* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Quantum Theory of Atomic and Molecular Structures and Interactions
Survey
Document related concepts
Quantum vacuum thruster wikipedia , lookup
Density of states wikipedia , lookup
Introduction to gauge theory wikipedia , lookup
Magnetic monopole wikipedia , lookup
Old quantum theory wikipedia , lookup
History of subatomic physics wikipedia , lookup
Hydrogen atom wikipedia , lookup
Renormalization wikipedia , lookup
Electromagnet wikipedia , lookup
Fundamental interaction wikipedia , lookup
History of quantum field theory wikipedia , lookup
Superconductivity wikipedia , lookup
Electromagnetism wikipedia , lookup
Time in physics wikipedia , lookup
Nuclear structure wikipedia , lookup
Aharonov–Bohm effect wikipedia , lookup
Transcript
A Dissertation entitled Quantum Theory of Atomic and Molecular Structures and Interactions by Constantinos Makrides Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics Dr. Bo Gao, Committee Chair Dr. Svetlana Kotochigova, Committee Member Dr. Lawrence Anderson-Huang, Committee Member Dr. Steven Federman, Committee Member Dr. Thomas Kvale, Committee Member Dr. Patricia Komuniecki, Dean College of Graduate Studies The University of Toledo August 2014 Copyright 2014, Constantinos Makrides This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author. An Abstract of Quantum Theory of Atomic and Molecular Structures and Interactions by Constantinos Makrides Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics The University of Toledo August 2014 This thesis consists of topics in two related areas of research that together provide quantum mechanical descriptions of atomic and molecular interactions and reactions. The first is the ab initio electronic structure calculation that provides the atomic and molecular interaction potential, including the long-range potential. The second is the quantum theory of interactions that uses such potentials to understand scattering, long-range molecules, and reactions. In ab initio electronic structure calculations, we present results of dynamic polarizabilities for a variety of atoms and molecules, and the long-range dispersion coefficients for a number of atom-atom and atom-molecule cases. We also present results of a potential energy surface for the triatomic lithium-ytterbium-lithium system, aimed at understanding the related chemical reactions. In the quantum theory of interactions, we present a multichannel quantum-defect theory (MQDT) for atomic interactions in a magnetic field. This subject, which is complex especially for atoms with hyperfine structure, is essential for the understanding and the realization of control and tuning of atomic interactions by a magnetic field: a key feature that has popularized cold atom physics in its investigations of few-body and many-body quantum systems. Through the example of LiK, we show how MQDT provides a systematic and an efficient understanding of atomic interaction in a magnetic field, especially magnetic Feshbach resonances in nonzero partial iii waves. iv This dissertation is dedicated to the memory of my grandfather Konstantinos Mousios. Acknowledgments I am forever indebted to my adviser, Dr. Bo Gao. I will always look back at my time as a graduate student and remember his inspiration and constant support. My interest in atomic and molecular physics is a result of his guidance and positive attitude. Beyond the discussions on physics, the insight and perspective on many aspects outside physics are things that I will keep close to my heart. I also want to express my deepest gratitude to Dr. Svetlana Kotochigova. The time I spent at the National Institute of Standards and Technology (NIST) has produced many positive experiences that I have cherished and I have learned quite a bit from all the projects we completed together. I would also like to thank Dr. Eite Tiesinga and Dr. Alexander Petrov for their assistance during my time at NIST. I am grateful for the patience and guidance from the other committee members, Dr. Lawrence Anderson-Huang, Dr. Steven Federman and Dr. Thomas Kvale. I would also like to thank the other members of the “Toledo Rat Pack”, Ming Li, Thomas Allen, Carl Starkey, Sam Ibdah, and Brad Hubartt. I am grateful for the lasting friendship over the past few years and I hope for many more years to come. Lastly, I would like to acknowledge my immediate family for their support of my graduate studies. My parents Demetrios and Agathi Makrides, my siblings Yiannis and Marianna Makrides and my grandmother Fereniki Mousios. It is always difficult to leave home, especially in a family as close as our own, but I have never felt alone or forgotten. As is evident by the mythological hero Odysseus, life often forces you to be far away from those you love. Thank you for always being there for me. vi Contents Abstract iii Acknowledgments vi Contents vii List of Tables ix List of Figures x List of Abbreviations xii 1 Introduction 1.1 1 Outline of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Ab initio Electronic Structure Calculations 2.1 2.2 2.3 2 4 Potential Energy Surface of LiY bLi . . . . . . . . . . . . . . . . . . . 6 2.1.1 Calculation of the Surface . . . . . . . . . . . . . . . . . . . . 6 2.1.2 Analytic Fit of the Data . . . . . . . . . . . . . . . . . . . . . 9 2.1.3 Technical Details and Results . . . . . . . . . . . . . . . . . . 12 2.1.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . 19 Dynamical Polarizability . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.2.1 Brief Theoretical Considerations . . . . . . . . . . . . . . . . . 21 2.2.2 Computational Details and Results . . . . . . . . . . . . . . . 23 Dispersion Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . 33 vii 2.4 2.3.1 Theoretical Background . . . . . . . . . . . . . . . . . . . . . 34 2.3.2 Results for Select Systems . . . . . . . . . . . . . . . . . . . . 39 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3 Ultracold Atomic Interaction 46 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.2 MQDT for Atomic Interaction in a Magnetic Field 50 3.2.1 . . . . . . . . . . Fragmentation Channels for Two Heteronuclear Group I Atoms in a Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . 50 3.2.1.1 A Single Group I Atom in a Magnetic Field . . . . . 50 3.2.1.2 Two Heteronuclear Group I Atoms in a Magnetic Field 57 3.2.2 MQDT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.2.3 The K c Matrix for Heteronuclear Group I Systems . . . . . . 59 3.3 MQDT for Magnetic Feshbach Resonances . . . . . . . . . . . . . . . 61 3.4 Results for the 6 Li40 K System . . . . . . . . . . . . . . . . . . . . . . 64 3.5 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4 Concluding Remarks 74 References 76 A Tables of Fitting Parameters for Atomic and Molecular Dynamic Polarizabilities 85 A.1 SrLi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 A.2 LiYb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 A.3 Yb∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 A.4 NaK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 viii List of Tables 2.1 Parameters for the three-body Potential of LiYbLi . . . . . . . . . . . . 19 2.2 Table of Dispersion Coefficients . . . . . . . . . . . . . . . . . . . . . . . 40 3.1 Comparison of Select Feshbach Parameters for 6 Li40 K . . . . . . . . . . . 64 3.2 Table of Feshbach Parameters for LiK System . . . . . . . . . . . . . . . 65 A.1 Ground State SrLi Dynamic Polarizability Fit Parameters . . . . . . . . 86 A.2 LiYb Dynamic Polarizability Fit Parameters . . . . . . . . . . . . . . . . 87 A.3 Metastable 3 P Ytterbium Dynamic Polarizability Fit Parameters . . . . 87 A.4 NaK ground state X1 Σ Dynamic Polarizability Fit Parameters . . . . . . 88 A.5 NaK a3 Σ Dynamic Polarizability Fit Parameters . . . . . . . . . . . . . . 88 ix List of Figures 2-1 Computational Plan of LiYbLi Surface . . . . . . . . . . . . . . . . . . . 8 2-2 Two-body Potentials of LiYb and Li2 . . . . . . . . . . . . . . . . . . . . 11 2-3 LiYbLi Potential Energy Surfaces: Isosceles Configuration . . . . . . . . 16 2-4 LiYbLi Potential Energy Surfaces . . . . . . . . . . . . . . . . . . . . . . 17 2-5 LiYbLi Three-Body Interaction Surfaces . . . . . . . . . . . . . . . . . . 18 2-6 Dynamic Polarizability of X2 Σ CaYb+ Molecule . . . . . . . . . . . . . . 26 2-7 Dynamic Polarizability of A2 Σ CaYb+ Molecule . . . . . . . . . . . . . . 27 2-8 Dynamic Polarizability of SrLi Molecule . . . . . . . . . . . . . . . . . . 30 2-9 Dynamic Polarizability of a3 Σ state of NaK . . . . . . . . . . . . . . . . 31 2-10 Dynamic Polarizability of X1 Σ state of NaK . . . . . . . . . . . . . . . . 32 2-11 Figure of Molecular Geometry . . . . . . . . . . . . . . . . . . . . . . . . 35 2-12 Vibrationally Dependent Dispersion Coefficients Between Similar 88 Sr7 Li Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2-13 LiYb∗ Long Range Potentials . . . . . . . . . . . . . . . . . . . . . . . . 43 3-1 Picture of Two Channel Magnetic Feshbach Resonance . . . . . . . . . . 48 3-2 Energy Diagram of Group I Atom with IA = 3/2 in a magnetic field . . . 52 3-3 Energy Diagram of Group I Atom with IA = 1 in a magnetic field . . . . 53 3-4 Energy Diagram of Group I Atom with IA = 4 in a magnetic field . . . . 54 3-5 Reduced generalized scattering lengths for LiK in the MF = −5 Manifold 70 3-6 Reduced generalized scattering lengths for LiK in the MF = −4 Manifold 71 3-7 Reduced generalized scattering lengths for LiK in the MF = −3 Manifold 72 x 3-8 LiK ab channel plot of sin2 δl for p wave Resonance . . . . . . . . . . . . xi 73 List of Abbreviations CC . . . . . . . . . . . . . . . . . . . . . . . CCSD . . . . . . . . . . . . . . . . . . . . CCSD(T) . . . . . . . . . . . . . . . . . CCSDT . . . . . . . . . . . . . . . . . . . Coupled Cluster Couple Cluster with Single and Double Excitations CCSD with perturbative Triplet Excitations Couple Cluster with Single, Double and Triple Excitations HF . . . . . . . . . . . . . . . . . . . . . . . . Hartree-Fock MCSCF . . . . . . . . . . . . . . . . . . . Multi Configuration Self Consistent Field MQDT . . . . . . . . . . . . . . . . . . . Multichannel Quantum Defect Theory PES . . . . . . . . . . . . . . . . . . . . . . Potential Energy Surface QDT . . . . . . . . . . . . . . . . . . . . . . Quantum Defect Theory xii Chapter 1 Introduction The ability to probe deeper into the quantum regime and control the interactions of the constituents has opened the door to many investigations into ultracold gases 1 . Since the late 1990’s many atomic species have been cooled to the ultracold regime, which is approximately defined as below a few µK. The ability to control the interactions between atoms has been a hallmark for the field of cold atomic physics. One of the first investigations in tuning ultracold atomic interactions was done by Inouye et al. [2] for a 23 Na Bose-Einstein condensate. In their experiment, the sign and strength of the interaction between the ultracold sodium atoms were shown to vary with the applied external magnetic field. This feature is called a Feshbach resonance and has since then been utilized and observed for a growing list of atoms, e.g. Rb [3], Li [4], K [5], Sr [6] and Yb [7, 8]. This ability to control ultracold atoms provides a control mechanism in the experimental studies of few and many-body systems. Essential to this control is having the best possible understanding of the atomic interactions around a Feshbach resonance. At the present moment, most of the previous theoretical understanding of Feshbach resonances have been limited to s-wave resonances[9], while support for non s-wave resonances has been lacking. One of the aims of this thesis is to provide some of this understanding to higher partial waves. 1 See Chin et al. [1] and references therein. 1 A particular application of Feshbach resonances is the creation and control of ultracold polar molecules. Polar molecules are exciting to use in ultracold molecular experiments since they posses a permanent dipole moment, which can be utilized for various purposes such as improving trap coherence times [10] and precision measurement studies [11]. Ultracold polar molecules have been created through Feshbach resonances; however, these “Feshbach molecules” are very weakly bound molecules. It should be noted that direct cooling of molecules to ultracold temperatures is problematic due to the internal structure of the molecules. These “Feshbach molecules” provide a successful alternative route in obtaining ultracold molecules. Subsequent techniques can later bring these weakly bound molecules to the rovibrational ground state. The successful creation of these ground state polar molecules has been realized by Ni et al. [12] for KRb and Takekoshi et al. [13] for RbCs. In both of those cases, a detailed knowledge of the molecular properties and structure is needed to bring these molecules to the ground state. This information for the most part is lacking in many of the contemporary systems and is another focus of this thesis. 1.1 Outline of the Thesis The present thesis is organized to address the two goals previously outlined. Chapter 2 focuses on ab initio electronic structure calculations for both atoms and molecules. The chapter aims on addressing the lack of important atomic and molecular properties. The discussion begins by calculating the Born-Oppenheimer potentials for the lithium-ytterbium-lithium system. This potential will be presented as part of a work on ultracold reactive scattering. This potential, commonly called a potential energy surface, is the higher dimensional analogue of the familiar diatomic potential energy curves. The aim of this work is to provide a surface for numerical computations, which in turn will be used as a test of alternative descriptions of re2 active scattering (e.g. Gonzlez-Lezana [14]). In connection to the discussion on polar molecules, LiYb has a large electric dipole moment and is being experimentally realized by [15]. Understanding and controlling the reactions between the LiYb molecules and background Li atoms can improve the production and stabilization of LiYb polar molecules. The discussion on electronic structure calculations then proceeds to the calculation of the dynamic polarizability for a number of atoms and molecules. The dynamic polarizability is important on its own merit; however in the present thesis it is necessary in the calculation of dispersion coefficients of neutral-neutral systems.The dispersion interactions of molecules are slightly different than their atomic (ground state) counterparts in that the anisotropy is quite significant. This anisotropy has gained much attention in recent years especially in the case of polar molecules [16]. In chapter 3 the discussion moves to ultracold atomic interactions in the framework of Multichannel Quantum Defect Theory(MQDT) for alkali systems. This chapter addresses the lack of non s-wave descriptions of magnetic Feshbach resonances, by providing precisely these descriptions. The focus is on providing this exact unified description for all resonances of arbitrary partial wave. The discussion in this chapter is applicable to all alkali systems; however, the emphasis will be on a test system, 6 Li40 K. We will depict these new characterization through this system. The final chapter outlines the key results from the previous chapters and will purpose some interesting directions to the current works. Unless otherwise noted, atomic units are used in all equations. In this system of units, the electron charge, electron mass, ~, and the Coulomb constant ke = 1/4π0 are set to unity. 3 Chapter 2 Ab initio Electronic Structure Calculations Ab initio is a Latin term that means “from the beginning”. In the present context of atomic physics, it corresponds to the numerical procedure of solving the electronic part of the Schrödinger Equation to make predictions without the aid of empirical information. There have been numerous packages that provide a number of features for one to use to conduct numerical electronic structure calculations. The field of ab initio electronic structure calculations have been somewhat merged with quantum computational chemistry as there is significant overlap between the two. Thus, many of these software packages for electronic structure calculations are refereed to as quantum chemical packages. These packages have had considerable growth since the 1950s when computer technology was becoming available for computational purposes. An early method in the computation of the Schrödinger Equation comes from Hartree [17] and Fock [18], which is called the Hartree-Fock Method1 (HF). Despite being a relatively old method, the method is still used today with a few improvements added along the way. The main limitation of the HF method is that electron correlations are completely neglected for many-electron atoms. Regardless, the HF method 1 This method is also called the Self Consistent Field (SCF) method. In the present literature there seems to be no distinction between the two. 4 is still used as a reference wavefunction for a number of other post HF calculations. The Møller-Plesset perturbation theory [19] and the heavily used Coupled-Cluster method [20] are two such examples that have been heavily used in these works. It is also note worthy to mention that the HF method results in a single determinant approximation to the multi-electron wavefunction. The alternative is to use multiple determinants, which in the literature is called the Multi-Configuration Self Consistent Field [21] method2 . The most basic results of the HF and post-HF methods are calculations of the energy of the system. In the case of an atom, this would correspond to the energy levels of the electrons. For diatomic molecules, the resulting calculations can produce a potential energy surface (PES) as a function of inter-nuclear distance. As the number of atoms grows, so does the dimension of the surface. The present discussion on ab initio single point energy calculations will focus on the calculation of the triatomic PES of lithium-ytterbium-lithium. The sections following the triatomic PES are devoted to the calculation of the dynamic polarizabilities and of the dispersion coefficients. The dynamic polarizability is another property that can be directly calculated using ab initio methods. It is a quantity associated with the response of the object to an electric field and is important in trapping and manipulating the internal states of atoms and molecules (e.g. [10, 22]). This is a quantity that is very relevant in the present study and will be one of the computations reported. The motivation for obtaining dynamic polarizabilities is that they can be used to determine the dispersion coefficients for neutral-neutral systems, which describe the long range behavior. These dispersion coefficients are in turn needed for describing atomic interactions and can be used in theories such as the Multichannel Quantum Defect Theory (MQDT) used later in chapter 3. 2 Historically this has also been called the Multi-Configuration Hartree-Fock method. 5 2.1 Potential Energy Surface of LiY bLi As mentioned in the Introduction, the motivation for producing a PES for this system is to approach ultracold reactions. For the present system we are interested in the exothermic reaction LiY b + Li → Li2 + Y b. It is a simple atom exchange reaction and has applications in controlling trap losses due to exothermic collisions. The numerical calculations of the reaction rate require as input the PES to solve the time independent Schrodinger equation3 . Once the rate can be produced from this surface, other alternative calculations on the reaction rate can be tested against this surface. The Statistical method of Gonzlez-Lezana [14] is one of the alternative approaches that is considered for this reaction. 2.1.1 Calculation of the Surface The potential surface was calculated at a large number of inter-particle separations using the computational chemistry package CFOUR [24]. The grid for which the system was calculated on is depicted in Fig. 2-1. Wernli et al. [25] have a similar scheme implemented for their system, LiHHe+ . A key difference is that the present system has additional geometric symmetry that reduces the computational domain. In our scheme, the origin is placed in between the two lithium atoms and the lithium atoms are restricted to move equally along the z axis. The addition of the ytterbium atom to the two lithium atoms defines a plane which is labeled as the Y Z plane. Computationally, the ytterbium atom is then placed only in the first quadrant. The calculations with ytterbium in the other quadrants can be determined from symmetry. The step by step procedure is then as follows. 1. Fix the distance between the two lithium atoms. Call this distance RLiLi . 3 For example of this calculation, see [23] for a calculation of H3 6 2. Place one of the lithium atoms at (0, 0, RLiLi /2) and the other at (0, 0, −RLiLi /2). 3. Place the ytterbium atom in the first quadrant in the y−z plane at (0, Y by , Y bz ), where Y by and Y bz are both positive. 4. Calculate the ground state energy of the system. 5. Repeat moving the lithium distance and the ytterbium atom around the quadrant. As a technical detail, placing the ytterbium atom along the z or y axis introduces another symmetry into the geometry and in this new symmetry convergence issues tend to arise in both the SCF and post-SCF calculations. Moving slightly off axis would break this symmetry and return to “normal” operation. For these cases along the axis, the symmetry of the system can once again be exploited to extrapolate to along the axis. This is again due to the fact that the calculations on the other side of the axis can be determined by reflections about the axis in question and in this situation one can expect the energy calculations with ytterbium along either axis to be either a local maximum or minimum in the immediate neighborhood. A simple quadratic fit to the near axial calculations can determine the axial calculations with sufficient accuracy. 7 15 10 Z axis (a.u.) 5 0 0 -5 -2 -4 -10 -6 -15 0 5 Y axis (a.u.) 10 15 -8 -10 -1 0 1 Figure 2-1: Plan of Single Point Energy Calculations of LiYbLi Potential Surface: In this figure a map is shown of the computational domain for the calculation of the PES of LiY bLi. As indicated in the text, the two lithium atoms are placed along the z axis and the ytterbium atom is placed in the colored first quadrant. The different colors represent regions of expected increasing computationally difficulty, with the yellow region expected to take the longest to converge, followed by green then blue. 8 2.1.2 Analytic Fit of the Data Following the plan depicted in Fig. 2-1, the grid was submitted to a computational cluster that calculated the energy of the system in the different geometries. In principle, the whole computational domain could be sampled and would create a rather large database with calculations at each grid point. The practicality of such a situation is not ideal as the computational resource demand would be large and the time constraints would not allow the full surface to be sampled at a reasonable mesh. This consideration prevents a spline type interpolation to be employed in this situation. The reason is that there would not have a small enough mesh for the spline interpolation to accurately represent the potential surface. An alternative approach is to fit the calculations to an analytic form that is robust enough to represent the entire surface. The analytic form of the potential can be written as a sum of two terms. V (rAB , rAC , rBC ) = Vpairwise + V3Body = V (2) (rAB ) + V (2) (rBC ) + V (2) (rAC ) + V (3) (rAB , rAC , rBC ) (2.1) V (2) (R) is a two-body potential corresponding to a diatomic potential curve and V (3) (rAB , rAC , rBC ) is the three-body term that describes deviations in the PES from the pairwise potential. The pairwise potential is the sum of the two-body potential for each pair of atoms. In the case of three atoms this corresponds to three terms. The advantage of this separation is that the two-body potentials may be calculated independently from the full three-body potential. This allows the usage of potentials from experiments if desired and available. Computationally, the two-body potentials are considerably less resource demanding. This allows one to calculate many points in the potential curve, which has the effect of having more sophisticated analytic 9 representations. For these reasons, the focus will not be on the analytic form of the two-body potentials. In the present case, the two-body potentials for both Li2 and LiY b were calculated to sufficient accuracy and quantity to directly use spline interpolation for the potential curves. To fit the three-body term, the analytic form suggested by Aguado and Paniagua [26] was used, which has a functional form V (3) (rAB , rAC , rBC ) = M X dijk ρiAB ρjAC ρkBC (2.2) i,j,k where for the three-body term, ρAB is defined as, (3) ρAB = RAB e−βAB RAB (2.3) The dijk serve as linear fit parameters and the three β (3) are non-linear fit parameters. These parameters are determined iteratively in the fitting routines. The input parameter M controls the number of fitting parameters. One of the key advantages of using this particular analytic form is the ability to dramatically increase the number of fit parameters and still respect the short, intermediate and long range behaviors of the potential [26]. The routine has two restrictions on M and the indices i, j, k. 1. i + j + k < M 2. i + j + k 6= i 6= j 6= k. These conditions enforce that the three-body term goes to zero when one of the inter-nuclear distances goes to zero [26]. In principle, increasing the value of M will achieve a better fit to the data at the cost of increasing the number of fit parameters. The increase in the number of linear and non-linear terms translates into longer times to determine the parameters; however, even for fairly large values of M ∼ 8, 10 Reactants Products Reactants 8000 LiYb + Li 6000 Li2 + Yb 4000 -1 Potentials (cm ) 2000 0 -2000 D0 LiYb -4000 -6000 -8000 Offset = D0 Li - D0 LiYb D0 Li 2 5 RLiYb [a.u.] 10 15 5 2 RLi [a.u.] 10 2 15 Figure 2-2: Two-body Potentials of LiYb and Li2 : The two-body potential curves for LiYb and Li2 are depicted in this figure. The Figure gives information on the practicality of a reaction involving the initial LiYb molecule interacting with a Li atom and exiting from the reaction as Li2 molecule with an excess ytterbium atom. The potential curves show that the reaction is an exothermic one since the lowest bound state of the LiYb system is higher in energy than the lowest bound state of the Li2 molecule. The potential curves in this figure are real calculations using the same basis as the LiYbLi potential surface, as opposed to the bound states which are used for illustrative purposes. 11 the computational time is modest on contemporary personal desktop computers. The more sensitive issue with respect to increasing M is that the number of fit parameters increases very rapidly. Thus, there is a balance between the number of fit parameters and the total number of calculated points available. Having too large an M can lead to very misleading fits. For example, if the number of fit parameters is on the order of the number of data points (but still less), one can imagine that the fitting routine can fit each of the individual data points resulting in a delta function-like behavior at each data point. In that case, the sum of residual squares will be close to zero indicating a “good” fit; however, it is evident that the function in between the data points will not be well represented. Thus, as a general principle, it is good to keep the value M as small as possible while maintaining a low sum of the residuals squared. 2.1.3 Technical Details and Results For the calculation of the ground PES for the LiYbLi system, a number of technical details need to be specified. The single point energy calculations were all done at the CCSD(T) level. As mentioned earlier, the perturbative treatment of the triple excitations is computationally cost efficient, as the full CCSDT treatment is a significant step up in performance time. For lithium we used the “aug-cc-pCVTZ” basis set of Prascher et al. [27], which is a contracted basis that includes a number of diffuse functions. For the calculation of the ytterbium atom we use a much larger contracted basis set similar to the one found in Cao and Dolg [28]4 . A relativistic pseudopotential5 is also used for the inner 28 electrons, which is common practice for higher Z atoms. The pseudopotential greatly simplifies the electron correlation part of the calculation for these atoms as these methods can scale very poorly (∼O(n7 ))[29]. 4 The actual basis set and pseudopotential used was sent via Private Communication by Dr. M. Dolg, one of the authors of the ytterbium basis set. 5 This is the common name in physics. On the physical chemistry side, this quantity is commonly refereed to as Effective Core Potentials. 12 To further simplify the calculation, a number of closed shell electrons are ignored from the correlation in a process that is known as Frozen-core calculations. For the present calculation, the lowest 15 orbitals are included in the core, which corresponds to the 1s electrons for each of the lithium atoms and the 4s,4p,4d,5s and 5p orbitals for the ytterbium atom. It is important to keep the 4f orbital in the correlation as these electrons are not as deeply bound and contribute to the present case. The exclusion of the core electrons greatly simplifies the post HF calculations that deal with electron correlation. To give a perspective on the computational runtime, the calculations were distributed over a number of modern Intel Xenon processors all of which were either quad core 2.4 GHz processors or 8 core 2.1 GHz processors. The multi-processor nodes allow multiple calculations to run concurrently on the same processor. On average 20 to 40 processors where used over a duration of two months to complete all of the purposed calculations. To further emphasize the choice of using CCSD(T) calculations over CCSDT, for diatomic single point energy calculations, the CCSDT calculation can take up to 50 times longer to converge over the corresponding CCSD(T) calculation. If one assumes this type of scaling for the three atom case, the resulting computation would take years to complete the same set of calculations. In addition to this scaling, the inclusion of more electrons in the correlation also scale the computation time greatly. The raw data points were passed through a screening procedure to eliminate data points that may have converged to a different (possibly excited) surface. These points should be sparse and are typically a result of the preceding SCF calculation converging to the wrong solution. In many cases, this difference can be rather large and abrupt and easily filtered out. The difficulty arises in certain regions where either (or both) the SCF or CC methods take longer than usual to converge. In these cases, the validity of the calculation can be difficult to discern by evaluating simple energy 13 differences with the neighboring data points. At worst, if other diagnostic information in the program output fails to conclusively rule out the calculation, the calculation is temporarily removed from the fitting process. The rigid selection criteria results in a large amount of data points being discarded. Of the original 1800 calculations submitted to the computational cluster, the selection criteria resulted in a final data set with 591 points. To a large degree, the difficulty is primarily due to the correlation of the 4f shell of the ytterbium atom. Even at the CCSD level of theory, the total computation time per data point can be rather large. With regard to the fitting parameters, the best representation of the calculated data points was found with M = 4, which corresponds to thirteen linear parameters and two non-linear parameters. The parameters used in the final version of the fit can be found in table 2.1. In the process of fitting the three-body potential, numerous fits were made with M going up to seven. In principle, increasing M will reduce the error between the fit and the data points; however, the resulting surface tends to have distortions and oscillations that are fictitious and artifacts of the fitting procedure. The reason for these fictitious features is difficult to discern. The root of the issue seems to be due to the fact that there are no constraints on the parameters in the regions where the distortions appear, thus they are free to take on any form. Another way to look at this issue is to consider the case where the number of fit parameters is comparable to the number of data points. In that scenario, it can be imagined that there would be some fit parameters that do not actually have any real constraints on them and will be able to “act” freely. As you reduce the number of fit parameters, i.e. lower M , the fit parameters start representing more and more data points; thus the number of free unrestricted parameters decrease until you have just enough parameters to fit the data. In the present case, the choice of M = 4 seems to not contain any fictitious oscillations in the sampled geometries. To prevent any unforeseen oscillations in the long range, the three-body term is forced to zero 14 via a smooth switching function past the point where the wavefunctions overlap. This choice is again determined by the diatomic potentials and is found to be at approximately 15 a0 for both potential curves. In addition to the computational complexities, the potential surface for three atoms cannot be shown in full as the surface requires a four dimensional plot. What is often useful is to plot a three dimensional cross section by restricting the independent variables of the system. In Fig. 2-3 The atoms are arranged in an Isosceles configuration, where the ytterbium to lithium distance is the same for both lithium atoms. The top panel shows the complete surface and the bottom panel shows the three-body contribution. This configuration allows us one to look at the angular profile of the molecule, which can give some insight on the nature of this interaction. There is a global minimum in the surface that seems to be primarily due to the pairwise interaction. The minimum in the surface occurs at about 6 a.u., which is close to the ytterbium-lithium two-body minimum as seen in Fig. 2-2. Figures 2-3, 2-4 and 2-5 show different aspects of the PES. The various panels in figure 2-5 show the three-body contributions to the potential at different molecular geometries. The three-body contribution has some of the typical features that one would expect. At sufficiently large distances, the three-body contribution tends to zero and is reflected in all four panels in figure 2-5. There is also some variation in a structure found in all four panels of figure 2-5. This feature is repulsive for more acute angles, while the opposite is true for the more obtuse angles. This structure in the three-body potential is also significant in magnitude. The plots in figure 2-5 show the maximum of the three-body term to be on the order of 10−2 a.u. (∼ 2000 cm−1 ). In comparison, Figure 2-2 shows the minimum of the LiYb diatomic potential to be slightly larger in magnitude. The Li2 diatomic potentials are a bit deeper, but not large enough to justify completely ignoring the three-body potential. 15 V(r12,r13,r23) (a.u.) A 0.02 0 -0.02 -0.04 60 120 θLiYbLi 180 -0.05 6 -0.025 14 8 10 12 rYbLi (a.u.) 0 0.025 V(r12,r13,r23) (a.u.) B 0.02 0 -0.02 -0.04 60 120 θLiYbLi 180 -0.05 6 -0.025 14 8 10 12 rYbLi (a.u.) 0 0.025 Figure 2-3: LiYbLi Potential Energy Surfaces: Isosceles Configuration: A cross section of the PES of LiYbLi. The top panel, the complete potential is shown. The bottom panel has the pairwise interaction removed leaving only the three-body interaction. The base of each figure shows a contour plot of the surface. 16 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 6 12 15 rYbLi1 (a.u.) 9 -0.025 0 0.025 C V(r12,r13,r23) (a.u.) V(r12,r13,r23) (a.u.) B 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 6 12 15 rYbLi1 (a.u.) 9 -0.025 0 0.025 D 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 -0.025 6 12 15 rYbLi1 (a.u.) 9 0 V(r12,r13,r23) (a.u.) V(r12,r13,r23) (a.u.) A 0.025 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 -0.025 6 12 15 rYbLi1 (a.u.) 9 0 0.025 Figure 2-4: LiYbLi Potential Energy Surfaces: Various plots of the potential energy surface of the LiYbLi system at different geometries. Each panel fixes the angle with ytterbium at the vertex. The angles are 45◦ , 90◦ , 135◦ and 180◦ for panels A, B, C and D, respectively. 17 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 6 12 15 rYbLi1 (a.u.) 9 -0.025 0 0.025 C V(r12,r13,r23) (a.u.) V(r12,r13,r23) (a.u.) B 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 6 -0.025 12 9 rYbLi1 (a.u.) 0 15 0.025 D 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 -0.025 6 12 15 rYbLi1 (a.u.) 9 0 V(r12,r13,r23) (a.u.) V(r12,r13,r23) (a.u.) A 0.025 0.02 0 -0.02 -0.04 6 9 12 rYbLi2 (a.u.) 15 -0.05 -0.025 6 12 9 rYbLi1 (a.u.) 0 15 0.025 Figure 2-5: LiYbLi Three-Body Interaction Surfaces: Various plots of the the three-body interaction for the LiYbLi system. Each panel fixes the angle with ytterbium at the vertex. The angles are 45◦ , 90◦ , 135◦ and 180◦ for panels A, B, C and D, respectively. 18 Table 2.1: Parameters for the three-body Potential of LiYbLi: Included here are the fitted parameters used to construct the threebody potential. The notation is as in the text with the dijk being the linear fit parameters and the β (3) being the non-linear parameters. The indices i and j correspond to powers of ρ for LiY b distances while the index k is for powers of ρ for the LiLi distance. i 1 1 1 2 2 0 2 1 2 2 3 3 0 j k dijk 1 0 -0.3791234233645178 × 100 0 1 -0.1207092112030131 × 102 1 1 0.6778574385332172 × 101 1 0 0.9609698323047215 × 100 0 1 0.1808003175501403 × 102 2 1 0.4946458265430991 × 100 1 1 -0.2785537833078476 × 102 2 1 0.2029448131083818 × 101 2 0 -0.8786730695382046 × 100 0 2 0.1047435507501138 × 103 1 0 0.7103760674735782 × 10−1 0 1 0.3961820620689986 × 102 3 1 -0.1402545726485475 × 100 (3) βLiY b = 0.7110242142956382 (3) βLiLi = 0.2079741859771922 2.1.4 Concluding Remarks This concludes the discussion of the PES for the LiYbLi system. Emphasized here were the computational details of computing the PES. Some of the technical details of fitting the data to an analytic formula were discussed as well. The fit parameters for the three-body potential can be found in table 2.1. The pairwise potentials require the diatomic potentials between the two lithium atoms and between lithium and ytterbium, which can be found by ab initio means as is depicted in figure 2-2, or experimentally. The structure of this potential does not seem to have any particular features much different from three-body potentials of other systems found in Wernli et al. [25]. The deciding factor on the accuracy of the PES will have to wait until an 19 experimental measurement of the reaction rate of this system can be made. Unfortunately at the time of this thesis the question on the reaction rate cannot be answered6 . Another question that would be of significant interest would be the effect of the three-body term on the reaction rate. Presumably, if the three-body term shows little to no difference on the rate, one could conceivably think about a class of reactions that depend solely on the long range interactions [30, 14] and the conditions to which such conditions apply. This is a question that we plan on addressing in the near future. 2.2 Dynamical Polarizability The dynamic polarizability is one of the most important quantities in ultracold experiment and theory. Its theoretical foundation is quite historic and is one of the most well known applications of perturbation theory. Since the dynamic polarizability describes the interaction between the atom (or molecule) and a laser at different frequencies, it is of great importance to the understanding of laser-atom interactions. Most experimental laser frequencies are not too large due to the financial cost of those type of lasers. A typical YAG laser has a wavelength of about λ = 1064 nm, which translates into a laser frequency of about 9400 cm−1 . A CO2 laser has a frequency on the order of about 1000 cm−1 . Quite often, the static polarizability is used to approximate the polarizability at these laser frequencies; however, there are cases where the two are quite different due to the presence of excited states nearby. In addition, as mentioned in the Introduction to the thesis, the polarizabilities are used in the calculation of the dispersion coefficients. This calculation is one of the main motivations for obtaining the dynamic polarizability. Most of the polarizabilities 6 The reaction rate for this PES is currently being calculated by the group of Dr. Naduvalath at UNLV. 20 for ground state atoms can be obtained by extracting the oscillator strengths from the NIST ASD[31] database, as the accuracy of such states are quite high. The situation is quite different for the molecules of interest, as the calculation of the all oscillator strengths for a particular diatomic molecule can be quite laborious. In the following section, the process of calculating the dynamic polarizability will be highlighted, describing the computational details in addition to touching upon some of the important theoretical aspects. Towards the end of the section, the fitting process will be explained and applied to a number of systems. Appendix A will contain the exact fitting data for the systems. 2.2.1 Brief Theoretical Considerations In the framework of perturbation theory, the dynamic polarizability at laser frequency ω can be determined though second-order perturbation theory7 and is given by, X D E 2 En0 − En 0 ~ α(ω) = 2 n d · ê n (En0 − En )2 − ω 2 . (2.4) n0 6=n where d~ is the dipole moment operator and ê is the polarization of the applied field. The connection between oscillator strengths and the dynamic polarizability is embedded in Eq. 2.4. Specifically, the oscillator strength fnn‘ is defined as, fnn‘ X D E 2 0 ~ == 2 n d · ê n (En0 − En ). (2.5) n0 6=n In principle the polarizability is a tensor that relates the components of the applied electric field to the induced dipole. For a linear molecule, the only non-zero elements 7 This qualitatively can be understood by the fact that the interaction energy of an induced dipole in a field is proportional to the field squared and relate this to the second order correction to the energy going also as the field squared. 21 of the polarizability tensor in the molecular frame is along the diagonal [32]. Taking the z axis as the internuclear axis, the polarizability tensor takes the form of α ⊥ 0 0 α= 0 α 0 ⊥ 0 0 αk (2.6) where αk is the polarizability component along the z axis and α⊥ is the component perpendicular to this axis. The molecule must also satisfy some atomic limits. If one were to imagine a diatomic molecule at very large internuclear distances, it would closely resemble two separate atoms. Thus for a diatomic molecule with two atoms, A and B, its polarizability at large internuclear separation is, R →∞ αAB −−AB −−−→ αA + αB . (2.7) This equation will be exploited later in the discussion on dispersion coefficients when large separation limits are taken and examined. It is also often useful to define for later discussions the isotropic polarizability for linear molecules, αiso ≡ 1 αk + 2α⊥ 3 (2.8) and the anisotropic polarizability for linear molecules, ∆α ≡ αk − α⊥ (2.9) For comparative purposes, consider the polarizability of a ground state alkali atom. The polarizability respects the symmetry of the atom and is thus completely atom isotropic, meaning that α̂atom = α · 1̂. As a result of this symmetry, α⊥ is zero. In addition, the origin of the parallel and perpendicular polarizabilities is slightly 22 different. In the case of a Σ ground state, the parallel polarizability is due to the dipole coupling to other Σ states. In the perpendicular case, it is due to dipole coupling to Π states. 2.2.2 Computational Details and Results Following the discussion in the previous section, the most direct way of calculating the dynamic polarizability would be to calculate the transition dipole moments between the different states and terminate the sum in equation 2.4. This approach is colloquially called the “sum-over-states” method. Aside from the difficulties in the actual calculation of the transition dipole moments, there is some difficulty in determining how many terms in the sum to use as it becomes increasingly more difficult to accurately calculate higher excited states. To get a good fit, certain features of the dynamic polarizability need to reflected, of which the static polarizability is the most straight forward check to make. In principle, as in the dynamic polarizability, all the terms in the sum contribute, but in practice the static polarizability can get reproduced by a few number of states. Two examples of this sum-over-states method are depicted in figures 2-6 and 2-7. They are both for the molecule CaYb+ . Fig. 2-6 is for the ground state (X2 Σ) while Fig. 2-7 is for the first excited state A2 Σ. The CaYb+ system is different from the other molecules in this thesis in that it is an ionic system. There is significant interest in the charge exchange rate of this system as it has been found to be four orders of magnitude larger than other ionic molecular systems [33]. As part of a theoretical investigation on this rate, the dynamic polarizability of both the ground and first excited state is required to completely describe the effects of the trapping lasers on the ions. The oscillator strengths and energy states were calculated at the MCSCF level of theory using the MOLPRO suite of ab initio programs [34]. In these calculations a 23 total of 20 different electronic states were calculated and used to calculate transition dipole moments between each other. In conjunction with the potential curves, the transition dipole moments can be inserted directly into equation 2.4. The MCSCF method is considered to be less accurate over the “gold-standard” CCSD(T) methods. The difference however is that the MCSCF is a multi reference theory and has better scaling than the CCSD(T) calculations. There are analogous post HF methods for MCSCF calculations; however these methods can be very computationally expensive and would be a massive undertaking to compute a similar number of oscillator strengths. Regardless, the sum-over-states approach is a straightforward way to compute the dynamic polarizability. The other approach involves perturbative methods by adding small frequency dependent dipolar fields to the computational Hamiltonian and to directly calculate the polarizability at various frequencies. Once this computation is done for a sufficient number of points, a fit can be applied and the corresponding parameters can be extracted. This approach will be referred to as “Independent-Frequency-Fit” in this work. For coupled-cluster calculations of polarizabilities the details of the process can be found in Kállay and Gauss [35]. The key aspects from the formulations are that the end results are analytic in terms of the basis function parameters and that the convergence times are “reasonable”. To give a small perspective on the computational runtime of some of these calculations, for the LiYb system, the calculation at ω = 24000cm−1 took approximately three weeks to converge at the CCSD level8 . Frequencies lower than that typically converged quicker, with the exception being if the particular frequency happened to be close to the energy of an excited state, in which case there would be a resonant feature in the dynamic polarizability. In the 8 It should be mentioned that the common CCSD(T) method cannot be used to calculate the dynamic polarizability. The computational cost of CCSDT methods for the dynamic polarizability also make it practically unusable. 24 cases near a resonant feature, the calculation may not converge or will converge very slowly. The independent-frequency-fit approach was used to find the dynamic polarizability for the diatomic strontium-lithium molecule. The SrLi molecule is a polar molecule that was strongly considered experimentally for polar molecule studies. Its dipole moment is relatively larger than that of KRb [36]; however, there was little experimental data on its molecular properties. Individually Sr and Li have both been cooled to the ultracold regime, thus making this system attractive to polar molecule studies. In figure 2-8, dynamic polarizability is plotted at various internuclear distances. The black circles and the red squares are calculations at various frequencies as was mentioned earlier. The solid curves are the corresponding fits to the calculations. These fits have a form of fg (ω) = N X i 1− Ai 2 . (2.10) ω ωgi In the above equation, it is understood that the molecule is in the state g. The functional form of each of the sums was chosen to closely resemble the equation of the dynamic polarizability from second order perturbation theory, eq. 2.4. The parameters, Ai and ωi , thus have direct interpretations in this framework. Ai is related to the oscillator strength, fgi , between the state g and the state i. The other parameter ωgi is the frequency corresponding to the difference in energy between states g and i. Implicitly, the state g is not included in the sum as is also the case in the perturbative equation. To highlight some of the more interesting features, the panel “c” in figure 2-8 shows that the parallel and perpendicular polarizability have become indistinguishable from one another. This is indicative that the atoms are sufficiently far apart such that 25 9000 αPara αPerp R=15.0 a.u. α (a.u.) 6000 3000 0 -3000 -6000 0 5000 10000 15000 20000 -1 ω (cm ) Figure 2-6: Dynamic Polarizability of X2 Σ CaYb+ Molecule: Depicted here is the dynamic polarizability of the X2 Σ state of CaYb+ at the MCSCF level of theory. The internuclear distance is 15 a0 . The polarizability here is calculated by the sum-overstates. As indicated, the black curve is the parallel polarizability and the red curve is the perpendicular polarizability. For this calculation, a total of 20 states are considered. 26 9000 αPara αPerp R=15.0 a.u. α (a.u.) 6000 3000 0 -3000 -6000 0 5000 10000 15000 20000 -1 ω (cm ) Figure 2-7: Dynamic Polarizability of A2 Σ CaYb+ Molecule: Depicted here is the dynamic polarizability of the A2 Σ state of CaYb+ at the MCSCF level of theory. The internuclear distance is 15 a0 . The polarizability here is calculated by the sum-overstates. As indicated, the black curve is the parallel polarizability and the red curve is the perpendicular polarizability. For this calculation, a total of 20 states are considered. The reason for the perpendicular polarizability being relatively flat in the above figure is due to the small oscillator strengths between the states considered. 27 they can be considered as separate atoms. As mentioned in the previous section, at this far distance the dynamic polarizability is the sum of the two polarizabilities of the separate atoms. The other two panels give an impression of the differences in the polarizability as the internuclear distance is changed. The NaK system is one the polar molecules currently being looked at by experimentalists. It has a rather large permanent dipole moment of 2.7D, which is desired for the purposed applications in quantum simulations of solids and quantum information studies. In addition, it is stable against reactions with other NaK molecules to form homonuclear dimers of sodium and potassium. As is the case with many polar systems, there are still efforts to get an ultracold gas of NaK molecules in the ground state. Figures 2-9 and 2-10 show our calculated dynamic polarizability of the NaK molecule at its equilibrium position. Also shown on those graphs is the position of a hypothetical trapping laser at 1550 nm. If the trapping laser is nearly resonant to molecular transition, there will be population transfers to the excited state. These molecules in the excited state will likely not be trapped and thus lead to trap losses. While, our polarizability calculation did not directly calculate the potentials of the excited states, the resonant features in the polarizability indirectly give us the location of some of these potentials. Gerdes et al. [37] have previously investigated some the lower potentials for this system. Our resonance positions agree with the differences in their potentials. The static polarizability for this system was also calculated by Deiglmayr et al. [38]. They calculated the average static polarizability by a finite field method using a configuration interaction type computational package. For comparative purposes, our static polarizability can be determined by evaluating our analytic functional fit at zero frequency. Our value for the average static polarizability is 348 a.u., while they obtained a value of 351 a.u. The difference between the two is quite small and is likely due to the usage of different basis sets in the calculation. 28 These polarizabilities complete the description of the interaction between polar NaK molecules and the trapping laser. One possible direction is to determine the optimal field geometries between all the applied laser and magnetic fields as was done for KRb [10]. The existence of “magic” geometries were found for KRb, which increased the trap coherence time by an order of magnitude . These “magic” geometries orient the trapping laser, an external applied electric field and an external applied magnetic field carefully such that the internal states of the molecule experience the same trapping potential. 29 α (a.u.) 8000 A 4000 0 -4000 α (a.u.) 8000 B 4000 0 -4000 α (a.u.) 8000 C 4000 0 -4000 5000 10000 15000 -1 ω (cm ) 20000 25000 Figure 2-8: Dynamic Polarizability of SrLi Molecule: The dynamic polarizability is depicted above at three different internuclear distances. The distances for panels A, B and C are 6.7, 11.3, and 333.7 a.u. respectively. In each plot, the black (red) curve circles (squares) represent ab initio calculations for the dynamic polarizability parallel (perpendicular) to the internuclear axis. The subsequent lines are fits using the canonical form of the dynamic polarizability. It is worth noting the dynamic polarizability in figure “c” has become fully symmetrical with respect to the parallel and perpendicular directions, indicating that the atomic limits have been reached. The calculated points were done with the computational chemistry package CFOUR. 30 α (a.u.) 6000 4000 2000 0 -2000 -4000 -6000 0 Parallel 1550 nm 5000 10000 15000 -1 20000 25000 30000 ω(cm ) 4000 Perpendicular α (a.u.) 2000 0 1550 nm -2000 -4000 0 5000 10000 15000 -1 20000 25000 30000 ω(cm ) Figure 2-9: Dynamic Polarizability of a3 Σ state of NaK: The polarizability of the a3 Σ state of NaK is shown here at the equilibrium distance, R= 6.59a0 . The top panel shows the parallel polarizability while the bottom shows the perpendicular polarizability. The dashed blue line is the frequency of 1550 nm laser. The calculated points were done with the computational chemistry package CFOUR. The QZVPP basis set was used for both sodium and potassium. 31 α (a.u.) 2000 Parallel 1000 1550 nm 0 -1000 -2000 0 5000 10000 15000 -1 20000 25000 30000 ω(cm ) α (a.u.) 2000 Perpendicular 1000 0 1550 nm -1000 -2000 0 5000 10000 15000 -1 20000 25000 30000 ω(cm ) Figure 2-10: Dynamic Polarizability of X1 Σ state of NaK: The polarizability of the X1 Σ state of NaK is shown here at the equilibrium distance, R= 6.59a0 . The top panel shows the parallel polarizability while the bottom shows the perpendicular polarizability. The dashed blue line is the frequency of 1550 nm laser. The calculated points were done with the computational chemistry package CFOUR. The QZVPP basis set was used for both sodium and potassium. 32 2.3 Dispersion Coefficients As eluded to in the preceding chapters, the full electronic structure sometimes can be very computationally expensive. That being the case, practical investigations into the dynamics of atoms and molecules in the early days of quantum mechanics would be severely limited by the computational abilities of the day. Most of the computational complexity resides in the correlation of many electrons in the inner region where one would expect large wavefunction overlap. In contrast, the situation at long range is far less complicated. The forces in the long range region are much weaker in comparison and can be treated perturbatively. In addition, the long range has very little wavefunction overlap, thus simplifying the situation even further. The issue of long range molecular interaction is also rather historic beginning with London investigating the general form of molecular interaction in the 1930s [39]. In the modern context, the long range dispersion coefficient plays an important role in the understanding of ultracold atomic and molecular scattering [16, 40, 1]. For neutral-neutral collisions, the dominant term in the long range potential is the dispersive van der Waals interaction, which is a second order dipole-dipole interaction. In the case of ground rotationless molecules, this interaction is purely isotropic [16] and has a − 1/r6 type of behavior at long range. The importance of obtaining accurate long range behavior appears in the subsequent chapter on ultracold scattering. While the early investigations were mainly focused on ground state atom-atom interactions, fairly recent experimental control of polar molecules [41, 12] and excited atoms [42] motivates theoretical investigations for these more complex systems. The long range interactions for these systems are no longer isotropic and have rather large anisotropy that cannot be neglected [16, 43]. The discussion on dispersion coefficients will be organized as follows. A theoretical derivation of the form of the potentials will first be presented. In the derivation, the connection between the previous work 33 on dynamic polarizabilities and the present work on dispersion coefficients will be highlighted. Following the derivation, a few results on experimentally relevant systems will then be presented. 2.3.1 Theoretical Background The dispersion coefficient between two neutral atoms can be determined by integrating the product of their dynamic polarizabilities over imaginary frequencies [32]. The long range dispersion interaction of two neutral molecules can be expressed as [44] 1 1 3 2 cos θA − =− 6 C6 + γ20 C6 + RAB 2 2 1 3 2 cos θB − γ02 C6 + 2 2 γ22 C6 (2 cos θA cos θB − sin θA sin θB cos φ)2 2 2 2 − cos θA − cos θB . 6 Udisp. (RAB , θA , θB , φ) (2.11) The coordinates R, θA , θB and φ = φA − φB are shown in figure 2-11. The first term in the curly brackets in equation 2.11 is called the isotropic dispersion interaction as it does not have any angular dependence. The other terms are all anisotropic interactions for analogous reasons. The second and third terms in the curly brackets closely resemble the form of the equation of two interacting dipoles [45]. Despite the lengthy appearance of the potential, there are in fact four dispersion coefficients to determine. The isotropic dispersion coefficient C6 and the anisotropic coefficients γ20 C6 , γ02 C6 and γ22 C6 . Thus, the task of characterizing the long range potential reduces to determining these four coefficients for each system of interest. 34 Figure 2-11: Figure of Molecular Geometry: Shown here are the coordinates in the molecular frame between two molecules A and B. The distance RAB is from the center of mass of A to the center of mass of B. The z axis is typically taken to be the internuclear axis. The angles θA and θB are the angles between the respective molecular axis and the internuclear axis. The angle φ = φA − φB is the relative rotational orientation about the intermolecular axis between the two molecules. 35 The derivation of these coefficients involve applying second order perturbation theory, treating the electrostatic interaction between the electrons of one atom or molecule with those of the other as the perturbation. The derivation is straight forward and can be found in many sources [32, 46, 47, 48]. Each formulation reduces the evaluation of the dispersion coefficients into integrals involving the dipole polarizability. For the isotropic coefficient, the expression is [32] 3 C6 = π Z ∞ dν ᾱA (iν)ᾱB (iν) (2.12) 0 For the current discussion, the frequency dependence of the polarizability is implicitly assumed. The previous section described the details of obtaining the polarizability. In principle, experimentally obtained polarizabilities may be used if available. There should not be too much emphasis on the physical interpretation of the polarizability at complex frequencies. Consider the integrand9 over real frequencies vs the integrand over complex frequencies: by using complex frequencies, the denominators all become positive in the domain ν ∈ (0, ∞), thus making the integrand a monotonically decreasing function. This has advantages if one were to determine the dispersion coefficient by numerical quadrature. The alternative, using real frequency ω, would be filled with singularities making numerical quadrature a much more difficult process. While it would have been sufficient to simply integrate the polarizabilities since they are fitted to a function described by equation 2.10, the specific functional form of the polarizability allows an analytic representation of the integral to be used. Substituting each atom’s polarizability with the functional form in equation 2.10 into equation 2.12 yields 9 In this case it would be simply the product of the two polarizabilities. 36 3 C6 = π Z ∞ dν 0 Ni X i=1 1+ Nj Z ∞ Ni X 3X dν = π i=1 j=1 0 Nj X Āi 2 · ν νAi 1+ Āi 2 · Z ∞ Nj Ni X 3X dν Āi B̄j = π i=1 j=1 0 1+ j=1 ν νAi 1+ 1+ 1 ν νAi B̄j 2 (2.13) ν νBj B̄j 2 2 · (2.14) ν νBj 1+ 1 ν 2 . (2.15) νBj In the above equations the coefficients Ai and Bj do not depend on the frequency ν; thus they may be pulled out of the integral. Once in this form, the integration can be done analytically. The details will not be included here. The outline of the solution involves splitting the integrand into two terms that have functional form 1/1 + ν 2 and then recognizing that each of those integrals can be done via contour integration or by relating it to a special function10 . The solution of the integral becomes Z ∞ dν I= −∞ 1+ 1 ν νAi 2 · 1+ 1 ν 2 = π νAi νBj π =1 . νAi + νBj /νBj + 1/νAi (2.16) νBj Since the integrand is symmetric over ν with respect to positive and negative values, we can take half since the original integral is from 0 to ∞, and substitute it back into the equation for C6 to get our final result. N Nj i X 3X Ai Bj C6 = 1 2 i=1 j=1 /νAi + 1/νBj (2.17) This result is analogous to the Slater-Kirkwood formula [49, 50]. To reiterate, 10 The special function is called the Student’s t-distribution with the characteristic parameter being ν = 1. The form of the integral is obtained by integrating the distribution from −∞ to +∞ and by definition it is equal to unity. 37 the main motivation here was to determine the analytic representation of the C6 coefficient in terms of the parameters used to fit the individual polarizabilities. The advantage is that there is no loss of precision due to numerical issues as the numerical quadrature is not used. In addition, the current formulation makes use of the fact that the integrand is a monotonically decreasing function of ν to minimize the error in C6 due to the cutoff in the sums for the polarizabilities. The cutoff reflects the computational difficulty in obtaining very high frequency calculations. The integrand in those regions has already decayed appreciably in comparison to low frequencies. Thus the contributions to the value of C6 from high frequencies is much less. The other anisotropic dispersion coefficients γ20 , γ02 and γ22 are derived in analogous manners to the isotropic case. The only difference is that the average polarizabilities, ᾱ, are replaced by the “difference” polarizability ∆α [32]. Z 1 ∞ dν ∆αA (iν)ᾱB (iν), γ20 C6 = π 0 Z 1 ∞ γ02 C6 = dν ᾱA (iν)∆αB (iν), π 0 Z 1 ∞ γ22 C6 = dν ∆αA (iν)∆αB (iν). π 0 (2.18) (2.19) (2.20) If the set of parameters for ᾱ can be denoted as {Āi , ν̄Ai } and the set of parameters for ∆α as {Ãi , ν̃Ai }, the resulting expressions for the anisotropic parameters become 38 N Nj i X 1X γ20 C6 = 2 i=1 j=1 N Nj i X 1X γ02 C6 = 2 i=1 j=1 N Nj i X 1X γ22 C6 = 2 i=1 j=1 Ãi B̄j 1 , + ν̃Ai ν̄Bj (2.21) Āi B̃j 1 , + ν̃Bj ν̄Ai (2.22) Ãi B̃j 1 . + ν̃Bj ν̃Ai (2.23) 1 1 1 In the case of ground state alkali and alkaline earth atoms, the polarizability tensor in the atomic frame is proportional to the identity matrix. As a result, the difference polarizability for these atoms is equal to zero. This concludes the derivation of the dispersion coefficients in terms of the fit parameters of the polarizability. The following section uses the polarizabilities and equations 2.17 and 2.21 to obtain the dispersion coefficient for a number of molecular systems. 2.3.2 Results for Select Systems The main results can be found in table 2.2. The parameters for the polarizability of the molecules and excited atoms can be found in appendix A. The systems in table 2.2 all involve polar molecules with SrLi and LiYb being mentioned in the previous subsections. The system LiYb∗ is a unique system as it involves a nonground state atom, Yb∗ . In addition, the Yb atom is about 30 times more massive than the Li atom making this a highly mass-mismatched mixture. The creation of these types of mass imbalanced molecules are important in the future investigations on Effimov trimers [51]. A few remarks can be made on the specific systems. In every case, the isotropic dispersion coefficient has the largest value, followed by the γ20 C6 or γ20 C6 . In both cases, SrLi and LiY b, the γ22 C6 coefficient is the weakest; however, it is large enough 39 and cannot be ignored. Table 2.2: Table of Dispersion Coefficients: Presented here are the values for the isotropic and anisotropic dispersion coefficient. For each system, the molecule is placed at the equilibrium position and its dynamic polarizability is calculated via ab initio methods. The ground state atoms use parameters found in the NIST Atomic Spectra Database [31] and conform to the functional form in the text. The Yb∗ polarizability is also calculated using ab initio programs. The value of the dispersion coefficient is in a.u. (Eh a60 ) System SrLi+Li SrLi+Li SrLi+Sr SrLi+Sr SrLi+SrLi SrLi+SrLi SrLi+SrLi LiYb+Li LiYb+Li LiYb+Yb LiYb+Yb LiYb+LiYb LiYb+LiYb LiYb+LiYb Li+Yb∗ Li+Yb∗ Quantity C6 γ20 C6 C6 γ20 C6 C6 γ20 C6 γ22 C6 C6 γ20 C6 C6 γ20 C6 C6 γ20 C6 γ22 C6 C6 γ02 C6 Value 3130.80 623.42 4683.19 854.61 7308.33 1419.91 989.18 3086.36 775.76 3469.65 779.79 7120.64 1731.36 1432.56 2219.73 240.34 In the case of SrLi the dispersion coefficient is determined for a number of internuclear radii. This permitted investigations into the vibrational dependencies of the dispersion coefficient. This process can be determined by evaluating matrix elements over the vibrational wavefunction. Figure 2-12 depicts this dependence of the isotropic and anisotropic dispersion coefficient for a pair of 88 Sr7 Li molecules in the same vibrational state. The calculation was done using the program VIBROT, which is a part of the MOLCAS quantum chemistry package [52]. The plot shows a few features that are consistent with some basic asymptotic limits. For example, 40 8000 C6 Iso γ20C6 γ22C6 Limit C6 (a.u.) 6000 4000 2000 0 0 2 4 6 Vibrational Quantum Number Figure 2-12: Vibrationally Dependent Dispersion Coefficients Between Similar 88 Sr7 Li Molecules: The leading coefficient of the dispersion interaction between two similar 88 Sr7 Li molecules is plotted as a function of vibrational quantum number ν. The solid black line is the isotropic dispersion coefficient, while the red and blue solid curves represent anisotropic coefficients as described in the text. The dashed line is a flat line at 8436 a.u. that represents the limit as the constituent atoms are all very far apart from one another. The value is determined from a combinatory relationship between all four of the atoms, C6limit = C6LiLi + 2C6SrLi + C6SrSr . 41 equation 2.7 explains that as the intermolecular distance increases, the dynamic polarizability reduces to the sum of the atomic values. Substituting this into the equation for the dispersion coefficient, Eq. 2.12 will obtain the large internuclear separation limit of the dispersion coefficient. For two SrLi molecules, the resulting limit is lim RSrLi →∞ C6,SrLi+SrLi Z 3 ∞ = lim dν ᾱSrLi ᾱSrLi RSrLi →∞ π 0 Z ∞ 3 dν (ᾱSr + ᾱLi ) (ᾱSr + ᾱLi ) = π 0 Z 3 ∞ = dν (ᾱSr ᾱSr + 2ᾱSr ᾱLi + ᾱLi ᾱLi ) π 0 (2.24) = C6,Sr2 + 2C6,SrLi + C6,Li2 . The result is a simple pairwise combinatory expression. In an analogous manner, one would expect the anisotropic values to also tend to zero as the ground state atomic anisotropic coefficients are equally zero. Figure 2-12 for the most part reflects both of these limits, although technical issues prevented results for highly vibrational states to explicitly show this. Specifically, there are certain hard coded parameters in the VIBROT program that prevented the calculation of highly vibrational states11 . To give an impression of the potentials at long range, figure 2-13 is an example of a number of long range potentials constructed using the calculated dispersion coefficients for the 6 Li174 Yb∗ system. The Zeeman and hyperfine interactions12 are also included in this figure to give a accurate impression of the potential landscape. It should be noted that this specific isotope of ytterbium does not have any hyperfine structure. The result of the Zeeman and hyperfine interactions greatly complicate the asymptotic channel structure, as there are many asymptotic channels in a small 11 These parameters were changed in the source code and the routine was recompiled; however, other instabilities occurred likely connected to these hard coded values. 12 See section 3.2.1.1 for more information on the Zeeman and Hyperfine interactions. 42 Figure 2-13: LiYb∗ Long Range Potentials: Depicted here are some of the long range potentials for the LiYb∗ system at a magnetic field of 100G. These potentials all have even ` ≤ 6 and are constructed using the long range coefficients found in table 2.2. Included in these potentials are the Zeeman and hyperfine interactions, which increase the number of asymptotic channels. There are 85 different potentials included in this figure. 43 energy range. The key feature of figure 2-13 is the energy scales associated at regions. The figure shows the long range on the order of 10−7 a.u., while the short range would be on the order of 10−2 a.u. This is not to imply that the long range is not important, but emphasizes that there is a dramatic difference in energy scales between the two regions. It is this physical picture that our MQDT formulation takes advantage of to simplify the theoretical description and will be discussed in chapter 3. 2.4 Chapter Summary Presented in the preceding chapter were some results in ab initio electronic structure calculations. The discussion is predominantly application driven with a variety of systems presented. The chapter was divided into three topics. The first went through the process of constructing a PES for the LiYbLi system. The motivation in that work was to produce a surface that would be used to explore reactive scattering. This surface would be used as a standard to test the validity of a number of theories that attempt to explain the reaction dynamics without relying on computationally expensive numerical techniques. Also presented were the dynamic polarizabilities for a variety of systems. Two separate methods of obtaining the polarizability were highlighted. The first, sumover-states way was shown for the CaYb+ system, which calculated the dynamic polarizability over a large number of states. The other method added dipolar fields to the computational Hamiltonian and calculated the polarizability at a number of frequencies. A fitting procedure is then employed to extract the parameters needed to characterize the polarizability. Lastly, the polarizabilities were employed to calculate the long range interaction between atoms and molecules. It was emphasized that while numerical quadrature techniques could be used, an analytical solution to the integral was used to obtain 44 dispersion coefficients. These computations permitted the construction of potentials that contained the correct long range behavior, which can then be used as input in numerical studies. 45 Chapter 3 Ultracold Atomic Interaction In this chapter a multichannel quantum defect theory for magnetic Feshbach resonances in the interaction of two heteronuclear alkali atoms is presented. Feshbach resonances are one of the principle mechanisms in controlling cold atomic gases. A Feshbach resonance is illustrated in Figure 3-1. It occurs when a bound state of a closed channel energetically crosses the scattering state of an open channel. The scattering state in the ultracold regime is energetically right above the threshold of the open channel, with zero energy being right at threshold. The open channel is sometimes called the entrance channel in some of the literature. In the present discussion, an external magnetic field is applied to the system to provide a control mechanism on the collisional process. The magnetic field controls the energy difference between the bound state and the collisional energy, as is indicated by the red arrow in figure 3-1. This ability to manipulate the energy difference allows one to tune to a magnetic field around a Feshbach resonance. Essential to this magnetic control would be to have the best possible description of the interaction around these resonances to probe different aspects. Historically, both numerical and other theoretical studies have been able to partially characterize s-wave Feshbach resonances with support for higher partial waves lacking in comparison. The present work provides a systematic and unified descrip- 46 tion for resonances in any partial waves. The following chapter will go through the MQDT treatment of the complexities caused by the inclusion of an external magnetic field and present the new description through a sample system, 6 Li40 K. The focus here is on the application of theory to zero energy magnetic Feshbach resonances. 3.1 Introduction Magnetic Feshbach resonance [53] plays an important role in studies and applications of cold atoms (see [1] and references therein). As mentioned, in the context of few-body and many-body physics, it is known for being the key mechanism for controlling and tuning of atomic interactions. In the context of more traditional atomic physics, it represents one of the most precise experimental measurements that can be carried out in the sub-thermal temperature regime and can be used to calibrate the understanding of atom-atom interactions (see, e.g., Refs. [54, 55, 56]). Being intrinsically a multichannel phenomenon that often involves a large number of channels, a thorough understanding of magnetic Feshbach resonances, or more generally the understanding of atomic interactions in a magnetic field, remains a complex task. Around an s wave resonance, the parametrization of the scattering length [57, 58, 1] al=0 (B) = abgl=0 ∆Bl=0 1− B − B0l=0 , (3.1) in terms of three parameters (the resonance position B0l=0 , the background scattering length abgl=0 , and the width parameter ∆Bl=0 ) has greatly simplified the description of ultracold atomic interaction in a magnetic field and greatly facilitated its applications in cold-atom physics [1]. For resonances in nonzero partial waves (e.g. Refs. [59, 3, 60, 61, 62, 63, 64, 65, 56]), however, similar parametrization has not existed until very recently [66], due to the fact that they cannot be described using the standard effect range theory [67, 68, 69]. Furthermore, even for an s wave resonance, the description 47 Figure 3-1: Picture of Two Channel Magnetic Feshbach Resonance: A simple two channel depiction of a Feshbach resonance. In this figure, the black channel is the open channel and the red channel is the closed channel. The closed channel supports a bound state right depicted by the red horizontal line. The Feshbach Resonance occurs when a bound state of the closed potential crosses the collisional energy. The collision energy for the ultracold regime is near the threshold of the open channel. The magnetic field comes into play by moving the position of the bound state relative the the threshold. This “tuning” of the closed bound state allows one to realize the Feshbach resonance. 48 in Eq. (3.1) is generally insufficient to fully characterize ultracold atomic interactions if the resonance is not of the “broad” type [70, 71, 58, 1, 66]. In Ref. [66], a uniform parametrization of magnetic Feshbach resonances in arbitrary partial waves is derived with analytical descriptions of ultracold atomic interactions around them. The theory is based on the MQDT of Ref. [72] and on the QDT expansion of Ref. [73]. It has been shown that a magnetic Feshbach resonance in an arbitrary partial wave l, and the ultracold atomic interactions around the resonance, can be fully characterized using a set of five parameters. They can either be the set c0 of B0l , Kbgl , gres , dBl , and sE (or C6 ), or the set of B0l , e abgl , ∆Bl , δµl and sE (or C6 ), with the second set being the more direct generalization of the s wave parametrization of Eq. (3.1) . Here e abgl is a generalized background scattering length, and is well defined for all l, δµl is a differential magnetic moment, and sE is the energy scale associated with the long-range van der Waals interaction, −C6 /R6 , which was discussed in the previous chapter. Used with the QDT expansion, or the generalized effective-range expansion derived from it [73, 66], they give accurate analytic descriptions of atomic interactions around a magnetic Feshbach resonance, not only of the scattering properties, but also of the binding energies of a Feshbach molecule and of scattering at negative energies [74]. The parametrization works the same for both broad and narrow resonances, and resonances of intermediate characteristics. The present work gives a more detailed presentation of MQDT for heteronuclear atomic interactions in a magnetic field [9, 66], with a focus on its application in the description of Feshbach resonances, especially in the determination of resonance parameters. Using the example of 6 Li40 K [65, 54, 9], it will be shown that MQDT provides an efficient and uniform description of a large number of resonances, in different partial waves and over a wide range of fields, from the same few parameters that characterize atomic interactions in the absence of the field [72]. All resonances 49 in aa, ab, and ba channels1 , in partial waves s (l = 0) through h (l = 5), and in fields of 0 G through 1000 G, are fully characterized using the parameters introduced in Ref. [66]. 3.2 MQDT for Atomic Interaction in a Magnetic Field In the subsequent sections, the MQDT formulation will be presented. the asymptotic channel structure will be defined and formalized incorporating the effects of an external magnetic field. The remainder of this section formulates the MQDT for atomic interaction accounting for the Zeeman splitting, hyperfine interaction and dispersion interaction. 3.2.1 Fragmentation Channels for Two Heteronuclear Group I Atoms in a Magnetic Field The following section will define and construct the channel structure used throughout. The channel structure will gradually be built up by first considering the case of a single atom in a magnetic field and then applying those results for the two atom channel description. 3.2.1.1 A Single Group I Atom in a Magnetic Field The Zeeman effect of a group I atom in a magnetic field B is well known [75]. The results will be summarized here with the intention of defining the notation. A group I atom, with an electronic angular momentum of JA = 1/2 and a nuclear 1 The channel labeling will be explained in the next section. 50 spin2 of IA in a magnetic field has 2(2IA + 1) states that are fully split. The states are labeled by |FA , mf i in a weak field, where FA = JA + IA , and by |JA , mJA ; IA , mIA i in a strong field. For intermediate fields, both labeling schemes can in principle still be used to uniquely identify a state, but are no longer ideal since quantum numbers other than mf = mJA + mIA are generally no longer good quantum numbers. The states will presently be identified using the notation of |(mf )α i, to emphasize that if there is interest in a wide range of magnetic fields, only mf is a good quantum number, whereas different FA states are generally coupled. Specifically, in the singleatom JA IA basis, with basis states |JA , mJA ; IA , mIA i to be denoted by a simpler notation of |mJA ; mIA i, the states of a group I atom in a magnetic field B are given, for |mf | = IA + 1/2, by |(IA + 1/2)0 i = |1/2; IA i , (3.2) |(−IA − 1/2)0 i = |−1/2; −IA i . (3.3) For mf = −IA + 1/2, · · · , IA − 1/2, there are two states corresponding to each mf . They are distinguished by α = 1, 2, and are given by |(mf )1 i = sin(θB /2) |1/2; mf − 1/2i − cos(θB /2) |−1/2; mf + 1/2i , (3.4) |(mf )2 i = cos(θB /2) |1/2; mf − 1/2i + sin(θB /2) |−1/2; mf + 1/2i . 2 (3.5) The nuclear spins have been well studied for many of the experimentally viable systems. All of the nuclear spins have been found in tables such as in [76]. 51 IA = 3 /2 1 .5 1 .0 g (+ 2 )0 (+ 1 )2 (0 )2 (-1 )2 h f 0 .5 0 .0 d E A / | ∆E h f A | e -0 .5 (-2 ) c b -1 .0 (-1 )1 (0 )1 (+ 1 )1 a -1 .5 0 .0 0 .5 0 1 .0 1 .5 2 .0 2 .5 |x | Figure 3-2: Energy Diagram of Group I Atom with IA = 3/2 in a magnetic field: Illustration of the energy diagram for a group I atom with IA = 3/2 in a magnetic field. It applies to 7 Li, 23 Na, 39 K, and 87 Rb, upon ignoring the small correction due to gI . Here x = (gJ −gI )µB B/∆EAhf is a scaled magnetic field. Both the alphabetic labeling and the more detailed labeling of states are illustrated. 52 IA = 1 1 .5 (+ 3 /2 ) 0 (+ 1 /2 ) 1 .0 (-1 /2 ) 2 2 0 .0 E A / | ∆E h f A | 0 .5 -0 .5 (-3 /2 ) 0 -1 .0 (-1 /2 ) (+ 1 /2 ) -1 .5 0 .0 0 .5 1 .0 1 .5 2 .0 1 1 2 .5 |x | Figure 3-3: Energy Diagram of Group I Atom with IA = 1 in a magnetic field: Illustration of the energy diagram for a group I atom with IA = 1 in a magnetic field, applicable to 6 Li. 53 IA = 4 1 .5 (-7 /2 )1 (-5 /2 )1 1 .0 (+ 5 /2 ) 1 (+ 7 /2 ) (+ 9 /2 ) 0 0 .0 E A / | ∆E h f A | 0 .5 1 -0 .5 (+ 7 /2 ) (+ 5 /2 ) 2 2 -1 .0 (-7 /2 )2 (-9 /2 )0 -1 .5 0 .0 0 .5 1 .0 1 .5 2 .0 2 .5 |x | Figure 3-4: Energy Diagram of Group I Atom with IA = 4 in a magnetic field: Illustration of the energy diagram for a group I atom with IA = 4 in a magnetic field, applicable to 40 K. 54 In these states, θB (mf ) takes values between 0 and π and is defined by r 1− tan θB (mf ) = 2mf 2IA +1 2mf 2IA +1 2 , +x (3.6) the quantity x is defined by x = (gJ − gI )µB B/∆EAhf , in which µB is the Bohr magneton and gJ and gI are the gyromagnetic ratios for the electron and the nucleus, respectively [77]. The hyperfine splitting ∆EAhf is defined as ∆EAhf ≡ EA (FA = IA + 1/2) − E(FA = IA − 1/2) in the absence of a magnetic field. It is positive for most alkali-metal atoms, but negative for 40 K. In defining these states, a phase convention is chosen such that B→0 |(mf )1 i ∼ |FA = IA − 1/2, mf i B→0 |(mf )2 i ∼ |FA = IA + 1/2, mf i . The energies of the states |(mf )α i are given, for |mf | = IA + 1/2, by EA ((IA + 1/2)0 ) = ∆EAhf EA ((−IA − 1/2)0 ) = ∆EAhf 1 gJ + 2IA gI IA + x , 2IA + 1 2 gJ − gI 1 gJ + 2IA gI IA − x , 2IA + 1 2 gJ − gI (3.7) (3.8) and for other states by [75] 1 1 2mf gI hf EA ((mf )1 ) = ∆EA − + x 2 2IA + 1 gJ − gI 1/2 # 4mf − 1+ x + x2 , 2IA + 1 55 (3.9) 1 1 2mf gI hf EA ((mf )2 ) = ∆EA − + x 2 2IA + 1 gJ − gI 1/2 # 4mf . x + x2 + 1+ 2IA + 1 (3.10) For ∆EAhf > 0, the state (mf )2 has higher energy than (mf )1 . The reverse is true for ∆EAhf < 0 The other common labeling of states of a group I atom in a magnetic field is alphabetic ordering, a, b, c, . . . , in the order of increasing energy [1]. This labeling is very concise and can be convenient, but is, at times, not sufficiently informative. Both sets of notation are utilized for convenience and for connection with earlier work, but I will defer to the |(mf )α i notation when necessary. Figure 3-2 illustrates the energy diagram for alkali-metal atoms 7 Li, 23 Na, 39 K, and 87 Rb, all with IA = 3/2. It shows that for ∆EAhf > 0, the lowest atomic state, the a state, is always |(IA − 1/2)1 i, and the b state is always |(IA − 3/2)1 i. Figure 3-3 depicts the energy diagram for a 6 Li atom, which has an IA = 1. Figure 3-4 depicts the energy diagram for a 40 K atom, which has an IA = 4. It illustrates that for ∆EAhf < 0, the lowest atomic state, the a state, is always |(−IA − 1/2)0 i, and the b state is always |(−IA + 1/2)2 i. Equations (3.2)-(3.5) define a unitary transformation UAB , with elements hmJA ; mIA |(mf )α i, that relates the single atom JA IA basis to the single-atom states in a magnetic field, which diagonalizes both the hyperfine and the magnetic interactions. It is a one dimensional unit matrix for |mf | = IA + 1/2, and is given by B B sin(θ /2) − cos(θ /2) UAB = , B B cos(θ /2) sin(θ /2) for mf = −IA + 1/2, · · · , IA − 1/2. 56 (3.11) 3.2.1.2 Two Heteronuclear Group I Atoms in a Magnetic Field The fragmentation channels for atomic interactions in a magnetic field are determined by two-atom states in a magnetic field. For two heteronuclear alkali-metal atoms, labeled 1 and 2, in a magnetic field, there are 2(2I1 + 1) × 2(2I2 + 1) number of channels for each partial wave l. They can each be identified by |(mf 1 )α1 , (mf 2 )α2 , li, with |(mf x )αx i characterizing the state of atom x in a magnetic field in a way as specified in Sec. 3.2.1.1. The channel threshold energies are given by E1 ((mf 1 )α1 ) + E2 ((mf 2 )α2 ). Ignoring the weak magnetic dipole-dipole [78, 57] and second-order spin-orbit interactions [79, 80, 81], MF = mf 1 + mf 2 is conserved. The interaction and the scattering matrices are block-diagonal with each block labeled by MF . The number of (coupled) channels in each block is independent of l, and is determined by |MF |. For instance, MF = ±(I1 + I2 + 1) blocks both have only a single channel, MF = ±(I1 + I2 ) blocks both have 4 channels, and blocks with |I1 − I2 | < |MF | < I1 + I2 have 4(I1 + I2 + 1 − |MF |) channels, etc. The total number of channels for each l adds up to 2(2I1 + 1) × 2(2I2 + 1). 3.2.2 MQDT MQDT for atomic interactions in a magnetic field is formally the same as MQDT for atomic interactions in the absence of external fields. The theory takes full advantage of the physics that both the energy dependence [82] and the partial wave dependence [83] of the atomic interaction around a threshold are dominated by effects of the long-range potential, which are described by a set of universal QDT functions [74]. The short-range contribution is isolated to a short-range K c matrix that is insensitive to both the energy and the partial wave. The details of the short range K c matrix will be explained in the next subsection. 57 For an N -channel problem and at energies where all channels are open, MQDT gives the physical K matrix, in the present case the K MF l , as [72] c c K MF l = −(Zfc c − Zgc K c )(Zfc s − Zgs K c )−1 , (3.12) c c c (si , l) being the Zxy s are N × N diagonal matrices with elements Zxy where Zxy functions [74] evaluated at scaled energies si = (E − Ei )/sE relative to the respective channel threshold Ei . Here sE = (~2 /2µ)(1/β6 )2 and β6 = (2µC6 /~2 )1/4 are the characteristic energy and the length scales, respectively, associated with the −C6 /2R6 van der Waals potential, with µ being the reduced mass. At energies where No channels are open, and Nc = N − No channels are closed, MQDT gives [72] c c c c −1 K MF l = −(Zfc c − Zgc Keff )(Zfc s − Zgs Keff ) , (3.13) c c c c −1 c Keff = Koo + Koc (χc − Kcc ) Kco , (3.14) where c c in which χc is a Nc × Nc diagonal matrix with elements χcl (si , l) [74], and Koo , Koc , c c , are submatrices of K c corresponding to open-open, open-closed, closed, and Kcc Kco open, and closed-closed channels, respectively. Equation (3.13) is formally the same c as Eq. (3.12), except that the K c matrix is replaced by Keff that accounts for the effects of closed channels. From the physical K matrix, the physical S matrix is obtained from [40] S MF l = (1 + iK MF l )(1 − iK MF l )−1 , (3.15) from which all the scattering properties can be determined. At energies where all channels are closed, the bound spectrum can be determined either from det[χc (E) − K c ] = 0 , 58 (3.16) where K c is the full N × N K c matrix, or from equivalent effective single-channel or effective multichannel problems, as discussed in Ref. [66]. The presence of a magnetic field changes the channel structure for atomic interactions, including both the channel wave functions and the channel threshold energies. It is through these changes that atomic interactions depend on the magnetic field. The change of channel threshold energies is discussed in Sec. 3.2.1. The change of channel wave functions leads to a field-dependent K c matrix. This aspect is discussed in the following subsection. 3.2.3 The K c Matrix for Heteronuclear Group I Systems The short-range K c matrix for two heteronuclear group I atoms in a magnetic field can be obtained through a frame transformation from the two single-channel K c : KSc (, l) for spin singlet X 1 Σ state, and KTc (, l) for spin triplet A3 Σ state, which are the same parameters describing atomic interactions in the absence of the magnetic field [72]. Specifically, the K c matrix, being a short-range K matrix, has much simpler representation in condensation channels that diagonalize the short-range interactions [40, 72]. For interactions of group I atoms, a convenient choice of condensation channels is the JI coupled basis, in which a state is labeled by |(JMJ , IMI )li where J = J1 + J2 and I = I1 + I2 . The short-range K c matrix is to an excellent approximation diagonal in this basis with elements given by h(J 0 MJ0 , I 0 MI0 )l|K c(JI) |(JMJ , IMI )li = KJc (, l)δJ 0 J δMJ0 MJ δI 0 I δMI0 MI , (3.17) c c where KJ=0 corresponds to KSc , and KJ=1 corresponds to KTc [72]. From K c(JI) , the K c matrix in fragmentation channels in a magnetic field, as 59 defined in Sec. 3.2.1, is given by K c (B) = U B† K c(JI) U B , (3.18) U B = U JI (U1B ⊗ U2B ) . (3.19) where Here U1B and U2B are unitary transformations, discussed earlier, describing the recoupling of atom 1 states and atom 2 states by the magnetic field, respectively. U JI is a unitary transformation with elements, hJMJ |J1 mJ1 , J2 mJ2 ihIMI |I1 mI1 , I2 mI2 i, given by the Clebsch-Gordon coefficients describing the coupling of J = J1 + J2 and I = I1 + I2 . The applicability of frame transformation depends critically on K c being a shortrange matrix. It is not applicable to the physical K matrix except at energies much greater than the hyperfine and the magnetic splittings. The K c matrix, when regarded as an operator associated with short-range atomic interactions, is dominated by electrostatic and exchange interactions, is to an excellent approximation not changed by external fields and other weak interactions such as hyperfine interaction. External fields and weak interactions only change the basis in which K c needs to be represented, not the operator itself. From the unitarity of U B , one can show rigorously that all off-diagonal elements of K c (B) are proportional to ∆K c ≡ KSc − KTc . It is a general feature applicable both in and in the absence of the magnetic field [72]. It implies that the coupling of different fragmentation channels, which is responsible both for the existence of Feshbach resonances and for inelastic scattering, is due to the difference between the singlet and triplet scattering, as reflected, e.g., in the difference in s wave scattering lengths ∆al=0 ≡ aSl=0 − aTl=0 , where aSl=0 and aTl=0 are the s wave scattering lengths for the singlet and the triplet state, respectively. 60 3.3 MQDT for Magnetic Feshbach Resonances The MQDT formulation of the previous section provides a systematic understanding of atomic interactions in a magnetic field, including both elastic and inelastic processes, and over a wide range of energies and fields. It works the same for all heteronuclear group I atoms, and uses the same parameters, KSc (, l) and KTc (, l), that one would use for interactions in the absence of the field [72]. The 6 Li40 K system will be used to illustrate the theory. To set the benchmark for future investigations and to keep the focus on concepts, the results will be limited to baseline MQDT results that ignore the energy and the partial-wave dependencies of the short-range parameters, namely in the approximation of KSc (, l) ≈ KSc ( = 0, l = 0) and KTc (, l) ≈ KTc ( = 0, l = 0). In this baseline description, all aspects of cold atomic interactions, including parameters for all magnetic Feshbach resonances in all partial waves, are determined from three parameters [72, 9]: the C6 coefficient, the singlet s wave scattering length aSl=0 , and the triplet s wave scattering length aTl=0 , in addition to well known atomic parameters such as the atomic mass and hyperfine splitting. 6 Li40 K is an interesting quantum system of two fermionic atoms. The magnetic Feshbach resonances have been investigated experimentally by Wille et al. [65], and theoretically by Hanna et al. [9]. 6 Li has I1 = 1 with a hyperfine splitting of ∆E1hf /h = 228.205 MHz and nuclear g-factor3 of gI1 = −4.477 × 10−4 [77]. 40 K has I2 = 4 with a hyperfine splitting of ∆E2hf /h = −1285.79 MHz and gI2 = 1.765 × 10−4 [77]. The three parameters used in the present baseline MQDT description are C6 = 2322 a.u. [84, 54], aSl=0 = 52.50 a.u., and aTl=0 = 63.70 a.u. The singlet and the triplet s wave scattering lengths are slightly adjusted from the values of Hanna et al. [9] for better agreement with experimental s wave Feshbach resonance positions at low 3 The nuclear g-factors are on the order of melectron /mproton times smaller than the electron g-factor. 61 fields. In terms of quantum defects, the adopted scattering lengths correspond to µcS ( = 0, l = 0) = 0.3938 for the singlet state and µcT ( = 0, l = 0) = 3202 for the triplet state. The s wave scattering lengths give KSc (, l) ≈ KSc ( = 0, l = 0) = −16.91 and KTc (, l) ≈ KTc ( = 0, l = 0) = 5.748 [72, 74], from which the K c matrix in a magnetic field is constructed as discussed in Sec. 3.2.3. From the K c matrix, the MQDT of Sec. 3.2.2 provides a complete description of cold atomic interactions in a B field. The C6 coefficient gives a length scale of β6 = 81.56 a.u. and a corresponding energy scale of sE /kB = 2.490 mK for 6 Li-40 K van der Waals interactions. In understanding magnetic Feshbach resonances, the focus is on atomic interactions near the lowest threshold for each MF . For all energies below the second lowest threshold, both scattering and bound states can be described by an effective single-channel problem with [66] c c c c −1 c Keff = K11 + K1c (χc − Kcc ) Kc1 , (3.20) or with its corresponding Klc0 matrix [66] Klc0 (, B) = c Keff (, l) − tan(πν0 /2) . c 1 + tan(πν0 /2)Keff (, l) (3.21) In Eq. (3.20), “1” refers to the lowest-energy channel for the MF manifold, and “c” refers to all other (closed) channels. In Eq. (3.21), ν0 = (2l + 1)/4, and = E − E1 is the energy relative to the lowest threshold. c In terms of Keff , scattering below the second threshold, where only elastic scat- tering is possible, is described by the physical K matrix c c c c −1 ) , tan δl = (Zgc Keff − Zfc c )(Zfc s − Zgs Keff 62 (3.22) from which the elastic scattering cross section can be calculated in the standard manner. In terms of the value of Klc0 (, B) at zero energy (namely at the lowest threshold), defined by Klc0 (B) ≡ Klc0 ( = 0, B) , (3.23) the resonance locations, B0l , of zero-energy magnetic Feshbach resonances in any partial wave l, can be determined by the roots of Klc0 (B), namely as the solutions of [66] Klc0 (B) = 0 . (3.24) The generalized scattering length as a function of magnetic field, for an arbitrary partial wave l, is given by [66] e al (B) = āl 1 (−1) + c0 Kl (B) l , (3.25) where āl = āsl β62l+1 is the mean scattering length (with scale included) with āsl = π2 , 24l+1 [Γ(l/2 + 1/4)Γ(l + 3/2)]2 being the scaled mean scattering length for partial wave l [73]. 63 (3.26) Table 3.1: Comparison of Select Feshbach Parameters for 6 Li40 K: Selective magnetic Feshbach resonances for the 6 Li-40 K system that have been found experimentally [65], and investigated in an earlier MQDT theory [9]. Before the work of Ref. [66], the width parameter ∆Bl was not rigorously defined for nonzero partial waves, and was therefore not provided in the previous theory [9] for p wave resonances. Identification Channel Mf ba -5 aa -4 aa -4 aa -4 ab -3 ab -3 ab -3 3.4 l 0 0 0 1 0 0 1 Present MQDT B0l (G) ∆Bl (G) 215.0 0.245 157.6 0.141 167.6 0.107 247.1 0.447 149.7 0.242 158.8 0.425 259.1 0.692 Previous MQDT [9] B0l (G) ∆Bl (G) 213.6 0.28 159.3 0.22 170.1 0.07 258.0 153.1 0.42 159.6 0.16 260.8 - Experiment [65] B0l (G) 215.6 157.6 168.2 249.0 149.2 159.5 263.0 Results for the 6Li40K System This MQDT-based formalism provides a unified and a uniform understanding of magnetic Feshbach resonances in all partial waves, as illustrated in Figs. 3-5-3-7 for 6 Li-40 K. Figure 3-5 depicts the reduced generalized scattering lengths, e al (B)/āl , versus magnetic field for 6 Li40 K in the ba channel, for all partial waves s (l = 0) through h (l = 5). Here ba refers to 6 Li in state b, namely the |(−1/2)1 i state (c.f. Fig. 3-3), and 40 K in state a, namely the |(−9/2)0 i state (c.f. Fig. 3-4). It is the lowest energy channel of the MF = −5 manifold with 4 coupled channels. Note that by plotting the dimensionless reduced generalized scattering lengths, the understanding of different partial waves are put on the same footing and can be depicted in the same figure. The plots are grouped into an even parity group and an odd parity group, as they are not coupled and can in principle be experimentally probed independently. Figures 3-6 and 3-7 show similar results for the aa channel (MF = −4 manifold with 8 coupled channels) and the ab channel (MF = −3 manifold with 11 coupled channels), 64 Table 3.2: Table of Feshbach Parameters for LiK System: A complete list of magnetic Feshbach resonances and their parameters for 6 Li40 K system in ba, aa, and ab channels, in partial wave s (l = 0) through h (l = 5), and in magnetic fields from 0 through 1010 G. A large number of resonances are predicted in the h wave (l = 5). The columns B0l , ∆B and dBl are all in Gauss while the other columns are unitless. Mf -5 -5 -4 -4 -4 -4 -4 -4 -4 -4 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 -3 l 0 5 0 0 1 5 5 5 5 5 0 0 0 1 1 5 5 5 5 5 5 B0l 215.0 206.7 157.6 167.6 247.1 18.22 184.3 201.1 981.7 999.8 149.7 158.8 165.2 7.721 259.1 1.265 22.43 211.7 228.1 237.1 1009 ∆B e abgl /āsl 0.245 1.631 2.189 3.509 0.141 1.648 0.107 1.608 0.447 3.415 0.506 3.477 0.053 3.705 1.100 3.468 0.028 3.578 0.596 3.458 0.242 1.697 0.425 1.579 0.004 1.490 0.092 1.557 0.692 3.284 0.0004 3.551 0.670 3.466 0.088 3.942 1.155 4.100 1.593 3.051 1.103 4.384 δµl /µB 1.776 1.769 1.660 1.800 0.152 2.854 1.678 1.830 1.890 1.956 1.588 1.758 1.815 1.005 0.163 2.308 2.372 1.510 1.672 1.760 1.901 65 c0 Kbgl 1.586 0.222 1.543 1.644 0.227 0.223 0.213 0.224 0.218 0.224 1.435 1.726 2.043 0.391 0.233 0.220 0.224 0.202 0.196 0.247 0.186 gres -0.03036 -0.08129 -0.01605 -0.01379 -0.00142 -0.03027 -0.00190 -0.04213 -0.00111 -0.02440 -0.02528 -0.05489 -0.00059 -0.00152 -0.00234 -0.00002 -0.03329 -0.00285 -0.04189 -0.05696 -0.04604 ζres dBl 0.006383 -0.6336 -0.003133 -1.703 0.003467 -0.3583 0.002795 -0.2839 -0.001255 -0.3460 -0.001158 -0.3932 -0.000077 -0.0420 -0.001609 -0.8535 -0.000043 -0.0217 -0.000930 -0.4624 0.005874 -0.5901 0.01060 -1.157 0.000096 -0.0121 -0.000776 -0.05597 -0.002008 -0.5308 -0.000001 -0.00034 -0.001271 -0.5203 -0.000121 -0.07011 -0.001826 -0.9286 -0.001972 -1.199 -0.002118 -0.8978 respectively. The resonance positions, determined from Eq. (3.24), are part of the parameters tabulated in Tables 3.1 and 3.2. Table 3.1 provides a subset of the present results for magnetic Feshbach resonances that have been measured experimentally by Wille et al. [65] and investigated theoretically by Hanna et al. [9]. The resonance positions, which are the only parameters measured, are found to be consistent with the previous theory and in good agreement with experiment. Around each resonance, the dependence of the generalized scattering length on the magnetic field can be characterized by [66] e al (B) = e abgl 1 − ∆Bl B − B0l , (3.27) where e abgl is the generalized background scattering length, and ∆Bl is one measure of the width of the resonance. They can be computed from e al (B = B0l + ∆Bl ) = 0 , (3.28) 1 e abgl = −e al (B = B0l + ∆Bl ) , 2 (3.29) and for an isolated resonance, and can be further refined, when necessary, by fitting e al (B) to Eq. (3.27) at B fields sufficiently close to B0l . Except for the case of a “broad” resonance, the set of parameters B0l , e abgl , and ∆Bl are generally still insufficient to characterize atomic interactions away from zero energy [70, 71, 58, 1, 66]. For a complete characterization of ultracold atomic interactions around a magnetic Feshbach resonance, another parameter is generally needed, the differential magnetic moment, δµl , which is the difference of the magnetic mo- 66 ments in interacting and non-interacting states [1, 66]. It is given by dl (B) δµl = , dB B=B0l (3.30) and can be found from solutions of Klc0 (, B) = 0 at energies slightly different from zero. The parameters, B0l , e abgl , ∆Bl , δµl , along with the C6 coefficient, form one set of parameters that provide a complete description of atomic interaction around a magnetic Feshbach resonance. It is the most direct generalization of the s wave characterization [1] to other partial waves, and is convenient at zero energy (the threshold). The parameters for 6 Li40 K in ba, aa, and ab channels are provided in Table 3.2 for all Feshbach resonances that have been found in partial wave s through h and in the field range of 0 G through 1009 G4 . For descriptions of atomic interactions at energies away from the threshold, it is c0 more convenient to use a different set of parameters, B0l , Kbgl , gres , dBl , and sE (or C6 ). They differ from the first set in three parameters that are related by [66] c0 Kbgl = gres 1 , e abgl /āl − (−1)l e abgl /āl =− e abgl /āl − (−1)l dBl = − δµl ∆Bl sE e abgl /āl ∆Bl . e abgl /āl − (−1)l (3.31) , (3.32) (3.33) With this set of parameters, which are also provided in Table 3.2, the Klc0 parameter 4 The reason for the unusual magnetic field range is due to a resonance for the Mf = 4 channel at 999.8 G with a width of 0.596 G. A larger range was needed to examine this particular Feshbach resonance. 67 is given by [66] Klc0 (s , Bs ) = c0 Kbgl gres 1+ s − gres (Bs + 1) , (3.34) around a Feshbach resonance, where Bs = (B −B0l )/dBl . This expression of Klc0 , as a function of both energy and magnetic field, goes into the QDT expansion of Ref. [66] to give an analytical description of atomic interactions in an arbitrary partial wave l and in a B field around B0l , not only at the threshold but also in a range of energies away from the threshold. Figure 3-8 depicts an example of the p wave phase shift, more precisely sin2 δl=1 , around a p wave resonance in the ab channel located around B0l ≈ 259 G. It shows that the results obtained using parameters of Table 3.2 and the analytical QDT expansion of Ref. [66] are in good agreement with those computed numerically from Eq. (3.22). Also tabulated in Table 3.2 is a derived parameter ζres ≡ gres , c0 (2l + 3)(2l − 1)Kbgl (3.35) which distinguishes a “broad” resonance (ζres 1) that follows the single-channel universal behavior from a “narrow” resonance (ζres 1) that deviates strongly from such behavior [66]. All resonances for 6 Li40 K are found to be narrow, a result that is closely related to the fact that ∆al=0 ≡ aSl=0 − aTl=0 , and thus ∆K c is small for 6 Li40 K. 3.5 Chapter Summary In conclusion, a multichannel quantum defect theory is presented for the interaction of heteronuclear group I atoms in a magnetic field, and applied to develop a theory for magnetic Feshbach resonances in such systems. The theory provides a systematic and a uniform description of resonances in all partial waves, and enables the descrip68 tion of a large number of resonances in terms of very few parameters. The simplified description and the parameters provided by the theory will facilitate further theoretical investigations of two-body, few-body, and many-body systems around magnetic Feshbach resonances, especially those in nonzero partial waves. These parametrizations were explicitly used in the presented 6 Li40 K system. In the future, it would be interesting to explore other heteronuclear alkali-metal systems to look for broad resonances in nonzero partial waves. 69 Reduced Generalized Scattering Length 4 Even Parity MF = -5 l=0 l=2 l=4 3 2 1 0 Odd Parity 6 l=1 l=3 l=5 4 2 0 -2 0 100 200 300 400 Magnetic Field (G) Figure 3-5: Reduced generalized scattering lengths for LiK in the MF = −3 Manifold: Reduced generalized scattering lengths, e al (B)/āl , versus magnetic field for 6 Li40 K in the ba channel (corresponding to the manifold of MF = −5), for partial waves s through h. 70 Reduced Generalized Scattering Length 4 Even Parity MF = -4 l=0 l=2 l=4 3 2 1 0 Odd Parity 6 l=1 l=3 l=5 4 2 0 -2 0 100 200 300 400 Magnetic Field (G) Figure 3-6: Reduced generalized scattering lengths for LiK in the MF = −4 Manifold: Reduced generalized scattering lengths, e al (B)/āl , versus magnetic field for 6 Li40 K in the aa channel (corresponding to the manifold of MF = −4), for partial waves s through h. 71 Reduced Generalized Scattering Length 4 Even Parity MF = -3 l=0 l=2 l=4 3 2 1 0 Odd Parity 6 l=1 l=3 l=5 4 2 0 -2 0 100 200 300 400 Magnetic Field (G) Figure 3-7: Reduced generalized scattering lengths for LiK in the MF = −3 Manifold: Reduced generalized scattering lengths, e al (B)/āl , versus magnetic field for 6 Li40 K in the ab channel (corresponding to the manifold of MF = −3), for partial waves s through h. 72 1 ab Channel B = 260.0 G p wave 0.6 2 sin δl=1 0.8 0.4 0.2 0 -3 4.020×10 -3 4.022×10 -3 4.024×10 -3 4.026×10 -3 4.028×10 εs Figure 3-8: LiK ab channel plot of sin2 δl for p wave Resonance: Plot of sin2 δl for p wave interaction in the ab channel as a function of scaled energy s at a magnetic field of 260.0 G. The solid line is the result of a numerical calculation from Eq. (3.22). The symbols are results from the analytical QDT expansion [66] using the parameters specified in Table 3.2 for the p wave resonance located at B0l ≈ 259 G. 73 Chapter 4 Concluding Remarks In the present thesis, two aspects of atomic and molecular investigations were conducted. Aspects of electronic structure calculations were discussed in chapter 2 and in the following chapter, MQDT of magnetic Feshbach Resonances were presented. In the chapter on ab initio electronic structure calculations a few topics were presented. A full PES for the ground state of lithium-ytterbium-lithium was constructed in an effort to explore ultracold reactive scattering. A calculation of the reaction rate is currently being carried out for the exchange reaction LiYb + Li → Li2 + Yb. Pending the results of this calculation, we hope to be able to address the applicability of less computationally intensive theories in the description of the reaction dynamics. In addition, control over this reaction would be of experimental importance. In the creation of ultracold molecules, there are often background atoms not used to form molecules. It would be of great interest to minimize the effects of these background atoms, which in most cases results in the loss of molecules from the trap. Dynamic polarizabilities were also calculated for a variety of atoms and molecules. The results of these fitting procedures are contained in appendix A. The last section used the new information on polarizabilities in addition to well known polarizabilities of ground state atoms to derive a number of dispersion coefficients for a variety of atom-atom, atom-molecule and molecule-molecule systems. These calculations, which 74 include anisotropic dispersion coefficients, provide a landscape for both numerical and analytic theories. A multichannel quantum defect theory was also presented for atoms in an external magnetic field. This enhanced understanding is manifested in the new description of Feshbach resonances for arbitrary partial waves. While the present work focused on the 6 Li40 K system, the presented material is systematic and aside from differences in a few number of parameters would apply to all alkali systems. In the future, our multichannel quantum defect theory can be used to better constrain some of the input parameters such as the dispersion coefficient and the two s-wave scattering lengths. In principle the resonance locations, and possibly the widths, are dependent on these parameters. Direct experimental measurement of a few resonance locations can be used with a fitting algorithm to fit the input parameters. This is one potential direction to apply our multichannel quantum defect theory. Another interesting direction would be to examine systems with energy and/or partial wave dependent s-wave scattering lengths. In our current formulation, the s-wave scattering lengths are assumed be energy and partial wave independent. In our test system, this assumption works well as all of the Feshbach resonances can be reproduced using our theory. This may not be the case for other systems, especially for some of these mass imbalanced systems. Addressing the energy and partial wave dependence can potentially open the door to higher temperature investigations. 75 References [1] C. Chin, R. Grimm, P. Julienne, and E. Tiesinga, Rev. Mod. Phys. 82, 1225 (2010). [2] S. Inouye, M. R. Andrews, J. Stenger, H.-J. Miesner, D. M. Stamper-Kurn, and W. Ketterle, Nature 392, 151 (1998). [3] A. Marte, T. Volz, J. Schuster, S. Dürr, G. Rempe, E. G. M. van Kempen, and B. Verhaar, Phys. Rev. Lett. 89, 283202 (2002). [4] J. Zhang, E. G. M. van Kempen, T. Bourdel, L. Khaykovich, J. Cubizolles, F. Chevy, M. Teichmann, L. Tarruell, S. J. J. M. F. Kokkelmans, and C. Salomon, Phys. Rev. A 70, 030702 (2004), URL http://link.aps.org/doi/10.1103/ PhysRevA.70.030702. [5] C. A. Regal, C. Ticknor, J. L. Bohn, and D. S. Jin, Phys. Rev. Lett. 90, 053201 (2003), URL http://link.aps.org/doi/10.1103/PhysRevLett.90.053201. [6] S. Stellmer, M. K. Tey, B. Huang, R. Grimm, and F. Schreck, Phys. Rev. Lett. 103, 200401 (2009), URL http://link.aps.org/doi/10.1103/PhysRevLett. 103.200401. [7] K. Enomoto, M. Kitagawa, K. Kasa, S. Tojo, and Y. Takahashi, Phys. Rev. Lett. 98, 203201 (2007), URL http://link.aps.org/doi/10.1103/PhysRevLett. 98.203201. [8] M. Kitagawa, K. Enomoto, K. Kasa, Y. Takahashi, R. Ciurylo, P. Naidon, and 76 P. S. Julienne, Phys. Rev. A 77, 012719 (2008), URL http://link.aps.org/ doi/10.1103/PhysRevA.77.012719. [9] T. M. Hanna, E. Tiesinga, and P. S. Julienne, Phys. Rev. A 79, 040701 (2009). [10] S. Kotochigova and D. DeMille, Phys. Rev. A 82, 063421 (2010), URL http: //link.aps.org/doi/10.1103/PhysRevA.82.063421. [11] A. Widera, O. Mandel, M. Greiner, S. Kreim, T. W. Hänsch, and I. Bloch, Phys. Rev. Lett. 92, 160406 (2004). [12] K.-K. Ni, S. Ospelkaus, M. H. G. de Miranda, A. Pe’er, B. Neyenhuis, J. J. Zirbel, S. Kotochigova, P. S. Julienne, D. S. Jin, and J. Ye, Science 322, 231 (2008), http://www.sciencemag.org/content/322/5899/231.full.pdf, URL http://www.sciencemag.org/content/322/5899/231.abstract. [13] T. Takekoshi, L. Reichsöllner, A. Schindewolf, J. M. Hutson, C. R. Le Sueur, O. Dulieu, F. Ferlaino, R. Grimm, and H.-C. Nägerl, ArXiv e-prints (2014), 1405.6037. [14] T. Gonzlez-Lezana, International Reviews in Physical Chemistry 26, 29 (2007), http://dx.doi.org/10.1080/03081070600933476, URL http://dx. doi.org/10.1080/03081070600933476. [15] A. H. Hansen, A. Khramov, W. H. Dowd, A. O. Jamison, V. V. Ivanov, and S. Gupta, Phys. Rev. A 84, 011606 (2011), URL http://link.aps.org/doi/ 10.1103/PhysRevA.84.011606. [16] S. Kotochigova, New Journal of Physics 12, 073041 (2010), URL http:// stacks.iop.org/1367-2630/12/i=7/a=073041. 77 [17] D. R. Hartree, Mathematical Proceedings of the Cambridge Philosophical Society 24, 426 (1928), ISSN 1469-8064, URL http://journals.cambridge.org/ article_S0305004100015954. [18] V. Fock, Zeitschrift für Physik 61, 126 (1930), ISSN 0044-3328, URL http: //dx.doi.org/10.1007/BF01340294. [19] C. Møller and M. S. Plesset, Phys. Rev. 46, 618 (1934), URL http://link. aps.org/doi/10.1103/PhysRev.46.618. [20] J. Čı́žek, The Journal of Chemical Physics 45, 4256 (1966), URL http:// scitation.aip.org/content/aip/journal/jcp/45/11/10.1063/1.1727484. [21] C. F. Fischer, J. B. Lagowski, and S. H. Vosko, Phys. Rev. Lett. 59, 2263 (1987), URL http://link.aps.org/doi/10.1103/PhysRevLett.59.2263. [22] A. Petrov, C. Makrides, and S. Kotochigova, Molecular Physics 111, 1731 (2013), http://www.tandfonline.com/doi/pdf/10.1080/00268976.2013.777812, URL http://www.tandfonline.com/doi/abs/10.1080/00268976.2013. 777812. [23] R. T. Pack and G. A. Parker, The Journal of Chemical Physics 87, 3888 (1987), URL http://scitation.aip.org/content/aip/journal/jcp/87/7/ 10.1063/1.452944. [24] J. Stanton, J. Gauss, M. Harding, P. Szalay, A. Auer, R. Bartlett, U. Benedikt, C. Berger, D. Bernholdt, Y. Bomble, et al., CFOUR, a quantum chemical program package (2014), for the current version, see http://www.cfour.de, URL http://www.cfour.de. [25] M. Wernli, E. Scifoni, E. Bodo, and F. Gianturco, International Journal of Mass Spectrometry 280, 57 (2009), ISSN 1387-3806, ion-molecule and ion-surface 78 interactions: A special issue honoring Zdenek Herman on the occasion of his 75th birthday, URL http://www.sciencedirect.com/science/article/pii/ S1387380608002819. [26] A. Aguado and M. Paniagua, The Journal of Chemical Physics 96, 1265 (1992), URL http://scitation.aip.org/content/aip/journal/jcp/96/2/ 10.1063/1.462163. [27] B. P. Prascher, D. E. Woon, K. A. Peterson, J. Dunning, Thom H., and A. K. Wilson, Theoretical Chemistry Accounts 128, 69 (2011), ISSN 1432-881X, URL http://dx.doi.org/10.1007/s00214-010-0764-0. [28] X. Cao and M. Dolg, The Journal of Chemical Physics 115, 7348 (2001), URL http://scitation.aip.org/content/aip/journal/jcp/115/ 16/10.1063/1.1406535. [29] K. Raghavachari, G. W. Trucks, J. A. Pople, and M. Head-Gordon, Chemical Physics Letters 589, 37 (2013), ISSN 0009-2614, {SPECIAL} {NOBEL} {ISSUE} Groundbreaking articles from Nobel Laureates published in Chemical Physics Letters, URL http://www.sciencedirect.com/science/article/ pii/S0009261413010749. [30] B. Gao, Phys. Rev. A 83, 062712 (2011). [31] A. Kramida, Y. Ralchenko, , and J. Reader, Nist atomic spectra database (2014), URL http://www.nist.gov/pml/data/asd.cfm. [32] A. Stone, The Theory of Intermolecular Forces, International Series of Monographs on Chemistry (Oxford University Press, 1997), ISBN 9780198558835. [33] W. G. Rellergert, S. T. Sullivan, S. Kotochigova, A. Petrov, K. Chen, S. J. 79 Schowalter, and E. R. Hudson, Phys. Rev. Lett. 107, 243201 (2011), URL http: //link.aps.org/doi/10.1103/PhysRevLett.107.243201. [34] H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby, M. Schütz, P. Celani, T. Korona, R. Lindh, A. Mitrushenkov, G. Rauhut, et al., Molpro, version 2012.1, a package of ab initio programs (2012), see http://www.molpro.net. [35] M. Kállay and J. Gauss, Journal of Molecular Structure: {THEOCHEM} 768, 71 (2006), ISSN 0166-1280, URL http://www.sciencedirect.com/science/ article/pii/S0166128006002727. [36] S. Kotochigova, A. Petrov, M. Linnik, J. Kos, and P. S. Julienne, The Journal of Chemical Physics 135, 164108 (2011), URL http://scitation.aip.org/ content/aip/journal/jcp/135/16/10.1063/1.3653974. [37] A. Gerdes, M. Hobein, H. Knckel, and E. Tiemann, The European Physical Journal D 49, 67 (2008), ISSN 1434-6060, URL http://dx.doi.org/10.1140/ epjd/e2008-00138-7. [38] J. Deiglmayr, M. Aymar, R. Wester, M. Weidemller, and O. Dulieu, The Journal of Chemical Physics 129, 064309 (2008), URL http://scitation.aip.org/ content/aip/journal/jcp/129/6/10.1063/1.2960624. [39] F. London, Trans. Faraday Soc. 33, 8b (1937), URL http://dx.doi.org/10. 1039/TF937330008B. [40] B. Gao, Phys. Rev. A 54, 2022 (1996). [41] J. M. Sage, S. Sainis, T. Bergeman, and D. DeMille, Phys. Rev. Lett. 94, 203001 (2005), URL http://link.aps.org/doi/10.1103/PhysRevLett.94.203001. 80 [42] A. Khramov, A. Hansen, W. Dowd, R. J. Roy, C. Makrides, A. Petrov, S. Kotochigova, and S. Gupta, Phys. Rev. Lett. 112, 033201 (2014), URL http: //link.aps.org/doi/10.1103/PhysRevLett.112.033201. [43] A. Petrov, E. Tiesinga, and S. Kotochigova, Phys. Rev. Lett. 109, 103002 (2012), URL http://link.aps.org/doi/10.1103/PhysRevLett.109.103002. [44] A. Stone and R. Tough, Chemical Physics Letters 110, 123 (1984), ISSN 0009-2614, URL http://www.sciencedirect.com/science/article/ pii/0009261484801608. [45] J. D. Jackson, Classical Electrodynamics (John Wiley & Sons, New York, 1999). [46] A. D. McLachlan, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences 271, 387 (1963), http://rspa. royalsocietypublishing.org/content/271/1346/387.full.pdf+html, URL http://rspa.royalsocietypublishing.org/content/271/1346/387. abstract. [47] C. Mavroyannis and M. Stephen, Molecular Physics 5, 629 (1962), http: //dx.doi.org/10.1080/00268976200100721, URL http://dx.doi.org/10. 1080/00268976200100721. [48] B. Linder, The Journal of Chemical Physics 37, 963 (1962), URL http:// scitation.aip.org/content/aip/journal/jcp/37/5/10.1063/1.1733253. [49] J. C. Slater and J. G. Kirkwood, Phys. Rev. 37, 682 (1931), URL http://link. aps.org/doi/10.1103/PhysRev.37.682. [50] K. T. Tang, Phys. Rev. 177, 108 (1969), URL http://link.aps.org/doi/10. 1103/PhysRev.177.108. 81 [51] E. Braaten and H.-W. Hammer, Physics Reports 428, 259 (2006), ISSN 0370-1573, URL http://www.sciencedirect.com/science/article/ pii/S0370157306000822. [52] F. Aquilante, L. D. Vico, N. Ferré, G. Ghigo, P. Malmqvist, P. Neogrády, T. Pedersen, M. Pitonak, M. Reiher, B. Roos, et al., Molcas 7.4 (2012), journal of Computational Chemistry, 31, 224, 2010, URL http://molcas.org/. [53] E. Tiesinga, B. J. Verhaar, and H. T. C. Stoof, Phys. Rev. A 47, 4114 (1993). [54] E. Tiemann, H. Knöckel, P. Kowalczyk, W. Jastrzebski, A. Pashov, H. Salami, and A. J. Ross, Phys. Rev. A 79, 042716 (2009), URL http://link.aps.org/ doi/10.1103/PhysRevA.79.042716. [55] S. Knoop, T. Schuster, R. Scelle, A. Trautmann, J. Appmeier, M. K. Oberthaler, E. Tiesinga, and E. Tiemann, Phys. Rev. A 83, 042704 (2011). [56] F. Wang, D. Xiong, X. Li, D. Wang, and E. Tiemann, Phys. Rev. A 87, 050702 (2013), URL http://link.aps.org/doi/10.1103/PhysRevA.87.050702. [57] A. J. Moerdijk, B. J. Verhaar, and A. Axelsson, Phys. Rev. A 51, 4852 (1995). [58] T. Köhler, K. Góral, and P. S. Julienne, Rev. Mod. Phys. 78, 1311 (2006). [59] C. Chin, V. Vuletić, A. J. Kerman, and S. Chu, Phys. Rev. Lett. 85, 2717 (2000). [60] C. A. Regal, C. Ticknor, J. L. Bohn, and D. S. Jin, Phys. Rev. Lett. 90, 053201 (2003). [61] J. Zhang, E. G. M. van Kempen, T. Bourdel, L. Khaykovich, J. Cubizolles, F. Chevy, M. Teichmann, L. Tarruell, S. J. J. M. F. Kokkelmans, and C. Salomon, Phys. Rev. A 70, 030702 (2004). 82 [62] C. H. Schunck, M. W. Zwierlein, C. A. Stan, S. M. F. Raupach, W. Ketterle, A. Simoni, E. Tiesinga, C. J. Williams, and P. S. Julienne, Phys. Rev. A 71, 045601 (2005). [63] J. P. Gaebler, J. T. Stewart, J. L. Bohn, and D. S. Jin, Phys. Rev. Lett. 98, 200403 (2007). [64] S. Knoop, M. Mark, F. Ferlaino, J. G. Danzl, T. Kraemer, H.-C. Nägerl, and R. Grimm, Phys. Rev. Lett. 100, 083002 (pages 4) (2008), URL http://link. aps.org/abstract/PRL/v100/e083002. [65] E. Wille, F. M. Spiegelhalder, G. Kerner, D. Naik, A. Trenkwalder, G. Hendl, F. Schreck, R. Grimm, T. G. Tiecke, J. T. M. Walraven, et al., Phys. Rev. Lett. 100, 053201 (2008), URL http://link.aps.org/doi/10.1103/PhysRevLett. 100.053201. [66] B. Gao, Phys. Rev. A 84, 022706 (2011). [67] J. Schwinger, Phys. Rev. 72, 738 (1947). [68] J. M. Blatt and D. J. Jackson, Phys. Rev. 76, 18 (1949). [69] H. A. Bethe, Phys. Rev. 76, 38 (1949). [70] M. Stoll and T. Köhler, Phys. Rev. A 72, 022714 (2005). [71] S. Simonucci, P. Pieri, and G. C. Strinati, Europhys. Lett. 69, 713 (2005). [72] B. Gao, E. Tiesinga, C. J. Williams, and P. S. Julienne, Phys. Rev. A 72, 042719 (2005). [73] B. Gao, Phys. Rev. A 80, 012702 (2009). [74] B. Gao, Phys. Rev. A 78, 012702 (2008). 83 [75] G. Breit and I. I. Rabi, Phys. Rev. 38, 2082 (1931), URL http://link.aps. org/doi/10.1103/PhysRev.38.2082.2. [76] N. Stone, Atomic Data and Nuclear Data Tables 90, 75 (2005), ISSN 0092-640X, URL http://www.sciencedirect.com/science/article/ pii/S0092640X05000239. [77] E. Arimondo, M. Inguscio, and P. Violino, Rev. Mod. Phys. 49, 31 (1977), URL http://link.aps.org/doi/10.1103/RevModPhys.49.31. [78] H. T. C. Stoof, J. M. V. A. Koelman, and B. J. Verhaar, Phys. Rev. B 38, 4688 (1988). [79] F. H. Mies, C. J. Williams, P. S. Julienne, and M. Krauss, J. Res. Natl. Inst. Stand. Technol. 101, 521 (1996). [80] S. Kotochigova, E. Tiesinga, and P. S. Julienne, Phys. Rev. A 63, 012517 (2000). [81] P. J. Leo, C. J. Williams, and P. S. Julienne, Phys. Rev. Lett. 85, 2721 (2000). [82] B. Gao, Phys. Rev. A 58, 4222 (1998). [83] B. Gao, Phys. Rev. A 64, 010701(R) (2001). [84] A. Derevianko, J. F. Babb, and A. Dalgarno, Phys. Rev. A 63, 052704 (2001), URL http://link.aps.org/doi/10.1103/PhysRevA.63.052704. 84 Appendix A Tables of Fitting Parameters for Atomic and Molecular Dynamic Polarizabilities Presented in the following sections are the parameters used in 2.10 for a number of diatomic molecules related to the present studies. Also included are the basis sets used for each atom and the level of theory the calculations were conducted at. As a reminder, the commonly used CCSD(T) level of theory does not permit dynamic polarizability calculations. In each case, the parallel and perpendicular polarizabilities are presented. A.1 SrLi See table A.1. A.2 LiYb See table A.2. 85 Table A.1: Ground State SrLi Dynamic Polarizability Fit Parameters: Def2-QZVPP basis set was used for both lithium and strontium. This basis set can be found in the standard CFOUR basis library. The parameters were fit to data points calculated at the CCSD level of theory. The associated pseudopotential for the inner 28 electrons was used for the strontium atom. i 1 2 3 4 1 2 3 4 A.3 Ai (a30 ) αk 144.016 130.659 271.714 21.6014 α⊥ 45.6356 8.73391 173.903 33.9479 Yb∗ See table A.3. A.4 NaK See tables A.4 and A.5 86 νi (cm−1 ) 9690.39 12663.6 14719.6 18887.1 11994.2 14383.8 22799.8 25031.7 Table A.2: LiYb Dynamic Polarizability Fit Parameters: A specialized basis set was used for ytterbium, sent via private communication from the authors of [28]. The Def2-QZVPP basis found in the CFOUR basis library was used for lithium. The parameters were fit to data points calculated at the CCSD level of theory. The associated pseudopotential for the inner 28 electrons was also used for the ytterbium atom. For this calculation, the molecule is at the equilibrium separation i 1 2 3 4 1 2 3 4 Ai (a30 ) αk 293.414 191.438 60.5147 42.3083 α⊥ 17.3537 40.3263 12.5494 165.183 νi (cm−1 ) 10596.9 15356.8 20016.0 23561.9 8420.48 15446.2 18391.7 26377.0 Table A.3: Metastable 3 P Ytterbium Dynamic Polarizability Fit Parameters: A specialized basis set was used for the ytterbium atom sent via private communication from the authors of [28]. The parameters were fit to data points calculated at the CCSD level of theory. The associated pseudopotential for the inner 28 electrons was also used. i 1 2 3 1 2 Ai (a30 ) αk 218.874 113.29 24.8098 α⊥ 162.31 80.7938 87 νi (cm−1 ) 10354.5 15076.7 27851.4 10442.3 27620.9 Table A.4: NaK ground state X1 Σ Dynamic Polarizability Fit Parameters: The QZVPP basis was used for both sodium and potassium. The parameters were fit to data points calculated at the CCSD level of theory. The molecule is at the equilibrium position for this calculation. i 1 2 1 2 Ai (a30 ) αk 495.192 21.3802 α⊥ 228.684 34.6618 νi (cm−1 ) 13322.2 19813.6 17683.6 21595.1 Table A.5: NaK ground a3 Σ Dynamic Polarizability Fit Parameters: The QZVPP basis was used for both sodium and potassium. The parameters were fit to data points calculated at the CCSD level of theory. The molecule is at the equilibrium position for this calculation. i 1 2 3 4 1 2 3 Ai (a30 ) αk 833.262 30.7737 6.15024 4.00357 α⊥ 25.0094 226.393 114.856 88 νi (cm−1 ) 9996.36 15694.0 23221.1 19027.4 4350.54 14330.7 20789.5