Article

Distributed quantum computing across an

optical network link

https://doi.org/10.1038/s41586-024-08404-x D. Main1 ✉, P. Drmota1, D. P. Nadlinger1, E. M. Ainley1, A. Agrawal1, B. C. Nichol1, R. Srinivas1,

G. Araneda1 & D. M. Lucas1
Received: 30 June 2024

Accepted: 14 November 2024

Distributed quantum computing (DQC) combines the computing power of multiple
Published online: xx xx xxxx
networked quantum processing modules, ideally enabling the execution of large
Open access
quantum circuits without compromising performance or qubit connectivity1,2.
Check for updates Photonic networks are well suited as a versatile and reconfigurable interconnect layer
for DQC; remote entanglement shared between matter qubits across the network
enables all-to-all logical connectivity through quantum gate teleportation (QGT)3,4.
For a scalable DQC architecture, the QGT implementation must be deterministic and
repeatable; until now, no demonstration has satisfied these requirements. Here we
experimentally demonstrate the distribution of quantum computations between
two photonically interconnected trapped-ion modules. The modules, separated by
about two metres, each contain dedicated network and circuit qubits. By using
heralded remote entanglement between the network qubits, we deterministically
teleport a controlled-Z (CZ) gate between two circuit qubits in separate modules,
achieving 86% fidelity. We then execute Grover’s search algorithm5—to our knowledge,
the first implementation of a distributed quantum algorithm comprising several
non-local two-qubit gates—and measure a 71% success rate. Furthermore, we
implement distributed iSWAP and SWAP circuits, compiled with two and three
instances of QGT, respectively, demonstrating the ability to distribute arbitrary
two-qubit operations6. As photons can be interfaced with a variety of systems, the
versatile DQC architecture demonstrated here provides a viable pathway towards
large-scale quantum computing for a range of physical platforms.

The potential of quantum computing to revolutionize various fields teleportation (QGT) efficiently implements non-local entangling gates
ranging from cryptography to drug discovery is widely recognized7,8. between qubits in separate modules, consuming only one Bell pair and
However, regardless of the physical platform used to realize a quantum the exchange of two classical bits14,15, as shown in Fig. 1b. Given arbitrary
computer, scaling up the number of qubits while maintaining precise single-qubit and two-qubit operations within each node, QGT com-
control and interconnectivity is a substantial technical challenge9–11. pletes a universal gate set for the distributed quantum computer13. The
The distributed quantum computing (DQC) architecture, shown in primary advantage of teleportation-based schemes over direct transfer
Fig. 1, addresses this challenge by enabling large quantum computa- is the exclusive use of the quantum channel for generating identical
tions to be executed by a network of quantum processing modules1,2. Bell states; channel losses can be overcome by repetition without los-
The modules each host a relatively small number of qubits and are inter- ing quantum information, and the distance between modules can be
connected through both classical and quantum information channels. increased by inserting quantum repeaters16. Furthermore, channel
By preserving the reduced complexity of the individual modules and noise may be suppressed using entanglement purification17. Because
transforming the scaling challenge into the task of building more mod- teleportation protocols are executed strictly after entanglement has
ules and establishing an interface between them, the DQC architecture been established, they enable continuous deterministic operation
provides a scalable approach to fault-tolerant quantum computing3,4. even if the entanglement is generated non-deterministically. This
The interface between modules could be realized by directly transfer- deterministic nature is crucial for scalability, eliminating the need for
ring quantum information between modules. However, losses in the post-selection of singular successful outcomes out of an exponentially
interconnecting quantum channels would lead to the unrecoverable large set of undesired ones.
loss of quantum information. Quantum teleportation offers a lossless Teleportation protocols are agnostic to the physical implementation
alternative interface, using only bipartite entanglement (for example, of the quantum channels, making them a versatile tool for DQC across
Bell states) shared between modules, together with local operations different platforms. In the trapped-ion quantum charge-coupled device
and classical communication to effectively replace the direct transfer (QCCD) architecture, qubits can be dynamically transported between
of quantum information across quantum channels12,13. Quantum gate modules within a single chip18—or even across chips19—and thus be

Clarendon Laboratory, Department of Physics, University of Oxford, Oxford, UK. ✉e-mail: [email protected]
1

Nature | www.nature.com | 1
Article
a M1 b c
M2
Local operations M1

Classical
M3 communication
M2

Non-local
operations
M3

M4
MN–1

Shared
entanglement MN
MN
Network
qubit Circuit
qubit

Fig. 1 | DQC architecture. a, Schematic of a DQC architecture comprising operations. QGT mediates non-local gate interactions (pink) between circuit
photonically interconnected modules. Entanglement is heralded between qubits in separate modules. These protocols require the resources of shared
network qubits through the interference of photons on beam splitters. entanglement, local operations and classical communication. c, A quantum
A photonic switchboard provides a flexible and reconfigurable network circuit distributed across a network of small quantum processing modules that
topology. b, The modules consist of at least one network qubit (purple) and at function together as a single, fully connected quantum computer.
least one circuit qubit (orange), which may directly interact by means of local

used to mediate entangling gates between different trap zones20,21. before the qubits were transported. Furthermore, there have been
Photons, however, make natural carriers of quantum information, as demonstrations of heralded non-local entangling gates across a pho-
they can travel long distances without substantial degradation of their tonic quantum network in which photons are used to directly trans-
quantum state. Photonic interconnects enable all-to-all connectivity fer quantum information between modules35,36. However, in these
between qubits distributed across the network, whose topology can demonstrations, unavoidable photon loss destroyed the states of the
be dynamically reconfigured without the need to open up complex circuit qubits, rendering these schemes non-deterministic. Until now,
vacuum and/or cryogenic systems. Moreover, optical components are there has been (1) no demonstration of deterministic QGT across a
widely available and can be operated under ambient conditions. These quantum network and (2) no demonstration of distributed circuits
properties make photonic interconnects particularly appealing for comprising several non-local entangling gates. In photonic platforms,
networking quantum computing modules, as shown in Fig. 1a. As shown this has been prevented by the inability to store the photons between
in Fig. 1b, we consider modules containing ‘network’ and ‘circuit’ qubits interactions30,31, whereas in the QCCD demonstration, this was limi­
with full interconnectivity through local quantum operations. Remote ted by the decoherence of the circuit qubits during the generation of
entanglement of network qubits in separate modules is generated by entanglement20.
the interference of photons, in which reconfigurability and flexibility In this work, we present, to our knowledge, the first demonstration
could be provided by means of a photonic switchboard. This entangle- of DQC across a network of two trapped-ion modules, each containing
ment can then be used to mediate multiqubit gates between the circuit a network qubit and a circuit qubit, and separated by a macroscopic
qubits in different modules through QGT, enabling the network to func- distance (about 2 m). We mediate deterministic two-qubit CZ interac-
tion as a single, fully connected quantum processor, as shown in Fig. 1c. tions between the circuit qubits through QGT, using entanglement
Quantum circuits can be partitioned freely in this architecture, down previously established across the network between the two network
to a minimum of one circuit qubit per module in the fully distributed qubits. By making use of the robust storage of quantum informa-
case. Heralded entanglement between spatially separated qubits has tion in the circuit qubits while generating subsequent rounds of
been achieved experimentally in a variety of platforms, including dia- entanglement between network qubits37, we execute distributed
mond colour centres22,23, superconducting qubits24, neutral atoms25,26 quantum circuits comprising several non-local two-qubit gates. We
and trapped ions27–29. demonstrate the distributed iSWAP and SWAP gates, which consist of
QGT has been implemented probabilistically in purely photonic sys- two and three instances of QGT, respectively. The actions of all tele­
tems, requiring passive optical elements and post-selection to perform ported gates are characterized using quantum process tomography.
the conditional rotations that complete the gate teleportation30,31. Chou Finally, we implement Grover’s algorithm on our distributed quantum
et al.32 demonstrated deterministic teleportation of a controlled-NOT computer.
gate between two qubits encoded in the modes of two superconducting
cavities on the same device, separated by about 2 cm, whereas a third
cavity enabled the deterministic generation of entanglement between Teleportation of a CZ gate
two transmon network qubits. Recently, there have been demonstra- Our apparatus, shown in Fig. 2a, consists of two trapped-ion modules,
tions of QGT between superconducting qubits within a single device, Alice and Bob, each co-trapping one 88Sr+ ion and one 43Ca+ ion (Meth-
demonstrating the viability of QGT to overcome nearest-neighbour ods). The Ca+ ion provides a magnetic-field-insensitive ‘circuit’ qubit,
constraints in this architecture33,34. In the trapped-ion QCCD archi- Q C := {∣0C⟩ ≡ ∣F = 4, mF = 0⟩, ∣1C⟩ ≡ ∣F = 3, mF = 0⟩}, in the ground hyperfine
tecture, Wan et al.20 demonstrated QGT between ions in two zones manifold, which has been used to demonstrate state-of-the-art
of the same trap, separated by about 840 μm; the entanglement was quantum logic38,39. The Sr+ ion, on the other hand, provides an effi­
deterministically generated between qubits through local operations cient interface to the optical quantum network 28. We define the

2 | Nature | www.nature.com
a b ideal ideal
Re( FCZ ) 10 × Im( FCZ )
Alice 0.25
II II
IY IY
C UA
Ca+ XI XI
X
XY XY
π YI YI 0
Sr+ N
2 y
YY YY
ZI ZI
ZY ZY
–0.25
Quantum link

Classical link
II IZ XY YX ZI ZZ II IZ XY YX ZI ZZ
~2 m

c exp
Re( FCZ ) exp
10 × Im( FCZ )
II II 0.25
IY IY
XI XI
π
N
2 x
XY XY
YI YI 0
X YY YY
C UB ZI ZI
ZY ZY
Bob –0.25
II IZ XY YX ZI ZZ II IZ XY YX ZI ZZ

Fig. 2 | Teleportation of a CZ gate between two trapped-ion modules. the gate mechanism. The outcomes of mid-circuit parity measurements of
a, The two modules, Alice and Bob, each hold a 88Sr+ ion (purple) and a 43Ca+ ion the network qubits are exchanged in real time through a classical (TTL) link
(orange). Sr+ provides a network qubit, Q N, whereas Ca+ provides both a long- connecting the control systems of the two modules. This information is used
lived circuit qubit, Q C, and an auxiliary qubit, Q X. Before the protocol, the to condition local feed-forward operations, UA and UB, on the circuit qubits—
circuit qubits are in some arbitrary state. The protocol begins by generating completing the teleportation of the CZ gate. b, Process matrix for an ideal CZ
entanglement between the network qubits through a photonic link. On gate. c, Measured process matrix, reconstructed through quantum process
heralding entanglement, each module applies a local CZ gate between the tomography, yielding an average gate fidelity of 86.2(9)% compared with an
network and circuit qubits, using the auxiliary qubit temporarily to mediate ideal CZ gate.

1 3
network qubit in Sr+ byQ N := {|0N⟩ ≡ |S 1/2, mJ = − 2 ⟩, |1N⟩ ≡ |D5/2, mJ = − 2 ⟩}. teleportation protocol (Methods). This implements the non-local
To implement local entangling operations between these two gate ∣ψinAB⟩ → UCZ
AB
∣ψinAB⟩.
species, we use the light-shift gate mechanism 40 between Q N We characterize the QGT protocol using quantum process tomog-
exp
and an auxiliary qubit in the ground hyperfine manifold of Ca +, raphy (Methods) to reconstruct the process matrix, χCZ , providing a
Q X := {∣0X ⟩ ≡ ∣F = 4, mF = +4⟩, ∣1X⟩ ≡ ∣F = 3, mF = +3⟩}, which—unlike the complete description of the action of the teleported CZ gate on the
Q C qubit—experiences the necessary light shifts (Methods). At the two circuit qubits. Compared with the ideal CZ process, shown in Fig. 2b,
points at which we want to perform the local entangling gate, we the reconstructed process matrix for the teleported gate, shown in
transfer the quantum information stored in Q C temporarily to Q X Fig. 2c, has an average gate fidelity of 86.2(9)%. The QGT protocol is
to perform the gate operations (Methods). completely self-contained—the input states of the circuit qubits are
The QGT protocol used here to mediate CZ gates between the circuit set before the execution of the non-local gate—and output states are
qubits in separate modules is shown in Fig. 2. We allow the circuit qubits available for further computation. With single-qubit rotations of the
to start in an arbitrary state ∣ψinAB⟩ ∈ Q ⊗2
C , which could be part of a larger, circuit qubits, this teleported CZ gate is a key element of a gate set for
long-running computation. We begin the QGT protocol by generating DQC, enabling the modules to act as a single, fully connected universal
the remotely entangled Bell state quantum processor.

|10⟩ + |01⟩
|Ψ +⟩ = ∈ Q ⊗2
N ,
2 DQC
In general, any arbitrary two-qubit unitary operation can be decom-
between the network qubits28, with a fidelity of 96.89(8)% (Methods). posed into at most three CZ gates6. We demonstrate our ability to per-
This entanglement is generated through a try-until-success process, form sequential rounds of QGT by executing the CZ decompositions of
for which a herald indicates a success. In contrast to the network qubits, the iSWAP and SWAP gates, shown in (1) in Fig. 3a,b, comprising two and
the circuit qubits provide a robust quantum memory37, enabling stor- three instances of QGT, respectively. As with the teleported CZ gate,
age of the encoded quantum information until the remote entangle- we characterize these circuits through quantum process tomogra-
ment is successfully heralded. At this stage, we map the state stored in phy (Methods); see (2) in Fig. 3a,b. From the reconstructed process
the circuit qubits (Q C) to the auxiliary qubits (Q X) in preparation for matrices, we measure average gate fidelities of 70(2)% and 64(2)% for
the local entangling operations (Methods). In each module, we perform the iSWAP and SWAP gates, respectively. By constructing circuits with
local CZ gates between the network and auxiliary qubits (Methods), several instances of QGT—enabled by our ability to perform QGT deter-
before transferring the auxiliary qubit back to the circuit qubit. We ministically and on demand—we demonstrate the ability to perform
then perform mid-circuit measurements of the network qubits in the universal DQC.
X and Y bases in Alice and Bob, respectively. The modules exchange the Finally, we implement Grover’s algorithm5,41,42 on our distributed
measurement outcomes in real time—using a classical (TTL) link quantum processor. This algorithm searches through a set of unsorted
between their control systems—and perform single-qubit feed-forward items, x ∈ L, to find a particular item, a ∈ L. The search problem is rep-
operations conditioned on the exchanged bits to complete the gate resented by the function

Nature | www.nature.com | 3
Article
a iSWAP gate c Grover’s algorithm
(1) Alice S H H H (1)
Oracle Diffusion
Bob S H H H
Alice H H X X H
Ua
(2) exp
Re( iSWAP ) exp
Im( iSWAP )
0.25 Bob H H X X H
II II
IY IY
XI XI Ideal Data
(2)
XY XY
0 1.0 1.0
YI YI
a = 00 a = 01
YY YY 0.8 0.8
ZI ZI
ZY ZY

Probability
Probability
–0.25 0.6 0.6

II IZ XY YX ZI ZZ II IZ XY YX ZI ZZ
0.4 0.4

b SWAP gate 0.2 0.2

(1)
Alice H H
0 0
00 01 10 11 00 01 10 11
Bob H H H H Outcomes Outcomes

1.0 1.0
(2) exp exp
Re( SWAP ) Im( SWAP )
II II 0.25 0.8 0.8
a = 10 a = 11
IY IY
Probability

Probability
XI XI 0.6 0.6
XY XY
YI YI 0 0.4 0.4
YY YY
ZI ZI 0.2 0.2
ZY ZY
–0.25 0 0
II IZ XY YX ZI ZZ II IZ XY YX ZI ZZ 00 01 10 11 00 01 10 11
Outcomes Outcomes

Fig. 3 | DQC results. a,b, (1) CZ decompositions of the distributed iSWAP (a) and Grover oracle call, which marks a particular state, a, and the second implements
SWAP (b) circuits, comprising two and three instances of QGT, respectively. the diffusion circuit. (2) Measurement outcomes from 500 repetitions of
(2) The reconstructed process matrices for the iSWAP (a) and SWAP (b) gates Grover’s algorithm per marked state; the average success probability is 71(1)%.
indicate average gate fidelities of 70(2)% and 64(2)%, respectively. c, Grover’s All error bars indicate one standard deviation.
algorithm. (1) Circuit comprising two instances of QGT: the first implements the

1 if x = a ,
fa (x) =  Discussion
0 otherwise.
The performance of our distributed quantum circuits is consistent with
In the two-qubit case, there are four items to search through. Classi- the errors from the teleported CZ gates. We summarize the leading
cally, the item a could be identified with, on average, two queries of error sources affecting our teleported CZ gate in Table 1. The measured
the function fa(x). Using the quantum circuit shown in (1) in Fig. 3c, the fidelity of our gate is slightly lower than that predicted by the error
same task can be accomplished with only one query. After preparing budget, which we attribute to drifts in the calibration of various com-
a superposition of all possible inputs with parallel Hadamard gates, an ponents over the duration of the data acquisition. Most of the identified
instance of QGT implements the oracle, which performs the mapping errors occur during local operations in each module. Our local errors do
Ua : ∣x ⟩ → (−1) fa (x )∣x ⟩, marking the state |a⟩. A second instance of QGT not represent the state of the art for trapped-ion processors; however,
implements the Grover diffusion circuit, which decodes the quantum local operations exceeding the approximately 99% fidelity threshold
information provided by the oracle into an observable. In the two-qubit for fault-tolerant quantum computing have been demonstrated in this
case considered here, the application of the Grover diffusion circuit platform39,40,43–48. Relevant to our implementation, Hughes et al.40 dem-
should leave the register in the state |a⟩, which is the solution to the onstrated mixed-species two-qubit gates between 88Sr+ and 43Ca+ ions
function fa, and thus a measurement of the register yields the solution with a gate fidelity of 99.8(1)%. We therefore conclude that the technical
to the search problem with unit probability. In the case of N items, limitations in our implementation can be overcome. The other notable
to approach unit probability of obtaining the solution, we would source of error is the remote entanglement of the network qubits across
require O( N ) iterations of the oracle–diffusion circuit, compared the photonic quantum network; we observe a fidelity of the remotely
with O(N /2) for a classical search. entangled network qubits to the desired |Ψ+⟩ state of 96.89(8)%. Unlike
The results of Grover’s algorithm—executed on our distributed the local operations, the performance of our remote entanglement is
quantum processor—are shown in (2) in Fig. 3c. For the marked states at the state of the art. To improve this, and hence enable the teleporta-
a ∈ {00, 01, 10, 11}, we obtain the correct result with an average success tion of high-fidelity entangling gates between modules, entanglement
rate of 71(1)%. To our knowledge, this represents the first deterministic purification could be used to distribute high-fidelity entangled states
execution of any algorithm on a distributed quantum computer. from several lower-fidelity entangled states17,49.

4 | Nature | www.nature.com
Table 1 | Error budget for CZ gate teleportation 7. Shor, P. W. Polynomial-time algorithms for prime factorization and discrete logarithms on
a quantum computer. SIAM J. Comput. 26, 1484–1509 (1997).
8. Cao, Y., Romero, J. & Aspuru-Guzik, A. Potential of quantum computing for drug discovery.
Source Error IBM J. Res. Dev. 62, 6:1–6:20 (2018).
Alice Bob 9. Bruzewicz, C. D., Chiaverini, J., McConnell, R. & Sage, J. M. Trapped-ion quantum
computing: progress and challenges. Appl. Phys. Rev. 6, 021314 (2019).
Raw entanglement 3.11(8)% 10. Bravyi, S., Dial, O., Gambetta, J. M., Gil, D. & Nazario, Z. The future of quantum computing
Mixed-species gate 2.4(2)% 2.0(2)% with superconducting qubits. J. Appl. Phys. 132, 160902 (2022).
11. Gill, S. S. et al. Quantum computing: vision and challenges. Preprint at https://arxiv.org/
Q C decoherence 1.9(4)% 1.8(5)% abs/2403.02240 (2024).
12. Bennett, C. H. et al. Teleporting an unknown quantum state via dual classical and
Q X ↔ Q C transfer 0.76(3)% 0.52(1)%
Einstein–Podolsky–Rosen channels. Phys. Rev. Lett. 70, 1895 (1993).
Mid-circuit measurement 0.091(3)% 0.122(2)% 13. Gottesman, D. & Chuang, I. L. Demonstrating the viability of universal quantum
computation using teleportation and single-qubit operations. Nature 402, 390–393
Q C rotations 0.016(1)% 0.015(1)% (1999).
Predicted total error 12.1(6)% 14. Eisert, J., Jacobs, K., Papadopoulos, P. & Plenio, M. B. Optimal local implementation of
nonlocal quantum gates. Phys. Rev. A 62, 052317 (2000).
The characterization of each error contribution is discussed in Methods. 15. Collins, D., Linden, N. & Popescu, S. Nonlocal content of quantum operations. Phys. Rev.
A 64, 032302 (2001).
16. Briegel, H.-J., Dür, W., Cirac, J. I. & Zoller, P. Quantum repeaters: the role of imperfect local
Our implementation features a single circuit qubit in each module; operations in quantum communication. Phys. Rev. Lett. 81, 5932 (1998).
17. Dür, W. & Briegel, H.-J. Entanglement purification for quantum computation. Phys. Rev. Lett.
however processors with larger numbers of qubits have been realized. 90, 067901 (2003).
With three circuit qubits (and one network qubit) per module, the 18. Kielpinski, D., Monroe, C. & Wineland, D. J. Architecture for a large-scale ion-trap quantum
purification of arbitrary quantum channels would be possible49. The computer. Nature 417, 709–711 (2002).
19. Akhtar, M. et al. A high-fidelity quantum matter-link between ion-trap microchip modules.
capabilities of the individual modules may be extended even further Nat. Commun. 14, 531 (2023).
by deploying the QCCD architecture. With recent demonstrations in 20. Wan, Y. et al. Quantum gate teleportation between separated qubits in a trapped-ion
both academic research50 and industry21 highlighting the power of processor. Science 364, 875–878 (2019).
21. Pino, J. M. et al. Demonstration of the trapped-ion quantum CCD computer architecture.
this approach, embedding these systems in a quantum network would Nature 592, 209–213 (2021).
combine their power with the reconfigurability and flexibility of the 22. Humphreys, P. C. et al. Deterministic delivery of remote entanglement on a quantum
network. Nature 558, 268–273 (2018).
DQC architecture. Conversely, computational bottlenecks associated
23. Knaut, C. M. et al. Entanglement of nanophotonic quantum memory nodes in a telecom
with ion transport overheads observed in the QCCD architecture21 network. Nature 629, 573–578 (2024).
could be mitigated using photonic interconnects integrated into a 24. Storz, S. et al. Loophole-free Bell inequality violation with superconducting circuits.
Nature 617, 265–270 (2023).
single device51.
25. Ritter, S. et al. An elementary quantum network of single atoms in optical cavities. Nature
Although the results presented here were achieved using trapped-ion 484, 195–200 (2012).
quantum processing modules, photons may be interfaced with a variety 26. van Leent, T. et al. Entangling single atoms over 33 km telecom fibre. Nature 607, 69–73
(2022).
of systems. The connectivity and reconfigurability enabled by photonic
27. Moehring, D. L. et al. Entanglement of single-atom quantum bits at a distance. Nature
networks provides a scalable approach for other quantum comput- 449, 68–71 (2007).
ing platforms, such as diamond colour centres and neutral atoms. 28. Stephenson, L. J. et al. High-rate, high-fidelity entanglement of qubits across an
elementary quantum network. Phys. Rev. Lett. 124, 110501 (2020).
Also, modules of different platforms could be connected by means of
29. Saha, S. et al. High-fidelity remote entanglement of trapped atoms mediated by time-bin
wavelength conversion, enabling a hybrid DQC platform. Furthermore, photons. Preprint at https://arxiv.org/abs/2406.01761 (2024).
teleportation protocols are not limited to qubits; they can be extended 30. Huang, Y.-F., Ren, X.-F., Zhang, Y.-S., Duan, L.-M. & Guo, G.-C. Experimental teleportation
of a quantum controlled-NOT gate. Phys. Rev. Lett. 93, 240501 (2004).
to higher-dimensional quantum computing models, such as qudits52
31. Gao, W.-B. et al. Teleportation-based realization of an optical quantum two-qubit
and continuous-variable quantum computing53,54, allowing these plat- entangling gate. Proc. Natl Acad. Sci. USA 107, 20869 (2010).
forms to benefit from the DQC architecture. Quantum repeater tech- 32. Chou, K. S. et al. Deterministic teleportation of a quantum gate between two logical
qubits. Nature 561, 368–373 (2018).
nology16 would enable large physical separation between the quantum 33. Bäumer, E. et al. Efficient long-range entanglement using dynamic circuits. PRX Quantum
processing modules, thereby paving the way for the development of 5, 030339 (2024).
a quantum internet55. The scope of these networks extends beyond 34. Hashim, A. et al. Efficient generation of multi-partite entanglement between non-local
superconducting qubits using classical feedback. Preprint at https://arxiv.org/abs/
quantum computing technologies; the ability to control distributed 2403.18768 (2024).
quantum systems, as enabled by this architecture, to engineer complex 35. Maunz, P. et al. Heralded quantum gate between remote quantum memories. Phys. Rev.
quantum resources has applications in multipartite secret sharing56, Lett. 102, 250502 (2009).
36. Daiss, S. et al. A quantum-logic gate between distant quantum-network modules.
metrology57 and examining fundamental physics58. Science 371, 614–617 (2021).
37. Drmota, P. et al. Robust quantum memory in a trapped-ion quantum network node. Phys.
Rev. Lett. 130, 090803 (2023).
Online content 38. Harty, T. P. et al. High-fidelity trapped-ion quantum logic using near-field microwaves.
Phys. Rev. Lett. 117, 140501 (2016).
Any methods, additional references, Nature Portfolio reporting summa- 39. Ballance, C. J., Harty, T. P., Linke, N. M., Sepiol, M. A. & Lucas, D. M. High-fidelity
ries, source data, extended data, supplementary information, acknowl- quantum logic gates using trapped-ion hyperfine qubits. Phys. Rev. Lett. 117, 060504
(2016).
edgements, peer review information; details of author contributions 40. Hughes, A. C. et al. Benchmarking a high-fidelity mixed-species entangling gate. Phys.
and competing interests; and statements of data and code availability Rev. Lett. 125, 080504 (2020).
are available at https://doi.org/10.1038/s41586-024-08404-x. 41. Brickman, K.-A. et al. Implementation of Grover’s quantum search algorithm in a scalable
system. Phys. Rev. A 72, 050306 (2005).
42. DiCarlo, L. et al. Demonstration of two-qubit algorithms with a superconducting quantum
1. Grover, L. K. Quantum telecomputation. Preprint at https://arxiv.org/abs/quant-ph/9704012 processor. Nature 460, 240–244 (2009).
(1997). 43. Gaebler, J. P. et al. High-fidelity universal gate set for 9Be+ ion qubits. Phys. Rev. Lett. 117,
2. Cirac, J. I., Ekert, A. K., Huelga, S. F. & Macchiavello, C. Distributed quantum computation 060505 (2016).
over noisy channels. Phys. Rev. A 59, 4249 (1999). 44. Erhard, A. et al. Characterizing large-scale quantum computers via cycle benchmarking.
3. Jiang, L., Taylor, J. M., Sørensen, A. S. & Lukin, M. D. Distributed quantum computation Nat. Commun. 10, 5347 (2019).
based on small quantum registers. Phys. Rev. A 76, 062323 (2007). 45. Srinivas, R. et al. High-fidelity laser-free universal control of trapped ion qubits. Nature
4. Monroe, C. et al. Large-scale modular quantum-computer architecture with atomic 597, 209–213 (2021).
memory and photonic interconnects. Phys. Rev. A 89, 022317 (2014). 46. Clark, C. R. et al. High-fidelity Bell-state preparation with 40Ca+ optical qubits. Phys. Rev.
5. Grover, L. K. A fast quantum mechanical algorithm for database search. In Proc. Twenty- Lett. 127, 130505 (2021).
Eighth Annual ACM Symposium on Theory of Computing, STOC ’96 212–219 (Association 47. Moses, S. A. et al. A race-track trapped-ion quantum processor. Phys. Rev. X 13, 041052
for Computing Machinery, 1996). (2023).
6. Vidal, G. & Dawson, C. M. Universal quantum circuit for two-qubit transformations with 48. Weber, M. A. et al. Robust and fast microwave-driven quantum logic for trapped-ion
three controlled-NOT gates. Phys. Rev. A 69, 010301 (2004). qubits. Phys. Rev. A 110, L010601 (2024).

Nature | www.nature.com | 5
Article
49. Nigmatullin, R., Ballance, C. J., de Beaudrap, N. & Benjamin, S. C. Minimally complex ion 58. Greenberger, D. M., Horne, M. A., Shimony, A. & Zeilinger, A. Bell’s theorem without
traps as modules for quantum communication and computing. New J. Phys. 18, 103028 inequalities. Am. J. Phys. 58, 1131–1143 (1990).
(2016).
50. Hilder, J. et al. Fault-tolerant parity readout on a shuttling-based trapped-ion quantum Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
computer. Phys. Rev. X 12, 011032 (2022). published maps and institutional affiliations.
51. Knollmann, F. W. et al. Integrated photonic structures for photon-mediated entanglement
of trapped ions. Opt. Quantum 2, 230–244 (2024). Open Access This article is licensed under a Creative Commons Attribution
52. Luo, Y.-H. et al. Quantum teleportation in high dimensions. Phys. Rev. Lett. 123, 070505 4.0 International License, which permits use, sharing, adaptation, distribution
(2019). and reproduction in any medium or format, as long as you give appropriate
53. Lloyd, S. & Braunstein, S. L. Quantum computation over continuous variables. Phys. Rev. credit to the original author(s) and the source, provide a link to the Creative Commons licence,
Lett. 82, 1784 (1999). and indicate if changes were made. The images or other third party material in this article are
54. Walshe, B. W., Baragiola, B. Q., Alexander, R. N. & Menicucci, N. C. Continuous-variable included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
gate teleportation and bosonic-code error correction. Phys. Rev. A 102, 062411 (2020). to the material. If material is not included in the article’s Creative Commons licence and your
55. Wehner, S., Elkouss, D. & Hanson, R. Quantum internet: a vision for the road ahead. intended use is not permitted by statutory regulation or exceeds the permitted use, you will
Science 362, eaam9288 (2018). need to obtain permission directly from the copyright holder. To view a copy of this licence,
56. Hillery, M., Bužek, V. & Berthiaume, A. Quantum secret sharing. Phys. Rev. A 59, 1829 visit http://creativecommons.org/licenses/by/4.0/.
(1999).
57. Kómár, P. et al. A quantum network of clocks. Nat. Phys. 10, 582–587 (2014). © The Author(s) 2025

6 | Nature | www.nature.com
Methods However, state-preparation and measurement errors would mani-
fest as errors in the reconstructed process. We therefore model the
Dual-species ion-trap modules imperfect state preparation by replacing the ideal input states, |ψi⟩,
Our apparatus comprises two trapped-ion processing modules, with the states
Alice and Bob. Each module, shown in Extended Data Fig. 1, consists
of an ultrahigh-vacuum chamber containing a room-temperature, ρi = Ri [(1 − ϵ)∣0⟩⟨0∣ + ϵ∣1⟩⟨1∣]Ri† , (4)
microfabricated surface Paul trap; the trap used in Alice (Bob) is
a HOA-2 (ref. 59) (Phoenix60) trap, fabricated by Sandia National in which ϵ is the state-preparation error. Note that this model assumes
Laboratories. In each module, we co-trap 88Sr+ and 43Ca+ ions. Each that imperfect state preparation leaves the ionic state within the qubit
species of ion is addressed by means of a set of lasers used for cool- subspace; however, imperfect state preparation often results in leakage
ing, state preparation and readout. A high-numerical-aperture outside this subspace. Nevertheless, for the purposes of our analysis,
(0.6 NA) lens enables single-photon collection from the Sr+ ions. this model is sufficient.
A roughly 0.5 mT magnetic field is applied parallel to the surface of Similarly, we model the imperfect qubit readout by replacing the
the trap to lift the degeneracies of the Zeeman states and provide a projectors P0,j and P1,j with the positive-operator-valued measures
quantization axis.
As outlined in the main text, the Sr+ ion provides an optical network M0, j = (1 − ϵ 0)Rj†∣0⟩⟨0∣Rj + ϵ 1Rj†∣1⟩⟨1∣Rj (5)
qubit, Q N, which is manipulated directly using a 674-nm laser. The
ground hyperfine manifold of the Ca+ ion provides a circuit qubit, Q C. M1, j = (1 − ϵ 1)Rj†∣1⟩⟨1∣Rj + ϵ 0Rj†∣0⟩⟨0∣Rj , (6)
At about 0.5 mT, the sensitivity of theQ C qubit transition frequency to
magnetic-field fluctuations is 122 kHz mT−1, that is, about two orders
of magnitude lower than that of the Q N qubit with a sensitivity of in which ϵ0 and ϵ1 are the computational basis readout errors. The values
−11.2 MHz mT−1, making it an excellent memory for quantum informa- for these errors are given in Extended Data Table 1.
tion37. Furthermore, we define an auxiliary qubit, Q X, in the ground To quantify the performance of a process, E , compared with an ideal
hyperfine manifold of Ca+ for implementing local entangling opera- unitary process, U, we make use of the average gate fidelity
tions, state preparation and readout. The measured state-preparation
and measurement errors for each qubit are presented in Extended FE, U = ∫ dψ ⟨ψ∣U †E(∣ψ ⟩⟨ψ∣)U∣ψ ⟩ (7)
Data Table 1.
The spectral isolation between the two species allows us to address as defined by Nielsen62, which corresponds to the fidelity averaged
one species without causing decoherence of the quantum informa- over all pure input states. We define the process E′ as the application
tion encoded in the other species. We make use of this property for of the process E followed by the inverse of the ideal process U, such
sympathetic cooling, mid-circuit measurement and interfacing with that
the quantum network during circuits.
E ′(ρ) = U †E(ρ)U . (8)
Quantum process tomography
The action of a quantum process acting on a system of N qubits may If χαβ
′ is the process matrix representing E′, as in equation (1), then
be represented by the process matrix χαβ such that the average gate fidelity can be expressed as
D −1 1 + dχ00
′
E(ρ) = ∑ χαβ PαρPβ†, (1) FE, U = , (9)
α , β =0
1+d

in which Pα ∈ P ⊗N are the set of single-qubit Pauli operators in which d is the dimension of the Hilbert space.
P = {I, σx , σy , σz } and D = dim(P ⊗N ) = 4N . Quantum process tomogra­ Resampling of the measurement outcomes is used to generate new
phy enables us to reconstruct the matrix χαβ, thereby attaining a datasets, which are analysed in the same way as the original dataset
com­plete characterization of the process. and are used to determine the sensitivity of the analysis to the statisti-
Quantum process tomography is performed by preparing the qubits cal fluctuations in the input data. The error bar on the average gate
in the states ρi = |ψi⟩⟨ψi|, in which |ψi⟩ are chosen from a tomographi- fidelity of a reconstructed process is quoted as the standard deviation
cally complete set of average gate fidelities of processes reconstructed from resampled
datasets.
 |0⟩ + |1⟩ |0⟩ + i|1⟩ 
|ψi ⟩ ∈ |0⟩, |1⟩, , , (2)
 2 2  Remote entanglement generation
The heralded generation of remote entanglement between network
performing the process, E , followed by measuring the output state qubits in separate modules, outlined in ref. 28, is central to our QGT
E[ρi ] in a basis chosen from a tomographically complete set. Using protocol. Spontaneously emitted 422-nm photons entangled with the
diluted maximum-likelihood estimation61, the outcomes of the meas- Sr+ ions are collected from each module using high-numerical-aperture
urements can be used to reconstruct the χ matrix representing the lenses and single-mode optical fibres bring the photons to a central
process. In practice, the input states are created by rotating |0⟩ to Bell-state analyser, in which a measurement of the photons projects
|ψi⟩ = Ri|0⟩ with the ions into a maximally entangled state28,63,64. This forms the photonic
quantum channel interconnecting the two modules. Following ref. 28,
 1 1  we use a 674-nm π-pulse to map the remote entanglement from the
Ri ∈ I, σx , (I − iσy), (I + iσx) . (3)
 2 2  ground-state Zeeman qubit to an optical qubit, which we refer to as
the network qubit, to minimize the number of quadrupole pulses in
Likewise, the tomographic measurements are performed by rotating subsequent operations. Successful generation of entanglement is
the output state E[ρi ] by Rj† (equation (3)) and subsequently measur­ heralded by particular detector click patterns and, after subsequent
ing it in the σz basis. Ideally, this sequence implements the projectors local rotations, indicate the creation of the maximally entangled Ψ +
P0,j = |ψj⟩⟨ψj| and P1, j = ∣ψ⊥, j ⟩⟨ψ⊥, j∣, in which ⟨ψ⊥, j∣ψj ⟩ = 0. Bell state
Article
|10⟩ + |01⟩ the appropriate Z rotations required to correct for the propagation
|Ψ +⟩ = ∈ Q ⊗2
N .
2 through the CZ gate. With this method, we suppress the dephasing
errors in the circuit qubits during entanglement generation while mini-
This process is executed while simultaneously storing quantum infor- mizing the time between successfully heralding the entanglement and
mation in the circuit qubits, which—as demonstrated in ref. 37—are consuming it for QGT.
robust to this network activity. We deploy Knill dynamical decoupling65,66 with a 7.4-ms interpulse
Each entanglement generation attempt takes 1,168 ns and it takes delay (corresponding to a pulse every three rounds of interleaved entan-
7,084 attempts to successfully herald entanglement on average, cor- glement attempts and recooling). We use quantum process tomogra-
responding to a success probability of 1.41 × 10−4. To mitigate heating phy to reconstruct the process of storing the quantum information
of the ion crystal, we interleave 200 μs of entanglement generation while generating entanglement; ideally, this process would not alter
attempts with 2.254 ms of sympathetic recooling of the Sr+–Ca+ crys- the quantum information stored in the circuit qubit. Quantum process
tal using the Sr+ ion. The sympathetic recooling comprises 1.254 ms tomography is implemented by choosing input states for the circuit
of Doppler cooling, followed by 1 ms of electromagnetically induced qubits from the tomographically complete set given in equation (2),
transparency cooling. Overall, this results in an average entangle­ generating remote entanglement between the network qubits
ment generation rate of 9.7 s−1 (equivalently, it takes, on average, while dynamically decoupling the circuit qubits, then—on successful
103 ms to generate entanglement between network qubits), although herald—completing the dynamical decoupling sequence and per­
this rate could be increased by optimizing the interleaved cooling forming tomographic measurements of the circuit qubits. The recon-
sequence. This rate is lower than the 182 s−1 rate previously reported structed process matrices for each module corresponding to the action
in our apparatus28 owing to the extra cooling. We characterize the of storing quantum information during entanglement generation are
remote entanglement using quantum state tomography; by perform- shown in Extended Data Fig. 2c. We observe fidelities to the ideal opera-
ing tomographic measurements on 2 × 105 copies of the remotely tion of 98.1(4)% and 98.2(5)% for Alice and Bob, respectively.
entangled state, we reconstruct the density matrix of the network
qubits, ρNAB, shown in Extended Data Fig. 2d. To isolate the fidelity of Local mixed-species entangling gates
the ‘quantum link’ in Fig. 2, we account for the imperfect tomographic The ability to perform logical entangling gates between ions of dif-
measurements in the reconstruction of the density matrix using the ferent species allows us to separate the roles of network and circuit
positive-operator-valued measures in equations (5) and (6). The fidel- ions. We implement mixed-species entangling gates following the
ity of the reconstructed state to the desired Ψ+ Bell state, given by approach taken in ref. 40, in which geometric phase gates are determin-
⟨Ψ +|ρNAB |Ψ +⟩, is 96.89(8)%. istically executed using a single pair of 402-nm Raman beams, as shown
We believe that the fidelity is predominantly limited by errors occur- in Extended Data Fig. 3. Here we apply the gate mechanism directly
ring during the generation of ion–photon entanglement in each mod- to the network qubit in Sr+—rather than the Zeeman ground-state
ule, rather than imperfections in the apparatus used to perform the qubit, as done by Hughes et al.40 and Drmota et al.37—at the cost of a
projective Bell-state measurement. In particular, we attribute the pri- slightly reduced gate efficiency that is compensated for by the use
mary sources of error to polarization mixing due to imperfections in of higher laser powers. This enables us to perform mixed-species CZ
the imaging systems used to collect single photons from each module gates between the network and auxiliary qubits. We characterize our
and to drifts in the birefringence of the optical fibres that form the mixed-species entangling gates using quantum process tomography
network link between the modules. in each module, reconstructing the process matrices χCZ representing
the action of the local CZ gate acting between the network and auxil-
Circuit qubit memory during entanglement generation iary qubits. The reconstructed process matrices for each module are
Because each instance of QGT requires the generation of entanglement shown in Extended Data Fig. 3d. Compared with the ideal CZ gate, we
between network qubits, it is necessary to ensure that the circuit qubits observe average gate fidelities of 97.6(2)% and 98.0(2)% for Alice and
preserve their encoded quantum information during this process. Bob, respectively.
Owing to their low sensitivity to magnetic-field fluctuations, the circuit
qubits have exhibited roughly 100 ms coherence times and, in previ- Hyperfine qubit transfer
ous work, we demonstrated these qubits to be robust to network acti­ Because the circuit qubit does not participate in the mixed-species
vity37. We further suppress dephasing through dynamical decoupling. gate, the gate interaction is performed on the network and auxiliary
Typically, dynamical decoupling is implemented over a fixed period of qubits. Consequently, we require the ability to map coherently between
time; however, the success of the entanglement generation process is the circuit and auxiliary qubit before and after the local operations. As
non-deterministic and would therefore leave the dynamical decoupling shown in Extended Data Fig. 4, this mapping is performed using a pair
sequence incomplete. of 402-nm Raman beams detuned by about 3.2 GHz, to coherently drive
One solution would be to complete the dynamical decoupling pulse the transitions within the ground hyperfine manifold of Ca+.
sequence once the entanglement has been generated. However, it is The transfer of the circuit qubit to the auxiliary qubit begins with
desirable to minimize the time between heralding the entanglement the mapping of the state |0C⟩ to the state |0X⟩. However, owing to the
generation and performing the QGT protocol, to prevent dephasing near degeneracy of the transition T 0 : |0C⟩ ↔ |F = 3, mF = +1⟩ and the
of the network qubits. Instead, we make use of the fact that the action transition T 1 : |1 C⟩ ↔ |F = 4, mF = +1⟩ (Extended Data Fig. 4), separated
of a dynamical decoupling pulse on one of the circuit qubits can be by only about 15 kHz, it is not possible to map the |0C⟩ state out of the
propagated through the teleported CZ gate as circuit qubit without off-resonantly driving population out of the |1C⟩
state. We suppress this off-resonant excitation using a composite pulse
(X ⊗ I)UCZ = UCZ(X ⊗ Z). (10) sequence, shown in (1) in Extended Data Fig. 4b, comprising three pulses
resonant with the T 0 transition, with pulse durations equal to the 2π
We therefore perform the dynamical decoupling pulses on the circuit time of the T 1 transition, and phases optimized to minimize the
qubits until we obtain a herald of remote entanglement, at which point off-resonant excitation. This pulse sequence allows us to simultane-
we immediately perform the QGT sequence—implementing a CZ gate ously perform a π-pulse on the T 0 transition and the identity on the
on the state of the circuit qubits at the point of interruption. Once this off-resonantly driven T 1 transition. Raman π-pulses are then used to
gate is completed, we perform the remaining dynamical decoupling complete the mapping to the |0X⟩ state. Another sequence of Raman
pulses (without any interpulse delay) and use equation (10) to apply π-pulses coherently maps |1C⟩ → |1X⟩, thereby completing the transfer
of the circuit qubit to the auxiliary qubit, Q C → Q X. To implement the |1N⟩ state67. We choose a mid-circuit measurement duration of 500 μs
mapping Q X → Q C, the same pulse sequence is applied in reverse. and estimate fluorescence detection errors of 6.6(1) × 10−4 and
We characterize our Q C ↔ Q X mapping sequence by performing a 5.51(2) × 10−4 for Alice and Bob, respectively. Combining these error
modification of single-qubit randomized benchmarking (RBM), in mechanisms, we estimate contributions to the teleported CZ gate error
which we alternate Clifford operations on the Q C and Q X qubits, as of 0.091(3)% and 0.122(2)% for Alice and Bob, respectively.
illustrated in Extended Data Fig. 4c. We assume that (1) the single-qubit
gate errors for the Q C and Q X qubits are negligible compared with
the Q C ↔ Q X transfer infidelity (we typically observe single-qubit Data availability
gate errors of around 1 × 10−4 for the Ca+ hyperfine qubits) and (2) the The datasets generated during this study are available from D.M. and
fidelity of the transfer Q C → Q X is similar to Q X → Q C . We therefore D.M.L. (email: [email protected]) on reasonable request.
we model the survival probability as

1 Code availability
S(m) = + Bp m
2 All analysis code that supports the plots in this paper and other findings
in which m is the number of hyperfine transfers, B accounts for of this study are available from the corresponding author on reason-
state-preparation and measurement error offsets and p is the depo- able request.
larizing probability for the transfer, related to the error per transfer as

1−p 59. Maunz, P. L. W. High Optical Access Trap 2.0 (US Department of Energy, Office of Scientific
ϵ C↔X = . and Technical Information, 2016); https://doi.org/10.2172/1237003.
2
60. Revelle, M. C., Phoenix and Peregrine ion traps. Preprint at https://arxiv.org/abs/2009.02398
The RBM results are shown in Extended Data Fig. 4c; we measure an (2020).
61. Řeháček, J., Hradil, Z., Knill, E. & Lvovsky, A. I. Diluted maximum-likelihood algorithm for
error per transfer of 3.8(2) × 10−3 (2.6(1) × 10−3) for Alice (Bob).
quantum tomography. Phys. Rev. A 75, 042108 (2007).
62. Nielsen, M. A. A simple formula for the average gate fidelity of a quantum dynamical
Conditional operations operation. Phys. Lett. A 303, 249–252 (2002).
63. Stephenson, L. Entanglement Between Nodes of a Quantum Network. DPhil thesis, Univ.
To complete the QGT protocol, the two modules perform mid-circuit
Oxford (2019); https://ora.ox.ac.uk/objects/uuid:5066d87a-55e8-4fd5-b5b8-df1878af97a8.
measurements of the network qubits, exchange the measurement 64. Nadlinger, D. P. Device-independent Key Distribution Between Trapped-ion Quantum
outcomes and apply a local rotation of their circuit qubits conditioned Network Nodes. DPhil thesis, Univ. Oxford (2022).
65. Wang, P. et al. Single ion qubit with estimated coherence time exceeding one hour. Nat.
on the outcomes of the measurements. By virtue of the spectral isola- Commun. 12, 233 (2021).
tion between the two species of ions, mid-circuit measurements of the 66. Souza, A. M., Álvarez, G. A. & Suter, D. Robust dynamical decoupling for quantum
network qubits can be made without affecting the quantum state of the computing and quantum memory. Phys. Rev. Lett. 106, 240501 (2011).
67. Letchumanan, V., Wilson, M. A., Gill, P. & Sinclair, A. G. Lifetime measurement of the
circuit qubits. The mid-circuit measurement outcomes, mA, mB ∈ {0, 1}, metastable 4d2D5/2 state in 88Sr+ using a single trapped ion. Phys. Rev. A 72, 012509 (2005).
are exchanged in real time through a classical communication channel 68. Bourdeauducq, S. et al. m-labs/artiq: 4.0 (4.0). Zenodo https://doi.org/10.5281/zenodo.
between the modules—in our demonstration, this is a TTL link connect- 1492176 (2021).

ing the control systems of the two modules. Following the exchange of
the measurement outcomes, the modules, Alice and Bob, perform the Acknowledgements We thank O. Băzăvan, S. Saner and D. Webb for maintenance of the
conditional rotations UA and UB, respectively, in which 674-nm laser system; C. Ballance and L. Stephenson for the design and construction of
the apparatus; J. Blackmore and P. Juhász for comments on the paper; Sandia National
Laboratories for supplying the ion traps used in this experiment; and the developers of the
  S † if mA ⊕ mB = 0, control system ARTIQ68. D.M. acknowledges support from the US Army Research Office
UA =  (ref. W911NF-18-1-0340). D.P.N. acknowledges support from Merton College, Oxford. E.M.A.
 S otherwise , acknowledges support from the UK EPSRC ‘Quantum Communications Hub’ EP/T001011/1.
 S if mA ⊕ mB = 0, R.S. acknowledges funding from an EPSRC fellowship EP/W028026/1 and Balliol College,


UB =  † Oxford. G.A. acknowledges support from Wolfson College, Oxford. This work was supported
by the UK EPSRC ‘Quantum Computing and Simulation Hub’ EP/T001062/1.
 S otherwise ,
Author contributions D.M., P.D., D.P.N., E.M.A., A.A., B.C.N., R.S. and G.A. built and operated
in which S = diag(1, i). the experimental apparatus. D.M. led the experimental work, with assistance from P.D. and
Errors in the mid-circuit measurements of the network qubits will D.P.N. D.M. performed the data analysis and prepared the paper, with input from all authors.
result in the application of the wrong conditional rotation; effectively, D.M.L. secured funding and supervised the work. All authors contributed to the discussion and
interpretation of results.
this would appear as a joint phase flip of the circuit qubits following
the teleported gate. The mid-circuit measurement errors arise from Competing interests R.S. is partly employed by Oxford Ionics Ltd. The other authors declare
the non-ideal single-qubit rotation of the network qubit to map the no competing interests.
measurement basis onto the computational basis and errors owing to
Additional information
the fluorescence detection of the network qubit. Using RBM, we meas- Supplementary information The online version contains supplementary material available at
ure single-qubit gate errors for the network qubits of 4.8(3) × 10−4 and https://doi.org/10.1038/s41586-024-08404-x.
9.8(3) × 10−4 for Alice and Bob, respectively. The error in the fluores- Correspondence and requests for materials should be addressed to D. Main.
Peer review information Nature thanks Tracy Northup and the other, anonymous, reviewer(s)
cence detection is estimated from the observed photon scattering for their contribution to the peer review of this work. Peer reviewer reports are available.
rates of Q N states, as well as the approximately 390 ms lifetime of the Reprints and permissions information is available at http://www.nature.com/reprints.
Article
  

 


 
 



  




 




 
 


 
 
  
  



  
  

 
 
    
   
 
 
  


 


 

 



 
 








 



 


 


  
   




  


 
   
    
  
 
 

Extended Data Fig. 1 | Outline of a trapped-ion module. a, An ultrahigh- the generation of entanglement. A high-numerical-aperture (0.6 NA) lens
vacuum chamber houses a microfabricated surface Paul trap, which co-traps couples the single photons into a single-mode optical fibre, which connects to
one 88Sr+ ion and one 43Ca+ ion. The ions are manipulated using lasers, which the optical quantum network. Both imaging systems are outside the vacuum
are delivered parallel to the surface of the trap. The Sr+ ion provides an optical chamber. c, Energy-level diagrams for cooling, state preparation and readout
network qubit, Q N, which is coherently manipulated using a 674-nm laser. The of each species. For state preparation and readout of the circuit qubit Q C in
ground hyperfine manifold of the Ca+ ion provides a circuit qubit, Q C, and an Ca+, we use the auxiliary qubit Q X. During state preparation, we prepare ∣0 X⟩
auxiliary qubit, Q X. The qubits in the ground hyperfine manifold are addressed through optical pumping and then transfer it to |0C⟩ using Raman π-pulses.
using a pair of 402-nm Raman beams. b, The rear-side imaging system is used to For readout, we transfer |0C⟩ to |0X⟩ before shelving to the D5/2 manifold.
perform fluorescence detection for qubit readout of both species. The front- Fluorescence detection is then used for both species; ions in the shelved state
side imaging system is used for single-photon collection from the Sr+ ion during do not scatter photons.
   
€   
­  ­
  ­
­

­
­

­­

ƒ
 

‚


­­
­
€ ­
€ ­ € ­
 


Œ
0 ns 1168 ns
    
 
  


 „    
„ †„
Š
†

‡
Š
‰

ˆ
‹Š
† ‡ ‰ ˆ † ‡ ‰ ˆ
„­ Œ
„ †„
Š
†

‡
Š
‰

ˆ
‹Š
† ‡ ‰ ˆ † ‡ ‰ ˆ

Extended Data Fig. 2 | Generation of remote entanglement and robust them on a beam splitter and perform a projective measurement on the
memory of the circuit qubits. a, Entanglement is generated between the two-photon polarization state. Particular detector click patterns occurring
network qubits using 200 μs of entanglement attempts interleaved with within the detection window herald the successful generation of remote
2.254 ms of sympathetic recooling using the Sr+ ion. This is repeated until the entanglement. We then exit the attempt loop and map the entanglement into
entanglement is successfully heralded by a particular detector click pattern. the optical network qubits, Q N, with an extra π-pulse on the 674-nm transition.
While attempting to generate entanglement between the network qubits, Knill c, Difference between the reconstructed process matrices, χmem, for the
dynamical decoupling pulses, K i, are used to preserve the state of the circuit process of storing the state of the circuit qubit in (1) Alice and (2) Bob while
qubits. b, Each entanglement attempt has a total duration of 1168 ns. We perform generating entanglement on the network qubits and the ideal process matrix,
a 320-ns state-preparation pulse (which has a switching latency of 500 ns), χideal = diag(1, 0, 0, 0). The reconstructed process matrices have fidelities
pumping the Sr+ ion into the lower ground Zeeman state. An approximately 98.1(4)% and 98.2(5)% for Alice and Bob, respectively. d, Reconstructed density
5-ps pulse excites the Sr+ ion to the upper P1/2 level (lifetime about 7 ns), which matrix of the remotely entangled network qubits. The state has a fidelity of
rapidly decays to one of the ground Zeeman levels, thereby generating ion– 96.89(8)% to the |Ψ+⟩ Bell state.
photon entanglement. We collect a photon from each of the modules, interfere
Article
  




 

 






 


 






 
 






 ƒ„ † ƒ„
‡ˆ

‚ƒ „ ƒ „ ‚ƒ „ ƒ „
 
­ ­
 
­ ­
­  ­ 
­­ ­­
€ €
€­ €­
 
€ ­ ­ € €€ € ­ ­ € €€ € ­ ­ € €€ € ­ ­ € €€
Extended Data Fig. 3 | Implementation of the local mixed-species CZ gates. experience a spin-dependent force that displaces the spin states in phase
a, Circuit element for the local CZ gate, implemented between the Q N and Q X space, as shown in the inset, thus enabling the implementation of geometric
qubits. b, Geometry for the mixed-species gate mechanism. A pair of Raman phase gates. c, Energy-level diagram for the gate mechanism acting on the Q N
laser beams are aligned orthogonal to one another, such that their relative and Q X qubits. By tuning the Raman lasers to 402 nm, we couple to both the
wavevector, Δk is along the trap axis. The interference of these beams leads 397-nm S1/2 ↔ P1/2 dipole transition in Ca+ and the 408-nm S1/2 ↔ P3/2 dipole
to a polarization travelling-standing wave, which induces spin-dependent transition in Sr+. d, Process matrices for the local, mixed-species CZ gates
light shifts oscillating at a frequency close to the frequency of the axial out-of- for (1) Alice and (2) Bob. The process matrices have average gate fidelities of
phase motional mode. Owing to the spatial gradient of the light shift, the ions 97.6(2)% and 98.0(2)% for Alice and Bob, respectively.
  

1.0 Alice
 Bob


Survival Probability
 
 0.9



 0.8


 
 0.7
2 26 51 75 100
Number of Transfers

  

Extended Data Fig. 4 | Transfer between the circuit and auxiliary qubits. experimentally. The subsequent transfer pulses (thin blue arrows) are π-pulses
a, Circuit element and level diagram showing the coherent transfer of quantum on the relevant transitions. This sequence therefore performs the mapping
information from the Q C qubit to the Q X qubit. The inverse transfer is |0C⟩ → |0X⟩, leaving the state |1C⟩ unaffected. (2) The second step comprises a
implemented by performing the same steps in reverse. b, The transfer pulse sequence of π-pulses that maps |1C⟩ → |1 X⟩. This completes the coherent transfer
sequence comprises two steps. (1) The first step maps the state |0C⟩ to |0X⟩. Q C → Q X. c, The performance of the transfer sequence is characterized using a
Owing to the near-degeneracy of the intended transition T 0 : ∣0C⟩ ↔ ∣✓⟩ (thick modified version of RBM, in which we alternately perform Clifford operations
blue arrow) and the unwanted transition T 1 : 1 C⟩ ↔ ∣✗⟩ (red dashed arrow), on the Q C and Q X qubits. By measuring the survival probability for different
separated by only about 15 kHz, we use a composite pulse sequence to suppress numbers of transfers, and neglecting the errors of the single-qubit gates Ci
off-resonant coupling to the T 1 transition. The composite pulse sequence, (which are about 1 × 10 −4), we extract the error per transfer, ϵC↔X, yielding
shown in the dashed box, comprises three pulses of duration τ resonant with 3.8(2) × 10 −3 and 2.6(1) × 10 −3 for Alice and Bob, respectively. All error bars
the T 0 transition with differing phases ϕi . The pulse duration, τ, is equal to indicate one standard deviation.
the 2π time of the T 1 transition, ϕ1 = ϕ3 = 0, and ϕ2 ≈ 2π × 0.231 is optimized
Article
Extended Data Table 1 | State-preparation and measurement
errors for all of the qubit states, in each module

   

  

     

   

  

   

    

The average state-preparation and measurement error is 5.0(2) × 10−3.