Researchers from MIT and the Chinese University of Hong Kong have developed a digital simulator that seeks to improve photolithography’s precision, often used in computer chips and optical devices manufacture. The process uses light to etch intricate designs onto surfaces, but minor discrepancies often cause devices’ final performance to deviate from designers’ initial intentions. The team’s new technique, called neural lithography, could enhance the accuracy and efficiency of device manufacture for use in mobile cameras, medical imaging, entertainment and telecommunications.
Photolithography projects a light pattern onto a surface to initiate a chemical reaction and etch designs onto the substrate. However, tiny deviations in the light’s diffraction and minor chemical reaction differences often lead to slight pattern anomalies in the finished product. Many traditional design approaches use physics-derived equations to predict these outcomes but struggle to accommodate a photolithography system’s specific anomalies, resulting in sub-par device performance.
The MIT researchers’ novel neural lithography technique involves using these equations as a base for building their photolithography simulator. They integrate a neural network trained with real, experimental data from users’ photolithography systems, which can compensate for any system-specific deviations. The team then uses the simulator in a broader design framework, which demonstrates how to create designs that meet specific performance objectives. Through post-fabrication calibration, they prove that their neural lithography model is far superior to previous methods.
The researchers demonstrated the proficiency of their technique by generating a butterfly image through a holographic device. Their method assembled a near-perfect butterfly closely matching the original design. Another test fabricating a multilevel diffraction lens resulted in superior image quality, testifying to the new technique’s potential.
The researchers aim to develop their algorithms further to model more complex devices and experiment with consumer cameras. They also hope to expand their method to include different photolithography systems such as those involving deep or extreme ultraviolet light. This research has been supported by the U.S. National Institutes of Health, Fujikura Limited, and the Hong Kong Innovation and Technology Fund, with some work completed using MIT.nano’s facilities.
