Phase-Randomized Laser Pulse Generation at 10 GHz for Quantum Photonic Applications

TL;DR

The paper addresses interpulse phase correlations that limit high-rate phase randomization in gain-switched lasers used for quantum cryptography and randomness generation. It introduces external injection of broadband amplified spontaneous emission (ASE) to the laser cavity, boosting the effective spontaneous emission and accelerating phase diffusion so that each pulse starts from a random phase even at 10 GHz. Experimentally, a CW SLD ASE source injected into a DFB laser recovers the expected arcsine intensity distribution and decorrelated traces at 10 GHz, with spectral evidence showing reduced inter-pulse coherence; timing jitter increases with ASE power, reflecting the seeding process. A min-entropy analysis suggests potential raw random-number generation rates above 40 Gbit/s (and over 20 Gbit/s after compression), indicating feasibility for 10-GHz QKD and high-rate QRNG, provided compatible detectors and high-bandwidth modulators are available.

Abstract

Gain-switching laser diodes is a well-established technique for generating optical pulses with random phases, where the quantum randomness arises naturally from spontaneous emission. However, the maximum switching rate is limited by phase diffusion: at high repetition rates, residual photons in the cavity seed subsequent pulses, leading to phase correlations, which degrade randomness. We present a method to overcome this limitation by employing an external source of spontaneous emission in conjunction with the laser. Our results show that this approach effectively removes interpulse phase correlations and restores phase randomization at repetition rates as high as 10 GHz. This technique opens new opportunities for high-rate quantum key distribution and quantum random number generation.

Figures (5)

• Figure 1: Experimental setup for phase randomization achieved via externally injected ASE photons. SLD, Superluminescent Diode; DFB, Distributed Feedback Laser; AMZI, Asymmetric Mach-Zehnder Interferometer; RF, Radio Frequency Driving Signal (1-10 GHz); BS, 50:50 beam splitter.
• Figure 2: Experimental results of the DFB laser under gain-switching at 1 GHz (top), 10 GHz (middle), and 10 GHz with external ASE light injection (bottom). Left column: Density plots of the oscilloscope traces, with the sampling points indicated by red lines. Middle column: The corresponding intensity distributions at the sampling point (expressed as raw analog-to-digital converter (ADC) output, ranging between 0 and 255). Right column: Autocorrelation functions of the intensity at the sampling point, computed between samples separated by a specified number of lags. The blue shaded region denotes the 99% confidence interval bounds and lag 0 is omitted for clarity. All analyses are performed over 10$^6$ samples.
• Figure 3: Optical spectra of gain-switched DFB laser at (a) 1 GHz without any injection and (b) 10 GHz under different injection powers. The high injection power of the ASE light arises from its broad spectral width of 33 nm. An optical filter centered at 1547.8 nm with a 0.5 nm bandwidth is used to filter the output.
• Figure 4: Experimental results of the DFB laser under gain-switching at (a) 5 GHz, (b) 5 GHz with external ASE light injection. Left column: Density plots of the oscilloscope traces, with the sampling points indicated by red lines. Middle column: The corresponding intensity distributions at the sampling point. Right column: Autocorrelation functions of the intensity at the sampling point, computed between samples separated by a specified number of lags. The blue shaded region denotes the 99% confidence interval bounds and lag 0 is omitted for clarity. Analyses are performed over $1 \cdot 10^6$ samples.
• Figure 5: Experimental results of the DFB laser under gain-switching at (a) 8 GHz, (b) 8 GHz with external ASE light injection. Left column: Density plots of the oscilloscope traces, with the sampling points indicated by red lines. Middle column: The corresponding intensity distributions at the sampling point. Right column: Autocorrelation functions of the intensity at the sampling point, computed between samples separated by a specified number of lags. The blue shaded region denotes the 99% confidence interval bounds and lag 0 is omitted for clarity. Analyses are performed over $1 \cdot 10^6$ samples.