velocity v absorb radiation
when the frequency of radiation is
 v
ω = ω0 1 + . (17)
c

Atoms in a gas move in all possible directions. As a result, the Doppler shift is different
for individual particles. When a thermal equilibrium is reached all directions are equally
probable; i.e., the velocity distribution of the particles is isotropic. The projection of the
particle velocity on any direction is given by a Maxwell velocity distribution [11]. The
number of atoms as a function of their velocity is then given by

Ni h  v 2 i
ni (v) = √ exp − , (18)
u π u
p
where Ni is the number of all atoms in the lower state, u = M/(π2kB T ) is the most
probable velocity, and kB is Boltzmann’s constant. The number of atoms as a function of
their absorption frequency, that is shifted from ω0 to ω, is acquired by using the relation
(16) between the velocity component and the frequency shift

Ni c h  c(ω − ω ) 2 i
0
ni (ω) = √ exp − . (19)
ω0 u π ω0 u

Since the absorbed light is proporsional to the number of atoms, the intensity profile of
Doppler broadened spectral line becomes
h  c(ω − ω ) 2 i
0
I(ω) = I0 exp − . (20)
ω0 u

This is a Gaussian profile with a full width at half maximum (FWHM)

11
√ ω0 p
∆ωD = 2 ln2ω0 u/c = 8kT ln2/m, (21)
c
this expression is known as the Doppler width.
A typical Doppler linewidth for 39 K at room temperature is around ∆ωD ≈ 4.8 GHz.
A huge difference in comparison with the natural linewidth of the 39 K D2 line, Γ/2π ≈ 6.03
MHz.

2.3.2 Doppler-free saturated absorption spectroscopy


In order to resolve hyperfine transitions in spectroscopy spectra, one needs to have a spec-
tral resolution that is not limited by the Doppler-width. Doppler-free saturated absorption
spectroscopy is a spectroscopy method that is capable of doing this. It will be explained
here, because it is an important part of the spectroscopy method that we used in this
project.
In saturated Doppler-free absorption spectroscopy, two beams are used; a weak probe
beam and a stronger pump beam. The beams propagate in opposite directions and overlap
in an atomic vapor cell. Both beams come from the same laser and the frequency of the
laser is scanned. After passing through the vapor cell the probe beam is sent to a photodi-
ode, so that the intensity of the transmitted probe beam can be monitored. If there was no
pump beam, and the laser would be scanned, an ordinary Doppler linewidth would be seen,
obscuring the hyperfine transitions. The addition of a pump beam, eliminates the Doppler
linewidth in atomic spectra and therefore resolves the hyperfine transitions. The pump
beam passes through the vapor cell, with an intensity high enough to saturate a certain
atomic transition. This means that the transmission is driven at a maximum rate which is
determined by the lifetime of the excited state and can not be driven faster. As a result,
the population density in the lower level is quickly depleted. The probe beam, with a lower
intensity, passes through the vapor cell coming from the other direction. Because of the
saturation of the absorption by the pump beam, there are few atoms left to interact with
for the probe beam. This results in a peak in the signal of the probe beam once it arrives
at the photodiode. A typical setup for saturated absorption spectroscopy is depicted in
figure 2.

12
Figure 2: A schematic representation of a typical saturation absorption spectroscopy setup.
A beamsplitter (BS) divides the power of the laser in a weak probe beam and a strong
pump beam. The angle between the two beams is strongly exaggerated, it is usually tried
to make the beams as counter-propagating as possible. Figure adapted from [12].

Depending on the frequency of the laser, three different situations can be distinguished.
First of all, if the laser has a frequency far from resonance, both beams interact with atoms
in different velocity classes due to the Doppler effect. If the pump beam interacts with
atoms with velocity, v, the counter-propagating probe beam interacts with atoms with ve-
locity, −v. The probe beam is not affected by the pump beam and a normal Doppler line
profile is seen. The second situation arises when the laser frequency is close to resonance,
both beams interact with atoms in the same velocity class, v = 0. As a result, the intensity
of the probe beam transmitted through the sample has a peak at the resonance frequency.
The last situation occurs when two atomic transitions have a common lower or upper level
and their frequency difference is less than the Doppler-width, |ω1 −ω2 | < ∆ωD . Then, extra
resonances will become visible when saturation absorption spectroscopy is applied. These
resonances are called cross-over resonances. Assume that the laser frequency, ω, is exactly
tuned halfway between two transitions ω1 and ω2 , (ω1 < ω2 ). Then, the pump beam will
be Doppler-shifted into resonance at ω1 for atoms in the velocity class, v = (ω − ω1 )c/ω1 ,
moving in the propagation direction of the pump beam. The returning probe beam will
be Doppler-shifted into resonance at ω2 for atoms in the velocity class, v = (ω2 − ω)c/ω2 ,
the atoms that move towards the probe beam; i.e., in the propagation direction of the
pump beam. If ω = (ω1 + ω2 )/2, the velocity classes that are interacted with are the same
and the saturating pump beam wil saturate these atoms and few atoms are left to interact
with for the probe beam. Therefore, besides the saturation signals at ω1 and ω2 where the
velocity class v = 0 is saturated, an additional saturation signal, the cross-over signal, can
be observed at the frequency (ω1 + ω2 )/2.

13
Figure 3: A schematic subtraction of the signals from the two photodiodes detecting the
probe beam and the reference beam (figure adapted from [13]).

A useful addition that can be made to a saturation absorption spectroscopy scheme,


is adding a third (weak) beam, called the reference beam. This reference beam only passes
through the vapor cell ones and does not overlap with any other beam. The reference
beam is sent to a photodiode and gives an ordinary Doppler-width. The photocurrents
from both photodiodes can be subtracted, resulting in a spectrum where the Doppler-
width is completely eliminated and only transition peaks and cross-over peaks are seen
(figure 3).

2.3.3 Frequency modulation spectroscopy


Doppler-free saturated absorption spectroscopy makes it possible to obtain a well-resolved
atomic spectrum at room temperature that would otherwise be hidden under the Doppler-
width. Intensity peaks are seen at the transition frequencies and cross-over frequencies.
Having a well-resolved atomic spectrum makes it possible to lock a laser; a well defined
absorption peak could be used as a locking point. However, atomic spectra can not be
directly used for laser locking for the following reason. To lock a laser, a dispersive signal
is needed. This so-called error signal should ’tell’ the feedback system how much the fre-
quency deviation is and in which direction, such that the feedback system can compensate
the change. If one wants to lock a laser to a certain atomic transition (i.e. ’on top’ of a
transmission peak), an intensity measurement gives no disclosure about the direction in
which the frequency has changed, since the measured voltage decreases in both directions.
Therefore, the change of the error signal has the same sign for opposing frequency shifts.
To circumvent this problem one can use the derivative of the original signal. The deriva-
tive signal has zero crossings at the peaks of the original signal, thus the sign of this signal
is different for opposite deviations of the laser frequency. The derivative signal makes it
possible to detect the amplitude and the direction of the change of the laser frequency, it
is therefore well suited to function as an error signal.
Frequency modulation (FM) spectroscopy is a technique that can be used to obtain

14
such an error signal. In FM spectroscopy the frequency of the laser is modulated with a
sinusoidal signal, typically with a small peak-to-peak amplitude of several mV and with
a high frequency of a about 10 kHz. This can be done either by modulating the current
of the laser or by letting the light pass through an electro-optic modulator (EOM) that
is driven at the required frequency. If the laser frequency is near a transition frequency,
the modulation of the frequency causes the absorption signal to modulate synchronously;
i.e., the laser frequency modulation is mapped onto the transmitted intensity. This change
in intensity allows the photodiode to detect a change in frequency. The conversion from
frequency modulation to intensity modulation is sensitive to where the center frequency,
around which the laser is modulated, is located on the absorption line. If it is located on
a steep slope of the absorption line, a small change in frequency will result in a significant
change in transmitted intensity. But if the center frequency is located at the top of the
absorption line, a change in frequency will result in a small change in transmitted intensity.
It is this principle that makes it possible to obtain the derivative of the absorption signal
(see figure 4).

Figure 4: In FM-spectroscopy, frequency modulation is converted into amplitude modu-


lation of the absorption signal, FM to AM. a) center frequency located at the negative
slope of the absorption profile (AM amplitude relatively large); b) center frequency located
near the point of maximal absorption (AM amplitude relatively small); c) center frequency
located at positive slope (same AM amplitude as in case a, but with AM/FM relationship
reversed; d) plot of AM/FM conversion ratio as a function of frequency (effectively the
derivative of the absorption profile) - figure adapted from [13].

15
The mathematical derivation of the derivative of the absorption signal can be under-
stood in the following way. Let a laser frequency ω be modulated with a frequency Ω with
amplitude m, the transmitted intensity IT that reaches the photodiode after passing the
vapor cell can be written as

IT (ω, t) = IT (ω + m sin(Ωt)). (22)

Here we assume that the amplitude of the modulation is small and that the frequency of
modulation is smaller than the linewidth Ω < Γ.
Now IT can be expanded as a Taylor series around ω

dIT
IT (ω + m sin(Ωt)) =IT (ω) + [m sin(Ωt)]+

1 d2 IT 2 1 d3 IT
[m sin(Ωt)] + [m sin(Ωt)]3 ... (23)
2! dω 2 3! dω 3

With the help of basic trigonometric identities we can combine the terms

h m2 d2 IT i h dI
T m3 d3 IT i
IT (ω + m sin(Ωt)) = IT (ω) + + ... + sin(Ωt) m + + ... +
4 dω 2 dω 8 dω 3
h m2 d2 I
T 
cos(2Ωt) − 2
+ ... + .... (24)
4 dω

It can be seen that the transmitted intensity contains a DC term, a term oscillating at Ω,
a term oscillating at 2Ω, etc. When phase sensitive detection is applied, the coefficients of
the different terms can be extracted. In this case phase sensitive detection is done with
the use of a lock-in amplifier. This amplifier is given the transmitted intensity signal and a
periodic reference signal. The amplifier responds only to the part of the signal that occurs
with the same frequency as the periodic reference signal. So if, in this case, the reference
signal is set at a frequency of Ω or a multiple of Ω, coefficients from different terms can be
extracted from the transmitted intensity signal. If m is sufficiently small, the coefficient
of the sin(Ωt) term is essentialy the first derivative of the absorption signal multiplied by
m. By the same token the second derivative can be extracted from the cos(2Ωt) term and
higher order derivatives extracted from higher order terms. This is why FM spectroscopy
is sometimes called derivative spectroscopy.

16
2.4 Gaussian optics
2.4.1 Gaussian beam
In optics, laser beams often occur in the form of Gaussian beams. They are a solution to
the paraxial form of the Helmholtz scalar equation, an equation that provides solutions
for the propagation of elecromagnetic waves. Gaussian beams are beams whose transverse
electric field and intensity distributions are well aproximated by Gaussian functions. The
transverse intensity profile of a Gaussian beam is described with the Gaussian function [14]
2
−2r2
  
w0
I(r, z) = I0 exp , (25)
w(z) w2 (z)

where I0 = I(0, 0) is the intensity at the center of the beam at its waist, w(z) is the beam
width (as function of the propagation direction) which is defined as the radius where the
intensity drops to 1/e2 of its maximum value and w0 is the beam width where the radius
is smallest, at the waist, see figure 5.
The beam radius w(z) varies along the propagation direction according to
s  2
z
w(z) = w0 1 + , (26)
zR

where the origin of the z-axis is defined to coincide with the beam waist and where zR is
the Rayleigh length, defined as zR = (πw0 2 )/λ, with λ being the wavelength of the light.
The Rayleigh length determines the length over which the beam can propagate without
significantly diverging. At a distance from the waist equal to zR , the radius of the beam is

w(zR ) = w0 2 . (27)

Figure 5: Gaussian beam width w(z) as a function of z. R(z) is the radius of curvature
of the wavefront. Notice that a wavefront at the waist is similar to that of a plane wave.
(Figure taken from [15])

17
Figure 6: Several different Laguerre-Gaussian transverse modes. These modes are electro-
magnetic normal modes in a Fabry-Perot cavity. The TEM00 mode has a simple Gaussian
beam profile. (Picture taken from [16])

2.4.2 Higher order transverse modes


Gaussian beams are not the only solution to the paraxial form of the Helmholtz scalar
equation. Several other solutions are used for modeling laser beams. For lasers with
cylindrical symmetry the transverse modes are described by a combination of a Gaussian
beam profile and a Laguerre polynomial. The intensity of a beam at a point r, φ (in polar
coordinates) is given by the formula [17]
h i2
Ipl (r, φ) = I0 ρl Llp ρ cos2 (lφ)e−ρ , (28)

where ρ = 2r2 /w2 and Llp is the Laguerre polynomial of order p and index l. In figure
6, several different modes are shown. These are so-called Transverse Electro Magnetic
(TEM) modes, which means that the associated beams have neither electric nor magnetic
field in the direction of propagation. The TEM00 mode has a simple Gaussian profile and
is referred to as the lowest order mode.

18
Figure 7: A plane Fabry-Perot etalon. If Θ is not 0, light ’walks-off’ in the direction parallel
to the mirrors. (Figure adapted from [18])

2.4.3 The Fabry-Perot cavity


Fabry-Perot cavities (also called etalons or interferometers) are widely used instruments
in optical physics. They are being used as sensitive wavelength discriminators, as stable
frequency references and for building up large field intensities with low input powers. Also,
optical cavities are a major component of lasers, surrounding the gain medium and pro-
viding feedback of the laser light. A Fabry-Perot cavity consists of two reflecting surfaces
or mirrors. The ’inner’ sides of the two mirrors (the sides facing each other) are coated
for high reflectivity. Curved mirrors are typically used because this configuration can trap
light in a stable mode. Flat mirrors can also make a cavity, but it is often not a stable
one, the light ’walks off’ if the incoming beam is not exactly perpendicular to the mirror,
see figure 7.
The transmission function of ligth that enters a cavity depends on the interference
between the multiple reflections of the light between the mirrors. If the transmitted beams
are in-phase, contructive interference occurs, creating a standing wave in the cavity and
resulting in a high transmission peak. If the reflected beams are out-of-phase, destructive
interference occurs, resulting in a transmission minimum. A detailed derivation of the
transmission function in cavities is given in [19]. Whether the transmitted beams are in-
phase or not depends on the wavelength of the light, the length of the cavity and the angle
the incoming beam makes with respect to the axis of the cavity. In order for constructive
interference to occur, the phase change after one round-trip in the cavity should be a
multiple of 2π. This is the case if the length of the cavity is an integer number of half-

19
wavelengths of the light, qλ/2 = L, or equivalently, ν = qc/2L, where L is the length of
the cavity, c is the speed of light and q is some integer. The transmission spectrum of a
Fabry-Perot cavity will have a series of peaks with frequencies νq−1 , νq , νq+1 . These peaks
are spaced by the free spectral range (FSR)
c
∆νF SR = νq+1 − νq = . (29)
2L

The width of the peaks with respect to the FSR (i.e., the quality of the etalon), depends
on the reflectivity of the mirrors. If the reflectivity is high, then the width of the peaks
will be narrow compared to the FSR. A quantity called the finesse, F , is defined as the
ratio of the FSR and the full-width-at-half-maximum (FWHM) of the peaks

∆νF SR
F = . (30)
∆νf whm

It can be shown [19] that the finesse of a cavity is approximately



π R
F ≈ , (31)
1−R

where R is the reflectivity of the mirrors in the cavity.

2.4.4 The confocal cavity


A special type of cavity is the confocal cavity. In this type of cavity the radius of curvature
of both mirrors is equal to the length of the cavity. Therefore, the focal points of the two
mirrors coincide at the center of the cavity. All light beams that enter the confocal cavity
parallel to its axis pass through the focal point and return to their entrance-point after
having made a bow-tie shape. Thus, light that enters a confocal cavity effectively traverses
the cavity twice before retracing its path, see figure 8. As a result, the transmission peaks
are spaced in frequency by ∆νF SR /2 = c/4L and various transverse modes are degenerate,
as will be explained in the next section. Another feature of the confocal cavity is that it is
relatively insensitive to the alignment of the incoming beam, since the bow-tie modes are
excited no matter where the beam enters the cavity.

20
Figure 8: The path of incoming light in a confocal Fabry-Perot cavity. The optical path
length is 4L; i.e., four times the radius of curvature of the mirrors.

A useful addition can be made to a Fabry-Perot cavity. Placing a piezoelectric spacer


between the two mirrors makes it possible to vary the separation of the two mirrors with
a few wavelengths. Therefore, the length of the cavity is scanned and hence the resonant
frequencies as well. This turns the Fabry-Perot into optical spectrum analyzer. If the laser
beam that is sent through the cavity contains frequencies in a range about ν0 , one can
record the laser spectrum.

2.4.5 Mode degeneracy in the confocal cavity


There is an important difference between the propagation of a Gaussian beam and the
propagation of a plane wave. A Gaussian beam acquires an additional phase shift, which
differs from that for a plane wave with the same frequency. This phase shift is called the
Gouy phase shift, named after the French physicist Louis Georges Gouy. The Gouy phase
shift is given by
 z 
φG (z) = − arctan , (32)
zR

where zR is the Rayleigh length and z = 0 corresponds to the position of the beam waist.
The result of this phase shift is that the distance between wavefronts, in comparison with
the wavelength defined for a plane wave of the same frequency, is slightly increased. This
means that the phase fronts have to propagate slightly faster, which leads to an increased
local phase velocity. The above written equation for the Gouy phase shift is only true
for Gaussian beams. In the case of higher order transverse modes, for example Laguerre-
Gaussian beams, the phase shift is even stronger. For TEMpl modes it is stronger by a
factor (1 + p + l), resulting in [20]
z
φG (z) = −(1 + p + l) arctan( ). (33)
zR

Which brings the total phase shift in Gaussian beams to

21
z
φq,p,l (z) = kq z − (1 + p + l) arctan( ), (34)
zR

where kq is the wavevector defined as 2π/λ and q is the longitudinal mode index defined
by an integer.
The Gouy phase shift has an important effect on the resonance frequencies for different
transverse modes in cavities. The frequencies in cavities are determined by the condition
that the length of the cavity should be equal to an integral number of half-wavelengths, in
order for contructive interference to occur. It is the same condition that requires that the
change in phase after a complete round-trip is equal to a multiple of 2π. Therefore, for
a spherical mirror cavity with mirrors at z1 and z2 , this condition for a p, l mode can be
written as

φq,p,l (z2 ) − φq,p,l (z1 ) = qπ. (35)

Now with the use of equation (34), we can write the resonance condition as
 
z1 z2
kq L − (p + l + 1) arctan( ) − arctan( ) = qπ. (36)
zR zR

here L is the distance between the mirrors, z2 − z1 . If we restrict ourselves to the special
case of the confocal cavity, where z2 = −z1 = zR , this equation simplifies to
π
kq L − (p + l + 1) = qπ. (37)
2

It follows that for two different longitudinal modes with the same value of the sum (p + l)
the following holds
π
kq+1 − kq = , (38)
L

or, using k = 2πν/c,


c
νq+1 − νq = . (39)
2L

This is the frequency spacing between different longitudinal modes, with the restriction
that the sum of the transverse modes indices (p + l) is constant. This spacing is the regular
free spectral range, seen before in section 2.4.3 .
Now, we can consider the effect of varying the transverse modes indices, but keeping the
longitudinal index q constant. From equation (37) it follows that the resonant frequencies

22
depend on the sum (p + l) and not on p and l separately. So for a given q, all the modes
with the same value of (p + l) have the same resonant frequency; i.e., are degenerate. For
modes with a constant q but with a different sum of (p + l), the frequency spacing follows
again from the resonance equation (37) for a confocal cavity,
c
νq,p,l − νq,p0 ,l0 = ∆(p + l) . (40)
4L

From this formula it is seen that in the confocal cavity the resonance frequencies of differ-
ent transverse modes either coincide or fall halfway between the resonance frequencies of
different longitudinal modes. The frequency spacing between the different modes is thus
half the free spectral range, ∆νF SR /2. Different transverse modes can be referred to as
even and odd modes, even when the sum (p + l) is an even number and odd when the
sum is odd. For instance, two consecutive TEM00 longitudinal modes are two even modes
separated by the FSR, but a TEM00 mode and a TEM01 mode are spaced by FSR/2.
The above gives rise to a method to align a confocal scanning Fabry-Perot cavity. If the
incoming beam is a pure TEM00 mode, then, if the cavity is well-aligned, the odd modes
can not be excited inside the cavity and one should only see even modes and a rejection of
odd modes.
Another feature that can be exploited is that if the incoming beam is not a pure TEM00
mode, but for instance a superposition of different transverse modes, odd modes are certain
to be seen. This can be used to give a qualitative statement about the mode structure of
the incoming light, e.g. studying the mode structure of gas lasers.
Since the frequency spacing between even and odd modes is known, it is possible to give a
quantitative measure for the stability of the laser, by measuring how much the transmission
peaks have shifted in time with respect to the distance between the peaks. This is done in
order to say anything about the quality of the laser frequency locking, findings are written
down in section 4.3.

2.5 Laser cooling and trapping


2.5.1 Overview
In this section the mechanism behind the magneto-optical trap is discussed. In order to do
this, the principle of the scattering force, optical molasses and the use of magnetic fields
in traps is explained first.

2.5.2 Scattering force


Consider the simple system of a two-level atom in its ground state at rest. If the atom is
irradiated by a resonant, single frequency laser light that propagates along the +î direction,

23
the atom will absorb a photon and will be excited into its excited state. In this process
the momentum of the atom changes from pi to pk = pi + ~k î. A bit later, the atom de-
excites, thereby re-emitting the absorbed photon either through spontaneous or stimulated
emission. If the emission is stimulated by the laser, the photon is emitted with momentum
+~k î, in the same direction as the laser. The result is that in this process of absorption
and stimulated emission, the atom undergoes no net momentum change. However, if the
emission of the photon is spontaneous, the direction of emission is random and isotropic.
Over many cycles of absorbing and spontaneously emitting photons the average momentum
due to spontaneous emission is zero, giving the atom a net push in the direction of the
laser beam. Although the recoil velocity is small, for the D2 transtition of 39 K at 766 nm
the recoil velocity is ~k/m = 1.34 cm/s, the cumulative effect of absorbing photons leads
to a significant velocity change. The force on an atom due to electro-magnetic radiation is
called the scattering force and its magnitude is proportional the rate at which the absorbed
photons impart momentum to the atom

Fsc = ~kΓsc , (41)

where Γsc is the scattering rate given by [12]

Γ I/Isat
Γsc = , (42)
2 1 + I/Isat + 4δ 2 /Γ2

here Γ is the natural linewidth of the atom, δ = ω − ω0 + kv is the detuning from resonance,
with the Doppler shift taken into account and Isat is the saturation intensity, given by

~ω03 Γ
Isat = . (43)
12πc2
Any additional intensity will not significantly increase the scattering force, because the
transition can not be driven faster than the maximum rate which is determined by the
lifetime of the exited state. If a really high intensity is used, assume I → ∞, the scattering
force has a limiting value of Fmax = ~kΓ/2. This means that the maximum acceleration
of an atom with mass m is given by

Fmax ~k Γ vr
amax = = = . (44)
m m2 2τ

For a 39 K atom amax = 2.5 × 105 m/s2 , which is over 104 times the gravitational acceler-
ation, which means that one can omit the effects of gravity when designing techniques for
trapping of these atoms.
Now, let us consider a second beam of equal intensity and frequency propagating in
the opposite direction; i.e., −î. If the frequency of both beams is tuned to a few linewidths

24
below the resonance peak, an atom at rest situated between both beams scatters relatively
few atoms because the lasers are not on resonance. However, if the atom is not at rest,
but has a velocity +v î in the direction of the +î-beam, the resonant frequency is Doppler
shifted. Therefore, the atom will scatter more photons from the −î-beam. This leads to
an inequality in scattering forces for moving atoms, which can be used to slow the atoms
down. In 1985 the group of S. Chu applied this so-called optical molasses technique to three
dimensions by using three orthogonal pairs of counter propagating beams. This resulted
in cooling of a sodium vapor to 240 µK [21]. It should be remarked that slowing atoms
down is not exactly the same as cooling atoms, but as the atoms are slowed down, their
velocity ditribution is not only shifted to lower values, but it also becomes narrower. This
leads to an increase in the phase-space density resulting in the cooling of atoms [22].

2.5.3 Magneto-Optical Trap


The optical molasses technique is important for cooling atoms but it is not able to confine
atoms in space. This technique does not provide a confining trapping potential, but rather
exerts a friction-like force on the moving atoms. Although cold, atoms in optical molasses
still follow a random-walk path due to all the scattering and will eventually diffuse out
of the molasses. A typical lifetime for atoms in molasses is about 0.1s [Chu et al. 1985].
The shortcoming of the optical molasses technique is that it only has a velocity-dependent
force but lacks a position-dependent force. Such a force was established in a magneto-
optical trap (MOT), for the first time in 1987 [Chu et al.]. The addition of a magnetic
field gradient with a zero-field at the origin of the trap was the key to trapping atoms with
both a velocity-dependent and a position-dependent force.
Let us consider the one-dimensional case again, with two counter-propagating slightly
red-detuned beams, now with the additional requirement that the beams have opposite
circular polarizations. And let us consider what occurs when a linearly varying magnetic
~
field of the form B(z) = (0, 0, Bz) is applied. At the center of the trap, z = 0, the field
is zero, but moving away from the center, the magnetic field grows linearly. Therefore,
the magnetic field gives rise to Zeeman splitting for atoms that are located away from the
origin. The difference in energy due to Zeeman-splitting is proportional to

∆Ez ∝ µB mJ B, (45)

here µB is the Bohr-magneton and mJ is the projection of the total angular momentum
along a specified axis. Thus, when the atoms are not situated at the origin the Zeeman
sublevels are split. The energy level of the ground state, J = 0, remains the same because
it has no angular momentum, l = 0. The energy levels of the excited state, J 0 = 1, split
for z 6= 0, according to equation (45), with mJ 0 = −1, 0, +1.

25
Figure 9: Principle of a 1-dimensional MOT for a two-level atom. A linearly varying
~
magnetic field of the form B(z) = (0, 0, Bz) is applied. Figure adapted from [23].

In order for the transition, J = 0 → |J 0 = 1, mJ 0 = +1i, to occur, it must be driven by


σ + -polarized light. This is a direct consequence of the conservation of angular momentum.
By the same token the transition J = 0 → |J 0 = 1, mJ 0 = −1i needs σ − -polarized light to
occur. It can be seen from figure 9, that an atom located at position z > 0, scatters more
photons from the σ − -polarized beam than from the σ + -polarized beam since the transition
to the |J 0 = 1, mJ 0 = −1i state is much closer to resonance than the |J 0 = 1, mJ 0 = +1i
transition. The same principle holds for an atom at position z < 0, but in that case the
atom scatters more photons from the σ + -polarized beam. This mechanism gives rise to a
position-dependent force with a direction toward the center of the trap where B = 0.
The above-described principle of the MOT in one dimension can be easily extended
to two or three dimensions with the use of respectively four and six laser beams and a
magnetic field that linearly varies along two or three cartesian axes. In any case, the
magnetic field should be zero in the center of a MOT. For a MOT in three dimensions,
the required magnetic field is established by the arrangement of two magnetic field coils in
an anti-Helmholtz configuration, meaning that the current in both coils flows in opposite
direction. In that case a so-called quadruple magnetic field is created (see figure 10).
The density of the atoms in the MOT is limited by collision processes. If two atoms
collide while one of them is in an excited state, there is a possibility that the atom de-excites
into the groundstate, transferring the energy to the other atom as kinetic energy, thereby
heating the atom. Another limitation comes from the spontaneous emitted photons that
heat the atoms by causing an ~k-recoil when they are emitted. These emitted photons can
be absorped again by other atoms, leading to a repulsion between the atoms, which poses a
limitation on the density that can be created. The high densities needed for Bose-Einstein
condensates or Fermi-degenerate gasses can not be created with merely a MOT. In order

26
to achieve these phase transitions, a technique called evaporative cooling can be applied.
In this technique the capturing potential of the atom trap is gradually reduced, such that
the fastest atoms can escape, subsequently lowering the average energy of the atoms that
stay trapped. With evaporative cooling it is possible to lower the temperature sufficiently
and achieve the required phase-space density. An extensive explanation of this technique
can be found in [8].

Figure 10: A three-dimensional MOT formed out of three orthogonal pairs of beams with
opposite polarization. The beams intersect in the center of two coils in anti-Helmholtz
configuration where the magnetic field is zero. (Figure adopted from [12])

2.5.4 MOT’s for multilevel atoms


In order to explain the principles of optical molasses and magneto-optical traps, a two-
level atom was used for simplicity. Luckily, the trapping scheme described in the previous
section is not only applicable to J = 0 → J 0 = 1 transitions but also to any J → J 0 = J + 1
transition. However, the trapping scheme becomes a bit more complicated when real
atoms are considered. In the case of nearly all alkali atoms, the ground states are split
by the hyperfine interaction with the nucleus, and the excited states are split in four
hyperfine levels. Cooling and trapping of alkali atoms is achieved by using the transition
F = I + S → F 0 = F + 1, here F = I + S and not F = I + J because L = 0 in the ground

27
state. This transition is a so-called closed transition because any spontaneous decay to the
ground state is always to the same F ground state due to the selection rule ∆F = 0, ±1.
However, since other excited hyperfine states are close by, atoms incidentally get excited
to the F 0 = F state. Atoms in this state can decay to the F = I − S ground state. Since
the hyperfine splitting in the ground state is usually large, the trapping-beam is out of
resonance and therefore not able to excite these atoms, the atoms are no longer being
cooled and trapped. The solution to this problem is the addition of a second beam, called
the repumper-beam. This beam should be tuned to the F = I − S → F 0 = I − S + 1
transition. Atoms excited in this state can then decay to the original F = I +S groundstate.
How the above mentioned process is related to the case of 39 K atoms is described in section
3.6.

2.6 MOT dynamics


The number of atoms in a MOT is determined by the balance between the loading into
the trap and losses out of the trap

dN (t)
= L − ΓN (t), (46)
dt
where N (t) is the number of atoms in the trap as a function of time, L is the loading rate
in atoms/sec and Γ is the total loss rate in s−1 .
The loading rate depends on the loading mechanism. The MOT can be loaded from a
background vapor, or, if one aims for a higher loading rate and a higher number of trapped
atoms, with a beam from an atomic beam source, e.g., a 2D MOT. This last method is
used in this project, as will be discussed in section 3.7. Therefore the rate equation (46)
will be specified for this loading method. The loading rate is proportional to the beam flux
φbeam , L = αφbeam , with α representing the capture efficiency of the MOT.
The loss rate consists of several different loss processes [24],

N2
Γ = (φbeam + γ)N + β . (47)
V
The first term within the brackets accounts for the probability that trapped atoms escape
out of the trap due to collisions with incoming atoms from the atomic beam. The coefficient
γ accounts for trap losses due to colissions of trapped atoms with hot background atoms.
In order to reduce γ, the background pressure should be minimized. The coefficient β
accounts for trap losses due to collisions between two cold trapped atoms. In general,
the losses due to collisions between trapped atoms only play a significant role in dense
MOT’s. For most traps these β-losses can be neglected when compared to the losses due
to collisions with much hotter background atoms and atoms from the atomic beam; i.e.,
βN 2 /V  (φbeam + γ)N .
With the above conventions and assumptions equation (46) becomes

28
dN (t)
= αφbeam − (φbeam + γ)N . (48)
dt
Several important properties of a MOT can be extracted from this equation. The lifetime
of a MOT is determined by stopping the loading mechanism and looking at the decay of
number of atoms in the trap. If φbeam = 0 and the boundary conditions N (t = 0) = N0
and N (t → ∞) = 0 are taken into account, the solution to equation (48) is an exponential
decay curve,

N (t) = N0 e−γt (49)

The vacuum lifetime of a MOT is defined as


1
τ= . (50)
γ

Also, the loading time and the loading rate L, can be determined from equation (48). With
the boundary conditions, N (t = 0) = 0 and N (t → ∞) = N0 , the solution to equation (48)
is given by

αφbeam
N (t) = N0 (1 − e−(φbeam +γ)t ), N0 = . (51)
(φbeam + γ)

The loading time is defined as


1
t0 = . (52)
(φbeam + γ)

The initial rate of change of trapped atoms is givens by

dN
= N0 (φbeam + γ) = αφbeam . (53)
dt |t=0

29
3 Experimental setup
3.1 Chapter overview
In this chapter the experimental setup used in this project is discussed in detail. The
first topic discussed in this chapter is the vacuum system in section 3.2. A description of
the master laser from which all the laser beams are derived is given in section 3.3. The
spectroscopy setup and laser stabilization are addressed in sections 3.4 and 3.5, respectively.
This is followed by a description of the whole laser system in section 3.6. Finally, the setup
of the two-dimensional and three-dimensional magneto-optical trap is discussed in sections
3.7 and 3.8.

3.2 Vacuum system


In this section, a summary of the most important ascpects of the vacuum system is given.
A more detailed description can be found in the theses of T. Tiecke [25] and A. Ludewig
[26].
The part of the vacuum system that was used in this project consists of two chambers:
one for the two-dimensional MOT (source MOT) and the other for the three-dimensional
MOT (recapture MOT). A schematic representation of the vacuum system is depicted in
figure 11. The 3D-MOT chamber is cylindrical, it has a port on the top and the bottom
and eigth ports on the side, labeled in figure 11. Six ports (number 2,4,6,8 and the top and
bottom) are used for the MOT laser beams. The windows in the ports are all uncoated
and of optical quality.
An ultra-high vacuum with a pressure of P < 6 × 10−10 mbar is created by using an
ion pump (55 l/s Varian, Vacion Plus 55 Starcell) and a titanium sublimation pump.
The 2D MOT cell is connected via a differential pumping section to the 3D-MOT chamber
in order to obtain a large difference in pressure, port 5 in figure 11. It is a custom-made
glass cell by Techglass and has the form of a four-way cross, see figure 12. The windows
in the glass cell are optical quality windows (30 mm diameter) that allow the two pairs of
the 2D-MOT beams to propagate through the cell. Along one of the sides of the 2D MOT
cell, an ampule containing potassium enriched to an abundance of 6% 40 K is connected. In
this project we only worked with 39 K, but in the experiments of T. Tiecke and A. Ludewig
the potassium isotope of interest was 40 K, therefore they used an enriched potassium-40
ampule (Trace Science International).
To create the desired vapor pressure in the 2D-MOT cell, it is heated with heating
tape and covered with aluminium foil to maintain a uniform temperature. Thermistors are
used to monitor the temperature, which is kept at T ∼ 50◦ C.

30
Figure 11: A schematic representation of the vacuum system seen from above. a) 3D MOT
chamber, b) 2D MOT source chamber, c) potassium containing ampule, d) tube leading
to the 55 l/s ion pump, e) titanium sublimation pump, f) differential pump that connects
the 2D MOT chamber with the 3D MOT chamber. (Figure adapted from [25])

3.3 Master Laser


The master laser used in this project is a high power tunable single-mode diode laser DLX
110 from Toptica. The DLX 110 is an external cavity grating laser with a high power laser
diode; a maximum power of 500 mW at 767 nm can be realised. The laser is set up in
a Littrow configuration [27]. This means that the first diffraction order of the grating is
reflected back towards the laser diode, effectively forming an external cavity between the
front facet of the diode and the grating mirror. While free-running laser diodes often emit
light with a typical linewidth of almost 100 MHz, this external cavity laser has a typical
narrow linewidth of around 1 MHz. This is a result of the significantly longer resonator
length of the external cavity (about 1 cm) compared to the resonator length in the diode
itself (100 mm). Also, the finesse of the external cavity is higher. It follows from the
equation νf whm = νF SR /F (see section 2.4.3), that the linewidth of the laser frequency is
strongly reduced for high finesse values. The diffraction grating that is placed inside can
be controlled by applying a voltage to a piezoelectric transducer that is mounted on the
back of the grating. The wavelength of the emitted light can therefore be tuned by rotating
this diffraction grating. The mode-hop free tuning range is > 15 GHz. The three main
parameters that determine the frequency of the laser, the temperature of the diode, the

31
Figure 12: The 2D-MOT chamber, a custom-made glass cell. Along one side, the potassium
containing ampule is visible. Also, the magnet that initially broke the glass of the ampule
such that the potassium could enter the vacuum system can be seen. (Picture taken from
[25])

current through the diode and the angle of the grating, can be adjusted with the control
modules DTC 110 (temperature controller), DCC 110 (current controller) and the SCC 110
(scan control box). The light output is collimated by a specially designed collimator and a
cylindrical lens resulting in a collimated and round output beam. This is needed since the
output of the laser diode is elliptical and not collimated. The laser has two output beams,
as shown in figure 13, both beams have a nice TEM00 mode.

3.4 Experimental setup spectroscopy


The primary light source (Toptica DLX 110) operates at a power of 185 mW. A small part
of the power of the laser (2.8 mW) is used for spectroscopy, laser locking and monitoring
the stabilization of the laser frequency. The schematic of the set-up is presented in figure
13.
The laser beam coming from the side of the master laser passes through an acousto-
optic modulator (AOM) (ISOMET 1205-C1). The zeroth order beam (1.8 mW) is sent to
a confocal Fabry-Perot (FP) cavity, with a length of 150 mm (built by J. Minet [28]). A
piezoelectric transducer (PZT) is attached to one of the mirrors of the FP cavity. This
PZT is driven by a signal generator (TTi TG 550) at 10.52 Hz with a triangular signal
with a peak-to-peak amplitude of 9.7 V, effectively adjusting periodically the length of the

32
Figure 13: Optical setup of frequency modulation spectroscopy. The Toptica DLX 110
provides a narrowband frequency that is actively stabilized with the use of a Digilock 110
module on the spectral line of the 39 K 2 S1/2 F = 2 to the unresolved 2 P3/2 transitions.

cavity. This makes it a scanning FP cavity, allowing one to observe different modes. At
the end of the FP cavity a photodiode (Thorlabs DET 110) is positioned and connected to
an oscilloscope, where the transmission peaks of different transverse modes are detected.
If well aligned, a rejection of odd transverse modes in the transmission signal occurs,
(see section 2.4.5). Figure 20 demonstrates this. Achieving the rejection of odd modes
is not essential for the purpose of the FP cavity in this set-up, which is to monitor the
stabilization of the laser frequency passively. There is no feedback from the cavity to the
system to compensate for changes in frequency, it is merely a tool to monitor the quality of
the stabilization of the laser frequency and to be sure that the laser operates in single-mode
and not in multi-mode regime. A drifting laser frequency results in the movement of the
transmission peaks on the oscillator.
The first-order beam (1.0 mW) that comes out of the AOM (ISOMET 2105-C1, -67
MHz) is used for spectroscopy and laser locking. In order to get a well-resolved absorption
spectrum with peaks to which the laser can be locked, Doppler-free FM spectroscopy is
performed. Before the laser beam passes through a heated (T ∼ 45◦ C) potassium vapor cell,
the beam passes an aperture that effectively halves the power (to ∼ 500µW) and is split in
two by a polarizing beam splitter (PBS). One of the beams serves as a reference beam (13.2
µW). It passes through the vapor cell and it is sent to a photodiode (OPT 101), resulting

33
in a Doppler-broadened spectrum of potassium for which the hyperfine transitions are not
visible. The other beam is used for Doppler-free saturation absorption spectroscopy. This
beam passes through the vapor cell and after the cell through a neutral density filter before
it is retro-reflected through the cell again where both beams overlap with each other. This
gives the stronger saturation beam (470 µW) and the retro-reflected weaker probe beam
(8 µW) needed for Doppler-free saturation absorption spectroscopy. Finally, the beam is
sent to a photodiode (OPT 101). In this signal a Doppler-broadened line profile is present.
In order to get rid of the Doppler-broadening, the signals from the two photodiodes have
to be subtracted. This occurs in a home-built electronic subtraction box. The result is a
spectrum of only transition and cross-over peaks, without a Doppler-broadened line-shape.

3.5 Laser locking setup


To adjust the laser frequency, one has three parameters to work with. The current through
the laser diode, the temperature of the diode and the angle of the grating with respect to
the incident beam. The current is adjusted with a current controller module (DCC 110
Topica) and the temperature is adjusted with a temperature controller module (DTC 110
Topica). The angle of the grating is adjusted by applying a voltage to a piezo element
that is attached to the back of the grating. This can be controlled with software that
comes with the Digilock 110 module. When locking a laser, usually the current and the
angle of the grating are the parameters that are being changed. Because of the more
fragile mechanical character of the piezo element that adjusts the angle of the grating, the
frequency of the grating modulations is in general much lower than the frequency of the
current modulations. For instance, scanning the laser frequency over a certain frequency
range is done by applying a periodically changing voltage to the piezo element with a
typical frequency of 10 Hz and a typical peak-to-peak ampllitude of 2 V. However, for
really fast modulations of the laser frequency, e.g. FM spectroscopy in our set-up requires
a frequency of around 13 kHz and a peak-to-peak amplitude of 0.003 V, current modulation
is used to modulate the frequency.
As mentioned in section 2.3.3, an error function is required for laser locking, it is
given by the derivative of the subtracted signal. The measured voltage change of the
error signal is an indicator for the amplitude and direction of the frequency drift. It is
obtained through FM spectroscopy and a lock-in amplifier set at the right frequency. The
error signal is then placed in an proportional-integral-derivative (PID) feedback loop and
coupled to the voltage applied to the piezo element of the laser and to the current that
runs through the diode. In this experiment the above is done with the help of a feedback
controller (Digilock 110) from the firm Toptica. It has a built-in lock-in amplifier and two
PID channels for separate piezo and current feedback. The laser is locked by both feedback
loops; one slow integrator feedback to the piezo (∼ 1 Hz) and one fast integrator feedback
to the laser current (∼ 4 kHz). The feedback to the piezo is meant for compensating
slow thermal drifts, whereas the feedback to the current is meant for stability on shorter

34
timescales. The Digilock module has a graphical user interface where one is able to select
the desired locking point and lock the laser. A graph showing the quality of the locking in
time is presented in section 4.3. In order to build a 39 K MOT, the chosen lockpoint is the
transition from the 39 K |2 S1/2 , F = 2i state to the unresolved |2 P3/2 i states.

3.6 Laser system


Potassium differs from other alkalis in that the nuclear magnetic moment of its isotopes
are relatively small, especially for the bosonic isotopes 39 K and 41 K [29]. This leads to
small hyperfine splittings of the optical transitions. The energy levels of potassium-39
are shown in figure 14. It is seen that the splitting between the |F 0 = 3i and |F 0 = 2i
states is only 21,1 MHz, whereas the natural linewidth of the D2 line is 6.03 MHz; i.e.,
the hyperfine level spacings are comparable to the natural linewidth. This causes some
difficulties. If the frequency of the trapping laser is chosen to be a bit lower than the
0
|2 S1/2 , F = 2i → |2 P3/2 , F = 3i transition, the laser will have a frequency just above the
0
F = 1, 2 levels, meaning that some atoms will not be decelerated and cooled, as is the
intention of a MOT, but accelerated and heated.
Another effect of the small hyperfine splittings is that some atoms will be excited to
the |F 0 = 1, 2i state instead of the |F 0 = 3i state and could then spontaneously decay to
the |2 S1/2 , F = 1i state. Since the splitting between the two ground states is too big for
the trapping beam to excite the atoms in the lowest state, these atoms are lost without
a repumper beam. A way to overcome the difficulties is to tune the frequency of the
trapping beam and the repumper beam below all the excited-state hyperfine levels. In
such a way that the beams only differ in frequency by the ground-state hyperfine splitting
and both beams provide trapping and cooling forces [31]. See figure 14 for the trapping
and repumping scheme. The beam with the red-detuned frequency with respect to the
0
|2 S1/2 , F = 2i → |2 P3/2 , F = 0i transition is referred to as the trap-beam and the beam
0
with the red-detuned frequency with respect to the |2 S1/2 , F = 1i → |2 P3/2 , F = 0i tran-
sition as the repump-beam. However, since the |F 0 = 1, 2i states are excited with a similar
rate as the |F 0 = 3i state, a depletion from the |F = 2i ground state to the |F = 1i ground
state is quickly realised. Therefore, a strong repump beam is required and the cooling force
arises from both the trapping beam and repumping beam. Thus, the distinction between a
trapping beam and a repumping beam makes not much sence, but because of a widespread
convention they are labeled as such. The difference in frequency between the two ground
states is only 461.7 MHz, therefore it is possible to derive the repump light and the trap
light from the same master laser with the use of AOM’s. In figure 15 an overview of the
laser setup is shown.

35
Figure 14: Optical transitions of the D1 and D2-lines of 39 K. The transitions used for the
trap-, repump- and push beam are indicated. Here, ∆, is the detuning of the frequencies
with respect to the unresolved 2 P3/2 state. The numerical values are taken from [9] and
[30].

Figure 15: Optical setup of the laser system for 39 K. The Toptica DLX 110 is the mas-
ter laser from which all beams are derived. Three tapered amplifiers (TA) function as
amplification stages. Beam shaping optics are not depicted in this schematic.

36
The Toptica DLX 110 serves as a master laser and is operated at an output power
of 185 mW. Its power is distributed over four beams. One beam is used for the laser
locking procedure described in section 3.4 and three beams are used for the injection of
tapered amplifiers (TA’s) that are needed since the power of the master laser alone is
not sufficient for the creation of a MOT. The output beams of the TA’s are used for the
trapping beams of the 2D MOT (coming from ”2D-trap TA”), the trapping beams of the
3D MOT (coming from ”3D-trap TA”), the repumping beams of the 2D and 3D MOT
and the push beam (coming from the ”repump TA”). The three TA’s have the same chip,
an EAGLEYARD, EYP-TPA-0765-01500-3006-CMT03-0000. The power supply for the
3D-trap TA is a Sacher Pilot 2000 laser controller. The current through the chip is 2.4 A.
We have used input (output) powers of 40 mW (550 mW). Through the repump TA runs a
current of 1.9 A, provided by a homebuilt powersupply. For this TA input (output) powers
of 15 mW (615 mW) were used. The current through the 2D-trap TA is 1.9 A. The input
(output) powers are 10 mW (540 mW). All TA’s are temperature stabilized.
Three AOM’s are used to shift frequencies. The AOM used in the spectroscopy part of
the setup and the one used to create the required trap frequency are ISOMET 1205-1C
AOM’s, respectively shifting the frequency -67 MHz and -100.5 MHz. The third AOM,
used to shift the repump frequency to its required value, is a Crystal Technologies 3200-124
AOM, operating a +187.5 MHz in a double pass configuration.
Due to the sensitivity of the laser locking set-up, e.g. a small vibration can cause
the laser to unlock, the apparatus and optics used to build a 39 K MOT, is spread over
two optical tables. The master laser and the TA’s standing on one table, the vacuum
system and MOT chambers standing on another table. In this way one can work on the
vacuum system without unlocking the laser. In order to transfer light from the laser-table
to the MOT-table, 10 m long single-mode polarization maintaining optical fibers were used,
Schäfter + Kirchhoff PMC 630. These fibers were used in combination with Schäfter +
Kirchhoff fiber couplers (60FC-4-A6,2S-02) for in- and out-coupling. In order for fibers to
maintain the polarization of light, the polarized beams should be coupled along the proper
axis of the fibers; every fiber has its own optimal polarization maintaining axis. This
was done with the help of a polarization analyzer from Schäfter + Kirchhoff, SK010PA.
By stressing the fibers and observing at the same time the quality of the maintaining of
polarization, we were able to couple the light along the right axis. The coupling efficiency
into the optical fibres had a maximum of about 50%. The significant loss of power is
due to the fact that the output beams of the TA’s are not optimal Gaussian modes, but
rather disturbed versions with several transverse modes that can not be coupled into the
single-mode fiber that functions as a filter and only couples pure Gaussian mode beams.

39
3.7 K 2D MOT
The purpose of the two-dimensional magneto-optical trap (MOT) is to serve as a cold
atom source for the three-dimensional MOT. Several methods for loading a 3D MOT were

37
developed over the years. For instance, if a high number of captured atoms in the 3D
MOT is not essential for the experiment, the atoms can be loaded directly from the ther-
mal vapor. For a large number of atoms (more than 107 -108 ), more sophisticated loading
methods are needed, e.g. Zeeman slower, Low Velocity Intense Source (LVIS), 2D MOT,
etc. Two-dimensional MOT’s have proven to be successful cold atom sources in terms of
high loading rates and high number of atoms [26]. Therefore, we have chosen to build a
2D MOT as a cold-atom source in this project.

Figure 16: Schematic axial side view of the 39 K 2D MOT. The beams are setup in a retro-
reflected contruction in order to create counterpropagating beams with opposite circular
polarizations. The 2D magnetic quadrupole field is formed by two permanent magnets.
Figure adapted from [25].

In figure 16, a schematic picture of the 2D MOT is shown. An enriched potassium


ampule is connected to the vacuum cell. To increase the vapor pressure and to prevent
the potassium atoms from sticking to the glass walls, the glass cell is heated to about
50o C. This is done with heating tape attached to the glass cell and two Thorlabs TED
200C temperature sensors that monitor the process. The trap and repump beams are
overlapped and sent through a polarizing beam splitter, creating two pairs of trap and
repump beams. The four circularly-porlarized beams needed for the 2D MOT are produced
in a retro-reflected setup. This means that a beam, after passing through the vacuum glass
cell, passes a λ/4 waveplate, is retro-reflected by a mirror and passes again through the
same λ/4 waveplate and vacuum glass cell. Hence, meeting the MOT-requirement of
counterpropagating beams with opposite circular polarization.
The 2D magnetic quadrupole field that is required for a 2D MOT is provided by two
sets of permanent magnets. The magnets are made of Nd2 Fe14 B (Eclipse magnets N750-

38
parameter value
detuning trap laser -33.5 MHz ' -5.5 Γ
detuning repump laser -19.7 MHz ' -3.3 Γ
beam waist 18 mm
trap power per beam 70 mW
repump power per beam 30 mW
push beam power 6 mW
magnetic field gradient 20 G/cm

Table 2: The optimal 2D MOT parameters at which the 2D MOT was operated. The
detunings are given with respect to the unresolved 2 P3/2 states.

RB) and have a measured magnetization of 8.8(1) × 105 A/m [25]. Each set consists of two
magnets with dimensions 25 × 10 × 3 mm. Separated from each other by 12 mm, each set
forms a magnetic dipole bar of 62 mm total length. Both sets are placed 35 mm away from
the symmetry axis, together forming a two-dimensional quadrupole field with a gradient
of 20 G/cm along the optical axis. An extra beam, called the push-beam, was added to
the setup. It has the same frequency as the repump beams. This beam is directed along
the axial direction of the MOT; i.e., the direction in which the atoms are not confined.
Its function is to ”push” the atoms, as a result of the scattering force, into the 3D MOT
chamber, thereby increasing the number of atoms that can be captured in the 3D MOT.
The 1/e diameter of the push beam is about 2 mm. The optimal parameters at which the
2D MOT was operated are shown in table 2.

39
3.8 K 3D MOT
The three-dimensional magneto optical trap in our setup was originally designed to func-
tion as a dual system, capable of capturing potassium and lithium atoms, [25] and [26].
Therefore, most waveplates and polarising cubes are dichroic. However, in our experiment
the 3D MOT is used solely as a trap for potassium-39 atoms. The repump and trap beams
in the 3D MOT are derived from the same laser sources as the beams used in the 2D
MOT. The six counter-propagating and counter-polarized beams are produced, as in the
2D MOT, in a retro-reflected setup. Each beam contains two frequencies, the repump light
and trap light, which have the same frequency and detuning as the 2D MOT beams. The
beams are clipped at a 1/e diameter of 18 mm. The total power of the trapping (repump-
ing) beams is 130 mW (60 mW). This power could not be equally distributed over the three
retro-reflected beams due to the fact that the trap- and repump beams have opposite po-
larizations while being mixed with polarizing beam splitters. We chose to give the vertical
MOT-beam one half of the power (trap-beam: 65 mW, repump-beam: 30 mW ) and the
two horizontal MOT-beams both one quarter of the power (trap-beam: 32.5 mW, repump-

39
beam: 15 mW). Since the corresponding intensities of the horizontal MOT-beams are still
higher than the saturation intensity (trap-beam: I = 7.3I sat , repump-beam: I = 3.4I sat )
this was not expected to pose a problem.
The magnetic quadrupole field in the 3D MOT is produced by two coils in an anti-
Helmholtz configuration, this means that the current in the two coils flow in opposite
direction. The centres of the two coils are separated by about 110 mm. A current of
4.8 A through the coils gives rise to a magnetic field gradient of B 0 = 8.5 G/cm along
the z-direction [26],pand the value of the magnetic field at a certain point in space is given
B(x, y, z) = (B 0 /2) x2 + y 2 + 4z 2 . The power supply used to drive the current in the coils
is a Delta Elektronika SM15-200D. To prevent the coils from over heating, water cooling
is used.

3.9 Fluorescence imaging


The fluorescence coming from the MOT is an important quantity to measure since several
parameters of the MOT (number of trapped atoms, lifetime, loading time, loading rate
etc.) can be extracted from it. We make use of a CCD camera (The imaging source DMx
21BF04) to make images and video’s (30 fps) of the MOT and its fluorescence. A single
positive lens is used to focus the fluorescence onto the chip of the camera. The camera and
lens are positioned according to the thin lens formula
1 1 1
= + , (54)
f A B
where f is the focal length of the lens (5 cm), A is the distance between the MOT and
the lens (13 cm) and B is the distance between the lens and the camera (8.2 cm). We use
Maxim DL pro 4 software to control the parameters of the camera (i.e., exposure time,
gain, etc.) and to analyse the resulting images.

40
4 Results
4.1 Chapter overview
In this chapter the results of measurements that haven been taken are presented and
discussed. In section 4.2, the obtained spectroscopy signals are shown and explained. The
quality of the laser stabilization is characterised in section 4.3. Quantitative statements
about several MOT parameters are given in the last four sections. A calculation of the
total number of trapped atoms is made in section 4.4. In section 4.5, the number of trapped
atoms is plotted as a function of the magnetic field gradient. The lifetime of the MOT is
determined in section 4.6 and the loading parameters are given in section 4.7.

4.2 Spectroscopy results


In order to lock the laser frequency to a certain transition and to stabilize it, frequency
modulation (FM) spectroscopy was used. The principles of this technique are explained in
section 2.3.3. In order to obtain a useful error signal, several steps have to be taken. As
a starting point we looked at the signal produced with Doppler-free saturation absorption
spectroscopy (section 2.3.2), see figure 17. As described in section 3.4 the frequency of
the laser beam is scanned and sent through a sample of potassium. A photodiode collects
the light and converts the transmitted intensity into an electrical signal. Several hyperfine
0
transitions of 39 K are seen in figure 17. The F = 2 → F = 1, 2, 3 transitions on the
0
left side of the dip, F = 1 → F = 0, 1, 2 transitions on the right side of the dip and the
crossover signal near the dip are visible. These transitions are certain to come from the 39
isotope, since the frequency of the probe beam is more resonant with the 39 K transitions
than with the transitions of the other isotopes.
The contribution of the Doppler-width to the lineshape is still present in the above
shown picture. Therefore, it is possible to calculate the temperature in the sample of
potassium with the use of formula (21). The Doppler-width, ∆ωD , which is defined as the
FWHM of the signal, was obtained by fitting a Gaussian function to the curves in figure
17. The temperature that was calculated is T = 48◦ C. This temperature for potassium
in the glass cell corresponds well with the temperature that was found in [25]. However,
to perform FM spectroscopy, the Doppler-profile should be eliminated. Otherwise, the
derivative of the Doppler line profile would be taken into account in the error signal,
whereas only the derivatives of the transition peaks are useful for the error signal. To get
rid of the Doppler-width and thereby obtaining a flat spectroscopy signal, the absorption
signal of the reference beam was subtracted from the saturation absorption signal of the
probe beam, see section 3.4. The result is shown in figure 18. Now, with the flat Doppler-
free absorption signal, FM spectroscopy could be performed and the error signal; i.e., the
derivative of the narrow absorption features was obtained, as described in section 3.5 (see
figure 19). In this error signal, not only 39 K transitions are visible, but the 41 K transitions
as well. In the saturated absorption signal (figure 17) and the subtraction signal (figure 18)

41
Figure 17: The absorption signal as a function of the frequency obtained from the photo-
diode that detects the saturation absorption signal. The 39 K F = 2 and F = 1 transitions
to the unresolved states are clearly visible, as is the cross-over signal. The offset of the
frequency is with respect to the F = 2 transition, which is the chosen lock point for laser
locking.

these transitions are respectively not and almost not observable. The error signal is more
sensitive to transitions, due to its dispersive character. The 40 K F = 9/2 and the 40 K
F = 7/2 transitions to the unresolved states are too far from resonance and respectively
870.6 MHz and -415 MHz away from the 39 K F = 2 transition, as can be seen in the
potassium level scheme (figure 1), therefore they are not visible in figure 19. Since the
39 K F = 2 transition is the strongest transition in this signal we chose it as the locking

transition in our laser setup. This transition is stronger than the F = 1 transition because
the F = 2 state has more magnetic substates than the F = 1 state and at room temperature
the atoms are equally distributed over their mF substates. Therefore, more atoms are in
the F = 2 state and a stronger absorption peak for the F = 2 transition is expected.

42
Figure 18: Doppler-free absorption signal, obtained by subtracting the signals from two
photodiodes detecting the absorption signal from the reference beam and the absorption
signal from the probe beam. The spectroscopic transitions are labeled by their quantum
number of the hyperfine ground state.

Figure 19: Error signal obtained with FM spectroscopy. The transitions are labeled by the
quantum number of the hyperfine groundstate. The excited states are unresolved. The
F = 2 transition of 39 K gives the strongest transition signal and it therefore serves as the
locking point in our laser setup. The frequency offset is with respect to this point.
43
4.3 Laser-locking performance
FM spectroscopy allows us to lock the laser to a certain frequency. In this experiment we
chose the 39 K 2 S1/2 F = 2 to the unresolved 2 P3/2 transitions as the locking point. As can
be seen in figure 19, this transition is strongly represented in the error signal, hence being
a suitable transition for laser locking. In the spectroscopy setup (figure 13), a confocal
Fabry-Perot interferometer is placed to see if the laser operates in single- or multi-mode
and to monitor the frequency of the laser, whether it is stable or if there are drifts. The
Fabry-Perot cavity is a scanning cavity, meaning that the length of the cavity periodically
varies, see section 2.4.5. Because of this scanning function, we are able to see different
transverse modes in the transmission spectrum of the FP-cavity, see figure 20. A rejection
of the odd transverse modes should be visible when the cavity is well aligned (see section
2.4.5), and this is indeed the case. The laser beam that enters the cavity is coming from
the master laser, which gives a Gaussian mode (TEM00 ), but also passes through an AOM
(see section 3.4). The AOM negatively affects the quality of the Gaussian mode, making it
unlikely not to excite other transverse modes inside the cavity. Therefore, the odd modes
are not as strongly rejected as one might expect if the laser beam was to come directly
from the master laser or from a sigle mode optical fiber, as was done in [32].

Figure 20: The transmitted intensity of a laser beam that passed a scanning Fabry-Perot
interferometer. A photodiode converted the intensity into a voltage. The rejection of the
odd modes is visible. The offset of the frequencies is chosen with respect to the left odd
mode. The frequency spacing between two adjacent peak is half the free spectral range.

The frequency difference between two subsequent even modes or two subsequent odd
modes is the free spectral range, FSR = c/2L. However, the frequency spacing between

44
two adjacent even and odd modes is a half of the free spectral range, FSR/2 = c/4L, see
section 2.4.5. This information gives rise to a method to monitor the stabilization of the
laser frequency. Once the frequency spacing between two peaks is known, the displacement
of a peak is a measure for the change in frequency. In figure 21, a graph is depicted that
shows the quality of the laser locking in time. It is seen that the frequency of the laser
drifts significantly in the long term. However, within one minute, no significant frequency
drifts were detected. Since a characteristic loading time of a MOT is of the order of several
seconds, the quality of the laser locking meets the requirements for this experiment. For
more advanced experiments where a long term stability of the laser frequency is needed
the quality of the laser locking should be optimized. A way to improve the locking quality
is to adjust the PID-settings in the software interface of the Digilock 110.

Figure 21: Typical drift of the laser frequency as a function of time. Caused by the master
laser, measured with the Fabry-Perot interferometer.

4.4 Fluorescence measurement with a CCD camera


In order to give a quantitative statement about the parameters of the MOT one can make
use of fluorescence of the trapped atoms. During the decay of atoms from an excited state
to a ground state, photons are being emitted. The collection of these photons can be
achieved in several ways. A photodiode can be used to convert the collected photons into a
current which then gives a measure for the amount of photons that is collected. A positive
lens is used to focus the fluorescence light onto the photodiode. The distance between the
MOT, the lens and the photodiode are determined by the thin-lens formula. However, we
used a charge-coupled device (CCD) camera instead of a photodiode to obtain the desired
fluorescence signal. One of the main advantages of the camera is that it allows for a live
image on a computer screen, which makes the search for the fluorescence signal less labo-
rious. An image of the 39 K MOT is seen in figure 22.

45
Figure 22: A CCD image of the fluorescence of the 3D MOT. The exposure time of the
camera was set to 0.033 s. The lower-limit number of atoms in this MOT is calculated to
be 2.9(9) × 106 . The size of the cloud is around 2 mm, the Gaussian width of the MOT is
∼ 0.96 mm. The magnetic field gradient is set to 8.3 G/cm.

A line profile of the MOT fluorescence makes it apparent that the atom distribution
in the MOT is Gaussian, see figure 23. From this data the Gaussian width of the MOT
can be determined, it is found to be ∼ 0.96 mm.

46
Figure 23: The upper curve shows the line profile of the MOT fluorescence that is depicted
in figure 22. The solid (red) line is a Gaussian fit resulting in the Gaussian width of the
MOT of 95.44(99) pixels = 0.9567(99) mm. The lower data points result from a line profile
taken from a picture of a ruler with the same imaging setup. The dips in the data points
represent milimeter markings on the ruler.

While calculating the number of trapped atoms in the MOT, the following assumptions
were made:

• As long as the chip of the CCD camera is not saturated, it has a linear response to
the number of photons that it collects.

• The emitted photons are not absorbed by other trapped atoms.

For large and dense MOT’s the last assumption will not be justified, emitted photons will
be reabsorbed by other atoms in the trap. However, with this assumption made, a lower
limit for the number of atoms can be obtained.

47
The relation between the number of detected photons and the number of atoms in the
MOT is given by

ηNcounted
Natoms = , (55)
ΩΓsc t
where, Ncounted = the number of digital counts by the CCD chip, η = the conversion factor
that relates the number of digital counts to the number of collected photons, Ω = the solid
angle that is covered with the image. It determines the fraction of photons the camera
captures. Γsc = the scattering rate and t = the exposure time of the camera
The solid angle, Ω, is determined by the distance between the MOT and the lens
(R = 13 cm) and the diameter of the lens (d = 2.54 cm),

π(d/2)2
Ω= ≈ 0.0024. (56)
4πR2
The CCD camera chip converts photons into an electrical charge, which is again converted
into digital counts. A conversion factor that relates these digital counts to the number of
collected photons has to be determined. This is done by calibrating the CCD chip. A laser
beam with a measured power and frequency is sent to the CCD chip and an image with
a known exposure time is taken. In this way, the number of photons that the ccd camera
has collected can be calculated and the number of digital counts can be read out from the
analysis software (see section 3.9). The ratio between the two quantities is the conversion
factor. We found a conversion factor of η ≈ 900(200).
The determination of the scattering rate, Γsc , poses some difficulties. With the scat-
tering rate (formula (42)) the detuning and the saturation intensity enter the calculation.
For alkalis with a better resolved hyperfine structure than 39 K, trapping is accomplished
with a strong trapping beam and a much weaker repumping beam. Therefore, these atoms
can essentially be treated as two-level atoms and the scattering rate is well-approximated
by only taking into account the detuning and saturation intensity of the trapping beam.
Potassium-39, however, requires a more equal balance in powers of the trapping- and re-
pumping beam and are detuned by different values, see section 3.6. In addition, both
beams are detuned below the entire hyperfine manifold, meaning that atoms in different
states are interacted with. Therefore, the atom can no longer be treated as a two-level
system. Nonetheless, a lower-limit for the number of trapped atoms is found if the max-
imum value of the scattering rate, Γsc = Γ/2 is used in the calculations. Here, Γ is the
natural linewidth of the D2-line. The lower-limit number of atoms in the MOT was found
to be 2.9(9) × 106 atoms. It must be remarked that there is a significant uncertainty in
this calculation of the number of trapped atoms, due to two reasons. The first reason
has its origin in the determination of the conversion factor, the relation between the CCD
digital counts and the real number of photons. It was observed that the CCD chip has a
certain non-linearity which makes that the conversion factor depends on the light power
it receives. Since we could not determine what the light power produced by the MOT

48
is, we can only say something about the range where the value of the conversion factor
has to lie in. Second, the scattering rate that is used in the calculations provides only a
lower-limit number, since the maximum value of the scattering rate is used. Therefore,
such calculation is a rather pessimistic approximation for the atom number. If one would
take into account only one laser frequency (trapping or repumping beam) contributing to
the scattering rate, a higher number of trapped atoms is obtained. For example, with the
detuning and the saturation parameter of the repump beam implemented in the scattering
rate, the calculated number of atoms becomes 1.2(2) × 107 . An even higher number is
found if the parameters of the trap beam are used in the scattering rate, then a number
of 1.6(3) × 107 atoms is calculated. Keeping in mind that our MOT has a waist of around
1 mm (see fit results in figure 23), this number appears to be a better approximation if
compared to the typical MOT density (see for example [8]), which is of the order of 1011
atoms/cm3 . In order to obtain a more acurate approximation, the parameters from both
beams should be combined in the scattering rate and a more complex model, e.g. a six-level
model is required to describe the different states that are used for trapping. A description
of such a model is given in [31].
In the figures and results presented in this chapter we shall use the lower-limit value
for the number of trapped atoms, unless stated otherwise.

4.5 MOT parameters as a function of the magnetic field gradient


An important quantity that affects the MOT parameters is the gradient of the applied
magnetic field. It can be adjusted by varying the current through the coils. In figure
24, a measurement of the number of trapped atoms as a function of the magnetic field
gradient is shown. It is seen that the trap population is optimized for a B-field gradient
around 9 G/cm. For lower field gradients the trap depth is smaller, which results in a lower
number of trapped atoms. For higher field gradients the Zeeman shift increases too rapidly
away from the center of the MOT, causing the atom trapping to be less effective, since
the effective detuning changes to a smaller value that is less optimal for cooling. These
findings correspond well with a similar measurement that was done in [23].

49
Figure 24: The plot shows the steady state number of trapped atoms as a function of the
magnetic field gradient. The numbers are calculated according to the method discussed in
section 4.4. The data points correspond to fluorescence images taken with a CCD camera.
The optimal magnetic field gradient is found to be around 9 G/cm. The error bars result
from the uncertainty in determining the conversion factor η (see section 4.4).

The Gaussian width of the MOT as a function of the magnetic field gradient is shown
in figure 25, which demonstrates that the size of the MOT decreases with an increasing
magnetic field gradient. This is expected since the Zeeman shift is larger for higher B-field
values, resulting in a smaller trapping volume and therefore a smaller size of the trapped
atomic cloud.

50
Figure 25: The plot shows the Gaussian width of the MOT as a function of the magnetic
field gradient. The error bars result from the found uncertainties in the Gaussian fits
(see fit results in figure 23). The error bars for the extreme values of the field gradient
are significantly broader than the error bars for the center values. This is a result of the
quality of the images which becomes worse for extreme values of the field gradient. In
the case of the highest field gradients, the number of atoms is the smallest (see figure 24),
resulting in a low signal-to-noise ratio and therefore a big uncertainty in the Gaussian fit.
For the lowest field gradients, the density in the MOT is low (see figure 26), which makes
the signal comparable to the noise.

The information in figures 24 and 25 can be combined to obtain a number for the
density in the MOT as a function of the magnetic field gradient (see figure 26). For
simplicity, the Gaussian width of the MOT is used as the spherical radius. From the image
of the MOT in figure 22 it is seen that this is not entirely justified, since the shape of
the MOT is not a perfect sphere. However, this approximation should still result in the
correct order of magnitude for the density in the MOT. It looks like the maximum density
is reached for a magnetic field gradient, around 11 G/cm.

51
Figure 26: The plot shows the density of the MOT as a function of the magnetic field
gradient. The error bars result from the propagating uncertainties in the number of atoms
and the Gaussian width of the MOT.

4.6 MOT lifetime


The lifetime of a MOT was determined by switching off the loading mechanism and filming
the decay in fluorescence with a CCD camera. We switched off the 2D MOT laser beams,
including the push beam, until the MOT appeared to be empty. The lifetime is an impor-
tant measure for the amount of background pressure present in the vacuum chamber. A
high background pressure results in a short lifetime of the MOT. In figure 27, the measured
decay of trapped atom number is shown. Compared to the lifetimes of 39 K MOT’s that
are reported in several articles and studies [23],[31],[33], this lifetime is seen to be short,
an indication that the background pressure is really high. Hence, the dominant process in
the trap losses is the collision between trapped atoms with hot background atoms. The
losses due to the colissions between two trapped atoms can be neglected, since the density
in the MOT is low, see section 2.6. Therefore, the rate equation (48) can be used

dN (t)
= αφbeam − (φbeam + γ)N . (57)
dt

52
If the loading mechanism is switched off, the flux from the atomic beam coming from the 2D
MOT is zero, φbeam = 0. The solution for the decaying number of atoms is N (t) = N0 e−γt .
The lifetime, τ = 1/γ, was found by fitting this function to the decaying curve, see figure 27.
The lifetime of the MOT that could be extracted from the fit is τ = 0.102(2) s.

Figure 27: A typical exponentially decaying fluorescence signal measured by a CCD camera
filming with 30 fps is seen. The loading mechanism is switched off at t = 0. The solid
(red) curve that lies on top of the data is a fit of the function N (t) = N0 e−γt , where the
loss rate γ = 9.8(2) s−1 and the lifetime τ = 1/γ = 0.102(2) s.

53
4.7 MOT loading time
The loading curve of the MOT was obtained by first switching the magnetic field off, to
be certain that no atoms are being trapped, and then switching it back on, while a CCD
camera films the process until the MOT is fully loaded, see figure 28. As discussed in
section 2.6, the solution to equation (57) in the case of loading the MOT is

αφbeam
N (t) = N0 (1 − e−(φbeam +γ)t ), N0 = . (58)
(φbeam + γ)

By fitting this function to the measured curve, see figure 28, the loading time, T =
1/(φbeam + γ) is found to be T = 0.107(6) s.
Also the initial loading rate, L = dN/dt|t=0 = N0 (φbeam + γ), could be deduced. A value
of L = 2.8(1) × 107 atoms/s was found.
The small difference between the lifetime and the loading time implies that the loss rate
formed by the collisions between atomic beam atoms and trapped atoms is significantly
smaller than the loss rate caused by the collision between the background gas and the
trapped atoms; i.e., φbeam  γ.

54
Figure 28: The MOT loading dynamics. The plot shows a typical exponentially growing
fluoresence measured by a CCD camera filming with 30 fps. The loading mechanism was
switched on at t = 0. The solid (red) curve that lies on top of the data is a fit of the
function N (t) = N0 (1 − e−(φbeam +γ)t ), where (φbeam + γ) = 9.34(40) s−1 and the loading
time T = 1/(φbeam + γ) = 0.107(6) s.

55
5 Conclusions and Outlook
In the course of this six-months project, a two-dimensional- and a three-dimensional
magneto-optical trap for the bosonic isotope potassium-39 has been realised. In order
to obtain a stable frequency output from the master laser a spectroscopy scheme was built.
Frequency modulation spectroscopy was applied to a sample of natural abundant potas-
sium providing a spectroscopy signal that met the requierements for a good feedback signal.
In combination with the Digilock system, a sufficiently stable locking of the laser frequency
was achieved. In addition, the spectroscopy scheme has proven to be a robust one, little
maintenance was needed in order to keep it functioning. A confocal Fabry-Perot interfer-
ometer was used to monitor whether the laser operated in single-mode or multi-mode and
allowed us to detect drifts in the frequency.
In our set-up, the 2D MOT was used as a cold atom source for the 3D MOT. The
non-confining axis of the 2D MOT was placed exactly in front of the differential pumping
stage which connects the 2D MOT and the 3D MOT chambers and ensures a sufficient
pressure difference. It was found that an additional push beam was required to load the 3D
MOT. A 3D MOT without a push beam was not observed. In order to give quantitative
statements about the MOT parameters, a CCD camera was used to image the MOT. The
CCD camera proved to be a good instrument in obtaining the characteristic loading and
decaying curves of the MOT fluorescence. It was, however, less suited for the calculation
of the steady state number of atoms in the trap. This was due to the non-linearity of the
camera that brought a significant uncertainty in the calibration. Therefore, for a more
accurate calculation of the number of trapped atoms the CCD camera should be replaced
by a device that allows for a more reliable calibration.
The values of the lifetime and loading time of the MOT were obained from the decay-
ing and growing curves of the MOT fluorescence and could be measured quite accurately.
The lifetime, which is mostly limited by the background pressure, appeared to be short
(τ = 0.102 s) in comparison with other reports on 39 K MOT’s [23],[31],[33]. The loading
time was found to be T = 0.107 s. The small difference between the lifetime and load-
ing time indicates that the loss process due to collisions between the atoms in the atomic
beam and the trapped atoms is small. This implies that the dominant part of the loss
process comes from the collisions between the background gas in the vacuum chamber and
the trapped atoms. Since the vacuum chamber was reported [25] to be pumped to below
5 × 10−10 mbar and since it is a closed system, the background pressure is most likely to
come from background potassium atoms that are adsorbed by the walls of the vacuum
chamber during the years that experiments were done in the chamber and now cause a
significant background pressure. A high background pressure also explains the low steady
state number of atoms in the trap, calculated to be ∼ 1.2(2) × 107 atoms.
An important property of the MOT is its temperature. In order to measure it, a trig-
gered camera is required to determine the expansion rate of the atomic cloud after the laser
beams and the magnetic field are switched off. From these time-of-flight measurements the

56
temperature of the cloud can be extracted [34].
It should be remarked that there are several parameters left to be optimized in terms
of the lifetime and steady state number of the MOT. The optimal powers for the trap and
repump beams should be looked into, the optimal detuning of the laser beams should be
investigated and the optimal power of the push beam could be determined more accurately.
Optimization of these parameters could improve the quality of the MOT. Another improve-
ment could come from changing the retro-reflective setup into a non-retro-reflective setup.
Since the retro-reflected beam has less intensity than its source beam, due to the absorption
of photons by atoms in the MOT, the counter-propagating beams exert forces that are not
exactly balanced. This has a negative effect on the MOT. However, this inequality plays
only a significant role in dense MOT’s.
Based on the short lifetime, it is expected that a major improvement to the quality
of the MOT will come from baking the vacuum system in order to remove the adsorbed
potassium atoms and to decrease the background pressure.

6 Update 22-04-2013
After finishing this project in April 2012, Dr. Steven Lepoutre has optimised the lifetime
of the MOT. As was thought to be the cause of the short lifetime, it turned out that the
vacuum system was not in an optimal condition. Probably due to residual argon inside
the vacuum system, the ion pump needed a bake-out. Argon saturates the ion pump
and reduces the vacuum. The argon was introduced into the system to provide for a
protective atmosphere when the vacuum system was assembled [25]. The bake-out lowered
the background pressure in the MOT significantly. As a result, the lifetime of the MOT
has been increased from ∼0.1 s to ∼10 s.

57
References
[1] A. Einstein. Quantentheorie des einatomigen idealen gases, 2. abhandlung. Sitzungs-
berichte der Preussischen Akademie der Wissenschaften zu Berlin, 1925.

[2] M.H. Anderson, J.R. Ensher, M.R. Matthews, C.E. Wieman, and E.A. Cornell. Ob-
servation of bose-einstein condensation in a dilute atomic vapor. Science, 269, 1995.

[3] K.B. Davis, M.O. Mewes, M.R. Andrews, N.J. van Druten, D.S. Durfee, D.M. Kurn,
and W. Ketterle. Bose-einstein condensation in a gas of sodium atom. Phys. Rev.
Lett, 75(3969), 1995.

[4] D. S. Jin and B. DeMarco. Onset of fermi degeneracy in a trapped atomic gas. Science,
285(5434), 1999.

[5] S. Chu, J. E. Bjorkholm, A. Ashkin, and A. Cable. Experimental observation of


optically trapped atoms. Phys. Rev. Lett., 57(314), 1986.

[6] T.G. Tiecke, M.R. Goosen, A. Ludewig, S.D. Gensemer, S. Kraft, S.J.J.M.F. Kokkel-
mans, and J.T.M. Walraven. Broad feshbach resonance in the 6 li-40 k mixture. Phys.
Rev. Lett., 104(053202), 2010.

[7] Tobias Tiecke. Properties of potassium. 2011.

[8] Harold J. Metcalf and Peter van der Straten. Laser Cooling and Trapping. Springer,
1999.

[9] S. Falke, E. Tiemann, C. Lisdat, H. Schnatz, and G. Grosche. Transition frequencies


of the d lines of 39k, 40k, 41k measured with a femtosecond laser frequency comb.
Physical Review A, 74(3)(032503), 2006.

[10] Daniel Steck. Rubidium 87 d line data.

[11] W. Demtroder. Laser Spectroscopy Basic Concepts and Instrumentation. Springer,


1981.

[12] C.J. Foot. Atomic Physics. Oxford University Press, 2004.

[13] Fm spectroscopy with tunable diode lasers. Technical report, New Focus, 2001.

[14] Gaussian beams http://www.rp-photonics.com/.

[15] Gaussian beams http://www.optique-ingenieur.org/en/courses/.

[16] http://en.wikipedia.org/wiki/file:laguerre-gaussian.png.

[17] Ph 77 advanced physics laboratory. Technical report, Caltech.

58
[18] Cavities http://www.loitz79.de/physik
[19] A. Yariv. Optical Elecronics. Saunders College Publishing, 1991.
[20] D. C. Sinclair. Scanning spherical-mirror interferometers for the analysis of laser mode
structure. Spectra-Physics laser technical bulletin, (6), 1968.
[21] S. Chu, L. Hollberg, J.E. Bjorkholm, A. Cable, and A. Ashkin. Three-dimensional
viscous confinement and cooling of atoms by resonance radiation pressure. Phys. Rev.
Lett., 55(1), 1985.
[22] Johanna Nes. Cold Atoms and Bose-Einstein Condensates in Optical Dipole Poten-
tials. PhD thesis, Technischen Universität Darmstadt, 2008.
[23] Thomas Uehlinger. A 2d magneto-optical trap as a high flux source of cold potassium
atoms. Master’s thesis, Swiss federal institute of technology Zurich, 2008.
[24] J.Fortagh, A. Grossmann, T.W. Haensch, and C. Zimmermann. Fast loading of a
magneto-optical trap from a pulsed thermal source. Journal of Applied Physics, 84(12),
1998.
[25] T. Tiecke. Feshbach resonances in ultracold mixtures of the germionic quantum gases
6 Li and 40 K. PhD thesis, Universiteit van Amsterdam, 6, 2009.

[26] A. Ludewig. Feshbach Resonances in 40 K. PhD thesis, Universiteit van Amsterdam,


2012.
[27] L. Ricci, M. Weidemüller, T. Esslinger, A. Hemmerich, C. Zimmermann, V. Vuletic,
W. König, and T.W. Hänsch. A compoact grating-stabilized diode laser system for
atomic physics. Opt. Commun., 117, 1995.
[28] J. Minet. Stabilization in frequency of a laser diode. 2007.
[29] J.T.M. Walraven. Atomic physics: Lecture course. 2008.
[30] E. Arimondo, M. Inguscio, and P. Violino. Experimental determinations of the hy-
perfine structure in the alkali atoms. Rev. Mod. Phys., 49(1), 1977.
[31] R.S. Williamson III. Magneto-optical trapping of potassium isotopes. PhD thesis,
University of Wisconsin-Madison, 1997.
[32] S.V. Beentjes, G.W.J. Oling, and J.S. Reinders. Obtaining odd-mode rejection and
displaying higher order transverse modes in a confocal fabry-perot interferometer.
2012.
[33] J. Catani, P. Maioli, L. De Sarlo, F. Minardi, and M. Inguscio. Intense slow beams of
bosonic potassium isotopes. PRA, 73(033415), 2006.

59
[34] T.M. Brzozowski, M. Mazynska, M. Zawada, J. Zachorowski, and W. Gawlik. Time-
of-flight measurement of the temperature of cold atoms for short trap-probe beam
distances. Journal of Optics B, 4, 2002.

60
Acknowledgements
In the past six months I have been incredibly lucky to have had the help of some people
who I would like to thank here.
First of all, I would like to thank my supervisor Prof. dr. Jook Walraven who gave
me the opportunity to take part in his research group and to work in his fantastic lab. I
am grateful for his inspriring lectures that were the reason for me to begin a project in
the field of laser cooling and trapping and for the fact that his door was always open and
questions could always be asked.
I was spoiled to have the full support from Dr. Vyacheslav Lebedev en Dr. Steven
Lepoutre. We had a great time in the lab, the coffee break, de polder and paradiso. I want
to thank Slava for treating me as an equal, for the patience he had while introducing me
to optics and lasers and for all the hours he invested in me. He showed me how a real
experimentalist works and let me believe that all experimental difficulties can be overcome.
I painfully discovered during his three-weeks absence, that this is certainly not always the
case. I also would like to thank Steven, who impressed me with his perserverance to build
the best experimental setup possible. I thank him for learning me that the quality of the
setup is the key to succes and for all those hours he invested in setting up the MOT.
Finally, I want to thank my dear friend Jannis Teunissen for saving me loads of time
with his superhuman computer skillz.

61

Anda mungkin juga menyukai