Compact Continuous-Variable Quantum Key Distribution System Employing Monolithically Integrated Silicon Photonic Transceiver

Compact Continuous-V ariable Quantum K e y Distribution System Emplo ying Monolithically Integrated Silicon Photonic T ransceiv er Denis F atkhiev , João dos Reis F razão , Alireza H. Derkani, Kadir Gümü ¸ s, Menno van den Hout, Aaron Albores-Mejia, and Chigo Ok onkwo High Capacity Optical T ransmission Lab , Eindhov en University of T echnology , The Netherlands d.f atkhie v@tue.nl Abstract W e demonstrate the first CV -QKD system f eaturing a custom-designed monolithic silicon photonic dual-polarisation transceiv er . Le ver aging PS-64-QAM, we achie ved 1.9 Mbit/s secret ke y r ate across 25 km of standard single-mode fibre , highlighting the potential of electronic-photonic integr ation f or practical QKD . © 2025 The Authors Introduction Owing to its high compatibility with conv entional telecommunication hardware infr astructure, which enhances its practical appeal, and increasingly robust security proofs [1,2], continuous-variab le quantum ke y distribution (CV -QKD) is a promising technology f or future inf or mation-theoretically secure communication netw orks. While discrete- variab le (D V) QKD uses single photons, CV - QKD offers adv antages by emplo ying coherent states and coherent detection, thus allowing the usage of components and digital signal processing (DSP) techniques similar to coherent optical transmission systems. Moreov er , CV -QKD with discrete modulation significantly simplifies the modulation scheme, reducing the requirement f or higher resolution digital-to-analogue conv er ters (D A Cs) [3] and consequently lowering system comple xity , thereby increasing the potential for its widespread adoption. Howe ver , practical deplo yment has been hindered not only b y the f easibility of specific components but also by the size , cost, and environmental sensitivity of b ulk-optic imple- mentations. Theref ore transition to photonic integrated circuits (PICs) is desirable for system miniaturisation, improv ed stability (especially Fig. 1: Schematic diagram of the transceiver chip . critical for phase-sensitiv e CV -QKD systems), and achie ving the economies of scale required f or industrialisation [4]. Recent de velopments in photonic integration f or CV -QKD demonstrate the realisation of core system elements, including light sources [5], single-polarisation [6,7] and dual-polarisation [8] tr ansmitters, and coherent receiv ers [9,10]. Although significant progress has been made, a complete, monolithic transceiv er im- plementation le v eraging polarisation multiple xing has not yet been demonstr ated. In this paper , we present the first CV -QKD system emplo ying a custom-designed monolithic- ally integrated silicon photonic dual-polar isation transceiv er co-packaged with trans-impedance amplifiers (TIAs). Utilising polar isation and phase diversity coherent detection, probabilistically shaped (PS) 64-QAM modulation, and a tailored DSP chain, the system reached 1.9 Mbit/s secret ke y r ate (SKR) across 25 km of standard fibre . For the security analysis, the asymptotic SKR bound under arbitrary modulation [11] is applied, while the trusted noise assumption and finite-size eff ects are considered as descr ibed in [12]. Our results demonstrate the potential roadmap tow ards practical and compact QKD systems e xploiting advanced photonic-electronic integ ration. Photonic Front-End The functional schematic of the transceiv er chip , detailing its core building bloc ks for both transmit and receive paths, is shown in Fig. 1. The transmitter takes four electr ical signals representing each polar isation state’ s in-phase (I) and quadrature (Q) components, which dr ive two independent IQ modulators. High-speed modulation is based on carr ier-depletion phase shifters within the Mach-Zehnder interf erometers, suppor ted b y thermal phase shifters f or bias control. An e xternal local oscillator (LO) source provides the optical carrier , which is split into both modulators. Note that the modulation stage is designed and optimised f or dr iverless operation Fig. 2: (a) Schematic of the experimental setup f eaturing transceiver and fibre link. (b) Normalised PSD estimate of the received signal. (c) Nor malised PSD estimates comparing the electronic and total noise measured at the receiver . (d) Receiver DSP chain. to reduce the system’ s e xcess noise. F ollowing modulation, two streams are combined with an integrated polarisation combiner-rotator (PCR). Monitor photodiodes (PDs) provide feedbac k signals f or automatic bias control (ABC) to set and maintain the desired operating points of the modu- lators. The receiv er divides the signal into its tw o constituent or thogonal polarisation components with a polar isation splitter-rotator (PSR). Coherent detection is performed f or each polarisation with two integr ated 90° h ybrid mixers , each combining one polarisation component with a por tion of the LO signal. The outputs from the optical h ybrids, corresponding to the I and Q components for each polar isation, are detected by high-speed PDs. TIAs conv er t the incoming photocurrents into amplified voltage signals representing the f our baseband wa vef or ms. Optical-to-electr ical conv ersion f or balanced coherent detection and monitoring utilises Ge-on-Si PDs optimised for telecom wa v elengths. System Implementation The e xperimental system, schematically detailed in Fig. 2a, centres around the silicon transceiv er chip as the core photonic quantum engine; hence , the transmit and receiv e paths perform the roles of Alice and Bob , respectiv ely . The signal generation process or iginates from the field-progr ammable gate arra y (FPGA), producing the digital baseband 250 MBaud PS-64- QAM streams time-interleav ed with QPSK pilot streams (power of pilot symbols 15 dB higher than powe r of QKD symbols) in a 50:50 ratio , which are subsequently pulse-shaped using a root-raised- cosine (RRC) filter with 10% roll-off . The power spectral density (PSD) estimate of the signal on the receiver side is presented on Fig. 2b . T o counteract signal quality deterioration induced b y low-frequency noise at the receiver (see Fig. 2c), the signals are also up-conv er ted by 300 MHz . These streams are fed to 14 bit D A Cs with a sampling rate of 4 GS/s , to generate the analogue electrical wav eforms for the modulators’ inputs. An external, narrow-line width laser operating at 1550 nm provides the continuous-w av e optical carrier . The modulated optical signal exits the transmitter por t and is attenuated b y a v ariable optical attenuator (V OA) f or state prepara tion. Part of the optical signal is tapped into an optical po wer meter (OPM) to estimate the modulation variance. The signal is then transmitted over 25.2 km of standard single-mode fibre (SSMF), bef ore being directed to the receiv er . An optical s witch is placed bef ore the receiver f or the noise calibration [13]. After coherent detection, the analogue electr ical signals are digitised by 14 bit analogue-to-digital conv er ters (ADCs) with a sampling rate of 2 GS/s and passed for fur ther offline processing using the DSP sequence outlined in Fig. 2d. The hardware implementation is depicted in Fig. 3, demonstrating a pr actical and relativ ely compact e xperimental setup with its ke y components. An additional control module, not shown in the figure , perf or ms ABC for the transmitter and automatic Fig. 3: Photograph of the experimental setup , highlighting ke y hardware components. Fig. 4: (a) Measured excess noise at the receiver o ver time , with the av erage v alue shown b y the red dashed line. (b) SKR versus transmission distance , illustrating theoretical perf or mance f or finite bloc k sizes (N) and the asymptotic limit (N → ∞ ), highlighting the experimental result (red star). gain control (A GC) f or the receiver . While based on con ventional coherent commu- nication techniques [14], the implemented DSP chain features some modifications. Specifically , a shor t data-aided least mean square (LMS) equalisation step precedes the constant modulus algorithm (CMA) equalisation stage to enhance conv ergence speed, and pilot symbols are used f or carr ier phase estimation. Fur thermore, the equaliser parameters (number of taps and step size) were optimised according to the approach proposed in [15]. Due to the use of the same transceiv er f or transmit and receive in the e xperiment – meaning that the carrier and LO light stem from the same laser source – the detection is effectiv ely homodyne, and thus a distinct frequency recov er y stage is omitted from the DSP chain. Results The system’ s performance was experimentally assessed through measurements of ke y para- meters and estimations of the achiev able SKRs. Figure 4a presents the measured excess noise at Bob ( ξ B ), e xpressed in shot noise units (SNU), monitored over 30 minutes. The excess noise fluctuates around an av erage value of appro ximately 0.005 SNU . The Alice’ s modulation variance V A f or the experiment was optimised and fix ed to 5.3 SNU. The analysis accounted for essential system parameters such as the shot noise clear ance (see Fig. 2c), the electronic noise contribution ( 0.34 SNU ), and the quantum efficiency of the receiver ( η = 0.7). The SKR was estimated based on the analytical bound f or arbitrar y modulation under the trusted noise assumption, incor porating finite-size eff ects according to: SKR = 2 R S n N (1 − FER)( β I AB − χ BE − ∆ ), where f actor of 2 is due to polarization m ultiple xing, R S is the system symbol rate, I AB is the mutual inf or mation between Alice and Bob , χ BE is the Hole vo information between Bob and Eve , and ∆ accounts for the penalty caused by the finite size eff ects of the pr ivacy amplification [16]. K e y operational parameters include a reconciliation efficiency β = 96% and a frame error rate (FER) of 50%, using an e xpanded protogr aph-based low- density parity check code ( R = 1 / 5) punctured to the appropr iate rate with an error correction bloc klength of 10 5 [17]. Given that 50% of the transmitted symbols are pilots and 50% of the remaining data symbols are utilised for parameter estimation, the fraction of symbols av ailable f or ke y generation is n / N = 0.25. Figure 4b displays the estimated SKR as a function of transmission distance. Theoretical SKR cur ves , calculated using the measured system parameters, are shown f or total b lock lengths ( N ) of 2 · 10 6 and 2 · 10 7 , alongside the asymptotic limit ( N → ∞ ). The achiev ed experimental SKR is highlighted by the red star , demonstrating a ke y rate of 1.9 Mbit/s ov er 25.2 km of standard fibre f or a practical b lock length of N = 2 · 10 7 . Conclusions W e demonstrated the first CV -QKD system built on top of a custom monolithically integrated dual-polarisation silicon photonic transceiv er , showcasing the potential of photonic-electronic integration to enable practical and compact quantum-secure communications. By tightly co-packaging the silicon PIC with TIAs, we achie ved a secret ke y rate of 1.9 Mbit/s ov er 25 km of standard single-mode fibre under finite-size analysis. Fur ther enhancement of the ke y rates is anticipated through dedicated optimisation of the receiver electronics to reduce excess noise and improv e shot noise clearance. In addition, the ne xt dev elopment milestones include investigating other PIC mater ial and packaging systems for laser source integration and deplo ying the system between two distinct transceivers , which will require mainly the addition of a frequency recov er y DSP stage. This work represents a ke y step tow ards realising lo w-comple xity CV -QKD systems within more compact f or m f actors. Ackno wledgements The Dutch Ministr y of Economic Affairs (EZ) suppor ted this work under the Quantum Delta NL and the PhotonDelta National Growth Funds Programme . We also ackno wledge the suppor t of CUbIQ T echnologies via the EIC T ransition Grant Project P AQAAL under g rant number 101213884. References [1] A. Leverrier, ‘Composable Secur ity Proof f or Continuous- V ar iable Quantum K ey Distribution with Coherent States’, Physical Revie w Letters , vol. 114, no. 7, p. 070 501, 2015. D O I : 10.1103/PhysRevLett.114.070501 [2] S. Pir andola and P . P apanastasiou, ‘Impro ved compos- able k ey rates f or CV -QKD’, Physical Re view Research , vol. 6, no. 2, p. 023 321, 2024. D O I : 10 . 1103 / PhysRevResearch.6.023321 [3] A. Leverrier and P . Grangier, ‘Unconditional Security Proof of Long-Distance Contin uous-V ar iable Quantum K ey Distribution with Discrete Modulation’, Physical Revie w Letters , v ol. 102, no . 18, p . 180 504, 2009. D O I : 10.1103/PhysRevLett.102.180504 [4] L. Labonté et al., ‘Integrated Photonics for Quantum Communications and Metrology’, PRX Quantum , vol. 5, no . 1, p . 010 101, 2024. D O I : 10 . 1103 / PRXQuantum . 5 . 010101 [5] L. Li et al., ‘Continuous-variable quantum k ey distribution with on-chip light sources’, Photonics Research , vol. 11, no . 4, pp . 504–516, 2023. D O I : 10.1364/PRJ.473328 [6] J. Aldama et al., ‘InP-based CV -QKD PIC T ransmitter’, in Optical Fiber Communication Conf erence (OFC) , 2023. D O I : 10.1364/OFC.2023.M1I.3 [7] A. A. E. Hajomer et al., Chip-based 16 gbaud continuous-v ariable quantum k ey distribution , 2025. D O I : 10.48550/arXiv.2504.09308 [8] S. Q. Ng, F . Kanitschar, G. Zhang and C. W ang, Gigabit- rate quantum key distr ibution on integrated photonic chips , 2025. D O I : 10.48550/arXiv.2504.08298 [9] Y . Piétri et al., ‘Exper imental demonstration of continuous-v ariable quantum key distribution with a silicon photonics integrated re ceiver’, Optica Quantum , vol. 2, no . 6, pp. 428–437, 2024. D O I : 10.1364/OPTICAQ. 534699 [10] S. Q. Ng, G. Zhang, C. Lim and C. W ang, ‘A chip- integrated homodyne detection system with enhanced bandwidth perf ormance f or quantum applications’, Quantum Science and T echnology , v ol. 9, no. 4, p . 045 010, 2024. D O I : 10.1088/2058- 9565/ad5d10 [11] A. Denys, P . Brown and A. Le v errier, ‘Explicit asymptotic secret ke y rate of continuous-v ariable quantum key distribution with an arbitrary modulation’, Quantum , vol. 5, p . 540, 2021. D O I : 10.22331/q- 2021- 09- 13- 540 [12] P . Jouguet, S. Kunz-Jacques, E. Diamanti and A. Lev errier, ‘Analysis of imperfections in practical continuous-v ariable quantum ke y distribution’, Ph ysical Revie w A , vol. 86, p. 032 309, 2012. D O I : 10 . 1103 / PhysRevA.86.032309 [13] S. v an der Heide, J. Frazão, A. Albores-Mejía and C. Okonkwo, ‘Receiver Noise Stability Calibration for CV -QKD’, in Optical Fiber Communication Conf erence (OFC) , 2023. D O I : 10.1364/OFC.2023.W2A.37 [14] M. S. F ar uk and S. J . Sav ory , ‘Digital Signal Processing for Coherent T ransceivers Emplo ying Multilevel Formats’, Journal of Lightwav e T echnology , vol. 35, no . 5, pp . 1125–1141, 2017. D O I : 10 . 1109 / JLT . 2017 . 2662319 [15] F . Roumestan et al., ‘Shaped Constellation Continuous V ar iable Quantum K ey Distribution: Concepts, Methods and Exper imental V alidation’, Jour nal of Lightwa ve T echnology , pp. 5182–5189, 2024. D O I : 10. 1109/ JLT. 2024.3391168 [16] M. Milicevic, C. Feng, L. M. Zhang and P . G. Gulak, ‘Quasi-cyclic multi-edge LDPC codes f or long-distance quantum cryptography’, npj Quantum Inf ormation , vol. 4, no . 1, 2018. D O I : 10.1038/s41534- 018- 0070- 6 [17] K. Gümü ¸ s and L. Schmalen, ‘Low rate protograph- based LDPC codes for continuous variable quantum ke y distribution’, in 17th Inter national Symposium on Wireless Communication Systems (ISWCS) , 2021. D O I : 10.1109/ISWCS49558.2021.9562244