Towards Fully Automated Characterisation of selfassembled Quantum Dots

Quantum dots (QDs) are nanoscale semiconductor heterostructures, made of common directbandgap semiconductors. QDs are able to confine the motion of electron-hole complexes (excitons) within their nanoscale structure. Those excitions behave as a few level quantum system, despite the solid state environment embedding of the QD. Therefore, they have been proposed as a potential source of entangled photon pairs for use in quantum communication, due to their ability to produce highly entangled photons on-demand. However, the growth process for QDs with the best optical properties to date is stochastic, making it difficult to locate QDs with the desired optical properties. This work proposes to utilize Machine Learning (ML) models as part of conventional automation software, to automate the process of searching for and characterizing QDs, allowing for more efficient experiments and potentially enabling the study of samples at large scale. The proposed software system consists of a communication software to control devices used in the optical measurement setup, software to detect QDs in images and correct distortions, a graphical user interface to execute and control measurements, and ML based assistance systems. The capabilities of ML models to provide decisions instead of a human at various tasks during automated measurements are investigated and compared in terms of performance. Finally, the proposed system is demonstrated by characterizing a sample of self-assembled GaAs QDs using automated photoluminescence measurements.