Parametric Excitation and Squeezing in a Many-Body Spin System

T.M. Hoang, M. Anquez, B.A. Robbins, X.Y. Yang, B.J. Land, C.D. Hamley, and M.S. ChapmanSchool of Physics, Georgia Institute of Technology, Atlanta, GA 30332-0430

(Dated: December 18, 2015)

We demonstrate a new method to coherently excite and control the quantum spin states ofan atomic Bose gas using parametric excitation of the collective spin by time varying the relativestrength of the Zeeman and spin-dependent collisional interaction energies at multiples of the naturalfrequency of the system. Compared to the usual single-particle quantum control techniques used toexcite atomic spins (e.g. Rabi oscillations using rf or microwave fields), the method demonstratedhere is intrinsically many-body, requiring inter-particle interactions. While parametric excitationof a classical system is ineffective from the ground state, we show that in our quantum system,parametric excitation from the quantum ground state leads to the generation of quantum squeezedstates.

Ultracold atomic gases with well-characterized col-lisional interactions allow new explorations of non-equilibrium dynamics of quantum many-body physicsand for synthesis of strongly-correlated quantum statesincluding spin-squeezed [1–4] and non-Gaussian entan-gled states [5–7] relevant for quantum sensing [8] andquantum information [9]. Here, we demonstrate a newmethod to coherently excite and control the quantumspin states of an atomic Bose gas using parametric exci-tation of the collective spin by time varying the relativestrength of the Zeeman and spin-dependent collisional in-teraction energies at multiples of the natural frequencyof the system. Compared to the usual single-particlequantum control techniques used to excite atomic spins(e.g. Rabi oscillations using rf or microwave fields), themethod demonstrated here is intrinsically many-body,requiring inter-particle interactions. While parametricexcitation of a classical system is ineffective from theground state [10], we show that, in our quantum system,this leads to exponential excitation and the generation ofquantum squeezed states.

Parametric excitation of an oscillating physical systemcan be achieved by periodically varying one of its param-eters to modulate the natural frequency of the oscillator,f0; a textbook example is a simple pendulum excited bymodulating its length, `, such that f0(t)∝1/

√`(t) [10].

A fundamental distinction between parametric excitationand direct excitation by periodic forcing is shown in Fig.1, which shows instantaneous phase space orbits of a sim-ple oscillator for the two cases. For direct excitation, theapplied force periodically displaces the equilibrium po-sition of the oscillator, leaving the orbits otherwise un-changed, and efficient excitation occurs when the exci-tation frequency matches the natural frequency of theoscillator, f=f0. For parametric excitation, the parame-ter modulation leaves the equilibrium location unchangedbut instead periodically distorts the phase orbits; in thiscase, efficient excitation occurs for excitation frequenciesf=2f0/n, n=1,2,3...

In ultracold atom traps, parametric excitation of theatomic motion, achieved by modulating the trapping po-

tential, is used to measure the trap frequency as well asin a variety of studies including the excitation of Bose-Einstein condensate (BEC) collective density modes [11–14], controlling the superfluid/Mott insulator transition[15, 16], and photon-assisted tunneling in modulated op-tical lattices and super-lattices [17–22].

In this work, we demonstrate parametric excitation ofthe internal states of a collection of atomic spins. Thespins are coherently excited to non-equilibrium states bya simple modulation of the magnetic field magnitude atvery low frequencies (<200 Hz) compared to the energydifference of the Zeeman states (∆E/h=0.7 MHz, whereh is Planck’s constant). The excitation spectrum is fullycharacterized and compares well to theoretical calcula-tions. Parametric excitation of the ground state is alsoinvestigated. Classically, parametric modulation of an os-cillating system does not excite the ground state. Here,we show that the finite quantum fluctuations of the col-lective spin leads to parametric excitation of the groundstate which manifest as exponential evolution of the fluc-tuations and the generation of non-classical squeezedstates. The exponential evolution and squeezing of thespin fluctuations are measured and agree qualitativelywith theory. Finally, we discuss how these techniquescan be applied to related systems including the double-well Bose-Hubbard model and interacting (psuedo)spin-1/2 ensembles.

The experiments use 87Rb Bose condensates with N=40000 atoms in the F =1 hyperfine level tightly confinedin optical traps such that spin domain formation is en-ergetically suppressed and dynamical evolution of thesystem occurs only in the internal spin variables. TheHamiltonian describing the evolution of this collectivespin system in a bias magnetic field B along the z-axis is[4, 23–25]:

H= cS2− 1

2qQz (1)

where S2 is the total spin-1 operator and Qz is propor-tional to the spin-1 quadrupole moment, Qzz. The co-efficient c is the collisional spin interaction energy per

arX

iv:1

512.

0564

5v1

[co

nd-m

at.q

uant

-gas

] 1

7 D

ec 2

015

2

x

p

a

x

p

b

c

ρ0

1

0θs

-π

π

Q⟂

Qz

S⟂

Time (ms )

ρ0

d

0 5 10 150

1

12

FIG. 1: Experimental concept. Momentum-position phasespace of a harmonic oscillator under (a) parametric and (b)direct excitation. The original orbits are shown in black, andthe modified orbits are shown in color at different instants ofthe periodic excitation. (c) The phase space of the conden-sate. The collective states (normalized to N) lie on a unitsphere with axes S⊥,Q⊥,Qz. The orbits of constant energyfor non-interacting (c=0) spins are lines of latitude shown inblack and the orbits for interacting spins with q=10|c| areshown in color. The ρ0,θs diagram is a Mercator projectionof the hemisphere. θs =θ+1 +θ−1−2θ0 is the relative phaseof the three Zeeman sub-levels, mF =0,±1 (d) Measurementsof the natural oscillations of ρ0 at B=1 G for different initialstates ρ0(t=0)∈ [0,1]. The experimental data (markers) arecompared to simulations (lines). Unless otherwise indicated,the uncertainty in the measurement of ρ0 is <1%.

particle integrated over the condensate and q=qzB2 is

the quadratic Zeeman energy per particle with qz=72Hz/G2 (hereafter, h=1). The longitudinal magnetiza-tion 〈Sz〉 is a constant of the motion (=0 for theseexperiments); hence the first order linear Zeeman en-ergy pSz with p∝B can be ignored. The spin-1 coher-ent states can be represented on the surface of a unitsphere shown in Fig. 1(c)with axes S⊥,Q⊥,Qz whereS⊥ is the transverse spin, S2

⊥=S2x+S2

y , Q⊥ is the trans-verse off-diagonal nematic moment, Q2

⊥=Q2xz+Q2

yz, andQz=2ρ0−1 where ρ0 is the fractional population inthe F =1,mF =0 state. In this representation, the dy-namical orbits are the constant energy contours of H=

12cS

2⊥− 1

2qQz where c=2Nc.The experiment is conducted at high fields where the

Zeeman energy dominates the spin interaction energy,q/|c|=10. In this regime, the lowest energy state is thepolar state (ρ0 =1) located at the top of the sphere,and the dynamical orbits of the excited states to lead-ing order are simple rotations about the Qz axis with afrequency f0≈q+cQz (Supplemental Information). De-spite the small relative magnitude of the spin interactionterm, it has the important effect of breaking the polarsymmetry and thereby slightly distorting the orbits fromthe latitudinal lines of the sphere. As the state orbitsthe sphere, the population ρ0 undergoes small periodicnutations at twice the orbit frequency, as shown in theρ0,θs projection in Fig. 1(c). The maximum nutationamplitude is ∆ρ0≈0.02 for ρ0 =0.5 and goes to zero forρ0 =0,1. Measurements of these distortions are shown inFig. 1(d) for different initial values of ρ0.Parametric excitation

Parametric excitation requires periodic modulation ofone of the parameters of the Hamiltonian; in a spin-1 condensate described by Eq. 1, this is convenientlyachieved by modulating the bias magnetic field and hencethe quadratic Zeeman energy term, q(t)∝B2(t). Thecondensate is first prepared in a coherent state withρ0,θs=(0.5,π) at a field of 1 G, corresponding to aninitial quadratic Zeeman energy q0 =72 Hz. The spinordynamical rate is c=−7(1) Hz, determined from mea-surements of coherent oscillations at low fields. To para-metrically excite the spins, the magnetic field is modu-lated for a duration of time, after which the spin popu-lations are measured to determine the final value of ρ0.The applied modulation is harmonic in q and has theform q(t)=q0[1+εsin(2πft−φ0)]. The measured exci-tation spectrum versus modulation frequency is shownin Fig. 2(a). The spectrum shows the characteristic fea-tures of parametric excitation, namely strong excitationat 2f0 =142 Hz and weaker excitation at f0. Other reso-nances are theoretically observable at smaller f=2f0/nvalues; however, they are dominated by the tails of themore prominent peaks making them difficult to detect.The experimental data (marker) are compared to a sim-ulation using Eq. 1 (solid line) and show good agreementoverall.

Beyond comparing the experimental results to numer-ical solutions of the quantum Hamiltonian, insight intothe parametric excitation is obtained by considering themean-field dynamical equations for ρ0 and the quadra-ture angle θ=θs/2 [26]:

ρ0 = 4πcρ0(1−ρ0)sin2θ

θ = −2π[q−c(1−2ρ0)(1+cos2θ)]

These equations are similar to bosonic Josephson junc-tion equations describing the double-well condensate[27, 28] and can be solved like-wise by integrating the

3

f (Hz)

ρ0

a

2f0f02f0/3

0 50 100 150 200 250 3000.2

0.4

0.6

0.8

f (Hz)

ρ0

b

100 120 140 160 1800.0

0.4

1.0

0.2

0.6

0.8

++

++ +

++ +

+++

+

+

+++ + + +

++++ + + +

* * * * * * * * * * * * * * * * * * * * * * * * * *

+ ϵ=0.61 ϵ=0.51 ϵ=0.34 ϵ=0.21 ϵ=0.1* ϵ=0

f (Hz)

c

ρ0

120 130 140 150 160 170

0.4

0.6

0.5

FIG. 2: Demonstration of parametric excitation. (a)Population ρ0 after 100 ms of parametric excitation for(ρ0(0),φ0, ε)=(0.5,π,ε=0.5). Data (markers) are plottedwith simulation (solid line) for comparison. The excitationspectrum shows clear resonances at frequencies 2f0 and f0(b) Populations ρ0 after 40 ms of parametric excitation fordifferent initial ρ0 populations for φ0 =π and ε=0.5. Data(markers) are compared to simulation (solid lines), and huecolors correspond to the initial ρ0∈ [0,1]. The dashed line isthe theoretical prediction for 2f0 resonance, and the squareboxes indicate the location of the measured resonance. (c)Population ρ0 after 40 ms of parametric excitation for differ-ent modulation amplitudes ε.

phase and using the Jacobi-Anger expansion (Supple-mental Information),

ρ0 = 4πcρ0(1−ρ0)∑n

Jn

(4πqmω

)sin[(nω−2ω0)t+φn]

where Jn is the Bessel function of order n, qm=εq0 is themodulation strength and φn depends on the initial condi-tions including the phase of the modulation (Supplemen-

tal Information). The parametric resonance frequenciesare obvious from this solution because the time-averageof ρ0 is zero unless ω=2ω0/n.

In the q|c| high field regime of these experiments,the collisional interactions shift the natural oscillationfrequency as f0 = |θ|/2π≈q+c(2ρ0−1) to lowest order.This shift is investigated in Fig. 2(b), where the excita-tion is measured for different initial values of ρ0. The huecolors correspond to the initial ρ0 values using the samescale as the coherent oscillation data in Fig. 1(d), andthe square markers indicate the positions of the measuredresonance frequencies where ∆ρ0≈0. The measured res-onance frequencies are in good overall agreement withthe expected dependence on the initial value of ρ0 shownwith the dashed line—the small discrepancy is attributedto an inductive delay in the excitation of ∼1 ms (seeMethods) that creates a phase offset δφ0∼0.9 rad in theexcitation. Indeed, the experimental data compare verywell to simulations (solid line) that include this phaseoffset.

The dependence of the excitation amplitude on thedrive strength qm=εq0 is reflected in the Bessel functionJn(4πqm/ω). In Fig. 2(c), the excitation is measured fordifferent modulation amplitudes. The experimental data(markers) are compared to the simulations (solid line)and show good agreement (see Methods). As expected,the modulation amplitude does not affect the resonancefrequency of the parametric excitation; however, increas-ing the modulation amplitude results in larger excitationof ρ0.

Parametric excitation is a coherent process, and henceit can be employed as a tool for quantum control of thecollective spin. However, accurate control requires de-tailed knowledge of the system response to the excitationparameters. In Fig. 3, we present a parameter variationmap that shows the excitation in the neighborhood ofthe 2f0 resonance for different values of excitation timet, phase φ0 and frequency f . For each measurement, weprepare the initial state ρ0,θs=(0.5,π) using an rf pulseand modulate the quadratic Zeeman energy at differentfrequencies and initial phases. The change in popula-tion ∆ρ0 is measured after excitation times of 40 ms,100 ms, 160 ms, and 220 ms as shown in Fig. 3 (fourvertical slices). The white/orange (black/green) regionsrepresent the positive (negative) changes in population.These two regions evolve and spiral to form a distinctive‘yin-yang’ pattern. This pattern in the measurementsagrees well with theoretical calculations of the excitation(see Supplemental Information). At the center of the pat-tern (f,φ0)=(2f0,π) where 2f0≈143 Hz, ρ0 remains un-changed and the excitation shows anti-symmetry aboutthis point (diamond marker). The inset of Fig. 3 showsthe temporal evolution of ρ0 for different f,φ0 pairs in-dicated in the vertical slice. Although in each case theinitial state is identical, the final state after excitationshows a strong dependence on both the frequency and

4

Time (ms )

f (Hz)

Pha

seϕ

0

Time (ms )

ρ0

0 200 400

0.3

0.5

0.7

Δρ0

0.3-0.3 a

b

FIG. 3: Excitation parameter map. The data show theexcitation of the condensate, ∆ρ0 =ρ0−ρ0(0) following para-metric excitation as a function of the excitation frequency fand initial phase φ0 of the modulation. The initial state isρ0(0),θs =(0.5,π) and the modulation strength is ε=0.5 Thefour vertical slices show ∆ρ0 after 40 ms, 100 ms, 160 ms, and220 ms of parametric excitation. The horizontal plot showsthe evolution of ∆ρ0 for an initial phase φ0 =0. The insetshows the temporal evolution of ρ0 for different f,φ0 pairsindicated in the vertical slice. The markers on the inset cor-respond to the markers on the main diagram.

phase of the applied modulation. Additionally, we showa map of the population dynamics for the initial phaseφ0 =0 (Fig. 3, horizontal slide). The two distinguishabledomains, white (orange) black (green), are separated bythe resonance frequency. The population dynamics ex-hibit oscillations during the excitation process, and, ap-proaching the resonance frequency, both the oscillatingperiod and the amplitude (∆ρ0) increase.

We now turn to measurements of uniquely quantumfeatures of the excitation. For a classical oscillator pre-pared in its stable equilibrium configuration, an impor-tant distinction between direct excitation and parametricexcitation is that former can efficiently excite the oscil-lator while the latter cannot. The equilibrium (ground)state is a stable fixed point in phase space and if the os-cillator is perfectly initialized it will remain unexcited byparametric modulation. However, for a quantum systemprepared in its ground state, intrinsic Heisenberg-limitedfluctuations of the state still allow for parametric excita-tion. In the semi-classical picture, the quantum fluctu-ations populate a family of orbits about the equilibriumpoint in the phase space that can be parametrically ex-cited.

In Fig. 4, we investigate parametric excitation fromthe quantum ground state of the condensate located atρ0 =1 and demonstrate that parametric excitation can be

Detection Limit

Time (s)

(ΔS⟂)2

dB

Min Max

0 0.1 0.2 0.3 0.4-10

0

10

20

30

FIG. 4: Parametric excitation from ρ0 =1. Evolution ofthe minimum and maximum values of the transverse spin fluc-tuations, ∆S2

⊥, following parametric excitation with ε=0.56and f=134 Hz starting from the initial state ρ0 =1. The mea-sured maximum (red markers) and minimum (blue markers)variance of the transverse magnetization S⊥ is compared withsimulation (solid red and solid blue).

used to generate quadrature squeezed states. Althoughthe population, ρ0, is largely insensitive to parametric ex-citation from the ρ0 =1 state (in contrast with the ρ0 6=1initial states), the fluctuations in the transverse coordi-nates, S⊥,Q⊥, evolve exponentially with time and showquadrature squeezing in the spin-nematic phase space. Incontrast to our previous demonstration of squeezing [4],in which the squeezing was generated by free dynamicalevolution following a quench that localized the state at aunstable (hyperbolic) fixed point, here, the squeezing isgenerated near a stable fixed point by periodic distortionof the phase space orbits produced by the modulation ofthe quadratic Zeeman energy. The essential difference isthe time-dependence in the Hamiltonian.

In Fig. 4, measurements of the minimum and maxi-mum values of the quadrature fluctuations of the trans-verse spin, ∆S⊥, are shown and compared with a quan-tum simulation. As evidenced by both the measurementsand the calculations, the fluctuations in the initial stateevolve exponentially at early evolution times and developinto quadrature squeezed states. The maximum squeez-ing measured is −5 dB, which is close to the detection-limited ceiling of −6 dB due to the photo-detection shot-noise and background scattered light [4]. The simulationssuggest that with technical improvements, the system iscapable of generating squeezing at the −20 dB level. Al-though the experiments data show the main effects pre-dicted by theory, the agreement of the measured fluctua-

5

tions with the theory is not perfect, particularly at longerevolution times. This is possibly due to effects of atomloss from the condensate, which has a lifetime of 1.5 sfor these experiments, and we plan to further investigatethis question in future work.Discussion

The parametric excitation can also be understood astransitions between eigentates of the many-body Hamil-tonian, which can be calculated by diagonalizing thetridiagonal matrix [5, 29]

Hk,k′ = 2ck(2(N−2k)−1)+2qkδk,k′

+ 2c(k′+1)√

(N−2k′)(N−2k′−1)δk,k′+1

+ k′√

(N−2k′+1)(N−2k′+2)δk,k′−1 (2)

written in the Fock basis |N,M ;k〉 where k is the numberof pairs ofmf =±1 atoms, N is the total number of atomsand M is the magnetization; both N and M are con-served by the Hamiltonian. Treating c as a perturbation,the eigenenergies are Ek=2qk+2ck(4N−4k−1) and theenergy difference between Fock states is ∂Ek/∂k=2q+2c(2ρ0−1) (Supplemental Information). Using this pic-ture, we note that the parametric excitation spectrumexcitation frequencies f=2f0/n, n=1,2,3... correspondsto many photon excitations of the system with f=2f0being the single photon transition.

It is interesting to contrast parametric excitation withthe usual Rabi excitation of 2-level atomic spins using rfor microwave magnetic fields. In parametric excitation,the time variation of a parameter modifies the Hamilto-nian without displacing the equilibrium (ground) stateof the system. As we have demonstrated, this can beachieved in a spin-1 condensate by simply modulatingthe magnitude of the bias magnetic field, which modu-lates the quadratic Zeeman energy term in the Hamilto-nian. As the field strength is varied, the ground stateremains at the pole of the sphere, while the shape of theorbits is modulated; this is similar to the case shown inFig. 1(a). Rabi excitation of 2-level atomic spins usingrf or microwave magnetic fields, on the other hand, isdirect excitation rather than parametric excitation. Al-though in both cases the excitation employs time-varyingmagnetic fields, in the Rabi case the oscillating magneticfield is transverse to the bias field, which leads to a os-cillation of the orientation of the total field. Because theground state of the spin aligns along the field direction,the addition of the time-varying transverse field leads toa periodic displacement of the ground state away fromthe pole in the usual Bloch sphere picture of the spinvector, similar to the case shown in Fig. 1(b).

We point out that the techniques we have demon-strated are applicable to related many-body systemsincluding the double-well Bose-Hubbard (or BosonicJosephon junction (BJJ) model), collisionally interact-ing pseudo-spin 1/2 two component condensates and en-sembles of spin-1/2 atoms with photon mediated interac-

tions. Each of these systems can be described by a ver-sion of the Lipkin-Meshkov-Glick (LMG) model Hamil-tonian [30], H=US2

x−KSz, whose the mean-field phasespace is functionally identical to the spin-nematic phasespace shown in Fig. 1(c) [4, 28]. At their heart, thesesystems feature competing energy terms (one of whichis non-linear) that give rise to a quantum critical point.The q>2|c| polar phase that we explore in this work cor-responds to the Rabi regime (K>U) in the BJJ system,which is the tunneling-dominated regime perturbed bythe interactions US2

x, and the modulation of q corre-sponds to a modulation of the tunneling coefficient K.Indeed, there have been numerous theoretical proposalsfor excitations of these systems using periodic modula-tions (e.g. see [31–34]), and many of the lattice-basedexperimental demonstrations mentioned previously real-ize closely related ideas generalized to multi-site systems[15–22].

In summary, we have demonstrated a new mechanismfor control and excitation of an ensemble of spins basedon parametric excitation. This is a many-body controltechnique that relies on spin-dependent collisional inter-actions, which we have characterized for a wide range ofcontrol parameters. We have shown that this method,when applied to the ground state, can be used to gener-ate squeezed states.

METHODS

The experiment utilizes 87Rb atomic Bose-Einstein conden-sates created in an optical trap containing N=40000 atomsinitialized in the |F =1,mF =0〉 hyperfine state in a high mag-netic field (2 G). To prepare the initial spin state, the conden-sate is rapidly quenched to a magnetic field of 1 G, and a Rabirf pulse resonant with the F =1 Zeeman transition is appliedto prepare the desired initial state ρ0,θs =(ρ0(t=0),π). Thenominal value of the magnetic field B=1 G is determined byusing rf and microwave spectroscopy, and the spinor dynam-ical rate is determined by measuring coherent spin dynam-ics oscillations using states prepared near the ferromagneticground state (c=−7(1) Hz).

Parametric excitation of the system is implemented by sinu-soidally modulating q, which is implemented by time-varyingthe magnetic field using external coils. Due to induced eddycurrents in the metal vacuum chamber and the inductanceof the magnetic field coils, the applied modulation is time-delayed relative to the intended control by 1 ms and reducedin amplitude by 15%. These effects are measured directly us-ing magnetically sensitive rf spectroscopy of the atoms, andthey are incorporated in all the simulations.

The final spin populations of the condensate are measuredby releasing the trap and allowing the atoms to expand ina Stern-Gerlach magnetic field gradient to separate the mF

spin components. The atoms are probed for 400 µs with threepairs of counter-propagating orthogonal laser beams, and thefluorescence signal is collected by a CCD camera is used todetermine the number of atoms in each spin component.

To measure the transverse spin fluctuations ∆S⊥, an rfRabi π/2 pulse is applied during the expansion to rotate the

6

transverse spin fluctuations (which are in the x,y plane) intothe z measurement basis. The fluctuations are then deter-mined from 30 repeated measurements of 〈Sz〉, the differencein the number of atoms measured in the mF =1 and mF =−1spin components.

Acknowledgements We acknowledge support from theNational Science Foundation Grant PHY–1208828

Author Contributions T.M.H., C.D.H. and M.S.C.jointly conceived the study. T.M.H, M.A., B.A.R. and X.Y.Y.performed the experiment and analyzed the data. T.M.H.,B.J.L. and C.D.H. developed essential theory and carried outthe simulations. M.S.C supervised the work.

[1] Gross, C., Zibold, T., Nicklas, E., Esteve, J., andOberthaler, M. K. Nonlinear atom interferometer sur-passes classical precision limit. Nature 464, 1165–1169(2010).

[2] Riedel, M. F., Bohi, P., Li, Y., Hansch, T. W., Sina-tra, A., and Treutlein, P. Atom-chip-based generation ofentanglement for quantum metrology. Nature 464, 1170–1173 (2010).

[3] Lucke, B., Scherer, M., Kruse, J., Pezz, L., Deuret-zbacher, F., Hyllus, P., Topic, O., Peise, J., Ertmer,W., Arlt, J., Santos, L., Smerzi, A., and Klempt, C.Twin matter waves for interferometry beyond the clas-sical limit. Science 334(6057), 773–776 (2011).

[4] Hamley, C., Gerving, C., Hoang, T., Bookjans, E., andChapman, M. Spin-nematic squeezed vacuum in a quan-tum gas. Nature Phys. 8, 305–308 (2012).

[5] Gerving, C., Hoang, T., Land, B., Anquez, M., Hamley,C., and Chapman, M. Non-equilibrium dynamics of anunstable quantum pendulum explored in a spin-1 Bose-Einstein condensate. Nat. Commun. 3, 1169 (2012).

[6] Gring, M., Kuhnert, M., Langen, T., Kitagawa, T.,Rauer, B., Schreitl, M., Mazets, I., Smith, D. A., Demler,E., and Schmiedmayer, J. Relaxation and prethermaliza-tion in an isolated quantum system. Science 337 (2012).

[7] Strobel, H., Muessel, W., Linnemann, D., Zibold, T.,Hume, D., Pezze, L., Smerzi, A., and Oberthaler, M. K.Fisher information and entanglement of non-gaussianspin states. Science 345, 424–427 (2014).

[8] Ma, J., Wang, X., Sun, C., and Nori, F. Quantum spinsqueezing. Physics Reports 509, 89–165 (2011).

[9] Braunstein, S. L. and van Loock, P. Quantum infor-mation with continuous variables. Rev. Mod. Phys. 77,513–577, Jun (2005).

[10] Landau, L. and Lifshitz, E. Mechanics. Butterworth-Heinemann, Third edition, (1976).

[11] Jin, D., Ensher, J., Matthews, M., Wieman, C., and Cor-nell, E. Collective excitations of a Bose-Einstein conden-sate in a dilute gas. Phys. Rev. Lett. 77, 420 (1996).

[12] Mewes, M.-O., Andrews, M., Van Druten, N., Kurn, D.,Durfee, D., Townsend, C., and Ketterle, W. Collectiveexcitations of a Bose-Einstein condensate in a magnetictrap. Phys. Rev. Lett. 77, 988 (1996).

[13] Engels, P., Atherton, C., and Hoefer, M. A. Observationof faraday waves in a Bose-Einstein condensate. Phys.Rev. Lett. 98, 095301 (2007).

[14] Jaskula, J.-C., Partridge, G. B., Bonneau, M., Lopes, R.,Ruaudel, J., Boiron, D., and Westbrook, C. I. Acoustic

analog to the dynamical casimir effect in a Bose-Einsteincondensate. Phys. Rev. Lett. 109, 220401 (2012).

[15] Stoferle, T., Moritz, H., Schori, C., Kohl, M., andEsslinger, T. Transition from a strongly interacting 1Dsuperfluid to a Mott insulator. Phys. Rev. Lett. 92,130403 (2004).

[16] Zenesini, A., Lignier, H., Ciampini, D., Morsch, O., andArimondo, E. Coherent control of dressed matter waves.Phys. Rev. Lett. 102, 100403 (2009).

[17] Sias, C., Lignier, H., Singh, Y. P., Zenesini, A., Ciampini,D., Morsch, O., and Arimondo, E. Observation ofphoton-assisted tunneling in optical lattices. Phys. Rev.Lett. 100, 040404 (2008).

[18] Alberti, A., Ivanov, V., Tino, G., and Ferrari, G. Engi-neering the quantum transport of atomic wavefunctionsover macroscopic distances. Nature Phys. 5, 547–550(2009).

[19] Ma, R., Tai, M. E., Preiss, P. M., Bakr, W. S., Simon,J., and Greiner, M. Photon-assisted tunneling in a bi-ased strongly correlated Bose gas. Phys. Rev. Lett. 107,095301 (2011).

[20] Chen, Y.-A., Nascimbene, S., Aidelsburger, M., Atala,M., Trotzky, S., and Bloch, I. Controlling correlated tun-neling and superexchange interactions with AC-drivenoptical lattices. Phys. Rev. Lett. 107, 210405 (2011).

[21] Haller, E., Hart, R., Mark, M. J., Danzl, J. G., Re-ichsollner, L., and Nagerl, H.-C. Inducing transport ina dissipation-free lattice with super Bloch oscillations.Phys. Rev. Lett. 104, 200403 (2010).

[22] Eckardt, A., Jinasundera, T., Weiss, C., and Holthaus,M. Analog of photon-assisted tunneling in a Bose-Einstein condensate. Phys. Rev. Lett. 95, 200401 (2005).

[23] Ho, T.-L. Spinor Bose condensates in optical traps. Phys.Rev. Lett. 81, 742–745 (1998).

[24] Ohmi, T. and Machida, K. Bose-Einstein condensationwith internal degrees of freedom in alkali atom gases. J.Phys. Soc. Jpn. 67, 1822–1825 (1998).

[25] Law, C. K., Pu, H., and Bigelow, N. P. Quantum spinsmixing in spinor Bose-Einstein condensates. Phys. Rev.Lett. 81, 5257–5261 (1998).

[26] Chang, M., Qin, Q., Zhang, W., You, L., and Chapman,M. Coherent spinor dynamics in a spin-1 Bose conden-sate. Nature Phys. 1, 111–116 (2005).

[27] Raghavan, S., Smerzi, A., Fantoni, S., and Shenoy, S. R.Coherent oscillations between two weakly coupled Bose-Einstein condensates: Josephson effects, π oscillations,and macroscopic quantum self-trapping. Phys. Rev. A59, 620–633 (1999).

[28] Vardi, A. and Anglin, J. R. Bose-Einstein condensatesbeyond mean field theory: Quantum backreaction as de-coherence. Phys. Rev. Lett. 86, 568–571 (2001).

[29] Mias, G. I., Cooper, N. R., and Girvin, S. M. Quantumnoise, scaling, and domain formation in a spinor Bose-Einstein condensate. Phys. Rev. A 77, 023616 (2008).

[30] Lipkin, H. J., Meshkov, N., and Glick, A. Validityof many-body approximation methods for a solvablemodel:(i). exact solutions and perturbation theory. Nu-clear Physics 62(2), 188–198 (1965).

[31] Teichmann, N., Esmann, M., and Weiss, C. Fractionalphoton-assisted tunneling for Bose-Einstein condensatesin a double well. Phys. Rev. A 79, 063620, Jun (2009).

[32] Boukobza, E., Moore, M. G., Cohen, D., and Vardi, A.Nonlinear phase dynamics in a driven Bose-Einstein junc-tion. Phys. Rev. Lett. 104, 240402, Jun (2010).

7

[33] Xie, Q., Rong, S., Zhong, H., Lu, G., and Hai, W.Photon-assisted tunneling of a driven two-mode Bose-Einstein condensate. Phys. Rev. A 82, 023616, Aug(2010).

[34] Jager, G., Berrada, T., Schmiedmayer, J., Schumm, T.,and Hohenester, U. Parametric-squeezing amplificationof Bose-Einstein condensates. Phys. Rev. A 92, 053632,Nov (2015).

[35] Zhang, W., Zhou, D. L., Chang, M.-S., Chapman, M. S.,and You, L. Dynamical instability and domain formationin a spin-1 Bose-Einstein condensate. Phys. Rev. Lett. 95,180403 (2005).

[36] Chang, M.-S., Qin, Q., Zhang, W., You, L., and Chap-man, M. S. Coherent spinor dynamics in a spin-1 Bosecondensate. Nature Phys. 1, 111 – 116 (2005).

8

PARAMETRIC EXCITATION OF A MANY-BODY SPIN SYSTEM: SUPPLEMENTARY INFORMATION

In this Supplementary Information, we provide ad-ditional details on the theoretical calculations of para-metric excitation in the spin-1 system using both semi-classical and quantum approaches, and we present ad-ditional simulations to compare with the data shown inFig. 3 of the main paper.

MEAN FIELD AND QUANTUMINTERPRETATION OF PARAMETRIC

EXCITATION

Coherent Oscillation Dynamics

We first discuss parametric excitation using semi-classical mean field theory. The excitation occurs whenthe quadratic Zeeman energy is modulated at integer di-visors of twice the natural coherent oscillation frequencyin the (θ,ρ0) phase space. The dynamics of the systemare governed by a set of differential equations for thefractional population ρ0 and the phase θ from [35, 36]

ρ0 = 4πcρ0√

(1−ρ0)2−m2 sin2θ (3)

θ = −2π(q−c(1−2ρ0)) (4)

−2πc(1−ρ0)(1−2ρ0)−m2√

(1−ρ0)2−m2cos2θ.

where m=〈Sz〉 and we have taken h→1. The spinorenergy of the system is given by

E = cρ0[(1−ρ0)+√

(1−ρ0)2−m2 cos2θ]+q(1−ρ0)

and has an oscillation period [35] of the form

T =1

π

√2√−qc

K(√

x2−x1

x3−x1)

√x3−x1

(5)

where K(k) is the elliptic integral of the first kind andxi are the roots of the differential equation

(ρ0)2 =(4π)2([E −q(1−ρ0)][(2cρ0 +q)(1−ρ0)−E ]−(cρ0m)2).

For a condensate prepared in mF =0, with magnetiza-tion conserved m=0, as is the case in our experiment,the roots xi are

xi ∈

2c−q+

√4c2−8cE+4cq+q2

4c,q−Eq

,

2c−q−√

4c2−8cE+4cq+q2

4c

.

In order to calculate the period T , we first calculate

√−qc√

2

√x3−x1 =

√q

2

(q2 +4qc(2ρ0−1)+4c2

)1/4≈√q2 +2qcQz

2

(1+

c2(1−Q2z)

(q+2cQz)2

)where Qz=2ρ0−1. The elliptical integral part of theperiod in Eq. (5)

K(x2−x1x3−x1

)≈K(0.01)≈ π2

(6)

Substituting these results back into the equation for theperiod, we obtain the natural coherent oscillation fre-quency in quadrature phase (θ,ρ0)

f0 =1

T=π√q2 +2qcQz

2

(1+

c2(1−Q2z)

(q+2cQz)2

)2

π

≈√q2 +2qcQz≈q+cQz (7)

where we have that q∼10|c| in our experiment.Parametric excitation of the system is achieved by pe-

riodically modulating the quadratic Zeeman term q inthe Hamiltonian. Efficient excitation occurs when themodulation frequency is an integer divisor of twice thenatural frequency of the system f= 2f0

n where n∈N. Inour system, the coherent oscillations occur at a magneticfield B=1 G and spinor dynamical rate c=−7.2(5) Hz.These parameters yield a range of natural frequencies

f0 =71.6×12−7.2x∈ [64.4, 78.8] Hz ∀ρ0∈ [0, 1]. (8)

with the most dominant excitation frequency correspond-ing to n=1

f=2f01∈ [128.8, 157.6] Hz ∀ρ0∈ [0, 1]. (9)

Parametric Excitation Theory

In our experiment, the system is prepared in the mF =0 state with magnetization m=0 and allowed to evolveat sufficiently high fields where q/|c|=10. Under theseconditions, Eq. (3) simplifies to

ρ0 = 4πcρ0(1−ρ0)sin2θ (10)

θ = −2π(q+c(2ρ0−1)).

Parametric excitation is then applied by modulating thequadratic Zeeman energy

q=q0 +εq0 sin(2πfmt−φ0)H[t−φ0/2πfm].

The Heaviside function and phase φ0 imply that themodulation starts at q0 and is active for t>φ0/2πfm=

9

φ0/ωm, where ωm=2πfm. We prepare an initial ρ0 athigh field using an rf pulse which initializes the quadra-ture phase θ0 =π/2. The system then freely evolvesfor t=φ0/ωm, advancing the quadrature phase ∆θ=−2π(q0 +2cx)φ0/ωm=−ω0φ0/ωm, followed by modula-tion of q, where we have implicitly assumed that for evo-lution at high field we can set ρ0 to be a constant so thatintegration of θ is straightforward. We therefore havetwo initial contributions to the quadrature phase priorto modulation, namely θ0 and ∆θ.

For evolution times t>φ0/ωm we can integrate θ inEq. (10) giving:

θ = θ0 +∆θ+

∫ t

0

dt′θ (11)

= π/2−ω0φ0/ωm−ω0t+2πεq0ωm

(cos(ωmt)−1)

where ω0 =2π(q0 +c(2ρ0−1)). Substituting the phaseinto the population dynamical equation 10, we obtain

ρ0 = 4πcρ0(1−ρ0)

×sin [2θ0−2ω0φ0/ωm−2ω0t (12)

+4πεq0ωm

(cos(ωmt)−1)

]= 4πcρ0(1−ρ0)

∑n

Jn(4πεq0ωm

) (13)

×sin((nωm−2ω0)t+δ+nπ/2)

where the phase δ=2θ0−2ω0φ0/ωm−4πεq0/ωm. InEq. 13, we have made use of the identity sin(A+B)=sin(A)cos(B)+cos(A)sin(B) along with the Jacobi-Anger expansions

cos(z sinα)=

∞∑n=−∞

Jn(z)cos(nα)

sin(z sinα)=

∞∑n=−∞

Jn(z)sin(nα).

Analyzing Eq. 13 gives us some insight into the pop-ulation dynamics as a function of the modulation pa-rameters. When ωm 6=2ω0/n, the time average of ρ0∝∑n sin(Ωt) is zero. When ωm=2ω0/n, the time aver-

age of the nth term in the expansion ρ0,n=4πcρ0(1−ρ0)Jn( 4πεq0

ωm)sin(δ+π/2) is non-zero. Therefore, only for

the case when ωm=2ω0/n is there sufficient couplingfrom the modulation to parametrically excite the system.

The behavior of the Bessel functions Jn

(4πεq0ωm

)also in-

dicates that the strongest coupling occurs for n=1 orωm=2ω0, a signature of parametric excitation.

The strength of the excitation is controlled by tuningε. When the system is modulated at ωm=ω0/2n we canfocus on short time dynamics (ωm−2ω0)t<π, since the

higher order terms are negligible due to time averaging,and expand Eq. 13 about ε=0. It can be shown thatthe coefficient in the expansion is exact up to O(ε)|nmax|.Using this fact, the population dynamical equation forρ0 can be rewritten as:

ρ0 = 4πcρ0(1−ρ0) (14)

×

[ ∞∑n=0

(−1)nx2n

(2n)!sinδ+

∞∑n=0

(−1)nx2n+1

(2n+1)!cosδ

]=4πcρ0(1−ρ0)(cosxsinδ+sinxcosδ)

=4πcρ0(1−ρ0)(sinδ+xcosδ+O(x)2

)where x=4πεq0/ωm. Depending on the value of δ, in-creasing the strength of the modulation ε will either en-hance or reduce the response of the system to the mod-ulation for short times. Furthermore, it is also apparentfrom Eq. 14, along with the expression for δ, that theresponse of the system is periodic with respect to the ini-tial phase of the modulation φ, and has fixed points forρ0 =0.5 and δ=nπ where n is an integer, both of whichagree with the experimental data shown in Fig. 3 of themain paper.

We now consider the strong coupling case where n=1 or ωm=2ω0, and the higher order terms |n|>1 ofJn( 4πεq0

ωm) are negligible, so that

ρ0 ≈ 4πcρ0(1−ρ0)(J−1(

4πεq02ω0

)sin(−4ω0t+2θ0−φ0−4πεq02ω0

− π2

)

+J0(4πεq02ω0

)sin(−2ω0t+2θ0−φ0−4πεq02ω0

)

+ J1(4πεq02ω0

)sin(2θ0−φ0−4πεq02ω0

+π

2)

)and after integration

ρ0 = 4πcρ0(1−ρ0)(J−1(

4πεq02ω0

)1

4ω0cos(−4ω0t+2θ0−φ0−

4πεq02ω0

− π2

)

+J0(4πεq02ω0

)1

2ω0cos(−2ω0t+2θ0−φ0−

4πεq02ω0

)

+ J1(4πεq02ω0

)sin(2θ0−φ0−4πεq02ω0

+π

2)t

)+ρ0(0).

The population ρ0(t)≈ρ0(0) if φ0 =1.37π and ωm=2ω0.

Fock States

An alternative way to view the parametric excitationis by considering transitions between the eigenstates ofthe many-body Hamiltonian. The energy correspondingto the oscillation frequency matches no single atom tran-sition. Rather, it approximately matches the energy dif-ference between two atoms in the mF =0 state and a pair

10

120 130 140 150 160 1700

π

2π

Frequency (Hz)

Pha

sede

lay(1/ω

m)

120 130 140 150 160 1700

π

2π

Frequency (Hz)P

hase

dela

y(1/ω

m)

120 130 140 150 160 1700

π

2π

Frequency (Hz)

Pha

sede

lay(1/ω

m)

120 130 140 150 160 1700

π

2π

Frequency (Hz)

Pha

sede

lay(1/ω

m)

FIG. 5: Excitation parameter map simulation Semi-classical simulations demonstrating parametric excitation forvarying modulation phase φ0 and modulation frequency f cor-responding to evolution times of 40 ms, 100 ms, 160 ms, and220 ms running clockwise.

of atoms in the mF =±1 states. Yet even this is not pre-cise enough as this energy separation also depends on thecollective state of the system varying from ∆E ≈2(q+c)for all atoms in the mF =0 state to ∆E ≈2(q−c) for allatoms in the mF =±1 states. These energy separationscan be calculated by diagonalizing the tridiagonal matrixgiven by

Hk,k′ = 2ck(2(N−2k)−1)+2qkδk,k′

+ 2c(k′+1)√

(N−2k′)(N−2k′−1)δk,k′+1

+ k′√

(N−2k′+1)(N−2k′+2)δk,k′−1 (15)

where c=c/2N and k is the number of pairs of mF =±1atoms in the enumeration of the Fock basis. The Fockbasis, |N,M,k〉, is also enumerated withN the total num-ber of atoms, and M the magnetization, both of whichare conserved by the Hamiltonian leaving all dynamicsin k. The off-diagonal contributions in Eq. 15 are due tothe many body interaction given by the cS2 term of theHamiltonian. This interaction results in mixing of theFock states, even in the high field limit. Without thisinteraction, there would be no transitions since the mag-

netic interactions, both linear and quadratic Zeeman, arediagonal in the Fock basis. In this picture, the integerdivisor frequencies of the spectrum correspond to manyphoton excitations of the system.

In the regime where q=qZB2>2c, we treat c as a per-

turbation and H(0)k,k′ =2qk with expansion coefficients

E(0)k = 2qk

E(1)k = 〈k |H′ |k〉=2ck (2(N−2k)−1)

and eigenenergy of the system

Ek=E(0)k +E

(1)k +O(2).

The resonance frequency between Fock states is the en-ergy difference between each Fock state

f =∂Ek∂k

=2q+2c(2N−8k−1)

≈ 2(q+cx)

where the last line corresponds to kN . To first order,the resonance frequency between Fock states is the sameas the frequency obtained from mean field theory Eq.7. The factor of two arises from the definition of theresonant frequency f=2f0.

PARAMETER VARIATION MAP SIMULATION

In order to compare the experimental data shown inFig. 3 of the main paper to theory, we perform foursemi-classical simulations at fixed evolution times thatdemonstrate excitation in the neighborhood of the 2f0resonance for different values of the modulation phaseφ0 and modulation frequency f . For each simulation,we initialize the state (ρ0,θs)=(0.5,π) and modulate thequadratic Zeeman energy q at different frequencies andphases. Details of the simulation method can be foundin [4, 5].

The change in population ∆ρ0 is calculated after exci-tation times of 40 ms, 100 ms, 160 ms, and 220 ms, asshown running clockwise in Fig. 5. These simulationscorrespond to the vertical slices shown in Fig. 3 of themain paper, and agree quite favorably. The white/orange(black/green) regions represent the positive (negative)changes in population which evolve and spiral to forma distinctive ‘yin-yang’ pattern. At the center of theplots (f,φ0)=(2f0,π) where 2f0≈143 Hz, ρ0 remains un-changed and the excitation shows anti-symmetry aboutthis point.