New Approaches to Quantum Cascade Laser Burn-In Testing and Waveguide Designs for Frequency Comb Operation

History has been shaped by humankind’s ability to tame light. From the first controlled fire to the first light bulb, innovations in light-science have paved the way for advanced technologies that have revolutionized human life. With this foundation spanning millennia of optical advances in place, the next critical steps will concentrate on improving the reliability and integration of light sources. Quantum cascade lasers are among the major achievements in laser technology. This dissertation makes significant advancements in the field of QCL reliability through two key research themes: improving reliability of the QCLs, and novel waveguide designs for achieving frequency comb operation. These advancements pave the way for the wider adoption and application of QCLs in various domains.

In the first part of the thesis, we design specialized QCL waveguides to enable future on-chip optical heterodyne detection in the mid-infrared part of the electromagnetic spectrum. We develop QCL active region designs targeting emission around 7 µm for potential sensing of highly energetic materials. Our design utilizes a double-phonon extraction scheme and ultra-strong coupling for maximizing gain. We design and optimize specialized waveguides to maximize optical mode overlap with the active gain region while still enabling sufficient interaction with a monolithically integrated

Schottky diode for future on-chip mixing. Our devices achieve room temperature electroluminescence at the target 7.56 µm wavelength under pulsed operation. Furthermore, we demonstrate frequency comb operation under low duty cycle pulsed operation, significant for enabling low power on-chip mixing.

In the second part of the thesis, we develop a novel accelerated burn-in testing approach that leverages support vector machine (SVM) models to analyze conventional electrical and optical measurements obtained from QCLs at elevated heatsink temperatures. We develop two generations of SVM models; the first one demonstrates accurate failure prediction for QCLs fabricated by the same process and quantum well design. The second one expands this approach to be more universal, applicable across varying fabrication conditions, facet coatings, and active region designs. We achieve this through feature reduction and ranking algorithms and filtering the experimental data to account for an early period of change in the laser operating characteristics. Our findings demonstrate the ability to detect up to six times more premature failures than traditional burn-in testing, all within a reduced 40-hour time frame. This allows prediction of devices prone to premature failure up to 319 hours in advance, significantly reducing the time and cost of conventional burn-in tests. The ability to predict failure in advance also enables deeper research into the mechanisms causing early QCL degradation. With differing emission wavelengths and varied fabrication processes of the quantum designs, our model’s robustness suggests a confident step towards universal performance across varying quantum designs. Additionally, our research reveals two distinct failure modes in the tested lasers during burn-in experiments, contributing to a deeper understanding of the underlying physics governing QCL failure mechanisms.