Yaseera Ismail

The realisation of scalable quantum communication networks hinges on overcoming the detrimental effects of atmospheric turbulence in free-space quantum channels. Turbulence-induced decoherence, photon loss, and reduced entanglement fidelity remain critical obstacles to high-rate, secure quantum key distribution (QKD) and long-distance quantum networking. In this work, we present a unified framework that integrates machine learning techniques and light engineering by partial spatial coherence to enhance the robustness and performance of free-space quantum communication systems. We first demonstrate how supervised machine learning models can be employed to characterise and predict the state of a turbulent quantum channel in real time. Using experimentally reconstructed density matrices of entangled photon states, we train regression models to estimate the Strehl ratio, a key indicator of atmospheric turbulence strength, with high precision. Ensemble methods, particularly random forest regressors, achieve prediction errors below 4%, enabling accurate monitoring of channel conditions and providing a pathway toward adaptive quantum communication protocols [1]. Complementing this data-driven approach, we investigate the role of structured partially coherent light in mitigating turbulence effects. Through theoretical and experimental studies of spontaneous parametric down-conversion, we show that Gaussian Schell-model (GSM) beams enable phase conjugation of optical fields, effectively reversing wavefront distortions. The coherence properties of the generated idler beam are directly controlled by those of the pump field, allowing tailored resilience against atmospheric perturbations. We analyse the propagation of entangled photon pairs generated using partially spatially coherent pump beams through turbulent free-space links. Our results reveal that biphoton states produced with reduced spatial coherence exhibit enhanced robustness against turbulence compared to those generated by fully coherent pumps. Specifically, coincidence detection rates degrade more slowly with increasing turbulence strength, indicating improved preservation of quantum correlations and entanglement. This robustness arises from reduced spatial correlations in the pump, which translate into decreased susceptibility to scattering-induced decoherence [2]. By combining predictive machine learning with physically engineered light fields, this work establishes a hybrid strategy for resilient quantum communication. Machine learning enables real-time channel estimation and adaptive control, while partially coherent light provides intrinsic resistance to environmental noise. Together, these approaches offer a promising route toward reliable, free-space quantum links.