Lithography Optimization With Artificial Intelligence for Thin Film Device Fabrication

In industrial production of thin film sensors, developing a new device can require time-consuming and expensive tests. These devices are characterized by a set of key performance indicators (KPIs) tailored to the device’s intended application. One of the critical design parameters to tune is the geometry of the sensing element, which is controlled through the lithography process. This work explores the use of machine learning (ML) algorithms and inverse analysis to find the optimal mask aligner (MA) lithography parameters (LPs) to obtain a desired pattern. First, an ML model is used to predict what pattern shape is obtained from a set of LPs. This model is then used by a search algorithm to find the LPs which lead to the shape closest to the one desired. The best model displayed an $R^{2}$ value of 0.85, which was enough to return a set of LPs that were validated in the cleanroom, obtaining high-quality $2 \times 10 \; \mu \text {m}^{2}$ ellipses.