Fachbereich Physik Momentum-Resolved Optical Lattice Modulation Spectroscopy on Bose-Fermi Mixtures Impulsaufgelöste Modulations-Spektroskopie an Bose-Fermi-Mischungen in optischen Gittern Diploma Thesis Bastian Hundt Universität Hamburg MIN-Fakultät, Department Physik Institut für Laser-Physik July 2011 Abstract Research on ultracold quantum gases has developed into one of the most dynamic fields in physics. Breakthroughs on a multitude of topics occur at daily basis. A cornerstone of this process was the development of optical lattices. The high degree of control over the experimental parameters makes ultracold quantum gases in optical lattices an ideal tool to simulate fundamental theories of solid state physics. But also fundamental processes of many-body physics are studied extensively. A central role is played by quantum phases and quantum phase transitions. Interaction and tunneling governs the behavior of the systems of interest. Momentum resolved spectroscopy on fermionic, bosonis and Bose-Fermi mixtures gives access to the underlying physics with high precision. A novel spectroscopy method has been developed at the “Bose-Fermi mixtures” experiment in the group of K. Sengstock in Hamburg. In the context of this thesis, the method based on modulating the lattice depth has been developed and characterized. The band structure of ultracold fermionic Potassium is probed with high accuracy. A significant influence of the harmonic confinement altering the band structure is observed and the lattice depth experienced by the atoms can be deduced very precisely. This allows the calculation of fermionic tunneling energies. The method has then been employed on mixtures of fermions and bosons. A central result of this thesis is the observation of a shift in the lattice depth experienced by fermions of up to 20%. This corresponds to a decrease in fermionic tunneling of up to 30%. The decrease depends on the bosonic occupation of the lattice sites and is explained in terms of an effective potential created by attractive interactions. The developed method and the obtained data can be helpful to further understand the bosonic Mott-insulator to superfluid transition. Employing the method on interacting fermionic mixtures could lead to the observation of new phenomena. Publication of the results is in preparation. Observing such new phenomena requires efficient and reliable detection of the atomic ensembles. To assure high quality absorption detection, this thesis presents a new optical design for the detection setup used at the “Bose-Fermi mixtures” experiment. The new design is developed using state of the art optical design software. Using optimization algorithms, a diffraction limited lens system, subject to very small optical errors is designed, tested and integrated into the experimental setup. Zusammenfassung Die Forschung an ultrakalten Quantengasen hat sich in den letzten Jahren zu einem der dynamischten Forschungsfelder der Physik entwickelt. Dazu beigetragen hat im großen Maße die Realisierung von ultrakalten Quantengasen in optischen Gittern. Die außergewöhnlich gute Kontrolle über eine Vielzahl physikalischer Parameter macht sie zu einem perfekten Werkzeug zur Simulation sowohl festkörpertheoretischer Fragestellungen als auch fundamentaler Fragen der Physik von stark wechselwirkenden Vielteilchensystemen. Von zentraler Bedeutung sind die auftretenden Phasen und Phasenübergänge. Diese werden sowohl in rein bosonischen oder rein fermionischen Systemen als auch in Bose-Fermi Mischungen durch die Wechselwirkungs- und Tunnelenergie bestimmt. Impulsaufgelöste Spektroskopie an solchen Systemen erlaubt dabei einen hochpräzisen Zugang. Am „Bose-Fermi-Mischungs“Experiment in der Gruppe von K. Sengstock wurde ein neuartiges, spektroskopisches Verfahren entwickelt, das bei voller Impulsauflösung das Anregungsspektrum fermionischer Atome in optischen Gittern untersucht. Im Rahmen dieser Arbeit wurde ein Verfahren basierend auf der Modulation der Gittertiefe entwickelt und charakterisiert. Daraufhin konnte die Bandstruktur von fermionischem Kalium in optischen Gittern mit hoher Präzision vermessen werden. Es wurden Daten bis einschließlich des vierten Bandes gewonnen. Dabei konnten signifikante Einflüsse des harmonischen Einschlusses auf die Bandstruktur beobachtet werden. Aus den vermessenen Bandstrukturen ist die Gittertiefe präzise abgeleitet worden. Dies erlaubt auf der einen Seite eine genaue Kalibrierung des optischen Gitters, auf der anderen Seite eine exakte Bestimmung der Tunnelenergie. Die entwickelte Methode wurde anschließend auf Mischungen von Fermionen und Bosonen angewendet. Als zentrales Ergebnis dieser Arbeit wurde eine Mischungsverhältnis abhängige Zunahme der fermionischen Gittertiefe beobachtet, die im Rahmen eines effektive Potentials erklärt werden kann. Für ein hohes Mischungsverhältnis zwischen Rubidium und Kalium wurde dabei eine Abnahme der fermionischen Tunnelenergie von 30% ermittelt. Die entwickelte Methode und die durchgeführten Messungen könnten sich beim Verständnis des bosonischen Mott-isolator zu superfluid Phasenübergangs in Bose-Fermi Mischungen als wertvoll erweisen. Die Anwendung auf wechselwirkende fermionische Systeme verspricht die Beobachtung neuartiger Phänomene. Eine Veröffentlichung der gewonnenen Ergebnisse ist in Vorbereitung. Im Rahmen dieser Arbeit wurde darüber hinausgehend ein neuer optischer Aufbau für die am Experiment verwendete Detektion mittels Absorptionsmessung entwickelt und implementiert. Dabei wurde das vorhandene Linsensystem durch einen neuen optischen Aufbau ersetzt, der sich durch eine starke Reduktion der Abbildungsfehler und ein hohes Maß an Flexibilität auszeichnet. Referenten Referent: Prof. Dr. Klaus Sengstock Universität Hamburg Fakultät für Mathematik, Informatik und Naturwissenschaften Department Physik – Institut für Laser-Physik „Quantengase und Spektroskopie“ Koreferent: Prof. Dr. Henning Moritz Universität Hamburg Fakultät für Mathematik, Informatik und Naturwissenschaften Department Physik – Institut für Laser-Physik „Quantenmaterie“ Erklärung zur Eigenständigkeit Ich versichere hiermit, dass ich die Diplomarbeit ohne fremde Hilfe selbstständig verfasst und nur die angegebenen Quellen und Hilfsmittel benutzt habe. Mit einer späteren Ausleihe meiner Arbeit bin ich einverstanden. Hamburg, den 05. Juli 2011 Bastian Hundt Contents 1 Introduction 2 Ultracold Bose-Fermi Mixtures in Optical Lattices 2.1 Experimental Setup and Capabilities . . . . . . . . . 2.2 Optical Lattice Potentials . . . . . . . . . . . . . . . 2.3 Non-Interacting Atoms in Optical Lattice Potentials 2.4 Ultracold Bosons in Optical Lattices . . . . . . . . . 2.5 Ultracold Fermions in Optical Lattices . . . . . . . . 2.6 Band Mapping . . . . . . . . . . . . . . . . . . . . . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5 6 7 10 13 15 3 Lattice Modulation Spectroscopy 3.1 Momentum Resolved Lattice Modulation Spectroscopy 3.2 Chracterization of Modulating the Lattice Amplitude . 3.3 Band Structure of Spin-Polarized Fermions . . . . . . 3.4 Spectroscopy on Pure Bosonic Samples . . . . . . . . . 3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 20 22 26 34 35 4 Lattice Modulation Spectroscopy with Bose-Fermi Mixtures 4.1 Ultracold Mixtures of Bosons and Fermions . . . . . . . . . . . . . . . . . . 4.2 Lattice Modulation Spectroscopy on Bose-Fermi Mixtures . . . . . . . . . . 4.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 37 38 41 5 Absorption Imaging with a Diffraction Limited Objective 5.1 Absorption Imaging . . . . . . . . . . . . . . . . . . . 5.2 Image Formation, Resolution and Abberrations . . . . 5.3 Raytracing and Computer Aided Optical Engineering 5.4 Old Detection Setup . . . . . . . . . . . . . . . . . . . 5.5 Design of a New Diffraction Limited Detection System 5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . 43 43 44 49 54 56 67 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Conclusion and Outlook 69 List of Figures 71 List of Tables 72 Bibliography 77 1 1 Introduction A major breakthrough of our understanding of nature was the development of quantum mechanics at the beginning of the 20th century and following this the formulation of different types of quantum-statistics: Fermi-Dirac statistics for particles subject to Pauli’s exclusion principle and Bose-Einstein statistics describing bosonic particles. The latter is the basis of A. Einstein’s proposal of Bose-Einstein condensation from 1924 [1], which has been realized in the Groups of E.A. Cornell [2] and W. Ketterle [3] in 1995. The first realization of a macroscopic, many-body quantum state allowed unprecedented insights into fundamental processes of nature and was awarded with the Nobel Prize in Physics 2001. With the realization of a quantum degenerate Bose gas also the interest in ultracold gases of fermions arose. But due to technical and physical complications it took until 1999 to realize a quantum degenerate Fermi sea. The breakthrough for reaching bosonic quantum degeneracy has been evaporative cooling. Due to Pauli blocking this technique could not be used to prepare a spin-polarized sample of fermions. The first quantum degenerate sample of Potassium was realized with a spin mixture in 1999 by B. DeMarco and D. S. Jin [4]. Using Feshbach resonances [5] the BEC-BCS crossover could be realized [6] triggering the hope for a deeper understanding of high-temperature superconductivity. Advances in preparation and the high degree of control over ultracold quantum gases led to proposals for using these gases as a tool to simulate other physical systems. Most notably the development of optical lattices allowed for the realization of the Hubbard Hamiltonian which plays a crucial role in solid-state theory [7]. A major breakthrough was the observation of the bosonic Mott insulator to superfluid phase transition in 2002 [8], where the transition from a strongly interacting and localized state to a weakly interacting delocalized state was realized. The recent simulation of classical magnetism [9] and quantum magnetism [10] in such systems showcases again the remarkable possibilities offered by quantum gases. Fermions in optical lattices are exceptionally well suited for simulating solid-state physics due to the resemblance to electrons in a crystal. Such a system was realized in 2005 and led to the observation of a crossover from a metallic phase to a band insulator [11]. In 2008 a fermionic Mott-insulator was observed using a spin-mixture of 40 K [12, 13]. The search for antiferromagnetic ordering of fermionic quantum gases in optical lattices is ongoing. With the realization of single species quantum gases it soon was realized that mixtures between different species offer new research possibilities. Especially the mixture of atoms obeying different quantum statistics led to new phenomena governed by intra- and interspecies interactionss. Using Bose-Fermi mixtures, it was possible to observe a collapse of a fermionic gas induced by bosons [14]. The creation of ultracold heteronuclear molecules [15] with a permanent dipole moment which are subject to long range, anisotropic interactions [16] and give access to ultracold chemistry [17]. Recently overlapping fermionic and 2 1 Introduction bosonic Mott-insulators were prepared, showing - among other phenomena - e.g. a complete phase separation [18]. During the first experiments using mixtures of bosons and fermions in optical lattices, a unexpected effect was observed. While examining the bosonic Mott-insulator to superfluid phase transition an admixture of fermions led to a decrease of the bosonic visibility [19–21]. This effect has been studied with great interest from theoretical and experimental side. It has been argued that adiabatic heating [22] results in a loss of coherence. Another model explains the visibility shift in terms of interaction-induced self-trapping of the bosons [23]. Until now it could not be decided which effect leads to the visibility shift. 50 pure fermions NRb / NK = 2 Emod / h [kHz] 45 NRb / NK = 10 40 7.6 EKr 35 8.1 EKr 9.2 EKr 0 0.2 0.4 0.6 0.8 1 q [kBZ] Figure 1.1: Fermionic dispersion relation of the third band with an admixture of bosons. A mixture of 40 K and 87 Rb has been prepared in an optical lattice. Modulation spectroscopy yields the dispersion relation of fermions. In red the pure fermionic band structure is shown. Green shows a moderate amount of bosons added to the system. Blue depicts the case of high admixtures of bosons. To gain further insight into the interplay between fermions and bosons, and to possibly understand the phase transition better, it is necessary to probe the tunneling rate of ultracold atoms exposed to a background field of another species. A tool to probe and access the underlying physical system are spectroscopical measurements. In the context of this thesis a novel spectroscopy method was developed at the “Bose-Fermi mixtures” experiment in Hamburg in the group of K. Sengstock. Employed on fermionic samples as well as Bose-Fermi mixtures in an optical lattice, it allows for momentum resolved spectroscopy. Based on modulation of the lattice depth, the method allows for a precise energy transfer to the atomic ensemble which excites atoms into higher bands. The occupation of the different bands is revealed applying the band mapping technique, allowing for the extraction of the full band structure. During the course of this thesis lattice modulation spectroscopy was characterized and employed on a mixture of fermionic Potassium-40 and bosonic Rubidium-87. As a central result the response of the system to lattice modulation with different atom number ratios is obtained. For large bosonic fillings a substantial shift 3 of the band structure is observed (see Fig. 1.1) which can be explained within an effective potential approach. This shift results in a decrease of the fermionic tunneling energy of up to 30% and depends on the bosonic lattice occupation. Structure of this thesis The second chapter of this thesis starts with an introduction to the experiment. After the description of creation and preparation of ultracold atoms in optical lattice potentials, the effect of periodic potentials on non-interacting particles is reviewed. The inclusion of interactions leads to the description of bosons in optical lattices in terms of the Bose-Hubbard model. Fermions can be described by a similar model. The knowledge of the behavior of fermions in an optical lattice allows us to characterize the band mapping technique which is of fundamental importance for the spectroscopy method introduced in the next chapter. The third chapter describes the method of lattice modulation. Starting with a general introduction we will obtain expressions for the transition probabilities between the lowest Bloch-band and higher bands. The method is further characterized to ensure accuracy and the avoidance of beyond linear-response effects. Thereafter spin-polarized fermions are studied. Obtaining the band structure for various lattice depths the influence of the harmonic confinement is studied. As a last step, bosons are examined in this context. Chapter four concentrates on a mixture of bosons and fermions. The mixture is described theoretically and an interaction-induced effective potential mechanism is introduced. Lattice modulation spectroscopy is employed and the band structure for different atom number ratios is measured. The fifth chapter concentrates on a more technical aspect of the experimental sequence which is absorption detection. First, absorption imaging after time of flight (TOF) is introduced. Then the process of image formation is explained and the errors, introduced by optical elements in the optical path of the detection laser, are explained. A new optical setup specifically engineered to minimize these errors is designed and evaluated. At the end of this thesis, proposals for further applications of the introduced spectroscopy method are given. 4 1 Introduction 5 2 Ultracold Bose-Fermi Mixtures in Optical Lattices 2.1 Experimental Setup and Capabilities All the experiments performed in the context of this thesis have been carried out at the “Bose-Fermi mixture” experiment in the group of Klaus Sengstock at the University of Hamburg. A sketch of the setup is shown in Fig. 2.1. 2D-MOT Pumping Stage & Vacuum System LA 3D-MOT D2 Figure 2.1: Experimental setup. The 3D-magneto-optical-trap (MOT) is loaded from a 2D-MOT which are connected via an differential pumping stage separating the upper vacuum system from the lower. The MOT/lattice laser beams denoted by D1, D2 and LA are tilted with respect to the detection axis. The experimental setup allows the trapping and preparation of ultracold mixtures of fermionic 6 2 Ultracold Bose-Fermi Mixtures in Optical Lattices 40 K and bosonic 87 Rb [24, 25]. The atoms are loaded from a 2D magneto optical trap (MOT) into a 3D-MOT. After cooling in an optical molasses and preparing the ensemble in one hyperfine state the atoms are transferred to a magnetic trap where evaporative cooling is performed. Potassium is sympathetically cooled due to collisions with Rubidium. The mixture is then loaded into an optical dipole trap which operates at the magic wavelength (808 nm) to compensate for different gravitational sags. The atoms reach quantum degeneracy where 87 Rb forms a Bose-Einstein-Condensate (BEC) and 40 K forms a Fermi-sea while reaching temperatures of T ≈ 0.1Tf where Tf is the Fermi-temperature. Depending on the experiments to conduct, one species can be removed by shining in resonant laser light for a short amount of time. Using radiofrequency (rf) or microwave pulses a high degree of control over the internal atomic states is possible. The interactions between the atoms can be tuned using intra- or interspecies Feshbach resonances. A cubic optical lattice can be superimposed. The lattice is generated by using three orthogonal pairs of laser beams denoted in Fig. 2.1 with D1, D2 and LA and operated at a wavelength of λL = 1030 nm. The radius of the atoms trapped is to small to be detected directly thus the atoms are released from all trapping potentials and allowed to expand freely. By shinning in resonant light a shadow of the density of the cloud is imaged onto a charge coupled device (CCD) camera. From the imaged shadow the density distribution of the expanded atoms is derived. This allows extraction of momentum distributions, correlation analysis etc. 2.2 Optical Lattice Potentials Atoms exposed to a laser field which is not on resonance with an atomic transition will experience a dipole force due to the coupling between the atomic energy levels and the photon field of the laser beam. A red detuning of the laser ωL with respect to the atomic resonance ω0 results in a force driving the atoms to the points of maximum intensity of the laser light. Blue detuning will push the atoms to the intensity minima. An optical lattice is generated by standing light waves formed by counter-propagating laser beams with the same frequency. For red detuning the resulting standing wave confines the atoms at the intensity maxima which occur at alat = λL /2. Three pairs of counterpropagating beams orthogonally aligned will result in a 3D-cubic-lattice. The resulting potential around the intersection of the beams can be expressed as (see [26]) V (x, y, z) = Vlat (x, y, z) + Vharm (x, y, z) ≈ V0 cos2 (kx x) + cos2 (ky y) + cos2 (kz z) (2.1) m 2 2 2 2 2 2 + (ω x + ωy y + ωz z ). 2 x V0 is the potential depth which is conveniently expressed in terms of the recoil energy Er = ~2 kL2 /2m. V0 is proportional to the power of the laser beams which allows a direct manipulation of the lattice depth in the experiment. The second term in (2.1) represents the additional harmonic confinement due to the spatial intensity variation of the laser beams. Instead of using all three pairs of beams, just overlapping two pairs or one pair will result in a 2D-lattice or 1D-lattice. These different lattice geometries are depicted in Fig. 2.2. At 2.3 Non-Interacting Atoms in Optical Lattice Potentials 7 Figure 2.2: Different Lattice Geometries. By using different geometries and multiple numbers of counter propagating laser beams it is possible to create a multitude of lattice geometries in various dimensions. Shown here is the case of pairs of counter-propagating beams creating confining lattice in 1D (pancake structure), 2D (tube structure) and 3D (cubic lattice structure). the Bose-Fermi mixtures project we use a laser which is far red detuned with respect to the atomic transitions which are at 767 nm for 40 K and 780 nm for 87 Rb. 2.3 Non-Interacting Atoms in Optical Lattice Potentials Considering optical lattices the question arises which eigenstates and energies will be possessed by non-interacting particles. This problem can be solved in a similar way in which the problem of electrons in a periodic potential created by ions in solid-state-physics is solved (see [27–29]). Considering the cubic lattice potential (2.1) without the additional harmonic confinement, inserting this potential into the time independent Schrödinger equation and making use of the fact that the potential is separable with respect to its spatial coordinates, yields a one dimensional equation of the following form ! p̂2 + Vlat (z) ψ(z) = Eψ(z). 2m (2.2) Bloch’s theorem [27] states that the eigenstates of a particle in a periodic potential can be written as ψq(n) (z) = eiqz u(n) q (z) (2.3) 8 2 Ultracold Bose-Fermi Mixtures in Optical Lattices (n) where uq (z) has the periodicity of the underlying lattice. The index n is called band index and q is the quasimomentum. The quasimomentum q is restricted to the first Brillouin zone −kBZ < q ≤ kBZ where R = 2π/alat = 2kL = 2kBZ is the reciprocal lattice vector. Because Figure 2.3: First Brioullin Zone of a Cubic Lattice. Here the first Brillouin zone of a square lattice geometry in reciprocal space is shown. The blue dots represent the atoms in reciprocal space. the potential as well as the eigenstates are periodic with respect to the lattice spacing alat a Fourier expansion is intuitive: ψq(n) (z) = eiqz X (n) cK,q eiKz K Vlat (z) = X vK eiKz . K (2.4) (2.5) Equation (2.4) is called a Bloch state and is denoted by |n, qi in Dirac notation. The index K in the sums has to be considered for all available reciprocal lattice vectors. Inserting equations (2.4) and (2.5) into (2.2) will result in a eigenvalue problem of the form (see [27]) ! X ~2 (n) (n) (q + K)2 − Eq(n) cK,q + vK 0 −K cK 0 ,q = 0. 2m K0 (2.6) The potential is proportional to cos2 (kz z) which can be rewritten to cos2 (kz z) = 1 2ikz z e + e−2ikz z + 2 . 4 (2.7) The only nonzero contributions of the Fourier expansion (2.5) are thus v0 = −V0 /2 and v1 = v−1 = −V0 /4 (see [28, 29]). Using this result and neglecting the v0 contribution equation (2.6) becomes 4Er V0 (n) V0 (n) (n) (n) (q + K)2 cK,q − cK−1,q − cK+1,q = Eq(n) cK,q . 2 R 4 4 (2.8) This equation can be solved numerically by restricting the calculation to a finite amount of different K. All calculations in this thesis are done with K = [−10R, 10R] and assume the same intensity of the laser beams forming the periodic potential. This calculation yields 2.3 Non-Interacting Atoms in Optical Lattice Potentials 9 V0 = 5ErK ~12.5ErRb 60 60 K 87 n=4 50 50 40 40 n=5 n=4 n=3 30 Rb 30 (n) Eq / h [kHz] 40 20 n=3 20 n=2 n=2 10 10 n=1 0 −1 0 1 n=1 0 −1 0 q [kBZ] 1 q [kBZ] Figure 2.4: Band structure for 40 K and 87 Rb. These band structures are calculated for the same intensity of the laser beams which form the periodic potential. Because of the larger mass of 87 Rb and the different wavelengths of the atomic transitions 87 Rb has a smaller recoil energy. The gray shaded areas depict the energies where the bands of both species overlap. (n) values for the energies Eq (n) as well as the Bloch coefficients cK,q which will be used later to (n) calculate the transitions probabilities between different bands. Plotting the energies Eq for different quasimomenta q and different band indices n will reveal a single particle band structure shown in Fig. 2.4. 87 Rb experiences a mRb /mK ≈ 87/40 ≈ 2.18 deeper potential because of different recoil energies. The atomic transitions for the two species are separated by 13 nm making an additional correction necessary which results in a factor of about 2.5 [30]. The relation between the potential depths of 87 Rb and 40 K is given by Rb EK r ≈ 2.5 Er . (2.9) This difference in the lattice depth experienced by the two species results in much smaller bandwidths (maximum to minimum energy of one band) and larger bandgaps (energy difference between the energy minimum of one band to the maximum energy of the next lower band) for 87 Rb in comparison to 40 K. Furthermore bands with different band indices of the two species overlap each other as shown by the gray shaded area in Fig. 2.4. For exmaple the region around 35 kHz where the third band of 40 K coincides with the fourth band of 87 Rb. 10 2 Ultracold Bose-Fermi Mixtures in Optical Lattices 2.4 Ultracold Bosons in Optical Lattices In section 2.3 the Bloch states were introduced to describe non-interacting single particles in an optical lattice potential. While it allows the calculation of eigenenergies and eigenstates and thus gaining insights into systems governed by the kinetics in periodic potentials, it completely neglects the fact that atoms interact with each other. For bosonic 87 Rb-atoms this interaction is repulsive thus the interplay between the interaction and the tunneling energy governs the behavior of the system. This interplay results in interesting effects like a quantum phase transition between the superfluid phase and the Mott-insulator phase [8] which will be described after a short presentation of the model used widely to describe this phases and the transition between them. 2.4.1 The Bose-Hubbard Model The behavior of interacting ultracold bosonic atoms in an optical lattice can be described in terms of a Bose-Hubbard model [7]. The Bose-Hubbard model considers only tunneling between adjacent lattice sites (tight-binding approximation) and restricts itself to the lowest Bloch band (see [7]). Further only on-site interactions are considered. The Bose-Hubbard Hamiltonian is derived from a full Hamiltonian given in second quantization [7, 31] and by changing from a Bloch basis to a Wannier basis. In terms of three dimensional Bloch states a Wannier state is expressed as wn (~r − ~ri ) = √ 1 X (n) ψ (~r) e−iq~ri . NL q q (2.10) Here NL is the total number of lattice sites and ~ri are the lattice site positions. The Wannier states are maximally localized at each lattice site. The Bose-Hubbard Hamiltonian now can be written as ĤBH = −J X † b̂i b̂j + hi,ji X UX i n̂i n̂i (n̂i − 1) − 2 i i (2.11) The first term describes the energy associated with tunneling (i.e. kinetic energy) from a lattice site i to a neighboring lattice site j. Lattice sites i and j are restricted to be nearest neighbors. The tunneling matrix element J is given by J= Z ! w1∗ (~r ~2 2 − ~ri ) − ∇ + Vlat (~r) w1 (~r − ~rj ) d~r 2m (2.12) The second term in (2.11) accounts for the interaction energy between the atoms where n̂i counts the numbers of bosons at lattice site i. U is expressed as U= 4πas ~2 m Z |w1 (~r)|4 d~r. (2.13) Here as is the s-wave scattering length between the bosons. The last term in (2.11) contributes an energy offset i for each lattice site. This offset accounts for the fact that the 2.4 Ultracold Bosons in Optical Lattices 11 0.3 0.2 U U U, J [Er Rb ] J 0.1 J 0 0 5 10 15 Rb Lattice Depth [Er ] Figure 2.5: Bose-Hubbard model. In the Bose-Hubbard model bosons are allowed to tunnel between nearest neighbor lattice sites and thereby gain kinetic energy. The model also includes the effect of interaction which in the case depicted here increases the potential energy of the system. The effect of harmonic confinement is not shown. The graph on the right shows the behavior of U and J depending on the lattice depth. laser beams forming the optical lattice have a Gaussian intensity distribution which yields an additional harmonic confinement (see Sec. 2.2). As outlined above the interplay between U and J leads to the emergence of a quantum phase transition as described in the following sections. 2.4.2 Superfluid Phase When the tunneling parameter J is much larger than the interaction U (which is the case in a shallow lattice), the bosonic gas tries to minimize its energy by tunneling. Because U is small compared to J the system is able to occupy lattice sites with more than one atom. This state is called a superfluid state where every atom is maximally delocalized over the entire lattice. A result of this delocalization is the emergence of interference peaks at ±2~k Figure 2.6: Superfluid Phase. The emergence of coherence peaks is characteristic for an ultracold bosonic gas in an optical lattice potential. during TOF expansion. This can be seen in Fig. 2.6. Here, the bosonic 87 Rb atoms were prepared in a superfluid state. Then the lattice has been switched off instantly and the atoms are allowed to expand freely. 12 2 Ultracold Bose-Fermi Mixtures in Optical Lattices 2.4.3 Mott-Insulator Phase Figure 2.7: Mott insulator phase. The destroyed coherence between the atoms at the different lattice sites result in a broad distribution. In contrast to the aforementioned situation, the case where U J results in a phase called Mott insulator which occurs at higher lattice depths. In this case it is very costly for the system to allow atoms to occupy a lattice site where another atom is located. So the system tries to minimize its energy by distributing the atoms evenly over the lattice. This results in an integer filling of the lattice sites. The fixed atom number per lattice site results in a uncertainty in the phase destroying the phase coherence. 2.4.4 Superfluid to Mott Insulator Transition The two phases explained above are connected via a quantum phase transition. By changing the lattice depth it is possible to change the ratio between U and J and observe the Superfluid to Mott insulator phase transition [32]. In Fig. 2.8 typical TOF pictures for the phase transition are shown. Increasing Lattice Depth Figure 2.8: Phase transition between superfluid and Mott-insulator.. The lattice depth where a transition between the two phases occurs can be estimated from a mean-field approach [32]. For the 3D cubic lattice used at the Bose-Fermi mixtures experiment, the transition is evaluated to a lattice depth of around 13.6 ERb r . One interesting feature of the Mott insulator to superfluid phase transition is the build up of insulating regions which are separated by superfluid domains. This so called wedding cake structure is shown in Fig. 2.9. A wedding cake structure has been observed experimentally with rf-spectroscopy [33] and with single site resolution microscopy [34, 35]. 2.5 Ultracold Fermions in Optical Lattices 13 (a) phase diagram (b) wedding cake structure 3 n= 3 superfluid 2 µ/U n= 2 n= 2 n= 1 1 n= 1 Mott insulator 0 0 0 .0 1 0 .0 2 0 .0 3 J/U Figure 2.9: Phase diagram for bosons in an optical lattice. In an experimental situation the Mottinsulating phases (blue) with fixed numbers of atoms per site n, are separated by Superfluid regions (grey). In a the phase diagram of a Bose-Hubbard model is shown. b shows the wedding cake structure. 2.5 Ultracold Fermions in Optical Lattices In the previous section interacting bosons in optical lattices where studied. While the behavior of bosons is governed by the repulsive interaction and tunneling, fermions are subject to Pauli’s exclusion principle and behave according to Fermi-Dirac statistics. 2.5.1 The Fermi-Hubbard Model for Spin-Polarized Fermions It is possible to describe a single component fermi gas in an optical lattice with a Hubbard type model as done before for bosons. Ultracold fermions prepared in the same spin state do not interact via s-wave scattering because of Pauli’s exclusion principle. P-wave scattering is highly suppressed [36] so interactions are neglected. The resulting Hamiltonian is ĤFH = −J X † ĉi ĉj − hi,ji X i n̂i . i (2.14) The first term describes tunneling between adjacent lattice sites while the former again describes the harmonic confinement of the atoms. For more details see Sec. 2.4.1. 2.5.2 Phase Diagram of Spin-Polarized Fermions The exclusion principle does not allow more than one fermion per lattice site in the same quantum state. According to Fermi-Dirac statistics, fermions fill up all available energy 14 2 Ultracold Bose-Fermi Mixtures in Optical Lattices states up to the fermi-energy beginning with the state with the lowest energy. So intuitively there are two situations one has to consider when looking at a fermi gas in an optical lattice. The first situation arises when the fermi-energy (i.e. chemical potential) is smaller than the bandwidth of the occupied band. This allows the atoms to minimize their energy by tunneling between lattice sites. This state belongs to a metallic phase because of the compressibility of the ensemble. If the fermi-energy is larger than the bandwidth of the lowest band, the fermi gas will occupy all available energy states up to the first bandgap. Thus the whole first Brillouin zone is occupied. This state is called band insulator. The dominant energy scales in a noninteracting fermi gas are the characteristic trap energy [13] Et = Vt γN 4π/3 2/3 (2.15) and the tunneling matrix element J which is connected to the bandwidth of a one dimensional lattice according to (2.16) 4J = max (Eq(n) ) − min (Eq(n) ). Vt denotes the minimal energy offset between two adjacent lattice sites, γ the aspect ratio of the trap, and N the number of atoms in the lattice. The characteristic trap energy denotes the Fermi energy of a noninteracting cloud in the limit of no tunneling [13]. 0.8 N=100000 0.7 energy [EKr ] N=70000 12J 0.6 band insulator N=40000 0.5 0.4 metal Et 0.3 0.2 0.1 0 6 8 10 12 14 16 18 20 lattice depth [EKr ] Figure 2.10: Phases of the non-interacting fermi-gas. In red the tunneling matrix element J is plotted for the 3D case (i.e. three times 4J). The dots represent the characteristic trap energy Et with different numbers of particles. In Fig. 2.10 a phase diagram for non-interacting fermions is shown. All measurements are performed in a three dimensional lattice, the bandwidth has to be considered in three dimensions: 3 · 4 J. The red curve shows 12 J depending on the lattice depth. The dots show the characteristic trap energy for different amounts of fermions. The point at which 2.6 Band Mapping 15 the characteristic trap energy gets larger than the tunneling bandwidth is the point at which all fermions are in a Band-insulating phase. It is important to note that a fraction of the atomic sample can be in a band insulating phase while others are in a metallic phase. This is because the trap energy depends on the position in the trap so the 12J boundary can be reached at much lower lattice depths. Using the band mapping technique (see Sec. 2.6) it is possible to distinguish the two phases. In a shallow lattice (for example V0 ≈ 3 EK r ) we expect to be in a metallic phase. We expect to see a washed out first Brillouin zone with a pronounced center towards small momentum components. In the case of deep lattices the Brillouin zone should be evenly filled and we expect to see sharp edges at the borders. Fig. 2.11 shows ultracold fermionic Potassium loaded into a cubic optical lattice of varying lattice depths. Because of the cubic lattice geometry, also the first Brillouin zone is cubic. The lattice in our experimental setup is tilted by 45 degrees with respect to the detection axis so we expect to see a rectangular Brillouin zone elongated in the direction of the tilting (see Fig. 2.1 and Fig. 2.11). The above explained behavior is observed. A shallow lattice leads to a pronounced center and a washed out Brillouin zone. Increasing the lattice depth results in a evenly filled Brillouin zone corresponding to fermions in a Band-insulating phase. It is not possible to find a distinct point where the phase transition occurs. We attribute this to imperfect Band-mapping and the fact that parts of the fermions can be in a band-insulating phase while other are not (see above). orientation of the 1.BZ V0 = 3ErK V0 = 6ErK V0 = 9ErK V0 = 12ErK Figure 2.11: Phase transition from metallic phase to band insulating phase. Shown is the orientation of the first Brillouin zone on the current experimental setup. On the right the resulting TOF pictures are presented for different lattice depths. 2.6 Band Mapping Band-Mapping Principles One experimental challenge is to observe and detect the momentum resolved population of occupied bands. Because q is restricted to the first Brillouin zone, and can always be transformed into this zone, the quasimomentum is not an eigenstate of the real momentum operator. This means that turning off the lattice will project the quasimomenta onto real momenta. Information about the population of bands is lost. This problem can be circumvented by ramping down the lattice with a ramp duration chosen such that no interband transitions are possible [37, 38] which means that the atoms will be mapped onto real momenta. The band mapping time (i.e. ramp down time) tBM has to be slower than the time 16 2 Ultracold Bose-Fermi Mixtures in Optical Lattices 45 40 35 E / h [kHz] 30 25 Eq(n) 20 15 ħ2k2 2m 10 5 0 −3 −2 −1 0 1 2 3 k [kBZ] Figure 2.12: Band mapping. Ramping down the lattice maps quasimomenta onto real momenta. associated with the energy of the bandgap, but has to be faster than the time it would take the atoms to redistribute their momentum within a band due to trap dynamics. With this the criterion for effective band mapping is given by [39] EBG h/tBM ~ωt . (2.17) Here EBG is the energy of the smallest bandgap (between bands to be observed) and ωt is the trap frequency. Useful band mapping times for our experimental conditions are in the order of a couple of hundred µs to some ms as discussed in the next section. These times are consistent with the observations made by other groups [39]. Mapping of the First Brillouin Zone with Fermion We can use the knowledge about the appearance of the first Brillouin zone discussed in Sec. 2.5.2 to determine useful band mapping times. We therefor prepare 40 K in a deep optical lattice (V0 ≈ 12 EK r ) and vary the time in which the lattice is ramped down between 200 µs and 10 ms. The resulting pictures are presented in Fig. 2.13. On the left side are the resulting TOF images shown. On the right the row- and column sums of these pictures are presented. The row sum shows the characteristic triangular structure because of the tilting of the camera axis with respect to the lattice axis (see Sec. 2.1 and Sec. 2.5.2). A mapping time of 200 µs results in a slightly rounded Brillouin zone as seen in the corresponding column sum. Mapping times on the order of 1 ms to 3 ms result in very homogeneous and sharp edged column sums and very pronounced triangular shapes in the row sum. After 5 ms a small peak starts to appear in the column sum, which further develops at a mapping time of 10 ms. For even longer mapping times the atomic distribution would transform back to an undisturbed fermi cloud governed by the optical dipole trap. From this series of pictures 2.6 Band Mapping 17 it is clear that mapping times longer than 5 ms are not practicable because the resulting momentum distribution is starting to be dominated by the momentum distribution of the trap. Good mapping times for our experimental setup thus are on the order of 1 ms to 3 ms whereby a reasonable mapping is also possible for 200 µs. TOF image row sum column sum 200µs 1ms 3ms 5ms 10ms Figure 2.13: Result of Different Band Mapping Times. Different band mapping times (i.e. ramp down time of the lattice) result in different mapping of the first Brillouin zone of a fermi gas loaded into an cubic optical lattice. 18 2 Ultracold Bose-Fermi Mixtures in Optical Lattices 19 3 Lattice Modulation Spectroscopy The high control of optical lattice systems together with advances in generation and preparation of ultracold fermions and bosons is a promising combination of tools to understand and explore fundamental processes and interaction phenomena of many body physics. Spectroscopical research on ultracold quantum gases has been proven as a key method to access the underlying physics. Different spectrosopical methods have been developed to access the available momentum states. It has been used to examine fundamental properties like the excitation spectrum a of single component BEC [40]. The method has been used to probe the Bogoliubov excitation spectrum of a bosonic superfluid in an optical lattice [41, 42]. Very recently, it could be used to detect the Amplitude Mode in the crossover region between a Superfluid and an Mott insulator [43]. For fermions Bragg spectroscopy has been used to probe the BEC-BCS crossover regime [44]. Photo emission spectroscopy based on radiofrequency (rf) pulses, has been used to observe the pairing gap of strongly interacting fermions and to probe the single particle excitation spectrum of fermions near the BEC-BCS crossover [45, 46]. In optical lattices, modulating the lattice depth will transfer energy to the system. This has developed into an easy to implement, multipurpose spectroscopy method. It has been used to study the excitations of bosonic atoms to higher bands [47], the two dimensional Mott-insulator to Superfluid transition [48] and is proposed to study the spectral function of ultracold bosons in optical lattices [49]. The detection of a gapped excitation mode in the fermionic Mott-insulating phase [12, 50] has been carried out by lattice modulation spectroscopy. Great effort is put into the realization of antiferromagnetic ordering and modulation spectroscopy is again proposed as a mean to detect and observe antiferromagnetic ordering [51, 52]. In the framework of this thesis a novel spectroscopy method has been developed. It is based on lattice amplitude modulation and band mapping, and allows momentum resolved spectroscopy of ultracold atoms in optical lattices. The method allows us to precisely determine the potential depth experienced by the atoms and from that accurate determination of tunneling energies and rates. In this chapter, first lattice modulation spectroscopy is described in a general way. Expressions for allowed transitions and their strength are theoretically derived. The influence of modulation amplitude and time is examined. Based on these considerations the band structure of non-interacting fermions is probed with momentum resolution, showing the accuracy of the presented method. Lastly the band structure for bosons is examined. 20 3 Lattice Modulation Spectroscopy 3.1 Momentum Resolved Lattice Modulation Spectroscopy Emod / h [kHz] The described method of lattice modulation spectroscopy consists of two experimental techniques: The first is the modulation of the lattice amplitude, which transfers atoms to higher bands creating particle-hole pairs. This creation of particle-hole pairs conserves the quasimomentum (see Fig. 3.1). Forming an band insulator, fermions occupy the whole first Brillouin zone. Bosons fill the first Brillouin zone when trapped in a deep lattice being in a Mott-insulator phase. Because of the different curvature of the bands, an specific amount of imprinted energy will have a resonance at one specific momentum. The second technique reveals the occupation of higher bands and is the band mapping procedure (see Sec. 2.6). The spectroscopical information about the system under study is extracted by varying the lattice modulation frequency and recording the position of the transferred atoms after time-of-flight (TOF). Particle Hole −1 0 1 q [kBZ] Figure 3.1: Effect of modulating the lattice amplitude. Modulating the amplitude of the optical lattice will excite atoms from the ground state to higher bands creating particle-hole pairs. The transfer will preserve the quasimomentum q reflected by the straight lines. Experimental Procedure The experiments performed always start with an ultracold sample of fermionic 40 K in the hyperfine state F = |9/2, 9/2i or a mixture of 40 K and bosonic 87 Rb (in hyperfine state F = |1, 1i) in an optical dipole trap. Thereon an optical lattice is superimposed. Afterwards the amplitude of two lattice laser beams is modulated. This is done by varying the power of the laser beams which form the corresponding axis of the optical lattice. The modulation is applied for 1 ms with a relative amplitude of 15% avoiding beyond linear response effects. The frequency of the amplitude modulation determines the imprinted energy onto the system. Thus the variation of the frequency allows addressing of all bands. Most of 3.1 Momentum Resolved Lattice Modulation Spectroscopy 21 the measurements performed in the context of this thesis have been done while transferring atoms to the third band. This band has the highest transition probability (see Sec. 3.2) and thus give best signal to noise ratios. After modulating, the lattice is ramped down during lattice ramp up 100ms latt. mod. 1ms latt. ramp down 200µs time of flight 20ms Figure 3.2: Experimental sequence of lattice modulation spectroscopy. Modulation spectroscopy consists of two key procedures namely the modulation of the lattice and mapping of the population of bands to reveal the momentum distribution by a fast lattice ramp down. 200 µs in order to map the quasimomentum onto real momenta. The band mapping time of 200 µs is chosen to prevent dynamics induced by the harmonic trapping potential (see Se. 6). Afterwards the atoms are released from all trapping potentials and expand freely for typically 20 ms thus revealing the momentum distribution. The whole procedure is sketched in Fig. 3.2. Effective Band Structure It is important to note that the first band is within our typically used lattice depths almost but not entirely flat. So the energy required to transfer atoms at a certain momentum to an higher band will be the difference between the energies of the higher band and the first band at that momentum. Hence an effective band structure is observed as shown in 3.3. This effective band structure is directly connected to the real band structure thus in the following the effective band structure will be denoted as the band structure. Eq(n) / h [kHz] 40 Band Str. Eff. Band Str. 30 20 10 0 −1 0 q [kBZ] 1 Figure 3.3: Effective band structure. Lattice modulation spectroscopy transfers atoms from the first band to higher bands. The first band is slightly curved even for higher lattice depths. This results in an effective band structure shown here. In light orange the original band structure for 40 K at a lattice depth of V0 = 7.5 EK r is shown. In red the effective band structure incorporating the energy offset from the first band is depicted. 22 3 Lattice Modulation Spectroscopy Deducing the Momentum from TOF Images without modulation with modulation Figure 3.4: TOF images obtained after lattice modulation and band mapping. Modulating the amplitude of the lattice results in creation of particle-hole pairs. Mapping the quasimomentum to real momenta and letting the atoms expand freely allows extraction of the population. The left side shows the images obtained without any modulation. The right side shows a typical picture where the lattice has been modulated in one lattice axis. The sharp peaks left and right of the first Brillouin zone are atoms transferred to the third band. The distance between the center of the first Brillouin zone and the excited atoms is directly related to the quasimomentum (see text). In a typical experimental situation the above outlined procedure results in TOF images shown in Fig. 3.4 (fermionic atoms). The left side shows the sequence without any modulation resulting in pictures of the first Brillouin zone. On the right side the effect of modulating the lattice is shown. The modulation is performed in one axis of the lattice. Atoms are transferred from the first Brillouin zone to a higher band (here the third) at a specific momentum (sharp peaks to the left and right). The momentum of the excited atoms is related to the distance observed on the image from a CCD camera by p= (n − 1 + q)~kL m ttof . mK rpixel (3.1) Here p is the distance between the center of the first Brillouin zone and the transferred atoms on the camera in pixel. rpixel is the size of a pixel on the camera and m the magnification of the detection optics. n is the band, q the quasimomentum and ttof the time the atoms expand freely. All parameters are known and p can be determined very accurately. Using this relation it is possible to reconstruct the momentum q from the TOF images. This data combined with the frequency of the lattice modulation gives access to the band structure of the lattice. By modulating another lattice axis or modulate more than one axis it is possible to address bands in other directions. 3.2 Chracterization of Modulating the Lattice Amplitude This section concentrates on induced transitions between different bands. In the first part we will derive expressions for the strength of the coupling between different bands. The second part characterizes the transfer to higher bands for different modulation amplitudes and different modulation times experimentally. 3.2 Chracterization of Modulating the Lattice Amplitude 23 3.2.1 Theory of Interband Transitions To understand the effect of modulating the amplitude of the lattice one has to consider the effect of an oscillating amplitude of the lattice potential. We can incorporate the modulation by adding a time-dependent oscillating perturbation to the Hamiltonian (2.2) p2 (3.2) + V0 cos (kz z)2 [1 + cos (ωt)] . 2m Here is a small number representing a small perturbation. In a periodic potential it is intuitive to use the Bloch states introduced in Sec. 2.4 as the eigenstates of the nonperturbed system: H= hz|n, qi = ψq(n) (z) = eiqz X (n) cK,q eiKz (3.3) K with (3.4) H0 |n, qi = Eq(n) |n, qi. where Ĥ0 is the non-perturbed Hamiltonian. To answer the question which transitions between the eigenstates of the system are possible we calculate the transition probabilities between two Bloch states. Therefore we express the perturbation as (3.5) V̂ (z, t) = V0 cos (kz z)2 cos (ωt). Using Fermi’s golden rule [53] the transition probability between two bloch states with momenta q, q 0 and band indices n, m is (3.6) nm 0 2 Wqq 0 ∝ |hn, q|V̂ (z)|m, q i| f (T, ω0 − ω). Here f (T, ω0 − ω) is a function describing the response of the coupling on the modulation time T and the modulation frequency ω. ω0 is the resonance frequency of the transition. For long modulation times, f (T, ω0 −ω) can be reduced to a Kronecker delta. The remaining part of the perturbation is now conveniently expressed as 2 p H0 − 2m V (z) = cos (kz z)2 = . V0 V0 (3.7) Using this and inserting into (3.6) yields nm 0 2 0 2 Wqq 0 ∝ | hn, q|Ĥ0 |m, q i −hn, q|p̂ |m, q i| . | {z (3.8) } 0 Here ω = ω0 and constant factors have been neglected. The first term is zero for n 6= m and/or q 6= q 0 . With p̂2 ∝ ∂z2 and inserting the Bloch states (3.3) nm Wqq 0 ∝ | X (n) (m) Z 0 0 e−iqz e−iKz ∂z2 eiq z eiK z dz |2 cK,q cK 0 ,q0 K,K 0 ∝| X (n) (m) 0 0 2 cK,q cK 0 ,q0 (q + K ) K,K 0 Z 0 0 −i(q−q )z −i(K−K )z 2 |e {z } e| {z } dz | . δq,q0 δK,K 0 (3.9) 24 3 Lattice Modulation Spectroscopy Using the Kronecker delta this becomes Wqnm ∝ | X (n) (m) cK,q cK,q (q + K)2 |2 (3.10) K This equation incorporates one very important fact: modulating the amplitude of the lattice only couples states with the same q. This means that no momentum transfer other than integral multiple of the reciprocal lattice vector are allowed. In this thesis the transi- Wnm q [a.u.] 1.0 1↔5 1↔4 1↔3 1↔2 1.0 0.5 0.5 0 -1 1↔5 1↔4 1↔3 1↔2 0 q [kBZ] 1 0 -1 0 q [kBZ] 1 Figure 3.5: Interband transition probabilities due to lattice modulation. Shown are the transitions probabilities Wqnm between n = 1 and m = 2, 3, 4, 5. Calculated for a lattice depth of 2.5 EK r on the left on the right. and 7.5 EK r tion probabilities using experimental parameter have been calculated using Eqn. (2.3) and Eqn. (3.10). The transitions possibilities for transitions between n = 1 and n = 2 to n = 4 are shown in 3.5. The transition to the third band is for every momentum much higher than for transitions to all the other bands. A very distinct feature is observed for even band indices at momentum zero. Here a transition is always forbidden. The curves for even bands will approach zero for every q for deep lattices. Thus in a deep lattice, modulation of the amplitude does not induce transitions between the first and any even numbered band. In a simple picture this can be understood by approximating a deep lattice site with an harmonic oscillator. The eigenstates of the harmonic oscillator have alternating parities. Lattice modulation is a even parity operation [47] thus only couples states with the same parity. Hence transitions between the first (i.e. lowest) eigenstate of the harmonic oscillator and the second, fourth etc. are not allowed. This explains the suppression of transitions to even numbered bands in deep lattices. 3.2.2 Time and Amplitude of the Modulation For effective spectroscopy several criteria have to be fulfilled. First a proper signal to noise ratio has to be obtained. Thus the number of excited atoms has to be sufficient. On the other hand it is necessary to keep the overall perturbation of the system small. A large perturbation can lead to beyond linear-response effects, for example final-state interaction and higher order processes. 3.2 Chracterization of Modulating the Lattice Amplitude 25 The behavior of fermions subject to lattice modulation is studied by changing the amplitude of the modulation and varying the modulation time. Both affect the number of atoms transferred to higher bands (here the third). In Fig. 3.6 the relative amount of excited atoms is shown. The amount is normalized to the total number of atoms. The data has been experimentally obtained for 40 K loaded into an optical lattice with a lattice depth of 7.5 EK r . Fitting a gaussian to the momentum distribution of the excited atoms yield an estimate for the width ∆k of the momentum distribution of the atoms in the third band. The momentum width is thereby deduced taking the FWHM of the Gaussian. Longer modulation times as 8 8 excitation fraction [%] (b) excitation fraction [%] (a) 6 4 2 0 5 0 10 15 20 25 30 35 modulation intensity [%] 40 6 4 2 0 45 (c) 0 0.2 0.2 Δk [kBZ] 0.25 Δk [kBZ] 1 .5 1 modulation time [ms] 2 1 1.5 modulation time [ms] 2 (d) 0.25 0.15 0.1 0 .5 0.15 0 5 10 15 20 25 30 35 modulation intensity [%] 40 45 0.1 0 0.5 Figure 3.6: Number of atoms and momentum width for different modulation amplitudes and modulation times. The amount of excited atoms depend on the modulation time and the amplitude of the modulation. Experimentally obtained data is shown here for different durations and relative amplitudes of modulation. Data is obtained for 7.5 EK r . Shaded are the parameters chosen for the following measurements. The black dotted curves are a guide to the eye. For a and c a modulation time of 1 ms is used. In b and d a realtive modulation amplitude of 15 % is applied. well as higher modulation amplitudes yield more excited atoms. Decreasing either the modulation time or the modulation amplitude will reduce the width of the momentum distribution. Considering Fig. 3.6a and Fig. 3.6b one can see a decrease in the slope of the curves. We attribute this behavior to a decrease in the number of atoms available in the first band which leads to a reduced number of atoms transferred to the third band. Higher order processes and transfers from the third band to higher bands could be also possible. To examine the coherence properties of lattice modulation, a high modulation intensity 26 3 Lattice Modulation Spectroscopy (40%) has been used while varying the modulation time. This measurement is presented in Fig. 3.7. Damped Rabi oscillations between the numbers of atoms in the first Brillouin zone and the third are observed. Rabi oscillations are a effect which is not described by linear-response theory. This means that a large modulation amplitude will substantially perturb the system which could result in a disturbed spectroscopy signal. atom fraction [%] 100 80 first band 60 40 third band 20 0 0 0.5 1 1.5 2 modulation time [ms] Figure 3.7: Coherence of the lattice modulation process. Lattice modulation drives coherent Rabi oscillations. The blue curve shows the amount of atoms in the first band. The green curve shows the relative amount of atoms the the third band. This data is obtained using a large modulation amplitude of 40%. The decrease of the slope in the curves for the number of transferred atoms and the emergence of Rabi oscillations for high modulation amplitudes are both signatures of beyond linear-response effects. To keep these effects small but still get a usable signal, a modulation time of 1 ms and a modulation amplitude of 15 % has been chosen for all measurements when not mentioned otherwise. 3.3 Band Structure of Spin-Polarized Fermions Lattice modulation spectroscopy can now be employed to measure the spectrum of a spinpolarized sample of fermionic 40 K. In Fig. 3.8 the experimentally obtained data for a lattice depth of 5 EK r is shown. Each horizontal line corresponds to one modulation frequency. The data is obtained by taking the column sum of the pictures obtained after TOF (see for example Fig. 3.4). The frequency is varied between 0 kHz and 55 kHz in steps of 500 Hz. The color represents the amount of atoms. The amount of atoms is normalized hence red is close to one and white zero. The x-axis shows the momentum calculated for a given time-of-flight (here 20 ms) in terms of the reciprocal lattice vector. Starting at 0 kHz the first Brillouin zone is clearly seen. No transfer to higher bands happen before reaching around 13 kHz. This corresponds to the first bandgap between the first and the second band (compare with Sec. 3.3). Increasing the energy (i.e. frequency) further, the first atoms are excited to the 3.3 Band Structure of Spin-Polarized Fermions 27 n=4 50 Emod / h [kHz] 40 30 n=3 20 n=2 10 0 -3 -2 -1 0 1 2 3 k [kBZ] Figure 3.8: Band structure for 40 K. The band structure for fermions at a lattice depth of 5EK r is depicted here. Shown is the energy imprinted onto the system in terms of the lattice modulation frequency depending on the obtained momentum distribution. A red color expresses a high amount of atoms where blue and lighter colors represent few or no atoms. Visible are the second, third and fourth band together with the corresponding bandgaps. second band. A very distinct feature arises here. Comparing with the bandstructure it can be seen that the second band has a negative slope. Thus the energy difference between the first and the second band is smaller for higher quasimomenta so the first atoms which are transferred to the second band are at q at the edge of the Brillouin zone. Imprinting higher energies result in particle-hole pairs at lower momenta. This behavior is clearly seen in the first Brillouin zone (corresponding to the reduced zone schema). The second bandgap between the second and third band occurs at around 22 kHz. The third band is again clearly seen while it is not possible to observe the third bandgap which is in the order of 300 Hz which is below the resolution of the measurement. Band Structure of Fermions for Different Lattice Depths This measurement has been repeated for different lattice depths shown in Fig. 3.9. The bandwidths decrease with deeper lattices while simultaneously the widths of the bandgaps increase. A number of distinct features are observable. For small lattice depths the number of atoms excited to the second and third band is nearly the same. At small momenta more atoms are transfered to the third band than to the second while for higher momenta this 28 3 Lattice Modulation Spectroscopy (a) 1.5 EK r (b) 5 EK r 50 50 40 40 30 30 20 20 10 10 0 −3 −2 −1 0 1 2 3 0 −3 −2 (c) 8 EK r (d) 11 EK r 50 50 40 40 30 30 20 20 10 10 0 −3 −2 −1 0 1 2 3 0 −3 −2 −1 0 1 2 3 −1 0 1 2 3 Figure 3.9: Band structures for different lattice depths. Shown is the data obtained from lattice modulation spectroscopy with fermions for different lattice depths. With increasing lattice depths the bandgap widths are increasing while the bandwidths are decreasing. The amount of transferred atoms to even bands is getting smaller while the width of the excited atom momentum distribution get larger. behavior turns around. This is consistent with the considerations from Sec. 3.2.1 in particular the left side of Fig. 3.5. At higher lattice depths it is expected that the amount of atoms excited to the second band is decreased which is again in agreement with the theoretical prognosis. Smaller bandwidths (i.e. flat bands) together with Fourier broadening of the imprinted energy yields a broader momentum distribution of the transferred atoms in the third band as well as in the second. This is very nicely shown in the band structure plot 3.9 for 11 EK r . Reduced Zone Scheme A more detailed view on the band structure of fermions is obtained by subtracting the atomic distribution in the first Brillouin zone without any excitation (energies less than 8 kHz) from the complete dataset shown in Fig. 3.9. The data obtained in such a way is shown in Fig. 3.10. The red areas show regions where more atoms are present in comparison 29 particles 3.3 Band Structure of Spin-Polarized Fermions n=4 50 30 n=3 0 Emod / h [kHz] 40 20 n=2 holes 10 0 -3 -2 -1 0 1 2 3 k [kBZ] Figure 3.10: Detailed band structure for 40 K. The data has been normalized to the amount of atoms present in the first Brillouin zone. In red colored areas atoms are present. The blue areas represent holes. Clearly visible is a reduced zone scheme in blue while in red the momentum distribution is shown. Lattice depth 5EK r . The black dotted circles mark regions where the band-mapping procedure failed. to the case without any excited atoms. The blue area corresponds to regions where less atoms are present. The dark blue region between +kBZ and −kBZ can be interpreted as the reduced zone scheme which is now clearly visible. This area represents holes. The dark red areas outside of the first Brillouin zone are the excited atoms (i.e. particles). The picture shows imperfections in the band-mapping process marked with black circles. These imperfections are not limiting because their signal is so low that no influence on the momentum distribution is observable. Dispersion relation for the third band By using equation (3.1) a detailed analysis of the energy-momentum relation for the third band is accessible. In Fig. 3.11 the lattice modulation frequency is shown depending on the momentum for a lattice depth of 11.3 EK r . The red points represent the mean momentum for the excited atoms. The red line shows the band structure for the third band calculated as shown in Sec. 3.3. The black dotted line shows a numerical simulation which incorporates the lattice potential as well as the harmonic confinement induced by the lattice laser beams. The light and dark shaded area shows a ±5% and ±10% deviation of the band structure shown by the red line. The bar graph on the right side shows the fraction of atoms transferred to the third band. 30 3 Lattice Modulation Spectroscopy experimental data 55 lattice + confinement for 11.3 EK r Emod / h [kHz] lattice for 11.3 EKr 50 45 ± 5% 40 ± 10% 0 0.2 0.4 0.6 0.8 q [kBZ] 1 0% 10% excited fraction Figure 3.11: Detailed analysis for fermions in the third band. Here the third band has been analyzed. Red points denote the mean momentum for a lattice depth of 11.4EK r . The red curve shows a fitted band structure where fitting has been limited to 0.4kBZ and 0.8kBZ . The black dashed line shows a simulation of the system incorporating the periodic potential and the lattice potential. The shaded areas show 5% and 10% deviations from the band structure at 11.4EK r . The red line shows the third band for a lattice depth of 11.3ErK . This depth has been deduced by a fit to the experimental data. The experimental data shows deviations in the shape of the curves between the expected band structure (i.e. red line) and the experimental data at high momenta and at momenta below 0.4 kBZ which is consistent with the simulation. The fit thus has been limited to the region between 0.4 kBZ and 0.8 kBZ . From this fit a lattice depth of 11.3EK r ± 0.1% is deduced. The error presented her is the standard deviation of the fit. It is important to note that the variation of the lattice depth through experimental imperfections is of the order of 2%. This precision thus governs the findings presented here. This analysis is performed for all data shown in Fig. 3.9. For each dataset a fit has been performed. The result is shown in Tab. 3.1. Image 3.9a 3.9b 3.9c 3.9d V0,fit [EK r ] 1.4 ±0.5% 5.0 ±0.6% 8.7 ±0.5 11.3 ±0.1% J [EK r ] 0.164 0.0658 0.026 0.0143 Table 3.1: Fitted lattice depth for spin polarized fermions with resulting tunneling energies. Presented are the lattice depths obtained from the band structures in Fig. 3.9. From the obtained lattice depths, the tunneling energy is calculated. 3.3 Band Structure of Spin-Polarized Fermions 31 The lattice depth data shown in Tab. 3.1 allows us to determine the tunneling energy of the fermionic sample. Using (2.12) the tunneling energy is calculated. Effect of harmonic confinement and finite modulation time The black dotted line in figure 3.11 shows a numerical simulation which includes the effect of harmonic confinement due to the Gaussian intensity distribution of the laser beams and the finite modulation time. The simulation has been performed for the deduced lattice depth of 11.3 EK r and shows three distinct features which all can be seen in the experimental data. First at small momenta and energies an overall shift to higher momenta is seen. Excitation of atoms at momenta near q = 0 is suppressed and excitations are possible before reaching the expected energy momentum relation. At low to medium momenta a shift to higher energies is observed while at high momenta a shift to higher energies is visible while excitations of atoms at high momenta are suppressed. The two effects leading to deviations are now explained more thoroughly. Finite modulation time. Exciting atoms with less modulation (i.e. energy) than required can be understood by considering the finite modulation time of the perturbation by lattice modulation. In first-order perturbation theory, the probability of transferring an atom to an excited state is given by (3.11) P (νmod ) ∝ sinc2 [tpulse (ν0 − νmod )]. Here νmod is the modulation frequency and ν0 the resonance frequency. In figure 3.12a the perturbation of the system depending on time is shown. The atomic response is plotted in figure 3.12b. This broadening mechanism is called Fourier broadening. From figure 3.12b (a) (b) 1 atomic response [a.u.] perturbation amplitude [a.u.] 1 0 .6 0 .2 − 0 .2 − 0 .6 −1 0.8 0.6 0.4 0.2 0 −1 0 time [ms] 1 38 40 42 frequency [kHz] Figure 3.12: Broadening due to finite modulation time. Due to the finite modulation time (a) the transitions probability (atomic response) behaves as shown in b. it can be deduced that the response peak is broadened with a full width at half maximum 32 3 Lattice Modulation Spectroscopy (FWHM) of approximately 1kHz. Further sidebands develop at ±1.5kHz around the center frequency. Due to sidebands and broadening the atoms can be transferred to another band is reached earlier for low momenta and is left later for high momenta. Harmonic confinement A second effect which leads to deviations from the theoretical expected band structure is the harmonic confinement induced by the lattice laser beams and the optical dipole trap. The lattice depth is varied by changing the power of the laser beams thus a different lattice depth leads to a different harmonic confinement. This spatial harmonic confinement induces a harmonic confinement in reciprocal space. The confinement alters the density of states (DOS) of the system [54, 55]. In reciprocal space this means that more states are available at medium to higher momenta and few states are available for low momenta. This affects the DOS and depends strongly on the lattice depth. This feature is seen when comparing the beginning of the third band for a very shallow lattice (Fig. 3.9a) and a deep lattice (Fig. 3.9d). In the shallow lattice it is possible to excite atoms at q = 0 while in the deep lattice we observe no transfer. So a shift of the density of states to higher energies has occurred. Because in the experiment mean momenta are detected the result is a deviation to higher momenta for low energies. The effect of harmonic confinement and the effect of broadening due to finite modulation times is depicted in Fig. 3.13. (a) shallow lattice Emod / h [kHz] Emod / h [kHz] (b) deep lattice −1 0 q [kBZ] 1 −1 0 1 q [kBZ] Figure 3.13: Effect of harmonic confinement and finite modulation time. The density of states of the first band is altered significantly due to the harmonic confinement from the optical dipole trap and the lattice laser beams. A deep lattice results in a smaller density of states for low and high momenta. This results in a shift to higher momenta. The dotted arrows represent the Fourier broadening. 3.3 Band Structure of Spin-Polarized Fermions 33 Lattice Calibration The precise determination of lattice depths makes the method of lattice modulation spectroscopy with fermions exceptionally well suited to determine the laser power necessary to create a lattice with a specific depth. The standard method [56] to calibrate an optical lattice has been the use of a BEC loaded into the lattice. Successive modulation of the lattice amplitude imprint energy onto the system. The modulation frequency is chosen such that a transfer into the third band is possible. Because the BEC has a momentum distribution sharply peaked at q = 0 a resonance at the corresponding energy is expected. On the resonance, the sample is substantially heated which can be observed as shown in 3.14. (a) (b) rel. atom number central peak 1 0.75 0.5 0.25 15 15.5 16 16.5 17 17.5 18 modulation frequency [kHz] Figure 3.14: Principles of lattice calibration using bosons. After preparing bosons in an optical lattice, the lattice amplitude is modulated. When the modulation frequency is such that a transfer to the third band is possible energy is imprinted onto the system. This energy can be observed as shown in a. Varying the modulation frequency yields a resonance as shown in b. The resonance position is then compared to the theoretically possible transition energies and thus a lattice depth is deduced. The interaction between the bosons lead to a shift of the resonance position. Further the influence of the harmonic confinement (see 3.3) leads to errors. By employing lattice modulation spectroscopy with fermions, these problems can be circumvented. This method is now standard at the "Bose-Fermi mixtures" experiment in Hamburg. 34 3 Lattice Modulation Spectroscopy 3.4 Spectroscopy on Pure Bosonic Samples In this section the method of lattice modulation spectroscopy is employed using a pure bosonic sample. Bosons are not subject to Pauli’s exclusion principle thus the lowest band is not filled in the same way it is for fermions. To allow momentum resolved spectroscopy the bosons have to occupy higher momentum states. This is achieved by preparing 87 Rb in a deep lattice (around 16 ERb r ) thus being in a Mott insulating phase. Due to the spatial localization the atoms essentially are found in a Fock state thus all momentum components are occupied. The result of this measurement is shown in Fig. 3.15. On the y-axis again the energy given by the modulation frequency of the lattice is shown. Each row shows the column sum of the momentum distribution obtained after time of flight imaging. It is nicely seen that the atoms do not homogeneously fill up the first Brillouin zone. Especially the middle to higher momentum parts are only populated with few atoms. We attribute this to interaction effects between the bosons. The interaction leads to a redistribution of the atoms resulting in a shift to lower momenta. For low energies no atoms are transferred to higher bands due to 50 Emod / h [kHz] 40 30 20 10 0 −5 −4 −3 −2 −1 0 1 2 3 4 5 k [𝛑/ a] Figure 3.15: Band Structure for bosons. Shown is the band structure for bosons. Lattice depht V0 = 16.5 ERb r . Nicely seen is the inhomogenous filling and the broad excitation. the first bandgap. A population of the second band is not observed which is consistent with the findings of Sec. 3.2.1 where for deep lattices less coupling between the first and even numbered bands is expected. At higher modulation frequencies the third band is populated. The momentum distribution is very broad and only very few atoms are excited resulting in 3.5 Summary 35 a very weak signal. Some but very few atoms are transferred to the fourth band so no clear signal can be obtained for higher bands then the third. The third band is further analyzed by extracting the momentum of the excited atoms for different energies. Plotting yields a band structure for the third band analogous to what has been shown for fermions. The result is shown in Fig. 3.16. Fitting a single particle 32 experimental data lattice for 16.4 ERb r Emod / h [kHz] 30 28 26 24 ± 5% 22 ± 10% 20 0 0.2 0.4 0.6 q [𝛑/ a] 0.8 1 0% 10% excited fraction Figure 3.16: Detailed band structure for bosons. Shown is the third band. Obtained using lattice modulation spectroscopy, large deviations from the ideal band structure are observed. band structure to the data (again limiting to the region of 0.4 kBZ to 0.8 kBZ ) yields a lattice depth of 16.6 ERb r ± 1%. The theoretical expected curve is shown in red. Deviations are again observed at low to medium and high momenta. The shifts are again induced by the harmonic confinement, the finite modulation time and the interaction. The small and broad signal results in a one order of magnitude higher error in the fit. This high error and the big deviations as wells as the low signal do not allow to extract useful informations from the measurement. 3.5 Summary In this chapter the principles of lattice modulation spectroscopy were presented. It has been shown that amplitude modulation of an optical lattice allows the creation of particle-hole pairs while conserving the quasimomentum. Band mapping reveals the quasimomentum distribution. This techniques allow sensitive probing of the band structure for fermionic samples. A significant influence of the harmonic confinement from the lattice laser beams and the finite modulation time has been observed. The dispersion relation for bosons could be extracted but because of the deep lattice and the inhomogenous filling of the first 36 3 Lattice Modulation Spectroscopy band precise measurements are difficult. The obtained data for fermions on the other hand allows a precise measurement of the lattice depth experienced by the fermionic sample. This information gives direct access to the tunneling energy and is crucial for most experiments performed with optical lattices. The next chapter of this thesis shows the application of modulation spectroscopy to a mixture of bosons and fermions. 37 4 Lattice Modulation Spectroscopy with Bose-Fermi Mixtures Mixtures of ultracold atoms obeying different quantum statistics have been developed into a powerful tool for simulating solid-state systems, as well as giving access to fundamental quantum mechanics. The interactions between bosons and fermions strongly influence the behavior of many-body systems. One of the first observations while studying Bose-Fermi mixtures in optical lattices, was a decrease in the bosonic visibility when mixed with fermions [19–21]. This observation has not been fully understood until today. Various theoretical proposals have been made. One reason for the observed shift could be an adiabatic heat-up of the mixture in the lattice [22]. Other models incorporate higher bands and yields an effective potential approach [23] reducing the tunneling rate of the bosons. This ansatz has been further developed, incorporating non-linear corrections to tunneling, renormalized two-body interactions and effective three body interactions [57]. In this thesis, lattice modulation spectroscopy is used to study the excitation spectrum of fermions with an admixture of bosons. The high precision of this spectroscopy method allows to determine the lattice depth with high accuracy. Because the tunneling rate is directly connected to the depth of the lattice the method is very promising to study variations in the tunneling rate of fermions with an admixture of bosons. In this chapter the band structure of fermions is probed with an admixture of different amounts of bosons. First, a theoretical description of Bose-Fermi mixtures in optical lattices is given. Afterwards the experiments conducted are presented and the results obtained are shown. 4.1 Ultracold Mixtures of Bosons and Fermions Bose-Fermi-Hubbard Hamiltonian Mixtures of bosons and fermions in an optical lattice are described by a Hubbard type Hamiltonian consisting of the single species Hamiltonians (see Sec. 2.4 and Sec. 2.5) and an 38 4 Lattice Modulation Spectroscopy with Bose-Fermi Mixtures interspecies interaction term [58]: ĤBF = − JB X † b̂i b̂j + hi,ji − JF X † ĉi ĉj − X UX B B B n̂i (n̂i − 1) − B i n̂i 2 i i X X (4.1) i hi,ji + UBF F F i n̂i B n̂F i n̂i . i Here the indices B and F denote the species. The interaction between bosons and fermions is given by UBF = gBF Z |w1B (~r)|2 |w1F (~r)|2 d~r (4.2) with gBF = 2π~2 aBF . µ (4.3) mF µ = mmBB+m is the reduced mass of two interacting atoms. aBF is the s-wave scattering F length between the bosonic and the fermionic atoms. For a mixture of 40 K and 87 Rb the interaction is attractive with a scattering length of aBF = −205a0 [59]. Effective Potential due to Interactions The attractive interaction gives rise to mutual effective potentials, determined by the density distribution ρ of the other species [23]: VBeff (~r) =V (~r) + gBF ρF (~r) VFeff (~r) =V (~r) + gBF ρB (~r). (4.4) The interplay between the two potentials can lead to a phenomenon called self trapping [23]. The fermions mediate a trapping between bosons. For higher bosonic fillings the experienced potentials become deeper. The trapping is mutual hence a deeper potential is experienced by both species. In Fig. 4.1 the effect of an effective potential depending on the occupation number is shown. The wedding-cake structure of a bosonic Mott-insulator results in site dependent occupation numbers. This in turn results in site dependent, effective potentials for the atoms. 4.2 Lattice Modulation Spectroscopy on Bose-Fermi Mixtures The method of lattice modulation spectroscopy introduced in the chapter 3.1 is now employed to probe the fermionic band structure with an admixture of bosons. 4.2 Lattice Modulation Spectroscopy on Bose-Fermi Mixtures 39 Figure 4.1: Interactions lead to an effective potential. Depicted is an optical lattice with harmonic confinement. In red fermions are shown. Small blue circles represent bosons. The effective potential created by attractive interaction is shown in blue for different boson occupations. In grey the potential without bosons is shown. Experimental Procedure The principle experimental procedure has been outlined in chapter 3.1. To tune the ratio between the atom numbers, different loading times of the MOT at the beginning of the experimental sequence are employed. This results in atom numbers of up to 2 × 105 Rubidium atoms forming a BEC and up to 1 × 105 Potassium atoms at 0.1 TF . The mixture of spinpolarized fermions and bosons is prepared in an optical lattice. A lattice depth of 7.5 EK r is chosen for all experiments. The bosons experience a deep lattice (approximately 19 ERb r ) and thus form a Mott-insulating phase due to strong localization. The lattice is modulated and the quasimomentum distribution is mapped onto real momenta. Both species are detected consecutively. The momentum distribution as well as the amount of 40 K atoms is extracted from the TOF images. Absorption imaging of the 87 Rb atoms yield the number of bosonic atoms. Spectroscopy of Bose-Fermi Mixtures The fermionic band structure has been probed for different ratios of fermions and bosons. The resulting dispersion relations are shown in figure 4.2. The unperturbed band structure is shown in red. Fitting a band structure to the data yields a lattice depth of 7.6 EK r . Admixture of about twice as much bosons as fermions results in the curve shown in green. A slight deviation starting at low momenta is seen over the whole third band. A lattice depth of 8.1 EK r is deduced. Large amounts of bosons result in even higher deviations from the original band structure shown in blue. The lattice depth is evaluated to 9.2 EK r which is a 20% deviation from the pure fermionic system. Some very interesting features are observable. First while the green curve shows only a small deviation from the pure band structure adding a large amount of bosons alters the shape of the resulting band structure significantly (blue curve). Further comparing the distribution of the amount of excited atoms, it can be seen that the amount of transferred atoms is smaller and the distribution of excited atoms has been altered significantly. We explain this behavior by the an effective potential explained in Sec. 4.1. The data for the green curve with an ratio NRb /NK ≈ 2 leads to a mean occupation of smaller than 40 4 Lattice Modulation Spectroscopy with Bose-Fermi Mixtures 50 pure fermions NRb / NK = 2 Emod / h [kHz] 45 NRb / NK = 10 40 7.6 EKr 35 8.1 EKr 9.2 EKr 0 0.2 0.4 0.6 0.8 1 q [kBZ] 0% 10% excited fraction Figure 4.2: Fermionic dispersion relation of the third band with an admixture os bosons. A mixture of 40 K and 87 Rb has been prepared in an optical lattice. Modulation spectroscopy yields the dispersion relation of fermions. In red the pure fermionic energy-momentum relation is shown. Green shows an moderate amount of bosons mixed into the system. Blue depicts the case of an atom number ratio of about 10. two. The bosons will build a Mott-insulator thus starting to localize the fermions. This localization results in a slightly increased lattice depth. Increasing the atom number ratio to NRb /NK ≈ 10, bosons will build up an extended wedding-cake structure with a maximal occupation number of three bosons per lattice site. This leads to localization of the fermions resulting in a deeper lattice. Deviations at small momenta are attributed to the effect of the harmonic confinement as discussed in Sec. 3.3. For large atom number ratios, this effect is increased due to the wedding-cake structure of the bosons which create a inhomogenous overlap between the fermions and the bosons. The localization yields a reduced tunneling rate which is dependent on the occupation number. NRb /NK 0 2 10 V0,fit [EK r ] 7.6 8.1 9.2 J [EK r ] 0.034 0.0301 0.0237 Table 4.1: Fermionic tunneling energy for different admixtures of bosons. Higher admixtures of bosons lead to a deeper lattice for the fermions. From the lattice depth the tunneling energy is calculated. With the obtained lattice depth and equation (2.12) the tunneling energy J can be calculated. The result of this calculation is shown in Tab. 4.1. The tunneling energy is reduced by 30%. 4.3 Summary 41 4.3 Summary In this chapter, lattice modulation spectroscopy has been employed to probe the band structure of fermionic samples with an admixture of bosons. A significant increase in the lattice depth experienced by fermions is observed. This behavior can be explained in terms of an effective potential which localizes the atoms. This localization leads to a decreased, site-dependent tunneling rate. The decrease has been evaluated to 30% for an atom number ration of NRb /NK ≈ 10. 42 4 Lattice Modulation Spectroscopy with Bose-Fermi Mixtures 43 5 Absorption Imaging with a Diffraction Limited Objective Efficient and reliable detection of atomic densities is a central requirement in all quantum gas experiments. A number of methods have been developed and are in use. One of the most widely used technique is a combination of time-of-flight (TOF) expansion with absorption imaging. After conducting the experiments, the atomic cloud is released from all confining potentials and allowed to expand freely. A resonantly tuned laser beam illuminates the atoms. The atoms absorb the light resulting in a shadow. This light distribution is imaged onto a CCD camera. The imaging onto the CCD is carried out by a high performance lens system. It is necessary that this lens system images the resulting shadow with as much detail as possible (i.e. high resolution with as much light collected as possible) with very few errors introduced (imaging only limited by diffraction). In this diploma thesis, a new high performance lens system has been designed, build and tested. The design process is based on theoretical considerations concerning the optical requirements of absorption imaging. We will show the relevant optical errors introduced by lenses. With the help of computer aided optical engineering tools, the present detection setup at the “Bose-Fermi mixtures” experiment is evaluated. Using automatic optimization algorithms, a complex new design for the detection setup is presented. After mechanical realization, optical tests reveal the performance of the system. 5.1 Absorption Imaging Detection according to the absorption imaging scheme is carried out by illumination of the atomic cloud with a resonant laser beam and successive imaging of the resulting shadow onto a CCD camera. The Beer-Lambert law is used to calculate the atomic density from the obtained picture. It relates the intensity of a light beam propagating through a medium with the distance passed. Consider a light beam with intensity I through a medium of density ρ in the x-direction. The Beer-Lambert law is then [60] dI = −σeff (I) ρ I. dx (5.1) Here σeff (I) is the intensity dependent absorption coefficient σeff (I) = σ0 1 + I/Isat + (2δ/Γ)2 (5.2) 44 5 Absorption Imaging with a Diffraction Limited Objective 3 ~ω with σ0 = 3λ2 /2π and Isat = 12πc 2 Γ the saturation intensity. Detection is carried out on resonance, thus the detuning is δ = 0. This equation is integrated along the direction of the detection laser beam passing the atomic cloud. It becomes ln I0 I0 − I = σ0 ρ + I Isat (5.3) where I0 is the initial intensity and I is the intensity after passing the ensemble. The right side of the equation is the optical depth of the cloud and is proportional to the number of atoms. In the experiment, three pictures are taken. The first picture is the absorption image IA where the shadow of the atoms is imaged. It represents the intensity after passing the cloud. A short amount of time later (on the order of milliseconds), a second picture is taken called reference image IR . This image is taken without any atoms present. It is used as an intensity reference. The third image taken is a dark picture ID where the detection beam is blocked. To account for stray light this image is subtracted from the reference and the absorption image. Using this intensities, relation (5.3) reads ln IR − ID IR − IA + = σ0 ρ. IA − ID Isat (5.4) Taking these three images allows precise determination of the density distribution of the atomic cloud. The imaging of the three pictures is carried out by a lens system and a CCD camera. A schematic of the detection system in use is shown in Fig. 5.1. The detection laser coming from the left illuminates the atoms. The resulting light is collimated by a lens an afterwards focussed onto the camera. A central part of the following chapter is the setup of the lenses used to create the image on the camera. imaging system detection laser g atomic sample b CCD camera Figure 5.1: Schematic of absorption imaging. Shown is a schematic of the absorption detection system used at the Bose-Fermi mixtures project. A resonant laser is passing through the atomic sample. Light is absorbed by the atoms and the shadow is imaged onto a CCD camera. 5.2 Image Formation, Resolution and Abberrations This part focusses on the process of image formation. We will see how the maximum achievable resolution is determined and which errors occur due to deviations from paraxial optics. 5.2 Image Formation, Resolution and Abberrations 45 5.2.1 Resolution A lens forms an image of an object onto an image surface. The finite size of the lens leads to diffraction at the edges. Hence the aperture (i.e. diameter for which light can pass the optical assembly) is the fundamental limit for the achievable quality of an image. An optical system only limited by diffraction is called diffraction limited. The resolution (i.e. smallest observable structures) of such a system is determined by the diffraction pattern on the image surface. For a spherical aperture a point source creates an intensity distribution on the image surface called an Airy disk (see Fig. 5.2a). The intensity distribution of an Airy disk in one dimension is [61] I(θ) = I0 2J1 (kD/2 sin θ) . kD/2 sin θ (5.5) I0 is the total intensity, J1 (x) the first order Bessel function, k = 2π/λ and D is the diameter of the aperture. θ is the angle between the optical axis and the point of interest on the image surface. Two point sources near each other (i.e. two features on an object to be imaged) create two Airy disks. The overlap of the two Airy disks determines if the two point sources are distinguishable. According to Rayleigh’s resolution criterion two points are considered resolved when the central peak of one of the Airy disks coincides with the first diffracted intensity maximum of the other disk (see Fig. 5.2d). According to this criterion, the maximum possible resolution of a system with focal length f and numerical aperture NA is given by [62] dRayleigh = 1.22 λ . 2 NA (5.6) Here d is the minimal displacement of two objects that can be resolved. The numerical aperture determines the amount of light accepted by the optical system. The Sparrow criterion considers an object resolved as long as there is a minimum between the central peaks of the two corresponding Airy disks (see Fig. 5.2c). With this the maximal achievable resolution of a lens setup illuminated with incoherent light is [62] dSparrow = 0.95 λ . 2 NA (5.7) If coherent light is used for imaging, Rayleigh’s criterion breaks down because it only considers the distance between maxima and minima of the Airy disk and neglects interference effects. The Sparrow criterion for coherent light is dcoh = 1.46 λ . 2 NA (5.8) 46 5 Absorption Imaging with a Diffraction Limited Objective (a) (b) Airy disk of one point source two point sources - not resolved (c) (d) two point sources resolved according to Sparrow two point sources resolved according to Rayleigh Figure 5.2: Resolving limits for an optical system. a A point source creates an Airy disk intensity pattern. b two point sources near each other are not resolved. c Sparrow’s criterion assumes two point sources as resolved, when there is a minimum between the two resulting Airy disks. d Rayleigh’s resolution criterion requires that the first diffracted intensity maximum coincides with the central maximum of the other Airy disk. 5.2.2 Aberrations In paraxial optics only small angles between the light rays and the optical axis, and small distances of these rays to the optical axis are considered. The height h of an object is then connected to the height y 0 of the image with the following relation: y 0 = mh. (5.9) Here m is the magnification of the optical system. For large angles or large distances from the optical axis, errors are introduced into the image . These deviations are called aberrations and degrade the resolution and overall image quality significantly. The equation describing a light ray passing through an rotational symmetric optical system incorporating aberrations is written as a power series of the following form [63]: y 0 = mh + A1 p + B0 p3 + B1 hp2 + B2 h2 p + B3 h3 + 5 X Ci hi p5−i (5.10) i=0 + ... The distance of a light ray incident on the aperture of the optical system is p and is measured from the optical axis. The linear term in p with the coefficient A1 describes defocus and is 5.2 Image Formation, Resolution and Abberrations 47 called a first-order aberration. Terms with coefficients Bi are called third-order aberrations or Seidel aberrations. The Seidel aberrations are considered the most important errors [64] and their coefficients represent the following aberrations: B0 spherical aberration (SA3), B1 coma (CMA3), B2 astigmatism (AST3) and Petzval field curvature (PTZ3) and B3 distortion (DIS3). Higher-order aberrations are highly suppressed [64] and will only play a minor role in the following chapters. The following section describes the monochromatic Seidel aberrations briefly. The section follows [62] where more in-depth explanations can be found. We neglect chromatic aberrations because the detection is carried out at two very sharp defined wavelengths. Further the detuning between the wavelengths is in terms of geometric optics very small (13 nm) thus only minor chromatic effects are expected. Spherical Aberration (SA3) The spherical surface of a lens leads to different angles of incidence for light rays passing the lens at distinct distances from the optical axis. The light rays experience different angles of refraction after propagating through the optical element. See Fig. 5.3. A blurred image is the result. SA3 depends only on p and not on the actual object size thus is the most influential aberration. Figure 5.3: Spherical Aberration. Spherical aberration occurs due to different refraction of light beams that enter a lens further away from the optical axis in comparison with rays passing the lens near the center. The result is a blurred image. Coma (CMA3) When light rays penetrate a lens under an angle with respect to the optical axis, light rays hitting the lens further away from the optical axis are imaged onto a different point than rays passing through the center of the lens. A point source imaged by an optical system with coma seems to have a blurred tail similar to a comet. See Fig. 5.4. 48 5 Absorption Imaging with a Diffraction Limited Objective image Figure 5.4: Coma. Coma is the result of rays hitting a lens with an angle with respect to the optical axis. The result is an image blurred and appearing to have a tail like a comet. Astigmatism (AST3) A point source which is not on the optical axis of the system will experience different focal length for the rays passing the optical system parallel (sagittal rays) or perpendicular (tangential rays) to the line between the point source and the optical axis. The result are different focal points. An image appears defocused in one direction while focused in the other. tangential ray sagittal ray lens point source - off axis Figure 5.5: Astigmatism. A point source off-axis results in different focal points in the image plane. Petzval Field Curvature (PTZ3) Rays emitted from distinct parts of the object are focused onto a spherical image surface due to different angles of incidence. See Fig. 5.6. 5.3 Raytracing and Computer Aided Optical Engineering 49 image surface Figure 5.6: Petzval field curvature. Light rays from an extended object have different angles of incidence on the imaging system. The result is a curved image surface. Distortion (DST3) When an off-axis point source is imaged closer to or further away from the optical axis, then the image is said to be distorted. Straight lines will be projected as curved ones. See Fig. 5.7. The reason for this aberration is variations of the magnification across the aperture of the optical system. object pincushion distortion barrel distortion Figure 5.7: Distortion. Varying magnification across the aperture of the optical system results in distortion. If the magnification around the center is larger than further away from the optical axis the distortion is called barrel distortion. Smaller magnification around the center yields a pincushion like distortion. Summary All the above presented aberrations degrade the performance of the detection setup. Aberrations are highly non-linear phenomena and can not be easily corrected. The help of dedicated simulation tools is necessary. This will be the topic of the next paragraph. 5.3 Raytracing and Computer Aided Optical Engineering The foregoing section showed that imaging systems will introduce aberrations (i.e. errors) into the formed image. These aberrations degrade the performance of any optical system. 50 5 Absorption Imaging with a Diffraction Limited Objective Creating a new optical system has the goal to deliver the needed performance in the parameters needed for the later application. This could be exact color representation, wide area of focus or high resolution. The ultimate goal is of course to create a design which is diffraction limited in all the areas of application. This is not always possible. For example lenses for digital-single-lens-reflex (DSLR) cameras are often not diffraction limited over the whole range1 . Still they deliver high performance in their area of application. In general such a high performance system (DSLR lens, high resolution optics, etc.) can not be build out of one or two optical elements [65] and even for such small amounts of lenses the number of adjustable parameters is huge (materials, thicknesses, curvatures, distances etc.). Thus designing of an well corrected optical system is nowadays done using specialized simulation and optimization programs. Popular applications are Code V by Synopsys, ZEMAX by Radiant ZEMAX LLC. or OSLO by Lambda Research. All these programs belong to the group of raytracers and incorporate functions for evaluation and design of optical setups. In the framework of this thesis OSLO light has been used. In this chapter the fundamentals of raytracing are briefly presented. Tools for evaluation of optical designs are shown and lastly an introduction to automatic optimization of optical designs is given. 5.3.1 Raytracing Raytracing programs simulate the propagation of light rays through optical media and are used in a multitude of applications. For example they are used to create high resolution computer graphics. Generally optical raytracers are based on geometrical optics and use Snell’s law for tracing [66]. A typical raytracing algorithm traces a given ray of light until hitting the first surface. The angle of incidence is calculated and using Snell’s law the angle of refraction is evaluated. The refracted ray is then traced until hitting the next surface. This procedure is repeated until the ray hits the image surface. By tracing rays throughout the whole aperture and under different angles, the properties of an optical design can be analyzed. 5.3.2 Evaluation of Optical Designs Raytracing allows the evaluation of optical systems with different techniques. First of all the aberration coefficients introduced in equation (5.10) can be calculated. With this coefficients it is possible to specifically optimize a system towards the needed performance. While this is very powerful and necessary for optimization, other tools are better suited to evaluate the overall performance (resolution, diffraction limitation, ...). Especially tools which represent the overall optical performance graphically allow a fast and reliable evaluation. In the following three tools are presented: the spot diagram gives a very intuitive picture of the image formed by a lens setup. The wavefront analysis allows quantification of the amount of errors introduced in comparison with an diffraction limited system. Finally the modulation 1 compare for example with http://www.pbase.com/samirkharusi/canon_mtf_curves 5.3 Raytracing and Computer Aided Optical Engineering 51 transfer function (MTF) incorporates the effect of diffraction and shows directly how well small structures are imaged by the system. Spot Diagram and Wavefront Analysis The following two sections describe the spot diagram and the wavefront analysis. Spot Diagram To generate a spot diagram a large number of rays is traced through an optical system. After propagating through the system, the rays hit the image surface. Each point where a ray hits the image surface is marked with a spot. This yields the spot diagram. A typical spot diagram for a well corrected lens system is shown in Fig. 5.8. The overall shape and density gives a very nice overview of the optical performance. Figure 5.8: Spot Diagram. Tracing a large number of rays through an optical system yields a spatial distribution of rays hitting the image surface. This spot diagram was calculated using OSLO for a well corrected lens system. The point source has been placed off-axis resulting in a deformed spot diagram. The black circle represents the Airy disk. Calculating the size of the Airy-disk gives an impression how well the system is diffraction limited. All spots lying within the calculated Airy-disk is a indication for an diffraction limited lens setup. However, the spot diagram does not show the real intensity distribution within the disk. Wavefront Analysis To decide whether an lens system is diffraction limited and how much the image is influenced by the aberrations is nicely seen in a wavefront analysis. A perfect imaging system images a point source onto a point on the image surface. Thus light waves leaving the imaging system form a wave with a spherical wavefront centered around the point on the image surface. Aberrations will deform this wavefront. The amount of deformation is measured as a peak-valley optical path difference (P-V OPD) as shown in Fig. 5.9a. It can be shown that Rayleigh’s diffraction criterion is fulfilled as long as the P-V OPD between the edge of the wavefront and the inner part is smaller than a quarter of a wavelength [62]. For a heavily distorted wavefront, this criterion can be reformulated: to fulfill Rayleigh’s criterion (i.e. being diffraction limited) the root mean square of the optical path difference (RMS OPD) has to be smaller than 7% of λ. In Fig. 5.9b a wavefront computed by OSLO is shown. A well corrected system with a point source off-axis has been analyzed. The RMS OPD has been evaluated to 0.06λ and the P-V OPD is 0.31λ. Hence the system under study is not diffraction limited but only by a small margin. 52 5 Absorption Imaging with a Diffraction Limited Objective (a) (b) OPD 0.23λ y distorted wavefront -0.14λ perfect wavefront x Figure 5.9: Wavefront analysis. The wavefront analysis shows the deviation of the wavefront from a perfect one. a shows the optical path difference (OPD) between a perfect wavefront and a distorted. In b shown is the wavefront for an well corrected optical system with a point source 1 mm off-axis (y-direction). Modulation Transfer Function The wavefront analysis presented in Sec. 5.3.2 is based on raytracing data. Hence the effect of diffraction is not incorporated. This shortcoming is circumvented by measuring the modulation transfer function (MTF). The MTF measures the contrast of an image depending on the reciprocal size of the object in front of the lens system. The reciprocal size is measured in lines/mm. The object is considered to have a sinusoidal dark-bright contrast. With that a line is one dark-bright cycle. The MTF is calculated by performing a Fourier transform on the intensity distribution of the image. The modulation transfer function (MTF) incorporates diffraction as well as refraction. The MTF is written as MTF(f ) = output(f ) input(f ) (5.11) where input is the object in front of the optical system and output is the resulting image. The input function is considered to be sinusoidal with ω = 2πf . The diffraction limited (i.e. optimal) MTF for a lens setup with spherical aperture is [67] λf 2 MTF(f ) = arccos π 2 NA − λf 2 NA s 1− λf 2 NA 2 . (5.12) The diffraction limited MTF reaches zero at 2 NA (5.13) . λ This is the fundamental limit for the resolving power of an optical system and is called cutoff frequency. The cutoff frequency resembles Sparrow’s resolution criterion (5.7). f0 = A MTF for a diffraction limited lens with λ = 780nm, NA = 0.1481 is shown in Fig. 5.10. For this setup the cutoff frequency is f0 ≈ 380 lines/mm Aberrations will reduce and deform the MTF. The universality of the MTF makes it one of the standard tools to evaluate lens designs. 5.3 Raytracing and Computer Aided Optical Engineering 53 modulation transfer 1 cutoff 0.8 0.6 0.4 0.2 0 0 100 200 300 spatial frequency [lines/mm] 400 Figure 5.10: MTF of an diffraction limited lens. Shown is the MTF for an diffraction limited lens with λ = 780 nm and NA = 0.1481. The green bar marks the cutoff frequency at f0 ≈ 380 lines/mm. 5.3.3 Optimization The last sections showed the tools used to evaluate the following optical designs. As outlined in the introduction to Sec. 5.3, the goal for a new lens design is to deliver the desired performance in the areas of application. In most cases optical systems consist of more than one or two elements, such that the number of variable parameters is large (materials, curvatures, distances, thicknesses, ...). Hence optimization is in most cases carried out by automatic optimization algorithms. The bases for an automatic optimization is an error function (sometimes also: merit function). The method of optimization used throughout this chapter is a damped least square (DLS) method with an error function φ of the following form [66]: φ(~c) = m X wi fi2 (~c). (5.14) i=1 Here the vector ~c represents all parameters of the system. These are thicknesses of optical elements, curvatures of surfaces, materials of lenses and index of refractions of space between lenses. Also all parameters allowed to vary are stored in ~c. fi is called an operand and returns a specific deviation from a target value. For example an operand could ensure that the focal length of the system is a specific value. m is the number of operands. The coefficients wi are weights to define the relative importance of the operands. The DLS optimization tries to minimize φ. The primary task for a lens designer is to define a proper error function, which incorporates the desired properties of a lens design. Unfortunately because aberrations can only be minimized and not completely removed, the error function has in general local minima. So during the optimization process it is necessary to manually check for other solutions by varying the starting conditions and impose restrictions on the variable parameters ~c. These restrictions are also necessary to make sure that a design is practical. For example a 2 m thick lens is not often very reasonable. The operands and weights used to develop the new detection objective is presented in Sec. 5.5.2. 54 5 Absorption Imaging with a Diffraction Limited Objective 5.4 Old Detection Setup The setup used so far at the "Bose-Fermi mixture" project in Hamburg consists of two achromatic lenses with focal length f1 = 120 mm and f2 = 250 mm as shown in Fig. 5.11. The first lens collimates the incoming light while the second lens focusses the image onto the camera. The atomic sample is prepared in a glass cell while detection is carried out outside of the apparatus. Thus the wall of the glass cell has to be incorporated into the evaluation process. This lens pair magnifies by approximately a factor of f2 /f1 = 250 mm/120 mm ≈ achromatic lens f=120mm atomic sample 11mm to inner cell wall 120mm glass cell - 2mm thick magnificaion ≅2.1 250mm achromatic lens f=250mm CCD camera - PCO.pixelfly Figure 5.11: Old detection setup. The so far used detection setup is based on two lenses which first collimate the light and than focus it onto the camera. 2.1. The used camera is a PCO pixelfly with a pixel size of 6.45 µm × 6.45 µm. The object side numerical aperture is NA ≈ 0.1638. The system presented above has been simulated and analyzed using OSLO. To compare with the later developed new lens design the aberration coefficients (see Sec. 5.2.2) are presented in Tab. 5.1. aberration SA3 CMA3 AST3 PTZ3 DIS3 computed value −0.3086 0.1627 −0.0879 −0.0015 0.0493 Table 5.1: Aberration coefficients for the two lens detection setup. Shown are the aberration coefficients for the two lens detection setup computed by OSLO for a point source 1mm off axis. A spot diagram for an on axis as well as an off axis point source is shown in Fig. 5.12. The resulting diagram (Fig. 5.12a) shows a large amount of rays hitting outside of the Airy disk. The radius of the Airy disk has been calculated to approximately 6 µm. The summed root mean square radius of the spot diagram has been evaluated to 15 µm. The off axis spot diagram (Fig. 5.12b) shows large amounts of coma and a much larger overall spot. The Airy disk is barely visible as a tiny spot in the upper part of the image. 5.4 Old Detection Setup 55 (a) on optical axis (b) 1mm off optical axis Figure 5.12: Spot diagrams for the two lens detection setup. a shows the spot diagram for a point source on the optical axis. The black circle represents the Airy disk. b shows the spot diagram for a point source 1mm off axis. The Airy disk is barely observable in the upper part. The resulting wavefront is shown in Fig. 5.13. The OPDs for the on axis source (Fig. 5.13a) are: P-V OPD = 1.232λ and RMS OPD = 0.298λ. For the off axis source (Fig. 5.13b) the OPDs are: P-V OPD = 2.355λ and RMS OPD = 0.475λ. Thus the system is even for an on axis point source subject to large amounts of aberrations. (a) on optical axis (b) 1mm off optical axis 1λ y 1λ y -1.9λ x -1.9λ x Figure 5.13: Wavefront analysis of the two lens detection setup. a shows the wavefront for a point source on the optical axis. b shows the same point source moved 1mm off axis. This is also consistent with the resulting MTF shown in Fig. 5.14. Because of geometrical restrictions of the experimental setup (magnetic coils reduce the numerical aperture in one direction) two curves are shown which represent two orthogonal axes (parallel and sagittal axis). Both curves show huge deviations from an diffraction limited system. 5.4.1 Summary In this chapter the old experimental setup used to perform absorption imaging has been simulated. It was shown that the two lens setup introduces large errors into the image. 56 5 Absorption Imaging with a Diffraction Limited Objective modulation transfer 1 diff. limited parallel sagittal 0.8 0.6 0.4 0.2 0 0 50 100 150 200 spatial frequency [lines/mm] Figure 5.14: MTF of the two lens setup. Shown is the MTF for the two lens detection setup. Because of geometrical restrictions (the magnetic coils obscure a small part reducing the numerical aperture in one direction) the sagittal and parallel directions have slightly different MTFs. The diffraction limit is shown in green. Even an point source placed on axis (so only SA3 influences the image) is subject to large aberrations such that diffraction limited imaging is not possible. Thus the maximal possible resolution is not achieved and detail is lost. 5.5 Design of a New Diffraction Limited Detection System In the foregoing chapter we have shown, that the old detection setup is subject to large amounts of aberrations. The resolving capabilities are thus very limited. Based on these considerations we decided to design and build a new optical setup which is subject to very small aberrations. 5.5.1 Requirements for the New Design The requirements for the new optical system are derived from typical experimental conditions. The atoms are, in most of the cases, not more than 3 mm displaced from the optical axis. Hence the off axis performance is not the determining factor but of course should still be good. Further we want to use the available NA. The available NA is limited by geometrical restrictions to approximately NA ≈ 0.17. This numerical aperture allows a absolute maximum resolution of 2.3 µm. The optical lattice used in the experiment has a lattice site distance of 515 nm. Thus it will not be possible to achieve single site resolution. Still the resolution should be as high as possible to perform high precision measurements after TOF expansion. This in turn means that the design should be diffraction limited for on axis sources as well as for sources placed slightly off-axis (up to 1 mm). For experimental convenience the magnification of the system should be easily changeable. 5.5 Design of a New Diffraction Limited Detection System 57 5.5.2 Underlying Design and Optimization Process The basis for the finally obtained lens design has been taken from [64] and [65]. The design presented in these references is based on a setup consisting of four lenses which are used to collimate the light from a source. These four lenses have curvatures of their surfaces chosen such that aberrations introduced by one lens is compensated by another. The focussing is then performed by a final optical element which in the following is called focussing lens. The ratio of the focal lengths of the four lens objective and the focussing lens determines the magnification. We decided to use an achromatic lens as the focussing lens to minimize focal shifts for the two different detection wavelengths. In a first step the principle design shown in [64, 65] has been adapted from one inch diameter lenses to 50 mm lenses simply by approximately doubling the lens diameters. This doubling essentially destroys the good optical performance of the original design. Nevertheless the basic informations like curvatures (concave, convex) and distance ratios are useful for the automatic optimization The second step was to define an error function for optimization (see Sec. 5.2.2). The new setup was expected to be a little bit bigger than the old two lens setup, so we decided to increase the distance between the atomic sample and the first lens by 1.5 cm. This results in a desired front focal length (FFL) of 135 mm. The FFL is the distance between the object (atoms) and the first surface of the imaging system. The effective focal length (EFL) is the optical focal length. It is not directly related to the FFL. The EFL directly influences the achievable magnification, so an effective focal length of 135 mm is the target focal length. To reach this target an operand to ensure an corresponding EFL is added to the error function. The weighting factor for the EFL operand has been chosen low. The low weighting factor allows the optimization algorithm to slightly vary the EFL when through this variation an aberration can be lowered. Because the atoms are not often displaced very much from the optical axis spherical aberration has been determined to be the most influential aberration. Therefore spherical aberration of third-order (SA3), fifth-order (SA5) and seventh-order (SA7) is incorporated into the error function. The other Seidel aberrations should be minimized, but with lower priority (weighting factors). The resulting error function is shown in Tab. 5.2. The so defined error function is minimized by allowing the variation of the curvatures of the surfaces, the distances between the optical elements and the thicknesses of the elements. For availability reasons BK7 has been chosen as the material of the lenses. The third step it to replace the first virtual lenses (as seen from the atoms) with a commercially available lens by CVI Melles Griot2 . Again the system is optimized and the last lens was replaced. This process is iterated for the remaining virtual lenses until all are replaced by real counterparts. 2 see http://www.cvimellesgriot.com/ 58 5 Absorption Imaging with a Diffraction Limited Objective weight wi 0.1 1.0 0.3 0.3 0.3 0.3 0.2 0.2 operation ensure 135 mm minimize minimize minimize minimize minimize minimize minimize operand fi EFL SA3 CMA3 AST3 PTZ3 DIS3 SA5 SA7 Table 5.2: Error function for new lens setup. Shown are the operands and weights used for all automatic optimizations of the new lens design. The weighting factor is given for each operand. The operand column shows which physical property of the system is manipulated and the operation column shows in which way this is done. For example all aberrations should be minimized but the focal length (EFL) should remain at 135 mm. Finally, a last optimization iteration is performed, allowing the variation of the distances between the real lenses. The in this way obtained design is shown in the next section. 5.5.3 New Design This optimization resulted in a final design shown in Fig. 5.15. The CVI Melles Griot part not to scale glass cell - 2mm thick atomic sample 11mm to inner cell wall LPX277 123.9mm LPX577 31.1mm LDX373 40mm lens diameters: 49mm LPK161 21.1mm interchangeable achromatic lens Figure 5.15: Diffraction limited detection setup. Shown here is the new design for a detection setup at the Bose-Fermi mixtures experiment. The image is not to scale. Used lenses including their CVI Melles Griot part numbers are shown as well as the distances between them. numbers are depicted as well as the distances between the lenses. In Tab. 5.3 the precise data for all lenses is shown. The materials, thicknesses and radii of curvatures (for the first and second surface) are shown as well as the focal length of the corresponding element. All lenses have a diameter of 49 mm ± 0/50 µm. The lens design has a numerical aperture of 0.1481 and an effective focal length of 135 mm. The FFL is 136.9 mm. For focussing an achromatic lens is chosen as explained above. To change the magnification of the imaging system, the last lens is interchangeable. In Tab. 5.4 the different achievable magnifications and the corresponding lenses are shown. 5.5 Design of a New Diffraction Limited Detection System part glass cell LPX277 LPX577 LDX373 LPK161 material silica glass BK7 BK7 BK7 BK7 d [mm] 2 6 4 4.9 3 r1 [mm] 336.37 -129.68 59 r2 [mm] 103.74 194.51 336.37 - f [mm] 200 375 325 -250 Table 5.3: Lenses used in the new detection setup. The table shows the first four lenses used to collimate the light with their relevant properties such as the thickness d, curvatures r1 and r2 , material and focal length f. 5.5.4 Evaluation of the New Design The newly developed design is now analyzed using the tools presented above. The analyses shown here has been performed for the twofold magnification. The results for the other magnifications are summarized at the end of this section. First all the aberration coefficients (see Sec. 5.2.2) are presented in Tab. 5.5 together with the difference to the aberration coefficients of the old system (shown in Tab. 5.1). A vary large decrease (90 percent and more) for SA3, CMA3, AST3 and DIS3 is observed. PTZ3 is reduced by nearly 30%. (a) on optical axis (b) 1mm off optical axis Figure 5.16: Spot diagrams for the new detection setup. a shows the spot diagram for a point source on the optical axis. The black circle represents the Airy disk. b shows the spot diagram for a point source 1mm off axis. magnification 1.1 2.2 3.0 4.4 5.1 part LAO551 LAO656 LAO690 LAO807 LAO809 f [mm] 150 300 400 600 700 Table 5.4: Lenses used for focussing the collimated light onto the camera. Shown are the achromatic lenses used to focus the light onto the CCD camera. Depending on the desired magnification a different lens can be mounted. 60 5 Absorption Imaging with a Diffraction Limited Objective aberration SA3 CMA3 AST3 PTZ3 DIS3 new system −0.0190 +0.0038 −0.0017 −0.0011 −0.0007 old system −0.3086 +0.1627 −0.0879 −0.0015 +0.0493 difference −94% −97% −98% −27% −99% Table 5.5: Aberration coefficients for the new lens detection setup. Shown are the aberration coefficients for the new lens design computed by OSLO for a point source 1mm off axis. Further the coefficients for the old system are shown together with the difference in percent. The resulting spot diagram for a point source on axis as well as off axis is shown in Fig. 5.16. The diagram shows a much more localized spot distribution in comparison to Fig. 5.12. Also the deformation due to off axis effects is much smaller. Note that the spots for an off axis point source are now very localized and most of the rays hit inside the Airy disk. This is a huge improvement over the old system where the Airy disk is barely recognizable (see Fig. 5.12b). The wavefront analysis confirms this finding. It is shown in Fig. 5.17. The OPDs are: P-V OPD = 0.1327λ and RMS OPD = 0.03595λ. For the off axis source (Fig. 5.13b) the OPDs are: P-V OPD = 0.3122λ and RMS OPD = 0.05762λ. This shows that the new design is very well diffraction limited on axis and thus is expected to reach the theoretically possible resolution. Moving the point source 1 mm off axis delivers a much better result than the old setup. (a) on optical axis (b) 1mm off optical axis 0.25λ y 0.25λ y -1.33λ x -1.33λ x Figure 5.17: Wavefront analysis of the new detection setup. a shows the wavefront for a point source on the optical axis. b shows the same point source moved 1mm off axis. A MTF for the new setup is shown in Fig. 5.18 and supports the findings from the wavefront analysis and the spot diagram. The resulting modulation is over a wide range near perfect and shows again the huge improvements over the old design. These evaluation steps have been performed for all magnifications. No substantial differences have been observed and all magnifications deliver a similar performance near to a perfect result. 5.5 Design of a New Diffraction Limited Detection System 61 1 diffraction limit modulation transfer 0.8 MTF new design 0.6 0.4 0.2 0 0 50 100 150 200 spatial frequency [lines/mm] Figure 5.18: MTF of the two lens setup. Shown is the MTF for the new optics design. The diffraction limit is shown in green whereas red shows the designed optical system. Summary In this part a new design for the optical components in the absorption detection setup is presented. Starting with a basic design and a carefully chosen error function, the system has been iteratively adapted and improved. Consisting of five optical elements, the new design shows much better optical performance and is due to interchangeable magnifications experimentally much more convenient. 5.5.5 Mechanical Realization The physical setup of the new detection setup was engineered using computer aided design software. A technical drawing is shown in Fig. 5.19. The basis for the system is a lens tube with a diameter of 49.05 mm. This tube holds the first four lenses (see 5.5.3). To achieve the desired distance between the lenses is maintained by precisely engineered distance rings with a diameter of 49.00 mm. The lenses are individually cut down to a diameter of 49.00 mm ± 0/50 µm. The diameter of the tube, lenses and distance rings needs to be precisely manufactured to achieve accurate axial alignment. The achromatic lenses used for collimating are housed separately in a small tube system. This focussing tube is screwed onto the large tube system and is aligned using precision alignment pins. This mounting allows simple changing of the magnification (i.e. changing the focussing lens). The whole optical assembly can be moved on a xy-translation stage. All tubes are made of aluminum to reduce overall weight. The distance rings are made out of stainless steel to improve stability. Brass is used to manufacture the middle part of the xy-translation stage. This adds additional weight but allows frictionless moving of the stage. 62 5 Absorption Imaging with a Diffraction Limited Objective interchangeable lens housing xy-translation stage Figure 5.19: Mechanical setup of the new lens design. The mechanical realization of the new lens setup is shown. It consists of a tube, housing the first four lenses used for collimation. The focussing lens is suported separately and is mounted with screws and alignment pins. The whole system is mounted on a xy-translation stage. 5.5.6 Optical Tests To characterize the optical system, a USAF 1951 resolution test chart (shown in Fig. 5.20) and a grid distortion test chart is used (see Fig. 5.26). The resolution test chart consists of dark-bright line pairs of different sizes. The sizes are specified and are identified by the surrounding numbers. Illumination of the chart is carried out by a laser tuned to 780 nm. The chart is imaged onto a CCD camera using the new lens system. The distance between the chart and the first surface of the lens system is chosen according to the expected distance between the lens and the atomic sample later in the experiment. A PCO pixelfly3 is used. This camera is of the same type as the one used later in the experiment. This allows us to directly transfer the experiences made outside of the apparatus to the situation where the lens is incorporated into the experimental setup. As a benchmark, the maximum achievable resolution for incoherent light is calculated using a NA of 0.1481: dSparrow ≈ 2.7 µm. (5.15) As a first step the resolution test chart is imaged with a perfectly aligned setup. The resulting pictures for the twofold and fivefold magnification are shown in Fig. 5.21. Because 3 pco.pixelfly, see http://www.pco.de/ 5.5 Design of a New Diffraction Limited Detection System 63 Figure 5.20: USAF resolution test chart. The test chart consists of dark-bright line pairs of different sizes. the theoretically achievable resolution is 2.7 µm the twofold magnification makes it necessary to distinguish structures off about 6 µm. This is approximately the size of one pixel of the camera. This means that we are pixel limited and are not able to observe the resolution limit. Obtained pictures are shown in Fig. 5.21a. The situation changes drastically when (a) 2x magnification (b) 5x magnification Figure 5.21: Maximum achievable resolution. Shown are the images obtained for the 2x and 5x magnification. The images for the 2x magnification is pixel limited. The 5x magnification shows a resolution limit of 2.8 µm marked in red. using the fivefold magnification which is shown in Fig. 5.21b. The last observable structure is the fourth line pair in group seven. This corresponds to a resolution of 2.8 µm which is indistinguishable to the expected limit. Thus the goal of diffraction limited imaging has been reached. Further characterization is now performed by moving the test chart from the optimal position. One key parameter is the depth of field (DOF) determining the distance which the object can be moved without loosing resolution. To determine the DOF the distance between the test chart and the first lens surface is varied and the resulting resolution is recorded. The results are shown in Fig. 5.22. The graphs show that the region of maximum 64 5 Absorption Imaging with a Diffraction Limited Objective (a) 2x magnification (b) 5x magnification 17.5 12.4 11 13.9 resolution [µm] resolution [µm] 15.6 12.4 11 9.8 8.8 7 5.5 4.4 −0.5 −0.3 −0.1 0 0.1 0.3 9.8 8.8 7.8 7 6.2 5.5 4.4 3.5 2.5 −0.3 −0.2 −0.1 0.5 0 0.1 0.2 0.3 test chart position [mm] test chart position [mm] Figure 5.22: Depth of field measurement. By varying the position of the test chart the depth of field (DOF) is measured. resolution is very small (below 0.1 mm) which on the one hand needs good alignment in the experiment but on the other hand reduces the impact of light coming from out of focus objects. Also it can be seen that the data obtained for the 5x magnification seem much more reliable because the curve shows a much smoother overall slope. We attribute this to the pixel limitation for the 2x magnification which makes reliable resolution measurements complicated. To determine how well the system has been assembled the DOF measurement has been adapted. The test chart position is again varied but this time the amount of displacement of the camera to refocus is recorded. The result is shown in Fig. 5.23. (a) 2x magnification (b) 5x magnification 150 measured fit 10 camera position [mm] camera position [mm] 15 5 0 −5 −10 measured fit 50 0 −50 −100 −150 −15 −20 −3 100 −2 −1 0 1 2 test chart position [mm] 3 −200 −5 −4 −3 −2 −1 0 1 2 3 4 5 test chart position [mm] Figure 5.23: Focus shift depending on test chart postition. Varying the position of the test chart and successive focussing allows the determination of the magnification by fitting a theory curve to the data points. Using 1/f = 1/b + 1/g and m = b/g a relation between the object (i.e. test chart) shift ∆g 5.5 Design of a New Diffraction Limited Detection System 65 and the focus shift ∆b is deduced: ∆b = m2 ∆g . 1+ m f ∆g (5.16) Here m is the magnification, f the focal length. f plays only a minor role for the fit thus it has been set to values calculated by OSLO. The fit resulted in: m2x,OSLO = 2.20 m2x = 2.19 (5.17) m5x,OSLO = 5.07 m5x = 4.98 Comparing with the values calculated by OSLO (without the glass cell because the tests have been performed outside of the experiment) yields deviations which are smaller than 2%. The remaining deviation is attributed to errors in finding the correct focus position and minor misalignments of the whole setup. Overall the very good agreement between the simulated and the measured parameters strongly suggests that the system is properly build and aligned. The field of view (FOV) is measured by varying the position of the test chart orthogonal to the optical axis and recording the resulting resolution. The result is shown in Fig. 5.24. Seen (b) 5x magnification 12.4 11 11 9.8 resolution [µm] resolution [µm] (a) 2x magnification 9.8 8.8 7.8 7 6.2 4.9 3.9 2.8 −3 8.8 7.8 7 6.2 5.5 4.9 3.9 2.8 −2 −1 0 1 2 test chart off axis position [mm] 3 −3 −2 −1 0 1 2 3 test chart off axis position [mm] Figure 5.24: Field of view measurement. Variation of the transversal position of the test chart allows the determination of the field of view. is a dependence of the maximum observable resolution on the off axis distance. Comparing with Sec. 5.5.4 we expect to remain diffraction limited close to 1 mm off axis shift. This is not observed but the measurement is influenced by systematic errors. The finite size of the resolution test chart is a major problem. Moving the chart into one direction, away from the optimal position, it can happen that the originally observable structure is not resolved anymore, but the next bigger structure is even further away and therefore also not observable. During the measurement the focus has not been realigned so that another possible problem could be some amounts of Petzval field curvature. Nevertheless, the resolution 66 5 Absorption Imaging with a Diffraction Limited Objective stays below 11 µm for a range of 6 mm which is the region in which the atoms typically are allowed to expand. In realistic experimental conditions, the region where the position of the atoms is varied stays below 1.5 mm which in turn means a resolution of below 4 µm. At a mixture experiment detection has to be carried out with two wavelengths: one for each species. We use 767 nm for Potassium and 780 nm for Rubidium. So we are interested in the behavior of the system when changing the wavelength of the detection laser. Figure 5.25 shows the shift in focus for four different wavelengths. Seen is a shift of 2 mm between 6 focus shift [mm] 4 2 0 −2 −4 750 767 780 wavelength [nm] 795 Figure 5.25: Shift of focus for different wavelengths with 5x magnification. Changing the wavelength of the detection light results in a focus shift. Potassium and Rubidium wavelengths. This shift is experimentally not limiting because often only one species is detected using the new lens setup while the other species is detected on another detection axis. To probe the behavior of the system regarding Petzval field curvature and distortion the grid distortion test chart is used. To image as many of the grid lines as possible (widest view angle) the 1x magnification is used here. The obtained image is shown in Fig. 5.26. Comparing the straight red lines with the lines imaged by the detection system, no Petzval field curvature or distortion is observed. The tests shown here have been performed on all available magnifications. All showed the predicted behavior and performed similarly. Summary The optical tests performed in this chapter are based on a standardized resolution test chart. The test results strongly suggest that the design worked the way it has been developed. By reaching the theoretically possible resolution it is shown that diffraction is the limiting factor and aberrations play only a minor role. 5.6 Summary 67 Figure 5.26: Grid chart imaged with 1x magnification. A periodic grid is imaged with the detection system. No Petzval field curvature or distortion is observed. In red, straight and orthogonal lines are shown. 5.5.7 The New Lens Build into the Experiment The new lens system has been successfully integrated into the experiment. As a alignment test, the magnification of the 1x and the 2x magnification is measured. This is done by preparing a BEC in an optical dipole trap and performing a TOF measurement. By varying the time between trap switch off and detection a free fall parabola is observed. The resulting data is fitted and thereby the magnification can be deduced. In Tab. 5.18 the theoretically obtained values, the values measured with the test chart and the magnifications obtained in the experimental setup are shown. m1x,OSLO = 1.11 m1x,measured = 1.10 m1x,freefall = 1.12 m2x,OSLO = 2.22 m2x,measured = 2.19 (5.18) m2x,freefall = 2.28 The results agree well with the expected values. This allows two conclusions: First, the alignment of the lenses within the optical assembly is accurate. Second, the alignment in the experimental setup is precise. 5.6 Summary This chapter showed the design and realization of a diffraction limited lens system to efficiently perform absorption detection. The system has been optimized using a dedicated error function. Simulation and experimental testing showed the predicted behavior. The objective has been built into the experiment and showed again good agreement between theoretical predictions and real behavior. Overall it was possible to build an high performance, diffraction limited lens system with a maximal resolution of 2.8 µm. In the near future a new camera with a high quantum efficiency (95%) will be build into the experiment. The 68 5 Absorption Imaging with a Diffraction Limited Objective new lens system together with the high performance camera will improve the quality of the detection especially for dilute fermionic clouds. 69 6 Conclusion and Outlook In the context of this thesis a novel momentum resolved spectroscopy method has been developed and employed. Based on modulating the lattice amplitude and successive band mapping, the band structure of fermions in optical lattices could be measured with high precision. A substantial effect of the underlying harmonic confinement has been observed, significantly altering the spectra. The high precision of the method allowed the direct observation of a shift in the potential experienced by fermionic atoms with an admixture of bosons. The shift is explained in terms of an effective potential induced by attractive interactions, leading to an occupation dependent decrease of the fermionic tunneling rate. This measurements could be a cornerstone to understand the Superfluid to Mott-insulator transition shift observed in Bose-Fermi mixtures. While developing the spectroscopy method some effects could be observed which will be studied in the near future. For example after creating a particle-hole excitation the dynamics shown below has been observed. Further the lifetime of the created holes is much smaller than the lifetime of the excitations. This features are attributed to the harmonic trapping potential but have not been fully understood. waiting time [ms] 20 0 momentum In the future the method of optical lattice modulation spectroscopy could be employed to study a multitude of systems. The use of Feshbach resonances between Rubidium and 70 6 Conclusion and Outlook Potassium could further develop the understanding of interaction-induced effective potentials. By modulation with different frequencies multiple excitations can be created. The dynamics of this excitations as well as collisions between such excitations promise interesting observations. Using mixtures of fermions in different hyperfine states allows sensitive probing of the physics of interacting fermions. Again interatomic Feshbach resonances can be used to change the interaction allowing experiments over a wide range of parameters. For example the interaction allows the creation of bound state of excited atoms and holes. This would allow the study of excitons. These opportunities for research on fundamental physics show the versatile nature of BoseFermi mixture experiments. The flexibility of Bose-Fermi mixtures together with the high degree of control in optical lattices promise further highly interesting studies in the near future. List of Figures 71 List of Figures 1.1 Fermionic band structure for a mixture of bosons and fermions . . . . . . . 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 Experimental Setup . . . . . . . . . . . . . . . . . . . . Different Lattice Geometries . . . . . . . . . . . . . . . . First Brillouin Zone of a Cubic Lattice . . . . . . . . . . Band Structure for Rubidium and Potassium . . . . . . Bose Hubbard Model . . . . . . . . . . . . . . . . . . . . Superfluid Phase . . . . . . . . . . . . . . . . . . . . . . Mott-Insulator Phase . . . . . . . . . . . . . . . . . . . . Phase Transition between Superfluid and Mott-Insulator Phase diagram and wedding cake structure for bosons in Phases of the Non-Interacting Fermi-Gas . . . . . . . . . Phase Transition from Metallic to Band Insulating . . . Band Mapping Technique . . . . . . . . . . . . . . . . . Band Mapping Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . lattice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 7 8 9 11 11 12 12 13 14 15 16 17 3.1 3.2 3.3 3.4 3.5 3.6 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 Effect of modulating the lattice amplitude . . . . . . . . . . . . . . . . . . Experimental sequence of lattice modulation spectroscopy . . . . . . . . . Effective band structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . TOF images obtained after lattice modulation and band mapping . . . . . Interband transition probabiities . . . . . . . . . . . . . . . . . . . . . . . Number of excited atoms depending on amplitude and time of modulation Fermionic band structure for 5EK r . . . . . . . . . . . . . . . . . . . . . . . Fermionic band structure for different lattice depths . . . . . . . . . . . . Detailed band structure for 5ErK . . . . . . . . . . . . . . . . . . . . . . . Detailed analysis for fermions in the third band . . . . . . . . . . . . . . . Broadening due to finite modulation time . . . . . . . . . . . . . . . . . . Effect of harmonic confinement and finite modulation time . . . . . . . . . Lattice calibration using bosons . . . . . . . . . . . . . . . . . . . . . . . . Band structure for bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . Detailed band structure for bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 21 21 22 24 25 27 28 29 30 31 32 33 34 35 4.1 4.2 Effective potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Band structure for a mixture of bosons and fermions . . . . . . . . . . . . . 39 40 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 Schematic of the absorption detection system Explanations of the different resolution limits Spherical aberration . . . . . . . . . . . . . . Coma . . . . . . . . . . . . . . . . . . . . . . Astigmatism . . . . . . . . . . . . . . . . . . Petzval field cuvature . . . . . . . . . . . . . Distortion . . . . . . . . . . . . . . . . . . . . Example of a spot diagram . . . . . . . . . . Example for a wavefront analysis . . . . . . . 44 46 47 48 48 49 49 51 52 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . an optical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 72 6 Conclusion and Outlook 5.10 5.11 5.12 5.13 5.14 5.15 5.16 5.17 5.18 5.19 5.20 5.21 5.22 5.23 5.24 5.25 5.26 MTF for an diffraction limited objective . . . . . . . . . . . . Setup of the old detection syste, . . . . . . . . . . . . . . . . Spot diagram for the old, two lens setup . . . . . . . . . . . . Wavefront analysis for the old, two lens setup . . . . . . . . . MTF for the old detection setup . . . . . . . . . . . . . . . . New diffraction limited detection lens setup . . . . . . . . . . Spot diagram for the new detection setup . . . . . . . . . . . Wavefront analysis for new optical design . . . . . . . . . . . MTF for the new detection setup . . . . . . . . . . . . . . . . Mechanical realization of the new lens . . . . . . . . . . . . . USAF resolution test chart . . . . . . . . . . . . . . . . . . . Maximum achievable resolution with the new detection setup Depth of field measurement . . . . . . . . . . . . . . . . . . . Focus shift depending on the test chart position . . . . . . . . Field of view measurement . . . . . . . . . . . . . . . . . . . . Focus shift for different wavelengths. . . . . . . . . . . . . . . Grid chart with 1x magnification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 54 55 55 56 58 59 60 61 62 63 63 64 64 65 66 67 3.1 Fitted lattice depth for spin polarized fermions . . . . . . . . . . . . . . . . 30 4.1 Fermionic tunneling rate for different admixtures of bosons . . . . . . . . . 40 5.1 5.2 5.3 5.4 5.5 Aberration coefficients for the two lens detection setup Error function for new detection optics . . . . . . . . . Lenses used in the new detection setup . . . . . . . . . Focussing lenses used in the new lens . . . . . . . . . . Aberration coefficients for the new lens setup . . . . . 54 58 59 59 60 List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography 73 Bibliography [1] A Einstein. Quantentheorie des einatomigen idealen gases. Sitzungsbericht Preussische Akademie der Wissenschaften, 1924. [2] 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, 1995. [3] K Davis, M Mewes, M Andrews, N van Druten, D Durfee, D Kurn, and W Ketterle. Bose-Einstein Condensation in a Gas of Sodium Atoms. Physical Review Letters, 1995. [4] B DeMarco and D. S. Jin. Onset of Fermi Degeneracy in a Trapped Atomic Gas. Science, 1999. [5] Cheng Chin, Rudolf Grimm, Paul Julienne, and Eite Tiesinga. Feshbach Resonances in Ultracold Gases. arXiv, 2008. [6] CA Regal and M Greiner. Observation of Resonance Condensation of Fermionic Atom Pairs. Physical Review Letters, 2004. [7] D Jaksch, C Bruder, J Cirac, C Gardiner, and P Zoller. Cold Bosonic Atoms in Optical Lattices. Physical Review Letters, 81(15):3108–3111, oct 1998. [8] M Greiner, O Mandel, T Esslinger, TW Hänsch, and I Bloch. Quantum phase transition from a superfluid to a Mott insulator in a gas of ultracold atoms. Nature, 2002. [9] Julian Struck, Rodolphe Le Targat, Parvis Soltan-Panahi, Maciej Lewenstein, Patrick Windpassinger, and Klaus Sengstock. Quantum simulation of frustrated magnetism in triangular optical lattices. arXiv.org, 2011. [10] J Simon, WS Bakr, R Ma, ME Tai, and PM Preiss. Quantum simulation of antiferromagnetic spin chains in an optical lattice. Nature, 2011. [11] Michael Köhl, Henning Moritz, Thilo Stöferle, Kenneth Günter, and Tilman Esslinger. Fermionic Atoms in a Three Dimensional Optical Lattice: Observing Fermi Surfaces, Dynamics, and Interactions. Physical Review Letters, 2005. [12] Robert Jördens, Niels Strohmaier, Kenneth Günter, Henning Moritz, and Tilman Esslinger. A Mott insulator of fermionic atoms in an optical lattice. Nature, 2008. [13] U Schneider, L Hackermuller, S Will, Th Best, I Bloch, T A Costi, R W Helmes, D Rasch, and A Rosch. Metallic and Insulating Phases of Repulsively Interacting Fermions in a 3D Optical Lattice. Science, 2008. 74 Bibliography [14] G Modugno, G Roati, F Riboli, and F Ferlaino. Collapse of a Degenerate Fermi Gas. Science, 2002. [15] C Ospelkaus, S Ospelkaus, L Humbert, P Ernst, K Sengstock, and K Bongs. Ultracold Heteronuclear Molecules in a 3D Optical Lattice. Physical Review Letters, 2006. [16] KK Ni, S Ospelkaus, MHG De Miranda, and A Pe’er. A High Phase-Space-Density Gas of Polar Molecules. Science, 2008. [17] S Ospelkaus, KK Ni, D Wang, and MHG De Miranda. Quantum-State Controlled Chemical Reactions of Ultracold Potassium-Rubidium Molecules. Science, 2010. [18] Seiji Sugawa, Kensuke Inaba, Shintaro Taie, Rekishu Yamazaki, Makoto Yamashita, and Yoshiro Takahashi. Interaction and filling-induced quantum phases of dual Mott insulators of bosons and fermions. Nature Physics, 2011. [19] K Günter, T Stöferle, H Moritz, and M Köhl. Bose-Fermi Mixtures in a ThreeDimensional Optical Lattice. Physical Review Letters, 2006. [20] S Ospelkaus, C Ospelkaus, O Wille, M Succo, P Ernst, K Sengstock, and K Bongs. Localization of Bosonic Atoms by Fermionic Impurities in a Three-Dimensional Optical Lattice. Physical Review Letters, 2006. [21] Th Best, S Will, U Schneider, L Hackermuller, D van Oosten, I Bloch, and D S Lühmann. Role of Interactions in Rb87-K40 Bose-Fermi Mixtures in a 3D Optical Lattice. Physical Review Letters, 2009. [22] M Cramer, S Ospelkaus, C Ospelkaus, K Bongs, K Sengstock, and J Eisert. Do Mixtures of Bosonic and Fermionic Atoms Adiabatically Heat Up in Optical Lattices? Physical Review Letters, 2008. [23] Dirk-Sören Lühmann, Kai Bongs, Klaus Sengstock, and Daniela Pfannkuche. SelfTrapping of Bosons and Fermions in Optical Lattices. Physical Review Letters, 2008. [24] Silke Ospelkaus. Quantum Degenerate Fermi-Bose Mixtures of 40K and 87Rb in 3D Optical Lattices. PhD Thesis, 2006. [25] Christian Ospelkaus. Fermi-Bose mixtures — From mean-field interactions to ultracold chemistry. PhD Thesis, 2006. [26] DC McKay and B DeMarco. Cooling in strongly correlated optical lattices: prospects and challenges. Reports on Progress in Physics, 2011. [27] Neil W. Ashcroft and N. David Mermin. senschaftsverlag, 2005. Festkörperphysik. Oldenbourg Wis- [28] Markus Greiner. Ultracold quantum gases in three-dimensional optical lattice potentials. PhD Thesis, 2003. [29] Dirk-Sören Lühmann. Multiorbital Physics in Optical Lattices. PhD Thesis, 2010. Bibliography 75 [30] Jannes Heinze. personal communication, 2011. [31] Christoph Becker. Multi component Bose-Einstein condensates. PhD Thesis, 2009. [32] D van Oosten, P Van der Straten, and H Stoof. Quantum phases in an optical lattice. Physical Review A, 2001. [33] G K Campbell. Imaging the Mott Insulator Shells by Using Atomic Clock Shifts. Science, 2006. [34] Jacob F Sherson, Christof Weitenberg, Manuel Endres, Marc Cheneau, Immanuel Bloch, and Stefan Kuhr. Single-atom-resolved fluorescence imaging of an atomic Mott insulator. Nature, 2010. [35] W S Bakr, A Peng, M E Tai, R Ma, J Simon, J I Gillen, S Folling, L Pollet, and M Greiner. Probing the Superfluid-to-Mott Insulator Transition at the Single-Atom Level. Science, 2010. [36] T Esslinger. Fermi-Hubbard physics with atoms in an optical lattice. Condensed Matter Physics, 2010. [37] A Kastberg, WD Phillips, and SL Rolston. Adiabatic Cooling of Cesium to 700 nK in an Optical Lattice. Physical Review Letters, 1995. [38] Markus Greiner, Immanuel Bloch, Olaf Mandel, Theodor Hänsch, and Tilman Esslinger. Exploring Phase Coherence in a 2D Lattice of Bose-Einstein Condensates. Physical Review Letters, 2001. [39] D McKay, M White, and B DeMarco. Lattice thermodynamics for ultracold atoms. Physical Review A, 2009. [40] J Stenger, S Inouye, A Chikkatur, D Stamper-Kurn, D Pritchard, and W Ketterle. Bragg Spectroscopy of a Bose-Einstein Condensate. Physical Review Letters, 1999. [41] PT Ernst, S Götze, JS Krauser, K Pyka, DS Lühmann, D Pfannkuche, and K Sengstock. Probing superfluids in optical lattices by momentum-resolved Bragg spectroscopy. Nature Physics, 6(1):56–61, 2009. [42] N Fabbri, D Clément, L Fallani, C Fort, M Modugno, K van der Stam, and M Inguscio. Excitations of Bose-Einstein condensates in a one-dimensional periodic potential. Physical Review A, 2009. [43] U Bissbort, Y Li, S Götze, and J Heinze. Detecting the Amplitude Mode of Strongly Interacting Lattice Bosons by Bragg Scattering. arXiv.org, 2010. [44] G Veeravalli, E Kuhnle, P Dyke, and C Vale. Bragg Spectroscopy of a Strongly Interacting Fermi Gas. Physical Review Letters, 2008. [45] Gretchen K. Campbell, Jongchul Mun, Micah Boyd, Patrick Medley, Aaron E. Leanhardt, Luis G. Marcassa, David E. Pritchard, and Wolfgang Ketterle. Observation of the Pairing Gap in a Strongly Interacting Fermi Gas. Science, 2004. 76 Bibliography [46] J T Stewart, J P Gaebler, and D S Jin. Using photoemission spectroscopy to probe a strongly interacting Fermi gas. Nature, 2008. [47] H Häffner, C McKenzie, A Browaeys, W D Phillips, and Rolst. A Bose-Einstein condensate in an optical lattice. Journal of Physics B: Atomic, Molecular and Optical Physics, 2002. [48] T Stöferle, H Moritz, C Schori, and M Köhl. Transition from a Strongly Interacting 1D Superfluid to a Mott Insulator. Physical Review Letters, 2004. [49] Rajdeep Sensarma, K Sengupta, and S Das Sarma. Momentum Resolved Optical Lattice Modulation Spectroscopy for Bose Hubbard Model. arXiv.org, 2011. [50] C Kollath, A Iucci, I P McCulloch, and T Giamarchi. Modulation spectroscopy with ultracold fermions in an optical lattice. Physical Review A, 2006. [51] Daniel Greif, Leticia Tarruell, Thomas Uehlinger, Robert Jördens, and Tilman Esslinger. Probing Nearest-Neighbor Correlations of Ultracold Fermions in an Optical Lattice. Physical Review Letters, 2011. [52] Zhaoxin Xu, Simone Chiesa, Shuxiang Yang, Shi-Quan Su, Daniel E Sheehy, Juana Moreno, Richard T Scalettar, and Mark Jarrell. The response to dynamical modulation of the optical lattice for fermions in the Hubbard model. arXiv.org, 2011. [53] Stephen Gasiorowicz. Quantenphysik. Oldenbourg Wissenschaftsverlag, 2005. [54] Ana Rey, Guido Pupillo, Charles Clark, and Carl Williams. Ultracold atoms confined in an optical lattice plus parabolic potential: A closed-form approach. Physical Review A, 2005. [55] Michael Köhl. Thermometry of fermionic atoms in an optical lattice. Physical Review A, 2006. [56] Jasper S. Krauser. Impulsaufgelöste Bragg-Spektroskopie an bosonischen Quantengasen in optischen Gittern. Diploma Thesis, 2009. [57] Alexander Mering and Michael Fleischhauer. Multi-band and nonlinear hopping corrections to the 3D Bose-Fermi Hubbard model. arXiv.org, 2010. [58] Alexander Albus, Fabrizio Illuminati, and Jens Eisert. Mixtures of bosonic and fermionic atoms in optical lattices. Physical Review A, 2003. [59] Francesca Ferlaino, Chiara D’Errico, Giacomo Roati, Matteo Zaccanti, Massimo Inguscio, Giovanni Modugno, and Andrea Simoni. Feshbach spectroscopy of a K-Rb atomic mixture. Physical Review A, 2006. [60] Andreas Bick. Detektion und Untersuchung von mehrkomponentigen Bose-EinsteinKondensaten in hexagonalen optischen Gittern. Diploma Thesis, 2010. [61] Wolfgang Demtröder. Experimentalphysik 2. Springer Wissenschaftsverlag, 2006. [62] WJ Smith. Modern optical engineering: the design of optical systems. McGraw-Hill Professional, 2008. [63] George Seward. Optical Design of Microscopes. SPIE Society of Photo-Optical Instrumentation, 2009. [64] Franziska Herrmann. Entwicklung optischer Elemente für das Atom Trap Trace Analysis Experiment zur Spurengasanalyse von Kryptonisotopen. Diploma Thesis, 2009. [65] W Alt. An objective lens for efficient fluorescence detection of single atoms. OptikInternational Journal for Light and Electron Optics, 2002. [66] OSLO Optics Reference, 2005. [67] G. D. Boreman. Modulation Transfer Function in Optical and Electro-Optical Systems. SPIE Press, 2001. 78 Bibliography Danksagung Diese Arbeit wäre ohne die Hilfe einiger Personen nicht möglich gewesen. An dieser Stelle möchte ich mich bei all jenen die zum gelingen dieser Arbeit beigetragen haben bedanken. Großer Dank geht an Prof. Dr. Klaus Sengstock für die Möglichkeit ein lehrreiches, spannendes und erlebnisreiches Jahr in seiner Gruppe verbringen zu dürfen. Insbesondere möchte ich mich für das entgegengebrachte Vertrauen, das Engagement und die alles andere als selbstverständlichen Möglichkeiten wie z.B. Konferenz-Besuche bedanken. Bedanken möchte ich mich weiter bei Prof. Dr. Henning Moritz für die freundliche übernahme der Zweitkorrektur und das anfängliche zur Verfügung stellen von OSLO. Ganz besonders möchte ich dem gesamten „BFM“-Team danken: „meinen“ Doktoranden Sören Götze, Jannes Heinze und Jasper Krauser für die Geduld und das Vertrauen was ihr mir entgegengebracht habt. Bei „meinem“ PostDoc Christoph Becker möchte ich mich für die großartige unterstützung das gesamte Jahr über bedanken. Dank geht auch an „meinem“ Diplomanden-Kollegen Nick Fläschner für das allseits offene Ohr für sämtliche Fragen. Dem gesamten „BFM“-Team danke ich für eine unglaublich tolle Atmosphäre und die vielen tollen Erlebnisse im Labor und außerhalb. Bei der gesamten Arbeitsgruppe Sengstock, sowie allen anderen Mitarbeitern des ILPs möchte ich mich für die offenheit und hilfsbereitschaft in jeder Situation bedanken. Last but not least: Einen riesen Dank möchte ich meiner Familie aussprechen: Ohne eure Untersützung wäre ich jetzt nicht da wo ich bin und nicht das was ich bin. Danke!