ARTICLE

https://doi.org/10.1038/s41467-022-28767-x OPEN

A programmable qudit-based quantum processor

Yulin Chi1,12, Jieshan Huang 1,12, Zhanchuan Zhang1,12, Jun Mao 1, Zinan Zhou1, Xiaojiong Chen1,
Chonghao Zhai1, Jueming Bao1, Tianxiang Dai 1, Huihong Yuan1,2, Ming Zhang3, Daoxin Dai3, Bo Tang4,
Yan Yang 4, Zhihua Li4, Yunhong Ding 5,6, Leif K. Oxenløwe 5,6, Mark G. Thompson 7,
Jeremy L. O’Brien8, Yan Li 1,9,10,11, Qihuang Gong 1,2,9,10,11 & Jianwei Wang 1,2,9,10,11 ✉
1234567890():,;

Controlling and programming quantum devices to process quantum information by the unit
of quantum dit, i.e., qudit, provides the possibilities for noise-resilient quantum commu-
nications, delicate quantum molecular simulations, and efficient quantum computations,
showing great potential to enhance the capabilities of qubit-based quantum technologies.
Here, we report a programmable qudit-based quantum processor in silicon-photonic inte-
grated circuits and demonstrate its enhancement of quantum computational parallelism. The
processor monolithically integrates all the key functionalities and capabilities of initialisation,
manipulation, and measurement of the two quantum quart (ququart) states and multi-value
quantum-controlled logic gates with high-level fidelities. By reprogramming the configuration
of the processor, we implemented the most basic quantum Fourier transform algorithms, all
in quaternary, to benchmark the enhancement of quantum parallelism using qudits, which
include generalised Deutsch-Jozsa and Bernstein-Vazirani algorithms, quaternary phase
estimation and fast factorization algorithms. The monolithic integration and high program-
mability have allowed the implementations of more than one million high-fidelity prepara-
tions, operations and projections of qudit states in the processor. Our work shows an
integrated photonic quantum technology for qudit-based quantum computing with enhanced
capacity, accuracy, and efficiency, which could lead to the acceleration of building a large-
scale quantum computer.

1 State Key Laboratory for Mesoscopic Physics, School of Physics, Peking University, 100871 Beijing, China. 2 Beijing Academy of Quantum Information

Sciences, 100193 Beijing, China. 3 State Key Laboratory for Modern Optical Instrumentation, College of Optical Science and Engineering, Ningbo Research
Institute, International Research Center for Advanced Photonics, Zhejiang University, 310058 Hangzhou, China. 4 Institute of Microelectronics, Chinese
Academy of Sciences, 100029 Beijing, China. 5 Department of Photonics Engineering, Technical University of Denmark, 2800 Kgs. Lyngby, Denmark.
6 Center for Silicon Photonics for Optical Communication (SPOC), Technical University of Denmark, 2800 Kgs. Lyngby, Denmark. 7 Quantum Engineering

Technology Labs, H. H. Wills Physics Laboratory and Department of Electrical and Electronic Engineering, University of Bristol, BS8 1FD Bristol, United
Kingdom. 8 Department of Physics, The University of Western Australia, Perth 6009, Australia. 9 Frontiers Science Center for Nano-optoelectronics &
Collaborative Innovation Center of Quantum Matter, Peking University, 100871 Beijing, China. 10 Collaborative Innovation Center of Extreme Optics, Shanxi
University, 030006 Taiyuan, Shanxi, China. 11 Peking University Yangtze Delta Institute of Optoelectronics, Nantong 226010 Jiangsu, China. 12These authors
contributed equally: Yulin Chi, Jieshan Huang and Zhanchuan Zhang. ✉email: jww@pku.edu.cn

NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x

N
atural quantum matters store rich multidimensional reconfigured and reprogrammed to implement different tasks has not
quantum information in a superposition of more than been realised to date, in any quantum system. Likely, it requires an
two electronic or mechanical modes. Engineering artificial integrated platform35,54,55, capable of initialising, manipulating and
multilevel quantum devices to mimic nature may allow funda- measuring qudit states and gates, in a fully controllable and highly
mental innovations and technological advances. Recently, though programmable manner. Realising a programmable qudit-based
the state-of-the-art qubit-based quantum technologies have demon- quantum processor therefore presents a significant step to transi-
strated revolutionary milestones, e.g., loophole-free Bell tests1,2, tion the technological advances of controlling qudit states and logic
satellite-relayed quantum communications3,4 and quantum compu- gates to the implementations of quantum tasks and quantum com-
tational advantages5,6, qudit-based quantum technologies might be putational algorithms, in d-ary.
able to further enhance quantum capabilities as they are intrinsically In this work, we demonstrate a programmable qudit-based
consistent with our natural quantum systems. For example, entangled quantum processing unit (d-QPU) on a large-scale silicon-pho-
qudit states can strengthen the Bell nonlocality7 and moderate the tonic quantum chip. The initialisation, manipulation and mea-
detection loophole8; distributing qudit states allows high-capacity surement of arbitrary single-qudit and two-qudit states, and
noise-resilient quantum cryptography9–11; by mapping Hamiltonians multi-value quantum-controlled logic gates can be implemented
into multilevel quantum devices, it can provide a direct solution for on the single d-QPU chip. Such a fully monolithic integration of
quantum simulations of complex molecular and physical all necessary functionalities allows the implementation of a top-
systems12–16; more importantly, universal quantum computation down hierarchy of programmable qudit-based quantum compu-
with qudits is possible in both of the circuit models17 and tation, as shown in Fig. 1. Different quantum tasks and compu-
measurement-based models18,19, requires less resource overhead in tational algorithms are implemented, all in quaternary, by
quantum error correction20,21, and can improve the execution of recompiling the qudit logic circuit in the software level, and then
quantum algorithms22,23. Heuristically, the exponential speed-up of executing the circuit by reprograming the configurations of the d-
many quantum algorithms is enabled by the quantum parallel eva- QPU chip in the hardware level. We then benchmark the
luation
of a function f(x) for all input x values simultaneously, as enhancement of quantum computational parallelism, by per-
∑x jxi
f ðxÞ , where the x input string is represented by a super-
position of quantum states. The adoption of qudit as the basic
quantum information unit in processing quantum algorithms24 offers
enhanced computational capacity that is represented by the size of
the Hilbert space of dn, where n is the number of qudits and d is the
local size of each qudit. Moreover, it can lead to higher computational
accuracy for example in implementing quantum Fourier transform
algorithms such as Shor’s fast factorisation25 and phase estimation26,
in which the computational accuracy is determined by the size of
auxiliary qudits. Processing the Kitaev’s version of quantum Fourier
algorithms26–29 with qudits may allow further speed-up of quantum
computing. These unique capabilities have strongly prompted the
development of qudit-based quantum computing in universal
models17–19,30,31, and very recently in experimental controls of qudit
states and logic gates in photonics32, solid-state15, trapped ion33, and
superconducting34 platforms. In particular, photons are intrinsically
multidimensional35, enabling flexible and reliable encoding of qudits
with their different degrees-of-freedom, e.g, path36,37, frequency38,39,
spatial mode40,41 and temporal mode11,42. Advances in the control of
quantum photonic devices have recently allowed remarkable
experimental progress. For example, multidimensional Greenberger-
Horne-Zeilinger (GHZ) states and cluster states prepared in the
frequency-bins and time-bins of two photons generated in a single
microring resonator43,44, have firstly shown enhancement in quan-
tum computation by providing increased quantum resources and Fig. 1 The top-down hierarchy of qudit-based quantum computation.
higher noise robustness compared to the qubit counterparts; An Users can define different quantum tasks and implement different quantum
integrated photonic chip for the generation, manipulation and algorithms in d-ary, e.g., the generalised Deutsch-Jozsa (DJ), Bernstein-
measurement of two-photon multidimensional Bell states has been Vazirani (BV), quantum phase estimation (QPEA) and Shor’s fast
demonstrated36, while the scaling capability has been verified by the factorization algorithms. In the software level, a multi-qudit quantum logical
generation of multiphoton multidimensional GHZ states45,46, and circuit for executing the algorithm is compiled with single-qudit gates (e.g,
the realisation of single-qudit quantum teleportation47,48. Despite of Zd, Xd and Hd) and multi-value controlled-unitary (MVCUd) gates. In the
these remarkable development of multidimensional quantum pho- hardware level, the logical circuit is physically implemented by an
tonic technologies that mainly focus on the preparation and control integrated photonic quantum device, i.e, the programmable d-QPU, and the
of qudit states and gates, a monolithically integrated quantum device gate operations are realised by optical waveguide devices, such as
that is able to initialise, manipulate and analyze qudit states and gates entangled photon sources, phase shifters, beamsplitters and
is lacking. Furthermore, the programmability of quantum hardware interferometers. Multiple quantum tasks and algorithms can be executed,
presents the major enabling capability of quantum computing tech- without the need of altering the device, only by reprogramming the
nologies. For example, several milestones in qubit-based quantum configurations of waveguide circuits. The outcome of the hardware is given
computing have been all realised in programmable quantum devices by photon coincidence counts, which are recorded and analysed by classical
of photons49,50, trapped ions51,52, superconductors5,6 and electronics and classical computer. Experimental outcomes can be feed-
semiconductors53. However, limited to the best to our knowledge, forwarded into the d-QPU for the implementation of Kitaev’s version of
such a qudit-based quantum computing device that can be fully quantum Fourier transform algorithms.

2 NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x ARTICLE

forming the generalised Deutsch-Jozsa and Bernstein-Vazirani where pi, qi are theoretical and measured distributions, respec-
pp ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffi pffiffiffiffiffi 2
algorithms, quaternary phase estimation and order finding algo- tively; the state Fq is defined as ðTr½ ρ0  ρ  ρ0 Þ , where ρ0, ρ
rithms. Our results show a proof-of-principle demonstration of are ideal and measured states, respectively; the process Fq is
qudit-based quantum computer with integrated optics, that defined as Tr[χ0χ], where χ0, χ are ideal and reconstructed process
allows improvement of the capacity, accuracy and efficiency of matrices, respectively.
quantum computing. We first characterised the single-ququart and two-ququart logic
gates. As examples, two single-ququart gates are characterised: the
Results


E


generalised d-level Pauli-Xd gate that is defined as X d ki ¼
kðid 1Þ
Scheme of multiqudit quantum processor. Figure 2 shows the
core of a multiqudit processor, i.e, the multiqudit multi-value con- where ⊕ d is addition module of d, and the d-level
quantum
 Fourier
trolled logic gate, which is realised by the following three steps: gate Fd that transforms the computational basis of
ki to the Fourier



generation of the multiphoton multidimensional Greenberger- basis
f i of p1ffiffid ∑d1 ij

j¼0 ω j where i, j ∈ Nd and ω = exp ði d Þ. When
2π
Horne-Zeilinger entangled state jGHZinþ1;d 45,46, which enables the
d is two, they return to the standard Pauli and Fourier (Hadamard)
entangling operations between the multiqudit states; Hilbert space gates for qubits. In Fig. 2c inset, it shows the measured mean Fc of
expansion of each qudit in y-register to form an entire space of d2n, 0.988(13) for the five X4 gates and 0.967(19) for the five F4 gates,
that locally allows individual and arbitrary single-qudit operations56; where the values in parentheses are uncertainty from photon
coherent compression of the entire state back to a dn space57. These statistics. Next, we characterised the two-qudit entangling gate:
sequences of operations result in a multiqudit

E Q multi-value


controlled-

E


1ffiffi d1
n

MVCU jxi
y ¼ jxiOx
y ; ð1Þ
unitary (MVCU) gate as d ∑j¼0
kj  i¼1 Oi;j ϕ i , where
kj
p


in the auxiliary x-register presents 59

where O can be arbitrarily operated on the jxi and y registers.

the logical state in the j-th mode
(for simplicity it is denoted as
j ), and Oi,j in the data
y-register Notably, the MVCU gate presents a coherent entanglement between

refers to an arbitrary local operation on the qudit state
ϕ i that is the auxiliary x-register and the data y-register. The processing of d-
initialised by the Pi qudit generator. Such multiqudit MVCU gate ary quantum algorithms relies on the multiple path interference in
works with a (1/d) success probability regardless of n (see Supple- the d-dimensional Fourier gate to obtain the desired solutions. Such
mentary Note 3 and Supplementary Fig. 1). The quantum circuits in coherent superposition of qudits ensures quantum parallelism, that is
Fig. 2a, b provide a scheme of implementing multiqudit quantum function evaluations for multiple inputs are executed in parallel. The
Fourier algorithms in the scalable Kitaev’s framework26–29. MVCU is thus a core logic enabling the quantum parallel evaluation
Figure 2c illustrates the integrated photonic quantum circuits for a of the function. For example, as the d-ary generalisation of the CNOT
two-ququart version of qudit-based quantum processing unit (d- gate24, the MVCXd gate allows the creation of a
complete  set of four-
level Bell states jΨii;j defined as 12 ∑3m¼0 ωmi jmi
md j , by inputing



E
QPU). It was fabricated in silicon using the complementary metal-
oxide-semiconductor (CMOS) process with the 248nm deep the
f i 
kj states into the logic, i, j = 0,1,2,3. Figure 3a shows the
ultraviolet lithography (see a device image in Fig. 2d). The processor
allows the generation of a path-encoded two-ququart entangled state reconstructed jΨi12 state, and Fig. 3c shows measured Fq for the 16
of jGHZi2;4 (i.e., the 4-dimensional generalised Bell state of jBelli4 ), Bell states with an averaged fidelity of 0.967(31). The state matrices
(ρ) represented as a linear combination of Gell-Mann matrices were
by a coherent excitation of four integrated spontaneous four-wave-
reconstructed by implementing compressed sensing quantum state
mixing (SFWM) sources. It is followed by the sequences of processes
tomography techniques60. In addition,
a fully  product state was
of “space expansion–local operation–coherent compression" for the
realisation of d-QPU, see Fig. 2b. The d-QPU chip monolithically created in Fig. 3b, given an input of
f 0 f 0 . Figure 3d shows the
integrates the core capabilities and functionalities, including arbitrary experimental process matrix (χ) of the MVCXd gate, by performing
single-ququart preparation (P), arbitrary two-ququart MVCU quantum process tomography with a full set of 256 state tomographic
operation (that presents a d-ary generalisation of two-qubit measurements61, and a process fidelity Fq of 0.952 was obtained. We
controlled-unitary operation), and arbitrary single-ququart measure- then characterised the MVCZd
gate  (Zd is the

 generalised d-level
ment (M). Though on-chip generation, manipulation and measure- Pauli-Zd gate) transforming jxi
y to jxiωxy
y , and the MVCHd
ment of entangled qudit states have been reported36, this work gate where Hd is the d-level Hadamard gate with elements of
demonstrate the key abilities to initialize, manipulate, and analyze p1ffiffi ð1Þij (i ⊙ j is the bitwise dot product, see Supplementary
d
qudit states and gates in a fully reconfigurable and reprogrammable Note 2). Instead of performing full process tomography, we adopted
manner, providing a major technological advance for qudit quantum an efficient characterisation by using complementary classical
computing. In Fig. 2d it shows one of the largest-scale programmable fidelity62. Figure 3e–j show measured input-output truth tables and
quantum photonic chip having 451 photonic components, including their classical fidelity (Fc1, Fc2) for the MVCU in two complementary
116 reconfigurable phase-shifters (see their characterisations in Fig. 2c {base I, base II}, from which the complementary classical fidelity is
insets). The two-photons detection rate at the magnitude of 103/s was upper and lower bounded by [Fc1 + Fc2 − 1, Min(Fc1, Fc2)].
measured in the two-ququart device, which is six orders higher than
that in a four-qubit device (note the detection rate depends on the Experimental implementation of d-ary Deutsch’s algorithms.
performance and loss of the quantum devices as well as their The class of Deutsch’s algorithms well identify quantum parallelism.
pumping and measurement apparatuses)58. Details of device A generalised d-ary Deutsch-Jozsa algorithm can determine whether
fabrication, state evolution and experimental setup are provided in a multi-value function f: {0, 1,..., d−1}n → {0, 1,..., d − 1} is constant
Supplementary Notes 1 and 3. or balanced by a single query of a quantum oracle63. Classically, it
however requires dn−1 + 1 queries. The quantum circuit performing
Characterisation of d-ary multi-value controlled-unitary gates. f(x) ⊕ dy is shown in Fig. 4a. In the case of d = 2, it returns to the
Before reporting experimental results, we first define classical original binary Deutsch-Jozsa64. We implemented the ququart
statistic fidelity (Fc) and quantum state (process) fidelity (Fq), Deutsch-Jozsa algorithm on the d-QPU for the case of n = 1 and
used in this work to quantify the qudit states, logic gates and
pffiffiffiffiffiffiffi d = 4. Figure 4b–h show the measured probability distributions of
algorithm implementations. The Fc is defined as ð∑i pi qi Þ2 , the x-register in the computational basis, when the multi-value

NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x

Fig. 2 A qudit-based programmable quantum processing unit in a photonic integrated circuit chip. a Quantum circuit, and b physical implementation of
the multiqudit QPU. It bases on multiphoton multidimensional entanglement of jGHZinþ1;d , where n + 1 is the number of photonic qudits and d is the local
dimensionality of each qudit. Pi is an arbitrary single-qudit gate; Fd is a generalised d-level Fourier gate; Mi is an arbitrary single-qudit projector; Oi,j
(i = 1,..., n, j = 1,..., d) is an arbitrary single-qudit logic gate that is locally performed on the i-th qudit of the y-register, and the Oi,j gates are coherently
entangled with the x-register state. The process of “space expansion--local operation--coherent compression" results in the multiqudit entangling gate,
with a success probability of 1/d, independent on n. c The simplified schematic of a two-ququart d-QPU: (I) generation of four-level entangled state in an
array of four integrated identical SFWM sources; (II) Hilbert space expansion and arbitrary single-qudit preparation of the y-register state; (III) arbitrary
single-qudit operation of the x-register state; (IV) arbitrary single-qudit operation (loading in the four layers) of the y-register state, in which the operations
are coherently entangled with the x-register state, thus forming the MVCU entangling gate, where the state-gate entanglement is indicated by the four
colourful links; (V) coherent compression of Hilbert space by an indistinguishable erasure of spatial information; (VI) and (VII) arbitrary single-qudit
projective measurement in the x and y registers. Insets: left top, measured resistance of all thermal-optic phase shifters (TOPSs); measured interference
visibility of all 2-dimensional Mach-Zehnder Interferometers (MZIs); bottom right, measured classical statistic fidelities (Fc) for the Pauli X4 gate with a
mean of 0.988(13) and Fourier F4 gate with a mean of 0.967(19). d A microscopy image of the d-QPU chip. It monolithically integrates 451 optical
components, including 4 SFWM sources, 116 reconfigurable TOPS, 131 multimode interferometer (MMI) beamsplitters, 4 wavelength-division multiplexing
(WDM) filters, 156 waveguide crossings and 40 grating couplers (GC). The d-QPU chip is wire bounded and can be flexibly controlled by classical
electronics, and can be reliably reprogrammed and reconfigured to benchmark a spectrum of different quaternary quantum algorithms.

function is chosen as constant (see Fig. 4b) and balanced (see in Fig. 4b, c, h, i are fully distinguishable. These imply an interesting
Fig. 4c–h), respectively. The d-QPU thus determines whether f is capability of computing a close expression for an affine function f:
constant or balanced, and the fidelity Fc of 0.967(2) was measured to A0 ⊕ A1x1 ⊕ . . . ⊕ Anxn. That presents the d-ary generalisation of the
quantify its success probability. Notably, the measured distributions Bernstein-Vazirani algorithm65, whose task is to compute the d-ary

4 NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x ARTICLE

Fig. 3 Characterisation of quaternary multi-value controlled-unitary logic operations. a, b Measured density matrices (ρ) for a four-level maximally
entangled Bell state and a fully product state. Column heights (colours) represent absolute values ∣ρ∣ (phases Arg(ρ)) of the elements. Quantum state
fidelity Fq was measured to be 0.983(4) and 0.953(3), respectively. c, Measured
quantum j i

 state fidelities for a complete set of four-level Bell states Ψ i;j ,
i, j = 0,1,2,3. The generalised Bell states are created by operating input states of
f i 
kj in the MVCXd gate. Shaded areas atop bars refer to ± 1σ error
bars. The Fq values in a–c were estimated by Monte Carlo the photon counts with photon Poissonian statistics. d Reconstructed process matrix (χ) of the
MVCXd gate. It was measured by quantum process tomography with in total 256 quantum state tomographic measurements. We obtained the quantum
process fidelity of 0.952, that is defined as Tr[χ0χ], where χ0 is the ideal matrix. The χ matrix is represented in the standard identity and Pauli basis {I2,
X2, Y2, Z2}. Blue and red colours are used to improve the clarity. e–j Measured truth tables (normalised photon counts) for three MVCU logic gates in two
complementary bases {I, II}: e, f, a multi-value controlled-Xd (MVCXd) gate;
g, h, a multi-value
controlled-Z
 d (MVCZ
d)gate; i,
j,a multi-value controlled-
Hd (MVCHd) gate. The





 definitions

 of basis are given as: computational basis ki ; Fourier basis f i ; Hadamard basis hi ; basis li is another eigenstate of
the Hadamard and
ai and
bi are given by rotations, which are provided in Supplementary (i = {0, 1, 2, 3}). Classical statistic fidelities (Fc1, Fc2) are
measured, which are adopted to estimate the lower and upper bound of the complementary classical fidelity: [0.891(2), 0.931(1)] for the MVCXd,
[0.912(2), 0.952(1)] for the MVCZd, and [0.865(1), 0.920(1)] for the MVCHd. In e–j, the probability distributions are colour coded (key is provided at the
right bottom). The values in parentheses of Fc and Fq refer to ± 1σ uncertainty from photon statistics.

NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x

Fig. 4 Implementations of generalised Deutsch-Jozsa and Bernstein-

Vazirani algorithms in quaternary. a Quantum logical circuit for
implementing the d-ary Deutsch-Jozsa and Bernstein-Vazirani algorithms.
This circuit can be implemented by the scheme in Fig. 1a, b with an
exchange of the x and y registers. The task of the d-ary Deutsch-Jozsa
algorithm is to determine an unknown multi-value function f: {0, 1,..., d−1}
n → {0, 1,..., d − 1} is either constant or balanced, while that of the d-ary

Bernstein-Vazirani algorithm is to compute the close expression of a multi-

value affine function f: A0 ⊕ A1x1. . . ⊕ Anxn, using only a single call of
quantum oracle. When d equals to 2, the two algorithms return to the
original Deutsch’s algorithms. The key part is the implementation of
f(x) ⊕ dy by the MVCU gate. The outcome of the algorithms is measured in
the computation basis of the x-register states. b–i Measured probability
distributions (normalised coincidence counts) of the x-register in the
computational basis. Results in b–h demonstrate that the d-ary Deutsch-
Jozsa algorithm allows the determination of whether f(x) is constant (b) or
balanced (c–h). Results in b, c, i, h show the d-ary Bernstein-Vazirani
algorithm can determine the expression of affine functions f: b, f(x) is
constant and A1=0; c, f(x) is affine and A1=1; i, f(x) is affine and A1=2;
h, f(x) is affine and A1=3; Dotted boxes in (b--i) refer to theoretical
probability distributions. Experimental probability distributions (coloured
bars) are obtained from photon coincidences, which are accumulated by
20s per measurement. The classical fidelity Fc presents the success
probability of each measurement. In order to make the small error bars
visible in the plots, they are plot by ± 3σ. The values in parentheses refer
to ± 1σ uncertainty. All error bars are estimated from photon Poissonian
statistics.

a single qubit in the auxiliary x-register, but at the expense of

repeating m-sequences of single-qubit measurement and single-
qubit feedforwarded operation, see quantum circuits in Fig. 5a. In
Kitaev’s phase estimation and order-finding algorithms, the
computational capacity is determined by the number of n-qubits
in the y-register, and the computational accuracy is determined
by the number of m-sequences in the x-register. In this respect,
one can see processing quantum algorithms with qudits results in
nontrivial advantages: a log2(d) larger computational capacity,
and log2(d) higher computational accuracy or log2(d)-less com-
putational steps to achieving the same precision, as shown in
Supplementary Fig. 3b, which could be important to quantum
computers with limited coherence time.
In the quantum phase estimation,
 i2πϕwe

 aim to compute the
eigenphase ϕ of an unitary as O
ψ =e
ψ , given the eigenstate of



ψ . The eigenphase of ϕ can be described in d-ary as 0. ϕ1ϕ2…
ϕm−1ϕm, where m denotes iterative steps determining the approx-
imation accuracy26,29, and each dit of the phase is in
[0, 1,..., d − 1]67,69,73. We take the s-th step as an example (see
coefficients

Ai. The output state
of the x register  can be derived as quantum
circuit in Fig. 5a). We prepare an input state of


ωA0
A1 ; A2 ; :::; An , where the
A1 ; A2 ; :::; An state can be directly p1ffiffi ∑d1
j
ψ and perform the MVCU gate, that results in a state
d j¼0


read out in its computational basis (A0 is lost as a global phase). d1 ij2πð0:ϕs ϕsþ1 :::ϕm Þ


of p1ffiffi ∑j¼0 e j ψ . Then, the x-register qudit state is
From the experimental results in Fig. 4b, c, h, i one can d
therefore determine the multi-value function with A1= {0, 1, 2, 3}, feed-forwardly rotated around the Pauli Zd basis as
respectively, by a single query of the oracle. Details of the diag[1; ei2πθs ; :::; ei2πðd1Þθs ], where the rotation angle θs of − 0.0ϕs
generalisation of the Deutsch’s algorithms are provided in +1ϕs+2…ϕm is given by previous measurement outcomes. Remark-
Supplementary Note 4. ably, implementing

 an inverse Fd in the x-register returns an output
state as
ϕs ¼ jsi (see Supplementary Note 5). Measuring the x-
Benchmarking of d-ary phase estimation and order finding. register in the computational basis of jsi therefore allows the
Quantum phase estimation and order finding are two of the most extraction of the s-th dit of the dit expansion. The algorithm
featured quantum Fourier transform ralgorithms, that are iteratively computes all m dits of the eigenphase backwardly, in
essential to molecular simulation66 and fast factorisation25. which, notably, each dit is once estimated with d-ary accuracy.
Kitaev’s scalable implementation of both algorithms (in Figure 5b–d report measured eigenphases of 4-dimensional unitary
binary)26–29 has been reported in several leading quantum matrices by quaternary phase estimation. We estimated the four
platforms67–72. The remarkable idea is to replace the 2n qubits by eigenphases for three logic gates, i.e., a phase gate Z4, a Fourier gate F4

6 NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x ARTICLE

Fig. 5 Implementations of quaternary quantum phase estimation and order-finding. a Their quantum logical circuit for
implementing Kitaev’s scalable

approaches. For the d-ary phase estimation, the task is to compute the eigenphase ϕ of a unitary O given its eigenstate of
ϕ . For the d-ary order-finding,

 is to find the order of a function as (a modN) = 1. The x-register single-qudit
the task r state is initialised by the Fourier gate Fd; the y-register is prepared in
the
ϕ eigenstate (j0in y
d state) for phase estimation (for order-finding). The F d terminates the x-register to output the desired solution in the
computational basis. In the s-th step, the Zd rotation is added with a feedback angle of θs= − 0.0ϕs+1ϕs+2…ϕm, that is determined by previous
measurements. The algorithm is iterated m times -- each step returns 1 dit result with d-ary accuracy, to obtain a m-dit estimation of the eigenphase of a
unitary or the order of a function. b–d Measured probability pie-distributions of the four eigenphases (ϕ1, ϕ2, ϕ3, ϕ4) for three different unitary matrices,
using the quaternary quantum phase estimation: b a generalised phase gate Z4 as diag[1; ei2πθ ; ei2π2θ ; ei2π3θ ] where θ = 1/4; c a generalised Fourier gate F4;
d a random gate Urandom (see form in Supplementary Note 5). Coloured sectors represent the experimental outcomes of {0,1,2,3} for each iteration,
measured in the computational basis of {j0i,j1i,j2i,j3i}, respectively. The measured dominating sector is used to obtain every dit of the eigenphases;
theoretical values for each dit are provided under the pies. The eigenphases are backwardly computed from the least significant dit from m = 12 to 1.
e, f Measured probability distributions for the quaternary order-finding algorithm with a setting of a = 4 and a = 2, respectively. From the distributions, the
order of r = 2 and r = 4 are experimentally computed with a 3-quart resolution (equivalent to 64-level), and with a classical statistic fidelity (Fc) of
0.909(9) and 0.922(9), respectively. The order-finding together with classical algorithm allows the factorisation of 15 = 3 × 5. Errors ( ± 1σ) arising from
photon Poissonian noise are indicated as red shaded caps. Dashed lines refer to theoretical predictions. Experimental probability distributions in b–f are
calculated from photon coincidences, which are accumulated by 20s per measurement.

and a randomised gate Urandom (see their explicit forms in computational accuracy of ± 4−12 requires a number of 24
Supplementary). Each pie chart presents one dit measurement computational interactions. And the achieved computational accura-
outcomes, and the area of each coloured sector denotes measured cies of 12 quarts are sufficient for the calculation of molecular eigen-
probability distributions in the computational basis of energies67,69. In Fig. 5, it shows experimental data are in good
{j0i,j1i,j2i,j3i}, respectively. In Fig. 5b, c, the eigenphases of Z4 agreement with theoretical predictions (indicated under each pie).
and F4 gates can be described by a single dit. Figure 5d shows the The task of quantum factoring is to efficiently compute the
computed eigenphases of the Urandom gate with an accuracy of 12 prime factors p and q from an integer N25. It can be reduced to
dits, by running the algorithm with a number of 12 interactions on the task of finding the order r of a module N, i.e., by computing a
the d-QPU. Instead, in the qubit-based device, achieving the same function f: armodN = 1 (a is a co-prime of N), and with a high

NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x

probability it returns a factor as gcd(ar/2 ± 1, N), where gcd(α, β) qudit. Regarding the dimension of units, though the ququart states
refers to the greatest common divisor of α and β. As the order- are implemented as an example in this work, it is straightforward
finding is just the phase estimation of a unitary having the to extend to a larger-d device36, which can be fabricated using
eigenphases of s/r, s ∈ [0, r − 1], one can directly adopt the d-ary the same CMOS fabrication techniques. Remarkably, this
phase estimation to determine the order of r in the d-ary format. entanglement-assisted d-QPU scheme works with a success
It can be considered as a generalisation of d-ary order-finding by probability of 1/d regardless of n (Supplementary Note 3). The
adopting Kiteav’s iterative techniques70–72 (see details in scaling of d-QPU therefore strongly relies on the generation of the
Supplementary Note 6). We then reprogrammed the d-QPU to qudit GHZ entangled states. Combing the state-of-the-art tech-
implement the order-finding in quaternary. The a ∈ [0, r − 1] nologies, including the techniques of generating multi-photon
satisfying gcd(a, 15) = 1 is randomly chosen. We chose a = 4 and qudit GHZ states45,46, on-chip high-fidelity control of qudit
2 as examples, and set the unitary of the MVCU gate as states36, high-quality photon-pair sources75,76, low-loss fibre-chip
{Id, Xd, Id, Xd} and fI d ; X d ; X 2d ; X 3d g, respectively, where Id is the d- interface75,77, and large-scale quantum integration57, we estimate
mode identity. In our experiment, the order-finding algorithm a 10-photons d-QPU is achievable in near term. Its further scaling
was iteratively implemented by three steps, and each step returns requires high-efficiency heralded multiplexing photon-pairs
quaternary outcomes in the computational basis, resulting in the sources78 and multiplexing qudit GHZ generators31. That being
3-quart (64-level) computational accuracy of the s/r eigenphase. said, given the efficient generation of the multi-photon multi-
Figure 5e, f show the measured output probabilities
 of the x- qudit GHZ states, the d-QPU scheme is scalable. Calculations and
register in the computational basis of
ijk , i, j, k = 0, 1, 2, 3. analysis are provided in Supplementary Note 9 and Supplemen-
Classical statistic fidelities Fc of 0.909(9) and 0.922(9) were tary Fig. 3d. Moreover, when scaling up the d-QPU, an interesting
obtained in comparison with theoretical distributions, showing concern is the required resources, in particular the number of
successful estimations of the order of r = 2 (Fig. 5e) and r = 4 classical controls. As an example, let us consider a processor with
(Fig. 5f), respectively. The d-QPU thus finds the order with one qudit in the auxiliary register and n qudits in the data register
double-enhanced computational accuracy; alternatively speaking, (see Fig. 2a). It requires a number of (n + 1) single-qudit gen-
it executes the task twice faster than a qubit-QPU, given the same erators for state preparation, (nd) local single-qudit operators for
estimation precision. The order-finding algorithm together with multi-qudit MVCU operation, and (dn + 1) single-qudit projec-
classical processing using the continued fraction algorithm tors. The physical resources, i.e, the number of phase-shifters,
returns the factor of gcd(ar/2 ± 1, N) = (3, 5). Implementing the scale with (d2 − d) for the qudit operators50,59, and 2(d − 1) for
d-ary algorithms in the d-QPU can therefore find the order of a the qudit generators and projectors36, as shown in Supplementary
function and compute the eigenphase of a unitary, with a log2(d)- Note 9 and Supplementary Fig. 2. Importantly, the required
faster computational speed. resources for classical controls scale polynomially with the number
of particles. In Supplementary Fig. 3c, it is shown that thousands
of phase-shifters are required for a 10-photon d-QPU. This large
Discussion amount of phase-shifters can be individually addressed and con-
We have reported a proof-of-principle experimental demonstration trolled, by using a co-integration technology of photonic and
of a programmable qudit-based quantum processor in photonic electronic circuits in silicon.
integrated circuits, and implementations of several generalised d-ary The highly flexible and reliable programmability of the qudit
quantum Fourier transform algorithms in the d-QPU chip. In processor, that is enabled by technological advances in a mono-
agreement with the references17–19,24,43–46, our experimental results lithic integration of all key functionalities and capabilities in a
show that qudit-based quantum computation with integrated pho- silicon chip, has allowed the implementations of more than one
tonics can enhance quantum parallelism in terms of the computa- million qudit generators, operators and projectors (see Supple-
tional capacity, accuracy and efficiency, in comparison with its qubit- mentary Note 8), and also the benchmarking of different gen-
based quantum computing counterpart. The computational capacity eralised quantum algorithms. Such programmability can
of the two ququart quantum processor is equivalent to that of a four- transition the advanced technologies in controlling qudit states
qubit processor, thus allowing the implementations of the Deutsch’s and gates36–48 to algorithm implementations, playing an enabling
algorithms for a function with longer-string. Keeping the same role in the roadmap of qudit-based quantum computations. The
number of photons n but encoding each qudit in a dimension d, not full chip-scale integration technologies also perfectly match the
only gives a larger Hilbert space74, but also significantly improve the top-down hierarchy of quantum computing, in which users can
detection rate of photons43,44. We obtained the detection rate of define and execute multiple quantum tasks by recompiling the
about 6 orders brighter than that in another device with the same software and reprogramming the quantum hardware. In general,
Hilbert space58. More analysis is provided in Supplementary Fig. 3. the programmable qudit-based quantum devices can find appli-
Moreover, multiple parallel evaluations of the function and multiple cations in noise-resilient quantum network9,10, quantum simu-
path interference in the d-ary quantum Fourier gate, allow the lation of complex chemical and physical systems12–15, and
enhancement of the computational efficiency and speed up of the universal quantum computing with qudit cluster states19–21.
determination of desired solutions. In the implementations of
Kitaev’s phase estimation and factorisation, a number of log2(d)-less Data availability
iterations are needed in the qudit processor, i.e., a log2(d) times speed The data that support the plots within this paper and other findings of this study are
up of quantum computation, compared with the qubit ones, given available from the corresponding author upon reasonable request.
the same computational accuracy (see Supplementary Fig. 3b).
As the multi-value quantum controlled gates are the result of Code availability
the entanglement in the generation stage and the gates are instead The codes used for the analysis included in the current study are available from the
corresponding authors upon reasonable request.
local operations that steer the state to collapse in the desired
outputs, our scheme can be straightforwardly generalised to multi-
qudit quantum computiation. Its scalability is naively dependent Received: 16 December 2021; Accepted: 11 February 2022;
on the number (n) of qudits and the dimensionality (d) of each

8 NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x ARTICLE

References 36. Wang, J. et al. Multidimensional quantum entanglement with large-scale

1. Giustina, M. et al. Significant-loophole-free test of bell’s theorem with integrated optics. Science 360, 285–291 (2018).
entangled photons. Phys. Rev. Lett. 115, 250401 (2015). 37. Li, L. et al. Metalens-array–based high-dimensional and multiphoton
2. Shalm, L. K. et al. Strong loophole-free test of local realism. Phys. Rev. Lett. quantum source. Science 368, 1487–1490 (2020).
115, 250402 (2015). 38. Kues, M. et al. Quantum optical microcombs. Nat. Photon. 13, 170–179 (2019).
3. Liao, S. K. et al. Satellite-to-ground quantum key distribution. Nature 549, 39. Hu, X. M. et al. Efficient generation of high-dimensional entanglement
43–47 (2017). through multipath down-conversion. Phys. Rev. Lett. 125, 090503 (2020).
4. Liao, S. K. et al. Satellite-relayed intercontinental quantum network. Phys. Rev. 40. Dada, A. C. et al. Experimental high-dimensional two-photon entanglement
Lett. 120, 030501 (2018). and violations of generalized Bell inequalities. Nat. Phys. 7, 677–680 (2011).
5. Arute, F. et al. Quantum supremacy using a programmable superconducting 41. Feng, L. F. et al. On-chip coherent conversion of photonic quantum entanglement
processor. Nature 574, 505–510 (2019). between different degrees of freedom. Nat. Commun. 7, 11985 (2016).
6. Wu, Y. et al. Strong quantum computational advantage using a 42. Martin, A. et al. Quantifying photonic high-dimensional entanglement. Phys.
superconducting quantum processor. Phys. Rev. Lett. 127, 180501 (2021). Rev. Lett. 118, 110501 (2017).
7. Collins, D. et al. Bell inequalities for arbitrarily high-dimensional systems. 43. Reimer, C. et al. High-dimensional one-way quantum processing
Phys. Rev. Lett. 88, 040404 (2002). implemented on d-level cluster states. Nat. Phys. 15, 148–153 (2019).
8. Vértesi, T., Pironio, S. & Brunner, N. Closing the detection loophole in bell 44. Imany, P. et al. High-dimensional optical quantum logic in large operational
experiments using qudits. Phys. Rev. Lett. 104, 060401 (2010). spaces. npj Quant. Inf. 5, 59 (2019).
9. Cerf, N. J. et al. Security of quantum key distribution using d-level systems. 45. Malik, M. et al. Multi-photon entanglement in high dimensions. Nat. Photon.
Phys. Rev. Lett. 88, 127902 (2002). 10, 248–252 (2016).
10. Bouchard, F. et al. High-dimensional quantum cloning and applications to 46. Erhard, M. et al. Experimental greenberger–horne–zeilinger entanglement
quantum hacking. Sci. Adv. 3, e1601915 (2017). beyond qubits. Nat. Photon. 12, 759–764 (2018).
11. Islam, N. T. et al. Provably secure and high-rate quantum key distribution 47. Hu, X. M. et al. Experimental high-dimensional quantum teleportation. Phys.
with time-bin qudits. Sci. Adv. 3, e1701491 (2017). Rev. Lett. 125, 230501 (2020).
12. Kaltenbaek, R. et al. Optical one-way quantum computing with a simulated 48. Luo, Y. H. et al. Quantum teleportation in high dimensions. Phys. Rev. Lett.
valence-bond solid. Nat. Phys. 6, 850–854 (2010). 123, 070505 (2019).
13. Neeley, M. et al. Emulation of a quantum spin with a superconducting phase 49. Zhong, H. S. et al. Phase-programmable gaussian boson sampling using
qudit. Science 325, 722–725 (2009). stimulated squeezed light. Phys. Rev. Lett. 127, 180502 (2021).
14. Blok, M. S. et al. Quantum information scrambling on a superconducting 50. Carolan, J. et al. Universal linear optics. Science 349, 711–716 (2015).
qutrit processor. Phys. Rev. X. 11, 021010 (2021). 51. Debnath, S. et al. Demonstration of a small programmable quantum computer
15. Choi, S. et al. Observation of discrete time-crystalline order in a disordered with atomic qubits. Nature 536, 63–66 (2016).
dipolar many-body system. Nature 543, 221–225 (2017). 52. Monroe, C. et al. Programmable quantum simulations of spin systems with
16. Wei, S. & Long, G. Duality quantum computer and the efficient quantum trapped ions. Rev. Mod. Phys. 93, 025001 (2021).
simulations. Quant. Inf. Process. 15, 1189–1212 (2016). 53. Watson, T. F. et al. A programmable two-qubit quantum processor in silicon.
17. Luo, M. & Wang, X. Universal quantum computation with qudits. Sci. China Nature 555, 633–637 (2018).
Phys. Mech. 57, 1712–1717 (2014). 54. Wang, J. et al. Integrated photonic quantum technologies. Nat. Photon. 14,
18. Zhou, D. L. et al. Quantum computation based on d-level cluster state. Phys. 273–284 (2020).
Rev. A. 68, 062303 (2003). 55. Pelucchi, E. et al. The potential and global outlook of integrated photonics for
19. Wei, T. C., Affleck, I. & Raussendorf, R. Affleck-kennedy-lieb-tasaki state on a quantum technologies. Nat. Rev. Phys. https://doi.org/10.1038/s42254-021-
honeycomb lattice is a universal quantum computational resource. Phys. Rev. 00398-z (2021).
Lett. 106, 070501 (2011). 56. Wang, J. et al. Experimental quantum Hamiltonian learning. Nat. Phys. 13,
20. Campbell, E. T. Enhanced fault-tolerant quantum computing in d-level 551–555 (2017).
systems. Phys. Rev. Lett. 113, 230501 (2014). 57. Chen, X. et al. A generalized multipath delayed-choice experiment on a large-
21. Bocharov, A., Roetteler, M. & Svore, K. M. Factoring with qutrits: Shor’s scale quantum nanophotonic chip. Nat. Commun. 12, 2712 (2021).
algorithm on ternary and metaplectic quantum architectures. Phys. Rev. A. 96, 58. Adcock, J. C. et al. Programmable four-photon graph states on a silicon chip.
012306 (2017). Nat. Commun. 10, 3528 (2019).
22. Gokhale, P. et al. Asymptotic improvements to quantum circuits via qutrits. In 59. Clements, W. R. et al. Optimal design for universal multiport interferometers.
Proceedings of the 46th International Symposium on Computer Architecture, Optica 3, 1460–1465 (2016).
554-566 (2019). 60. Riofrío, C. A. et al. Experimental quantum compressed sensing for a seven-
23. Wang, D. S., Stephen, D. T. & Raussendorf, R. Qudit quantum computation qubit system. Nat. Commun. 8, 15305 (2017).
on matrix product states with global symmetry. Phys. Rev. A. 95, 032312 (2017). 61. Mohseni, M., Rezakhani, A. T. & Lidar, D. A. Quantum-process tomography:
24. Wang, Y. et al. Qudits and high-dimensional quantum computing. Front. resource analysis of different strategies. Phys. Rev. A. 77, 032322 (2008).
Phys. 8, 479 (2020). 62. Hofmann, H. F. Complementary classical fidelities as an efficient criterion for
25. Shor, P. W. Algorithms for quantum computation: Discrete logarithms and the evaluation of experimentally realized quantum operations. Phys. Rev. Lett.
factoring. Proc. 35nd Annual Symposium on Foundations of Computer Science, 94, 160504 (2005).
IEEE Computer Society Press (1994). 63. Fan, Y. A generalization of the deutsch-jozsa algorithm to multi-valued
26. Kitaev., A. Y. Quantum measurements and the abelian stabilizer problem. quantum logic. In 37th International Symposium on Multiple-Valued Logic,
Electronic Colloquium on Computational Complexity. 3 (1996). 12–12 (IEEE, 2007).
27. Griffiths, R. B. & Niu, C. S. Semiclassical fourier transform for quantum 64. Deutsch, D. & Jozsa, R. Rapid solution of problems by quantum computation.
computation. Phys. Rev. Lett. 76, 3228–3231 (1996). Proc. R. Soc. Lond. Ser. A Math. Phys. Sci. 439, 553–558 (1992).
28. Parker, S. & Plenio, M. B. Efficient factorization with a single pure qubit and 65. Bernstein, E. & Vazirani, U. Quantum complexity theory. SIAM J. Comput.
log N mixed qubits. Phys. Rev. Lett. 85, 3049–3052 (2000). 26, 1411–1473 (1997).
29. Dobšíček, M. et al. Arbitrary accuracy iterative quantum phase estimation 66. Aspuru-Guzik, A. et al. Simulated quantum computation of molecular
algorithm using a single ancillary qubit: A two-qubit benchmark. Phys. Rev. A. energies. Science 309, 1704–1707 (2005).
76, 030306 (2007). 67. Lanyon, B. P. et al. Towards quantum chemistry on a quantum computer. Nat.
30. Adcock, M. R. A., Høyer, P. & Sanders, B. C. Quantum computation with Chem. 2, 106–111 (2010).
coherent spin states and the close hadamard problem. Quant. Inf. Process. 15, 68. Zhou, X. Q. et al. Calculating unknown eigenvalues with a quantum
1361–1386 (2016). algorithm. Nat. Photon. 7, 223–228 (2013).
31. Paesani, S. et al. Scheme for universal high-dimensional quantum 69. Santagati, R. et al. Witnessing eigenstates for quantum simulation of
computation with linear optics. Phys. Rev. Lett. 126, 230504 (2021). hamiltonian spectra. Sci. Adv. 4, aap9646 (2018).
32. Erhard, M., Krenn, M. & Zeilinger, A. Advances in high-dimensional 70. Smolin, J. A., Smith, G. & Vargo, A. Oversimplifying quantum factoring.
quantum entanglement. Nat. Rev. Phys. 2, 365–381 (2020). Nature 499, 163–165 (2013).
33. Ringbauer, M. et al. A universal qudit quantum processor with trapped ions. 71. Martin-Lopez, E. et al. Experimental realization of shor’s quantum factoring
arXiv:2109.06903 [quant-ph] (2021). algorithm using qubit recycling. Nat. Photon. 6, 773 (2012).
34. Cervera-Lierta, A. et al. Experimental high-dimensional greenberger-horne- 72. Monz, T. et al. Realization of a scalable Shor algorithm. Science 351,
zeilinger entanglement with superconducting transmon qutrits. 1068–1070 (2016).
arXiv:2104.05627 [quant-ph] (2021). 73. Lu, H. H. et al. Quantum phase estimation with time-frequency qudits in a
35. Chen, X. et al. Quantum entanglement on photonic chips: a review. Adv. single photon. Adv. Quant. Technol. 3, 1900074 (2020).
Photon. 3, 064002 (2021).

NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-28767-x

74. Vigliar, C. et al. Error-protected qubits in a silicon photonic chip. Nat. Phys. Additional information
17, 1137–1143 (2021). Supplementary information The online version contains supplementary material
75. Llewellyn, D. et al. Chip-to-chip quantum teleportation and multi-photon available at https://doi.org/10.1038/s41467-022-28767-x.
entanglement in silicon. Nat. Phys. 16, 148–153 (2020).
76. Paesani, S. et al. Near-ideal spontaneous photon sources in silicon quantum Correspondence and requests for materials should be addressed to Jianwei Wang.
photonics. Nat. Commun. 11, 2505 (2020).
77. Paesani, S. et al. Generation and sampling of quantum states of light in a Peer review information Nature Communications thanks Jacqui Romero and Taira
silicon chip. Nat. Phys. 15, 925–929 (2019). Giordani for their contribution to the peer review of this work. Peer reviewer reports are
78. Kaneda, F. & Kwiat, P. G. High-efficiency single-photon generation via large- available.
scale active time multiplexing. Sci. Adv. 5, eaaw8586 (2019).
Reprints and permission information is available at http://www.nature.com/reprints

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
Acknowledgements published maps and institutional affiliations.
We acknowledge X.Wang and S.Tao for useful discussions and assistance of experiment.
We acknowledge support from Beijing Natural Science Foundation (Z190005), the
National Key R&D Program of China (nos 2019YFA0308702, 2018YFB1107205,
2018YFB2200403, and 2018YFA0704404), the National Natural Science Foundation of Open Access This article is licensed under a Creative Commons
China (nos 61975001, 61590933, 61904196, 61675007, and 61775003), and Key R&D Attribution 4.0 International License, which permits use, sharing,
Program of Guangdong Province (2018B030329001). adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
Author contributions material in this article are included in the article’s Creative Commons license, unless
J.W. conceived the project. Y.C., J.H., Z.C.Z., J.M., Z.N.Z, X.C., C.Z., J.B., T.D., H.Y., indicated otherwise in a credit line to the material. If material is not included in the
M.Z., B.T., and Y.Y. implemented the experiment. Y.C., X.C., and J.B. designed the article’s Creative Commons license and your intended use is not permitted by statutory
device. Y.C., J.H., Z.C.Z., J.M. and Z.N.Z provided theoretical analysis. D.D., Z.L., Y.D., regulation or exceeds the permitted use, you will need to obtain permission directly from
L.K.O., M.G.T., J.L.O., Y.L., Q.G., and J.W. managed the project. All authors discussed the copyright holder. To view a copy of this license, visit http://creativecommons.org/
the results and contributed to the manuscript. licenses/by/4.0/.

Competing interests © The Author(s) 2022

The authors declare no competing interests.

10 NATURE COMMUNICATIONS | (2022)13:1166 | https://doi.org/10.1038/s41467-022-28767-x | www.nature.com/naturecommunications