Inverse Design of Copolymer Self-Assembly
Open Access
- Author:
- Kleeblatt, Devon
- Area of Honors:
- Materials Science and Engineering
- Degree:
- Bachelor of Science
- Document Type:
- Thesis
- Thesis Supervisors:
- Wesley Reinhart, Thesis Supervisor
Robert Allen Kimel, Thesis Honors Advisor - Keywords:
- Inverse materials design
self-assembly
polymers
generative model - Abstract:
- This thesis is an investigation into the inverse design of copolymer self-assembly. Co-polymers self-assemble into a diverse set of aggregate structures depending on their chemical sequence. Using the paradigm of inverse design, the co-polymer aggregate can be designed by dictating a target structure and using computational techniques to find possible chemical sequences which are likely to form the desired structure. Two approaches are taken to attempt inverse design of the aggregate structures, design by regression and design by generation. The regression technique involves using random forest regression and k-nearest neighbor regression to predict the structure of a copolymer from its chemical sequence. This effort was unsuccessful, as the margin of error was too large to make practical predictions. The generation technique involved training a conditional, long-short term memory autoencoder and using its decoder to generate sequences which are highly likely to reflect the values of the input condition. The autoencoder was not successful at generating sequences whose structure had a linear correlation with its input condition. However, the autoencoder was successful at generating sequences with aggregate structures that qualitatively reflected the desired target structure.