Abstract
Integrated entangled photon pair source is an essential resource for both fundamental investigations and practical applications of quantum information science. Currently there have been several types of entanglement, among which the transversemode entanglement is becoming attractive because of its unique advantages. Here, we report an onchip transversemode entangled photon pair source via the spontaneous fourwave mixing processes in a multimode silicon waveguide. Transversemode photon pairs are verified over multiple frequency channels within a bandwidth of ~2 THz, and a maximally entangled Bell state is also produced with a net fidelity of 0.96 ± 0.01. Our entangled photon pair source is the key element for quantum photonics based on transversemode, and also has the possibility to extend to higherdimensional Hilbert space. Furthermore, the transversemode entanglement can be converted coherently to path and polarization entanglement, which paves the way to realizing highly complex quantum photonic circuits with multiple degrees of freedom.
Introduction
Quantum photon pair sources are not only critical to advancing our fundamental understanding of quantum mechanics, but also play a key role in many applications for quantum technologies.^{1} Compared with freespace ones, integrated photon pair sources have attracted much attention owing to their compactness, scalability, and stability.^{2} For example, path or polarization entangled photon pairs were demonstrated by using the nonlinear processes in a singlemode waveguide or in a microring cavity.^{3,4,5,6,7} Aside from these degrees of freedom, transversemode in a multimode optical waveguide becomes more and more attractive, and shows advantages in modulation of wavelength, dispersion and polarization compared to singlemode ones.^{8,9,10,11} It is useful for parallel encoding and enables information processing in a more compact way for nextgeneration optical communications.^{12,13,14} On the other hand, via the nonlinear optical medium, photons in different orthogonal transversemodes can interact with each other, which is useful for wavelength conversion and opens the door to many more nonlinear phenomena.^{15,16,17,18,19}
Recently, transversemode has been applied for quantum photonics, for example, implementing quantum interference in one multimode waveguide and onchip entanglement conversion between path, transversemode and polarization.^{20,21,22} Also several works have demonstrated the photon pair generation with multiple modes in few mode fiber,^{23} as well as in periodically poled KTiOPO_{4} multimode waveguide.^{24} These transversemode photon pairs in one multimode waveguide can be extended to higherdimensional entangled states,^{25} and therefore benefit highlyefficient quantum logic gates and noiseresilient communications.^{26,27} Besides, transversemode entanglement can be converted coherently to path and polarization entanglement,^{21} which showed the possibility to control multiple degrees of freedom of photons simultaneously and to realize chipscale hyperentangled states.^{28} So far, onchip transversemode entangled photon pair source has not yet been demonstrated.
Here, we study the spontaneous fourwave mixing (SFWM) processes in a multimode silicon nanophotonic waveguide thoroughly, and realize for the first time an onchip transversemode entangled photon pair source. Transversemode entangled photon pairs are generated and experimentally verified with a bandwidth of ~2 THz. By adjusting the energy ratio and phase difference between different transversemode terms of the pump light, a maximally transversemode entangled Bell state is produced with a net fidelity of 0.96 ± 0.01. Our protocol can definitely be generalized to cases of higherdimensional entangled photon pair sources with more transversemodes, thus showing potential for realizing complex quantum states with high dimensionality and multiple degrees of freedom.
Results
Entangled photon pair source chip
As shown in Fig. 1a, the device includes three parts: part I, the mode modulator; part II, the multimode waveguide; and part III, the output analyzer. The mode modulator is used to modulate the pump power ratio for different transversemodes, and the output state analyzer is used to export the state for measurement. The multimode waveguide is 3 mm long, with a crosssection of ~760 × 220 nm^{2}, so that it can support two transverseelectric (TE) modes, i.e., TE_{0} and TE_{1}. The pump light is coupled to the device through a grating coupler and then divided into two parts through a directional coupler, with one part being converted to the transversemode TE_{0} and the other to the transversemode TE_{1} through a mode multiplexer.^{12} After nonlinear interactions in the multimode waveguide, a mode demultiplexer is used to separate the generated photons into different paths, which are then coupled out using two identical grating couplers for further measurement.
The SFWM processes in our device can be divided into three types (see Supplementary Materials, Section 1). For Type I, the four photons involved are in the same transversemode. For Type II, the two pump photons are in the same transversemode, while the signal and idler photons are in the other transversemode. For Type III, the two pump photons are in different transversemodes (i.e., one in TE_{0} and the other in TE_{1}), and the signal and idler photons are also in different transversemodes. The SFWM process of Type I is intramodal, while the SFWM processes of Type II and Type III are intermodal. Here, the phase mismatching for the Type II process is so large that it hardly occurs. Therefore, we focus on Type I and Type III processes, where the signal and idler photon pairs have four combinations of transversemodes, i.e., \(\left {{\mathrm{TE}}_0{\mathrm{TE}}_0} \right\rangle\), \(\left {{\mathrm{TE}}_1{\mathrm{TE}}_1} \right\rangle\), \(\left {{\mathrm{TE}}_0{\mathrm{TE}}_1} \right\rangle\) and \(\left {{\mathrm{TE}}_1{\mathrm{TE}}_0} \right\rangle\) (Fig. 1b). We calculated the firstorder dispersions β_{1} and secondorder dispersions β_{2} of the multimode waveguide for the TE_{0} and TE_{1} modes, respectively (Fig. 1c). The signal gain \(\hat S\) of the generated photon pairs for those four transversemode combinations is shown in Fig. 1d. All data are normalized to the maximum value. The signal and idler photons are time correlated, and their frequencies are equally separated from the central pump frequency.
Photon pair source verification
According to the size of waveguide at different parts, we calculated the SFWM efficiencies within parts I, II and III, and concluded that the detected photon pair counts mainly came from the multimode waveguide part (see Materials and Methods). To ascertain the generated quantum states, we measured the photon pair coincidence for different combinations of the frequency and transversemode. The experimental setup is shown in Fig. 2a, an amplified continuouswave (CW) laser (central wavelength 1550.11 nm) was coupled to the device via a grating coupler by a singlemode fiber array after passing through two cascaded 100 GHz bandwidth prefilters. At the output end, the pump light was blocked by two cascaded offchip 200 GHz bandwidth postfilters. Two 40channel densewavedivisionmultiplexers (DWDMs) were used to separate the signal and idler photons. Each channel has a 100 GHz bandwidth, such that we can select any frequency and transversemode combinations for photon pairs with a frequency detuning of ~2 THz from the central pump frequency (see Supplementary Materials, Table S3). The generated photon pairs were finally detected by two superconducting nanowire single photon detectors (SCONTEL, dark count rate 100 Hz, detection efficiency 85%).
We measured single photon counts in different frequency channels, as shown in Fig. 2b, c. The differences in single photon count arise from the different gains for the SFWM processes and the uneven Raman scattering and losses for different DWDM channels. To further characterize the photon source in the multimode waveguide, we also measured the coincidence counts for the processes of Type I (Fig. 2d, e) and Type III (Fig. 2f, g) processes. All coincidences are the raw data without subtracting the background and accident counts. As seen, the coincidences for the Type I processes are almost constant but accompanied by oscillation and the coincidences for the Type III processes decrease as the frequency detuning increases, which agrees well with the theoretical calculations (curves in Fig. 2d–g). The minor discrepancy may originate from the correlation between intramodal and intermodal SFWM processes and the difference in propagation loss of different transversemodes that we ignored in theoretical calculation. Note that the photon pairs were collected with a longer integral time for the Type III process due to its lower collection efficiency.
Coincidencetoaccidental ratio (CAR) and source brightness are two key parameters for characterizing the photon pair source. We measured CARs for correlated photon pairs in different transversemode combinations (see Supplementary Materials, Fig. S1). Intramodal (Type I) photon pairs show high CARs (generally higher than 200 for \(\left {{\mathrm{TE}}_0{\mathrm{TE}}_0} \right\rangle\) and 100 for \(\left {{\mathrm{TE}}_1{\mathrm{TE}}_1} \right\rangle\)), while the CARs for the intermodal photon pairs (Type III) are not high (lower than 10 in most cases) because of the large phase mismatching. It is possible to obtain higher CARs by broadening the frequency spectra with a wider multimode waveguide, benefiting from a smaller difference in the firstorder dispersion between signal and idler photons.
In a multimode waveguide, the lower energy density enables a better tolerance to nonlinear noise; thus, one can simultaneously achieve high brightness and CARs by simply increasing the input pump power. In the experiment, we measured the intramodal photon pair \(\left {{\mathrm{TE}}_0{\mathrm{TE}}_0} \right\rangle\) \(\left( {\left {{\mathrm{TE}}_1{\mathrm{TE}}_1} \right\rangle } \right)\) for channel ±8 with a calculated generation rate ranging from 19 kHz (5.5 kHz) to 530 kHz (180 kHz). The CARs did not suffer severe reduction even at the highest generation rates (see Supplementary Materials, Fig. S2). It also applies for the intermodal photon pair \(\left {{\mathrm{TE}}_0{\mathrm{TE}}_1} \right\rangle\) \(\left( {\left {{\mathrm{TE}}_1{\mathrm{TE}}_0} \right\rangle } \right)\) for channel ±14, with a calculated generation rate ranging from 2.7 kHz (2.9 kHz) to 54 kHz (53 kHz) (see Supplementary Materials, Fig. S3). These results show that the present device provides a flexible platform for quantum information processing. It is worth pointing out that our multimode photon pair source could also be used as heralded single photon sources. Secondorder correlation measurements show g^{(2)}(0) = 0.13 ± 0.02 for photon pair \(\left {{\mathrm{TE}}_0{\mathrm{TE}}_0} \right\rangle\) and g^{(2)}(0) = 0.19 ± 0.06 for photon pair \(\left {{\mathrm{TE}}_1{\mathrm{TE}}_1} \right\rangle\), respectively. More details are provided in the Supplementary Materials (Section 3 and Fig. S4).
In fact, silicon platform has been used for photon source generation with various structures.^{29,30,31} A brief summary is provided in Supplementary Materials (Table S4). Compared to those single mode ones, the multimode waveguides have similar performance and could be further improved. Additionally, we also calculated the nonlinear efficiencies in a multimode waveguide as a function of the waveguide width (see Supplementary Materials, Fig. S5). For a waveguide with larger size, although the nonlinear efficiencies are decreased, a series of SFWM processes involving higherorder modes occur, and provides extra dimensions in quantum state generation.
Entanglement verification and Bell state generation
After demonstrated the photon pair source generation, we then went further to demonstrate the generation of quantum entanglement, which lies at the heart of quantum information studies.^{1} As described above, the quantum state in the multimode waveguide is given as
where η_{1}, η_{2} and η_{3} are relative proportions of different components; δ_{1}, δ_{2} and δ_{3} are the phase differences, which come from the mode dispersions in the multimode waveguide. By using a twodimensional (2D) grating coupler,^{32,33} the transversemode entangled state is coherently converted into the polarization entangled state, which is expressed as \(\left {\mathrm{\Phi }} \right\rangle = \frac{1}{{\sqrt {1 + \eta _1^2 + \eta _2^2 + \eta _3^2} }}\) \(\left( {\left {HH} \right\rangle + \eta _1e^{i\delta _1}\left {HV} \right\rangle + \eta _2e^{i\delta _2}\left {VH} \right\rangle + \eta _3e^{i\delta _3}\left {VV} \right\rangle } \right)\).
According to Figs. 1d and 2d–g, when the frequencies of the signal and idler photons are close to the pump frequency, all four transversemode combinations are generated. As an example, we performed the quantum state tomography measurement with the experimental setup shown in Fig. 3a for the photon pairs in the frequency channels ±2. Using {H, V, D, R}^{⊗2} as the measuring basis, total 16 combinations are required to reconstruct the state density matrix, where \(\left H \right\rangle = (1,0)^T\), \(\left V \right\rangle = (0,1)^T\), \(\left D \right\rangle = \frac{1}{{\sqrt 2 }}(1,1)^T\), \(\left R \right\rangle = \frac{1}{{\sqrt 2 }}(1,i)^T\) and T is the transpose operator. Here we used the maximumlikelihoodestimation method^{34} to reconstruct the density matrix \(\hat \rho _{{\mathrm{mea}}}\), as shown in Fig. 3b. Assuming the state is pure, the experimental data are fitted with the state expressed as \(\left {\mathrm{\Phi }} \right\rangle = 0.60\left {{\mathrm{TE}}_1{\mathrm{TE}}_1} \right\rangle\) + \(0.29e^{  2.8i}\left {{\mathrm{TE}}_1{\mathrm{TE}}_0} \right\rangle\) + \(0.48e^{0.29i}\left {{\mathrm{TE}}_0{\mathrm{TE}}_1} \right\rangle\) + \(0.57e^{0.40i}\left {{\mathrm{TE}}_0{\mathrm{TE}}_0} \right\rangle\). According to the definition \(F = {\mathrm{Tr}}\left( {\hat \rho _{{\mathrm{mea}}}\hat \rho _{{\mathrm{pure}}}} \right)\), where Tr represents the trace and \(\hat \rho _{{\mathrm{pure}}} = \left {\mathrm{\Phi }} \right\rangle \left\langle {\mathrm{\Phi }} \right\), the net fidelity between these two density matrices is estimated to be 0.99 ± 0.02 (raw fidelity 0.93 ± 0.02). The high fidelity unambiguously shows that both intramodal and intermodal photon pairs are generated and that they constitute a complex but pure quantum state. The errors in fidelity were obtained by 100 times Monto Carlo calculation, with the experimental data subject to Gaussian statistics.
More interestingly, in some channels, intermodal photon pairs are negligible such that the entangled state
can be acquired. Furthermore, by adjusting the energy ratio and the phase difference between the TE_{0} and TE_{1} terms for the pump light, one can achieve a maximally transversemode entangled state as follows:
We took the frequency channels ±8 to test the quality of the entanglement. First, we remove the QWPs, and adjust the HWP in signal path to 0° or 45°, respectively. Then, we rotate the HWP in idler path from 0° to 360°. In this way, we obtained the coincidence between signal and idler paths as shown in Fig. 4a. The coincidence fringes can be fitted as 1 + V sin[2π(ϕ − ϕ_{c})/T], where V is the fringe visibility, ϕ_{c} is the initial phase, and T is the oscillation period. The fringe visibility V is defined as V = (d_{max} − d_{min})/(d_{max} + d_{min}), where d_{max} and d_{min} are the maximum and minimum of the fitted data, respectively. The net visibilities for ϕ_{s} = 0° (solid black line) and ϕ_{s} = 45° (solid red line) bases are 97 ± 2% (raw visibility 95 ± 2%) and 100 ± 2% (raw visibility 99 ± 2%), which are both greater than \(\frac{1}{{\sqrt 2 }} \approx 71\%\) and show the presence of Bell nonlocality between the signal and idler photons.
Then, quantum state tomography was performed to measure the state density matrix \(\hat \rho\). The ideal density matrix of the maximally entangled state in Eq. (3) and the measured density matrix of the output state from frequency channels ±8 are shown in Fig. 4b, c, respectively. The fidelity is defined as \(F = {\mathrm{Tr}}\left( {\hat \rho _{{\mathrm{mea}}}\hat \rho _{{\mathrm{ideal}}}} \right)\), where \(\hat \rho _{{\mathrm{ideal}}}\) is the ideal density matrix. We obtained a net fidelity of 0.96 ± 0.01 (raw fidelity 0.93 ± 0.01), confirming that the generated quantum state is of high quality and very close to the ideal maximally entangled state. The deviation of the fidelity from unity was mainly due to the errors in rotating the angles of the wave plates. Through use of a cascading onchip MachZehnder interferometer with thermal tuning to regulate the energy distribution and phase of different transversemodes, any level of biphoton entanglement can be achieved.
Discussion
We have shown the photon pair and entangled Bell state generation based on transversemode in a multimode waveguide. Using larger waveguide that support more transversemodes enables highcapacity information processing within a more compact chip than singlemode ones. Very recently, we have successfully demonstrated a 10channel transversemode (de)multiplexer in a 2.3 μmwide waveguide.^{35} We believe that a higherdimensional entangled state preparation will become feasible in the near future, even though several issues need to be addressed. For example, the efficiency of Type III SFWM processes have very narrow bandwidth and decreases sharply. And the dispersion of different transversemodes also leads to the decoherence of the quantum superposed state. Nevertheless, for higherdimensional Bell state generation, we just need to use Type I SFWM processes, and should avoid the cross terms induced by the Type III processes by selecting frequency channels. As for the dispersion, it can be compensated using delay line or by switching the transversemodes with grating.^{22} Taking a 4dimensional quantum state preparation as an example, we discussed about the feasibility of realizing higherdimensional entangled quantum state with one multimode waveguide (see Supplementary Materials, Section 4 and Figs. S6–S8). Also, we could hybridize the degrees of freedom of both path and transversemode for higherdimensional quantum state generation. For example, a 16dimensional spatial mode requires only 4 multimode waveguides that support 4 transversemodes.
The fact that transversemode entangled state can be coherently converted into path and polarization entanglement provides convenience for largescale quantum photonic integrated circuits (QPICs)^{36,37,38,39,40,41,42,43,44} and hyperentanglement generation.^{28} Due to the nonuniform intermodal nonlinear process, photon pairs in different frequency channels are in different quantum states; thus, we can choose the quantum state as desired by selecting frequency channels, which is unimaginable for singlemode ones. This frequencymultiplexed transversemode entangled photon pair source in a multimode waveguide offers high selectivity and flexibility for realizing quantum applications. In principle, we can excite and measure arbitrary transversemodes with the mode multiplexing technique, special quantum state also could be engineered with the intermodal nonlinear processes.
The intermodal SFWM process, where both the firstorder and secondorder dispersions can be engineered for phase matching, also can be used to realize frequency conversion between widely separated wavelengths. According to our calculations, in a multimode silicon waveguide with a crosssection of ~1600 × 220 nm^{2}, conversion between wavelengths separated by 600 THz (~800 nm) can be achieved (see Supplementary Materials, Fig. S9). On the other hand, the high quality cavities with multiple transversemodes have also been explored to enhance the SFWM processes.^{45,46,47}
In conclusion, we have demonstrated the transversemode entangled photon pairs and quantum states generation using a multimode silicon waveguide, which has potential in extending further to higherdimensional space. Combined with previous studies on coherent conversion and manipulation of transversemode entanglement,^{21} the present work shows the potential for realizing complex quantum information processing with high dimensionality and multiple degrees of freedom. Moreover, the multimode waveguide enables to tailor quantum state with mode multiplexing technique and shows much richer nonlinear phenomena than the singlemode one, thus providing an attractive platform for quantum information applications.
Methods
System efficiency
We ascertained the efficiencies for all components using laser light measurements (see Supplementary Materials, Fig. S10). The onedimensional grating coupler and 2D grating coupler show coupling losses of 5 dB and 8 dB, respectively. The chip excess losses in our device are 5 dB for the TE_{0} mode and 6 dB for the TE_{1} mode. The mode multiplexer and demultiplxer induces an excess loss of 0.5 dB on the TE_{1} mode. The postfilters and DWDM show an excess loss of 6 dB. The system used for state tomography shows a loss of 2.12 dB and both detectors have an efficiency of 85% (−0.7 dB).
Photon pairs generated in different parts
Using the photon pair \(\left {{\mathrm{TE}}_0{\mathrm{TE}}_0} \right\rangle\) as example, we estimated the contribution of the finally detected photon pairs from parts I, II, and III. In our device, the part I and part III consist of singlemode waveguides with length of about 450 and 250 μm, respectively. Assuming that the pump power is the same and the conversion processes are lossless, the ratio of photon counts is C_{I} : C_{II} : C_{III} = 1 : 188 : 0.1. When a transmission loss of 2 dB/mm is considered in the multimode waveguide (part II), the ratio became C_{I} : C_{II} : C_{III} = 1 : 2.3 × 10^{4} : 5.6 × 10^{−8}. Therefore, the photons generated in part I is mostly dissipated as propagation loss in part II. And in both cases, the contribution from singlemode waveguide (parts I and III) is negligible.
Optical apparatus
We used a CW tunable laser (linewidth 10 kHz, central wavelength 1550.11 nm) as the pump light. The laser was amplified by an erbiumdoped fiber amplifier and filtered (100 dB extinction ratio). The remaining power, which was coupled to the device after a singlemode fiber and an onchip grating coupler, was 6.96 mW, which was then divided into two parts by an onchip directional coupler to excite the two transversemodes. Fiber alignment was maintained using a piezocontrolled fourdimensional displacement table for the adjustment of the position, as well as the coupling angle. The coupling angle was set as 15° and 4° for the onedimensional grating coupler and 2D grating coupler, respectively. Two offchip postfilters (100 dB extinction ratio) were used to remove the pump photons, and two DWDMs (with an extinction ratio of 30 dB for adjacent channels and 50 dB for nonadjacent channels) were used to separate the signal and idler photons. The correlated photons were recorded by two superconducting nanowire single photon detectors. The electrical signals were collected and analyzed through a timecorrelated single photon counting (TCSPC) system, with the coincidence window set as 0.8 ns.
Device fabrication
Chips were fabricated on the most commonly used 220 nm SOI platform. The pattern of the waveguides was exposed on maN 2403 negative tone photoresist using electronbeam lithography. The pattern was then transferred to the top silicon layer by an inductively coupledplasma etching process. Grating couplers were fabricated using a second etching process to achieve efficient fiberchip coupling. Finally, 1.2 μm plasma enhanced chemical vapor deposition (PECVD) SiO_{2} was deposited on the top of the waveguide to form the uppercladding.
Data availability
Data are available from the authors upon reasonable request.
References
 1.
Horodecki, R., Horodecki, P., Horodecki, M. & Horodecki, K. Quantum entanglement. Rev. Mod. Phys. 81, 865–942 (2009).
 2.
Caspani, L. et al. Integrated sources of photon quantum states based on nonlinear optics. Light Sci. Appl. 6, e17100 (2017).
 3.
Matsuda, N. et al. A monolithically integrated polarization entangled photon pair source on a silicon chip. Sci. Rep. 2, 817 (2012).
 4.
Silverstone, J. W. et al. Onchip quantum interference between silicon photonpair sources. Nat. Photon. 8, 104–108 (2014).
 5.
Silverstone, J. W. et al. Qubit entanglement between ringresonator photonpair sources on a silicon chip. Nat. Commun. 6, 7948 (2015).
 6.
Kues, M. et al. Onchip generation of highdimensional entangled quantum states and their coherent control. Nature 546, 622–626 (2017).
 7.
Reimer, C. et al. Generation of multiphoton entangled quantum states by means of integrated frequency combs. Science 351, 1176–1180 (2016).
 8.
Selvas, R. et al. Tunable multimodeinterference bandpass fiber filter. Opt. Express 13, 9439–9445 (2005).
 9.
AntonioLopez, J. E., CastilloGuzman, A., MayArrioja, D. A., SelvasAguilar, R. & LiKamWa, P. Tunable multimodeinterference bandpass fiber filter. Opt. Lett. 35, 324–326 (2010).
 10.
Dai, D. & Bowers, J. E. Novel ultrashort and ultrabroadband polarization beam splitter based on a bent directional coupler. Opt. Express 19, 18614–18620 (2011).
 11.
Diebold, E. D. et al. Giant tunable optical dispersion using chromomodal excitation of a multimode waveguide. Opt. Express 19, 23809–23817 (2011).
 12.
Wang, J., He, S. & Dai, D. Onchip silicon 8channel hybrid (de) multiplexer enabling simultaneous modeand polarizationdivisionmultiplexing. Laser Photonics Rev. 8, L18–L22 (2014).
 13.
Luo, L. W. et al. WDMcompatible modedivision multiplexing on a silicon chip. Nat. Commun. 5, 3069 (2014).
 14.
Stern, B. et al. Onchip modedivision multiplexing switch. Optica 2, 530–535 (2015).
 15.
Wright, L. G., Christodoulides, D. N. & Wise, F. W. Controllable spatiotemporal nonlinear effects in multimode fibres. Nat. Photon. 9, 306–310 (2015).
 16.
Wright, L. G. et al. Selforganized instability in gradedindex multimode fibres. Nat. Photon. 10, 771–776 (2016).
 17.
Pourbeyram, H., Nazemosadat, E. & Mafi, A. Detailed investigation of intermodal fourwave mixing in SMF28: bluered generation from green. Opt. Express 23, 14487–14500 (2015).
 18.
Ding, Y., Xu, J., Ou, H. & Peucheret, C. Modeselective wavelength conversion based on fourwave mixing in a multimode silicon waveguide. Opt. Express 22, 127–135 (2014).
 19.
Qiu, Y. et al. Modeselective wavelength conversion of OFDMQPSK signals in a multimode silicon waveguide. Opt. Express 25, 4493–4499 (2017).
 20.
Defienne, H., Barbieri, M., Walmsley, I. A., Smith, B. J. & Gigan, S. Twophoton quantum walk in a multimode fiber. Sci. Adv. 2, e1501054 (2016).
 21.
Feng, L. T. et al. Onchip coherent conversion of photonic quantum entanglement between different degrees of freedom. Nat. Commun. 7, 11985 (2016).
 22.
Mohanty, A. et al. Quantum interference between transverse spatial waveguide modes. Nat. Commun. 8, 14010 (2017).
 23.
GarayPalmett, K. et al. Photonpair generation by intermodal spontaneous fourwave mixing in birefringent, weakly guiding optical fibers. Phys. Rev. A. 93, 033810 (2016).
 24.
Karpiński, M., Radzewicz, C. & Banaszek, K. Dispersionbased control of modal characteristics for parametric downconversion in a multimode waveguide. Opt. Lett. 37, 878–880 (2012).
 25.
Wang, J. et al. Multidimensional quantum entanglement with largescale integrated optics. Science 360, 285–291 (2018).
 26.
Lanyon, B. P. et al. Simplifying quantum logic using higherdimensional Hilbert spaces. Nat. Phys. 5, 134–140 (2009).
 27.
Gisin, N., Ribordy, G., Tittel, W. & Zbinden, H. Quantum cryptography. Rev. Mod. Phys. 74, 145–195 (2002).
 28.
Ciampini, M. A. et al. Pathpolarization hyperentangled and cluster states of photons on a chip. Light Sci. Appl. 5, e16064 (2016).
 29.
Harada, K. et al. Frequency and polarization characteristics of correlated photonpair generation using a silicon wire waveguide. IEEE J. Sel. Top. Quantum Electron. 16, 325–331 (2010).
 30.
Ma, C., Jere, M., Wang, X. & Mookherjea, S. High CAR and low g ^{(2)}(0) of 1.55 μm entangled photonpairs generated by a silicon microring resonator. CLEO: QELS_Fundamental Science. Opt. Soc. Am. FTh1G. 4 (2018).
 31.
Xiong, C. et al. Photonic crystal waveguide sources of photons for quantum communication applications. IEEE J. Sel. Top. Quantum Electron. 21, 205–214 (2015).
 32.
Taillaert, D. et al. A compact twodimensional grating coupler used as a polarization splitter. IEEE Photon. Tech. Lett. 15, 1249–1251 (2003).
 33.
Wang, J. et al. Chiptochip quantum photonic interconnect by pathpolarization interconversion. Optica 3, 407–413 (2016).
 34.
James, D. F. V., Kwiat, P. G., Munro, W. J. & White, A. G. Measurement of qubits. Phys. Rev. A 64, 052312 (2001).
 35.
Dai, D. et al. 10channel mode (de) multiplexer with dual polarizations. Laser Photonics Rev. 12, 1700109 (2018).
 36.
Harris, N. C. et al. Largescale quantum photonic circuits in silicon. Nanophotonics 5, 456–468 (2016).
 37.
Spring, J. B. et al. Boson sampling on a photonic chip. Science 339, 798–801 (2013).
 38.
Metcalf, B. J. et al. Quantum teleportation on a photonic chip. Nat. Photon. 8, 770–774 (2014).
 39.
Carolan, J. et al. Universal linear optics. Science 349, 711–716 (2015).
 40.
Harris, N. C. et al. Integrated source of spectrally filtered correlated photons for largescale quantum photonic systems. Phys. Rev. X 4, 041047 (2014).
 41.
Harris, N. C. et al. Quantum transport simulations in a programmable nanophotonic processor. Nat. Photon. 11, 447–452 (2017).
 42.
Santagati, R. et al. Silicon photonic processor of twoqubit entangling quantum logic. J. Opt. 19, 114006 (2017).
 43.
Wang, J. et al. Experimental quantum Hamiltonian learning. Nat. Phys. 13, 551–555 (2017).
 44.
Paesani, S. et al. Experimental Bayesian quantum phase estimation on a silicon photonic chip. Phys. Rev. Lett. 118, 100503 (2017).
 45.
Gabrielli, L. H., Liu, D., Johnson, S. G. & Lipson, M. Onchip transformation optics for multimode waveguide bends. Nat. Commun. 3, 1217 (2012).
 46.
Jiang, X., Wu, H. & Dai, D. Lowloss and lowcrosstalk multimode waveguide bend on silicon. Opt. Express 26, 17680–17689 (2018).
 47.
Reimer, C. et al. Crosspolarized photonpair generation and bichromatically pumped optical parametric oscillation on a chip. Nat. Commun. 6, 8236 (2015).
Acknowledgements
This work was supported by the National Natural Science Foundation of China (NSFC) (Nos. 61590932, 11774333, 61725503, 61431166001), the Anhui Initiative in Quantum Information Technologies (No. AHY130300), the Strategic Priority Research Program of the Chinese Academy of Sciences (No. XDB24030601), the National Key R & D Program (No. 2016YFA0301700), the Zhejiang Provincial Natural Science Foundation of China (Z18F050002), and the Fundamental Research Funds for the Central Universities. This work was partially carried out at the USTC Center for Micro and Nanoscale Research and Fabrication.
Author information
Author notes
Affiliations
Contributions
All authors contributed extensively to the work presented in this paper. M.Z. and D.X.D. prepared the samples. L.T.F., M.Z., D.X.D., and X.F.R. performed the measurements, data analyses and discussions. X.X., Y.C., H.W., M.L., G.P.G., and G.C.G. conducted theoretical analysis. X.F.R. and D.X.D. wrote the manuscript and supervised the project.
Corresponding authors
Correspondence to DaoXin Dai or XiFeng Ren.
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, 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 material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Received
Accepted
Published
DOI
Further reading

Generation of multiphoton quantum states on silicon
Light: Science & Applications (2019)

Quantum experiments and graphs II: Quantum interference, computation, and state generation
Proceedings of the National Academy of Sciences (2019)