Arxiv:2108.01022V1 [Quant-Ph] 2 Aug 2021 to Strongly Interacting Gauge ﬁeld Theories of the Stan- Dard Model of Particle Physics [6–9]

UMD-PP-021-05

Engineering an Eﬀective Three-spin Hamiltonian in Trapped-ion Systems for Applications in Quantum Simulation

1 2 1 3, 4 5 6, BárbaraAndrade, Zohreh Davoudi, Tobias Graß, Mohammad Hafezi, Guido Pagano, and Alireza Seif ∗ 1ICFO-Institut de Ciencies Fotoniques, The Barcelona Institute of Science and Technology, Castelldefels (Barcelona) 08860, Spain. 2Maryland Center for Fundamental Physics and Department of Physics, University of Maryland, College Park, MD 20742, USA. 3Joint Quantum Institute and Department of Physics, University of Maryland, College Park, MD 20742, USA. 4Department of Electrical and Computer Engineering and Institute for Research in Electronics and Applied Physics, University of Maryland, College Park, MD 20742, USA. 5Department of Physics and Astronomy, Rice University, Houston, TX 77005, USA. 6Pritzker School of Molecular Engineering University of Chicago, Chicago, Illinois 60637, USA. (Dated: August 3, 2021) Trapped-ion quantum simulators, in analog and digital modes, are considered a primary candidate to achieve quantum advantage in quantum simulation and quantum computation. The underlying controlled ion-laser interactions induce all-to-all two-spin interactions via the collective modes of motion through Cirac-Zoller or Mølmer-Sørensen schemes, leading to effective two-spin Hamiltonians, as well as two-qubit entangling gates. In this work, the Mølmer-Sørensen scheme is extended to induce three-spin interactions via tailored first- and second-order spin-motion couplings. The scheme enables engineering single-, two-, and three-spin interactions, and can be tuned via an enhanced protocol to simulate purely three-spin dynamics. Analytical results for the effective evolution are presented, along with detailed numerical simulations of the full dynamics to support the accuracy and feasibility of the proposed scheme for near-term applications. With a focus on quantum simulation, the advantage of a direct analog implementation of three-spin dynamics is demonstrated via the example of matter-gauge interactions in the U(1) lattice gauge theory within the quantum link model. The mapping of degrees of freedom and strategies for scaling the three-spin scheme to larger systems, are detailed, along with a discussion of the expected outcome of the simulation of the quantum link model given realistic fidelities in the upcoming experiments. The applications of the three-spin scheme go beyond the lattice gauge theory example studied here and include studies of static and dynamical phase diagrams of strongly interacting condensed-matter systems modeled by two- and three-spin Hamiltonians.

I. INTRODUCTION freedom of the physical system, can be taken into account in enhancing and extending the simulator’s toolkit. With Quantum simulation is expected to be a promising av- a focus on the analog mode of operation of the quan- enue to revealing the rich dynamics of quantum many- tum device, such an objective is applied in this work to body systems, from strongly correlated electron systems trapped-ion quantum simulators [10–12]. These systems in material [1], to dense nuclear matter in the interior have demonstrated their value in recent years in success- of neutron stars [2,3], to coherent neutrino propagation ful simulation of quantum many-body systems [13–23], from core-collapse supernovae and early universe [4,5], some of which have started to challenge classical numerical methods. These simulations were mostly based on arXiv:2108.01022v1 [quant-ph] 2 Aug 2021 to strongly interacting gauge field theories of the Stan- dard Model of particle physics [6–9]. Quantum simu- an effective two-spin Ising Hamiltonian with single-spin lators can perform in analog, digital, or hybrid modes, field terms. Motivated by the need for an enhanced dy- with the underlying hardware architecture ranging from namics as exhibited in certain exotic spin systems as well condensed-matter-based systems to optical, atomic, and as in lattice gauge theories, we generalize this effective molecular systems. As the vibrant field of quantum sim- Hamiltonian to a Hamiltonian with simultaneous single-, ulation moves toward making large-scale noise-resilient two-, and three-spin dynamics, and examine its accuracy quantum simulators and quantum computers a reality, it through a numerical simulation with realistic experimen- is crucial to pursue a co-design process in which the speci- tal parameters. fications of the theoretical problem subject to the simula- While low-energy descriptions of interactions in na- tion, such as characteristics of interactions and degrees of ture primarily rely on two-body couplings, three- and higher-body couplings occur in many physical systems and impact the dynamics in a nontrivial manner. In high- ∗ The authors’ list is alphabetically ordered. energy physics, quantum field theories exhibit interacting 2 terms beyond two-field couplings, including gauge-matter scheme [102]. Effective interactions proportional to + + + + and gauge-field self interactions in gauge theories of the σi σj− + h.c. and σi σj σk + h.c. are generated, along z Standard Model [24–27]. In nuclear physics, three and with a single-spin Hamiltonian proportional to σi . Off- higher-nucleon interactions that are effectively generated resonant contributions constitute interactions that entan- from quantum chromodynamics (QCD) have proven to gle the spin and phonon degrees of freedom and are en- be important in providing an accurate description of nu- sured to be kept small with a careful choice of laser and clei and dense matter [28–32]. While present-day quan- trap parameters as will be discussed. Our strategy for tum simulators are not yet suited for simulating quan- tuning the relative importance of the two- and three-spin tum fields and their complex interactions in general (see dynamics is different from Ref. [100] and does not require e.g., Refs. [7–9, 23, 33–76] for some recent progress), the addressing the axial modes of motion. In particular, by possibility of inducing simultaneous interactions among introducing an almost “mirror” copy of the drives that more than two degrees of freedom of a quantum simulator induce effective single-, two-, and three-spin interactions, can open up many interesting possibilities. A more im- the single- and two-spin Hamiltonians can be significantly mediate application is in the realm of condensed matter suppressed. To validate the effective dynamics, we go physics where it is argued that spin systems with two- beyond a qualitative discussion of requisite experimen- and three-spin Hamiltonians exhibit novel phase dia- tal parameters and present a thorough numerical simu- grams and quantum phase transitions, signified by unique lation of a three-spin coupling scheme in the trapped- entanglement characteristics that are absent in systems ion system by examining the exact dynamics (within a with only two-spin couplings [77–90]. Such dynamics are highly accurate rotating-wave approximation), including interesting in the context of spin glass physics [86–88] the effects of spin-phonon contamination. High fidelities and quantum statistics [91]. Furthermore, quantum link for the three-spin dynamics, as well as the effectiveness models that exhibit local symmetry constraints, and are of the “mirror” drive in suppressing lower-body interac- argued to approximate infinite-dimensional gauge field tions, are established for realistic experimental values for theories in certain limits [92, 93], are spin systems (after laser intensities and frequencies, and for the axial trap mapping fermions to hard-core bosons) with multi-spin frequency that controls the relative strength of contribu- interactions. tions from multiple normal modes to the dynamics. Given such rich physics exhibited in systems that are To demonstrate an application of our scheme in simpli- governed by multi-spin interactions, and given the in- fying near-term quantum simulations of physical models terest in demonstrating the near-term benefits of ana- of interest, the example of the lattice Schwinger model log (and hybrid) quantum simulators, multitude of pro- within its simplest quantum-link-model representation is posals and strategies are put forward to directly engi- examined and feasible strategies are proposed for scaling neer correlated three-qubit interactions. These ideas in- the simulations given undesired contamination from pro- clude the use of Floquet engineering in superconduct- liferation of the normal modes of motions. Furthermore, ing qubits [94], optimization of periodic drives in weakly the question of the severity of local gauge-symmetry vi- driven quantum systems such as superconducting cir- olations in light of imperfect fidelity of realistic exper- cuits and molecular nanomagnets [95], perturbative gen- iments is analyzed through a crude model of interac- eration of multi-spin interactions with tunable couplings tions. The underlying scheme for Hamiltonian engineer- in a triangular configuration of an optical lattice of two ing of this work, through analog or gate-based implemen- atomic species [96], natural generation of three-body cou- tations, can be extended analogously to engineer three- pling in polar molecules driven by microwave fields [97], spin Hamiltonians with higher-dimensional spin degrees adiabatic passages with tunneling interactions among of freedom, as well as four- and higher-body spin inter- atoms in a one-dimensional optical lattice [98], coupling actions, with direct applications in lattice gauge-theory Rydberg-pair interactions and collective motional modes simulations. In the context of bosonic quantum field the- in trapped Rydberg ion systems [99], and resonantly driv- ories such as gauge theories, another interesting extension ing three-spin interactions through an adiabatic elimina- of the scheme is the engineering of effective spin-spin- tion of the off-resonant transitions involving spin-phonon phonon dynamics which will be explored in future work. couplings in trapped-ion systems [100]. The present paper is organized as follows. SectionII More specifically, the possibility of engineering a three- contains a detailed description of the scheme and its spin Hamiltonian in a trapped-ion quantum simulator numerical simulation. In particular, Sec.IIA offers a was proposed by Bermudez et al. in Ref. [100], where the semi-qualitative description of the underlying mechanism use of first and second order spin-phonon couplings en- for the generation of an effective three-spin Hamiltonian abled the generalization of the two-spin “phase gate”[101] without full technical details. A detailed derivation of z z z z z Hamiltonian σi σj to a three-spin Hamiltonian σi σj σk, the effective Hamiltonian for single- and multi-mode sce- where i, j, k are∼ the ion indices. An analogous strategy is narios and using single- and multi-drive schemes to con- applied in this work to induce effective correlated spin- trol the three-spin dynamics is provided in Sec.IIB. Sec- flipping transitions among (quasi)spins of three ions us- tionIIC presents the numerical simulations of the exact ing nearly resonant single- and double-sideband transi- unitary time evolution of the three-ion system, includ- tions, extending upon the well-known Mølmer-Sørensen ing phonon contributions to the dynamics, and explores 3 the range of laser intensities and frequencies, as well ions used in present-day quantum simulators. In a linear as trap characteristics, that result in maximum fideli- trap with a common confining potential and with long- ties. To present an application of the engineered three- range Coulomb interactions among the ions, the motion spin Hamiltonian in the context of quantum simulation, of the ions can be described in terms of a set collec- the example of the quantum-link-model representation of tive normal modes with quantized excitations, i.e., the U(1) lattice gauge theory is studied in Sec.III, and in- phonons. As a result, the displacement of each ion from cludes a discussion of an exact mapping of the degrees its equilibrium position, ∆ri, can be expressed in terms of freedom and the Hamiltonian and the scaling specifi- of phonon creation (am† ) and annihilation (am) operators cations and limitations in Sec.IIIA, an analysis of the associated with the motion along the three orthogonal expected dynamics in the quantum link model assum- principal axes of the trap, x, y and z, with commutation ing imperfect gauge-violating interactions in Sec.IIIB, relations [am, am0 ] = [am† , am† 0 ] = 0 and [am, am† 0 ] = δm,m0 . and a brief note on how the fully digital implementation Here, x and y denote the most-confined directions in the of the model may compare with its fully analog simula- trap, which will have the same normal-mode spectra for tion within the scheme of this work in Sec.IIIC. Conclu- symmetric traps commonly used. These will be denoted sions and a discussion of future directions are presented as transverse directions. The least-confined direction is in Sec.IV. denoted as z and is named the axial direction.2 In general, the indices m run from 1 to 3N, but for the coupling scheme described below, only the modes along one spa- II. ENGINEERING AN EFFECTIVE tial direction, selected by a light field, will be relevant. THREE-SPIN HAMILTONIAN WITH TRAPPED The index m will then run from 1 to N to denote the IONS relevant set of modes. One or multiple pairs of counter-propagating laser The effective Hamiltonian that couples three spins simul- beams, indexed by L, can be introduced with wave- taneously can be engineered using tailored ion-laser inter- vector difference ∆kL, frequency difference (beatnote) actions. SectionIIA presents basic theoretical aspects of ∆ωL, and phase difference ∆φL, to stimulate the two- the ion-laser dynamics in a linear Paul trap [103], along photon Raman transition among the two states of the with a qualitative description of an enhanced Mølmer- qubit with a Rabi frequency ΩL. The beams can ad- Sørensen scheme that leads to an effective three-spin dress the ions globally, or for more flexibility in devising Hamiltonian. SectionIIB presents the exact relations for interactions individually, in which case both amplitude the effective Hamiltonians, including the desired three- and frequency control of the beams are required (and an spin Hamiltonian and accompanying single- and two-spin index i must be attached to laser’s parameters). Hamiltonians, along with their detailed derivation and a The Hamiltonian describing the ion-laser interactions description of the errors. SectionIIC presents a numer- can be written as [12] ical study of the full evolution compared with the one anticipated from the effective picture in an experimen- N nL i∆ωLt+i∆ϕL+i∆kL ∆ri tally realistic three-ion system, and identifies regimes in Hint. = ΩLe− · the tunable parameters of the scheme that are best suited i=1 L=1 X X x y z for suppressing the single- and two-spin terms, giving rise (α0Ii + α1σi + α2σ + α3σi ) + h.c. (2) × i to predominantly three-spin dynamics. Here, index L runs over nL pairs of Raman beams. α0, α1, α2, and α3 are constants related to spin- A. Ion-laser dynamics and a qualitative description dependent forces on the two states of the qubit [12] of the scheme and are controlled by the intensity, geometry, and po- larization of the laser beams. These are set to α0 = α = α = 0 and α = 1 , associated with a com- Consider a system of N ions confined in a radio- 2 3 1 2 mon choice. The operator ∆kL ∆ri can be written frequency Paul trap. The Hamiltonian of the system in 3N · as m=1 ηm,j(am + am† ), where ηm,i are Lamb-Dicke pa- absence of any interactions with external lasers can be 2 1 ∆k written as rameters defined as η = | | b . Here, b P m,i 2Mionωm m,i m,i N 3N are the (normalized) normal-modeq eigenvector compo- ω0 z 1 H = σ + ω (a† a + ). (1) nents between ion i and mode m, and Mion denotes free 2 i m m m 2 i=1 m=1 the mass of the ion. For optimal control, the ion-laser X X system must operate in the Lamb-Dicke regime where Here, σ is a Pauli operator acting on the space of the ions’ (quasi)spin, namely the qubit. ω0 is the qubit frequency which is of the order of a few GHz for typical 2 x, y and z in san-serif font correspond to spatial physical coor- dinates of the trap, hence labeling the Cartesian components of the wave-vector of the laser beams. x, y, and z in italic font 1 Planck’s constant ~ is set to unity throughout. correspond to the Bloch-sphere axes in the qubit Hilbert space. 4

H xx Carrier (a) / i j (b)

Beatnote frequency

+ + + H i j (c) / k (d) Carrier

Beatnote frequency

FIG. 1. (a,b) Traditional Mølmer-Sørensen scheme based on a pair of bichromatic laser beatnotes off-resonantly driving first- order spin-phonon couplings with symmetric detuning (±δ), giving rise to an effective spin-spin interaction. The two-ion case is shown for simplicity. (c,d) Generalized Mølmer-Sørensen scheme to generate an effective three-spin coupling. A second-order blue sideband is driven with twice the detuning (2δ) as the first-order red (−δ) sideband. As shown in (c), this process creates two virtual phonons with a second-order process and annihilates the same number of phonons through two first-order processes. Note that only two out of several possibilities are depicted. In all subfigures, Ωr and Ωb are the Rabi frequencies of the red and blue beatnotes, respectively. ω0 is the qubit frequency, and ω [≡ ωcom] is the transverse center-of-mass frequency.

2 1/2 3 η (a + a† ) 1. To exclusively couple the Ra- arise interactions proportional to m,i h m m i man beams to one set of normal-mode excitations, ∆kL + + can be set along only a single principal axis. σi am, σi−am† , σi am† , σi−am, (4) 1 x y In the interaction picture with respect to the free where σ± = (σ iσ ). These describe spin excitation i 2 i ± i Hamiltonian Hfree, the interacting Hamiltonian becomes of ion i accompanied by either absorption or emission of a quantum of motion in the mth mode, the so-called blue and red sideband transitions, respectively. On the N nL P3N −iωmt † iωmt ΩL i η ,i(a e +a e ) other hand, when µ = ωm but µ ωm ηm,iΩL, the H0 = e m=1 m m m 6 ± | − | int. 2 × motional modes are only virtually excited, and the ﬁrst- i=1 X LX=1 order contributions in the Lamb-Dicke parameter give i(∆ωL ω0)t+i∆φL + e− − σi + h.c., (3) rise to an eﬀective Hamiltonian proportional to

σxσx. (5) where it is assumed that ∆ω ω ω . The prime i j | L − 0| 0 on H is to denote that this Hamiltonian is in the inter- As will be shown in Sec.IIB, in the limit where the de- action picture. Tuning the Raman-beams beatnote to ω0 tuning from the center-of-mass mode, δ, satisfies: δ | | ≡ (ω0 ωm) leads to carrier (sideband) transitions, which µ ωcom ∆m for a typical mode splitting ∆m, the two- ± 0 | − | 2 can be derived by expanding Eq. (3) to (η )( (η)), ηcomΩr Ωb O O spin coupling can be approximated as δ , neglect- where η is the short-hand notation for the Lamb-Dicke 2 th ∆k parameter defined above. Similarly, an n -order side ing counter-rotating terms. Here, η = | | com 2MionωcomN band transition can be achieved by expanding Eq. (3) to q (ηn) and setting ∆ω = ω nω . O L 0 ± m Consider one set of Raman beams. Setting the detun- 3 ing µ ∆ω ω of the Raman beams to ω , there Upon tunig the Raman-beam phase differences properly. ≡ L − 0 ± m 5 is the Lamb-Dicke parameter associated to the center-of- we introduce two pairs of Raman beams, as described in mass normal mode, and Ωr(Ωb) are the Rabi frequencies the previous section. The Rabi and beatnote frequencies associated with the red and blue sideband beatnotes, re- associated with pair I are denoted as Ω and µ ∆ω r r ≡ 0 − spectively. This is the familiar Mølmer-Sørensen scheme, ωI , respectively, and those associated with pair II are that is schematically described in Fig.1(a-b). In par- denoted as Ωb and µb ω0 ∆ωII . Additionally, the ticular, it is shown how phonon emission and absorption following relations ≡ − cooperate to create a coherent coupling between the |↓↓i and two-ion states. µr = ωcom δ, (8) |↑↑i − − The key feature of this work, which resembles that µb = 2µr = 2ωcom + 2δ, (9) first proposed in Ref. [100], is the generalization of the − Mølmer-Sørensen scheme to create an effective three-spin will be applied to ensure resonant three-spin transitions, Hamiltonian. This scheme relies on second-order contri- as will be seen shortly. This choice of beatnote frequen- butions in the Lamb-Dicke parameter, interactions that cies justifies the use of ‘r’ and ‘b’ subscripts for each pair, are proportional to corresponding to (single) red and (double) blue sideband + + + + transitions of the qubit, respectively. With this setup, σi aman, σi aman†, σi am† an, σi aman, and ignoring contributions from counter-rotating waves σi−am† an†, σi−am† an, σi−aman†, σi−am† an†. (6) oscillating with a frequency of ωm or higher, the Hamil- tonian of the system at (η) in the interaction picture Here, a spin excitation can be associated with the cre- becomes O ation or annihilation of two phonons at the same time. As shown in Fig.1(c-d), this scheme amounts to driv- N N RWA i i(ωm+µr )t + ing the ion chain using two beatnotes with asymmetric H = ηm,iΩre a† σ + h.c., int. (η) 2 m i detunings (for example, 2δ for the blue and δ for the O i=1 m=1 − X X red sideband) so that a single second-order spin-phonon (10) excitation process can be combined with two first-order and that at (η2) reads absorption processes, giving rise to a coherent resonant O coupling between and states. This leads to |↓↓↓i |↑↑↑i N N N an effective three-spin Hamiltonian proportional to RWA 1 i(ω +ω µb)t H = η η Ω e− m n− int. (η2) −4 m,i n,i b + + + O i=1 m=1 n=1 σi σj σk , σj−σj−σk−, (7) X X X + amanσ + h.c., (11) 4 2 i ηcomΩr Ωb with a coupling that can be approximated by δ2 , in the limit of small detuning δ from the center-of-mass where RWA in the superscripts refers to the rotating- mode. This process generates resonant single- and two- wave approximation applied. Only the transverse normal spin interactions as well, and if the goal is to achieve a modes along the x direction of the trap are addressed with pure three-spin Hamiltonian, these contributions must be both pairs of the beams, hence the sums over modes m suppressed, particularly given that their strength is larger and n run from 1 to N. than that of the three-spin interactions. For this reason, Performing a Magnus expansion [104] of the time- a second drive is added to cancel out the undesired single- evolution operator, keeping only contributions that 1 are resonant and further ignoring contributions from and two-spin contributions that are proportional to δ and will enhance the three-spin contribution that is propor- counter-rotating terms (see discussions below), give rise 1 to an effective (time-independent) spin Hamiltonian. Ex- tional to δ2 . This additional drive is composed of two sets of Raman beams with asymmetric detunings of op- plicitly, posite sign compared with the first drive i.e., 2δ for i R t H0 (τ) +H0 (τ) dτ the blue and δ for the red sideband. This cancellation∼ − U(t, 0) = e− 0 int. O(η) int. O(η2) ∼ T is only exact for the center-of-mass mode contribution. i H(σ)+H(σσ)+ H(σσσ) t e− eff eff eff , (12) The exact relations in the next section will provide fur- ≈ ther essential detail of the single- and two-drive schemes where denotes time ordering. The single-spin Hamil- in the single- and multi-mode scenarios, and the numeri- tonianT has the form cal study of Sec.IIC will investigate the accuracy of the 2 effective three-spin dynamics. (σ) 1 Ω 1 H = η2 r (n + ) eff 4 m,i ω + µ m 2 i m m r X X 1 Ω2 B. Derivation of the effective Hamiltonian b η2 η2 (n + n + 1) + − 4 ω + ω + 2µ m,i n,i m n m,n m n r To obtain the exact relations for the effective dynamics X z 0 0 † 0 †0 of the ion-laser system within the scheme just described, ηm,iηn,jηm ,iηn ,jδωm0 +ωn0 ,ωm+ωn am an aman σi . one can start from the interaction-picture Hamiltonian in m0,n0 X Eq. (3). To engineer an effective three-spin Hamiltonian, (13) 6

The two-spin Hamiltonian reads certain lattice gauge theories. The single- and two-spin dynamics arising from Eqs. (13) and (14) can in gen- 2 (σσ) 1 Ωr 1 eral be significant compared with the three-spin dynam- Heff = ηm,iηm,j + −4 ωm + µr 2 ics from Eq. (15). Nonetheless, there is a simplifying i i=j m m,n X X6 X X limit in which a scheme can be devised to cancel the un- 2 Ωb + desired single- and two-spin contributions entirely. Be- ηm,iηn,iηm,jηn,j (nm + nn + 1) σi σj−, ωm + ωn + 2µr fore introducing this scheme, let us consider the following (14) simplification. If the spacing among the modes is much larger than δ, the only mode contributing to the effective and the desired three-spin Hamiltonian takes the form dynamics is that close to the lasers beatnote frequency. This can be achieved for the transverse modes if trap’s (σσσ) 2 + + + Heff = ηm,iηn,jηm,kηn,k ΩrΩb σi σj σk axial confinement is increased. In the scheme adopted m,n i,j,kX X above, this is the center-of-mass mode, see Eqs. (8) and 3µ + ω + 2ω (9). Ignoring contributions from all other modes, the ef- r m n . (15) 24(µr + ωm)(µr + ωn)(2µr + ωm + ωn) fective single-, two-, and three-spin Hamiltonians become 2 (σ) 1 ηcom 2 1 In these relations, nm(n) a† am(n) is the occupation- H = Ω (n + ) ≡ m(n) eff −4 δ r com 2 number operator for the corresponding phonon mode, i X and the condition µb = 2µr is imposed. In an 1 − η2 Ω2 (n2 + n + 1) σz, (16) individual-addressing scheme, the Rabi frequencies can − 8 com b com com i be set independently at the location of each ion, hence the following replacements must be taken into account: 2 2 2 2 (σσ) 1 η Ω Ω in Eq. (13), Ω Ωr(b),iΩr(b),j in com 2 r(b) r(b),i r(b) Heff = Ωr + → 2 → 4 δ Eq. (14), and Ω Ωb Ωr,iΩr,jΩb,k in Eq. (15). i j=i r X X6 Corrections to the→ above picture can be attributed to 1 2 2 1 + at least three sources. First, off-resonant contributions, η Ω (ncom + ) σ σ−, (17) 2 com b 2 i j i.e., those introducing oscillatory time dependence in the exponent of U(t, 0) (instead of linear time dependence 4 that is a feature of resonant contributions), introduce (σσσ) η H = com Ω2Ω σ+σ+σ+ + h.c., (18) corrections to the effective Hamiltonians. They include eff 16 δ2 r b i j k first-order terms in the Lamb-Dicke parameter and can i,j,kX iδt have a time-dependence as slow as e . At long times and respectively. ncom is the occupation-number operator for 4 as long as ηΩr(b)/δ remains small, these contributions the center-of-mass mode, i.e., ncom acom† acom. can be ignored. Such a slow time dependence occurs at The above single-mode approximation≡ makes it clear higher orders in the Lamb-Dicke parameter as well, but that the single- and two-spin couplings are odd under δ they will average to zero compared with the resonant δ while the three-spin coupling is even. Therefore, the→ contributions at long times, i.e., when t 1/δ. Second, net− single- and two-spin contributions can be canceled, there are higher-order resonant and off-resonant terms by introducing two extra Raman-beam pairs (I0 and II0) that are suppressed by positive powers of Lamb-Dicke pa- each with Rabi and beatnote frequencies Ωr0 = Ωr, µr0 = rameter compared with the contributions retained. The ωcom + δ, and Ωb0 = Ωb, µb0 = 2ωcom 2δ, respectively. third source of corrections to the effective Hamiltonians −However, such detunings give rise to resonant− sideband are from counter-rotating terms. While being resonant, transitions when applied along with the first set of lasers. these generally scale with inverse powers of mode fre- To avoid such undesired contributions and still achieve quency or mode-frequency separation (and combinations the desired cancellation, one can set the detuning of the of), which are suppressed compared with those included second set of lasers from the center-of-mass mode, δ0, to in Eqs. (13)-(15), which scale as inverse powers of δ, probe incommensurate with respect to that of the first set, δ. vided that δ is the smallest frequency scale in the set- The cancellation of single- and two-spin couplings can be ting. Given these sources of corrections to the effective still achieved by scaling the Rabi frequencies accordingly. picture above, it is important to benchmark the scheme Explicitly, assuming δ0 = qδ for non-rational constant q in a complete numerical simulation, and obtain the range Ω0 = √qΩ , µ0 = ω + qδ, of experimental parameters with which the effective dy- r r r − com namics above can be achieved. Such an investigation is Ωb0 = √qΩb, µb0 = 2ωcom 2qδ, (19) conducted and described in detail in the next section. − While the simultaneous one-, two-, and three-spin dynamics may be relevant in describing a generic quantum 4 1 Note that for the center of mass mode, bcom,i = √ {1, 1, ··· , 1}, many-body system, the overarching goal of our study is N to achieve a pure three-spin Hamiltonian for applications hence ηcom,i can be replaced with ηcom for each i, where ηcom is in quantum computing, and in quantum simulation of defined after Eq. (5) 7

Once again, given the approximation made, it is impor- the two pairs needs to be introduced through an appro- tant to investigate how close the simplified dynamics is to priate scale factor q = 1. With this, the Hamiltonian the full dynamics, as will be studied in the next section. including both drives6 is given by The original scheme introduced earlier will be denoted N in the following as a “single-drive” scheme while the en- com iΩr iδt iqδt + H = η e− + √qe a† σ hanced scenario is denoted as a “two-drive” scheme. 2drive 2 com com i i=1 X Ωb 2 2iδt 2iqδt 2 + ηcom e + √qe− acomσ + h.c. (21) C. Numerical simulations − 4 i For convenience, we choose Ωb = 2Ωr/ηcom, such that The effective Hamiltonian obtained from the Magnus the red- and blue-sideband processes appear with equal expansion relies on the assumption that only the reso- strengths. Additionally, the scale factor is set to q = 1.3, nant combination of second- and third-order processes which as will be evident shortly, appears to be a good contribute to the system dynamics. In the present sec- choice. Assuming a transverse trap frequency ωx = 5 tion, we validate and corroborate the analytical results 2 × 2π MHz, and a recoil energy of ωrec ∆k /(2Mion) = of Sec.IIB by performing a numerical simulation of the 26 2π kHz, the Lamb-Dicke parameter≡ | for| N = 3 ions system dynamics. × 5 evaluates to ηcom = ωrec/(Nωx) 0.0416. For dif- The numerical treatment requires discretization of ≈ ferent choices of detuning δ and Rabi frequency Ωr, the time, with time steps smaller than the time scale of the dynamics are simulatedp under the two-drive single-mode fastest processes in the Hamiltonian. Therefore, to fa- Hamiltonian in Eq. (21), with the results shown in Figs.2 cilitate the simulation, the rotating-wave approximation and3. in Eqs. (10)-(11) is assumed, which is justified by the Specifically, Fig.2 shows in black the evolution of the fact that δ µ ωcom ωm. The dynamics gener- z | | ≡ | − | spin expectation value σi as a function of time if the ated by these Hamiltonians can differ from the desired system is initially preparedh i in a state 0 corre- effective spin dynamics due to off-resonant terms which sponding to all spins pointing down and|↓↓↓i the phonon ⊗ | i vac- are neglected in the effective picture. These terms man- uum. In all cases, this evolution exhibits an oscillatory ifest themselves by changes in the phonon number dur- behavior which is a manifestation of the three-spin inter- ing the time evolution. Moreover, as explained above, actions. In addition, blue and red lines in Fig.3 show the effective dynamics becomes a pure three-spin inter- the evolution if the system is initially prepared in the action when a second drive cancels the undesired res- 0 state. In this case, the dynamics is strongly onant single- and two-spin contributions in Eqs. (16)- suppressed,|↑↓↓i ⊗ | i demonstrating that the cancellation of two- (17). However, this cancellation scheme is complete only body terms indeed works very well. with respect to the contributions from the center-of-mass To compare more quantitatively the three-spin dy- mode. Therefore, the presence of more than one mode namics in Fig.2 with the prediction from Eq. (18), yields another source of error if the purpose is the design one can consider the period of the oscillations in the of a pure three-spin Hamiltonian. full and the effective dynamics. In the effective Hamil- tonian, the three-spin coupling from a single drive is 4 2 1drive ηcomΩr Ωb J 2 , and therefore, from the combination 1. Single-mode approximation 3 16δ of the two≡ drives, one expects three-spin Rabi oscillations with a period In order to discern between these two sources of errors, one can first study the effect of off-resonant terms π T3 = 2drive , (22) in the single-mode approximation, where the contribu- 6J3 tions from all modes except for the center-of-mass mode 2drive 1drive 1 are discarded. Up to second order in η, the Hamiltonian where J J (1+ q− ) and the factor 6 in the 3 ≡ 3 of one Raman pair reads denominator accounts for all permutations of the three spins. For the parametersp specified in Fig.2(a-d), the N expected oscillations periods are: (a) T3 22.8 ms, (b) com iΩr iδt + ≈ H = ηcome− a† σ T 19.7 ms, (c) T 11.4 ms, (d) T 19.2 ms. These 1drive 2 com i 3 ≈ 3 ≈ 3 ≈ i=1 values agree well with the observed period, showing for X Ω all cases the presence of three-spin interactions with ap- b η2 e2iδt a2 σ+ + h.c. (20) − 4 com com i proximately the strength predicted by Eq. (18). Another A second Raman pair with opposite detuning would cancel the undesired single- and two-spin terms from the first pair, but it would also produce a new resonant second- 5 Values that are accessible in current experimental trapped-ion order term from the combination of the two pairs. There- systems operating using Ytterbium ions in radio-frequency Paul fore, as introduced in Eq. (19), an asymmetry between traps. 8

1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 -0.8 -0.8 -1 -1 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

z com FIG. 2. Evolution of hσi i under the single-mode two-drive Hamiltonian H2drive in Eq. (21), where i = 1, 2, 3 is the ion index, for different choices of detuning δ and Rabi frequency Ωr. The evolution of the spins from an initial state |↓↓↓i ⊗ |0i probes the system’s three-spin interactions, and it is shown by the black curve (for ion 1), as well as fully hidden (identical) curves for ions 2 and 3. feature of the evolution plots is that, qualitatively, the os- spin dynamics occurs for the largest detuning employed, cillations are quite different in the four scenarios shown in which is consistent with the highest fidelity of three-spin Fig.2. Clean sinusoidal oscillations, with an amplitude oscillations in Fig.2(d). ranging from -1 to 1, are seen in panel (d), which among For a direct comparison between single- and two-drive all shown cases has the largest detuning (δ = 4 2π kHz). × schemes, in Fig.4 we have simulated the dynamics of In this case, the suppression of two-body terms, as shown the Hamiltonian with only one Raman pair, Eq. (20). in Fig.3(d), also works best. In panels (a) and (b) of As expected, when the system is initialized in a state Fig.2, similarly fast three-spin dynamics as in panel (d) with all-parallel spins, the single-drive scheme in Fig.4(a) are achieved, but at smaller values of the detuning. It is leads to similar three-spin oscillations as the two-drive seen that for δ = 2 2π kHz in panel (b) the results are × scheme in Fig.2(d), yet with approximately twice the still in good accordance with sine-like oscillations, but period. However, when the initial state contains anti- the fidelity of the evolution becomes significantly poorer parallel spins, as considered in Fig.4(b), the single-drive at δ = 1 2π kHz in panel (a). Similarly, the fidelity × Hamiltonian exhibits very fast two-spin dynamics which decreases if the oscillation are sped up by increasing Ωr. is absent in the two-drive scheme in Fig.3(d). In this respect, one can compare panels (b) and (c), that This interpretation of the spin dynamics is corrobo- exhibit identical detuning, but different Rabi frequencies. rated by the number of phonons generated during the Clearly, the price for the speed-up in (c) is a significant evolution, as a quantitative measure for the deviation loss of fidelity, also resulting in an increase of spurious from the ideal evolution in which the systems remains in two-body interactions, as shown in Fig.3(c). Specifi- the phonon vacuum throughout. The average number of cally, in Fig.3 we demonstrate the two-spin dynamics by phonons over a time duration of 40 ms is given in Ta- initializing the system in a mixed-spin state 0 . blesI andII for the scenarios shown in Figs.2,3 and4. As is seen in Fig.3(d) the maximum suppression|↓↑↓i of ⊗ two- | i Smaller detunings and/or larger Rabi frequencies lead to 9

1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 -0.8 -0.8 -1 -1 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

z com FIG. 3. Evolution of hσi i under the single-mode two-drive Hamiltonian H2drive in Eq. (21), where i = 1, 2, 3 is the ion index, for diﬀerent choices of detuning δ and Rabi frequency Ωr. The evolution of the spins from an initial state |↓↑↓i ⊗ |0i probes the system’s two-body dynamics, and it is shown by the yellow, blue and red curves for ions 1, 2 and 3, respectively. Here, the yellow curve is fully hidden by the red one. larger deviations from the phonon vacuum. The number As a result of these considerations, one might conclude of phonons oscillates around this time-averaged value but that a larger detuning is the better choice. However, the its peak remains below one at all times. anticipation from the more realistic multi-mode scheme is that a small detuning of the center-of-mass mode, to-

1 2π {δ, Ωr}[kHz] {1, 6}{2, 10}{2, 12}{4, 16}

1 2π {δ, Ωr}[kHz] {4, 16} |↓↓↓i ⊗ |0i 0.43 0.19 0.44 0.08

|↓↓↓i ⊗ |0i 0.05 |↓↑↓i ⊗ |0i 0.14 0.08 0.14 0.05

|↓↑↓i ⊗ |0i 0.05 TABLE I. The average values of the center-of-mass phonon occupation over a time duration of 40 ms for the three-spin evolution revealed using an all-spin-parallel initial state and TABLE II. The average values of the center-of-mass phonon the two-spin evolution revealed using a mixed-spin initial occupation over a time duration of 40 ms for the three- and state. The evolution occurs under the single-mode two-drive two-body evolutions revealed by diﬀerent initial states under com com Hamiltonian H2drive in Eq. (21) for the detunings and Rabi the single-mode single-drive Hamiltonian H1drive in Eq. (20) frequencies used in Figs.2 and3. The Hilbert space is re- for the parameters considered in Fig.4. The Hilbert space is stricted by truncating the phonon occupation of the center- restricted by truncating the phonon occupation of the center- of-mass mode at 6. of-mass mode at 6. 10

Dicke parameter reads

N N 1 i 0 multi iωmt iµr t iµr t + H2drive = ηm,ie Ωre + Ωr0 e am† σi 0.8 2 i=1 m=1 0.6 X X N 0.4 0 1 iµbt iµ t ηm,iηn,i Ωbe + Ω0 e b 0.2 − 4 b n=1 0 X -0.2 i(ωm+ωn)t + e− amanσi + h.c., (23) -0.4 × -0.6 -0.8 Here, Ωr0 ,Ωb0 , µr0 , and µb0 concern the second Raman- -1 0 5 10 15 20 25 30 35 40 beam pair. They are defined in Eq. (19) in terms of the parameter q. Simulations of the evolution under the multi-mode Hamiltonian were performed under the same conditions 1 as for the single-mode case, i.e., with q = 1.3, ω = 0.8 rec 0.6 26 2π kHz, and N = 3 ions. In addition to the transverse× trap frequency, ω = 5 2π MHz, the axial trap 0.4 x × 0.2 frequency ωz also needs to be fixed, as it determines 0 the bandwidth of the phonon spectrum, i.e., transverse -0.2 modes separations. The results for different ωz values are -0.4 shown in Fig.5 for an all-spin-parallel initial spin state, -0.6 and in Fig.6 for different mixed-spin initial states. For -0.8 the detuning and the Rabi frequencies, those parameters -1 which worked well in the single-mode approximation have 0 5 10 15 20 25 30 35 40 been chosen, i.e., δ = 2 2π kHz with Ωr = 10 2π kHz and δ = 4 2π kHz with× Ω = 16 2π kHz. × × r × From an initial state 0, 0, 0 (where the latter |↓↓↓i ⊗ | i z ket corresponds to the vacuum of three phonon modes), FIG. 4. Evolution of hσi i under a single-mode (center- of-mass) Hamiltonian with a single drive corresponding to one observes three-spin oscillations with good fidelity, com H1drive in Eq. (20), where i = 1, 2, 3 is the ion index, for de- as shown in Fig.5(a)-(c). This is not surprising, since tuning and Rabi frequency shown. (a) The evolution from an two-spin interactions are automatically suppressed by the initial state |↓↓↓i ⊗ |0i probes three-spin interactions. They choice of initial state, and off-resonant contributions from give rise to three-spin oscillations, represented by the black the other modes are negligible due to their large detun- curve for ion 1. Ions 2 and 3 behave identically. (b) The two- ing. Notably, the evolution of the three ions is almost spin dynamics is probed by considering the evolution from an exactly the same in these cases, that is, one can hardly initial state |↓↑↓i ⊗ |0i. Yellow, blue, and red curves corre- discern the black curve in all panels of Fig.5 for ion 1, spond to ions 1, 2, 3, respectively. Note that the yellow curve from the green and the purple curves for ion 2 and ion 3. is fully hidden by the red one. These plots should be compared with Figs.2(d) and3(d), that is the evolution under An interesting observation is that the oscillation am- two-drive scenario using the same laser parameters. plitude when evolving under the multi-mode scenario is larger than that for the single-mode approximation with similar parameters (compare e.g., Fig.2(b) and Fig. gether with a large phonon bandwidth, will be a feasi- 5(b)). This might seem counterintuitive given the ex- ble strategy to suppress the undesired effect of the other act cancellation of single- and two-spin dynamics in the modes. In particular, within the two-drive scheme these single-mode approximation. An explanation of this be- modes spoil the cancellation of the two-spin dynamics. havior could be a destructive interference of the two fac- The optimal set of parameters for an experimental im- tors which contribute to the suppression of the oscilla- plementation should, therefore, be concluded from the tions’ amplitude hence decrease in fidelity: i) the effective analysis of the multi-mode scenario, as will be presented single-spin terms in Eq. 13, and ii) noise, i.e., any terms next. which are not accounted for by the effective Hamiltonian, that involve various spin-phonon couplings. In the single- mode calculation, only the latter contribution appears as the former is fully suppressed in the two-drive scheme. 2. Multi-mode simulation In the multi-mode case, both factors contribute, and it is plausible that they can partially cancel each other out. Next, let us turn our attention to the multi-mode Let us now probe the two-body dynamics by initializ- Hamiltonian, which up to second order in the Lamb- ing the system in a mixed-spin state. Since the second 11

=2 2⇡=2kHz,2⇡⌦rkHz= 10, ⌦ 2=⇡ 10kHz 2⇡ kHz =2 2⇡ kHz=2, ⌦2r⇡=kHz 10, ⌦2⇡=kHz 10 2⇡ kHz =4 2⇡=4kHz, 2⌦⇡rkHz= 16, ⌦r2=⇡ 16kHz 2⇡ kHz ⇥ ⇥ ⇥r ⇥ ⇥ ⇥ ⇥ r ⇥ ⇥ ⇥ ⇥ ⇥ !z =1.0!z =12⇡ .MHz0 2⇡ MHz !z =1.5 !2z⇡=1MHz.5 2⇡ MHz !z =1.5!z 2=1⇡ MHz.5 2⇡ MHz 1 ⇥ ⇥ 1 ⇥ ⇥ ⇥1 ⇥ 1 1 0.8 1 1 1 0.81 1 1 0.81 0.8 0.8 0.60.8 0.8 0.8 0.60.8 0.8 0.8 0.80.6 0.6 0.6 0.40.6 0.6 0.6 0.40.6 0.6 0.6 0.60.4 0.4 0.4 0.20.4 0.4 0.4 0.20.4 0.4 0.4 0.40.2 0.2 0.2 00.2 0.2 0.2 0.20 0.2 0.2 0.20 0 0 -0.2 0 0 0 -0.20 0 0 -0.20 -0.2 -0.2 -0.4-0.2 -0.2 -0.2 -0.4-0.2 -0.2 -0.2 -0.2-0.4 -0.4 -0.4 -0.6-0.4 -0.4 -0.4 -0.6-0.4 -0.4 -0.4 -0.4-0.6 -0.6 -0.6 -0.6 -0.8-0.6 -0.6 -0.6 -0.8-0.6 -0.6 -0.6 -0.8 -0.8 -0.8 -0.8 -1-0.8 -0.8 -0.8 -0.8-1 -0.8 -0.8 -1 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40 -1 -1 -1 -1 -1 -1 -1 -1 -1 0 5 010 515 10 0 20 15 5 25 10 20 30 15 25 35 2030 40 2535 3040 35 40 0 5 010 515 010 20 515 25 10 20 30 15 25 35 20 30 40 25 35 30 40 35 40 0 5 010 515 0 10 20 5 15 25 10 20 30 15 25 35 20 30 40 25 35 30 40 35 40 (a) t[ms] (b) t[ms] (c) t[ms]

z multi FIG. 5. Evolution of hσi i under the multi-mode Hamiltonian H2drive in Eq. (23) from an initial state |↓↓↓i ⊗ |0, 0, 0i. The evolution of the spins probes the system’s three-spin interactions. Different panels consider different choices of detuning δ, Rabi frequency Ωr, and axial trap frequency ωz. Different colors (black, green, purple) correspond to the evolution of each of the different spins (i = 1, 2, 3), which, in contrast to the single-mode case, take different numerical values, but still can barely be discerned from one another.

1 1 1 0.8 0.8 0.8 0.6 0.6 0.6 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 -0.2 -0.2 -0.2 -0.4 -0.4 -0.4 -0.6 -0.6 -0.6 -0.8 -0.8 -0.8 -1 -1 -1 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

z multi FIG. 6. Evolution of hσi i under the multi-mode Hamiltonian H2drive in Eq. (23) from the initial state |↓↑↓i⊗|0, 0, 0i. (first row) and from the initial state |↓↓↑i ⊗ |0, 0, 0i (second row). The evolution of the spins from these initial states probes the system’s two-spin interactions. Different columns consider different choices of detuning δ, Rabi frequency Ωr, and axial trap frequency ωz. In each panel, different colors (yellow, blue, red) correspond to the evolution of each of the different spins (i = 1, 2, 3). In the case of a parity-symmetric initial state (first row), the yellow curves (i = 1) is hidden by the red curve (i = 3).

pair of Raman beams is designed to only cancel the two- , shown in Fig.6(d)-(f). The “tilt” mode induces spin terms stemming from the center-of-mass mode, one two-spin|↑↓↓i interactions only if the first and the third spin observes an enhanced amount of two-spin dynamics in are different, and therefore this mode, which is next to the multi-mode scenario, but one also has to differen- the center-of-mass mode in the phonon spectrum, does tiate between parity-symmetric initial states, like not contribute in the parity-symmetric case. Therefore, shown in Fig.6(a)-(c), and asymmetric initial states|↓↑↓i like the two-spin dynamics in Fig.6(a)-(c) remains strongly 16

6 12 4

2 0.98 1 2π {δ, Ωr, ωz}[kHz] {2, 10, 1000}{2, 10, 1500}{4, 16, 1500} 16

0.9414

|↓↓↓i ⊗ |0, 0, 0i 0.20 0.19 0.08 12 0.90

max 10 F |↓↑↓i ⊗ |0, 0, 0i 0.07 0.08 0.04 8 0.86 6 |↓↓↑i ⊗ |0, 0, 0i 0.09 0.08 0.05 0.824 7 8 91011 12 2 TABLE III. The average values of the total phonon occupa- 0.9816 tion in all modes over a time duration of 40 ms for the three 14 and two-body evolutions under the multi-mode Hamiltonian 0.94 multi 12 H2drive in Eq. (23). In all cases, the average values for occupation of the non-center-of-mass modes are less than 0.01. The 10 [ms] Hilbert space is restricted by truncating the phonon occupa- 0.90 8 tion of the center-of-mass mode at 6. The phonon occupation max t of the other two modes is truncated at 2. 6 0.86 4

0.822 suppressed. However, this is not the case for asymmet- 0.987 8 91011 12 ric initial states, and much faster two-spin dynamics is observed, see Fig.6(d)-(f). ⌦r/(2⇡) [kHz] 0.94 Nevertheless, by increasing the bandwidth of the phonon spectrum, the detuning of the modes other than FIG. 7. Maximum fidelity for the GHZ-state preparation, the center-of-mass mode can be increased, and thereby along with the0.90 time to reach such a fidelity, for different val- their effect will decrease. This strategy works best if the ues of Rabi frequency Ωr, at a fixed detuning δ = 2 × 2π kHz and axial frequency ω = 1500×2π kHz. The GHZ state con- detuning of the center-of-mass mode is kept small. To 0.86 z sidered here is defined as |GHZ( 3π )i = √1 (|↓↓↓i − i |↑↑↑i), enhance the phonon bandwidth, the axial confinement 2 2 and the phonon degrees of freedom are traced out in the can be increased, e.g. by changing ωz from 1 2π MHz × intermediate-states0.82 density matrices. to 1.5 2π MHz, see Figs.5 and6. While this change has 7 8 91011 12 almost× no effect on the three-spin evolution, the two-spin dynamics is significantly slowed down. as keeping the detuning small. As an optimal lower limit Moreover, from comparison between panels (b)-(e) and for the detuning, the simulation carried out in this sec- (c)-(f) of Fig.6, one can recognize the effect of an increase tion determines the value δ 2 2π kHz. Limitations of the detuning from 2 kHz to 4 kHz (together with a to the axial frequency are more≈ × general, given by zig- proportional increase of the Rabi frequency to keep the zag deformation of the ion chain, which happens when three-spin dynamics similarly fast in both cases). This ωx 0.86 ωz > 0.73 N − [105]. For the case of N = 3 and the increase of detuning, while slightly improving the fidelity transverse frequency of this simulation, the bound on the of the three-spin evolution (as already seen in the single- allowed axial frequency is 2.67 MHz. mode case), leads to significantly faster two-spin dynam- ≈ ics. As anticipated earlier, this shows that in order to achieve a quantum simulator with purely three-spin in- 3. Preparation of a GHZ state teractions, one must keep the detuning as small as possible. To conclude this section, we consider the possibility For completeness, the average phonon number over a of using the three-spin coupling scheme to prepare a duration of 40 ms for the different combinations of pa- maximally entangled state of three spins. Specifically, rameters and initial states as used in Figs.5 and6 are by subjecting an all-parallel spin state to the three- reported in TableIII. As is seen, the average phonon spin Hamiltonian for the proper amount of time, i.e., occupation remains well below one. Furthermore, not 2drive π/(2J3 ), a GreenbergerHorneZeilinger (GHZ)-like surprisingly in all cases the most occupied mode is the state≈ GHZ(ϕ) = 1 ( +eiϕ ) can be obtained, center-of-mass mode, as it the beams are more closely √2 | i | ↑↑↑i | ↓↓↓i 3π 6 tuned to the center-of-mass mode than the rest. where in our scheme the phase angle is ϕ = 2 . In summary, the implementation of the three-spin Hamiltonian is possible with an appropriate choice of the experimental parameters. Mainly, this requires increas- 6 To set the relative phase of the two states equal to unity, a trivial ing the axial trap frequency as much as possible, as well single-qubit phase gate can be applied to one of the ions at the 13

To determine how well this GHZ state is reached, one QCD and has long served as a testbed for benchmarking can calculate the fidelity F as the system evolves defined quantum simulation of lattice gauge theories on quan- as tum hardware [23, 63, 71, 72]. In the Kogut-Susskind 2 formulation [107], the model consists of fermion-gauge F (t) tr √ρGHZ ρ(t) √ρGHZ , (24) interactions, electric-field contribution, and a staggered ≡ q fermion mass term: where ρGHZ denotes the density matrix corresponding to Nstag 1 Nstag 1 3π − − the state GHZ( ) . Since this is a pure state, the sim- 2 2 H = x ψ†Uiψi+1 + h.c. + E + | i 3π 3π U(1) i i plified relation F (t) = GHZ( 2 ) ρ(t) GHZ( 2 ) can also i=1 i=1 h | | i X X be used. ρ(t) denotes the reduced density matrix of the Nstag time-evolved state after tracing out the phonon degrees i µ ( 1) ψ†ψi. (25) of freedom. The maximum fidelity F is plotted in − i max i=1 Fig.7 for different choices of the Rabi frequency Ω . We X r Fermions ψ sit on site i, while the conjugate-variable also plot the time t at which the maximum fidelity i max pairs E ,U belong to the link originating from site i, is reached. It can be seen that a fidelity > 0.96 can be i i see Fig.{ 8. N} denotes the number of staggered sites. reached within a time less than 10 ms for Ω = 8 2π kHz. stag r Rescaled dimensionless couplings x and µ8 are used such Slight increases of the Rabi frequencies allow× to reduce that the Hamiltonian is expressed in dimensionless unit. the preparation time to smaller than 5 ms, keeping F max Open boundary conditions will be considered such that still as large as approximately 0.92. A comparison of the the electric-field flux into the lattice, E , is fixed. The fidelities achieved in digital universal implementations of 0 continuum limit of the expectation values of the observ- quantum circuits using only two-qubit entangling gates able O for a fixed mass over coupling ratio is obtained compared with the three-qubit entangling operation pro- via a double-ordered limit: lim lim O , a posed here will be discussed more closely in Sec.IIIC. x Nstag limit that will not be considered→∞ here. The→∞ statesh i are characterized by bosonic and fermionic quantum num- (g) III. APPLICATIONS IN QUANTUM bers. Explicitly, bosonic quantum number ni is associ- SIMULATION: A LATTICE GAUGE THEORY ated with the discrete spectrum of a quantum rotor sat- EXAMPLE (g) (g) (g) (g) (g) isfying Ei ni = ni ni and Ui ni = ni + 1 (g) | i | i | i | (f) i with ni Z. The fermionic quantum number ni is Quantum simulation of quantum many-body systems can associated∈ with the fermionic occupation at each site, potentially benefit from a direct implementation of three- (f) which satisfies ni 0, 1 . An unoccupied (occupied) spin interactions. Examples include simulating certain odd (even) site represents∈ { } the presence of an electron lattice gauge theories whose dynamics can be mapped to (positron). The physical states are those annihilated by multi-spin Hamiltonians [92, 106], as well as spin systems 1 the Gauss’s law operator Gj = Ej Ej 1 ψj†ψj + 1 with two- and three-spin interactions which can exhibit − − − 2 − ( 1)j at each site. nontrivial phase diagrams [77–90]. Among the features −A finite-dimensional formulation of the same U(1) the- of the scheme of this work that can be useful in simulating such physical systems are the possibility of a direct ory is described by a quantum link model [92, 93]. Within simulation of the three-spin dynamics without the need this formulation, and upon performing a Jordan-Wigner for digitalization of the evolution, as well as the ability to transformation of the staggered fermions, the Hamilto- tune the relative strength of two- and three-spin interac- nian becomes Nstag 1 Nstag 1 tions. In this section, an example in the context of lattice − − 7 + + 2 gauge theory will be presented. This section also inves- HQLM = J τi Si τi−+1 + h.c. + Sz + tigates the benefit of implementing three-qubit gates in i=1 i=1 X X place of two-qubit gates to achieve higher-fidelity decom- Nstag position of certain time-evolution operators in quantum µ ( 1)iσz, (26) − i simulation. i=1 X where τi are Pauli matrices at each site, and S is a spin operator. While the exact Schwinger model is recovered A. A U(1) quantum link model with the three-spin in the limit of S , for low-energy quantities, small S coupling scheme values provide accurate→ ∞ approximations to the exact theory, as shown in e.g., Ref. [108].9 To map this Hamilto- The U(1) lattice gauge theory in 1+1 dimensions, i.e., 1 nian to the trapped-ion effective Hamiltonian, the spin- 2 the lattice Schwinger model, is a valuable prototype of

8 Not to be confused with the lasers’ frequency in the previous beginning of the operation. section. 7 We refer the reader to Refs. [77–90] for applications in the context 9 Note that only in the limit of S → ∞, J approaches the original of condensed-matter theory. coupling x, see Ref. [108]. 14

i i+1

E ,U E ,U { i i} { i+1 i+1} Lattice Schwinger model ......

⌧i Si ⌧i+1 Quantum link model ......

Ions in a linear trap 2i 1 2i 2i+1 am ......

Collective normal modes used to Internal states of the ion are used to encode engineer three-spin couplings. the dynamic of fermions and gauge links.

FIG. 8. The mapping of the degrees of freedom of the lattice Schwinger model to a quantum link model, that is ultimately mapped to (quasi)spin degrees of freedom of a linear chain of trapped ions, which are addressed with given sets of laser beams as described in the text. The ion-laser interactions induce an eﬀective three-spin Hamiltonian via virtually-excited phonons.

limit of the quantum links will be considered, and Si in as well as on the last ion in the chain. Explicitly, Eq. (26) will be replaced by the Pauli operators at each N stag 1 site. The electric-ﬁeld contribution is then trivial and one 2 − is left with matter-gauge interaction in form of a near- x x x σ2Nstag 1 σ4i 1σ4i (28) U ≡ − − est neighbor three-spin operator, as well as the staggered i=1 mass term. Y Such a Hamiltonian can be mapped to the eﬀective is applied at the beginning and the end of the simulation. Hamiltonian of the trapped-ion simulator. Explicitly, The Hamiltonian in the transformed basis is

Nstag 1 Nstag − Nstag 1 Nstag (rot.) + + + z − HIon = J σ2i 1σ2iσ2i+1 + h.c. µ σ2i 1. + + i z − − − HIon = J σ2i 1σ2iσ2−i+1+h.c. +µ ( 1) σ2i 1. i=1 i=1 − − − i=1 i=1 X X (29) X X (27) For an analog simulation of the dynamics governed by Note that N = 2Nstag 1 ions are needed to encode this Hamiltonian, one can use the two-drive scheme dis- − the dynamics with open boundary conditions. The ba- cussed in the previous section, in addition to a Stark- sis states are simply the direct product of eigenstates of shift beam to induce the evolution under the mass the Pauli operator σz, while the physical states are those term. Specifically, the gauge-matter term proportional annihilated by the Gauss’s law operator, which in this to three-spin interactions can be mapped to the effec- 1 z z z i formulation reads: Gi = 2 σ2i σ2i 2 σ2i 1 ( 1) tive Hamiltonian in Eq. (15). Acting on three ions, the z − − − − − − with σ0 fixed by (open) boundary conditions. A dia- two-drive scheme produces the desired three-spin inter- grammatic representation of the mapping between de- actions, while suppressing single- and two-spin terms. grees of freedom in the original lattice Schwinger model, For N 3, however, the ions at locations 2i + 1 for ≥ the quantum link model, and the trapped-ion simulator i 1, ,Nstag 2 need to be addressed with two is shown in Fig.8. sets∈ of{ Raman··· beams,− } as these ions participate in three- In order to apply the scheme of the previous section spin interactions with their left and right neighbors si- to simulate the dynamics governed by Eq. (27), one first multaneously. This can be achieved through local ad- needs to perform a local unitary transformation consist- dressing of the ions, hence individual control of Rabi fre- ing of π-rotations on every other odd-even pairs of ions, quencies. Besides the need for individual addressing, one 15

needs to mitigate the problem of deviating from the ef- The mass term can be implemented by applying, si- fective three-spin dynamics given multiple drives on each multaneously with the lasers that induce the three-spin third ion in the chain. One strategy is to apply lasers coupling, a Stark shift on participating ions, i.e., ions that address different transverse modes of the chain in 2i 1 for i 1, ,N , corresponding to a longi- − ∈ { ··· stag} an alternate pattern. Explicitly, ions 2i 1, 2i, 2i + 1 tudinal effective magnetic field Bz applied to the ions. can be addressed by beams with ∆k {= ∆−kˆx while ions} Alternatively, properly shifted red and blue sideband 2i + 1, 2i + 2, 2i + 3 can be addressed with beams with transitions can effectively implement such a magnetic- { } 10 ∆k = ∆kyˆ, for i 1, 3, ,Nstag 3 , see Fig.8. field interaction, as described, e.g., in Refs. [11, 51, 109]. In this scheme,∈ there { ··· will still be− two} types of unde- However, arbitrarily large masses cannot be simulated sired contributions to the dynamics even with two sets if long evolution times are desired, as undesired cross- of alternate drives. First, there will be terms propor- coupling terms with the lasers implementing the three- tional to phonon creation and annihilation operators in spin coupling occurs. By requiring the model parameters i(δ δ˜)t both modes, that go as e − or slower, where δ and to match those in experiment for the effective magnetic- δ˜ are the detuning from the center-of-mass frequency of field term throughout the evolution, that is by setting each set of the modes. Thus, multiple drives with sim- µtQLM = Bzt, the undesired terms will be small for 2 ilar detunings on a single ion will give rise to undesired BztηΩr/δ = µtQLMηΩr/δ 1 and Bztη Ωb/δ = | 2 | | | | | spin-phonon entanglement. These contributions can be µtQLMη Ωb/δ 1. | | made off-resonant by setting the two detunings suffi- ciently differently, and by compensating for the required equal couplings by adjusting the corresponding Rabi fre- B. A numerical study of inexact dynamics with a quencies. The second type of terms can result from cross- crude model coupling among e.g., ions in the sets 2i 1, 2i, 2i + 1 and 2i + 3, 2i + 4, 2i + 5 , as the beams{ aligned− with the} { } It is interesting to investigate if presently-available same principal axis of the trap are used for the two non- trapped-ion simulators can reveal nontrivial nonper- adjacent sets. Such cross-couplings can result in long- turbative features of strong dynamics in the lattice range effective spin interactions along the chain. They Schwinger model despite the imperfect fidelity of imple- can be made off-resonant by adopting different detun- menting effective three-spin dynamics as quantified in our ings for each set of the beams. To increase the distance numerical study. Let us consider a 7-ion and an 11-ion among the ions addressed by lasers along the same prin- system, as these system sizes are currently available in cipal axes, axial modes of motion can be addressed too, both digital and analog modes (see e.g., Refs. [11, 110– but these modes are lower in frequency, and the second 112]). These systems can encode a quantum link model sideband drive with respect to the center-of-mass mode with Nstag = 4 and Nstag = 6 staggered matter sites, would unavoidably induce the first sideband transitions respectively. The numerical study of Sec.IIC with re- of several modes other than the center-of-mass mode, alistic experimental parameters established that the pe- which complicates the effective dynamics. Increasing the riod of the undesired two-spin evolution is at worse 20 axial trapping frequency, as well as detuning close the times larger than that of the desired three-spin evolution,∼ highest frequency (“zig-zag”) axial mode, can help with considering contributions from all the modes. This indi- this limitation. cates an effective three-spin coupling that is at least 20 Another complication with scaling up the scheme of time larger than the two- (and one-) spin couplings—a∼ the previous section to simulate a large instance of the consequence of the effectiveness of the two-drive scheme quantum link model is related to the proliferation of the in suppressing the undesired contributions. The accu- normal modes and the increased contribution from modes racy of this scheme, of course, drops as a function of other that the center-of-mass mode to the dynamics—an time, particularly as a result of noise, coupling to envi- effect that is anticipated to worsen the cancellation of ronment, and decoherence—effects that will be left out the undesired single- and two-spin contributions. As a of our analysis and can only be carefully quantified in an result, the fidelity of the simulation compared with the experimental implementation. An accurate prediction of exact effective three-spin dynamics degrades as a func- the dynamics in the quantum link model with the full tion of the number of the trapped ions. As performing Hamiltonian requires inputting a comprehensive model a numerical study of the full dynamics to estimate the of interactions including phonons in chains longer than expected outcome is costly and reaches the limits of clas- three ions, and hence is beyond a simple numerical in- sical methods quickly, experimental implementations on vestigation. Nonetheless, one can still come up with a the quantum simulator are the only means by which to crude model of interactions to describe the quantum link assess the performance of the scheme for larger system model under an imperfect Hamiltonian that qualitatively sizes. resembles that of the full Hamiltonian. For this purpose, we consider the Hamiltonian in Eq. (26) with the addition of an all-to-all two-spin interactions that are uniform 10 With open boundary conditions, ions 1 and 2Nstag − 1 do not in strength, and to be conservative, are taken to be 10 need to be addressed with two sets of orthogonal Raman pairs. times and 5 times weaker than the desired three-spin in- 16

2 4

e + e e + e e e e+ e+ (a) (b) e e+ 2 4 Nstag =4 Nstag =6 �� mesons mesons �� ! �� ! �� string string 3 P �� 1 P �� tQLM tQLM (c) (d)

�� i i

| Nstag =4 | Nstag =6

i �� i ��

G �� G �� | | �� i i X X

h �� h �� -�� -��

tQLM tQLM e

e 3 e+ + 1 e

FIG. 9. (a) The (absolute value) of the overlap between a time-evolved string state |ψstri and a fully-occupied mesonic state −iHQLMtQLM |ψmesi, Pstring→mesons ≡ | hψmes|e |ψstri |, as a function of the (scaled dimensionless) time tQLM for a lattice with Nstag = 4 fermion sites, corresponding to N = 7 ion sites. For the dashed and dotted curves, the Hamiltonian must be 0 00 replaced with the non-exact HQLM and HQLM Hamiltonians, respectively. (b) The same quantity plotted for Nstag = 6 fermion sites, corresponding to N = 11 ion sites. The graphical representations of states, both in terms of the electron, positron, and electric-ﬁeld strings and in terms of the (quasi)spins of each corresponding ions are shown for the initial string state, and for a fully-occupied mesonic state whose probability amplitude is maximum at the points denoted. (c) The expectation P value of the lattice sum of the (absolute value of) the Gauss’s law operator between a time-evolved string state, h i |Gi|i ≡ iHQLMtQLM 1 PNstag−1 −iHQLMtQLM hψstr|e |Gi|e |ψstri for Nstag = 4 fermion sites, corresponding to N = 7 ion sites. For the 2Nstag−3 i=1 0 00 dashed and dotted curves, the Hamiltonian must be replaced with the non-exact HQLM and HQLM Hamiltonians, respectively. (d) The same quantity as in (c) for Nstag = 6 fermion sites, corresponding to N = 11 ion sites. The maximum breakdown of P the Gauss’s law corresponds to h i |Gi|i = 1.

teractions. Additionally, single-spin interactions on all mass. Explicitly, ions are included to modify the mass term with uniform

coeﬃcients that are 10 and 5 times weaker than the true 2Nstag 1 − ( ) x + H0 00 = H + σ σ− + h.c. + QLM QLM g ( ) j k 0 00 j,k=1 Xj6=k

2Nstag 1 µ − σz, (30) g ( ) j 0 00 j=1 X 17

with g0 = 10 and g00 = 5. One can now ask, assum- HQLM0 , the gauge-symmetry violation remains at a few- ing this crude model captures the imperfection of the percent level at early times. For HQLM00 , the violation can effective model obtained in Sec.IIB, whether interesting be significant, for both lattice sizes considered, although nontrivial features of the quantum link model of the lat- it remains well below one at early times. Despite the tice Schwinger model, such as string breaking and pair breakdown of gauge invariance, the qualitative behavior creation after a quench, can be observed in experiment. of the evolution under exact Hamiltonian appears to be Furthermore, given the gauge-symmetry violating inter- robust with respect to the perturbations introduced (for actions in play, will the degree of symmetry breaking in a few meson-state revival cycles), making this model a the evolved states be small with a gauge-invariant ini- suitable first case study in the upcoming implementa- tial states, such that qualitative features of a constrained tions.12 dynamics are not fully lost? Figure9 shows the result of one such study. The initial gauge-invariant state consisting of positive and negative C. Digital versus analog electric charges at the far ends of lattice with an electric- field string (quantum-link) attaching them, the so-called As demonstrated in Sec.IIC, the three-spin(qubit) “string” state, is evolved in time, and its overlap is mea- coupling scheme can be used to prepare a maximally- sured with a fully-occupied “mesonic” state consisting entangled three-qubit state, such as the GHZ state, with of Nstag/2 electron-positron pairs, connected by proper high fidelity. The two-qubit entangling gates, i.e., the electric-field fluxes to satisfy Gauss’s law. Explicitly, the conventional Mølmer-Sørensen (MS) gate, have achieved quantity increasingly high fidelities ( 0.995 for a three-qubit sys- ∼ iHQLMtQLM tem for comparison [112]) while the fidelity of the three- Pstring mesons ψmes e− ψstr (31) → ≡ | h | | i | spin coupling obtained here is at best 96% for realistic experimental parameters. Furthermore,∼ the GHZ-state is calculated via exact diagonalization of the Nstag = 4 preparation only requires two MS gates compared with and Nstag = 6 site theories, where ψstr denotes the | i a single three-spin gate. Therefore, one will not find a string state and ψmes denotes the mesonic state. Be- | i greater benefit in using the three-spin gate in creating sides the evolution with the exact Hamiltonian HQLM (in GHZ states. There are, however, other situations for solid curves), the two modified Hamiltonian H0 and QLM which a direct implementation of a unitary of the type H00 are also considered (in dashed and dotted curves, + + + QLM iα(σ σ σ +h.c.) respectively). The plots indicate that the generation of e− i j k can potentially reduce the time of the a significant amplitude for a mesonic state out of an ini- operation, and hence the fidelity of the implementation. tial string state after a quantum quench is retained for An example is the quantum simulation of the U(1) imperfect Hamiltonians, and is only slowly diminished quantum link model presented in this section. The po- as function of time, particularly for the smaller pertur- tential of a fully analog implementation of the correlated bation. The reason for the decline in the meson-state three-spin dynamics in the quantum link model is a pri- amplitude is the leakage to the vast unphysical Hilbert mary motivation for the three-spin scheme of this work, space that can occur since the modified Hamiltonians do so that digitalization error of the time-evolution opera- not respect the Gauss’s law constraints. tors could be avoided. Nonetheless, while the implemen- iα(σ+σ+σ++h.c.) To reveal the degree of gauge-symmetry violation in tation of the operator e− i j k on three qubits, the nonperturbative dynamics, Fig.9 also depicts the with a decomposition to MS gates as

quantity y y iα(σ+σ+σ++h.c.) iα σ σ σx iα σy σxσy e− i j k = e 4 i j k e 4 i j k x y y iHQLMtQLM iα σxσxσx iα σ σ σ Gi ψstr e e− 4 i j k e 4 i j k h | |i ≡ h | y y i iπ σ σz iα σxσx iπ σ σz X = e− 4 i j e 4 j k e 4 i j Nstag 1 iπ y y iα z y iπ y y − 4 σi σj 4 σj σk 4 σi σj 1 iHQLMtQLM e− e e G e− ψ , (32) y y i str iπ σxσ iα σz σx iπ σxσ 2Nstag 3 | | | i e− 4 i j e 4 j k e 4 i j − i=1 − X iπ σxσx iα σz σy iπ σxσx e− 4 i j e 4 j k e 4 i j , (33) with the Gauss’s law operator Gi defined above in the 1 spin- 2 formulation of the quantum link model. For a is exact and incurs no digitalization errors, at least twelve gauge-invariant state, therefore, Gi = 0, while for two-qubit MS gates are needed, in addition to a number h i | |i a state with maximum violation of the Gauss’s law con- of single-qubit gates. The fidelity of this operation, given 11 P straint, Gi = 1. As is seen in the figure, for the best MS gates reported, is 0.99512 = 0.942. There- h i | |i fore, here a single-shot implementation∼ of the operator in P

11 Hence the normalization 1 adopted in the definition of 2Nstag−3 P 12 h i |Gi|i. Note that with open boundary conditions, the last The robustness of gauge-invariant dynamics in latice gauge the- fermionic site is left open and is not counted toward the sum. ories is thoroughly studied in Refs. [113, 114]. 18 the left-hand side of Eq. (33) appears to be comparable tions, a thorough check of the scheme is presented by or even slightly higher in fidelity compared with the dig- conducting a numerical simulation of the exact dynam- itized scheme, at least for the first few cycles of the pure ics (that from all orders in the Lamb-Dicke parameter three-spin evolution, and for certain laser and trap pa- and including phonons dynamics) in a rotating-wave ap- rameters identified in our numerical study. proximation. This investigation reveals the experimen- Performing side-by-side comparison of the fidelity of tal parameters for Rabi and beatnote frequencies of the operations with only MS gates versus using three-spin Raman beams, as well as the axial trapping frequency, gates, however, cannot be rigorously done as this stage such that with the use of the second drive, the fidelity for the following reasons. First, while our numerical of generating a GHZ state with only three-spin dynam- study gives a reliable estimate of the expected perfor- ics reaches & 95%, and single- and two-spin dynamics mance, it does not include the anticipated experimental remains at least 20 times slower that the three-spin ∼ imperfections arising from noise (e.g., laser and trap in- dynamics. Furthermore, the phonon occupation in par- stabilities) and coupling to environment (e.g., heating, ticipating modes remains far below one for simulations and undesired emission and absorption processes). Such that start in a phonon-less state. These promising ana- effects are increasingly controlled and diminished in ex- lytic and numerical results motivate future experimental perimental platforms, as is evident from achieved high- implementation of this scheme, and can potentially sim- fidelity MS operations in chains of ions with various sizes, plify quantum simulation of spin systems with multi-spin but are nonetheless needed to be taken into account for interactions. The important case of a lattice gauge theory a fair comparison with the MS gate-based performances. is investigated in this context, and the specifications of Second, a high degree of optimization and adjustments is an upcoming implementation of the quantum link model applied to devise complex pulse-shaping protocols in the of the U(1) gauge theory in 1 + 1 dimensions within this case of a MS gate to minimize the spin-phonon entan- scheme are detailed. In particular, it is shown through glement (i.e., that can occur already at leading order in a crude model of interactions that the near-term demon- the Lamb-Dicke parameter) throughout the gate opera- strations for systems of 10 ions can still reveal inter- ∼ tion [115–119], hence increasingly high-fidelities reported esting constrained dynamics of the lattice gauge theory, in recent state-of-the-art demonstrations. Such an opti- such as string breaking, despite the anticipated undesired mization has not been investigated in this work and hence gauge-violating interactions revealed by our numerical the fidelity reported for the three-spin gate corresponds simulation. The comparative performance of analog sim- to the simplest implementation. Devising a pulse-shaping ulation using engineered three-spin dynamics and that protocol for the three-spin gate is far more complex as based on solely MS two-spin gates in digitized dynamics it requires a simultaneous minimization of both the first- depends on the task and model, and a decisive conclusion and second-order contributions in the Lamb-Dicke pa- will need to await experimental benchmarks. rameter, but can potentially lead to the same improve- A few extensions and improvements over the scheme ment in the three-spin gate fidelity as with the MS gate. developed here can be enumerated, along with potential Such an avenue will be left to future studies. applications: While our effective Hamiltonian realizes maximally spin-flipping transitions in a given basis, it is easy IV. CONCLUSION AND OUTLOOK z z z to see that interactions proportional to σi σj σk can also be realized in an analogous way, extending two- Correlated evolution of the (quasi)spin of three ions in a qubit geometric phase gates [120, 121] to correlated trapped-ion quantum simulator can be achieved via an three-qubit operations. In fact, a number of para- extension of the standard Mølmer-Sørensen scheme. Ex- sitic undesired single- and higher-spin interactions plicitly, a resonant combination of two single-sideband coupled to phonon operators will be absent in the and one double-sideband transitions can be combined to three-qubit phase-operation scheme since couplings effectively couple three spins via virtual phonons. Res- among the spin operators at the same ion will be onant single- and two-spin effective transitions are also absent in the Magnus expansion, i.e., the commu- z induced, and can either be used to simulate spin sys- tation among σi operators at the same site is van- tems with competing two- and three-spin interactions ishing.13 in a magnetic field, or if only pure three-spin dynam- The three-spin Hamiltonian of this work is engi- ics is desired, be eliminated with another set of Raman beams driven with carefully tuned Rabi and beatnote neered with semi-global beams on every triple of frequencies. The effective Hamiltonian in the single- nearest-neighbor ions such that with each opera- mode approximation (when the lasers are detuned closely tion, interactions are of short range. On the other to the single and double sidebands of a single mode so that the contribution to the dynamics from the nearby 13 modes is small) and in the multi-mode approximation To achieve interactions proportional to σz only, a bias term must are derived, and the leading corrections to the effective be eliminated [12] through setting Raman beam detunings and picture are qualitatively identified. Given these correc- polarizations properly, as discussed in Refs. [51, 122]. 19

hand, global Raman beams that address all the ions theories, extending the recent hybrid analog-digital in the chain can generate three-spin couplings with proposals [75, 127, 128]. A detailed theory and nu- all-to-all interactions. Various coupling profiles for merical investigation will be required to establish if the interactions, both the single-, two-, and three- this scheme can be a viable path toward this goal. spin couplings, can be engineered by tuning the Ra- man beatnote frequencies, see Eqs. (13)-(15), and While an experimental demonstration of the scheme if individual addressing of the ions is a possibility, of this work is the next immediate goal, all these afore- by tuning the Rabi frequencies as well. Such an mentioned directions will be valuable to investigate in optimization of parameters to achieve certain cou- future studies. These align with the overarching goal of pling profiles is customary in MS-based analog sim- enhancing and expanding trapped-ion simulator toolkit ulation schemes, see e.g., Refs. [51, 123, 124]. Fur- for applications beyond what is possible today. thermore, for digital gate-based applications, pulse- shaping techniques such as those applied to generate optimized MS gates [115–119] can be employed in the three-spin scheme to minimize spin-phonon entanglement, and optimize the operation of the ACKNOWLEDGMENTS associate three-qubit gate, as mentioned before. Quantum-simulation and quantum-computing pos- We are grateful to Christopher Monroe for valuable dis- sibilities can be expanded in trapped-ion systems cussions. We acknowledge Andrew Shaw’s involvement by addressing more than two internal hyperfine at early stages of this work. BA and TG further ac- levels of the ions, hence effectively introducing a knowledge discussions with Maciej Lewenstein. BA ac- higher-spin degree of freedom for encoding infor- knowledges funding from the European Unions Hori- mation, i.e., a qudit. The success of this encoding zon 2020 research and innovation programme under the has been already demonstrated in Ref. [125] and Marie Skodowska-Curie grant agreement No. 847517. its scope is analysed further in Ref. [126]. In par- BA and TG acknowledge funding from Fundacio Privada ticular, two-spin entangling operations have been Cellex, Fundacio Mir-Puig, Generalitat de Catalunya realized in such a setting. It is then straightfor- (AGAUR Grant No. 2017 SGR1341, CERCA pro- ward to extend the scheme of this work to three- gram, QuantumCAT U16-011424, co-funded by ERDF spin transitions, where one or more ions exhibit a Operational Program of Catalonia 2014-2020), Agen- higher (quasi)spin. Among the applications of this cia Estatal de Investigacion (Severo Ochoa Center of capability is in approaching the continuum limit of Excellence CEX2019-000910-S, PlanNational FIDEUA quantum link model which recovers the U(1) lattice PID2019-106901GB-I00/10.13039/501100011033, FPI), gauge theory by increasing the spin of the quantum MINECO-EU QUANTERA MAQS (funded by State link [108]. The fidelity of the effective three-spin Research Agency (AEI) PCI2019-111828-2/10.13039/ dynamics with higher-spin encodings will need to 501100011033), EU Horizon 2020FET-OPEN OPTO- be both numerically and experimentally quantified. Logic (Grant No 899794), ERC AdGNOQIA, and the National Science Centre, Poland-Symfonia Grant No. It is interesting to explore the viability of engineer- 2016/20/W/ST4/00314. TG further acknowledges a fel- ing higher-spin effective interactions, following the lowship granted by la Caixa Foundation (ID100010434, strategy of this work. Explicitly, resonant transi- fellowship code LCF/BQ/PI19/11690013). ZD is sup- tions involving multi-spin flips, assisted by a num- ported in part by the U.S. Department of Energy’s Of- ber of virtual phonons, can be induced, but such fice of Science Early Career Award, under award no. processes will be higher-order contributions in the DE-SC0020271, for theoretical developments for map- Lamb-Dicke parameter, and hence exhibit slower ping lattice gauge theories to quantum simulators, and dynamics. Furthermore, for effective four-spin in- by the DOE Office of Science, Office of Advanced Sci- teractions, for example, the lower-order resonant entific Computing Research (ASCR) Quantum Com- transitions will be proportional to both 1/δ and puting Application Teams program, under fieldwork 1/δ2 in the single-mode approximation, and the proposal number ERKJ347, for algorithmic develop- simple two-drive scheme of this work with almost ments for scientific applications of near-term quantum opposite detunings cannot eliminate these lower- hardware. MH and GP acknowledge support by the order spin transitions. Nonetheless, more com- DOE Office of Science, Office of Nuclear Physics, un- plex schemes can be potentially devised. Perhaps der Award no. DE-SC0021143, for designing hardware- more interestingly is the possibility of inducing res- specific simulation protocols for applications in nuclear onant spin-spin-phonon transitions by borrowing physics. GP is further supported by the Army Re- ideas from the extended MS scheme of this work, search Office (W911NF21P0003), Army Research Lab hence opening up the possibility of analog quan- (W911QX20P0063), and the Office of Naval Research tum simulation of coupled fermion-boson models, (N00014-20-1-2695). AS is supported by a Chicago Prize including gauge-fermion couplings in lattice gauge Postdoctoral Fellowship in Theoretical Quantum Science. 20

[1] Anton Mazurenko, Christie S. Chiu, Geoffrey Ji, systems with long-range interactions,” Nature 511, 198 Maxwell F. Parsons, Márton Kanász-Nagy, Richard (2014). Schmidt, Fabian Grusdt, Eugene Demler, Daniel Greif, [16] J Zhang, PW Hess, A Kyprianidis, P Becker, A Lee, and Markus Greiner, “A cold-atom fermi–hubbard an- J Smith, G Pagano, I-D Potirniche, Andrew C Potter, tiferromagnet,” Nature 545, 462 EP – (2017). A Vishwanath, et al., “Observation of a discrete time [2] D. Beck, M. J. Savage, et al., “Nuclear Physics and crystal,” Nature 543, 217 (2017). Quantum Information Science: A Report by the NSAC [17] PW Hess, P Becker, HB Kaplan, A Kyprianidis, Quantum Information Science Subcommittee,” (2019). AC Lee, B Neyenhuis, G Pagano, P Richerme, C Senko, [3] Ian C Cloët, Matthew R Dietrich, John Arring- J Smith, et al., “Non-thermalization in trapped atomic ton, Alexei Bazavov, Michael Bishof, Adam Freese, ion spin chains,” Philosophical Transactions of the Alexey V Gorshkov, Anna Grassellino, Kawtar Hafidi, Royal Society A: Mathematical, Physical and Engineer- Zubin Jacob, et al., “Opportunities for nuclear physics ing Sciences 375, 20170107 (2017). & quantum information science,” arXiv preprint [18] Harvey B. Kaplan, Lingzhen Guo, Wen Lin Tan, Ar- arXiv:1903.05453 (2019). injoy De, Florian Marquardt, Guido Pagano, and [4] Michael J. Cervia, Amol V. Patwardhan, A. B. Bal- Christopher Monroe, “Many-body dephasing in a antekin, ‡ S. N. Coppersmith, and Calvin W. John- trapped-ion quantum simulator,” Phys. Rev. Lett. 125, son, “Entanglement and collective flavor oscillations in 120605 (2020). a dense neutrino gas,” Phys. Rev. D 100, 083001 (2019), [19] Guido Pagano, Aniruddha Bapat, Patrick Becker, arXiv:1908.03511 [hep-ph]. Katherine S. Collins, Arinjoy De, Paul W. Hess, Har- [5] Benjamin Hall, Alessandro Roggero, Alessandro Baroni, vey B. Kaplan, Antonis Kyprianidis, Wen Lin Tan, and Joseph Carlson, “Simulation of Collective Neu- Christopher Baldwin, Lucas T. Brady, Abhinav Desh- trino Oscillations on a Quantum Computer,” (2021), pande, Fangli Liu, Stephen Jordan, Alexey V. Gor- arXiv:2102.12556 [quant-ph]. shkov, and Christopher Monroe, “Quantum approxi- [6] John Preskill, “Simulating quantum field theory with a mate optimization of the long-range ising model with quantum computer,” PoS LATTICE2018, 024 (2018), a trapped-ion quantum simulator,” Proceedings of the arXiv:1811.10085 [hep-lat]. National Academy of Sciences 117, 25396–25401 (2020). [7] M.C. Banuls et al., “Simulating Lattice Gauge Theories [20] W. L. Tan, P. Becker, F. Liu, G. Pagano, K. S. Collins, within Quantum Technologies,” Eur. Phys. J. D 74, 165 A. De, L. Feng, H. B. Kaplan, A. Kyprianidis, R. Lund- (2020), arXiv:1911.00003 [quant-ph]. gren, W. Morong, S. Whitsitt, A. V. Gorshkov, and [8] Monika Aidelsburger et al., “Cold atoms meet lat- C. Monroe, “Domain-wall confinement and dynamics tice gauge theory,” (2021), arXiv:2106.03063 [cond- in a quantum simulator,” Nature Physics 17, 742–747 mat.quant-gas]. (2021). [9] Natalie Klco, Alessandro Roggero, and Martin J. Sav- [21] Antonis Kyprianidis, Francisco Machado, William Mo- age, “Standard Model Physics and the Digital Quantum rong, Patrick Becker, Kate S Collins, Dominic V Else, Revolution: Thoughts about the Interface,” (2021), Lei Feng, Paul W Hess, Chetan Nayak, Guido Pagano, arXiv:2107.04769 [quant-ph]. et al., “Observation of a prethermal discrete time crys- [10] Rainer Blatt and Christian F Roos, “Quantum simula- tal,” Science 372, 1192–1196 (2021). tions with trapped ions,” Nature Physics 8, 277 (2012). [22] Ben P Lanyon, Cornelius Hempel, Daniel Nigg, Markus [11] Christopher Monroe, WC Campbell, L-M Duan, Z- Müller,Rene Gerritsma, F Zähringer,Philipp Schindler, X Gong, AV Gorshkov, PW Hess, R Islam, K Kim, Julio T Barreiro, Markus Rambach, Gerhard Kirch- NM Linke, G Pagano, et al., “Programmable quantum mair, et al., “Universal digital quantum simulation with simulations of spin systems with trapped ions,” Reviews trapped ions,” Science 334, 57–61 (2011). of Modern Physics 93, 025001 (2021). [23] E. A. Martinez et al., “Real-time dynamics of lattice [12] Ch Schneider, Diego Porras, and Tobias Schaetz, “Ex- gauge theories with a few-qubit quantum computer,” perimental quantum simulations of many-body physics Nature 534, 516–519 (2016), arXiv:1605.04570 [quant- with trapped ions,” Reports on Progress in Physics 75, ph]. 024401 (2012). [24] Ho-Kim and Pham, “Elementary particles and their in- [13] R Islam, EE Edwards, K Kim, S Korenblit, C Noh, teractions,” (1998), 10.1007/978-3-662-03712-6. H Carmichael, G-D Lin, L-M Duan, C-C Joseph Wang, [25] Chris Quigg, “Gauge theories of the strong, JK Freericks, et al., “Onset of a quantum phase tran- weak, and electromagnetic interactions,” (2013), sition with a trapped ion quantum simulator,” Nature doi:10.1515/9781400848225. communications 2, 377 (2011). [26] Matthew D. Schwartz, “Quantum field theory and the [14] Petar Jurcevic, Ben P Lanyon, Philipp Hauke, Cor- standard model,” (Cambridge University Press, 2014). nelius Hempel, Peter Zoller, Rainer Blatt, and Chris- [27] P.A. Zyla et al. (Particle Data Group), “Review of Par- tian F Roos, “Quasiparticle engineering and entangle- ticle Physics,” PTEP 2020, 083C01 (2020). ment propagation in a quantum many-body system,” [28] David B. Kaplan, “Five lectures on effective field the- Nature 511, 202 (2014). ory,” (2005) arXiv:nucl-th/0510023. [15] Philip Richerme, Zhe-Xuan Gong, Aaron Lee, Crys- [29] R. Machleidt and D. R. Entem, “Chiral effective field tal Senko, Jacob Smith, Michael Foss-Feig, Spyridon theory and nuclear forces,” Phys. Rept. 503, 1–75 Michalakis, Alexey V Gorshkov, and Christopher Mon- (2011), arXiv:1105.2919 [nucl-th]. roe, “Non-local propagation of correlations in quantum 21

[30] H. W. Hammer, S. König, and U. van Kolck, “Nu- [45] Alexei Bazavov, Yannick Meurice, S-W Tsai, Judah clear effective field theory: status and perspectives,” Unmuth-Yockey, and Jin Zhang, “Gauge-invariant im- Rev. Mod. Phys. 92, 025004 (2020), arXiv:1906.12122 plementation of the abelian-higgs model on optical lat- [nucl-th]. tices,” Physical Review D 92, 076003 (2015). [31] J. Carlson, S. Gandolfi, F. Pederiva, Steven C. Pieper, [46] T.V. Zache, F. Hebenstreit, F. Jendrzejewski, M.K. R. Schiavilla, K. E. Schmidt, and R. B. Wiringa, Oberthaler, J. Berges, and P. Hauke, “Quantum simu- “Quantum Monte Carlo methods for nuclear physics,” lation of lattice gauge theories using Wilson fermions,” Rev. Mod. Phys. 87, 1067 (2015), arXiv:1412.3081 Sci. Technol. 3, 034010 (2018), arXiv:1802.06704 [cond- [nucl-th]. mat.quant-gas]. [32] Ingo Tews, Zohreh Davoudi, Andreas Ekström,Jason D. [47] Jesse R. Stryker, “Oracles for Gauss’s law on digital Holt, and Joel E. Lynn, “New Ideas in Constrain- quantum computers,” Phys. Rev. A 99, 042301 (2019), ing Nuclear Forces,” J. Phys. G 47, 103001 (2020), arXiv:1812.01617 [quant-ph]. arXiv:2001.03334 [nucl-th]. [48] Indrakshi Raychowdhury and Jesse R. Stryker, “Solv- [33] Tim Byrnes and Yoshihisa Yamamoto, “Simulating lating Gauss’s Law on Digital Quantum Computers with tice gauge theories on a quantum computer,” Phys. Rev. Loop-String-Hadron Digitization,” Phys. Rev. Res. 2, A 73, 022328 (2006), arXiv:quant-ph/0510027. 033039 (2020), arXiv:1812.07554 [hep-lat]. [34] J. Ignacio Cirac, Paolo Maraner, and Jiannis K. Pa- [49] Natalie Klco and Martin J. Savage, “Digitization of chos, “Cold atom simulation of interacting relativistic scalar fields for quantum computing,” Phys. Rev. A 99, quantum field theories,” Phys. Rev. Lett. 105, 190403 052335 (2019), arXiv:1808.10378 [quant-ph]. (2010), arXiv:1006.2975 [cond-mat.str-el]. [50] Julian Bender, Erez Zohar, Alessandro Farace, and [35] L. Lamata, J. Casanova, R. Gerritsma, C. F. Roos, J. J. J. Ignacio Cirac, “Digital quantum simulation of lat- Garcia-Ripoll, and E. Solano, “Relativistic quantum tice gauge theories in three spatial dimensions,” New J. mechanics with trapped ions,” New J. Phys. 13, 095003 Phys. 20, 093001 (2018), arXiv:1804.02082 [quant-ph]. (2011), arXiv:1106.6222 [quant-ph]. [51] Zohreh Davoudi, Mohammad Hafezi, Christopher Mon- [36] Stephen P. Jordan, Keith S. M. Lee, and John Preskill, roe, Guido Pagano, Alireza Seif, and Andrew Shaw, “Quantum Algorithms for Quantum Field Theories,” “Towards analog quantum simulations of lattice gauge Science 336, 1130–1133 (2012), arXiv:1111.3633 [quant- theories with trapped ions,” Phys. Rev. Res. 2, 023015 ph]. (2020), arXiv:1908.03210 [quant-ph]. [37] Stephen P. Jordan, Keith S. M. Lee, and John Preskill, [52] Henry Lamm, Scott Lawrence, and Yukari Yamauchi “Quantum Computation of Scattering in Scalar Quan- (NuQS), “General Methods for Digital Quantum Sim- tum Field Theories,” Quant. Inf. Comput. 14, 1014– ulation of Gauge Theories,” Phys. Rev. D 100, 034518 1080 (2014), arXiv:1112.4833 [hep-th]. (2019), arXiv:1903.08807 [hep-lat]. [38] J. Casanova, L. Lamata, I. L. Egusquiza, R. Ger- [53] Siddhartha Harmalkar, Henry Lamm, and Scott ritsma, C. F. Roos, J. J. Garcia-Ripoll, and E. Solano, Lawrence (NuQS), “Quantum Simulation of Field “Quantum Simulation of Quantum Field Theories in Theories Without State Preparation,” (2020), Trapped Ions,” Phys. Rev. Lett. 107, 260501 (2011), arXiv:2001.11490 [hep-lat]. arXiv:1107.5233 [quant-ph]. [54] Alexander F. Shaw, Pavel Lougovski, Jesse R. Stryker, [39] L. Tagliacozzo, A. Celi, A. Zamora, and M. Lewen- and Nathan Wiebe, “Quantum Algorithms for Simu- stein, “Optical Abelian Lattice Gauge Theories,” An- lating the Lattice Schwinger Model,” Quantum 4, 306 nals Phys. 330, 160–191 (2013), arXiv:1205.0496 [cond- (2020), arXiv:2002.11146 [quant-ph]. mat.quant-gas]. [55] Dmitri E. Kharzeev and Yuta Kikuchi, “Real-time chi- [40] D. Banerjee, M. Dalmonte, M. Muller, E. Rico, P. Ste- ral dynamics from a digital quantum simulation,” Phys. bler, U. J. Wiese, and P. Zoller, “Atomic Quan- Rev. Res. 2, 023342 (2020), arXiv:2001.00698 [hep-ph]. tum Simulation of Dynamical Gauge Fields coupled to [56] Bipasha Chakraborty, Masazumi Honda, Taku Fermionic Matter: From String Breaking to Evolution Izubuchi, Yuta Kikuchi, and Akio Tomiya, “Digital after a Quench,” Phys. Rev. Lett. 109, 175302 (2012), Quantum Simulation of the Schwinger Model with arXiv:1205.6366 [cond-mat.quant-gas]. Topological Term via Adiabatic State Preparation,” [41] Erez Zohar, J. Ignacio Cirac, and Benni Reznik, “Cold- (2020), arXiv:2001.00485 [hep-lat]. Atom Quantum Simulator for SU(2) Yang-Mills Lattice [57] Junyu Liu and Yuan Xin, “Quantum simulation of Gauge Theory,” Phys. Rev. Lett. 110, 125304 (2013), quantum field theories as quantum chemistry,” JHEP arXiv:1211.2241 [quant-ph]. 12, 011 (2020), arXiv:2004.13234 [hep-th]. [42] Erez Zohar, J. Ignacio Cirac, and Benni Reznik, [58] JoãoBarata, Niklas Mueller, Andrey Tarasov, and “Quantum simulations of gauge theories with ultra- Raju Venugopalan, “Single-particle digitization strat- cold atoms: local gauge invariance from angular mo- egy for quantum computation of a φ4 scalar field the- mentum conservation,” Phys. Rev. A88, 023617 (2013), ory,” (2020), arXiv:2012.00020 [hep-th]. arXiv:1303.5040 [quant-ph]. [59] Danny Paulson et al., “Towards simulating 2D effects in [43] P. Hauke, D. Marcos, M. Dalmonte, and P. Zoller, lattice gauge theories on a quantum computer,” (2020), “Quantum simulation of a lattice schwinger model in a arXiv:2008.09252 [quant-ph]. chain of trapped ions,” Phys. Rev. X 3, 041018 (2013). [60] Anthony Ciavarella, Natalie Klco, and Martin J. Sav- [44] A. Mezzacapo, E. Rico, C. Saban, I.L. Egusquiza, age, “A Trailhead for Quantum Simulation of SU(3) L. Lamata, and E. Solano, “Non-Abelian SU(2) Yang-Mills Lattice Gauge Theory in the Local Multi- Lattice Gauge Theories in Superconducting Circuits,” plet Basis,” (2021), arXiv:2101.10227 [quant-ph]. Phys. Rev. Lett. 115, 240502 (2015), arXiv:1505.04720 [61] Raka Dasgupta and Indrakshi Raychowdhury, “Cold [quant-ph]. Atom Quantum Simulator for String and Hadron Dy- 22

namics in Non-Abelian Lattice Gauge Theory,” (2020), [75] Zohreh Davoudi, Norbert M. Linke, and Guido Pagano, arXiv:2009.13969 [hep-lat]. “Toward simulating quantum field theories with con- [62] Rene Gerritsma, Gerhard Kirchmair, Florian Za- trolled phonon-ion dynamics: A hybrid analog-digital ehringer, E Solano, R Blatt, and CF Roos, “Quan- approach,” (2021), arXiv:2104.09346 [quant-ph]. tum simulation of the dirac equation,” Nature 463, 68 [76] Jesse R. Stryker, “Shearing approach to gauge invariant (2010). Trotterization,” (2021), arXiv:2105.11548 [hep-lat]. [63] N. Klco, E. F. Dumitrescu, T. D. McCaskey, A. J. and- [77] Frank Pollmann, Ari M Turner, Erez Berg, and Masaki mMorris, R. C. Pooser, M. Sanz, E. Solano, P. Lougov- Oshikawa, “Entanglement spectrum of a topological ski, and M. J. Savage, “Quantum-classical computation phase in one dimension,” Physical review b 81, 064439 of Schwinger model dynamics using quantum comput- (2010). ers,” Phys. Rev. A98, 032331 (2018), arXiv:1803.03326 [78] Ruben Verresen, Roderich Moessner, and Frank Poll- [quant-ph]. mann, “One-dimensional symmetry protected topologi- [64] Hsuan-Hao Lu et al., “Simulations of Subatomic Many- cal phases and their transitions,” Physical Review B 96, Body Physics on a Quantum Frequency Processor,” 165124 (2017). Phys. Rev. A 100, 012320 (2019), arXiv:1810.03959 [79] Ruben Verresen, Nick G Jones, and Frank Pollmann, [quant-ph]. “Topology and edge modes in quantum critical chains,” [65] Natalie Klco, Jesse R. Stryker, and Martin J. Savage, Physical review letters 120, 057001 (2018). “SU(2) non-Abelian gauge field theory in one dimen- [80] Adam Smith, Bernhard Jobst, Andrew G Green, and sion on digital quantum computers,” Phys. Rev. D 101, Frank Pollmann, “Crossing a topological phase tran- 074512 (2020), arXiv:1908.06935 [quant-ph]. sition with a quantum computer,” arXiv preprint [66] Yasar Atas, Jinglei Zhang, Randy Lewis, Amin Ja- arXiv:1910.05351 (2019). hanpour, Jan F. Haase, and Christine A. Muschik, [81] CH Tseng, S Somaroo, Y Sharf, Emanuel Knill, Ray- “SU(2) hadrons on a quantum computer,” (2021), mond Laflamme, Timothy F Havel, and David G arXiv:2102.08920 [quant-ph]. Cory, “Quantum simulation of a three-body-interaction [67] Michael Kreshchuk, Shaoyang Jia, William M. Kirby, hamiltonian on an nmr quantum computer,” Physical Gary Goldstein, James P. Vary, and Peter J. Love, Review A 61, 012302 (1999). “Light-Front Field Theory on Current Quantum Com- [82] GH Wannier, “Antiferromagnetism. the triangular ising puters,” (2020), arXiv:2009.07885 [quant-ph]. net,” Physical Review 79, 357 (1950). [68] Christian W. Bauer, Marat Freytsis, and Benjamin [83] Xinhua Peng, Jingfu Zhang, Jiangfeng Du, and Dieter Nachman, “Simulating collider physics on quantum Suter, “Quantum simulation of a system with compet- computers using effective field theories,” (2021), ing two-and three-body interactions,” Physical review arXiv:2102.05044 [hep-ph]. letters 103, 140501 (2009). [69] Frederik Görg,Kilian Sandholzer, Joaqu´ın Minguzzi, [84] Jean Christian AnglèsdAuriac and Ferenc Igloi, “Level Rémi Desbuquois, Michael Messer, and Tilman statistics of multispin-coupling models with first-and Esslinger, “Realization of density-dependent Peierls second-order phase transitions,” Physical Review E 58, phases to engineer quantized gauge fields coupled to ul- 241 (1998). tracold matter,” Nature Phys. 15, 1161–1167 (2019), [85] KA Penson, JM Debierre, and L Turban, “Conformal arXiv:1812.05895 [cond-mat.quant-gas]. invariance and critical behavior of a quantum hamilto- [70] Christian Schweizer, Fabian Grusdt, Moritz Berngru- nian with three-spin coupling in a longitudinal field,” ber, Luca Barbiero, Eugene Demler, Nathan Gold- Physical Review B 37, 7884 (1988). man, Immanuel Bloch, and Monika Aidelsburger, [86] David J Gross and Marc Mézard, “The simplest spin “Floquet approach to Z2 lattice gauge theories with glass,” Nuclear Physics B 240, 431–452 (1984). ultracold atoms in optical lattices,” arXiv preprint [87] E. Gardner, “Spin glasses with p-spin interactions,” Nu- arXiv:1901.07103 (2019). clear Physics B 257, 747–765 (1985). [71] Alexander Mil, Torsten V. Zache, Apoorva Hegde, Andy [88] Theo M Nieuwenhuizen and Felix Ritort, “Quantum Xia, Rohit P. Bhatt, Markus K. Oberthaler, Philipp phase transition in spin glasses with multi-spin inter- Hauke, JürgenBerges, and Fred Jendrzejewski, “A actions,” Physica A: Statistical Mechanics and its Ap- scalable realization of local U(1) gauge invariance in plications 250, 8–45 (1998). cold atomic mixtures,” Science 367, 1128–1130 (2020), [89] Dimitris I Tsomokos, Juan Jose Garcia-Ripoll, Nigel R arXiv:1909.07641 [cond-mat.quant-gas]. Cooper, and Jiannis K Pachos, “Chiral entanglement [72] Bing Yang, Hui Sun, Robert Ott, Han-Yi Wang, in triangular lattice models,” Physical Review A 77, Torsten V. Zache, Jad C. Halimeh, Zhen-Sheng 012106 (2008). Yuan, Philipp Hauke, and Jian-Wei Pan, “Observa- [90] Wen X., Quantum Field Theory of Many-body Systems tion of gauge invariance in a 71-site Bose–Hubbard (Oxford Univ. Press, New York, 2004). quantum simulator,” Nature 587, 392–396 (2020), [91] Francesco, Mathieu, and Senechal, arXiv:2003.08945 [cond-mat.quant-gas]. Conformal Field Theory (Springer-Verlag New York, [73] Zohreh Davoudi, Indrakshi Raychowdhury, and An- 1997). drew Shaw, “Search for Efficient Formulations for [92] Shailesh Chandrasekharan and U-J Wiese, “Quantum Hamiltonian Simulation of non-Abelian Lattice Gauge link models: A discrete approach to gauge theories,” Theories,” (2020), arXiv:2009.11802 [hep-lat]. Nuclear Physics B 492, 455–471 (1997). [74] Sarmed A. Rahman, Randy Lewis, Emanuele Mendi- [93] R. Brower, S. Chandrasekharan, and U.J. Wiese, “QCD celli, and Sarah Powell, “SU(2) lattice gauge theory on as a quantum link model,” Phys. Rev. D 60, 094502 a quantum annealer,” (2021), arXiv:2103.08661 [hep- (1999), arXiv:hep-th/9704106. lat]. 23

[94] Wuxin Liu, Wei Feng, Wenhui Ren, Da-Wei Wang, and [111] Laird Egan, Dripto M Debroy, Crystal Noel, An- Haohua Wang, “Synthesizing three-body interaction of drew Risinger, Daiwei Zhu, Debopriyo Biswas, Michael spin chirality with superconducting qubits,” Applied Newman, Muyuan Li, Kenneth R Brown, Marko Physics Letters 116, 114001 (2020). Cetina, et al., “Fault-tolerant operation of a quantum [95] Francesco Petiziol, Mahdi Sameti, Stefano Carretta, error-correction code,” arXiv preprint arXiv:2009.11482 Sandro Wimberger, and Florian Mintert, “Quan- (2020). tum simulation of three-body interactions in weakly [112] Colin D Bruzewicz, John Chiaverini, Robert McConnell, driven quantum systems,” Physical Review Letters 126, and Jeremy M Sage, “Trapped-ion quantum computing: 250504 (2021). Progress and challenges,” Applied Physics Reviews 6, [96] Jiannis K Pachos and Martin B Plenio, “Three-spin 021314 (2019). interactions in optical lattices and criticality in clus- [113] Jad C. Halimeh, Valentin Kasper, and Philipp Hauke, ter hamiltonians,” Physical review letters 93, 056402 “Fate of Lattice Gauge Theories Under Decoherence,” (2004). (2020), arXiv:2009.07848 [cond-mat.quant-gas]. [97] HP Büchler, A Micheli, and P Zoller, “Three-body in- [114] Maarten Van Damme, Haifeng Lang, Philipp Hauke, teractions with cold polar molecules,” Nature Physics and Jad C. Halimeh, “Reliability of lattice gauge 3, 726–731 (2007). theories in the thermodynamic limit,” (2021), [98] Jiannis K Pachos and Peter L Knight, “Quantum com- arXiv:2104.07040 [cond-mat.quant-gas]. putation with a one-dimensional optical lattice,” Phys- [115] S.-L. Zhu, C. Monroe, and L.-M. Duan, “Trapped ion ical review letters 91, 107902 (2003). quantum computation with transverse phonon modes,” [99] Filippo M Gambetta, Chi Zhang, Markus Hennrich, Phys. Rev. Lett. 97, 050505 (2006). Igor Lesanovsky, and Weibin Li, “Long-range multi- [116] Christian F Roos, “Ion trap quantum gates with body interactions and three-body antiblockade in a amplitude-modulated laser beams,” New Journal of trapped rydberg ion chain,” Physical Review Letters Physics 10, 013002 (2008). 125, 133602 (2020). [117] Todd J. Green and Michael J. Biercuk, “Phase- [100] A Bermudez, Diego Porras, and MA Martin-Delgado, modulated decoupling and error suppression in qubit- “Competing many-body interactions in systems of oscillator systems,” Phys. Rev. Lett. 114, 120502 trapped ions,” Physical review A 79, 060303 (2009). (2015). [101] D. Leibfried, B. DeMarco, V. Meyer, D. Lucas, M. Bar- [118] Pak Hong Leung, Kevin A. Landsman, Caroline Fig- rett, J. Britton, W. M. Itano, B. Jelenković,C. Langer, gatt, Norbert M. Linke, Christopher Monroe, and T. Rosenband, and et al., “Experimental demonstra- Kenneth R. Brown, “Robust 2-qubit gates in a linear tion of a robust, high-fidelity geometric two ion-qubit ion crystal using a frequency-modulated driving force,” phase gate,” Nature 422, 412–415 (2003). Phys. Rev. Lett. 120, 020501 (2018). [102] Klaus Mølmer and Anders Sørensen, “Multiparticle en- [119] Reinhold Blümel,Nikodem Grzesiak, Nhung H. Nguyen, tanglement of hot trapped ions,” Physical Review Let- Alaina M. Green, Ming Li, Andrii Maksymov, Nor- ters 82, 1835 (1999). bert M. Linke, and Yunseong Nam, “Efficient, stabi- [103] Wolfgang Paul, “Electromagnetic traps for charged and lized two-qubit gates on a trapped-ion quantum com- neutral particles,” Reviews of modern physics 62, 531 puter,” (2021), arXiv:2101.07887. (1990). [120] Juan I Cirac and Peter Zoller, “Quantum computations [104] Wilhelm Magnus, “On the exponential solution of differ- with cold trapped ions,” Physical review letters 74, 4091 ential equations for a linear operator,” Commun. Pure (1995). Appl. Math. 7, 649–673 (1954). [121] Diego Porras and J Ignacio Cirac, “Effective quantum [105] A. Steane, “The ion trap quantum information proces- spin systems with trapped ions,” Physical review letters sor,” Applied Physics B 64, 623–643 (1997). 92, 207901 (2004). [106] Pablo Sala, Tao Shi, Stefan Kühn,Mari Carmen Banuls, [122] Joseph W Britton, Brian C Sawyer, Adam C Keith, Eugene Demler, and Juan Ignacio Cirac, “Variational C-C Joseph Wang, James K Freericks, Hermann Uys, study of u (1) and su (2) lattice gauge theories with Michael J Biercuk, and John J Bollinger, “Engi- gaussian states in 1+ 1 dimensions,” Physical Review neered two-dimensional ising interactions in a trapped- D 98, 034505 (2018). ion quantum simulator with hundreds of spins,” Nature [107] John Kogut and Leonard Susskind, “Hamiltonian for- 484, 489 (2012). mulation of wilson’s lattice gauge theories,” Physical [123] Simcha Korenblit, Dvir Kafri, Wess C Campbell, Ra- Review D 11, 395 (1975). jibul Islam, Emily E Edwards, Zhe-Xuan Gong, Guin- [108] Torsten V Zache, Maarten Van Damme, Jad C Halimeh, Dar Lin, Lu-Ming Duan, Jungsang Kim, Kihwan Kim, Philipp Hauke, and Debasish Banerjee, “Achieving the et al., “Quantum simulation of spin models on an arbi- continuum limit of quantum link lattice gauge theories trary lattice with trapped ions,” New Journal of Physics on quantum devices,” arXiv preprint arXiv:2104.00025 14, 095024 (2012). (2021). [124] Yi Hong Teoh, Marina Drygala, Roger G Melko, and [109] Michael L Wall, Arghavan Safavi-Naini, and Ana Maria Rajibul Islam, “Machine learning design of a trapped- Rey, “Boson-mediated quantum spin simulators in ion quantum spin simulator,” Quantum Science and transverse fields: X y model and spin-boson entangle- Technology 5, 024001 (2020). ment,” Physical Review A 95, 013602 (2017). [125] C Senko, P Richerme, J Smith, A Lee, I Cohen, A Ret- [110] K Wright, KM Beck, S Debnath, JM Amini, Y Nam, zker, and C Monroe, “Realization of a quantum integer- N Grzesiak, J-S Chen, NC Pisenti, M Chmielewski, spin chain with controllable interactions,” Physical Re- C Collins, et al., “Benchmarking an 11-qubit quantum view X 5, 021026 (2015). computer,” arXiv preprint arXiv:1903.08181 (2019). 24

[126] Pei Jiang Low, Brendan M White, Andrew A Cox, models in trapped ions,” EPJ Quant. Technol. 1, 9 Matthew L Day, and Crystal Senko, “Practical trapped- (2014), arXiv:1312.2849 [quant-ph]. ion protocols for universal qudit-based quantum com- [128] A Mezzacapo, J Casanova, L Lamata, and E Solano, puting,” Physical Review Research 2, 033128 (2020). “Digital quantum simulation of the holstein model [127] L. Lamata, A. Mezzacapo, J. Casanova, and E. Solano, in trapped ions,” Physical review letters 109, 200501 “Eﬃcient quantum simulation of fermionic and bosonic (2012).