Electron-to-photon noise transfer in mid-infrared semiconductor lasers

TL;DR

This work investigates how electrical current fluctuations propagate to intensity noise in mid-infrared semiconductor lasers (QCLs and ICLs) by using a dedicated sub-shot-noise current driver to drive the devices. The authors quantify the current-to-light transfer with a transfer function $T$ by measuring relative intensity noise (RIN) and its dependence on bias current, finding a linear relationship $y = A x + B$ where the transfer coefficient $A$ peaks near threshold and decreases as the bias moves away from threshold. They report that intrinsic laser noise and detector matching limit observation of sub-Poissonian light in the MIR, and that the two laser types show different transfer sensitivities, with ICLs being more responsive than QCLs. The results underscore the need to improve laser quantum efficiency and detection-system matching to approach quantum-limited intensity-noise performance in MIR cascaded lasers, and provide a practical methodology for future MIR quantum-noise studies.

Abstract

Noise characteristics of state-of-the art light sources are crucial parameters in understanding their limitations towards quantum applications. This work describes a method to study the electrical noise transfer of current driver sources to the intensity noise of mid-infrared emission by commercial quantum and interband cascade lasers (QCLs and ICLs, respectively). A current driver with sub-shot electrical noise in a specific frequency range (up to 10 dB below the shot noise level) was developed for this purpose. This enables testing the performance of mid-infrared lasers when driven via such a quiet pump source. By using this novel current driver, we identify the fundamental noise of a QCL and an ICL, that is the laser intensity noise resulting solely from the internal dynamics of the laser under test. The proposed methodology allows us to retrieve the noise transfer function from current to light, showing that the main limitations in observing the quantum properties of the emitted photons come from laser excess noise and poor matching between laser and detection system in terms of bandwidth and optical power. From the analysis of the measured parameters, we highlight current technological limitations and suggest which key features should be optimized in mid-infrared systems for matching the performance required by quantum applications.

Figures (9)

• Figure 1: CW LIV curves of the two lasers. The turquoise circles represent voltage, while the pink circles represent optical power, both as functions of the applied current. (a) refers to the QCL at T = $\mathrm{20\degree C}$, (b) to the ICL at T = $\mathrm{20\degree C}$. In both plots, the orange-shaded region represents the range of bias current that has been investigated in the subsequent noise analysis. The orange area highlights the profile of the LI curve, showing that the slope of the line is not constant in the region under test for the ICL. The laser temperature is the same as the one used later in the noise analysis.
• Figure 2: Scheme of the setup used for the measurement of the lasers' intensity noise power spectral density (INPSD). The tested laser is supplied with the low-noise current driver. The current is modulated by means of the driver modulation unit, with white noise generated by an arbitrary function generator. While the laser's operating point is tuned across the respective orange range in the plots from Fig. \ref{['fig:LIV']}, for each value of bias current the peak-to-peak amplitude of the white noise is varied. The laser light is then collected onto a detector whose signal is processed by a spectrum analyzer to retrieve the wanted INPSD.
• Figure 3: Setup for the measurement of the driver's INPSD. The bias current is converted to a voltage signal via an home-made amplification circuit, whose core consists of two op-amps (AD797N and AD8001), and which is battery-supplied to extinguish the electric noise from the wall-plug. The amplified signal is sent to the spectrum analyzer. For each value of bias current, the latter is modulated with a white noise of varying amplitude by means of a function generator. In the same fashion, the lasers' bias current has been modulated for the measurements of the lasers' INPSD.
• Figure 4: Panel (a): RIN spectrum of the driver supplying 180 mA of bias current (RIN$_{\mathrm{driver}}$). The RIN$_{\mathrm{driver}}$ was measured for different amplitudes of applied modulation, similarly to the graphic in panel (c). But for clarity, the plot only shows the trace corresponding to the non-modulated current signal, together with the computed corresponding shot noise level symbolized by the black dashed line. The orange-shaded area indicates the frequency range where the spectra were averaged. Panel (b): Average RIN$_{\mathrm{driver}}$ plotted against the corresponding modulation amplitude. The red dashed line is the average RIN$_{\mathrm{driver}}$ of the non-modulated current signal, taken as a reference, while the average RIN of the modulated signal is represented by blue dots. The black dashed line represents the corresponding shot noise level. Panel (c): RIN spectra of the QCL pumped at 180 mA, with an attenuation of the optical power $\mathrm{\alpha}$ of 99.1% (RIN$_{\mathrm{QCL}}$). Traces of different colors correspond to different peak-to-peak amplitudes of the white noise added to the bias current, as reported in the legend, while the term "no modulation" stands for the non-modulated signal. The old-rose line represents the detector's background, while the black dashed line is the computed laser's shot noise. The orange-shaded area shows the frequency region where the average on the RIN spectra has been performed. Panel (d): Similarly to panel (b), this plot represents the average RIN$_{\mathrm{QCL}}$ versus the average RIN$_{\mathrm{driver}}$, i.e. the transfer function from the driver's bias current to the photon flux. The transfer function is a line whose slope A and offset B are stated in the legend.
• Figure 5: Noise analysis for the ICL at 90 mA of bias current. The panels (a), (b), (c), and (d) are similar to those of Fig. \ref{['fig:RIN_QCL']}, where (a) and (b) refer to the RIN$_{\mathrm{driver}}$ at 90 mA, while (c) and (d) to the RIN$_{\mathrm{ICL}}$ when the ICL is driven at 90 mA, with an optical attenuation $\mathrm{\alpha}$ of 69.7%. Here, the RIN$_{\mathrm{ICL}}$ has been averaged in the range between 200 and 400 kHz.