Key points are not available for this paper at this time.
Random number generation is a fundamental task for modern cryptography, secure communications, and stochastic computing. Quantum random number generators (QRNGs) provide inherently unpredictable data derived from quantum physical processes and therefore offer stronger security guarantees than classical approaches. However, some optical QRNGs rely on blue light-emitting diodes (LEDs), which suffer from reduced spectral matching with commonly used silicon avalanche photodiodes (APDs). This mismatch restricts the achievable signal-to-noise ratio (SNR) and limits the extractable quantum entropy. Here, we demonstrate a spontaneous emission-based QRNG employing a green InGaN LED coupled to a silicon APD. The improved spectral overlap between the emitted light and the APD responsivity results in a significantly higher SNR than with previously reported blue LEDs operating at similar electrical power levels. The measured signal is first filtered with a high-pass filter to suppress low-frequency noise, then randomness is extracted using the SHAKE256 hash function. A detailed statistical and spectral analysis is performed to evaluate the extractable entropy of the generated data. A physically reasonable estimation yields a quantum entropy generation rate of 2.78 Gbit/s. These results establish longer-wavelength nitride-based LEDs as a more suitable entropy source for QRNG systems, using inexpensive and widely available silicon photodiodes.
Kotov et al. (Mon,) studied this question.