Authors:

Rohit Jammula

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

The ability to 3D-print nanoscale objects has lead to innovations in fields as diverse as microfluidics, electronics, and medicine. A particularly scalable process involves projection printing with two-photon polymerization. By using a digital micro-mirror device to focus specified patterns of light through a polymer, one can induce chemical reactions in a sheet of highly localized volumes to rapidly print the desired layer. Precise, nanoscale 3D structures result from altering the pattern at each layer and vertically adjusting the printing stage. Unfortunately, owing to extraneous physical factors, the resultant print often deviates from the intended pattern described by the incident light. Thus, we attempt to build predictive models relating the input light pattern and output structure, investigating both partial and complete pattern representations. Small-scale models such as Gaussian Random Processes are computationally efficient for partial parametrizations, though their inability to capture all facets of a pattern necessitates an accurate understanding of print deformation effects. On the other hand, convolutional neural networks are more computationally intensive but can capture the more robust spatial information inherent in multimodal, full-image representations.

Source:

Purdue University / 2023

Topics:

No topics listed

Co-authors:

Rohit Jammula