A Wavelength-Insensitive, Multispecies Entangling Gate for Group-2 Atomic Ions

Brian C. Sawyer∗ and Kenton R. Brown† Georgia Tech Research Institute, Atlanta, Georgia 30332, USA

We propose an optical scheme for generating entanglement between co-trapped identical or dis- similar alkaline earth atomic ions (40Ca+, 88Sr+, 138Ba+, 226Ra+) which exhibits fundamental error rates below 10−4 and can be implemented with a broad range of laser wavelengths spanning from ultraviolet to infrared. We also discuss straightforward extensions of this technique to include the two lightest Group-2 ions (Be+, Mg+) for multispecies entanglement. The key elements of this wavelength-insensitive geometric phase gate are the use of a ground (S1/2) and a metastable (D5/2) electronic state as the qubit levels within a σzσz light-shift entangling gate. We present a detailed analysis of the principles and fundamental error sources for this gate scheme which includes photon scattering and spontaneous emission decoherence, calculating two-qubit-gate error rates and dura- tions at fixed laser beam intensity over a large portion of the optical spectrum (300 nm to 2 µm) for an assortment of ion pairs. We contrast the advantages and disadvantages of this technique against previous trapped-ion entangling gates and discuss its applications to quantum information processing and simulation with like and multispecies ion crystals.

I. INTRODUCTION the effect of terms outside the Lamb-Dicke approxima- tion when trap confinement is weak or motional cooling A variety of entangling gate schemes has been demon- is imperfect, or more extreme (i.e. counter-propagating) strated in trapped-ion systems. To date, two-qubit gates laser beam geometries can be used in ion traps with con- have been implemented most commonly with two dis- straints on optical access. tinct techniques, the light-shift (LS) gate [1–4] and the Two-qubit gates between trapped-ion qubits take ad- “Mølmer-Sørensen” (MS) gate [5, 6]. While MS gates vantage of the ions’ motional degrees of freedom to have been used with optical, hyperfine, and Zeeman imprint quantum phases onto the system in a spin- transition qubits, LS gates have been implemented only dependent way. In the MS gate [5], a laser beam with on the latter two. Alternatively, it should be possi- two tones is tuned near the upper and lower motional ble perform a light-shift gate with a dipole force on sidebands of the ions’ qubit transition frequency. This an optical-transition qubit (an “optical-transition dipole- induces a force onto the ions whose strength depends on force” (OTDF) gate), which is less sensitive to several their states in the σxσx or σyσy basis. With appropri- important sources of error encountered in the more com- ate detunings from the motional sidebands the system mon LS or MS gates. A related idea has been explored acquires the desired quantum geometric phase. The MS in Refs. [7, 8] which present a σzσz interaction induced gate is often convenient, because it can be implemented via beams at a specific, small detuning near the electric- with the same set of laser beams used to achieve one- dipole-forbidden S − D optical transition. qubit interactions (often at visible or IR wavelengths). In contrast, the OTDF gate presented in this Arti- Furthermore, it is compatible with qubits based on clock cle relies on a state-dependent optical-dipole force and transitions with long coherence times [12]. Unfortu- functions in the regime of laser wavelengths detuned far nately, especially for optical-transition qubits, the MS from the S − D transition. Our OTDF gate permits an scheme suffers from the presence of unwanted AC Stark unprecedented range of feasible laser wavelengths, such shifts during the laser interaction which can fluctuate that reliable commercial laser systems, including at the with small changes in laser intensity and result in σz telecommunications-relevant 1550 nm wavelength, could errors that do not commute with the two-qubit gate op- be employed. This broad range of wavelengths also makes eration [13]. Moreover, the speed of the MS gate here arXiv:2010.04526v1 [quant-ph] 9 Oct 2020 the OTDF gate straightforward to implement in systems varies only as the square root of laser intensity, making of disparate-species ions [9–11]. In particular, a second faster gates impractical and increasingly prone to errors species can be used for sympathetic cooling, for measure- from AC Stark shifts. Drifts in magnetic field or in qubit ment of a nearby ion’s state without photon-scattering laser frequency also lead to (non-commuting) σz errors decoherence, or for photonic-qubit wavelength conver- in these systems. Calibrations for these drifts are time- sion. Another powerful consequence of broad wavelength intensive and may be difficult to perform quickly enough tuneability is the possibility to tailor the OTDF Lamb- for accurate compensation of their effects. Dicke parameter for a given ion species and trap con- The LS gate, thus far only experimentally demon- figuration. As an example, one can choose to minimize strated on hyperfine or Zeeman qubits, operates in a re- lated manner but in a different basis [1]. Here two laser beams, each far detuned from any transition, intersect to form an optical lattice at the ions’ positions. The beams ∗ [email protected] do not directly induce qubit transitions, but their pres- † [email protected] ence shifts the internal energy of the ions via the AC 2

transitions to suppress photon scattering errors without greatly suppressing the gate interaction strength [3]. Fur- thermore, for these types of qubits the desired gate laser wavelengths typically lie within the UV range, a serious disadvantage from the perspectives of optical power gen- eration, compatibility with optical fibers [14] and waveg- uides [15], and the accumulation of stray charges on trap surfaces [16, 17]. Before 2015, the highest fidelity two-qubit gate had been performed with an MS interaction on an “opti- cal” qubit in 40Ca+, with qubit levels separated by an S − D transition wavelength [13]. In comparison with hyperfine and Zeeman qubits, an optical qubit benefits from higher detection fidelities [18]. Ions with low-lying metastable levels also have the advantage of several aux- iliary states which can be used to “hide” one ion from decoherence incurred while detecting the state of another nearby ion [19]. The wavelengths for manipulation of an optical qubit typically lie in the visible domain making the light easier to generate, easier to transport in fibers and waveguides, and less prone to creating stray charge in ion traps than UV light. Unfortunately, an optical qubit generally suffers more from decoherence caused by magnetic-field variation and beam intensity fluctuations, especially during MS two-qubit gate operations. Mølmer- Sørensen gates on optical qubits also impose stringent requirements on gate laser phase and frequency stabil- ity, where stabilization techniques are usually adapted Figure 1. (Color online) (a) Schematic level structure for directly from state-of-the-art optical clock experiments the iso-electronic species 40Ca+, 88Sr+, 138Ba+, and 226Ra+ [20]. showing low-lying electronic states and the specific mJ qubit To date, a LS gate of the type described above has not levels (red) assumed for this work. (b) Example illustration of z z been implemented in optical-transition qubits such as the a σ σ entangling gate experiment showing entangling laser 40 + 88 + ◦ long-lived S1/2 − D5/2 superpositions in Ca , Sr , beams intersecting at 90 (red) with linear polarizations out 138 + 226 + of the page, trap electrodes (gray), and difference-wavevector Ba , and Ra . Here we propose such an OTDF k∆ along the trap axis. Here we assume that the two-qubit gate acting on the S1/2−D5/2 transitions in these species. gate is performed using the axial modes of a pair of identical or Our restriction to nuclear-spin-free isotopes throughout disparate co-trapped ions (blue circles). More extreme laser this Article is for the sake of simplicity; extension to ions beam geometries (counter-propagating) are feasible at long with hyperfine structure — even including optical-clock- (i.e. infrared) wavelengths. transition qubits [8] — is straightforward. Expanding on recent dual-species entanglement work [11, 21], the OTDF gate naturally operates without modification on + Stark effect. Because the size of this shift depends on pairs of the heterogeneous alkaline-earth ion species Ca , + + + the internal spin state, and because the ions reside in Sr , Ba , and Ra in superpositions of their respective an optical lattice (with its accompanying optical inten- S and D levels. Furthermore, due to the broad wave- sity or polarization gradient), there is a force induced length tunability of the OTDF gate, the lighter species + + whose strength depends on the ions’ states in the σzσz Be and Mg may each be paired with a heavier al- basis. With appropriate detunings from the motional kaline earth for dual-species entangling gates combining + + sidebands the system acquires the desired quantum ge- traditional hyperfine or Zeeman qubits (Be , Mg ) with + + + + ometric phase. Because the LS gate uses a σzσz inter- optical qubits (Ca , Sr , Ba , Ra ). action, it may be paired with straightforward dynamical The OTDF gate can leverage large laser detunings (be- decoupling pulses [1, 3] (a Hahn spin-echo pulse, for ex- cause of the > 100 THz qubit splitting) to suppress pho- ample) to eliminate σz errors caused by unwanted AC ton scattering errors and can be implemented over a wide Stark shifts or drifts in qubit frequency. Furthermore, range of laser wavelengths even into the near- and short- the speed of the gate varies linearly with laser inten- wavelength infrared. Beyond their technological advan- sity, so that faster gates can be achieved. Unfortunately, tages, longer laser wavelengths yield smaller Lamb-Dicke because of the relatively small qubit splitting (MHz to factors, thereby relaxing ion temperature and trap con- GHz) in typical Zeeman and hyperfine systems, the laser finement requirements. beams cannot be sufficiently detuned from nearby optical The remainder of this Article is structured as follows. 3

Sec. II provides a general description of the OTDF gate dient [1, 4]. In our proposed scheme, the ODF beams may operations. Sec. III summarizes calculation of the AC be detuned far (> 100 THz) from any S−P or D−P tran- Stark shifts and photon scattering rates for the four sitions, and it is not possible to eliminate the differential species under consideration. Sec. IV presents gate dura- Stark shift by any choice of polarization. Instead, one tions and intrinsic error rates for identical ion pairs based can choose an arbitrary polarization (speed is maximized on the calculations of Sec. III, and Sec. V presents the when the two beam polarizations are matched: each lin- gate durations and error rates for heterogeneous pairs. ear and orthogonal to the laser-beam k-vectors) and rely The full derivations for AC Stark shifts, photon scatter- on an optical intensity gradient rather than on a polar- ing rates, gate durations, and intrinsic error rates are ization gradient to create the necessary ODF. However, presented in the three appendices. the resulting differential Stark shift induces unwanted σz rotations onto the ions beyond the desired phase-gate in- teractions. In principle these could be accounted for with II. GATE OVERVIEW careful control and calibration of laser intensity, but in practice the use of a spin echo ensures that unwanted The basic theory of LS gate operation has been well rotations from the first gate pulse will be canceled by explored in the literature [1, 3, 21]. For completeness, we analogous rotations from the second. The dependence of provide a detailed derivation of OTDF gate operation in these unwanted rotations on the absolute optical phase of a multispecies system in Appendix C. To perform entan- the lattice can be eliminated by turning the laser pulses gling gates with axial modes of motion, we consider two on and off adiabatically [22]. optical-dipole-force (ODF) beams at angular frequencies ω and ω +2ω propagating 90◦ to one another to form L L ∆ III. AC STARK SHIFTS AND SCATTER RATES a traveling optical lattice overlapping a two-ion crystal with a motional mode frequency Ωk (see Fig. 1(b)). At The speed of the OTDF gate outlined above is gov- any instant in time, the beams√ form an optical lattice erned by the magnitude of the AC Stark shift induced with wavelength λeff = 2 2πc/ωL along the ion trap symmetry axis, where c is the speed of light. by the gate beams, which is dominated by D − P and Because the two beams are detuned from each other, S − P electric dipole transitions. The minimum entan- gling gate duration for two phase-space loops (derived in the optical lattice moves through a distance λeff in the Appendix C and reproduced here) is interval 2π/(2ω∆), thereby modulating the optical dipole force that each ion experiences. For a constant detuning s 2 δk from the gate mode 2ω∆ ≡ ΩM + δk, the mode is 2π τg = . first excited and then de-excited and follows a circle in |ηk1ηk2 cos(φ0,2 − φ0,1)∆0,∆(x0,1)∆0,∆(x0,2)| position-momentum (x-p) phase space. After a duration (1) 2π/δk the motion has returned to its initial phase-space In addition to its dependence on the effective Lamb-Dicke location, but the quantum state has acquired an overall parameters ηkj and ion spacing (captured via φ0,j), the geometric phase proportional to the area enclosed by the duration is inversely proportional to the geometric mean path in phase space. of the beams’ differential Stark shifts at the two ion lo- The ODF strength depends on the internal cations, ∆0,∆(x0,j). state (|↑i or |↓i) of each ion, so that the geo- Two unavoidable sources of error are spontaneous pho- metric phase is state-dependent: the four pos- ton scattering and metastable-state decay. A complete sible states {|↑↑i , |↑↓i , |↓↑i , |↓↓i} acquire phases theoretical description of the former, including both elas- {Φ↑↑, Φ↑↓, Φ↓↑, Φ↓↓}. To symmetrize the interaction tic and inelastic terms, is presented in Appendix B. Sum- we imagine performing two such pulses separated by marizing here, the total intrinsic two-ion decoherence a spin echo pulse such that the total acquired phases (2) rate, Γtot, is become {Φ↑↑↓↓, Φ↑↓↓↑, Φ↑↓↓↑, Φ↑↑↓↓}. For an appropriate choice of ODF strength, the ideal two-qubit phase gate 2   π (2) X 1 (Φ↑↓↓↑ = Φ↑↑↓↓ ± ) is realized [1]. This symmetrization Γ = Γ + Γ + A , (2) 2 tot el,j in,j 2 D,j enables the use of ions of disparate species, qubit states j=1 with vastly differing AC Stark shifts, and ions within an ODF intensity gradient (among other benefits; see where Γel,j (Γin,j) is the elastic (inelastic) photon scatter- Appendix C for more details). ing decoherence rate of ion j due to the OTDF gate laser Because previous LS gate implementations used hyper- beams as defined in Appendix B. The final term in Eq. 2 fine or Zeeman qubits and gate laser wavelengths tuned describes spontaneous decay of the metastable superpo- near the S1/2 −P1/2 and S1/2 −P3/2 transitions, these ex- sition of S and D levels, which is one-half of the D-state periments could leverage ODF laser polarizations which decay rate of ion j. Given the finite lifetime of the D lev- partially or completely nulled the differential AC Stark els, it might be favorable to adopt a ‘hybrid’ qubit encod- shift between |↑i and |↓i states while still achieving finite ing whereby population is cycled between Zeeman qubits optical-dipole forces via a non-vanishing polarization gra- for storage and optical qubits for two-qubit interactions 4

2 7 2 7 10 10 10 10 Ca Sr ) ) 1 1 Stark shift magnitude (Hz) Stark shift magnitude (Hz) s s

( 1 6 ( 1 6

e 10 10 e 10 10 t t a a r r

g g n n i i r r e e t t

t 0 5 t 0 5

a 10 10 a 10 10 c c s s

s s u u o o e e n n a a

t 1 4 t 1 4

n 10 10 n 10 10 o o p p S S

2 3 2 3 10 10 10 10 500 1000 1500 2000 500 1000 1500 2000 Wavelength (nm) Wavelength (nm)

2 7 2 7 10 10 10 10 Ba Ra ) ) 1 1 1 Stark shift magnitude (Hz) Stark shift magnitude (Hz)

s 10 s ( 6 ( 1 6

e 10 e 10 10 t t a a r r

g 0 g n n i 10 i r r e e t t

t 5 t 0 5

a 10 a 10 10 c c s s 1 s s

u 10 u o o e e n n a a

t 4 t 1 4

n 2 10 n 10 10 o o

p 10 p S S

3 3 2 3 10 10 10 10 500 1000 1500 2000 500 1000 1500 2000 Wavelength (nm) Wavelength (nm)

Figure 2. (Color online) Spontaneous photon scattering decoherence rate (orange line, left axis) and AC Stark shift magnitude (blue line, right axis) versus laser wavelength for (a) 40Ca+, (b) 88Sr+, (c) 138Ba+, and (d) 226Ra+. All curves shown assume a single ion at the center of a single laser beam with a total power of 100 mW, a Gaussian waist of 10 µm (peak intensity 637 MW/m2), and linear (equal σ+ and σ−) polarization. A gate would generally be operated far enough from the resonant wavelengths to reduce spontaneous scattering to an acceptable level. as recently discussed in Ref. [23]. We exclude Yb+ from quency difference comparable to or larger than the gate our OTDF gate analysis because of the relatively short laser beam detuning from the nearest intermediate state < lifetimes (∼ 60 ms) of its metastable states [24, 25]. will suffer from non-negligible elastic photon scattering decoherence. Reference [26] describes the process by which elastic photon scattering causes decoherence of a superposition Both the AC Stark shift and the scattering deco- state. In summary, if the elastic photon scattering am- herence rate for a single 40Ca+, 88Sr+, 138Ba+, and plitudes are different between the |↑i and |↓i states, then 226Ra+ion are depicted in Fig. 2 over the wavelength the elastic scattering process will cause dephasing of the range 300 − 2000 nm at a single-laser-beam intensity superposition at a rate proportional to the square of the of 637 MW/m2 (e.g. 100 mW in a 10 µm Gaussian difference between the scattering amplitudes from |↑i and beam waist). We assume linear (equal σ+ and σ−) po- |↓i (see Eq. B5). In practice, qubit states with a fre- larization, although polarization effects are negligible for 5 the large detunings considered here. The narrow elec- tion with high bandwidth, experimentally removing op- tric quadrupole resonances are not visible in the curves tical lattice phase fluctuations that can limit precision of Fig. 2. We neglect their contribution to the AC Stark measurement experiments with trapped ions [28]. shifts and scattering rates, assuming that the OTDF gate Despite the larger gate errors for all species in the laser frequencies are not directly resonant with any S−D infrared regime, quantum simulation experiments (e.g. transitions [27] variational quantum eigensolvers [29, 30], quantum spin model simulations [31–33], and quantum approximate op- timization [34]) may benefit from the more mature op- IV. SAME SPECIES ENTANGLEMENT tical technologies at longer wavelengths. Furthermore, the larger Lamb-Dicke parameters at infrared gate laser Given Eqs. 1 and 2, it is straightforward to calculate wavelengths permit weaker trap confinement or imperfect (2) ground-state cooling. Efficient ground state cooling of the gate durations, τg, and intrinsic errors,  = Γtotτg, for various ion pairs. We begin with a comparison of the relevant motional modes of many-ion crystals is par- pairs consisting of the same ion species (40Ca+, 88Sr+, ticularly challenging, a problem which further motivates 138Ba+, and 226Ra+) in Figs. 3 and 4. Here, we assume entangling gate schemes utilizing infrared wavelengths in the pair is trapped in a harmonic electrostatic potential both radiofrequency and Penning ion trap quantum sim- providing a 2 MHz axial frequency to a single 40Ca+ ion ulation experiments. (we neglect the small changes in potential that would be required to achieve ion spacing optimized for gate speed). 2 The chosen qubit states (Fig. 1(a)) are | S1/2(mJ = 1/2)i 2 and | D5/2(mJ = +3/2)i. The ion pair is illuminated by two beams each with a power of 100 mW, a 10 µm V. MULTISPECIES ENTANGLEMENT Gaussian waist (peak intensity 637 MW/m2 per beam), ◦ + an angle of 45 to the crystal axis, and linear (equal σ We performed similar calculations for pairs of het- − and σ ) polarization as in Fig. 1(b). erogeneous ion species and present the results for Note that at a given wavelength, the photon scatter- (40Ca+−88Sr+, 88Sr+−138Ba+, and 138Ba+−226Ra+) in ing decoherence rate (Eq. B2) is proportional to the laser Fig. 5. Other combinations yield similar gate errors and intensity while the gate duration (Eq. 1) is inversely pro- speeds but are not presented here due to the large differ- portional to intensity. For a gate duration with a negligi- ence in Mathieu stability parameter for ions with widely ble contribution of D-state decay error, the fundamental disparate masses [35]. This challenge is not insurmount-  (2)  two-qubit-gate error Γtotτg is independent of laser in- able but reduces the appeal of such mass combinations. tensity. Near the S − P and D − P resonant wavelengths The trapping potential and laser beam characteristics are the gate speeds grow quickly, but the photon scattering the same as for the homogeneous case (Section IV). The error rates increase faster still, so that wavelengths opti- dissimilar-ion behavior is more complex as a function of mized for minimum error occur at relatively large detun- wavelength than the homogeneous case as there are twice ings from these resonances. Similarly, the “magic wave- as many resonant wavelengths that must be avoided. We lengths” where the differential AC Stark shift approaches reiterate that a spin flip between the two ODF pulses is zero must also be avoided. essential to symmetrize the interaction here because of All four ions considered in this Article offer two qubit the mismatched ion Stark shifts. In general, the gate er- gate wavelengths in the visible range where the error is rors and speeds are comparable to the same-species gates below 10−4 and the gate duration is a few 10’s of µs. Cal- over a wide range of wavelengths. We note that there cium presents the lowest error in this range (2.4 × 10−5) are many regions where gate performance is comparable 138 + between a pair of identical ions and a pair of heteroge- while Ba gates are particularly fast with durations + + < 10 µs throughout the visible spectrum. Near the 1550- neous ions. For example, Ca − Sr gates perform well 40 + in the range ∼ 450 − 800 nm which is also favorable for nm telecom wavelength Ca attains a gate error of + + 1.9 × 10−4 and 138Ba+ achieves < 10−5 (the other two the Ca − Ca combination. Accordingly, gates on both species have error > 10−3 so appear less useful in this configurations could be achieved in a multi-ion system range). Although the gate speeds here are slower than via the same OTDF laser beams. in the visible by an order of magnitude (due to a com- The OTDF gate may also be extended to multispecies bination of smaller Lamb-Dicke parameters and smaller crystals that include Be+ or Mg+. This is possible since Stark shifts) and the gate errors (overwhelmed here by all of the alkaline earth species considered in Fig. 2 metastable decay) are correspondingly higher, the possi- have favorable AC Stark shifts and photon scatter rates bility to generate and modulate higher levels of infrared near the traditional LS gate wavelengths of ∼ 313 nm power mitigates these deficiences to some extent. The for Be+ [1] and ∼ 280 nm for Mg+. Extreme mis- use of a phase-modulated retro-reflected beam geome- match of atomic masses remains a concern for co-trapped try could recycle the incoming laser power and would ions, therefore Be+ and Mg+ would be best paired with allow in-situ lattice phase characterization and stabiliza- 40Ca+or 88Sr+. 6

3 3 10 10 (a) CaCa Gate duration (s) 4 10 4 10

Gate error 5 10

5 10 6 10 400 600 800 1000 1200 1400 1600 1800 2000 Wavelength (nm)

3 3 10 10 (b) SrSr Gate duration (s) 4 10

4 10

Gate error 5 10

5 6 10 10 400 600 800 1000 1200 1400 1600 1800 2000 Wavelength (nm)

Figure 3. (Color online) Total fundamental error (solid orange line, left axis), photon scattering error (dotted orange line, left axis) and gate duration (blue line, right axis) for an OTDF two-qubit entangling gate between identical, co-trapped ion species (a) 40Ca+ −40 Ca+ and (b) 88Sr+ −88 Sr+. Total error includes both elastic and inelastic photon scattering decoherence as well as metastable decay. All curves assume a two-ion crystal confined in a potential that would yield a 2 MHz axial frequency for a single 40Ca+ ion and centered upon two laser beams, each with a power of 100 mW, a 10 µm Gaussian waist (peak intensity 637 MW/m2), an angle of 45◦ to the crystal axis, and linear (equal σ+ and σ−) polarization.

VI. EXTRINSIC GATE ERRORS demonstrated gates [13]. A unique class of errors is that associated with the Beyond fundamental physics constraints, realistic tech- non-zero Stark shift required for OTDF operation. For a nical limitations to our proposed two-qubit gate must pulse intensity profile rectangular in time, this Stark shift also be considered. Some important technical noise varies sinusoidally about its mean. Because the OTDF sources include gate laser intensity instability, limitations gate incorporates a spin echo, it is not necessary to cali- to spin-echo dynamical decoupling, and trap frequency brate the precise phase rotation imprinted by this Stark variations caused by dielectric charging or drift in the shift, but it is necessary to ensure that the time inte- trapping potential electronics. We briefly discuss each of gral of the second pulse’s intensity matches that of the these potential error sources below. first. We consider the integral of the sinusoidal compo- Assuming the laser intensity remains fixed during the nent separately from that of the quasi-DC component. gate but varies from shot to shot with a fractional varia- The sinusoidal component can in principle be controlled tion ∆I away from the optimum value I, to lowest order I via a tight phase lock between the lattice optical phase the associated error is and the pulse envelope, although a simpler solution em- π2 ∆I 2 ploys an adiabatic ramp up and down of the pulse so  = . (3) 4 I that the integral becomes negligibly small. The quasi-DC term leads to the requirement that the time-averaged in- An identical error is present in other experimentally tensity of the second pulse matches that of the first. We 7

3 3 10 10 (a) BaBa Gate duration (s) 4 4 10 10

Gate error 5 5 10 10

6 6 10 10 400 600 800 1000 1200 1400 1600 1800 2000 Wavelength (nm)

3 3 10 10 (b) RaRa Gate duration (s) 4 10

4 10

Gate error 5 10

5 6 10 10 400 600 800 1000 1200 1400 1600 1800 2000 Wavelength (nm)

Figure 4. (Color online) Total fundamental error (solid orange line, left axis), photon scattering error (dotted orange line, left axis) and gate duration (blue line, right axis) for an OTDF two-qubit entangling gate between identical, co-trapped ion species (a) 138Ba+ −138 Ba+ and (b) 226Ra+ −226 Ra+. Assumptions are the same as in Fig. 3. note that during a pulse the quasi-DC Stark term from be a stringent requirement in terms of absolute intensity each beam (assuming optimum ion spacing) leads to a stability, the demand here is only on the relative stabil- phase ity between first and second pulses, which is significantly more feasible [36]. In different terms, the gate is sensitive τg φ0 = ∆0,∆ to the Allan deviation of the unwanted σz rotation over 2 π (4) the gate time scale, not its absolute value. = √ 2ηk1ηk2 imprinted onto each ion. A fractional intensity difference If either the gate mode frequency or the lattice travel- ∆I12 ing wave frequency changes from the intended value by I between the first and second pulses therefore leads ∆ω, then the detuning δk of the oscillating dipole force to an uncompensated phase shift ∆φ0 with associated error from the gate mode will change to δk + ∆ω leading to errors. This source of error is not unique to the OTDF 2  = ∆φ0 gate; it can be reduced through appropriate Walsh modu-  ∆I 2 lation of the OTDF beam phase difference [37], although = φ 12 not below the value (to lowest order) 0 I (5) π2 ∆I 2 = 12 . 2ηk1ηk2 I

∆I12 < 2  2 For a representative ηki ∼ 0.07 this requires I ∼ π ∆ω −4 −4  = . (6) 3 × 10 to achieve  < 10 . While this appears to 4 δk 8

3 3 10 10 (a) CaSr Gate duration (s) 4 10 4 10

Gate error 5 10

5 10 6 10 400 600 800 1000 1200 1400 1600 1800 2000 Wavelength (nm)

3 3 10 10 (b) SrBa Gate duration (s) 4 10

4 10

Gate error 5 10

5 6 10 10 400 600 800 1000 1200 1400 1600 1800 2000 Wavelength (nm)

3 3 10 10 (c) BaRa Gate duration (s) 4 10

4 10

Gate error 5 10

5 6 10 10 400 600 800 1000 1200 1400 1600 1800 2000 Wavelength (nm)

Figure 5. (Color online) Total fundamental error (solid orange line, left axis), photon scattering error (dotted orange line, left axis) and gate duration (blue line, right axis) for an OTDF two-qubit entangling gate between pairs of heterogeneous-species, co-trapped ions (chosen for favorable mass ratios) (a) 40Ca+ −88 Sr+, (b) 88Sr+ −138 Ba+, and (c) 138Ba+ −226 Ra+. Assumptions are the same as in Fig. 3.

VII. CONCLUSIONS optical-transition qubits capable of an error rate below 10−4 over a wide range of wavelengths. An entangling gate with wavelength tunability over a sizeable fraction We have presented a detailed analysis of a two-qubit of the laser-accessible electromagnetic spectrum gives an entangling gate based on optical-dipole forces acting on 9 unprecedented degree of flexibility with respect to, for ωki ≡ (ωk − ωi) is the transition angular frequency be- example, Lamb-Dicke confinement, required laser power, tween states i and k. From the Wigner-Eckart theorem, and gate fidelity, parameters that may be traded in dif- the matrix elements are ferent experimental circumstances. 2 The OTDF gate scheme is naturally suited for en- | hk| µq |ii | = tangling ions of different species or for optical-clock- 3  2 3π0¯hc Ji 1 Jk transition qubits (in contrast with some earlier LS gates). 3 AJkJi (2Jk + 1) , (A3) ω mi q −mk In systems where entangling gates between both similar ki and dissimilar pairs of ions is required, the same beams whereh ¯ is the reduced Planck constant and 0 is the can be used for both gates. Future work could include permittivity of free space. The coefficient AJkJi is the extension of this OTDF gate technique to atom-molecule spontaneous emission rate between an excited state |ki or molecule-molecule entanglement using metastable su- and the state |ii, while the total electronic angular mo- perpositions of molecular ion states, augmenting recent mentum of a state |ai and its projection along the quan- molecular ion gate proposals [38, 39] and demonstra- tization axis are represented by Ja and ma, respectively. tions [40]. The incorporation of spin-echo sequences into For this work we assume that all excited, intermediate the gate further improves the viability of magnetic-field- states |ki are sublevels of the P1/2 and P3/2 manifolds. sensitive optical-transition qubits. We anticipate that For small laser detunings the first term in Eq. A1 dom- this wavelength-insensitive OTDF gate will significantly inates, but for larger detunings the second term (the expand the experimental toolbox available to ion trap Bloch-Siegert shift caused by the counter-rotating elec- quantum computing researchers. tric field) must also be included. Making the substitution 2 E0 = 2I/(0c), which relates the electric field amplitude to the intensity I, and summing over all intermediate ACKNOWLEDGMENTS states |ki gives the total energy shift: ( 2 2 !  ) X Iˆqµki,q 1 1 We acknowledge funding from GTRI for preparation ∆E = + . 20¯hc ωki − ωL ωki + ωL of this manuscript. We also thank David Hayes, Adam k,q Meier, and Nicholas Guise for useful discussions. (A4)

Appendix B: Intrinsic Decoherence due to Appendix A: AC Stark Shifts Spontaneous Decay and Photon Scattering

The OTDF gate relies on a differential AC Stark shift The OTDF entangling gate is performed on a between an S and a D electronic state. For a given state metastable superposition of S and D electronic states |ii, the AC Stark shift induced by a laser beam with and is subject to two intrinsic sources of decoherence: electric field amplitude E0 and polarization components off-resonant photon scattering from the gate laser beams ˆq (q = 0, ±1) due to off-resonant dipole interaction with and spontaneous decay from D → S. The S − D transi- another state |ki is [41] tions of 40Ca+, 88Sr+, 138Ba+, and 226Ra+are at optical wavelengths spanning from to 674 nm to 2050 nm. The 2 2 2 !   OTDF gate laser beams will induce off-resonant sponta- E0 ˆqµki,q 1 1 ∆Ek = + , (A1) neous photon scattering events via the excited P1/2 and 4¯h ωki − ωL ωki + ωL P3/2 levels, which will cause qubit decoherence during the entangling operation. With such a large qubit splitting where relative to gate laser beam detuning from the P manifold, both elastic and inelastic photon scattering decoherence 2 2 µki,q ≡ | hk| µq |ii | (A2) must be considered. To treat inelastic (Raman) scattering, we begin with a and µq are the components of the electric-dipole opera- general expression for the differential photon scattering tor. Here ωL is the angular frequency of the laser and cross section, dσ/dΩ, from Ref. [42]:

2 dσ i→f ω (ω − ω )3 hf| ˆ · ~µ |ki hk| ˆ· ~µ |ii hf| ˆ· ~µ |ki hk| ˆ · ~µ |ii L L fi X s s = 2 + . (B1) dΩ (4π )2¯h c4 ωki − ωL ωL + ωkf 0 k

In the above expression, ~µ is the electric dipole opera- tor. As in Appendix A, transition angular frequencies 10 between states a and b are written as ωba ≡ (ωb − ωa). photon couples state |ii to |ki and the gate laser photon Photon scattering occurs between an initial state |ii and connects |ki to |fi. The ‘sum-frequency’ term may be final state |fi via intermediate excited (i.e. P ) atomic neglected in experiments where ωki − ωL is much smaller states |ki. The gate laser polarization unit vector is ˆ (for some k) than ωL + ωkf . However, this term is non- and the polarization of the scattered photon is ˆs. The negligible in the far-detuned gate laser regimes presented first (‘difference-frequency’) term in the sum of Eq. B1 in this Article. corresponds to off-resonant scattering through state |ki, We integrate Eq. B1 over all solid angle and multiply with the spontaneously-emitted photon leaving the atom by the incoming photon flux (I/(¯hωL)) to obtain the total in state |fi. The second (‘sum-frequency’) term describes scatter rate from state |ii to |fi through all intermediate the opposite process, whereby a spontaneously-emitted states |ki:

s ! 2 3πc2I A A hf| T 1 |ki hk| ˆ· T~ |ii hf| ˆ· T~ |ki hk| T 1 |ii i→f 3 X X JkJi JkJf s s Γ = (ωL − ωfi) (2Jk + 1) 3 3 + , 2¯h ω ω ωki − ωL ωL + ωkf s={0,±1} k ki kf (B2) where we have used the Wigner-Eckart theorem to re- the total electronic spin of the given level (S = 1/2 for all place the electric dipole operators with unitless spher- states considered in this Article). The photon polariza- ical tensors, T . The coefficients AJkJi and AJkJf are tion state for electric dipole transitions is q ∈ {−1, 0, +1}. the spontaneous emission rates between an intermediate The rightmost element of Eq. B3 is a Wigner-3j symbol. level with total electronic angular momentum Jk and an The total inelastic scattering rate of a superposition of initial or final level with angular momentum Ji and Jf , |↑i and |↓i states of a single ion (assuming equal popu- respectively. Here we assume that all intermediate states lations in each state) is are sublevels of the P and P excited states and that 1/2 3/2   the S , D , and D sublevels can be initial or final 1/2 3/2 5/2 1 X X states. In other words, only electric-dipole-allowed scat- Γ = Γ↑→f + Γ↓→f . (B4) in 2   tering channels are considered. Furthermore, note that f6=↑ f6=↓ a Raman scattering event can only occur if ωL > ωfi, otherwise the scattering rate is negative. To obtain the elastic scattering decoherence rate, we The matrix elements of Eq. B2 may be evaluated as take the scattering amplitude from Eq. B2 and apply the result of Ref. [26]. For a superposition of |↑i and |↓i 1 ha| Tq |bi = (B3) states of a single ion (assuming equal populations in each state),  J 1 J  (−1)Ja−ma (−1)La+S+Jb+1 a b , −ma q mb 2 3πc I X 2 Γ = χ↑→↑ − χ↓→↓
, (B5) el 4¯h s s where ma is the projection of state a with angular mo- s={0,±1} mentum Ja on the external magnetic field axis, La is the electronic orbital angular momentum of state a, and S is where we define

s 3 1 ~ ~ 1 ! i→i X ωL hi| Ts |ki hk| ˆ· T |ii hi| ˆ· T |ki hk| Ts |ii χs ≡ (2Jk + 1)AJkJi 6 + . (B6) ω ωki − ωL ωL + ωki k ki

(2) Finally, the total intrinsic decoherence rate, Γtot, of a Appendix C: Geometric Phase Gate Theory two-ion system where both ions are in a superposition of their respective S and D states (with equal populations The system Hamiltonian is given by in each state) is   H = Hm + Hs + HLI . (C1) (2) X 1 Γtot = Γel,j + Γin,j + AD,j , (B7) 2 Hm represents the ions’ motional degrees of freedom, Hs j their spins, and HLI the laser-ion interactions which can where j is the ion index and AD,j is the spontaneous couple spin to motion. In the following discussion we will decay rate of the chosen D level of ion j. make successive transformations to the spin and spin- 11 motion interaction frames with Hamiltonians Hs and This is the Hamiltonian for a collection of harmonic os- s,m H in these frames: cillator modes each with position qk, with unit mass (in s s this choice of units), and with (angular) frequency Ωk H = Hm + HLI , (C2) 2 given by Ωk = βk. If we introduce raising and lowering operators a† and Hs,m = Hs,m. (C3) k LI ak for each mode in the collection in the usual way, then we can write the position operators

1. Ion Motions from Mode Amplitudes r ¯h † qk = (ak + ak). (C9) Our first task is to determine effective Lamb-Dicke pa- 2Ωk rameters for a multi-species, multi-ion, optical-lattice in- teraction. The following discussion was inspired by [43]; This expression gives us the transformation from normal- here we expand that treatment slightly. For a general mode amplitudes to real-space coordinates: configuration of N ions in harmonic wells with linear cou- ! plings, we have the following Hamiltonian: Xi 1 X xi = √ = √ qkbk mi mi 1 X 2 X k Hm = mix˙i + Aijxixj. (C4) i (C10) 2 r i ij X ¯h † = bki (ak + ak). 2miΩk Here, each ion has mass mi, and its displacement from k equilibrium along the principal axis of its local potential is x (there are three of these variables for each ion, one i √ Here we have defined the ith component of bk as (bk)i ≡ for each principal axis direction). We define Xi ≡ mixi bki. What remains is to determine the matrix B and its Aij (corresponding vector X) and Bij = √ (correspond- eigenvectors for a particular configuration of ions. mimj ing matrix B). Then Here we choose to consider motion only along the trap axis of linear symmetry, although our discussion could be 1 X 2 X H = X˙ + B X X . (C5) expanded to encompass radial directions in a straightfor- m 2 i ij i j ward manner. We assume that ion i by itself has fre- i ij quency νi in the potential near equilibrium (in the ab- Let bk be an orthonormal basis of eigenvectors of B with sence of couplings), and we assume the equilibrium posi- corresponding eigenvalues βk, so that B · bk = βkbk, tion of each ion along the axis is given by Ri, such that and define components qk of X in this basis such that Rj > Ri for j > i. Then P X = k qkbk. Then 1 X 1 X X X H = m x˙ 2 + m ν2x2 + V (x). (C11) BijXiXj = BijXjXi m 2 i i 2 i i i eq ij ij i i = (B · X) · X ! Here Veq(x) represents the potential between the ions at X equilibrium, assumed to be second-order in the ion dis- = B · q b · X k k placements. We assume that the first-order dependence k has already been captured in the equilibrium positions ! (C6) X of the ions. To determine Veq, we expand the Coulomb = β q b · X k k k interaction: k X   = βkqkbk · qjbj e2 X 1 kj V (x) = 0 Coulomb 4π R + x − R − x  X 2 0 i,j:j>i j j i i = βkqk. k   e2 X 1 1 = 0  xi−xj  Therefore 4π0 Rj − Ri 1 − i,j:j>i Rj −Ri 1 X ˙ 2 X 2 " Hm = Xi + βkqk. (C7) e2 X 1 2 = 0 i k 4π R − R 0 i,j:j>i j i 2 Now P X˙ = X˙ · X˙ = P q˙ 2 , so !# i i k k x − x  x − x 2 × 1 + i j + i j . 1 X R − R R − R H = q˙ 2 + β q2. (C8) j i j i m 2 k k k k (C12) 12

As discussed above, Veq keeps only the second-order term: and eigenvectors of B are   p 2 2 Ω2 1 + µ − 1 − µ + µ ν1 e2 X 1 β = 1 = , V (x) = 0 x2 − 2x x + x2
1 2 2µ eq 4π |R − R |3 i i j j 0 i,j:j>i j i   p 2 2 2 1 + µ + 1 − µ + µ ν1 2 Ω2 1 e0 X 1 2 2
β2 = = , = 3 xi − 2xixj + xj . 2 2µ (C17) 2 4π0 |Rj − Ri| i,j:j6=i  q  2 (C13) b1 = (b11, b12) = b0, 1 − b0 , q  b = (b , b ) = 1 − b2, −b , To second order our Hamiltonian becomes 2 21 22 0 0

where we have defined 1 X 2 1 X 2 2 H = m x˙ + m ν x p 2 m 2 i i 2 i i i 1 − µ + 1 − µ + µ i i b0 = . (C18) r  2 2 p 2 1 e0 X 1 2 2
µ + 1 − µ + 1 − µ + µ + 3 xi − 2xixj + xj . (C14) 2 4π0 i,j:j6=i |Rj − Ri| This gives all the information needed to determine real- space motions from normal-mode amplitudes via Equa- Comparing this with Equation C4 we determine that tion C10. 2. Two-laser Stark Shifts

 e2 1 2 0 P 1 Assume that the two optical-dipole-force beams are  miν + 3 , if i = j. 2 i 4π0 k6=i |Rk−Ri| Aij = 2 . approximately plane-waves with frequencies ω1 and ω2 −e0 1  3 , if i 6= j. 4π0 |Rj −Ri| and wave-vectors k1 and k2. The laser-beam amplitudes (C15) are assumed constant over the extent of the ions’ mo- tions during the gate, but we allow for different inten- Now we specialize to the situation with only two ions, sities at the two ion positions. We assume that they possibly with different masses. In this case have the same polarization, but the discussion can easily be broadened to arbitrary polarizations by considering each polarization component (π, σ , and σ ) indepen-  2 1/3  2 1/3 + − e0 e0 dently. In this more general case one would take the sum R2 − R1 = 2 = 2 , 2π0m1ν1 2π0m2ν2 of the Stark shifts of the three polarization components to determine the overall optical-dipole force. The electric fields of the two beams along the polarization direction and we find can be expressed as

( E1(r, t) = E1(r) cos(k1 · r − ω1t − φ1), m ν2, if i = j. (C19) i i E (r, t) = E (r) cos(k · r − ω t − φ ). Aij = 1 2 1 2 . (C16) 2 2 2 2 2 − 2 miνi = − 2 mjνj , if i 6= j. Here r and t represent the space and time coordinates of the fields. We define E0 = E1(r), ∆E = E2(r) − E1(r), In terms of the mass ratio µ = m2/m1, the eigenvalues so that the total electric field can be expressed as

E(r, t) = E1(r, t) + E2(r, t)

= E1(r) cos(k1 · r − ω1t − φ1) + (E2(r) − E1(r)) cos(k2 · r − ω2t − φ2) + E1(r) cos(k2 · r − ω2t − φ2)

= E0(cos(k1 · r − ω1t − φ1) + cos(k2 · r − ω2t − φ2)) + ∆E cos(k2 · r − ω2t − φ2) 1 1 (C20) = 2E cos( [k · r − ω t − φ + k · r − ω t − φ ]) cos( [k · r − ω t − φ − k · r + ω t + φ ]]) 0 2 1 1 1 2 2 2 2 1 1 1 2 2 2 + ∆E cos(k2 · r − ω2t − φ2)

= 2E0 cos(kΣ · r − ωΣt − φΣ) cos(k∆ · r − ω∆t − φ∆) + ∆E cos(k2 · r − ω2t − φ2).

1 1 1 1 1 We have defined kΣ = 2 (k1 + k2), k∆ = 2 (k1 − k2), ωΣ = 2 (ω1 + ω2), ω∆ = 2 (ω1 − ω2), φΣ = 2 (φ1 + φ2), 13

1 and φ∆ = 2 (φ1 − φ2). Here we are interested in the AC- compared to optical frequencies but short compared to Stark shift induced by these beams, so we take the square the frequency difference between the beams. The Stark of the electric field and time-average over a duration long shift is then

∆(r, t) = γhE(r, t)2i 2 2 2 ∆(r, t) = γh4E0 cos (kΣ · r − ωΣt − φΣ) cos (k∆ · r − ω∆t − φ∆) + 4E0∆E cos(kΣ · r − ωΣt − φΣ) cos(k∆ · r − ω∆t − φ∆) cos(k2 · r − ω2t − φ2) 2 2 + ∆E cos (k2 · r − ω2t − φ2)i 2 2 = γ(2E0 cos (k∆ · r − ω∆t − φ∆) + h4E0∆E cos(kΣ · r − ωΣt − φΣ) cos(k∆ · r − ω∆t − φ∆) cos(k2 · r − ω2t − φ2)i 1 + ∆E2) 2 2 2 (C21) = γ(2E0 cos (k∆ · r − ω∆t − φ∆) + h2E0∆E cos(k∆ · r − ω∆t − φ∆)(cos(kΣ · r − ωΣt − φΣ + k2 · r − ω2t − φ2) 1 + cos(k · r − ω t − φ − k · r + ω t + φ ))i + ∆E2) Σ Σ Σ 2 2 2 2 2 2 = γ(2E0 cos (k∆ · r − ω∆t − φ∆) 1 + h2E ∆E cos(k · r − ω t − φ ) cos(k · r − ω t − φ )i + ∆E2) 0 ∆ ∆ ∆ ∆ ∆ ∆ 2 1 = γ[2(E2 + E ∆E) cos2(k · r − ω t − φ ) + ∆E2]. 0 0 ∆ ∆ ∆ 2

γ represents the constant of proportionality between in- interaction frame, the laser-ion Hamiltonian is given by tensity and Stark shift (see Eq. A1). Defining ∆0(r) = q γ 2 γ 2 γ 2 X1 (E0 + E0∆E) = sign(γ) ( E1 )( E2 ) (the geometric s 2 2 2 HLI = (∆↑(rj, t) − ∆↓(rj, t))σz,j 0 2 mean of the individual beam Stark shifts) and ∆0(r) = j γ 2 1 γ 2 γ 2 4 ∆E + ∆0 = 2 ( 2 E1 + 2 E2 ) (the arithmetic mean), we 1  can further simplify this to + (∆↑(rj, t) + ∆↓(rj, t)) , (C23) 2 2 ∆(r, t) = γ[(E + E0∆E) 0 where the sum runs over the two ions. Defining the differ- 1 2 × {1 + cos(2k∆ · r − 2ω∆t − 2φ∆)} + ∆E ] ential Stark shift ∆∆ = ∆↑ −∆↓ and (twice) the common 2 Stark shift ∆ = ∆ + ∆ we find 0 Σ ↑ ↓ = 2∆0(r) cos(2k∆ · r − 2ω∆t − 2φ∆) + 2∆0(r). (C22) X 1 1  Hs = ∆ (r , t)σ + ∆ (r , t) LI 2 ∆ j z,j 2 Σ j j 3. Optical Dipole Force Hamiltonian X = [∆0,∆(rj)σz,j + ∆0,Σ(rj)] (C24) j The first term in Eq. C22 represents the desired  0 optical-dipole potential. The second (2∆0(r)) term rep- × cos(2k∆ · rj − 2ω∆t − 2φ∆) . resents additional Stark shifts. These must be accounted for during operation of the gate, either by the addition of appropriate phase rotations or through a spin-echo pulse. Considering here motion only along the trap axis, we can This term also leads to optical-dipole forces, but these write k∆·rj = k∆(x0,j +xj) (where x0,j is the equilibrium are generally negligible unless there is significant spa- position). We define tial variation in beam intensity. Furthermore, for slowly s time-varying laser pulses this term is off-resonant with ¯h ηkj = 2k∆bkj (C25) any ion motion, further reducing its impact. We ignore 2mjΩk it in the following discussion. Each ion’s possible states (|↑i and |↓i) will experience a different Stark shift ∆↑(r, t) and ∆↓(r, t). In the spin φ0,j = 2φ∆ − 2k∆x0,j (C26) 14 and substitute Eq. C10 to find to obtain

X 1 Hs,m = [∆ (x )σ + ∆ (x )] LI,k 2 0,∆ 0,j z,j 0,Σ 0,j  j s X HLI = [∆0,∆(rj)σz,j + ∆0,Σ(rj)] × exp(i{−2ω∆t − φ0,j}) j   −iΩkt † iΩkt × {iηkj(ake + ake )} + H.c. × cos(2k∆[x0,j + xj] − 2ω∆t − 2φ∆) (C29) X = [∆0,∆(x0,j)σz,j + ∆0,Σ(x0,j)] X i = a† exp(iδ t) exp(−iφ )η j k k 2 0,j kj 1  j × exp(i{2k [x + x ] − 2ω t − 2φ }) + H.c.  2 ∆ 0,j j ∆ ∆ × [∆0,∆(x0,j)σz,j + ∆0,Σ(x0,j)] + H.c. X 1 = [∆ (x )σ + ∆ (x )] 2 0,∆ 0,j z,j 0,Σ 0,j j To more easily understand the dynamics we write out all  four components of this Hamiltonian for two ions (diag- X † × exp(i{ [ηkj(ak + ak)] − 2ω∆t − φ0,j}) + H.c. onal in the σz basis). We define k i (C27) F = [η exp(−iφ )∆ (x ) k,∆,+ 2 k1 0,1 0,∆ 0,1 + ηk2 exp(−iφ0,2)∆0,∆(x0,2)] Here we have assumed that the beam intensities do not i Fk,∆,− = [ηk1 exp(−iφ0,1)∆0,∆(x0,1) vary appreciably over the extent of the ions’ motions. In 2 (C30) the Lamb-Dicke regime (keeping terms only to first order − ηk2 exp(−iφ0,2)∆0,∆(x0,2)] in {ηkj}) i F = [η exp(−iφ )∆ (x ) k,Σ,+ 2 k1 0,1 0,Σ 0,1 + ηk2 exp(−iφ0,2)∆0,Σ(x0,2)] X 1 Hs = [∆ (x )σ + ∆ (x )] in terms of which LI 2 0,∆ 0,j z,j 0,Σ 0,j j s,m † H (↑↑) = (Fk,∆,+ + Fk,Σ,+) exp(iδkt)a + H.c. × exp(i{−2ω∆t − φ0,j}) LI,k k  s,m † X HLI,k(↑↓) = (Fk,∆,− + Fk,Σ,+) exp(iδkt)ak + H.c. × exp(i{ [η (a + a† )]}) + H.c. kj k k s,m † H (↓↑) = (−Fk,∆,− + Fk,Σ,+) exp(iδkt)a + H.c. k (C28) LI,k k X 1 s,m † ≈ [∆ (x )σ + ∆ (x )] HLI,k(↓↓) = (−Fk,∆,+ + Fk,Σ,+) exp(iδkt)ak + H.c. 2 0,∆ 0,j z,j 0,Σ 0,j j (C31) × exp(i{−2ω∆t − φ0,j}) This is a spin-dependent force Hamiltonian, and it drives X  the usual closed trajectories in motional phase space for × (1 + i{ [η (a + a† )]}) + H.c. kj k k appropriately chosen pulse durations and detunings. As- k suming a fixed detuning and intensity, after an interval 2π τg = n the mode has undergone n circles in phase δk space and the state has acquired a geometric phase

Each of the terms in this sum (over k) commutes with 0 2πn 2 Φ (↑↑) = sign(δk) 2 |(Fk,∆,+ + Fk,Σ,+)| the other terms, so we can consider the dynamics of δk each term separately, writing Hs = Hs + P Hs . LI LI,0 k LI,k 0 2πn 2 The motion-independent term Hs represents an oscil- Φ (↑↓) = sign(δk) |(Fk,∆,− + Fk,Σ,+)| LI,0 δ2 lating Stark shift whose effect can be accounted for ei- k (C32) ther through precise timing, in which case it leads to a 0 2πn 2 Φ (↓↑) = sign(δk) 2 |(−Fk,∆,− + Fk,Σ,+)| phase shift (this would be difficult), or through an adia- δk batic ramp up and down of the laser pulses (in which 2πn Φ0(↓↓) = sign(δ ) |(−F + F )|2. case the accumulated phase can be made negligible). k 2 k,∆,+ k,Σ,+ δk We now move to the interaction frame of the motion −iΩkt † † iΩkt (ak → ake , ak → ake ) and define mode detun- We note that the interaction as derived cannot be used ings δk = Ωk − 2ω∆. We neglect counter-rotating terms in the naive way to achieve a phase gate with a single 15

4π pulse, because all four states in general acquire different ing is δk = , which leads to τg phases. Rather we imagine a scheme with two pulses sep- arated by a spin echo, so that the total phase acquired s by |↑↑i and by |↓↓i are equal, and the phase acquired by 2π2 |↑↓i and by |↓↑i are also equal. This has the added ben- τg = . |η η cos(φ − φ )∆ (x )∆ (x )| efit of cancelling the quasi-static Stark shifts neglected k1 k2 0,2 0,1 0,∆ 0,1 0,∆ 0,2 earlier as well as increasing the resilience of the gate to (C35) slow qubit frequency drifts. In this case the acquired This duration is minimized when the two ions are spaced geometric phase is by a half-integer multiple of the lattice wavelength: φ0,2− φ0,1 = mπ. 2π 2 Note that the chosen detuning δk and gate duration τg Φ(↑↑) = sign(δk) 2 (|(Fk,∆,+ + Fk,Σ,+)| δk do not in general close the trajectories in phase space for 2 + |(−Fk,∆,+ + Fk,Σ,+)| ) the spectator mode. For example, if the gate is performed 2π with a detuning close to the two-ion breathing mode, the Φ(↑↓) = sign(δ ) (|(F + F )|2 center-of-mass-mode phase-space trajectories may not be k δ2 k,∆,− k,Σ,+ k closed. However, given the much larger detuning from 2 + |(−Fk,∆,− + Fk,Σ,+)| ) the this mode, one can choose from several possible gate (C33) durations and detunings close to the desired values for 2π 2 Φ(↓↑) = sign(δk) 2 (|(Fk,∆,− + Fk,Σ,+)| which both sets of trajectories are closed. δk 2 Gate sensitivity to mode-frequency variations is often + |(−Fk,∆,− + Fk,Σ,+)| ) reduced through appropriate Walsh modulation of the 2π Φ(↓↓) = sign(δ ) (|(F + F )|2 OTDF beam phase difference [37]. In the simplest sit- k δ2 k,∆,+ k,Σ,+ k uation this is achieved automatically when Fk,Σ,+ = 0. 2 + |(−Fk,∆,+ + Fk,Σ,+)| ). This will be the case only in special scenarios, such as for two ions of the same species, illuminated at the same To implement the ideal phase gate we want intensity, spaced by an integer lattice wavelength mul- π |Φ(↑↑) − Φ(↑↓)| = 2 , so tiple, driven near the breathing mode (ηk1 = −ηk2) (see Eq. C30). Nevertheless, Walsh modulation can still

4πηk1ηk2 cos(φ0,2 − φ0,1)∆0,∆(x0,1)∆0,∆(x0,2) π achieve the desired effect in more general situations via 2 = . δk 2 a sequence consisting of four pulses: reverse the OTDF (C34) phase between pulses one and two, flip the spins between For a gate at fixed detuning incorporating two loops in pulses two and three, and again reverse the phase be- phase space within a duration τg, the appropriate detun- tween pulses three and four.

[1] D. Leibfried, B. DeMarco, V. Meyer, D. Lucas, M. Bar- D. Leibfried, and D. J. Wineland, Nature 528, 380 rett, J. Britton, W. M. Itano, B. Jelenkovic, C. Langer, (2015). T. Rosenband, and D. J. Wineland, Nature 422, 412 [10] C. D. Bruzewicz, R. McConnell, J. Stuart, J. M. Sage, (2003). and J. Chiaverini, npj Quantum Information 5, 1 (2019), [2] L. Aolita, K. Kim, J. Benhelm, C. F. Roos, and number: 1 Publisher: Nature Publishing Group. H. H¨affner,Physical Review A 76, 040303 (2007). [11] A. C. Hughes, V. M. Sch¨afer, K. Thirumalai, D. P. [3] C. J. Ballance, T. P. Harty, N. M. Linke, M. A. Sepiol, Nadlinger, S. R. Woodrow, D. M. Lucas, and C. J. and D. M. Lucas, Physical Review Letters 117, 060504 Ballance, Physical Review Letters 125, 080504 (2020), (2016). publisher: American Physical Society. [4] C. H. Baldwin, B. J. Bjork, M. Foss-Feig, J. P. Gae- [12] C. Langer, R. Ozeri, J. D. Jost, J. Chiaverini, B. De- bler, D. Hayes, M. G. Kokish, C. Langer, J. A. Sedlacek, Marco, A. Ben-Kish, R. B. Blakestad, J. Britton, D. B. D. Stack, and G. Vittorini, arXiv:2003.01102 [physics, Hume, W. M. Itano, D. Leibfried, R. Reichle, T. Rosen- physics:quant-ph] (2020), arXiv: 2003.01102. band, T. Schaetz, P. O. Schmidt, and D. J. Wineland, [5] A. Sørensen and K. Mølmer, Physical Review A 62, Physical Review Letters 95, 060502 (2005). 022311 (2000). [13] J. Benhelm, G. Kirchmair, C. F. Roos, and R. Blatt, [6] J. P. Gaebler, T. R. Tan, Y. Lin, Y. Wan, R. Bowler, Nat Phys 4, 463 (2008). A. C. Keith, S. Glancy, K. Coakley, E. Knill, D. Leibfried, [14] Y. Colombe, D. H. Slichter, A. C. Wilson, D. Leibfried, and D. J. Wineland, Physical Review Letters 117, 060505 and D. J. Wineland, Optics Express 22, 19783 (2014). (2016). [15] R. J. Niffenegger, J. Stuart, C. Sorace-Agaskar, D. Kha- [7] C. F. Roos, New Journal of Physics 10, 013002 (2008). ras, S. Bramhavar, C. D. Bruzewicz, W. Loh, R. Mc- [8] K. Kim, C. F. Roos, L. Aolita, H. H¨affner,V. Nebendahl, Connell, D. Reens, G. N. West, J. M. Sage, and and R. Blatt, Physical Review A 77, 050303 (2008). J. Chiaverini, arXiv:2001.05052 [physics, physics:quant- [9] T. R. Tan, J. P. Gaebler, Y. Lin, Y. Wan, R. Bowler, ph] (2020), arXiv: 2001.05052. 16

[16] S. X. Wang, G. Hao Low, N. S. Lachenmyer, Y. Ge, P. F. H. Shen, P. Jurcevic, B. P. Lanyon, P. Love, R. Babbush, Herskind, and I. L. Chuang, Journal of Applied Physics A. Aspuru-Guzik, R. Blatt, and C. F. Roos, Phys. Rev. 110, 104901 (2011). X 8, 031022 (2018). [17] D. T. C. Allcock, T. P. Harty, H. A. Janacek, N. M. [30] Q. Wang, C. Monroe, and Y. Nam, arXiv:2004.04151 Linke, C. J. Ballance, A. M. Steane, D. M. Lucas, R. L. [quant-ph] (2020). Jarecki, S. D. Habermehl, M. G. Blain, D. Stick, and [31] J. W. Britton, B. C. Sawyer, A. C. Keith, C.-C. J. D. L. Moehring, Applied Physics B 107, 913 (2012). Wang, J. K. Freericks, H. Uys, M. J. Biercuk, and J. J. [18] A. H. Myerson, D. J. Szwer, S. C. Webster, D. T. C. Bollinger, Nature 484, 489 (2012). Allcock, M. J. Curtis, G. Imreh, J. A. Sherman, D. N. [32] P. Jurcevic, H. Shen, P. Hauke, C. Maier, T. Brydges, Stacey, A. M. Steane, and D. M. Lucas, Physical Review C. Hempel, B. Lanyon, M. Heyl, R. Blatt, and C. Roos, Letters 100, 200502 (2008). Phys. Rev. Lett. 119, 080501 (2017). [19] H. Haffner, C. Roos, and R. Blatt, Physics Reports 469, [33] J. Zhang, G. Pagano, P. W. Hess, A. Kyprianidis, 155 (2008). P. Becker, H. Kaplan, A. V. Gorshkov, Z.-X. Gong, and [20] J. Benhelm, G. Kirchmair, R. Gerritsma, F. Z¨ahringer, C. Monroe, Nature 551, 601 (2017). T. Monz, P. Schindler, M. Chwalla, W. H¨ansel,M. Hen- [34] G. Pagano, A. Bapat, P. Becker, K. S. Collins, A. De, nrich, C. F. Roos, and R. Blatt, Physica Scripta T137, P. W. Hess, H. B. Kaplan, A. Kyprianidis, W. L. Tan, 014008 (2009). C. Baldwin, L. T. Brady, A. Deshpande, F. Liu, S. Jor- [21] C. J. Ballance, V. M. Sch¨afer,J. P. Home, D. J. Szwer, dan, A. V. Gorshkov, and C. Monroe, Proc Natl Acad S. C. Webster, D. T. C. Allcock, N. M. Linke, T. P. Harty, Sci USA , 202006373 (2020). D. P. L. Aude Craik, D. N. Stacey, A. M. Steane, and [35] D. J. Wineland, C. Monroe, W. M. Itano, D. Leibfried, D. M. Lucas, Nature 528, 384 (2015). B. E. King, and D. M. Meekhof, Journal of Research of [22] The accumulated qubit phase after a gate pulse may be the National Institute of Standards and Technology 103, expressed as the Fourier transform of the pulse envelope 259 (1997). at the optical beatnote frequency. Adiabatic ramps lead [36] J. Thom, B. Yuen, G. Wilpers, E. Riis, and A. G. Sin- to negligible Fourier components at a given beatnote fre- clair, Appl. Phys. B 124, 90 (2018). quency. [37] D. Hayes, S. M. Clark, S. Debnath, D. Hucul, I. V. Inlek, [23] K. K. Mehta, C. Zhang, M. Malinowski, T.-L. Nguyen, K. W. Lee, Q. Quraishi, and C. Monroe, Physical Review M. Stadler, and J. P. Home, arXiv:2002.02258 [physics, Letters 109, 020503 (2012). physics:quant-ph] (2020), arXiv: 2002.02258. [38] E. R. Hudson and W. C. Campbell, Phys. Rev. A 98, [24] P. Taylor, M. Roberts, S. V. Gateva-Kostova, R. B. M. 040302 (2018). Clarke, G. P. Barwood, W. R. C. Rowley, and P. Gill, [39] W. C. Campbell and E. R. Hudson, arXiv:1909.02668 Phys. Rev. A 56, 2699 (1997). [physics, physics:quant-ph] (2019), arXiv: 1909.02668. [25] N. Yu and L. Maleki, Phys. Rev. A 61, 022507 (2000). [40] Y. Lin, D. R. Leibrandt, D. Leibfried, and C.-w. Chou, [26] H. Uys, M. J. Biercuk, A. P. VanDevender, C. Ospelkaus, Nature 581, 273 (2020). D. Meiser, R. Ozeri, and J. J. Bollinger, Physical Review [41] Dmitry. Budker, Atomic physics: an exploration through Letters 105, 200401 (2010). problems and solutions (Oxford University Press, Oxford, [27] Given the extremely large ratio of A-coefficient values 2004). between P and D levels of ∼ 108 for all species considered [42] R. Loudon, The quantum theory of light, 3rd ed. (Uni- here, a gate laser detuning of > 100 MHz from the nearest versity Press, Oxford, 2000). S − D transition is sufficient to neglect this effect for all [43] D. Kielpinski, B. E. King, C. J. Myatt, C. A. Sack- species. ett, Q. A. Turchette, W. M. Itano, C. Monroe, D. J. [28] K. A. Gilmore, J. G. Bohnet, B. C. Sawyer, J. W. Britton, Wineland, and W. H. Zurek, Physical Review A 61 and J. J. Bollinger, Physical Review Letters 118, 263602 (2000), 10.1103/PhysRevA.61.032310. (2017). [29] C. Hempel, C. Maier, J. Romero, J. McClean, T. Monz,