Researchers at MIT and the Chinese University of Hong Kong have developed a machine learning tool to emulate photolithography manufacturing processes. Photolithography is commonly used in the production of computer chips and optical devices, manipulating light to etch features onto surfaces. Variations in the manufacturing process can cause the end products to deviate from their designs. The research team addressed this inconsistency by creating a digital simulator that models photolithography systems using real data. The simulator has been coupled with another digital simulator used to emulate photolithography systems’ performance. The combination of these two simulators allows users to create optical devices that closely match their design, optimising task performance.
This development could help to improve the accuracy and efficiency of the production process for optical devices, such as mobile cameras and devices used for medical imaging, telecoms, entertainment and augmented reality. The innovative pipeline for learning this digital simulator uses real-world data, making it adaptable to a wide variety of photolithography systems. Researchers collected data by designing and fabricating a wide range of features and sizes through photolithography systems. They then paired data measuring the physical characteristics and actual structures with design specifications to create a base for the neural network of the digital simulator.
The digital simulator has two components – an optics model that shows how light is projected onto the device surface, and a resist model that demonstrates how features are produced on the surface through photochemical reactions. Connecting these two parts to a physics-based simulator helps predict the device’s performance. The simulated device achieved better performance compared to other techniques in tests, closely resembling the intended design.
The researchers are considering improving their algorithms to replicate more complex devices, and test the system on consumer cameras, and photolithography systems using more involved forms of ultraviolet light. The research was supported in part by the US National Institutes of Health, Fujikura Limited, and the Hong Kong Innovation and Technology Fund. Facilities at MIT.nano were partly used for the study.
