**Abstract:** This paper introduces a novel approach to high-precision pattern generation in gold microfabrication utilizing dynamic optimization of chemical etching parameters. Leveraging a Bayesian Adaptive Network (BAN) integrated with a multi-modal data ingestion system and a self-evaluating feedback loop, our method minimizes feature deviations and enhances etch uniformity across large substrate areas. The system achieves a 15-20% improvement in feature accuracy compared to conventional etching techniques by proactively adjusting etchant composition, temperature, and agitation based on real-time feedback from in-situ optical monitoring and finite element analysis (FEA) simulations.

**Abstract:** Plasma etching is a cornerstone of microfabrication, yet achieving optimal etch performance—high etch rate, precise feature profile, and minimal damage—remains a complex, multi-objective challenge. This work proposes an innovative process optimization framework employing Adaptive Multi-Objective Reinforcement Learning (AMORL) coupled with Bayesian hyperparameter tuning for enhanced control over Plasma Etching processes. Our approach dynamically learns the optimal parameter set for specific material and feature requirements, surpassing traditional statistical experimental design methods by adapting to real-time process feedback and predicting subtle, non-linear etch behavior.

**Abstract:** This paper introduces a novel approach to high-precision gold patterning for microfabrication applications, leveraging stochastic diffusion-controlled etching (SDCE) within a dynamically optimized feedback loop. Traditional photolithography faces challenges in achieving sub-10nm resolution and minimizing edge roughness, especially with metals like gold. This method controls the etching process by strategically introducing a spatially varying gas flow and modulating plasma parameters, guided by a real-time feedback system analyzing etching progress via high-resolution optical microscopy and machine learning prediction.

**Abstract:** The successful commercialization of lithium-metal batteries (LMBs) hinges on overcoming the challenges of dendrite formation and solid electrolyte interphase (SEI) instability, significantly hindering cycle life and safety. This paper introduces a novel, computationally-driven approach for accelerating the discovery and validation of high-performance solid electrolytes (SEs) for LMBs. We leverage a combination of density functional theory (DFT) calculations, machine learning (ML) interpoation, and automated microfabrication to rapidly screen candidate SE compositions and architectures, focusing on lithium garnet (Li₇La₃Zr₂O₁₂) derivatives with targeted dopant incorporation.

Research Assistant (Fixed Term) in the Department of Chemical Engineering and Biotechnology at the University of Cambridge.

Enabling & Support