Abstract
The power of quantum random number generation is more than just the ability to create truly random numbers—it can also enable self-testing, which allows the user to verify the implementation integrity of certain critical quantum components with minimal assumptions. In this work, we develop and implement a self-testing quantum random number generator (QRNG) chipset capable of generating 15.33 Mbits of certifiable randomness in each run (an expansion rate of $5.11\times 10^{-4}$ at a repetition rate of 10 Mhz). The chip design is based on a highly loss-and-noise tolerant measurement-device-independent protocol, where random coherent states encoded using quadrature phase shift keying are used to self-test the quantum homodyne detection unit: well-known to be challenging to characterise in practice. Importantly, this proposal opens up the possibility to implement miniaturised self-testing QRNG devices at production scale using standard silicon photonics foundry platforms.
The Trust Problem in Randomness
Random number generation (RNG) is a cornerstone of modern cryptography. While classical "true" RNGs leverage physical phenomena like thermal noise, they are inherently vulnerable if the underlying physical model changes or if a manufacturer introduces a hidden hardware flaw.
Quantum RNGs (QRNGs) offer a superior alternative by utilizing the fundamental unpredictability of quantum mechanics. However, most commercial QRNGs still require users to trust that the hardware matches the manufacturer’s specifications. "Device-Independent" (DI) protocols can remove this trust requirement but are notoriously difficult to implement, requiring near-perfect detectors and complex entangled sources. This research bridges the gap by implementing a Measurement-Device-Independent (MDI) protocol on a mass-producible silicon chip, providing high-security randomness without the extreme hardware demands of full DI systems.
Methods and Hardware Innovation
The team implemented an MDI-QRNG protocol based on a "prepare and measure" game. In this setup, a trusted encoder (Alice) prepares quantum states using Quadrature Phase Shift Keying (QPSK)—four distinct phases of light. These states are then measured by an untrusted homodyne detector (Bob).
To ensure the system's integrity, a scoring rule compares the detector's output against expected quantum statistics. If the outputs align with quantum mechanical predictions, the system certifies the resulting bits as truly random.
Key hardware innovations include:
- Silicon Photonics Integration: The entire system, excluding the laser, is integrated onto a silicon-on-insulator (SOI) chip.
- Push-Push Modulation: To overcome "phase-loss dependency" (where changing phase accidentally alters intensity), the authors developed a novel modulation scheme to ensure the prepared quantum states remained pure.
- Tunable Homodyne Detector: The chip features a homodyne detector with a tunable beam splitter, achieving a Common-Mode Rejection Ratio (CMRR) of >65 dB for extremely precise measurement balance.
Key Insights: Resilience through Binning
A major breakthrough in this work was the realization that increasing the number of discretization bins for the homodyne output significantly improves tolerance for optical loss. By using 6 bins instead of the standard 2, the system achieved a total efficiency of ~69.1%, whereas previous protocols often required efficiencies exceeding 86% to be viable.
Results and Future Outlook
The experimental implementation successfully generated 15.33 Mbits of certifiable randomness per run with a net expansion rate of $5.11 \times 10^{-4}$ at a 10 MHz repetition rate. The integrated detector demonstrated a shot-noise clearance of 18.0 dB, ensuring quantum signals were clearly distinguishable from electronic noise.
While the current setup uses an external laser, the authors identify full laser integration as the next step toward a completely self-contained device. Furthermore, while the current 10 MHz speed is a proof-of-concept, silicon photonic components are capable of GHz speeds, suggesting that future iterations could be orders of magnitude faster. Because the design is compatible with standard silicon photonics foundry processes, these self-testing QRNGs are well-positioned for mass production in existing semiconductor facilities.
Summary updated