Spin 1 Micrcondensates Everything you ever wanted to know about the single mode approximation (and more) Austen Lamacraft University of Virginia May 15, 2011 Technion Quantum Magnetism Workshop Phys. Rev. A 83, 033605 (2011) Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 1 / 42 What is Quantum Magnetism in Ultracold Atoms? Is it this? Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 2 / 42 What is Quantum Magnetism in Ultracold Atoms? Or this? Cs Fe8 Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 2 / 42 Where are we? (rash overgeneralizations) Equilibrium states take a backseat to Dynamics Theory is only now coming to grips with real experiments We don’t understand what happens when you cool the simplest spinor system (spin 1 Bose gas) Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 3 / 42 Outline 1 Spin 1 Microcondensates 2 The (semi-)classical limit 3 Some interesting features of integrable classical motion 4 Quantizing the Microcondensate Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 4 / 42 Outline 1 Spin 1 Microcondensates 2 The (semi-)classical limit 3 Some interesting features of integrable classical motion 4 Quantizing the Microcondensate Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 5 / 42 What is a Microcondensate?1 Convenient shorthand for all atoms are in the same spatial state System size L ξd,c , the healing lengths for density and spin If spinor condensate ←→ magnet then microcondensate ←→ single domain 1 Coinage due to Fabrice Gerbier and Jean Dalibard, I believe. Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 6 / 42 Microcondensate Hamiltonian – Just two ingredients Contact interactions Hint = X i<j = X i<j δ(ri − rj ) [g0 P0 + g2 P2 ] δ(ri − rj ) [c0 + c2 Si · Sj ] c0 = (g0 + 2g2 ) /3 Austen Lamacraft (University of Virginia) c2 = (g2 − g0 ) /3 Spin 1 Microcondensates May 15th, 2011 7 / 42 Microcondensate Hamiltonian – Just two ingredients Zeeman energy of atom with z-component m = −1, 0, 1: both linear 2 (∝ 3 m) and quadratic (∝ m ) contributions DEUTSCHE PHYSIKALISCHE GESELLSC 39 K atoms in a magnetic field. Levels are lab using zero-field quantum numbers. Figure 1. Energy levels of Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 7 / 42 Microcondensate Hamiltonian For N particles is HSMA = N̂ = 1 X e0 e2 : N̂ 2 : + : Ŝ·Ŝ : +HZ 2N 2N A∗m Am Ŝ = A∗m Smm0 Am0 , m,m0 m=−1 HZ = 1 X m=−1 Austen Lamacraft (University of Virginia) X A∗m pm + qm2 Am Spin 1 Microcondensates May 15th, 2011 8 / 42 In gory detail... Dropping the density-density and linear Zeeman pieces, we have Hred = e2 2 Sz + 2 A∗1 A∗−1 (A0 )2 + (A∗0 )2 A1 A−1 + 2N 1 ∗ ∗ ∗ A1 A1 + A−1 A−1 + q A∗1 A−1 + A∗−1 A−1 . 2 A0 A0 − 2 ... and probably you are none the wiser! Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 9 / 42 Related models Dicke–Jaynes–Cummings HDJC = 2Sz + ωb † b + g (bS+ + b † S− ) Lipkin–Meshkov–Glick (shape transitions in nuclei, magnetic molecues, Bose–Hubbard dimers...) HLMG = − 1 γx Sx2 + γy Sy2 − hSz N Interacting Boson Model of the nucleus HIBM = s ns +d nd + X l1 ,l2 ,l10 ,l20 ,L Austen Lamacraft (University of Virginia) L L vlL1 l2 l 0 l 0 bl†1 × bl†2 · b̃ l10 × b̃ l20 Spin 1 Microcondensates 1 2 May 15th, 2011 10 / 42 Intriguing features in the spectrum... HSMA = e2 > 0, q > 0 (e.g. Austen Lamacraft (University of Virginia) 23 e2 : Ŝ · Ŝ : +q [N1 + N−1 ] 2N Na) e2 > 0, q < 0 Spin 1 Microcondensates May 15th, 2011 11 / 42 Outline 1 Spin 1 Microcondensates 2 The (semi-)classical limit 3 Some interesting features of integrable classical motion 4 Quantizing the Microcondensate Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 12 / 42 These coupled equations are nonlinear Josephson-type 3(b) equations and point to the equivalency of spin mix0 ms ing in a spin-1 condensate to Josephson systems re(a)in superconductors [27] and other superfluids alized [11, 14, 39, 40, 41, 42, 43, 44]. The non-linearity of these 70 ms equations provides a rich manifold of dynamical trajectories that can be accessed experimentally by choice of 0 components +1 of the spin -1 initial populations and phases 140 ms and the strength of the applied magnetic field. To investigate the coherent dynamics of this system, we 87 87 begin with Rb condensates created using an improved mf 1 0 -1 version of the all-optical trapping technique we have pre1.0 (c) viously reported [16, 19]. Using a dynamical compression 0.25 technique (b) and just a single focused laser beam, conden0.20 sates with up to 300, 000 atoms are created after 2 s of 0.8 0 ms 0.15 evaporative cooling. The condensates created in this op0.10 tical trap are generally in a mixture of all F = 1 spin 0.6 0.2 0.4 0.6 0.8 states and reveal complicated spatial domains. To create !" a well-characterized initial condition, we first prepare a 70 ms 0.4 condensate in the |F = 1, mF = −1" state by applying a magnetic field gradient during the evaporative cooling. To initiate spin dynamics, a coherent superposition of 0.2 spin states with non-equilibrium spin populations is cre140 ms ated by applying a sequence of phase-coherent microwave 0.0 pulses tuned to F = 1 ↔ F = 2 transitions. Follow0.0 0.1 0.2 0.3 0.4 0.5 0.6 ing this state preparation, the condensate is allowed to m 1 0 -1 f Time (s) freely evolve in the optical trap. A typical evolution is shown1.0 in Fig. 1c for an initial spin configuration of (c) ρ(1,0,−1) $ (0, 3/4, 1/4). Up to0.25 four distinct oscillations FIG. 1: Coherent spin mixing of spin-1 Bose condensate in an optical trap. Coherent spin mixing producing oscillations are observed in this example before the spin populations 0.20 in the populations of the F = 1, mF = 0, ±1 spin states of damp to0.8 a steady state. These oscillations demonstrate 87 Rb condensates confined in an optical trap starting from a 0.15 the coherence of the spin mixing process. superposition of condensate spin components at t = 0 that is We have measured the spin0.10 oscillation frequency for subsequently allowed to evolve freely. a) Schematic indicates different0.6 initial spin populations. These shown 0.8 fundamental spin mixing process. b) Absorptive images of the 0.2 data 0.4are Austen Lamacraft of Virginia) Spin0.6 1with Microcondensates May 15th, 2011 13 / 42 in the inset of(University Fig. 1c. and show good agreement Experiments in the single mode regime Period (s) lations Rb Period (s) Spin populations Chapman group (GA Tech, 2005) working with Experiments in the single mode regime (NIST, P H Y SLett I C A group L REV I E W L2009) E T T Eworking RS 23 Na week ending 27 MARCH 2009 1.0 Faraday Signal (arb. units) d state populations as a edicted phase transition spinor BEC. ntaining up to 6 ! 109 cal dipole trap derived nm is then loaded, folalization. A weak magg 6 s forced evaporation 1:5 ! 105 atoms in the trap frequencies are d the mean Thomasp up the magnetic field the field gradient. The 3 and 60:7 "T with an d uncertainties are estined statistical and sysstate with any desired m atomic spin followed by n mF state. The rf pulse splitting, and its ampli- with B = 26 µT < Bc 0.8 B = 40 µT > Bc (a) (b) 0.6 0.4 0.2 0.0 0 20 40 60 80 100 0 Hold Time (ms) 20 40 60 80 100 Hold Time (ms) FIG. 1 (color online). Faraday signal (proportional to hFx i2 ) taken from a single measurement for m ¼ 0 at two magnetic fields 26 and 40 "T starting with #0 ¼ 0:5, $ ¼ 0. The solid line is a fit with a damped sinusoid. The signals show (a) an oscillating phase and (b) a running phase at B below and above Bc , respectively, as evidenced by the signal reaching zero or not. backactionSpin in the present experiments. Austen Lamacraft (University of other Virginia) 1 Microcondensates In the absence May 15th, 2011 13 / 42 Dynamics in the classical limit Gross-Pitaevskii dynamics2 i da±1 ∗ = e2 ±sz a±1 + (a∓1 + a±1 )(1 − |a1 |2 − |a−1 |2 )− dt ∗ (a1∗ a−1 + a1 a−1 + |a1 |2 + |a−1 |2 )a±1 + qa±1 . ...can be solved in terms of elliptic functions, but what do you learn? √ Lower case throughout means ‘per particle’ quantities e.g. a1 = A1 / N, sz = Sz /N, etc. 2 Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 14 / 42 A strange choice of variables: Hyperbolic Spins 1 ∗ A1 A1 + A∗−1 A−1 + 1 2 K+ = A∗1 A∗−1 , K− = A1 A−1 , 1 1 B0 = A∗0 A0 + , 2 4 1 ∗ ∗ 1 B+ = − (A0 A0 ) , B− = − (A0 A0 ) . 2 2 K0 = Invariant under the rotations about the z-axis generated by the conserved quantity Sz Compare K ’s with Anderson spins in the fermionic case Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 15 / 42 Hyperbolic (or SU(1, 1) spins) K ’s and B’s are two reps of the non-compact group SU(1, 1) [K0 , K± ] = ±K± [K+ , K− ] = −2K0 and likewise for the {B0 , B+ , B− } Quadratic Casimir operators 1 1 2 (K− K+ + K+ K− ) = Sz − 1 2 4 1 3 CB = B02 − (B− B+ + B+ B− ) = − . 2 16 CK = K02 − Classical limit is a hyperboloid instead of a sphere! Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 16 / 42 Return on investment in strange variables3 Hred = e2 2 Sz + 8B0 K0 − 4B− K+ − 4B+ K− + 2qK0 . 2N Pair of coupled Hyperbolic spins Note ‘Minkowski’ scalar product in coupling K0 + B0 is just (half) total number of particles 3 N.M. Bogoliubov, J. Math. Sci. 136, 1, 3552-3559 (2006) Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 17 / 42 Classical limit Treat Am , A∗m as classical amplitudes Am , A∗m of ‘order √ N’ Sz2 4 CB ∼ 0. CK ∼ N ∼ 2 (B0 + K0 ) B degrees of freedom are restricted to a cone B0 = |B+ | Note that K0 , B0 ≤ Austen Lamacraft (University of Virginia) N 2 Spin 1 Microcondensates May 15th, 2011 18 / 42 Final stage of reduction Hred = e2 2 Sz + 8B0 K0 − 4B− K+ − 4B+ K− + 2qK0 . 2N Exploit the symmetry under rotations about the z-axis in hyperbolic spin space, B+ to be real and positive. Then B+ = B0 = N/2 − K0 and we can eliminate the B’s Hclass = e2 2 Sz + 8 (N/2 − K0 ) (K0 − Kx ) + 2qK0 . 2N Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 19 / 42 Geometric construction 1 h = sz2 + 2(1 − 2k0 ) (k0 − kx ) + 2q̃k0 2 h = const. fixes a hyperbola with asymptotes k0 = Austen Lamacraft (University of Virginia) Spin 1 Microcondensates 1 2, q̃ ≡ q/e2 k0 = kx + q̃2 May 15th, 2011 20 / 42 Geometric construction 1 h = sz2 + 2(1 − 2k0 ) (k0 − kx ) + 2q̃k0 2 q̃ ≡ q/e2 h = const. fixes a hyperbola with asymptotes k0 = 12 , k0 = kx + Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 q̃ 2 20 / 42 Two regimes q̃ > 0 Separatrix at h = 12 sz2 + q̃ Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 21 / 42 Two regimes Separatrix in the spectrum Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 21 / 42 Two regimes q̃ > 0 : topologically distinct trajectories at sz = 0 Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 21 / 42 Two regimes Signature in the spectrum. What’s happening at the origin? Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 21 / 42 Two regimes For q̃ < −2 origin moves through top boundary Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 21 / 42 Evidence for separatrix in 23 Na (q̃ < 0) week ending 17 AUGUST 2007 PHYSICAL REVIEW LETTERS m equilibrium, the full q. (1) are revealed. We e rf pulse as above, but cond time scales. ibed by the Hamilton 2 @E "_ ! : @ @!0 (2) Oscillation Period (ms) 40 Oscillation Amplitude ization are obviated, as each point as the differ! !" # !# . The meathe predictions, except n region. There the syse ground state, and the ce. We observe equilifrom 200 ms at high ds, by which time atom (a) 30 20 0.6 0.4 ρ 0 10 0.2 m 0 0.30 0 0.25 (b) 0 40 Time (ms) 80 0.20 0.15 0.10 0.05 e double-well ‘‘bosonic 0 50 0 10 20 30 40 and exhibits a regime of Magnetic Field (µT) near a critical field Bc , rmonic oscillations. At FIG. 2 (color online). Period (a) and amplitude (b) Black of spin oscillations applied magnetic field, following a #$B c&$1 # !(University Austen Lamacraft of Virginia) as a function Spin 1ofMicrocondensates c% ! 0% " et al. PRL 2007 May 15th, 2011 22 / 42 Relation to mean-field phase diagram4 e2 < 0 4 Stenger et al. Nature 396, 345-348 (1998) Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 23 / 42 Relation to mean-field phase diagram4 e2 < 0 4 Stenger et al. Nature 396, 345-348 (1998) Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 23 / 42 Relation to mean-field phase diagram4 e2 > 0 4 Stenger et al. Nature 396, 345-348 (1998) Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 23 / 42 Outline 1 Spin 1 Microcondensates 2 The (semi-)classical limit 3 Some interesting features of integrable classical motion 4 Quantizing the Microcondensate Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 24 / 42 Classical mechanics of the spin-1 gas System of three degrees of freedom {A∗m , Am0 } = iδmm0 . m = −1, 0, +1 There are three (commuting) conserved quantities 1 2 3 The energy Nh The angular momentum S z The particle number N The system is integrable in the sense of Liouville–Arnol’d Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 25 / 42 The Liouville–Arnol’d theorem Liouville–Arnol’d theorem For a system of N degrees of freedom integrable in the above sense, can find N conjugate pairs of action-angle variables (Ii , ϕi ), such that evolution of angles is trivial ϕi = ωt + ϕi,0 ϕ̇i = ∂H ∂Ii Submanifold of phase space at fixed {Ii } is N-Torus T N Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 26 / 42 The angles Can use each of the N commuting conserved quantities Fi in place of H in Hamilton’s equation to generate ‘flow’ x(t1 , . . . , tN ) Motion in general quasiperiodic: N incommensurate frequencies Set of values of t = (t1 , . . . , tN ) for which x(t1 , . . . , tN ) = x(0, . . . , 0) is a lattice (the period lattice) with vectors e1 , . . . , eN Define angular variables in terms of the reciprocal lattice vectors 1 , . . . , N satisfying i · ej = 2πδij . Then ϕi ≡ i · t, i = 1, . . . , N, in particular ϕi (t1 ) = ωi t1 + ϕi (0) ωi ≡ (i )1 , Austen Lamacraft (University of Virginia) i = 1, . . . , N. Spin 1 Microcondensates May 15th, 2011 27 / 42 The actions Ii ≡ 1 2π I γi p · dqi , γi around the i th circle of the torus. Actions are conjugate to the angles ∂H ωi = ∂Ii Ik 6=Ii . fixed Matrix of angular ‘velocities’ (i )j = ∂Fj ∂Ii or equivalently period lattice vectors ∂Ii (ei )j = 2π ∂Fj Fk 6=Fj Austen Lamacraft (University of Virginia) , Ik 6=Ii fixed Spin 1 Microcondensates . fixed May 15th, 2011 28 / 42 Actions for the spin 1 Microcondensate I1 = N I2 = Sz A I3 = πSz dA = Sz 2 2 sinh θ dθ dψ Sz iψ e sinh θ 2 Sz K0 = cosh θ 2 S2 K02 − |K+ |2 = z , 4 Area A on hyperboloid induced the ‘Minkowski’ metric 2 Austen Lamacraft (University of Virginia) Spin 1 Microcondensates K+ = May 15th, 2011 29 / 42 Trajectories on the half-plane y= Austen Lamacraft (University of Virginia) 2ky , k0 − kx z= Spin 1 Microcondensates sz . k0 − kz May 15th, 2011 30 / 42 q̃ > 0: Separatrix on the half-plane Circles project to circles −→ phase space is a disc Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 31 / 42 Properties of the action: q̃ < 0 Topologically distinct trajectories at sz = 0 Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 32 / 42 Properties of the action: q̃ < 0 A πSz 2 dydz Sz dA = 2 z2 I3 = Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 32 / 42 What is going on at the origin? Recall period lattice (ei )j = 2π ∂Ii ∂Fj Fk 6=Fj fixed 1 = 2π 0 ∂I3 ∂N 0 1 ∂I3 ∂Sz 0 0 ∂I3 ∂H ∂I3 ∂I3 ∂I3 e3 = 2π ∂N tells us how to execute a closed orbit around the ∂Sz ∂H third circle of the three-torus 1 ∂I3 Evolve for a time 2π ∂H 2 ∂I3 Change overall phase of the spinor by 2π ∂N 3 ∂I3 rotate about the z-axis by 2π ∂S z Rotation angle Φ(Sz , H) ≡ −2π ∂I3 ∂Sz not single-valued for q̃ < 0: increases by 2π around the origin H = Sz = 0 Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 33 / 42 Hamiltonian Monodromy φ 4π T −Φ 2π T Φ 2T t Non trivial mapping of period ei → the e0i =origin Mij ejin Sz , FIG. 13. (Color online)lattice (Left)toAsitself we circle M is monodromy matrix H space for q̃ < 0 the period lattice is deformed continuously, but after shifting the lattice returning to its original form, 1 0 0 vector corresponding to I3 by 2π in the φ direction. (Right) 1 0 Schematic illustrationMof=the 0rotation angle. While executing 1 1reduced phase space the a single period T of motion 0 on the system matrix rotatesofbyunit an angle Φ integer-valued determinant (an element of the group SL(3, Z)) Austen Lamacraft (University of Virginia) May 15th, 2011 �Spin 1 Microcondensates� Note riod l corres scaled Wh Recal the re jector From state rotati this p φ has there the or of non singu mathe 34 / 42 What is special about Sz = H = 0? Tip of the cone corresponds to the state (a1 , a0 , a−1 ) = (0, e −iχ , 0), invariant under rotations about the z-axis Thus the torus is pinched at this point: the circle corresponding to EFSTATHIOU, JOYEUX, AND SADOVSKIÍ rotations through φ has contracted to nothing M i c d FIG. 16. Two possible plots of a pinched torus; cf. Chap. IV.3, Fig. 3.5 on p. 163 of Ref. [4]. Both representations are equivalent in the four-dimensional phase space. Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 35 / 42 What is special about Sz = H = 0? Tip of the cone corresponds to the state (a1 , a0 , a−1 ) = (0, e −iχ , 0), invariant under rotations about the z-axis Thus the torus is pinched at this point: the circle corresponding to EFSTATHIOU, JOYEUX, AND SADOVSKIÍ rotations through φ has contracted to nothing M i c d FIG. 16. Two possible plots of a pinched torus; cf. Chap. IV.3, Fig. 3.5 on p. 163 of Ref. [4]. Both representations are equivalent in the four-dimensional phase space. Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 35 / 42 sis of a pendulum system reveals the relevance to he classical concept of ‘monodromy’ — why a falling right way up. Falling cats a slightly unstable state. The vertical ‘spring mode’ motion quickly becomes a ‘swing mode’ oscillation, just like a clock pendulum swinging in some vertical plane. However, this swing state is transient and the system returns once more to its spring mode, then back to a swing mode, and so on indefinitely. The surprise is that the successive planes in which it swings are different at each stage. Moreover, the angle through which the swing plane turns, from one occurrence to the next, depends sensitively on the amplitude of the original spring mode. The apparent paradox here is that the initial state has zero angular momentum — the this net spin about the vertical axis is zero.Yet the es, or swing state rotates from one instance to the attice, next. Analogously, a falling cat that starts d, the upside down has no angular momentum h the about its own longitudinal axis, yet it can They invert itself, apparently spinning about that od in axis. The resolution of the paradox, for a cat, nom- is that the animal changes its shape by movtum- ing its paws and tail in a particular way. At cat, each stage of the motion, angular momensition tum remains zero and is thus conserved, but e and the overall effect of the shape changes is to uces a invert the cat. The final upright state also romy has zero angular momentum, so there is no s feet contradiction of conservation. This effect new is known as the ‘geometric phase’, or monouanti- dromy, and is important in many areas of ple of physics and mathematics. The central topic of the paper is this: how amics Austen Lamacraft (University of Virginia) does monodromy show up when the system NHPA hysics s for systions ntum is the ntum ssical R. H. ntum whose nrico el for Rev. Spin 1 Microcondensates May 15th, 2011 36 / 42 Evolution of rotation angle Rotation angle could be extracted from Faraday rotation spectroscopy Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 37 / 42 From another cold atom lab... PRL 103, 034301 (2009) PHYSICAL REVIEW LETTERS week ending 17 JULY 2009 Experimental Demonstration of Classical Hamiltonian Monodromy in the 1:1:2 Resonant Elastic Pendulum N. J. Fitch,1 C. A. Weidner,1 L. P. Parazzoli,1 H. R. Dullin,2 and H. J. Lewandowski1 2 1 JILA and Department of Physics, University of Colorado, Boulder, Colorado 80309-0440 School of Mathematics and Statistics, The University of Sydney, Sydney, NSW 2006, Australia (Received 7 April 2009; published 15 July 2009) The 1:1:2 resonant elastic pendulum is a simple classical system that displays the phenomenon known as Hamiltonian monodromy. With suitable initial conditions, the system oscillates between nearly pure springing and nearly pure elliptical-swinging motions, with sequential major axes displaying a stepwise precession. The physical consequence of monodromy is that this stepwise precession is given by a smooth but multivalued function of the constants of motion. We experimentally explore this multivalued behavior. To our knowledge, this is the first experimental demonstration of classical monodromy. DOI: 10.1103/PhysRevLett.103.034301 PACS numbers: 45.05.+x, 02.30.Ik, 45.50.!j week e intriguing After more than 300 years since the formulation ofS I103, HE YR S ISC Athat L Rthe E Vsingular I E W L E T T17 E RJULY S PHY CThere A034301 L isRthus E(2009) V Ithe EW L E TPpossibility T PRL PRL 103, 034301 (2009) behavior of monodromy may be a common feature of Newton’s laws of motion, one would expect that a system ∼ 5 thresholds [17]. dynamics near chemical isomerization as simple as a mass on a spring would have been fully enclosing loop ($3) motion are A quantum analog of the resonant elastic pendulum understood for some time. In fact, an in-depth investigation former loop does not c under consideration here (Fig. 1) is the Fermi of even a subset of its possible dynamics produces a returning H% H2 resonance Lzto the initial #¼ ; $It¼ : in the CO2 molecule, whose monodromatic number of surprises. Chief among these is a phenomena is a simple distin 3=2features have H2 "H 2 stepwise precession ad known as Hamiltonian monodromy, which was introduced been thoroughly investigated theoretically [14,16]. Despite behavior by Duistermaat in 1980 as a topological obstruction to the the large number of systems in which monodromy is along a loop With this scaling, the thread pierces the 2D lem the key to our first existence of global action-angle variables [1]. In the resotheoretically predicted, there have been no previousYet, clas∼3 ð$;and #Þ ¼ ð0;a0Þ, producing singularity [Fig. 2(a)]. monodromy is our abi nant elastic pendulum, monodromy has easily observable sical experiments only single quantumaexperiment is developing responsiblea more for the existence of monod resonant elastic pendu physical consequences. Specifically, the observed stepwise [18] of which wesingularity are aware. In heuristic be p The presence the monodromy-producing precession of the elliptical swinging major axis is given by understanding of monodromy in of quantum systems,sical it isexample cansingu to build the intuition ne rotationtonumber of the integrable approxim a smooth, but multivalued function of the constants of useful to have acauses classicala example guide one’s intuition. by no means rare, motion. This functional form results in loops of valuesFIG. of 4. Thus we designed experiment on abetween readily realized to beour multivalued. This rotation number correspondsin Experimental measurements of the step angle FIG. 3. Measured mass positions as projected onto the XY foundly influences Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 2011 38 /plane 42 thd successive swinging motions loops #) the space are mapped the constants of motion having differing overall behavior, classical system inaswhich the(",consequences of15th, monodromy step size in !% of stepwise-precessing swing Outline 1 Spin 1 Microcondensates 2 The (semi-)classical limit 3 Some interesting features of integrable classical motion 4 Quantizing the Microcondensate Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 39 / 42 Semiclassical (EBK) quantization Ii = (ni + µi ) ~, ni ∈ Z µi are Maslov indices. Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 40 / 42 Monodromy as a Quantum Defect5 EBK gives lattice in space of conserved quantities with vectors ∂Fj ~ = ~ (i )j ∂Ii Ik 6=Ii fixed Just reciprocal lattice of period lattice Accompanying monodromy period lattice is the corresponding map on the ‘quantum’ lattice i → 0i = M −1 ji j , 5 Due to Boris Zhilinskii Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 41 / 42 Exact solution using Algebraic Bethe Ansatz6 An eigenstate of the Hamiltonian is written as |Ψi = NR + Y B j K+ + λj λj + N q̃/2 |0, νK iK ⊗ |0, νB iB The λj j = 1, . . . NR satisfy the equations N 1− R X νK 1 νB − = . λj λj + N q̃/2 λj − λl l6=j With energy NR e2 4e2 X E ({λj }) = N(N − 1) + q|Sz | − λj 2N N j=1 6 N.M. Bogoliubov, J. Math. Sci. 136, 1, 3552-3559 (2006) Austen Lamacraft (University of Virginia) Spin 1 Microcondensates May 15th, 2011 42 / 42 e to Josephson systems re[27] and other superfluids 4]. The non-linearity of these 70 ms manifold of dynamical trajecexperimentally by choice of ases of the spin components Single of ms a Microcondensate holds a few surprises lied magnetic field. mode dynamics 140 t dynamics of this system, we Stillandon’t understand real experiment. What is-1 the origin of damping? es created using improved mf 1 0 pping technique we have pre(c) 1.0 ing a dynamical compression 0.25 focused laser beam, conden0.20 toms are created after 2 s of 0.8 0.15 ndensates created in this op0.10 a mixture of all F = 1 spin 0.6 0.2 0.4 0.6 0.8 d spatial domains. To create !" condition, we first prepare a 0.4 F = −1" state by applying a ng the evaporative cooling. , a coherent superposition of 0.2 rium spin populations is creof phase-coherent microwave 0.0 F = 2 transitions. Follow0.0 0.1 0.2 0.3 0.4 0.5 0.6 the condensate is allowed to Time (s) l trap. A typical evolution initial spin configuration of p to four distinct oscillations FIG. 1: Coherent spin mixing of spin-1 Bose condensate in Small size necessary but probably insufficient condition for validity of an optical trap. Coherent spin mixing producing oscillations e before the spin populations in the populations(doesn’t of the F = 1,refer mF =to 0, ±1 spin states of single mode approximation temperature) hese oscillations demonstrate 87 Rb condensates confined in an optical trap starting from a ixing process. superposition of condensate spin components at t = 0 that is pin oscillation frequency for subsequently allowed to evolve freely. a) Schematic indicates Austen Lamacraft (University of Virginia) Period (s) Spin populations Summary Spin 1 Microcondensates May 15th, 2011 43 / 42