Professional Documents
Culture Documents
A thesis submitted for the degree of Bachelor of Science Advanced with Honours in the School of Physics of Monash University
November 5, 2010
ii
iii
Abstract
An integral part of a Bose-Einstein condensate (BEC) machine is the magnetooptical trap (MOT), where a cloud of atoms can be trapped and cooled by a spatially modulated force from atom-photon momentum exchange. From this cloud of atoms, a BEC can be formed by evaporative cooling in a conservative trap. This thesis presents three components required for BEC production: variable beam expanders for wide, intensity balanced laser trapping beams; an ultra-high vacuum system; and a high resolution objective lens for BEC imaging. Six beam expanders for magneto-optical trapping were constructed, in which the magnication of the beam can be continuously varied between 4.7 5.3, without losing collimation, by adjusting the position of a single lens. This allows the peak intensity to be varied by up to 24 % with no loss in total power. The construction of an ultra-high vacuum system is discussed, which reached a pressure of 2 108 Torr. Having obtained this pressure, it is expected that after bakeout the pressure will reach 1012 Torr. Lastly the design of a high resolution objective lens is presented, which has a resolution of 0.9 m with a diraction limited eld of view of 340 m. Unlike commercial microscope objectives, this design corrects for spherical aberration caused by the glass cell wall, which allows the objective lens to be located outside of the vacuum system. The objective lens contains four elements made of SF11 glass, and is insensitive to element spacing variations of < 0.1 mm.
iv
Acknowledgements
This thesis would not exist without the help, support and contribution of many people. I would like to thank: Lincoln Turner for his unfailing energy and enthusiasm, Russell Anderson for all of his assistance throughout the year The mechanical workshop, especially Stephen Downing, for their hard work in constructing various items for the project. Without Stephen, the vacuum system would have literally fallen apart! Nino Benci and the electronic workshop, Everyone in the Monash BEC group for providing a supportive team environment, Phil, for his love and encouragement, and nally all the Honours students for making this such a fun year!
vi
Contents
1 Introduction 1.1 Trapping in the MOT . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 Optical molasses . . . . . . . . . . . . . . . . . . . . . . 1.1.2 The MOT - a damped harmonic oscillator . . . . . . . . 2 Variable beam expanders for magneto-optical trapping 2.1 The MOT trapping beams . . . . . . . . . . . . . . . . . . 2.2 Beam expansion methods . . . . . . . . . . . . . . . . . . 2.2.1 A single collimation lens . . . . . . . . . . . . . . . 2.2.2 Keplerian and Galilean beam expanders . . . . . . 2.2.3 Galilean variable beam expanders . . . . . . . . . . 2.3 Design of the MOT beam expanders . . . . . . . . . . . . 2.3.1 Theoretical design . . . . . . . . . . . . . . . . . . 2.3.2 Simulations in OSLO . . . . . . . . . . . . . . . . . 2.4 Construction of the prototype . . . . . . . . . . . . . . . . 2.4.1 Diculties in obtaining straight beam propagation 2.4.2 Diculties in bre collimation . . . . . . . . . . . 2.4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Construction of the MOT beam expanders . . . . . . . . . 3 The ultra-high vacuum system 3.1 The preliminary vacuum system . . . . . . . . 3.1.1 Vacuum pressure before bakeout . . . . 3.1.2 Vacuum pressure after bakeout . . . . . 3.2 Design of the UHV system . . . . . . . . . . . . 3.3 Construction of the UHV system . . . . . . . . 3.3.1 Evacuating the system . . . . . . . . . . 3.3.2 Incorporating the MOT beam expanders
1 2 2 5 9 9 10 10 11 12 14 14 16 19 20 20 21 22 25 25 26 27 28 29 31 33
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
4 A high resolution BEC imaging system 35 4.1 Imaging system requirements . . . . . . . . . . . . . . . . . . . 36 4.1.1 Spherical aberration . . . . . . . . . . . . . . . . . . . . 36 4.1.2 Lens design software . . . . . . . . . . . . . . . . . . . . 38 vii
viii 4.2 4.3 An imaging system using catalogue 4.2.1 Initial results . . . . . . . . A custom objective lens . . . . . . 4.3.1 The nal design . . . . . . lenses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contents . . . . . . . . . . . . 39 40 41 44 49 51 54
ix
Chapter
Introduction
The study of Bose-Einstein condensation is currently an active eld of physics. The 2010 International Conference on Atomic Physics (ICAP) demonstrated that there is intensifying interest in this area, both experimental and theoretical. This thesis examines the construction of part of a machine used to create Bose-Einstein condensates, known as the magneto-optical trap. A Bose-Einstein condensate (BEC) is a state of matter rst predicted in 1924 by Nath Bose and Albert Einstein [1, 2][3, p.1]. Seven decades later, technology and cold atom physics advanced to the stage where the rst BEC was created. This was achieved in 1995 by Carl Wieman, Eric Cornell and Wolfgang Ketterle [4]. If a dilute gas of weakly interacting bosons in an external trapping potential is cooled to several hundred nanokelvin, the bosons begin to occupy the lowest energy level of the trap potential. Bosons, unlike fermions, can multiply occupy the same energy level. Once one boson occupies the lowest energy level, it becomes energetically favourable for more bosons to occupy the same level. In this manner, when a critical temperature is reached a cascade occurs where the majority of atoms in the trap condense into the one energy level, forming a BEC. This process is shown in Figure 1.1(a). Once a BEC has formed in the ground state of the trapping potential, it can then be raised to an excited state of the potential. An example of a BEC in an excited state is shown in Figure 1.1(b), where quantised vortices are observed to form a regular lattice pattern [6]. The goal of the Spinor BEC Laboratory at Monash University is to perform magnetic resonance imaging (MRI) on BECs to observe topological defects in the trapped condensate. To form a BEC, a cloud of cold atoms must rst be held in a trap, from where the temperature can be lowered further. This is achieved using a magneto-optical trap (MOT). The MOT was developed by Steven Chu et al. in 1987 [7], and consists of three pairs of orthogonal, counterpropagating laser beams and a magnetic eld gradient increasing in strength from the centre of the trap outwards. Many MOTs have been made which trap 108 atoms, which results in a BEC of 105 atoms [8, 9]. Larger atom number MOTs of 1010 atoms have been made, from which a BEC of 107 atoms can form [10, 11]. Our aim was to build a MOT which can trap 1010 atoms with a high loading 1
Chapter 1. Introduction
(a)
(b)
Figure 1.1: (a): The velocity-distribution graph of the rst Bose-Einstein condensate (BEC), showing the characteristic sharp peak at the origin, created in 1995 [4, 5]. (b): Observation of a quantised vortex lattice in a BEC [6].
rate, which can reliably and rapidly produce a BEC of 107 atoms held in an optical dipole trap. We require a large BEC because they produce a higher signal-to-noise ratio, resulting in high-precision measurements, and rapid production of BECs allows more experiments to be performed in a shorter space of time. While we consider MOTs only in the context of creating BECs, this is not the only application of a MOT. Large densities of cold atoms are also required in studies of photon localisation, cold collisions and the formation of molecules in cold atoms [11]. The work discussed in this thesis can be grouped into three main sections: the Galilean variable beam expanders used to form the wide trapping beams for the MOT; the construction of the ultra-high vacuum system within which the atoms are trapped; and the design of an imaging system to resolve structure in the BEC on a scale of 0.9 m. This chapter introduces the theory behind trapping atoms in the MOT.
(a)
(b)
Figure 1.2: (a) Schematic of a typical MOT, showing six counter-propagating, orthogonal, circularly polarised laser beams, with two coils in an anti-Helmholtz conguration producing a magnetic eld [12, p.192]. This magnetic eld is zero at the centre and linearly increases in strength with radial distance. (b) Contour lines from two anti-Helmholtz coils [13]. Here Z is the distance along the axis of the two coils, the vertical direction shown in (a), and R is the radial distance from the z-axis. At the centre of the trap, the magnitude of the magnetic eld increases outwards linearly and has a steeper gradient along the z-axis. This steep gradient is usually oriented vertically, so that the greater trapping force counteracts the force of gravity.
this occured when the atom was travelling towards the photon, then the atom would be slowed by this momentum exchange. The frequency of the photon determines whether this interaction occurs in the correct direction for slowing. If the photon frequency corresponds to an atomic transition, then the atom will absorb the photon. This is known as the resonance frequency. Due to the Doppler eect, the photon frequency will be dierent in the reference frame of the atom, depending on the direction and velocity the atom is travelling with respect to the photon. If the photon frequency is tuned to the atomic resonance only when the atom is moving towards the photon, then the atom will be slowed. Now consider a laser beam incident on an atom. The rate of atom-photon interaction will increase with the beam intensity, and will increase as the photons approach the resonance frequency in the atomic rest frame. The net force on an atom due to the laser beam is F = hk s0 /2 , 1 + s0 + [2( + D )/]2 (1.1)
where is the reduced Planck constant, s0 = I/Is is the ratio of the laser h
Chapter 1. Introduction
Figure 1.3: Prole of the force in an optical molasses as a function of atom velocity [15, p.88]. For small velocities, the force is linear, as shown by equation 1.4, however for larger velocities the slowing force decreases. When we include a magnetic eld, the force on the atoms will have a similar prole, but will be spatially dependent rather than velocity dependent [15, 157].
intensity I to the saturation intensity Is of the atomic transition, is the linewidth of the atomic transition, is the laser detuning from resonance and D is the Doppler shift of the laser in the atomic reference frame [15, 16]. The laser detuning = l a is the dierence between the laser angular frequency l and the atomic resonance angular frequency a . The Doppler shift is D = k v where k is the wavenumber of the laser and v is the velocity of the atom in the laboratory frame. We see from this equation that if two identical counterpropagating laser beams were directed onto a stationary atom, then the net force on that atom would be zero, since the wavevectors k have opposite direction. However, the net force is non-zero for the case where the lasers are tuned slightly below resonance, and the atom is moving slowly along the axis of the beam. As the atom moves in one direction, the oncoming laser beam is blue-shifted (in the atomic rest frame) towards resonance, while the other laser beam is red-shifted. This means that the atom will absorb more photons from the oncoming laser than from the co-propagating laser, and as such a greater force will be exerted by the oncoming laser, resulting in a non-zero damping force, or optical molasses force, FOM . If the lasers were tuned above resonance rather than below, then the atom would be accelerated instead of slowed, which would not be the desired eect in an optical molasses. Equation (1.1) can be extended to include the two directions of the beams, to give s0 /2 F = k h , (1.2) 1 + s0 + [2( |D |)/]2 such that the optical molasses force is the sum of the forces in both directions; FOM = F+ + F . (1.3)
1.1. Trapping in the MOT Expanding the denominator of (1.2) as a Taylor series, we then obtain FOM = 8 k 2 s0 v h (1 + s0 + (2/)2 )2 v,
(1.4)
where we have assumed small velocities [15, p.88]. This is a velocity dependent force which is linear for small velocities (Figure 1.3). While this analysis only considered a one-dimensional case, it can be extended into three dimensions using six counter-propagating orthogonal laser beams instead of two.
Chapter 1. Introduction
(a)
(b)
Figure 1.4: (a) The ground and rst excited state of 87 Rb, where each level has then been magnied to show the hyperne structure [18]. The trapping beams in the MOT are tuned to (i) the cyclic transition, while a repump laser (ii) causes atoms to enter the cyclic transition. (b) The splitting of the hyperne structure with increasing magnetic eld strength [19].
and hence the net force on the atom, increases as the atom moves out from the centre of the trap, and is always directed towards the centre. Once the detuning reverses sign, the probability decreases. The restoring force on the atom therefore has a similar form to the force due to the optical molasses (Figure 1.3), except this force acts in position space rather than momentum space. Notice that as an atom crosses the centre of the trap, it will no longer be in the necessary state to be trapped by the oncoming laser. This does not result in a problem, because the atom will be optically pumped into the trapping transition after several photons have been absorbed. The total force on the atoms due to both the optical molasses and the magnetic eld is F = v r, (1.5) where was dened in equation (1.4), r is the distance from the trap centre and A = , (1.6) h k where is the magnetic moment of the atomic transition and A is the magnetic eld gradient. We see from this equation why, in a rst order approximation, the MOT can be regarded as a damped harmonic oscillator, with
(a)
(b)
Figure 1.5: (a) The energy levels of 87 Rb used for trapping in the MOT, with increasing magnetic eld strength. The scale is arbitrary, and the diagram indicates the relative dierence in splitting between the energy levels. Using this diagram, and the knowledge that the magnetic eld strength is linear with distance in the MOT, the change in detuning with distance can be found, and is shown in (b), again on an arbitrary scale. The background shading indicates the magnitude of the trapping force on the atoms, with darker shading indicating stronger force. This force is directed towards the centre of the trap.
as the damping coecient and as the spring constant. To construct a MOT, wide trapping beams and an ultra-high vacuum system are required. When the MOT is constructed, and BECs are condensed from the cloud of trapped atoms, a method of imaging the BEC is also needed. In chapter 2, we present the Galilean variable beam expanders used to form the wide, intensity balanced MOT trapping beams. The ultra-high vacuum system was constructed, discussed in chapter 3, and nally the design of a high resolution BEC imaging system is presented in chapter 4.
Chapter 1. Introduction
Chapter
10
Figure 2.1: The number of atoms N trapped in a MOT increases with beam diameter . The straight line shows the relation N 3.3 , which agrees with the data for up to 1010 atoms [11].
(2.1)
where w is the 1/e2 beam radius1 , w0 is the beam waist (radius of the beam at the bre exit) and z is the propagation distance from the beam waist (see
1
11
Figure 2.2: A Gaussian beam will diverge from a waist of 1/e2 radius w0 to a radius of w(z) after a propagation distance of z, as described by equation (2.1).
Figure 2.2)[21, p.595]. If a convex lens is placed on the optical axis one focal length from the beam waist, or the end of the bre, then the lens will form a collimated beam. In our case we require a beam diameter of 30 mm, as will be discussed in section 2.3.1. For light with wavelength 780 nm exiting a bre with a mode eld diameter of 5.3 m [Thorlabs P3-780-PM-FC-2], the required focal length to obtain a collimated beam with a 30 mm diameter is 160 mm. Not only does this method result in a long collimation length, it means that the width of the beam cannot be adjusted without losing collimation, unless the lens is changed to one with a dierent focal length and re-positioned with respect to the beam waist accordingly. This is not a simple operation. To get a circularly polarised beam, a quarter-wave plate needs to be placed in a region where the beam is collimated, otherwise dierences in optical path lengths across the beam would result in a non-uniform elliptical polarisation. We are using polarisation maintaining (PM) bres which ensure that the output is linearly polarised, so we only need to use a quarter-wave plate to convert to circular polarisation. In this case, the quarter-wave plate would be located after the collimating lens, and would have to be greater than 30 mm in diameter. A quarter-wave plate of this size becomes expensive for example, CVI Melles Griot sell a multiple-order quarter-wave plate only 30 mm in diameter for $730 [22]. So although using a single lens to produce a wide collimated beam is a very simple design, the size of the beam cannot be adjusted, the distance of free-space propagation takes up unnecessary room in the experimental set-up, and including a quarter-wave plate makes the design expensive.
12
Figure 2.3: A standard (a) Keplerian beam expander and (b) Galilean beam expander. In the Keplerian system, the focus allows for spatial ltering, but can cause air breakdown in high power cases. The total length of the system is shorter in a Galilean design.
collimated beam is initally needed, so if the optical system did not already produce a collimated beam, an additional collimating lens would need to be placed before the beam expander. This lens could have a much shorter focal length than required if no beam expander was used, so the total length of the system would be shorter than when using a single collimation lens. Using a Keplerian expander has the advantage that the focus allows easy spatial ltering of the beam, while the Galilean expander produces a wide beam over a smaller propagation distance than either the Keplerian expander or a single collimation lens. We chose to use a Galilean beam expander both for the shorter propagation distance and because air turbulence at the focus of a Keplerian expander can produce scintillation of the beam. However, we modied the design to allow a variable output beam width which maintains collimation. This is called a Galilean variable beam expander.
13
(a)
(b)
Figure 2.4: Schematics of a (a) symmetric and (b) non-symmetric Galilean variable beam expander. If both rear lenses are moved as shown, the beam magnication can be continuously changed while maintaining collimation. If the input and output lens have equal focal lengths, then the system is centred at a magnication of one. If the input lens has a shorter focal length than the output lens then the system is centred at a magnication greater than one.
a symmetric variable beam expander. When the total length of the system is at a maximum, then the output beam has the same size as the input beam, and the system is said to be centred at a magnication of one. Such a system is shown in Figure 2.4(a). If the rear lens has a shorter focal length than the output lens, then the system is non-symmetric. When the system is at a maximum length, the output beam will be wider than the input beam, and the system is then centred at a magnication greater than one. An example of this is shown in Figure 2.4(b). In practice, it is preferable to keep the number of variable positions or degrees of freedom to a minimum. Ideally, only one lens should need to be adjusted to change the beam width with a minimum of collimation loss. To achieve this, we designed a lens system that produced the desired beam magnication when the rear lens is located at the maximum distance from the xed front lens. In this position, in order to produce a small change in the beam magnication the rear lens would only need to move position slightly to keep the beam collimated. But in our design, the rear lens is kept xed for simplicity, and only the central lens is adjustable. When the magnication is changed by moving the single lens by a small distance (dened in 2.3.2), the beam size does not change visibly over a distance much greater than the
14
Figure 2.5: A diagram of the optical components of the Galilean variable beam expanders (not to scale).
length of the MOT, and so it is considered to remain well collimated. The details of the design of our MOT beam expanders are discussed in the next section.
where d1 , d2 are lens spacings and M = r1 /r2 where r1 , r2 are beam radii as shown in Figure 2.6 [25]. However, these two equations do not specify whether the design is symmetric and centred about a magnication of one, or non-symmetric and centred about a magnication of ve as required. When the design is centred about the desired magnication, the total length of the system is maximised as the magnication is adjusted. This sets the condition that (d1 + d2 ) = 0, (2.4) M
15
Figure 2.6: Diagram showing the lens focal lengths f , separations d and beam radii r of a Galilean variable beam expander. These quantities are related by equations (2.2) and (2.3).
which, when applied to equations (2.2) and (2.3) yields the relation f1 = M f3 . (2.5)
Notice that when M = 1, the focal length of the output lens is the same as that of the input lens, as is indeed the case for a symmetric variable beam expander system. For the system to instead be centred on a magnication of ve, the focal length of the output lens must be ve times longer than the input lens. This equation does not contain f2 the focal length of the concave lens determines the overall length and distances between the lenses. To our knowledge, this simple relation between the focal lengths of the rst and third lens in a Galilean variable beam expander has not been documented elsewhere. Using equations (2.2) and (2.3) and the condition (2.5), it was possible to then nd a combination of standard lenses to produce a magnication of ve with reasonable distances between the lenses. The lenses in the nal design have focal lengths of f1 = 150 mm, f2 = 9 mm, and f3 = 30 mm,
which were available as catalogue items [Thorlabs LA1417-B, LD2568-B and LA1027-B]. Equations (2.2) and (2.3) give corresponding separation distances of d1 = 132 mm, and d2 = 12 mm. The distance between the lenses was the main limiting factor in the design process. Often a combination of lenses would be found that would result in either d2 being negative, or so small that once the lens thickness was accounted for there was little or no space left between the lenses for movement. If d2 was increased, then d1 would be so large that it oered little improvement in propagation distance over using a single collimation lens. These problems resulted in a magnication of 6 being rejected, as a satisfactory combination could not be found. For example, with a magnication of 6, three possible lenses were f1 = 150 mm, f2 = 10 mm and f3 = 25 mm, however the distances between these lenses were d1 = 130 mm and d2 = 5 mm. The distance d2 was far too small to be achieved, since the distance from the optical centre to the front of
16
Figure 2.7: A graph of the total length (d1 + d2 ) of our Galilean variable beam expanders as a function of magnication, using lenses with focal lengths f1 = 150 mm, f2 = 9 mm and f3 = 30 mm. With this lens combination, the total length is maximised at a magnication of ve. This means that for small variations around a magnication of ve, the total length can remain constant, and hence the rear lens can be xed with minimal loss to collimation.
a 1 Thorlabs f = 25 mm lens is 7.7 mm. This would require the two lenses to be overlapping, a slight physical impossibility! It was conrmed that the design was centred at a magnication of 5 by graphing the total length of the system as a function of magnication, for the chosen values of f1 , f2 and f3 . Figure 2.7 conrms that the total length of the system is maximised at a magnication of ve, and that there is little change in the total length at this position for small variations in magnication. While nding the desired combination of lenses, certain relations were observed between the focal lengths of the lenses and the distances between them. If M and f2 were xed, and f1 and f3 satised the relation (2.5), then it was observed that f1 = d2 = M d1 . (2.6) If instead f2 was the only focal length allowed to vary, then it was observed that d2 d1 f2 = = . (2.7) 2 2 With this knowledge, the best combination of lenses was found eciently.
17
In OSLO, a surface is specied by its radius of curvature, material, aperture radius and thickness to the next surface along the optic axis. This means that to correctly simulate the beam expander system the distances between the surfaces of the lenses (s1 and s2 ) need to be known - this is in contrast to d1 and d2 , which are the distances between the optical centres of the lenses. For example, the beam expanders use a plano-convex f = 30 mm lens. The distance from the focal point of this lens to the optical centre is 30 mm, while the distance from the focal point to the centre of the curved surface is 27.1 mm. Calculating the surface separations of the lenses yielded s1 = 125.7 mm and s2 = 7.6 mm. Using these separations, the simulated beam expander did not give satisfactory results. The output beam was not collimated but rather converging to a focus 4 m from the system, and the beam was magnied at the surface of the nal lens by a factor of only 3.3 (Figure 2.8(a)). Interestingly, these results were better than if the rear lens was reversed so that the at side faced the concave lens. In that case, when the new separation was calculated, the output beam converged to a focus 3.8 m away with a magnication of only 2.3 (Figure 2.8(b)). This is contrary to the convention of placing the convex side of the lens facing the collimated beam, and the at side of the lens facing the focused beam, to minimise the angle between the rays and the surface. These results are not surprising when we consider that the equations used to nd d1 and d2 assume thin, ideal lenses, and do not take into account the thickness of the lenses used. With this knowledge it seemed likely that the separations of these lenses could be adjusted, without changing curvatures, so that a collimated beam with a magnication of ve could be achieved. This was done manually in OSLO until the best result was found. When the lenses are separated by s1 = 118.2 mm and s2 = 2.2 mm, the beam is magnied by a factor of 4.97 and is well collimated, coming to a focus 148 m away from the system (Figure 2.9(a)). The specications of this design are shown in Table 2.1. Again, if the rear lens is reversed, the performance is diminished, with a focal length of only 8 m and a magnication of 3.1 (Figure 2.9(b)).
Having found the ideal combination of focal lengths and lens separations, the eect of varying the magnication was simulated. The bi-concave lens is mounted on a translation mount [Thorlabs SM1Z] with a 1.5 mm range of movement. By keeping the total length of the system constant, and moving the bi-concave lens forward and backwards by 0.75 mm, the changing characteristics of the lens system was calculated in OSLO. These results are shown in Figure 2.10. When the central lens was moved 0.75 mm backwards, the beam remained fairly collimated, with a focal length of 18 m, and the magnication increased to 5.3. When the lens was moved forward by 0.75 mm the beam had a focal length of 16 m, and the magnication decreased to 4.7. In both cases, after the change in magnication the beam size did not change visibly across a distance of 4 m. As this is much greater than the distance across the MOT ( 0.3 m), the beam can be considered to remain well collimated. This total
18
(a)
(b)
Figure 2.8: OSLO raytrace of the three chosen lenses with optical centres separated by the distances calculated using equations (2.2) and (2.3). We found that with these separations, the system did not produce a collimated beam, and the magnication of the beam near the output lens was less than ve. The rst lens is plano-convex and was simulated in both possible orientations. We found that (a) placing the convex side towards the concave lens yielded a higher magnication and longer focal length than when (b) the plane side was placed towards the concave lens. These simulations caused us to alter the lens separations to achieve better results.
(a)
(b)
Figure 2.9: The results of manually adjusting the lens separations in OSLO to give a magnication as close to ve as possible, with the best collimation. In Figure (a), the system has a very long focal length of 148 m, and as such is well collimated over the distance of a MOT trapping beam, and gives a magnication of 4.97. In Figure (b), the orientation of the plano-convex lens was reverse, and we again found that the quality of the system was diminished, having a focal length of only 8 m and a magnication of 3.1.
Surface 1 2 3 4 5 6
Table 2.1: The surface data entered into OSLO giving the best beam expander design shown in Figure 2.9(a). These are from left to right Thorlabs lenses LA1027B, LD2568-B and LA1417-B. BK7 is the most common type of glass for optical lenses, with a refractive index of 1.51.
19
(a) backwards
(b) forwards
Figure 2.10: Simulating the magnication range available in the nal design. In Figure (b) the central lens was moved forwards by 0.75 mm and the focal length decreased to 18 m while the magnication increased to 5.3. In Figure (a) the central lens was moved 0.75 mm backwards, resulting in a magnication of 4.7 and a focal length of 16 m. This gives a total magnication range of 0.6 while maintaining satisfactory collimation.
range in magnication corresponds to a change in intensity of 13% to +11%. The bre splitters used to form the six MOT beams do not split the power of the input light source perfectly evenly; there can be a variation in power between two MOT beams of up to 1%. The range of intensity available to us through the use of the variable beam expander is sucient to counteract this inbalance of the beam powers, such that we can form an intensity-balanced optical molasses.
20
Once the initial collimation of the beam was achieved, the positions of the three beam expander lenses were adjusted to achieve a magnied, collimated output beam. It was during this process that a number of challenges began to appear in the construction of this prototype.
21
(a)
(b)
Figure 2.11: Comparison of the Gaussian quality of bre output collimated using (a) an asphere with a focal length of 15.3 mm and (b) a Thorlabs bre collimator containing an air spaced doublet. We observed that the asphere introduced error into the beam, which can be seen as faint rings in the image, while the bre collimator successfully produced a Gaussian prole.
objective lens were both used to collimate the bre output, but no signicant improvement was observed. Instead, we decided to purchase bre collimators [F810APC-780], consisting of an air-spaced doublet lens positioned at a xed distance from the bre connector, which are able to collimate the bre output to a beam 7.5 mm in diameter. A comparison of the beam quality using the asphere and bre collimator is shown in Figure 2.11. Unfortunately, only one bre collimator arrived before the conclusion of this project. As such, the prototype was completed using the original asphere despite the irregular prole it introduced.
2.4.3 Results
The prototype was constructed as shown in Figure 2.12(a) and measurements were made to determine the magnication. We found that the distance between the input and central lens set the size of the beam immediately exiting the system, while the relative position of these two lenses to the output lens determined the collimation. The magnied beam was too wide to use the beam proler to measure the beam width, so a knife-edge measurement was performed. Instead of using a knife-edge, a piece of thin cardboard was deemed sucient for the accuracy required. To measure the width of the beam, the beam was gradually blocked by the cardboard edge, and the decreasing intensity of the beam was recorded by focussing the beam onto a power meter. The intensity measured as a function of the edge position for a Gaussian beam is given by P I(x) = 2 1 erf 2(x x0 ) (2.8)
where P is the total power of the beam, x0 is the beam centre and is the 1/e2 width of the beam. This equation is the result of dierentiating a Gaussian
22
(a)
(b)
Figure 2.12: (a) The beam expander prototype containing the (i) bre connector, (ii) initial collimating lens with f = 30 mm, (iii) the beam expander input lens (f = 30 mm), (iv) variable lens (f = 9 mm), held in a translation mount and (v) the output lens (f = 150 mm). (b) The data from a knife-edge measurement near the output lens, from which the magnication was calculated to be 5.3 0.1.
intensity prole. A model was constructed using this formula, and tted to the recorded data using the Mathematica routine NonLinearModelFit to nd the beam width. In order to adjust the beam expander to produce a magnication of ve, several knife-edge measurements were performed and analysed, and the lens positions adjusted accordingly. A graph of the knife-edge data and the line of best t is shown in Figure 2.12(b), for the beam immediately exiting the prototype, showing the nal result of a magnication of 5.3 0.1. Collimation was conrmed by viewing the beam through a infra-red viewer. Having found that the beam expander could produce a magnication which was close to ve and easily adjustable, measurements of the magnication range were reserved for the nal beam expanders.
23
expanders t neatly around the glass cell with minimum space, but this also provided full directional tilt adjustment to the output beams to make precise alignment possible. A close-up of one beam expander is shown in Figure 2.13(a), and the collection of all six beam expanders is shown in Figure 2.13(b). Duplicating a system by six can be time consuming. For the sake of eciency, one of the six beam expanders was carefully adjusted while performing knife-edge measurements until the output beam was magnied and collimated as required. The size of this output beam was viewed on a card through an infra-red viewer, and the apparent size was marked. This card was then used as a reference to adjust the remaining ve beam expanders. Measurements of the rst beam expander revealed that the beam was magnied by 5.2 0.2, and translation of the central lens gave a total change in magnication of 0.8 0.3. This agrees within uncertainty to the expected magnication range of 4.7 to 5.3. These uncertainties were calculated from the condence interval of the model tted to the data. Any degradation of collimation during this movement was visually imperceptible over a distance of four metres. We have constructed six Galilean variable beam expanders centred on a magnication of ve to produce 28 mm diameter trapping beams for the MOT. This magnication can be continuously varied from 4.7 to 5.3 by translating the single central lens along the optic axis, corresponding to a change in intensity of 24%, which is more than adequate to compensate for the slightly unbalanced power splitting in each of the bre splitters. In chapter 3, we discuss the construction of the ultra-high vacuum system. Once this was completed, the six MOT beam expanders were arranged around the glass cell, and a photograph of the nal arrangement is shown in Figure 3.8.
24
(a)
(b)
Figure 2.13: (a) One of the six Galilean variable beam expanders constructed for the MOT, with the (i) bre connector, (ii) f = 30 mm initial collimating lens, (iii) mirrors providing full tilt adjustment of the output beam, (iv) quarter-wave plate on a rotation mount, (v) f = 30 mm lens, (vi) f = 9 mm lens held in a custom brass holder, mounted in a (vii) translation mount to vary the beam magnication, and (viii) the output f = 150 mm lens. (b) The MOT beam expander collection. The lower two beam expanders have been built such that the input and output beams are orthogonal, to t the beam expander above and below the glass cell.
Chapter
26
(a)
(b)
Figure 3.1: (a) The preliminary vacuum system, with (i) a roughing pump and (ii) turbomolecular pump connected to (iii) a tee tting with (iv) a blanked-o ange and (v) an RGA. (b) An internal view of a turbomolecular pump, showing the slanted blades which force molecules down towards the exhaust [26].
exhaust is connected to a roughing pump [Varian IDP3A01], which can achieve pressures of 103 Torr and does not use oil, as many roughing pumps do, which lowers the risk of vacuum system contamination. The RGA [Stanford Research Systems RGA100] consists of a heated lament and anode which ionises the residual gas in the system and identies the partial pressures of the dierent gases and contaminants present. To achieve a high vacuum seal, all open ends of the vacuum components we used end in a ConFlat ange. To seal two anges together, a copper gasket (or ring) is placed between the anges into the grooves provided. When the anges are then bolted together, a knife-edge circling each ange cuts into the copper gasket, forming a metal seal. When bolting components together, the bolts must be gradually tightened in a star formation (see Figure 3.2) to ensure that uniform pressure is applied. If the bolts are tightened in a circular pattern, then as the copper gasket is indented the resulting slight bulge will travel around the system to meet the rst tightened bolt. This will not result in a vacuum seal and can damage the knife-edges of the anges. All bolts used were silver plated, to prevent steel-on-steel seizing after bakeout.
27
(a)
(b)
(c)
Figure 3.2: Recommended star formations for evenly tightening bolts on ConFlat anges with (a) 6, (b) 8 and (a) 16 holes [27].
1.2 1. 8. 6. 4. 2. 10 10 10 10 10 10
6 6 7 7 7 7
Pressure Torr
(a)
2 3 Time hours
(b)
Figure 3.3: (a) Data recorded using the RGA of the pressure inside the vacuum system after the turbo pump was turned on. The pressure drop was exponential with time, as expected, however there appear to be two distinct regions where the pressure dropped at dierent rates. This was perhaps caused by the turbo pump working at a dierent eciency once the pressure fell below 2 107 Torr. (b) An analysis of the composition of the residual gas in the vacuum system before bakeout, when the pressure reached 1.6 108 Torr. Before bakeout the pressure was dominated by (i) hydrogen and (ii) water, with peaks at (iii) masses 39, 41, 43, 55 and 57 which indicate the presence of machine oil [28]
if the machine oil was from the turbo pump, or from the walls of the tee tting, and so we decided to bake the vacuum system to further reduce the pressure and hopefully remove the contamination.
28
(a)
(b)
Figure 3.4: (a) To bake out the vacuum system to remove the remaining water and machine oil, the tee was wrapped in heater tape and thermocouples were placed around the system to measure temperature. The tee was then wrapped in aluminium foil and heated to 120 C. (b) The results from the RGA after bakeout, when the system reached 1.1 109 Torr. The scale on this graph is an order of magnitude less than that in Figure 3.3(b), and shows that while there is a small amount of hydrogen (2 u) and water (18 u) remaining, the machine oil was successfully removed.
room temperature, the pressure reached a minimum value of 1.1 109 Torr, an order of magnitude lower than before bakeout. The mass spectrum from the RGA after bakeout is shown in Figure 3.4(b), where we can see that the oil peaks have disappeared. During bakeout the turbo pump was only heated to 30 C, as it has a maximum operating temperature of 90 C. This temperature would not have been sucient to remove oil contamination, and so we concluded that the oil had been on the surface of the tee tting, despite the fact that it had come straight from the manufacturer. As a result, we realised that every component of the ultra-high vacuum system would need to be thoroughly cleaned before assembly.
29
a turbo pump, which can be isolated from the UHV system using a right-angle all-metal valve [Huntington MV-250-T], an ion pump [Gamma Vacuum 75S], to reduce the pressure from 109 Torr to 1012 Torr, an RGA to detect residual gas components, a hot cathode, dual lament, nude ion gauge, which can measure the total pressure of the system to a higher degree of accuracy than the RGA, a getter [Alvatec AS-3-Rb-10-V vapour source], a temporary atom source containing 10 g of rubidium, which will be used until the rubidium oven and atomic beam system is completed, and a gate valve which allows later extensions to be made to the vacuum system, without needed to expose the baked system to atmosphere. The UHV system will eventually be joined to the vacuum system housing the rubidium oven, which forms a collimated beam of rubidium which is slowed by the Zeeman slower before entering the MOT [19, 29]. As this system is not yet completed, we will temporarily be using a getter as our atom source. The gate valve on the opposite side to the atom source is currently shut, and allows another chamber to later be added to the vacuum system. This chamber is commonly known as a science chamber, and a BEC can be moved into it to perform additional experiments that may not have been planned into this existing system design.
30
(a)
(b)
Figure 3.5: (a) A drawing of the vacuum system constructed for the BEC machine. On the left is the rubidium oven and atomic beam source [29], with the ultra-high vacuum system discussed in this chapter to the right, separated by the Zeeman slower coils. (b) A close-up of the UHV system.
31
can leave lint behind. If this occurred, a lens tissue soaked in methanol was used to remove all traces of lint. For the smaller components, an ultrasonic bath was used instead of Kimwipes, in which vibrations were coupled through water to a beaker of methanol to thoroughly clean all contaminants from the part. Most components however were too large to t into the bath, or would have used several litres of methanol, so cleaning with Kimwipes was sucient in most cases. As each component was cleaned, they were wrapped in aluminium foil and stored until required. To assemble the components a copper gasket was placed between each pair of anges, which were then bolted together in the star formation described earlier. Where possible, a torque wrench was used to ensure than an even torque of 20 Nm was applied to each bolt on the larger 6 or 4.5 anges, and 15 Nm on the smaller 2.75 anges. To later join the UHV and atomic beam systems, we decided that the rubidium beam would be positioned 200 mm above the table. To hold the UHV system such that the centre of the glass cell was at this exact height, several high-load bearing support structures were designed and constructed. A photograph of the completed UHV system is shown in Figure 3.6.
32
Figure 3.6: Two views of the UHV system, showing (i) the getter, (ii) the glass cell, (iii) the ion gauge, (iv) the right-angle valve, (v) the turbo pump, (vi) the ion pump and (vii) the RGA.
33
Figure 3.7: The mass spectrum of the vacuum system at 2 108 Torr, indicating that the system is dominated by (i) hydrogen (2 u) and (ii) water (18 u), with traces of (iii) nitrogen (28 u), (iv) argon (40 u) and (v) carbon dioxide [30, p.494].
108 Torr and continued to slowly drop. For such a large system with a high surface area, this is an expected pressure to reach before bakeout. Figure 3.7 shows mass spectrum at this pressure. We can see that the pressure is dominated by hydrogen and water, and no oil peaks were detected. This indicated that our cleaning regime had been appropriate. The next stage will be to bake out the system, after which we expect to reach ultra-high vacuum at 1012 Torr.
34
Figure 3.8: Two dierent views of the UHV system surrounded by the six MOT beam expanders presented in chapter 2. One beam expander was held above the glass cell using a suspended breadboard, and (i) the temporary coils were attached to the upper and lower beam expander, with the glass cell at the centre.
Chapter
36
where 1.22 is Rayleighs constant and is the wavelength of light used (in our case, 780 nm) [23, p.548]. For example, to achieve a resolution of 1 m, the numerical aperture of the system must be at least 0.48. In air or vacuum, this corresponds to a widest angle of scattered light of approximately 30 . One of the main factors limiting the numerical aperture of a system is the physical lens size a larger diameter lens can capture a wider angle of scattered light than a smaller lens. We will be imaging the top of the BEC, so that when the BEC is released it does not move out of the eld of view. This means that our imaging system will need to use lenses with a maximum diameter of 50.8 mm, to t within the magnetic eld coils at the top of the glass cell (see Figure 4.1). The BEC will be located at the centre of the cell, 15 mm from the inside surface of the glass wall. However, we want to be able to image the BEC after it has fallen by up to 2 mm, requiring the objective lens to have a working distance of 17 mm.
37
Figure 4.1: A schematic of the imaging system in position near the glass cell (not to scale). The angle of scattered light collected by the objective lens denes the numerical aperture and resolution of the system (equations (4.1) and (4.2)). An achromat will be used to focus the light collimated by the objective lens to form an image.
38
(a)
(b)
Figure 4.2: (a) An exaggerated example of a lens causing positive spherical aberration. Rays passing though near the edge of the lens are brought to a closer focus than rays passing through near the lens centre. (b) A microscope objective lens with a numerical aperture of 0.9 [23, p.590]. The light is rst bent outwards by the rear element, then gradually bent inwards by a series of dierent lenses. This process minimises aberrations in the system, and is characteristic of microscope objective lenses. In our case, we require a system in which the focal point is much further from the last lens, in order to position our imaging system outside of the glass cell.
(see 2.4.2). Spherical aberration can occur in two directions; either the light is focussed in a spread before the expected focal point, called positive spherical aberration, or it is focussed in a spread after the expected focal point, called negative spherical aberration [32, p.226]. If a lens produces positive spherical aberration, then it can be corrected by using a second lens which introduces equal, negative spherical aberration [33, p.73]. This is the approach we used to minimise spherical aberration in our imaging system. An example of a high numerical aperture objective lens which minimises spherical aberration is shown in Figure 4.2(b). The 5 mm silica wall of the glass cell causes a signicant amount of spherical aberration which we needed to correct for using our objective lens. This correction requirement was the reason we cannot use a commercial long working distance microscope objective.
4.2. An imaging system using catalogue lenses Author W. Alt [34] T. Ottenstein [35] D. Weiss [36] S. Johnstone [37] L. Bennie NA 0.29 0.4 0.55 0.49 0.52 Elements 4 3 ? 5 4 FOV (m) 1000 400 140 10 340 Glass BK7 SF11 ? BK7 & SF11 SF11
Table 4.1: A comparison of the performance of our objective lens to other systems designed for high resolution imaging of BECs or cold atoms. The nal design of our system (e) has a numerical aperture of 0.52 with a eld of view of 340 m, and will be discussed in more detail in 4.3.1. It is similar in performance to entry (b), which has a similar eld of view, but a lower NA of 0.4.
design which was both diraction limited and manufacturable. Using this software, it was logical to dene the width of a collimated input beam, which would then be focussed by the array of lenses, rather than dening rays emanating from a point source at a certain angle. For this reason, although our imaging system will be used to collect light scattered from the BEC, in the following simulations we will consider the reverse situation collimated light entering the system and being focused at the point where the BEC will be located. The performance of the lens system was measured by the wavefront error both on- and o-axis. The wavefront error is the dierence between the shape of the wavefront produced by the lens system at the exit pupil of the lens (in our case the glass cell) and a perfectly spherical wavefront, which would be produced if the system exhibited zero aberrations. If the average, or rootmean-squared (RMS), wavefront error is less than 0.07 , then the system is considered diraction limited [33, p.191].
The lens system we used as a basis for our design is shown in Figure 4.3(a), and is the rst entry in Table 4.1. This system used four 25.4 mm diameter lenses to give a diraction limited eld of view of 1 mm, which is much wider than we require, however it only had a numerical aperture of 0.29, whereas we required at least 0.48. After replicating this design in OSLO, using the data provided in reference [34], the system was scaled to use 50.8 mm lenses. The
40
(a)
(b)
Figure 4.3: (a) The lens design listed in entry (a) of Table 4.1 which we initially used as a basis for our design [34]. It used 25.4 mm Thorlabs catalogue lenses to achieve an NA of 0.29 and a diraction limited eld of ew of 1 mm. (b) A lens system we designed using Thorlabs catalogue lenses [from left to right: LC1315, LA1384 (twice), LE1076]. It has a NA of 0.4, but could not be increased to greater than 0.48 as required. As the analysis in Figure 4.4 shows, this design was not diraction limited on axis, and so it was rejected.
curvature of the last surface before the glass cell was set as a solved variable such that the numerical aperture of the system was set to 0.4. This decreased the focal length of the system, and the glass cell was positioned such that the focus was 17 mm from the inner cell surface. Interestingly, we observed that the distance between the glass cell and the lens closest to the focus had no eect on the performance of the system. This meant that in the nal design the BEC can be imaged from two positions; rst when held in a trap 15 mm from the glass cell, and then after it has fallen over a distance of up to 2 mm. Introducing these changes to the system meant that the system was no longer diraction limited. The optimisation routine was then performed to minimise the error in the system. We found that the best way to do this was to incrementally allow curvatures, and then thicknesses, to vary optimising the system between each new introduction. If all lens curvatures and thicknesses were allowed to vary from the start, then although the error in the system would become very small, the result would not be manufacturable with a combination of overlapping lenses, distances of several metres between lenses, and lens thicknesses either greater than 10 cm or less than 0.5 mm. This could have been controlled by setting the range over which these quantities could vary, however this feature was not discovered until later, as discussed in 4.3.
41
was a very limited choice of catalogue meniscus lenses. Once all the lenses were specied as catalogue lenses, the air spaces were xed one at a time, again performing optimisation each time, until nally there were no variables left in the system. If we had been able to produce a satisfactory design using catalogue lenses, then the imaging system would have been very inexpensive (compared to manufacturing custom lenses) and could have been tested in the laboratory in a relatively short amount of time. Unfortunately, we did not manage to achieve this. The best result obtained using Thorlabs catalogue lenses is shown in Figure 4.3(b). As the simulations of spot size and wavefront error shows, the system is not diraction limited (Figure 4.4). This means that even though the numerical aperture is 0.4, this objective lens could not be used to form an image with a resolution of 1.2 m as expected from equation (4.2). In addition to this, we observed that with 50.8 mm diameter lenses made of BK7 glass the numerical aperture could not be made higher than 0.4. When the NA of the system was increased, the rays would begin to pass outside of the 2nd or 3rd lens. The NA of the system also depended on the diameter of the input beam, so even though we could prevent the marginal rays from passing outside of the lens, this also had the eect of decreasing the numerical aperture.
42
(a)
(b)
Figure 4.4: Performance simulations of the lens system shown in Figure 4.3(b), which used Thorlabs catalogue lenses. (a) Ray tracing results showing that the light (green dots) did not fall within the diraction limited Airy disc (black circle), thereby showing that the system was not diraction limited. This was simulated on-axis (bottom row), 120 m o-axis (middle row) and 170 m o-axis (top row), from a distance along the optic axis 1 m before the focus (left column) to 1 m after the focus (right column). All values shown are in millimetres. (b) A calculation of the wavefront error showing both the peak-to-valley (P-V) and RMS error in units of one wavelength, for a position 170 m o-axis (left), 120 m o-axis (middle) and on-axis (right). In all cases, the RMS error was greater than 0.07 , and hence the system was not diraction limited.
43
Figure 4.5: The wavefront error in an optical system as a function of refractive index, for dierent numbers of elements. We see that for a four element system, increasing the refractive index from 1.5 (BK7) to 1.77 (SF11) can result in a system with an order of magnitude lower wavefront error [33, p.69].
The numerical aperture of the system was again controlled by the curvature of the front surface of the lens closest to the glass cell. It was found that a numerical aperture of 0.6 was the highest value that could be achieved with 50.8 mm lenses made of SF11 glass. A similar optimisation routine was followed several times, each time introducing variables in a dierent order to nd a lower local minimum of the error function. A range of maximum and minimum thickness were also specied for each surface, to prevent lenses either overlapping or becoming unreasonably thick. Generally if the simulation produced a lens more than 5 cm thick, the lens thickness could be reduced and a lens separation could be adjusted to produce the same overall performance. The system was also optimised o-axis. This involved specifying a second collimated ray bundle that entered the system at a slight angle, to be focussed at a point approximately the same distance away from the glass cell, but a few hundred micrometers radially away from the optic axis. The optimisation routine then produced a minimum error function at both focal points. This resulted in the on-axis performance being slightly diminished numerically, but still diraction limited, as will be discussed in 4.3.1. As the second focal point was moved further away from the optic axis, the error remaining after optimisation increased, so there was a limit to the diraction limited eld of view which could be achieved in our design.
44
Figure 4.7 shows the wavefront error on axis, 120 m and 170 m o-axis. The RMS error at these positions is 0.042 , 0.034 and 0.067 respectively. The wavefront error at several other positions between 0 and 170 m was calculated, and in all cases the RMS error was less than 0.07 , and so the system is diraction limited to a radial distance of 170 m, giving a total eld of view of 340 m. The appearance of the beam around the focus is shown in Figure 4.8. The numerical aperture of our lens system is 0.52, corresponding to a resolution of 0.9 m. This design surpassed the original requirements of our imaging system, having a resolution of 0.9 m instead of 1 2 m, and a eld of view of 340 m instead of only 100 m. A quote from LENS-Optics suggests that these lenses will cost roughly $3000 to manufacture (without mounting). Before committing to this design, we will explore the option of using a combination of Thorlabs lenses and custom, high refractive index lenses, to see if a system can be designed which will be cheaper and within our requirements. It seems
1 OSLO calculates the NA of this objective lens to be 0.6, using a small angle approximation which becomes invalid for higher NA systems. When this lens was simulated in a dierent program called Zeemax, which used equation (4.1), the numerical aperture was 0.52. 2 This surface was kept at to accommodate the 10 surface limit in the free version of OSLO. In Zeemax this surface was allowed to curve, however we did not nd signicant improvement in performance and the objective lens became less compact.
45
Figure 4.6: The nal design of our objective lens. It has a NA of 0.52, corresponding to a resolution of 0.9 m when used with 780 nm light, and is diraction limited to 170 m o axis, or a total eld of view of 340 m. The surface curvatures and thicknesses are listed in Table 4.2.
Surface 1 2 3 4 5 6 7 8 9 10
Curvature Radius (mm) 64.42 114.03 -178.65 45.6 214.8 25.35 32.28
Diameter (mm) 50.8 50.8 50.8 50.8 50.8 50.8 21 18 50.8 50.8
Glass SF11 Air SF11 Air SF11 Air SF11 Air Silica Vacuum
Table 4.2: The specications of the nal design of our objective lens, shown in Figure 4.6. We found that the spacing between the elements could be rounded to the nearest 0.1 mm while still maintaining a diraction limited system.
46
Figure 4.7: The wavefront error in the objective lens design, from 170 m o-axis (left), 120 m o-axis (middle) and on-axis (right). In all cases, the RMS error was less than 0.07 , and so the system is diraction limited for a eld of view of 340 m.
Figure 4.8: Spot size analysis of the objective lens design. At the focus (central) column, the rays fall within the diraction limit circle. Away from the focus, both on and o axis, we see that more error occurs, and the system is no longer diraction limited after a distance of 1 m away from the focus. This shallow depth-of-eld is a consequence of having a high NA lens. In particular we can observe the eect of some residual spherical aberration in the bottom row, with the spot size growing larger with a negative focal shift, and smaller with a positive focal shift. We can also see the eects of coma aberration occurring in the top row, far from the optic axis, as the spot size becomes elliptical. However, since in most cases this occurs within the Airy disc, we can attribute these eects mainly to artefacts in ray tracing approximations, rather than physical behaviour of the system.
47
unlikely however that a numerical aperture of 0.5 or greater could be achieved with such a system. The quote also provided tolerance values for the thicknesses and radius of curvatures. One nal aspect of the design which has not yet been explored is the eect of variations within these tolerance values on the performance of the system. However, we are condent that the objective lens will be insensitive to < 0.1 mm variations in the lens spacing, based on the diraction limited performance after the spacings were rounded. The main concern would be the eect of tilt or centration of the lenses, which could decrease the performance of the objective lens. This would have to be considered before manufacturing this design.
48
Chapter
Conclusion
As a result of the work presented in this thesis, the Spinor BEC Laboratory is now several steps closer towards producing condensates. A set of Galilean variable beam expanders were made to produce wide trapping beams for the MOT. The ultra-high vacuum system was constructed, within which rubidium atoms will be trapped and cooled to form a BEC. In preparation for the near future, a high resolution imaging system was designed to image BECs both when they are held in a trap and when the condensate has been released and expanded. Chapter 2 discussed the design and construction of the beam expanders used to trap atoms in the MOT. We derived the relation between the focal lengths of the rst and last lens in a three element variable beam expander to the magnication of the laser beam. Using this relation, we designed a nonsymmetric variable beam expander centred on a magnication of ve. Using such a system, the magnication of a collimated beam can be continuously varied between a range of 4.7 5.3 by moving only the central lens along the optic axis. This is in contrast to standard, symmetric variable beam expanders in which both the central and the input lens must be moved at dierent rates to vary the beam size around a magnication greater than one. Six beam expanders were constructed which produce collimated beams 28 mm in diameter, approaching the maximum beam size which can pass through the MOT apparatus and glass cell. These trapping beams will allow the MOT to trap a high number of atoms, on the order of 1010 , which will result in a condensate of 107 atoms. In three of the six beam expanders the central lens position can be adjusted. This allows the relative intensity in each MOT beam pair to be balanced by changing the width of one beam. Equal intensities in each beam pair is particularly important during optical molasses cooling stages after atoms have been trapped in the MOT. To form a MOT, the trapping region must be located within ultra-high vacuum, to prevent background thermal atoms transferring kinetic energy to trapped atoms via collisions. The construction of the UHV system was discussed in chapter 3. The pressure within the UHV system reached 2 108 Torr, and will be reduced further by baking out the system, a process 49
50
Chapter 5. Conclusion
which will occur in the near future. During bakeout, the temperature of the vacuum surfaces is raised to around 300 C, and the layer of water which forms on the vacuum surfaces during exposure to water vapour in air is then removed from the system. Due to our successful cleaning procedures, were are condent of achieving ultra-high vacuum of 1012 Torr. In chapter 4, we presented the design of a high resolution objective lens to operate at 780 nm for absorption imaging of a BEC. Simulations show that the lens system has a numerical aperture of 0.52 and is diraction limited across a eld of view of 340 m. Using this objective lens, a BEC which has been dropped from the trap and expanded can be imaged with a resolution of 0.9 m when imaged with 780 nm light. The performance of the lens system exceeds that required for our purposes, with a resolution smaller than 1 m and a eld of view wider than 100 m. The objective lens contains four elements, 50.8 mm in diameter, made of SF11 glass (n = 1.77); one bi-concave, two meniscus, and one plano-convex lens. Unlike commercially available microscope objective lenses, our design corrects for the spherical aberration introduced by the wall of the glass cell. The system has a working distance of 17 mm from the glass cell, and a total length of 68.7 mm. While these lenses are not mass produced standard elements, they can be produced by a custom lens manufacturer. Our BEC machine will include all three components presented in this thesis; the variable beam expanders, the ultra-high vacuum system, and the high resolution objective lens. We expect during 2011 to have BECs in production. We will be developing a new technique using magnetic resonance imaging (MRI) to non-destructively image structures such as vortices in trapped BECs, which can not normally be resolved unless the BEC is expanded.
Appendix
51
52
(a)
(b)
(c)
Figure A.1: The (a) front and (b) back view of the custom lens holder made for the f = 9 mm beam expander lens. The lens slots in from the back and is held between the lip of the entrance and a ring (not shown) which screws in from the back. There is an external and internal thread which is not shown in this model. (c) The translation mount into which the custom lens holder screws into. To ensure that the lens is held centrally on-axis by the translation mount, the six screws shown in the picture can be disassembled, and reassembled, releasing any possible tension on the translation mechanism.
53
Figure A.2: The technical specications used to manufacture the six custom lens holders.
54
Bibliography
[1] S. Bose, Plancks Gesetz und Lichtquantenhypothese, Zeitschrift fur Physik 26 (1), 178181 (1924). [2] S. Bose, Plancks law and the light quantum hypothesis, Journal of Astrophysics and Astronomy 15 (1), 37 (1994). [3] L. Pitaevskii and S. Stringari, Bose-Einstein Condensation, Oxford University Press (2003). [4] M. H. Anderson, J. R. Ensher, M. R. Matthews, C. E. Wieman and E. A. Cornell, Observation of Bose-Einstein Condensation in a Dilute Atomic Vapor, Science 269 (5221), 198201 (1995). [5] NIST, Bose-Einstein Condensate, http://patapsco.nist.gov/ imagegallery/details.cfm?imageid=193, accessed 11/10/09. [6] J. R. Abo-Shaeer, C. Raman, J. M. Vogels and W. Ketterle, Observation of vortex lattices in Bose-Einstein condensates. Science 292 (5516), 476 479 (2001). [7] E. L. Raab, M. Prentiss, A. Cable, S. Chu and D. E. Pritchard, Trapping of neutral sodium atoms with radiation pressure. Physical Review A 59 (23), 26312634 (1987). [8] Y.-J. Lin, A. Perry, R. Compton, I. Spielman and J. Porto, Rapid production of Rb87 Bose-Einstein condensates in a combined magnetic and optical potential, Physical Review A 79 (6), 063631 (2009). [9] E. W. Streed, A. P. Chikkatur, T. L. Gustavson, M. Boyd, Y. Torii et al., Large atom number Bose-Einstein condensate machines, Review of Scientic Instruments 77 (2), 023106 (2006). [10] H. Yan-Xu, L. Yong-Hong, Z. Chun-Hong and L. Shu, Realization of High Optical Density Rb Magneto-optical Trap, Chinese Physics 26 (2), 023201 (2009). [11] G. L. Gattobigio, T. Pohl, G. Labeyrie and R. Kaiser, Scaling laws for large magneto-optical traps, Physica Scripta 81 (2), 025301 (2010). 55
Bibliography
[13] W. D. Phillips, Nobel Lecture: Laser cooling and trapping of neutral atoms, Reviews of Modern Physics 70 (3), 721741 (1998). [14] S. Chu, L. Hollberg, J. Bjorkholm, A. Cable and A. Ashkin, Threedimensional viscous connement and cooling of atoms by resonance radiation pressure, Physical Review Letters 55 (1), 4851 (1985). [15] H. J. Metcalf and P. van der Straten, Laser Cooling and Trapping, Springer-Verlag New York (1999). [16] C. J. Dedman, J. Nes, T. M. Hanna, R. G. Dall, K. G. H. Baldwin et al., Optimum design and construction of a Zeeman slower for use with a magneto-optic trap, Review of Scientic Instruments 75 (12), 5136 (2004). [17] P. Zeeman, On the inuence of magnetism on the nature of the light emitted by a substance, Philosophical Magazine 43 (262), 226239 (1897). [18] D. A. Steck, Rubidium 87 D Line Data, (2009), http://steck/us/ alkalidata. [19] L. Bennie, Design and Construction of a Zeeman Slower for 85 Rb and 87 Rb, Research Project, School of Physics, Monash University (2010). [20] P. Starkey, An optical, experiment control and data aquisition system for a BEC lab, Honours thesis, School of Physics, Monash University (2010). [21] E. Hecht, Optics 4th Ed., Addison Wesley (2002). [22] CVI Melles Griot, http://www.cvimellesgriot.com/Products/ Multiple-Order-Quartz-Waveplates.aspx, accessed 03/11/10. [23] H. Gross, F. Blechinger, and B. Achtner, Handbook of Optical Systems Vol. 4, Wiley-VCH (2008). [24] D. Malacara, Z. Malacara, Handbook of Optical Design 2nd Ed., Marcel Dekker, New York (2004). [25] M. Yeh, S. Shiue and M. Lu, First-order analysis of a three-lens afocal zoom system, Optical Engineering 36 (4), 12491258 (1996). [26] Image source: A cut through turbomolecular pump http://commons.wikimedia.org/wiki/File:Cut_through_ turbomolecular_pump.jpg, accessed 03/11/10. (2005),
[27] K. M. Birnbaum and the Caltech Quantum Optics Group, Ultra-High Vacuum Chambers, (2005), unpublished, http://commons.wikimedia. org/wiki/File:Cut_through_turbomolecular_pump.jpg, accessed 03/11/10.
Bibliography
57
[28] Stanford Research Systems, Vacuum Diagnosis with an RGA, http://www.thinksrs.com/downloads/PDFs/ApplicationNotes/Vac_ diag_RGA.pdf, accessed 31/09/10. [29] A. Wood, A rubidium cold atom beam source, Honours thesis, School of Physics, Monash University (2010). [30] J. F. OHanlon, A users guide to vacuum technology 3rd Ed., John Wiley & Sons (2003). [31] R. Anderson, Nonequilibrium dynamics and relative phase evolution of two-component Bose-Einstein condensates, Ph.D. thesis, Swinburne University (2010). [32] H. Gross, H. Zugge, M. Peschka and F. Blechinger, Handbook of Optical Systems Vol. 3, Wiley-VCH (2007). [33] R. E. Fisher and B. Tadic-Galeb, Optical System Design, McGraw-Hill (2000). [34] W. Alt, An objective lens for ecient uorescence detection of single atoms, Optik 113 (3), 142144 (2002). [35] T. B. Ottenstein, A New Objective for High Resolution Imaging of BoseEinstein Condensates, Ph.D. thesis, Kirchho Institute of Physics (2010). [36] K. D. Nelson, X. Li and D. S. Weiss, Imaging single atoms in a threedimensional array, Nature Physics 3, 556560 (2007). [37] S. Johnstone, Design of an optical system for the uorescence imaging of tracer particles in a BEC, Honours thesis, School of Physics, Monash University (2010).