Abstract
We propose and demonstrate a scheme to realize a high-efficiency truly quantum random number generator (RNG) at room temperature. With the effective time bin encoding method, the avalanche pulses of an avalanche photodiode (APD) are converted into intrinsically self-unbiased random number bits that are robust to slow-varying noise. A light source is not necessary in this scheme. Furthermore, experimental results indicate that a high-speed RNG chip based on this scheme is potentially available with an integrable APD array.
Introduction
True Random Number Generators (TRNGs) are essential for modern cryptography and secure communications. While pseudo-random number generators are deterministic, quantum processes—such as the arrival of a single photon—provide a source of fundamental unpredictability. However, practical implementation of Quantum Random Number Generators (QRNGs) faces challenges. Most physical systems are inherently biased due to hardware imperfections or environmental fluctuations like temperature shifts. Traditionally, "post-processing" algorithms are used to correct this bias, but they are computationally expensive and can significantly reduce the bit-generation rate. This work presents a "self-unbiased" encoding method that is robust enough for real-world conditions.
Methodology
The proposed system utilizes the avalanche pulses of an APD operating in Geiger mode. These pulses can be triggered by incident photons or thermal fluctuations (dark counts). The core innovation is the Time-Bin Encoding Method.
Time is divided into blocks, each containing $N$ small detection windows (bins). Within a block, if $k$ pulses occur, there are $\binom{N}{k}$ possible arrangements. Since pulses follow a Poisson distribution, each specific arrangement of $k$ pulses is equally likely. The system maps these arrangements to unique integers.
Because the encoding occurs over microsecond durations, physical parameters like temperature remain essentially constant. Even if the environment changes slowly over longer periods, the relative probability of pulse arrangements within a block remains uniform, making the output immune to "slow-varying noise."
Experimental Results
The system was tested in three scenarios: a single APD with a light source, dual APDs with a light source, and APDs using only dark counts. The proof-of-principle system achieved generation rates of 0.69 Mbps with two APDs and 0.34 Mbps with a single APD.
A key advantage of this $N_k$-ary encoding method is its high efficiency. Unlike the classic von Neumann method, which has an efficiency limit of 0.25, the proposed method's efficiency increases with the block size $N$, theoretically approaching the Shannon entropy limit.
The generated bits passed the NIST Statistical Test Suite, a standard benchmark for detecting non-random patterns. The calculated Min-entropy was approximately 0.998 per bit, indicating nearly perfect randomness. To demonstrate robustness, researchers introduced 0.05 Hz sinusoidal noise to the light source; the encoding method successfully filtered out the resulting bias.
Conclusion and Future Work
The primary limitation identified is "afterpulsing," where an APD fires a false pulse shortly after a real one, introducing slight bit correlations. While current speeds are modest, the simplicity of the encoding method allows for implementation on an integrable APD array chip, which could potentially push generation rates into the Gbps range. This scheme offers a practical path toward high-speed, chip-integrated quantum randomness without the need for complex post-processing or external light sources.
Summary updated