Qudit entanglers using quantum optimal control

Sivaprasad Omanakuttan,1, ∗ Anupam Mitra,1 Michael J. Martin,2, 1 and Ivan H Deutsch1, †

1
Center for Quantum Information and Control (CQuIC), Department of Physics and Astronomy,
University of New Mexico, Albuquerque, New Mexico 87131, USA
2
Materials Physics and Applications Division, Los Alamos National Laboratory, Los Alamos, New Mexico 87545
(Dated: 2022-12-20)
We study the generation of two-qudit entangling quantum logic gates using two techniques in
quantum optimal control. We take advantage of both continuous, Lie-algebraic control and digital,
Lie-group control. In both cases, the key is access to a time-dependent Hamiltonian which can
generate an arbitrary unitary matrix in the group SU(d2 ). We find efficient protocols for creating
arXiv:2212.08799v1 [quant-ph] 17 Dec 2022

high-fidelity entangling gates. As a test of our theory, we study the case of qudits robustly encoded
in nuclear spins of alkaline earth atoms and manipulated with magnetic and optical fields, with
entangling interactions arising from the well-known Rydberg blockade. We applied this in a case
study based on a d = 10 dimensional qudit encoded in the I = 9/2 nuclear spin in 87 Sr, controlled
through a combination of nuclear spin-resonance, a tensor AC-Stark shift, and Rydberg dressing,
which allows us to generate an arbitrary symmetric entangling two-qudit gate such as CPhase.
Our techniques can be used to implement qudit entangling gates for any 2 ≤ d ≤ 10 encoded in the
nuclear spin. We also studied how decoherence due to the finite lifetime of the Rydberg states affects
the creation of the CPhase gate and found, through numerical optimization, a fidelity of 0.9985,
0.9980, 0.9942, and 0.9800 for d = 2, d = 3, d = 5, and d = 7 respectively. This provides a powerful
platform to explore the various applications of quantum information processing of qudits including
metrological enhancement with qudits, quantum simulation, universal quantum computation, and
quantum error correction.

I. INTRODUCTION transmon[2, 25, 26] as well as in trapped ions [27, 28] up

to dimension d = 7. In these experiments, one imple-
In the standard paradigm of quantum information pro- ments qudit gates using constructive methods through
cessing (QIP) one encodes information in qubits, the a prescribed set of Givens rotations [24]. While there
quantum analog of classical bits, by isolating two well- has been substantial progress, much work remains to be
chosen energy levels of the system. In many platforms, done to efficiently implement a high-fidelity universal qu-
one has access and control over multiple levels, which can dit gate set.
enhance our ability to do QIP in a variety of ways [1–5]. In this article, we study an alternative approach based
In particular, one can encode information in base-d > 2 on quantum optimal control. Quantum optimal control
using d-level qudits [1]. With a larger state space per was originally developed in NMR [29] and for coherent
subsystem, qudits offer potential advantages for quan- control of chemistry [30, 31], and has been extensively
tum communication [6], quantum algorithms [7–10], and used in quantum information and quantum computa-
topological quantum systems [11–13] . Quantum com- tion [32]. We consider both continuous Hamiltonian con-
putation with qudits can also reduce circuit complexity trol (Lie algebraic) and digital gate-based control (Lie
and can be advantageous in a variety of NISQ-era ap- group). Quantum optimal control ideas have been used
plications [8–10, 14, 15]. Qudits may also provide sig- widely in quantum information processing and have been
nificantly advantages in quantum error correction and experimentally implemented in a wide range of platforms
fault-tolerant quantum computation [16–20]. ranging from ion traps [33], neutral atoms [34–36], super-
In the gate-based approach to quantum computation conductors [37, 38], and nitrogen vacancy (NV) centers
with qubits, a universal gate set consists of single-qubit [39, 40]. Its use in implementing single qudit gates in
gates in the group SU(2) and one entangling two-qubit the hyperfine states of cesium was demonstrated in the
gate, such as CNOT [21]. This generalizes simply for seminal experiments of Jessen [41], and by the authors
qudits. The universal gate-set consists of single-qudit when encoding qudits in the nuclear spin of alkaline-earth
gates in SU(d) and an entangling two-qudit gate [22– atoms [42]. In this work, we extend these techniques to
24]. Unlike qubits, where native Hamiltonians can be the implementation of entangling gates between qudits.
used to naturally implement the desired gate set, qu- We study qudit entangling gates for any k ≤ d within the
dits require more complex protocols. The gates neces- d-dimensional Hilbert space of each subsystem.
sary for the implementation of the universal gate set have
been recently implemented for qudits in superconducting As a concrete example that demonstrates the power of
the method, we present here an optimal control scheme
to implement entangling gates in qudits encoded in the
nuclear spin of 87 Sr atoms. The nuclear spin is a good
∗ somanakuttan@unm.edu memory for use in quantum information processing given
† ideutsch@unm.edu its weak coupling to the environment and resilience to
 2

other background noise [43–45]. Also, the recent signif- {H0 , H1 , H2 , . . . , Hk }, are generators of the desired Lie
icant achievements of quantum information processing algebra,
h  e.g., su(d). Then i ∃ c(t) such that U [c, T ] =
using the Rydberg blockade [46–48] make this an ideal RT
T exp −i 0 H[c(t)]dt = Utar for any target unitary
platform for exploring quantum computation. Using a
combination of a tunable radio-frequency magnetic field in desired Lie Group, e.g., Utar ∈ SU(d). In addition,
and interactions that arise when atoms are excited to we require T ≥ T∗ , where T∗ is known as the “quantum
high-lying Rydberg states, the atomic qudit is fully con- speed limit,” which sets the minimal time needed for the
trollable. The analysis here extends our previous work on system to be fully controllable.
single-qudit SU(d) gates to the case of two-qudit entan- We consider here open-loop control determined by a
gling gates. We find that one can use quantum optimal well-defined Hamiltonian of the general form,
control to implement high-fidelity entangling qudit gates
H(t) = H (1) (t) + H (2) (t) + Hent , (3)
even in the presence of decoherence arising from the finite
Rydberg-state lifetime. where H (i) (t) are time-dependent Hamiltonians acting
The remainder of this article is organized as follows. In on the individual subsystems, and Hent is the interac-
Sec. (II) we review the fundamentals of quantum control tion that entangles them. Here we include the time de-
and define two approaches that we use here: the Lie alge- pendence in the Hamiltonian that acts on the individual
braic and Lie group theoretic protocols for the generation system as these will be generally easier to implement ex-
of any arbitrary qudit entangling gates. In Sec. (III), we perimentally. In this formulation, Hent = H0 , is the drift
study how control is achieved using numerical optimiza- Hamiltonian. However, one could in principle include
tion based on the well-known GRAPE algorithm [49] and time dependence in the entangling Hamiltonian as well
obtain control waveforms using the Lie-algebraic method. and this may achieve faster gates.
We also use a gradient-based approach to find a digital
sequence of unitary maps that achieves the desired gate
using a Lie group theoretic method. Finally, we study B. Lie-group approach
how decoherence affects the fidelity of these gates. We
give conclusions and outlook of our approach in Sec. (IV).
In the digital, Lie-group approach to quantum con-
trol, we consider a family of unitary maps in the desired
II. CONTROLLABILITY
group that are easily implementable, U (λj ), where {λj }
are the parameters that specify the unitary matrices at
our disposal. The relevant Lie group of interest here is
A complete universal gate set for qudits requires one SU(d2 ), the group of two-qudit unitary matrices in d2
entangling gate. A standard choice is the CPhase gate, dimensions, where the overall phase is removed. The
which is the generalization of CZ gate for qubits, defined system is controllable if ∀Utar ∈ SU(d2 ), ∃{λi } such that
Qk
CPhase |ji |ki = ω jk |ji |ki , (1) j=1 U (λj ) = Utar . Similar to the Lie-algebraic quan-
tum control approach, the goal is to find {λj } through
where ω = exp(2πi/d), the d-th primitive root of identity numeric optimization, e.g., via gradient-based methods.
for a subsystem of dimension d. We can see that for d = 2 For the case of two-qudit gates, a controllable Lie group
we recover the CZ gate. This gate is equivalent to the structure is given as,
qudit-analog of the CNOT gate, known as CSUM gate,
Uλj = Uent ∗ (U1 ⊗ U2 ), (4)
CSUM |ii |ji = |ii |i ⊕ j(mod d)i (2)
where U1,2 ∈ SU(d) and Uent = exp(−iHent t) ∈ / SU(d) ⊗
by Pthe local Hadamard gate for qudits, Hd |ji = SU(d). Thus, we can achieve the target gate to the de-
√1 ij
d i ω |ii. Previous works have studied how to im- sired fidelity by intertwining a sequence of local SU(d)
plement these gates through a well-defined sequence of gates and the available entangling interaction in alter-
maps generated by one-qudit and two-qudit Hamiltoni- nating layers of single qudit gates and entangling gates,
ans [24, 50–52]. We study here the use of numerical op- as shown in Fig. 1(b). This approach is similar to the
timization and the theory of optimal control. construction based on Givens rotation [27]. Here, the
possibility of accessing arbitrary local SU(d) gates makes
this protocol very powerful. A schematic comparison of
A. Lie-algebraic approach both these approaches is shown in Fig. (1).

In the Lie-algebraic approach to quantum control,

we consider a Hamiltonian of the form H[c(t)] = C. Physical Platform: Rydberg atoms
Pk
H0 + j=1 cj (t)Hj , where c(t) = {cj (t)} is the
set of time-dependent classical control waveforms, and To make these ideas concrete, we consider the im-
H0 is called the drift Hamiltonian. The system is plementation of entangling gates in neutral atoms using
said to be “controllable” if the set of Hamiltonians, the strong van der Waals interactions between atoms in
 3

FIG. 1. Comparison of Lie algebra versus Lie group approach for quantum control. (a) Schematic of the continuous-
time Lie-algebraic Papproach for quantum control. The physical systems are governed by the time-dependent Hamiltonian,
H[c(t)] = Hent + kj=1 cj (t)Hj , here with a time-dependent entangling Hamiltonian, Hent . The time-dependent waveforms
{cj (t)} are found through numerical optimization, and this defines the target unitary map of interest through the solution to
the time-dependent Schrödinger equation. (b) Schematic for a digital, Lie-group approach to quantum control of entangling
two-qudit gates. The target unitary is achieved through a discrete series of layers consisting of unitary maps from a given family.
One layer of the scheme consists of single-qudit gates on each subsystem and an entangling interaction between them, applied
for a given time tj . Through numerical optimization, one finds the parameters of the local SU (d)-gates and the entangling
time tj in each layer.

high-lying Rydberg states. We use the Rydberg dress- magnetic dipole moment, which substantially increases
ing paradigm in which one adiabatically superposes the the speed of gates, but this comes at the cost of increased
Rydberg state into the ground states to introduce inter- sensitivity to background magnetic fields and residual
actions between dressed ground states [53–58]. Rydberg tensor light shifts from the trapping lasers. By choosing
dressing has been studied with multiple applications in- the metastable 3 P0 clock state, one retains the robustness
cluding the dynamics of interacting spin models [56–58] of having J = 0 at the cost of much slower gates, since
as well as to prepare metrologically-useful states [59]. the magnetic coupling is now solely to the nuclear spin.
Entanglement between neutral atoms via Rydberg dress- We consider here coherently transferring qudits from the
1
ing has been theoretically proposed for creating qubit S0 ground state to the F = 9/2 state hyperfine states
entangling gates [54, 60, 61] and experimentally imple- of the 3 P2 manifold, which provides for faster and more
mented [55, 62, 63]. flexible control [64], putting technical noise aside.
We study here encoding a qudit in the spin of 87 Sr. In To achieve the entangling interaction, we consider Ry-
the ground state there is neither orbital nor spin angu- dberg dressing, generalizing the mechanism discussed
lar momentum in the electrons, J = 0, and only nuclear in [54, 55, 60]. The AC Stark shift (light shift) asso-
spin, I = 9/2, giving us ten possible levels in which to ciated with a dressed state when a laser is tuned near a
encode our qudit, labeled from |0i = |mI = 9/2i , |1i = Rydberg resonance is modified for two atoms because of
|mI = 7/2i , · · · , |9i = |mI = −9/2i. The nuclear spin is the Rydberg blockade. The deficit between the two-atom
highly isolated from the environment and thus serves as a light shift and twice the one-atom light shift determines
robust memory for quantum information. In [42] we stud- the entangling energy [54]. For the case of qudits, the
ied how we could implement single qudit gates in these same physics holds, but now with a multilevel structure
systems through a combination of a laser-induced ten- and a spectrum of entangling energies. When the spec-
sor light-shift and radio frequency (rf)-induced Larmor trum is nonlinear, the system is controllable.
precession. We generalize to the two-qudit case here. Figure (2) depicts the basic scheme. The levels of the
To implement entangling two-qudit control, we will qudit that will participate in the gate are excited to the
make use of the excitation to the 5sns 3 S1 Rydberg se- first excited 3 P2 state. The Rydberg states in 87 Sr have
ries from the metastable 5s5p 3 PJ first excited states. well-resolved hyperfine splitting. We consider tuning a
The choice of energy levels depends on practical consid- UV dressing laser from the 3 P2 , F = 9/2 magnetic sub-
erations. By choosing the metastable state 5s5p 3 P2 , the levels to the 3 S1 Rydberg states near resonance with a
total electron angular momentum gives rise to a large well-chosen hyperfine manifold, e.g., F ′ = 11/2. In the
 4

FIG. 2. Schematic for designing two-qudit entangling interactions in 87 Sr neutral atoms. (a) A k ≤ d-dimensional
qudit is encoded in memory in the nuclear spin with d = 10 magnetic sublevels in the electronic ground state (5s2 ) 1 S0 . When the
gate is to be performed, the k levels (here k = 3) are transferred coherently to the metastable clock states (5s5p) 3 P2 , F = 11/2 in
the presence of a bias magnetic field. The system becomes controllable by adiabatically dressing the 3 P2 with Rydberg character
through the application of a near-resonant laser with Rabi frequency ΩL and detuning ∆L with respect to the hyperfine manifold
(5sns) 3 S1 , F ′ = 9/2 in the Rydberg series. Control is then achieved through the application of a phase-modulated rf-field with
Rabi rate Ωrf which acts on the dressed states to generate a nonlinear Larmor precession. The entanglement arises due to the
Rydberg blockade. The coupling of the state of two qudits for a perfect blockade as depicted in (b), where i is a state from the
first qudit and j is from the second qudit, excited by E two Rabi frequencies and detunings determined by the Clebsch-Gordan
e
coefficients and Zeeman shifts. The state |iji → ij is the dressed state given in Eq. (6). The spectrum of eigenvalues of the
entangling Hamiltonian Eq.(5) is given in (c) as a function of i and j where the function chosen is f (i, j) = 10i + j; 0 ≤ i, j < 10.
The spectrum indicates 10 parabolas, where each parabola corresponds to the effect of a single state in the first atom sees due
to all the states in the second atom. This nonlinear spectrum arises through a combination of the tensor AC Stark shift and
the Rydberg blockade, making the system controllable, allowing us to implement any symmetric two qudit gate in this system
of interest.

presence of a bias magnetic field, due to the difference combination of the tensor light shift and Rydberg block-
in the g-factors, the manifolds will be differently Zeeman ade. This nonlinearity makes the Hamiltonian control-
shifted. The different magnetic sublevels that define the lable; further details are discussed in Appendix (B).
qudit will thus be differently detuned to the Rydberg The time-dependent Hamiltonian necessary for the
magnetic sublevels. Due to this and the Clebsch-Gordan Lie-algebraic control can be chosen as phase-modulated
coefficients associated with the different transitions, each Larmor precession, Hmag = −µ · B(t), with µ = gF µB F
sublevel will be differently dressed (equivalently, there is the magnetic dipole
 vector operator, and
 where B(t) =
a tensor light shift). When two atoms are dressed, the Bk ez + BT Re (ex + iey )e−i(ωrf t+φ(t)) . Defining the
effect of the Rydberg blockade modifies the spectrum as auxilary subspace, a, for the levels in hyperfine mani-
discussed above. An example of two sublevels (one from fold {5s5p 3 P2 , F = 9/2} and the subspace, r, for the
each atom) is shown in Fig. (2 b). Diagonializing this levels {5sns 3 S1 , F ′ = 11/2} in the Rydberg hyperfine
atom-laser Hamiltonian under the approximation of a manifold, we have gF (r)/gF (a) ≈ 2. Thus defining the
perfect Rydberg blockade yields the representation Zeeman shift ω0 = gF (a)Bk , the Larmor precession fre-
X ED quency Ωrf = gF (a)BT , and choosing rf drive on res-
Hent = e
E ij ij e ,
ij (5) onance in the a-manifold, ωrf = ω0 , in the co-rotating
ij frame at ω0 , the Hamiltonian is
(a)
 
where the tilde indicates dressed states, Hmag (t) = Ωrf cos φ(t)Fxa + sin φ(t)Fya ,
  (7)
E (r)
Hmag (t) = 2Ωrf cos φ(t)Fxr + sin φ(t)Fyr + ω0 Fzr ,
e = Cij |iji + Cri j |ri ji + Cirj |irj i ,
ij (6)
where Fia , Fir are the spin angular momentum operators
and E ij are the light shifts originating from these in- in the respective subspaces along axis i ∈ {x, y, z}.
teractions. The spectrum of the entangling Hamiltonian As the Hmag acts on the laser-dressed states defined in
shown in Fig. (2c) gives us insight into the controlla- Eq. (6), which are superpositions of a and r states that
bility of the system. As ordered, the spectrum reveals have different g-factors, one needs to find the action of
the structure of 10 quadratic potentials arising from a the magnetic interaction in the dressed basis. Due to
 5

the nonlinearity, the action of the rf-magnetic driving on Considering a global rf-magnetic interaction, the Hmag
the dressed states is no longer simple Larmor precession. acts on both qudits as

E h i h i
e = Cij Hmag
(Hmag (t) ⊗ 1 + 1 ⊗ Hmag (t)]) ij (a) (a)
(t) ⊗ Hmag (r)
(t) |iji + Cri j Hmag (a)
(t) ⊗ Hmag (t) |ri ji
h i (8)
(a) (r)
+ Cirj Hmag (t) ⊗ Hmag (t) |irj i .

Thus in the dressed basis, the Hamiltonian is H(t) = is symmetric with respect to the exchange of the qu-
He [φ(t)] + Hent , where the action of the magnetic field in dits, we consider here symmetric gates, with global con-
the dressed basis is given by the Hamiltonian, trol. We seek, through numerical optimization, the time-
dependent rf-phase, φ(t). To achieve this we employ
e [φ(t)]
H the well-known GRAPE algorithm [49]. To implement
X D E ED GRAPE we discretize the control waveform, φ(t), and
= ij e ij
e Hmag (t) ⊗ 1 + 1 ⊗ Hmag (t)] kl e kle . numerically maximize the fidelity by gradient ascent.
i,j,k,l We choose here a piecewise constant parameterization
(9) (as in [41]) and write the control waveform as a vec-
By modulating the phase φ(t) one can generate any tar- tor c = {φ(tj )/π | j = 1, . . . , n} where t = j∆t and
get unitary gate. n = T /∆t. The waveform is thus a series of square rf-
pulses with constant amplitude and phase over the dura-
tion ∆t.
III. NUMERICAL METHODS The minimum number of elements in the control vec-
tor c is determined by the number of parameters needed
We consider encoding a k-dimensional qudit in the to specify the target isometry. A K-dimensional par-
d = 10 dimensional Hilbert space associated with 10 mag- tial isometry is defined by the K columns in a D × D-
netic sublevels of the nuclear spin of 87 Sr. To implement dimensional unitary matrix, as can be seen from Eq. (10).
gates based on optimal control for k < 10, we use tech- Hence, to find the number of free parameters for a
niques based on the structure of partial isometries. A K-dimensional isometry one can count the number of
partial isometry of dimension k ≤ d in a physical system parameters needed to specify K orthonormal vectors
of dimension d is defined as, uniquely in a D-dimensional vector space. This is given
by
k
X K
X
Vtar = |fi i hei | (10) nmin (K, D) = 2(D − j) − 1 + K − 1
i=1
j=1
  (13)
where {|ei i}, {|fi i} are two orthonormal bases for the K(K + 1)
=2 KD − +K −1
qudit. This gives the unitary of interest to us in the form 2
Utar = Vtar + V⊥ , (11) =2KD − K 2 − 1,
where in the first line, we subtracted one from the pa-
where V⊥ is acting on the orthogonal subspace of dimen- rameter count in since the overall phase of the isometry
sion d − k. To find the control waveform, one then opti- is neglected. Equation (13) recovers well-known limits.
mizes the objective function When K = 1 and D = d, nmin = 2d − 2, which is the
 2 number of free parameters needed to specify a pure state
†
FV [c, T ] = Tr Vtar V [c, T ] /k 2 , (12) in a d-dimensional Hilbert space. When K = D = d,
nmin = d2 − 1, which is the number of free parameters
the fidelity between the target isometry and the isometry needed to specify a special unitary map in d-dimensions.
generated using quantum control [65]. In the Lie-algebraic protocol for designing entangling
gates, the control Hamiltonian, as well as the target uni-
tary matrices, are symmetric under the exchange of qu-
A. Numerical results for Lie-algebraic approach dits. In this case, one can work in the symmetric sub-
space for two qudits. Using the hook length formula [66],
As discussed in Sec. (II c), one can implement an ar- the dimension of the symmetric subspace of the total vec-
bitrary entangling gate through a combination of Ryd- tor space and isometry is,
berg dressing and phase-modulated Larmor precession d(d + 1) k(k + 1)
driven by rf-fields. Because our control Hamiltonian D= ,K = . (14)
2 2
 6

k nmin (K, D) can parameterize these according to

2 320  2

dX −1
3 623 (j)
α(j) ) = exp−i
Ui (~ αi Λi , (16)
5 1424
i=1
7 2295
where Λ is the generalized Gell-Mann matrices that span
TABLE I. The minimum number of parameters required for the Lie algebra su(d). The elements can be categorized
encoding a partial isometry of dimension k in the d = 10
as,
dimensional Hilbert space according to Eq. (13) for the prime
dimensions k ≤ 10 with K and D given by Eq. (14) symmetric: Λxjk = |jihk| + |kihj| ,
anti-symmetric: Λyjk = −i |jihk| + i |kihj| ,
l (17)
Thus, using Eq. (13), we find the number of free param- X
eters required for the two-qubit entangling unitary given diagonal: Λzl = |jihj| − l |l + 1ihl + 1| .
in Table. (I). j=1

Proof-of-principle numerical examples of waveforms

The task of the numerical optimization, thus, is to find
that generate the CPhase gate are given in Fig. (3). The
the set of times of the entangling interaction {tj }, and the
figure gives the φ(t) as a piecewise constant function of (j)
time, obtained using the GRAPE algorithm. We con- expansion coefficients of the Gell-Mann matrices {αi }
(j)
sider prime-dimensional qudits, the cases of most inter- and {βi }. We denote this whole set of parameter as
est in quantum algorithms. Figure. (3a) shows the case {λj } = {tj , α ~ (j) }.
~ (j) , β
of the k = 3, a qutrit encoded in d = 10. The total We define one layer of the control as consisting of a pair
time is T = 50π/Ωrf , which is divided into 700 intervals of local SU(d) gates followed by the entangling Hamilto-
for the quantum control. Figure (3b) shows an exam- nian for a time tj . The total number of free parame-
ple waveform for the case of k = 5. Here, the time is ters for a CPhase gate is d2 (d2 + 1)/2, as follows from
T = 240π/Ωrf and we divided the time into 1600 inter- Eq. (14) for a symmetric gate in SU(d2 ). Thus, the mini-
vals. Similarly, Fig. (3c) shows the case of k = 7 in our mum number of layers one requires to obtain the CPhase
d = 10 level system. The total time is T = 400π/Ωrf , di- gate is given as
vided into 2500 time intervals. This controllable Hamil-
tonian can also be used to generate other two-qudit gates.
d2 (d2 + 1)
Nmin 2(d2 − 1) + 1 =
The qudit generalization of the Mølmer-Sørensen gate, as 2
is given in the Appendix (C). (18)
d2 (d2 + 1)
The waveforms found here are a proof-of-principle set Nmin = .
2(2d2 − 1)
of square pulses and are not intended to be taken as the
best choice for experimental implementation. In prac- In practice, we find that one needs more than this min-
tice, one can design and optimize for much smoother imum to implement the target unitary gate with high
waveforms using well-known techniques by imposing ad- fidelity as this improves the optimization landscape for
ditional constraints on bandwidth and slew rate. Alter- gradient ascent [68]. The numerical results for the mini-
natively, one can optimize in the Fourier domain or in any mum number of layers needed in the system are given in
other complete basis of functions using the techniques of Table (II) for the cases of d = 3, 5, and d = 7.
gradient optimization of analytic controls (GOAT) [67]. For our case of study, we chose the same entangling
Hamiltonian as we used in the Lie-algebraic approach,
given in Eq. (5). However, unlike that approach, we
interleave the entangling interaction with local single-
B. Numerical results for Lie-group approach qudit SU(d) gates. Implementation of this requires an-
other layer of optimization. As we do not have access
In the Lie group control protocol discussed in Sec.IIc to native Hamiltonians proportional to the Gell-Mann
we parameterize the target unitary map as matrices, to implement local qudit gates we can employ
local SU(d) optimal control, as discussed in [42]. From
Y
Utar = Uλj , a practical perspective, this might be implemented di-
j rectly in the 3 P2 manifold, either through a combination
Y (15) of tensor-light shift and rf-driven Larmor precession sim-
= α(j) ) ⊗ U2 (β
e−iHent tj U1 (~ ~ (j) ).
ilar to [42], or alternatively through a combination of
j microwave-driven Rabi oscillations between different hy-
perfine levels in 3 P2 and rf-driven Larmor procession as
The control parameters {λi } consist of the set of times in [41]. In either case, optimal control can be used to
~ (j) and β~ (j)
{ti } as well as the 2(d2 − 1) parameters α find the relevant experimental waveform that generates
that specify each of the local SU(d) unitary maps. We the desired local SU(d) gates.
 7

FIG. 3. Waveforms of the CPhase gate. Quantum control is achieved by modulating the phase of an rf-field as a function
of time, φ(t). We parameterize this by a piecewise constant waveform. The figure shows proof-of-principle examples of φ(t)
that generate the CPhase gate, optimized using the GRAPE algorithm for different qudit dimensions. (a) The case of the d = 3
for a total time of T = 50π/Ωrf with 700 piecewise constant steps. (b) The case of the d = 5 for a total time of T = 240π/Ωrf
with 1600 piecewise constant steps. (c) The case of the d = 7 for a total time of T = 400π/Ωrf with 2500 piecewise constant
steps. For all of these calculations, the rf-field is on resonance with the Zeeman splitting ωrf = ω0 and we choose the rf-Larmor
frequency Ωrf = ωrf . Control is achieved by Rydberg dressing with laser Rabi frequency ΩL = 6Ωrf .

d Nmin Nlocal Nglobal employ symmetric global control of the local unitaries,
3 3 6 7 ~ (j) = β~ (j) , but to reverse the sign of the entangling
α
5 7 10 12 Hamiltonian Hent → −Hent in alternating layers. This
7 13 14 15 allows for effective optimization, and the corresponding
result is given in Table (II).
TABLE II. The number of layers of primitive gates in the Lie-
group approach required to achieve the CPhase gate. The
theoretical minimum is Nmin according to Eq. (18). If we
allow locally addressable single qudit gates, the number of C. Decoherence
layers required is Nlocal . If we have only global control, but
allow for a sign change in the entangling Hamiltonian, the In a closed quantum system, quantum optimal control
number of layers required is Nglobal employing either the Lie-algebraic or the Lie-group ap-
proaches can be used in principle to implement any qudit
entangling gate to any desired fidelity. In our numerical
In this analysis, we included locally addressable control optimization, we took the target infidelity to be 10−3 . In
on each qudit. Though the CPhase gate is symmetric un- the absence of decoherence, we could achieve that tar-
der exchange, we find that this symmetry breaking is nec- get in a reasonable time for d ≤ 5. For d = 7, more
essary for effective optimization of this parameterization, time is required. However, fundamentally the fidelity one
similar to that seen in [69]. An alternative protocol is to can achieve is limited by decoherence associated with the
 8

particular physical platform. For the system at hand,

decoherence occurs due to the finite lifetime of the Ry- 100
dberg states, which predominantly leads to leakage and
loss outside the computational basis. In that case, we
can model the gate as generated by a non-Hermitian ef-
fective Hamiltonian, Heff [c(t)], where the Hermitian part 10-1
is the control Hamiltonian and the anti-Hermitian part
governs the decay out of the Rydberg states. The fidelity
of interest is given by
 2 10-2
†
FV [c, T ] = Tr Vtar Veff [c, T ] /d2 , (19)
h  R i
T
where Veff [c, T ] = T exp −i 0 Heff [c(t)]dt . Here
ij 10-3
the decay amplitude from a dressed state is γdecay =
2 2 10 20 30 40 50 60
|Cri j | Γri + |Cirj | Γrj , which in turn gives the effective
Hamiltonian as
X  ij ij
 ED
eff
Hent = ELS2 − iγdecay e
/2 ij e .
ij (20) FIG. 4. Infidelity as a function of time. Simulated in-
ij fidelity with and without decoherence as a function of con-
trol time divided by the dimension d for CPhase gate with
With this model for decoherence in hand, the numer- different prime dimensions with d ≤ 10, as found using Lie-
ical results for the Lie-algebraic approach are given in algebraic quantum control and the GRAPE algorithm. De-
Fig. (4), which shows the infidelity as a function of time coherence due to Rydberg decay outside the computational
for a CPhase gate for different dimension isometries. We basis is included through an imaginary part of the Hamilto-
focus here on the case of the prime dimensional qudits. In nian. We take the Rydberg lifetime to be 140µs and choose
contrast to closed-system control, in the presence of de- the rf-Larmor frequency to be Ωrf /2π = 10 MHz. In the ab-
coherence, infidelity decreases at first and then increases. sence of decoherence (dashed lines), for a time greater than
the “quantum speed limit” (the time required to obtain ideal
This is due to the fact there is an optimal time of evo-
fidelity) we achieve a minimal error (infidelity) of 10−3 due to
lution, larger than the quantum speed limit, but not too our threshold in the numerics for d ≤ 5. This speed-limit time
large when compared to the coherence time of the sys- increases as we increase the qudit dimension, which in turn
tem. As expected, one needs more time as the qudit results in an increased decay in maximum fidelity. For the
dimension increases, which in turn results in an increase CPhase gate, we obtain a fidelity of 0.9985, 0.9980, 0.9942,
in the minimum infidelity one could achieve in each of and 0.9800 for d = 2, d = 3, d = 5, and d = 7 respectively.
these cases as shown in Fig. (4). We obtain a maximum For all of these calculations, we have taken the dressing laser
fidelity of 0.9985, 0.9980, 0.9942, and 0.9800 for d = 2, Rabi frequency to be ΩL = 6Ωrf and the lifetime of the Ryd-
d = 3, d = 5, and d = 7 respectively for the CPhase gate. berg states to be 140µs.
Note, the values of fidelity for different dimensional qu-
dits should be considered in the context of a particular
application. For example, the threshold for fault toler- The comparison of the fidelities achieved for the Lie-
ance for qudits, in general, is larger for larger d [70, 71]. algebraic and Lie-group approaches are given in Fig. (5)
For the particular scheme considered in [70], the thresh- for d = 3, 5, and d = 7. The results suggest that the
old for d = 2, d = 3, d = 5, and d = 7 are close to Lie-algebraic approach slightly outperforms the Lie group
0.008, 0.012, 0.0135, and 0.015 respectively. Hence, the approach in the presence of decoherence. However both
proof-of-principle fidelity obtained here is promising and approaches yield high fidelities in large dimensional qu-
can be further optimized. dits, and the Lie group approach may be preferable in
In the Lie-group approach, we can use the effective terms of the complexity necessary for experimental con-
Hamiltonian to describe the evolution when the Rydberg trol.
dressing is employed. In this case, we have,
Y
eff IV. CONCLUSION AND OUTLOOK
Utar = Uλj ,
j
Y (21)
=
eff
~ (j) ).
α(j) ) ⊗ U2 (β
e−iHent tj U1 (~ Quantum computation with qudits has potential ad-
j
vantages when compared with architectures employing
qubits. Implementing gates for qudit-based quantum
We neglect here any decoherence associated with the lo- computation is fundamentally more difficult, as the gen-
cal SU(d) gates. Thus the fidelity including the decoher- erators for these gates are not available on physical plat-
ence effects is given as, forms. One way to overcome this challenge is to use the
 2 tools of quantum optimal control, whereby we employ na-
† eff
Feff = Tr Utar Utar /d2 , (22) tive Hamiltonians and design time-dependent waveforms
 9

Finally, while we have studied a particular case study

1 in the context of neutral-atom quantum computing, the
general methods we have developed here can be applied
in other platforms, including trap ions and transmon qu-
0.99 dits, which also have natural encoding and control Hamil-
tonians.

0.98
ACKNOWLEDGMENTS

0.97 This work was supported by the Laboratory Directed

Research and Development program of Los Alamos Na-
tional Laboratory under project numbers 20200015ER,
and the NSF Quantum Leap Challenge Institutes pro-
0.96
10 20 30 40 gram, Award No. 2016244. The authors acknowledge
fruitful discussions with Sri Datta Vikas Buchemmavari,
Milad Marvian, Pablo Poggi, Jonathan Gross, Irfan Sid-
FIG. 5. A comparison of the optimized fidelity, F of the
diqi, and Noah Goss during various stages of this work.
CPhase gate achieved for the Lie-algebraic and Lie-group ap-
proaches (including both local single-qudit control and only
global control) is plotted as a function of the total Hilbert
space dimension d2 , for the qudits of dimension d = 3, 5, and
d = 7. For all of these simulations, we have taken the param-
eters given in Fig. 4.

that drive the system in order to implement a universal

gate set with high fidelity.
In this work, we introduced two classes of numerical
methods of quantum optimal control for implementing
the qudit entangling gates, an essential component of
the universal gate set. The first approach is based on
continuous-time driving given a controllable Hamiltonian
with tunable parameters and uses the Lie-algebraic struc-
ture of the control problem. The second approach is more
“digital,” using the Lie-group structure to design a fam-
ily of unitary maps that can be applied in sequence to
achieve any nontrivial entangling gate of interest.
As a specific example, we studied encoding a qudit in
the nuclear spin of 87 Sr, a species of atoms that is par-
ticularly important in quantum information processing.
The nuclear spin can accommodate a qudit of dimension
d ≤ 10. We have previously studied protocols for imple-
menting single-qudit gates in SU(d). To implement en-
tangling gates we studied how we make two atoms inter-
act using the well-known Rydberg blockade mechanism,
and in particular, we studied Rydberg dressing schemes.
Using this we are able to generate any two-qudit entan-
gling gate, both using the Lie-algebraic and Lie-group
based approaches.
We also studied how the fundamental effects of de-
coherence introduced by the finite lifetime of the Ryd-
berg states reduce the gate fidelity. To model this we
used a non-Hermitian Hamiltonian and found that even
when including decoherence, one could achieve very high
fidelity for these qudit entanglers. The flexibility of ar-
bitrary control provides avenues to explore the best ap-
proach to encoding qudits and mitigating errors.
 10

1.8

1.6

1.4

1.2

1
-4 -2 0 2 4

FIG. 6. Relative Rabi frequency, Ωri /ΩL , plotted as a func-

FIG. 7. Autler-Townes splitting of the three dressed states
tion of mF for π polarized light for the (5s5p)3 P2 F = 9/2 →
as a function of detuning for the Hamiltonian in Eq. (A2),
(5sns)3 S1 F ′ = 11/2 transition to the Rydberg state. The
where i = 0, j = 1, such that |0i ≡ 3 P2 , mF = 9/2 and
quadratic function arises due to the tensor polarizability. p p
|1i ≡ 3 P2 , mF = 7/2 . Here α = 7/16 and β = 9/16.
The dashed line shows the AC Stark shift (light shift) in the
Appendix A: Hyperfine structure of Rydberg states absence of a perfect Rydberg blockade. The blue curve adia-
and Clebsch-Gordan coefficients batically connects to the clock states for large blue detuning
and the red curve for large red detuning. The black curve is
a dressed superposition that does not adiabatically connect
As described in the Sec. (II c), to create entanglement to the clock states. The dashed lines show the light shifts in
we promote the population from the ground state 1 S0 to the absence of van der Waals interactions between the atoms.
the excited 3 P2 clock state, with the hyperfine quantum The difference between the solid line and the dashed line is the
number F = 9/2, and then consider a UV laser to excite entangling power of the Hamiltonian H212 defined in Eq.(A2).
the atoms to the 3 S1 Rydberg series to implement the
interaction between atoms with adiabatic dressing (see
Fig. (2). The Rabi frequency characterizing the coupling on the three dressed states after diagonalizing this Hamil-
of the different mF levels in the 3 P2 hyperfine manifold tonian. The dressed ground state is shown in red; the
to the 3 S1 Rydberg states will be different due to the other two dressed states represent Autler-Townes split-
Clebsch-Gordon Coefficients for these transitions. Let ΩL
ting. In the absence of the van der Waals interaction the
be the Rabi frequency on the |0a i → |0r i (mF = −9/2
transition). The Rabi frequency experienced by the other AC Stark shift (light shift) is the sum of the light shifts
levels is then of each atom independently (dashed line in Fig. (7). The
difference between these is the entangling energy.
hF, mF = −9/2 + i|1, 0; F ′ , mF = −9/2 + ii
Ωri = ΩL , One can understand the entangling power of the
hF, mF = −9/2|1, 0; F ′ , mF = −9/2i Hamiltonian by studying the properties of the dressed
(A1)
energy levels as a function of detuning. Figure (7) shows
where we have chosen F = 9/2 and F ′ = 11/2, and
the particular case of i = 0,j = 1 for the Hamiltonian in
a π-polarized light. In Fig. (6) the Rabi frequencies of
Eq. (A2), where |0i ≡ |mF = 9/2i and |1i ≡ |mF = 7/2i.
the different levels are given as a function of mF , whose
On the red side of detuning and for large detuning, as
parabolic shape describes the tensor light shift, thus giv-
we start with the bare state and we adiabatically sweep
ing a natural nonlinearity which arises solely due to well-
through resonance, the state maps to the superposition
defined hyperfine structure of 87 Sr.
of the two Rydberg states. Note, this is not an equal su-
Consider the Rydberg dressing scheme in Fig. (2). In
perposition as seen in [60] due to the fact that the states
the perfect blockade regime, the two-atom Hamiltonian
|0i and |1i couple with different Rydberg Rabi frequency
coupling of two magnetic sublevels labeled i and j is de-
and detuning to the Rydberg states.
scribed by a three-level system, governed by the Hamil-
tonian,
Ωr i Appendix B: Controllability
H2ij = − ∆i |ri ji hri j| + (|ri ji hij| + |iji hri j|)
2
Ωr j The quantum system is said to be controllable if,
− ∆j |irj i hirj | + (|irj i hij| + |iji hirj |) , given a time-dependent Hamiltonian H[c(t)], there ex-
2
(A2) ist a time-dependent set of waveforms c(t), such that
where ∆i determines the detunings due to the differential the one can generate an arbitrary unitary map. Here
Zeeman shit. Fig. (7) shows the resulting AC Stark shifts we consider those two-qudit unitary maps generated by
 11

an entangling Hamiltonian that is symmetric under the where the total angular momentum operator for the two
exchange of the qudits and thus does not require local qudits is
addressing. To show that a Hamiltonian is controllable,
we use the operator basis of irreducible spherical tensors Jz = 1 ⊗ jz + jz ⊗ 1. (C2)
on spin j defined as [72, 73],
We employ the same procedure for optimal control as
s we discussed in the main text in designing the waveforms
2k + 1 X to implement the CPhase gate. Numerical examples of
Tq(k) = hj, k + q|k, 1; j, mi |j, m + qihj, m| .
2j + 1 q the waveforms that create the Mølmer-Sørenson gate for
θ = π/2 are given in Fig. 9. The figure shows φ(t), the
(B1)
These satisfy the fundamental commutation relations, piecewise constant of the control waveform, obtained us-
ing the GRAPE algorithm. Fig. 3(a) shows the case of
h i
2
jz , Tq(k) =qTq(k) ,
h i p (B2)
(k) 1
j± , Tq(k) = k(k + 1) − q(q ± 1)Tq±1 .
0
(k)
The set of operators form a complete orthonormal
Tq -1
operator basis. Merkel et al. [74] showed that given a
generating set of Hamiltonians {hi }, if -2

 -3
Tr hi , Tqk 6= 0 (B3)
-4
for k > 2, the system is fully controllable. That is, the
set generates the whole Lie algebra of interest, which thus -5
0 2000 4000 6000 8000 10000
allows us to implement an arbitrary unitary map on the
spin of the system using quantum control.
We consider two-qudit systems, where the relevant Lie FIG. 8. The decomposition the entangling Hamiltonian Hent ,
Group is SU(d2 ); here d2 = 100. We expand the en- (K)
Eq. (5) in different orders of spherical tensors, Tq , for
tangling Hamiltonian in the operator basis of spherical j = 99/2, an operator basis of dimension D = 2j + 1 = 100,
tensors with j = 99/2, spanning the space of dimension spanning the two-qudit space for d = 10. The expansion co-
D = 2j + 1 = 100. Fig. 8 shows operator decomposition (K) (K)†
2
efficients are given by Cq = Tr(Hent Tq ) . We have
of the entangling Hamiltonian Hent in different orders of
spherical tensors. One can see in this figure that there ordered the expansion coefficients according to g(K, q) =
(k + 1)2 − 1 + q, where 0 ≤ k ≤ j, and −k ≤ q ≤ k. The
are contributions from higher rank tensors, making the
existence of contributions from higher-rank tensors makes the
system controllable. system controllable when combined with time-dependent rf-
fields that act locally on the atoms.

Appendix C: Creating other symmetric qudit the k = 3 the qutrit encoded in d = 10. The total time
entanglers for the Lie-algebraic approach is T = 50π/Ωrf and we divide the time into 700 time
steps for the quantum control. In Fig. 3(b) we plot an
Since the Hamiltonian described in Eq.(3) can be used example waveform for the case of the d = 5 into our 10
to create any symmetric two-qudit Hamiltonian, we can level system. We have a total time of T = 240π/Ωrf and
also generate the Mølmer-Sørenson gate for qudits de- we divide the time into 1600 time steps for the quantum
fined as, control. In Fig. 3(c) we plot an example for the case of
the d = 7 into our 10 level system. We have a total time
  of T = 400π/Ωrf and we divide the time into 2500 time
Jz2
UMS (θ) = exp −iθ . (C1) steps for the quantum control.
2

[1] Yuchen Wang, Zixuan Hu, Barry C Sanders, and Sabre J. M. Kreikebaum, D. Dahlen, A. Morvan, B. Yoshida,
Kais, “Qudits and high-dimensional quantum comput- N. Y. Yao, and I. Siddiqi, “Quantum information
ing,” Frontiers in Physics 8, 589504 (2020). scrambling on a superconducting qutrit processor,”
[2] M. S. Blok, V. V. Ramasesh, T. Schuster, K. O’Brien, Phys. Rev. X 11, 021010 (2021).
 12

FIG. 9. The figure gives the φ(t) that generates the Mølmer-Sørenson gate as a function of time for θ = π/2 using the piecewise
constant quantum control approach for the Hamiltonian given in Eq.(5). In (a) the case of the d = 3 for a total time of
Ωrf T = 50π with 700 piecewise constant steps. In (b) the case of the d = 5for a total time of Ωrf T = 240π with 1600 piecewise
constant steps. And in (c) the case of the d = 7for a total time of Ωrf T = 240π with 2500 piecewise constant steps. For all of
these calculations we have taken ΩL = 6Ωrf .

[3] Jonathan A. Gross, “Designing codes versal quantum computation with qudits,”
around interactions: The case of a spin,” Science China Physics, Mechanics & Astronomy 57, 1712–1717 (2014
Phys. Rev. Lett. 127, 010504 (2021). [9] Bin Li, Zu-Huan Yu, and Shao-Ming Fei, “Ge-
[4] Shruti Puri, Lucas St-Jean, Jonathan A. Gross, ometry of quantum computation with qutrits,”
Alexander Grimm, Nicholas E. Frattini, Pavithran S. Scientific reports 3, 2594 (2013).
Iyer, Anirudh Krishna, Steven Touzard, Liang [10] Hsuan-Hao Lu, Zixuan Hu, Mohammed Saleh Alshaykh,
Jiang, Alexandre Blais, Steven T. Flammia, and Alexandria Jeanine Moore, Yuchen Wang, Poolad Imany,
S. M. Girvin, “Bias-preserving gates with stabi- Andrew Marc Weiner, and Sabre Kais, “Quantum
lized cat qubits,” Science Advances 6, eaay5901 (2020), phase estimation with time-frequency qudits in a single
https://www.science.org/doi/pdf/10.1126/sciadv.aay5901. photon,” Advanced Quantum Technologies 3, 1900074,
[5] Daniel Gottesman, Alexei Kitaev, and John https://onlinelibrary.wiley.com/doi/pdf/10.1002/qute.201900074.
Preskill, “Encoding a qubit in an oscillator,” [11] Shawn X Cui and Zhenghan Wang, “Universal
Phys. Rev. A 64, 012310 (2001). quantum computation with metaplectic anyons,”
[6] Mikio Fujiwara, Masahiro Takeoka, Jun Mizuno, Journal of Mathematical Physics 56, 032202 (2015).
and Masahide Sasaki, “Exceeding the classical [12] Shawn X Cui, Seung-Moon Hong, and
capacity limit in a quantum optical channel,” Zhenghan Wang, “Universal quantum com-
Phys. Rev. Lett. 90, 167906 (2003). putation with weakly integral anyons,”
[7] Ming-Xing Luo, Xiu-Bo Chen, Yi-Xian Yang, and Xi- Quantum Information Processing 14, 2687–2727 (2015).
aojun Wang, “Geometry of quantum computation with [13] Alex Bocharov, Shawn X. Cui, Martin Roetteler, and
qudits,” Scientific reports 4, 1–5 (2014). Krysta M. Svore, “Improved quantum ternary arith-
[8] MingXing Luo and XiaoJun Wang, “Uni- metics,” (2015), arXiv preprint arXiv:1512.03824.
 13

[14] Jean-Luc Brylinski and Ranee Brylinski, “Universal [31] Quantum Control of Molecular Processes (John Wi-
quantum gates,” Mathematics of quantum computation ley & Sons, Ltd, 2011) Chap. 17, pp. 463–490,
79 (2002). https://onlinelibrary.wiley.com/doi/pdf/10.1002/9783527639700.ch17
[15] VE Zobov and AS Ermilov, “Implementation of a quan- [32] Christiane P Koch, Ugo Boscain, Tommaso Calarco,
tum adiabatic algorithm for factorization on two qudits,” Gunther Dirr, Stefan Filipp, Steffen J Glaser, Ron-
Journal of Experimental and Theoretical Physics 114, 923–932 (2012).
nie Kosloff, Simone Montangero, Thomas Schulte-
[16] Earl T. Campbell, “Enhanced fault-tolerant Herbrüggen, Dominique Sugny, et al., “Quantum opti-
quantum computing in d-level systems,” mal control in quantum technologies. strategic report on
Phys. Rev. Lett. 113, 230501 (2014). current status, visions and goals for research in europe,”
[17] Wim van Dam and Mark Howard, “Noise thresholds EPJ Quantum Technology 9, 19 (2022).
for higher-dimensional systems using the discrete wigner [33] Uffe V. Poulsen, Shlomo Sklarz, David Tannor, and
function,” Phys. Rev. A 83, 032310 (2011). Tommaso Calarco, “Correcting errors in a quan-
[18] Daniel Gottesman, “Fault-tolerant quantum com- tum gate with pushed ions via optimal control,”
putation with higher-dimensional systems,” in Phys. Rev. A 82, 012339 (2010).
Quantum Computing and Quantum Communications, [34] Philipp Treutlein, Theodor W. Hänsch, Jakob Re-
edited by Colin P. Williams (Springer Berlin Heidelberg, ichel, Antonio Negretti, Markus A. Cirone, and Tom-
Berlin, Heidelberg, 1999) pp. 302–313. maso Calarco, “Microwave potentials and optimal con-
[19] Earl T. Campbell, Hussain Anwar, and Dan E. trol for robust quantum gates on an atom chip,”
Browne, “Magic-state distillation in all prime Phys. Rev. A 74, 022312 (2006).
dimensions using quantum reed-muller codes,” [35] Michael H. Goerz, Eli J. Halperin, Jon M. Aytac, Chris-
Phys. Rev. X 2, 041021 (2012). tiane P. Koch, and K. Birgitta Whaley, “Robustness of
[20] Eliot Kapit, “Hardware-efficient and fully autonomous high-fidelity rydberg gates with single-site addressabil-
quantum error correction in superconducting circuits,” ity,” Phys. Rev. A 90, 032329 (2014).
Phys. Rev. Lett. 116, 150501 (2016). [36] Nathan K. Lysne, Kevin W. Kuper, Pablo M. Poggi,
[21] David P. DiVincenzo, “Two-bit gates Ivan H. Deutsch, and Poul S. Jessen, “Small, highly ac-
are universal for quantum computation,” curate quantum processor for intermediate-depth quan-
Phys. Rev. A 51, 1015–1022 (1995). tum simulations,” Phys. Rev. Lett. 124, 230501 (2020).
[22] Ashok Muthukrishnan and C. R. Stroud, “Mul- [37] P. Rebentrost, I. Serban, T. Schulte-Herbrüggen,
tivalued logic gates for quantum computation,” and F. K. Wilhelm, “Optimal control of a
Phys. Rev. A 62, 052309 (2000). qubit coupled to a non-markovian environment,”
[23] D. L. Zhou, B. Zeng, Z. Xu, and C. P. Sun, Phys. Rev. Lett. 102, 090401 (2009).
“Quantum computation based on d-level cluster state,” [38] Michael H Goerz, Daniel M Reich, and Chris-
Phys. Rev. A 68, 062303 (2003). tiane P Koch, “Optimal control theory for a
[24] Gavin K. Brennen, Dianne P. O’Leary, and unitary operation under dissipative evolution,”
Stephen S. Bullock, “Criteria for exact qudit universal- New Journal of Physics 16, 055012 (2014).
ity,” Phys. Rev. A 71, 052318 (2005). [39] Gerald Waldherr, Y Wang, S Zaiser, M Jamali,
[25] Noah Goss, Alexis Morvan, Brian Marinelli, Bradley K T Schulte-Herbrüggen, H Abe, T Ohshima, J Isoya,
Mitchell, Long B Nguyen, Ravi K Nail, Larry JF Du, P Neumann, et al., “Quantum error
Chen, Christian Jünger, John Mark Kreikebaum, correction in a solid-state hybrid spin register,”
David I Santiago, et al., “High-fidelity qutrit Nature 506, 204–207 (2014).
entangling gates for superconducting circuits,” [40] Jochen Scheuer, Xi Kong, Ressa S Said, Jeson
Chen, Andrea Kurz, Luca Marseglia, Jiangfeng Du,
arXiv preprint arXiv:2206.07216 (2022), 10.48550/arXiv.2206.07216.
[26] Laurin E Fischer, Alessandro Chiesa, Francesco Philip R Hemmer, Simone Montangero, Tommaso
Tacchino, Daniel J Egger, Stefano Carretta, and Calarco, Boris Naydenov, and Fedor Jelezko, “Precise
Ivano Tavernelli, “Towards universal gate syn- qubit control beyond the rotating wave approximation,”
thesis and error correction in transmon qudits,” New Journal of Physics 16, 093022 (2014).
arXiv preprint arXiv:2212.04496 (2022), 10.48550/arXiv.2212.04496.
[41] Brian Eric Anderson, “Unitary transformations in a large
[27] Martin Ringbauer, Michael Meth, Lukas Postler, Roman hilbert space,” (2013).
Stricker, Rainer Blatt, Philipp Schindler, and Thomas [42] Sivaprasad Omanakuttan, Anupam Mitra, Michael J.
Monz, “A universal qudit quantum processor with Martin, and Ivan H. Deutsch, “Quantum optimal
trapped ions,” Nature Physics 18, 1053–1057 (2022). control of ten-level nuclear spin qudits in 87 Sr,”
[28] Pavel Hrmo, Benjamin Wilhelm, Lukas Gerster, Mar- Phys. Rev. A 104, L060401 (2021).
tin W van Mourik, Marcus Huber, Rainer Blatt, Philipp [43] Katrina Barnes, Peter Battaglino, Benjamin J Bloom,
Schindler, Thomas Monz, and Martin Ringbauer, “Na- Kayleigh Cassella, Robin Coxe, Nicole Crisosto,
tive qudit entanglement in a trapped ion quantum pro- Jonathan P King, Stanimir S Kondov, Krish Kotru,
cessor,” arXiv preprint arXiv:2206.04104 (2022). Stuart C Larsen, et al., “Assembly and coher-
[29] L. M. K. Vandersypen and I. L. Chuang, “Nmr ent control of a register of nuclear spin qubits,”
techniques for quantum control and computation,” Nature Communications 13, 1–10 (2022).
Rev. Mod. Phys. 76, 1037–1069 (2005). [44] Andrew J. Daley, Martin M. Boyd, Jun Ye, and Peter
[30] Herschel Rabitz, Regina de Vivie-Riedle, Zoller, “Quantum computing with alkaline-earth-metal
Marcus Motzkus, and Karl Kompa, atoms,” Phys. Rev. Lett. 101, 170504 (2008).
“Whither the future of controlling quan- [45] Andrew J Daley, “Quantum computing and
tum phenomena?” Science 288, 824–828 (2000), quantum simulation with group-ii atoms,”
https://www.science.org/doi/pdf/10.1126/science.288.5467.824. Quantum Information Processing 10, 865–884 (2011).
 14

[46] Harry Levine, Alexander Keesling, Giulia Semeghini, squeezing algorithms on programmable quantum sen-
Ahmed Omran, Tout T. Wang, Sepehr Ebadi, Hannes sors,” Phys. Rev. Lett. 123, 260505 (2019).
Bernien, Markus Greiner, Vladan Vuletić, Hannes Pich- [60] Anupam Mitra, Michael J. Martin, Grant W. Bie-
ler, and Mikhail D. Lukin, “Parallel implementation dermann, Alberto M. Marino, Pablo M. Poggi, and
of high-fidelity multiqubit gates with neutral atoms,” Ivan H. Deutsch, “Robust mølmer-sørensen gate for
Phys. Rev. Lett. 123, 170503 (2019). neutral atoms using rapid adiabatic rydberg dressing,”
[47] Dolev Bluvstein, Harry Levine, Giulia Semeghini, Phys. Rev. A 101, 030301 (2020).
Tout T Wang, Sepehr Ebadi, Marcin Kalinowski, [61] Anupam Mitra, Sivaprasad Omanakuttan, Michael J
Alexander Keesling, Nishad Maskara, Hannes Pich- Martin, Grant W Biedermann, and Ivan H
ler, Markus Greiner, et al., “A quantum processor Deutsch, “Practical and fundamental limits of neu-
based on coherent transport of entangled atom arrays,” tral atom entanglement using rydberg dressing,”
Nature 604, 451–456 (2022). arXiv preprint arXiv:2205.12866 (2022).
[48] TM Graham, Y Song, J Scott, C Poole, L Phut- [62] Michael J Martin, Yuan-Yu Jau, Jongmin Lee, Anu-
titarn, K Jooya, P Eichler, X Jiang, A Marra, pam Mitra, Ivan H Deutsch, and Grant W Biedermann,
B Grinkemeyer, et al., “Multi-qubit entanglement and “A mølmer-sørensen gate with rydberg-dressed atoms,”
algorithms on a neutral-atom quantum computer,” arXiv e-prints , arXiv–2111 (2021).
Nature 604, 457–462 (2022). [63] Nathan Schine, Aaron W Young, William J Eckner,
[49] Navin Khaneja, Timo Reiss, Cindie Kehlet, Thomas Michael J Martin, and Adam M Kaufman, “Long-
Schulte-Herbrüggen, and Steffen J. Glaser, “Opti- lived bell states in an array of optical clock qubits,”
mal control of coupled spin dynamics: design of Nature Physics 18, 1067–1073 (2022).
nmr pulse sequences by gradient ascent algorithms,” [64] Jan Trautmann, Dimitry Yankelev, Valentin
Journal of Magnetic Resonance 172, 296–305 (2005). Klüsener, Annie Jihyun Park, Immanuel Bloch,
[50] Ashok Muthukrishnan and C. R. Stroud, “Mul- and Sebastian Blatt, “The 1 S0 − 3 P2 magnetic
tivalued logic gates for quantum computation,” quadrupole transition in neutral strontium,”
Phys. Rev. A 62, 052309 (2000). arXiv preprint arXiv:2211.02470 (2022).
[51] Alexander Yu Vlasov, “Noncommutative tori [65] Line Hjortshøj Pedersen, Niels Martin Møller, and
and universal sets of nonbinary quantum gates,” Klaus Mølmer, “Fidelity of quantum operations,”
Journal of Mathematical Physics 43, 2959–2964 (2002). Physics Letters A 367, 47–51 (2007).
[52] Ranee K Brylinski and Goong Chen, [66] J Sutherland Frame, G de B Robinson, and Robert M
Mathematics of quantum computation (CRC Press, Thrall, “The hook graphs of the symmetric group,”
2002). Canadian Journal of Mathematics 6, 316–324 (1954).
[53] J. E. Johnson and S. L. Rolston, “Inter- [67] Shai Machnes, Elie Assémat, David Tannor, and
actions between rydberg-dressed atoms,” Frank K. Wilhelm, “Tunable, flexible, and efficient
Phys. Rev. A 82, 033412 (2010). optimization of control pulses for practical qubits,”
[54] Tyler Keating, Robert L. Cook, Aaron M. Hankin, Yuan- Phys. Rev. Lett. 120, 150401 (2018).
Yu Jau, Grant W. Biedermann, and Ivan H. Deutsch, [68] Martı́n Larocca, Pablo Poggi, and Diego Wisniacki,
“Robust quantum logic in neutral atoms via adiabatic “Quantum optimal control: Landscape structure and
rydberg dressing,” Phys. Rev. A 91, 012337 (2015). topology,” arXiv preprint arXiv:1802.05683 (2018).
[55] Y-Y Jau, AM Hankin, T Keating, IH Deutsch, [69] Alexandre Choquette, Agustin Di Paolo, Panagio-
and GW Biedermann, “Entangling atomic spins tis Kl. Barkoutsos, David Sénéchal, Ivano Taver-
with a rydberg-dressed spin-flip blockade,” nelli, and Alexandre Blais, “Quantum-optimal-control-
Nature Physics 12, 71–74 (2016). inspired ansatz for variational quantum algorithms,”
[56] Johannes Zeiher, Rick Van Bijnen, Peter Schauß, Phys. Rev. Research 3, 023092 (2021).
Sebastian Hild, Jae-yoon Choi, Thomas Pohl, Im- [70] Hussain Anwar, Benjamin J Brown, Earl T Campbell,
manuel Bloch, and Christian Gross, “Many-body and Dan E Browne, “Fast decoders for qudit topological
interferometry of a rydberg-dressed spin lattice,” codes,” New Journal of Physics 16, 063038 (2014).
Nature Physics 12, 1095 (2016). [71] Fern H. E. Watson, Hussain Anwar, and Dan E. Browne,
[57] Johannes Zeiher, Jae-yoon Choi, Antonio Rubio- “Fast fault-tolerant decoder for qubit and qudit surface
Abadal, Thomas Pohl, Rick van Bijnen, Immanuel codes,” Phys. Rev. A 92, 032309 (2015).
Bloch, and Christian Gross, “Coherent many-body [72] J Sakurai and J Napolitano, “Modern quantum me-
spin dynamics in a long-range interacting ising chain,” chanics. 2-nd edition,” Person New International edition
Phys. Rev. X 7, 041063 (2017). (2014).
[58] V. Borish, O. Marković, J. A. Hines, S. V. Ra- [73] A. B. Klimov, R. Guzmán, J. C. Retamal, and C. Saave-
jagopal, and M. Schleier-Smith, “Transverse-field dra, “Qutrit quantum computer with trapped ions,”
ising dynamics in a rydberg-dressed atomic gas,” Phys. Rev. A 67, 062313 (2003).
Phys. Rev. Lett. 124, 063601 (2020). [74] Seth Merkel, “Quantum control of d-dimensional quan-
[59] Raphael Kaubruegger, Pietro Silvi, Christian Kokail, tum systems with application to alkali atomic spins,”
Rick van Bijnen, Ana Maria Rey, Jun Ye, Adam M. arXiv preprint arXiv:0906.4790 (2009), 10.48550/arXiv.0906.4790.
Kaufman, and Peter Zoller, “Variational spin-