Session: 12-06-02: Scientific Machine Learning (SciML) for Characterization, Modeling, and Design of Structures and Materials

Paper Number: 150696

150696 - Programmable Strainscapes in a Two-Dimensional (2d) Material Monolayer 

Two-dimensional (2D) materials are a group of materials that are super thin in one dimension while expanding in the other two dimensions. These materials exhibit unique electronic transport properties, which are sensitive to deformation and strain distribution in 2D materials. By depositing dissimilar materials (i.e., stressors), such as metal oxide, onto surfaces of 2D materials, tensile or compressive strains could be introduced to 2D materials. These stressor-based methods offer a powerful and versatile approach to manipulating strain and the properties of 2D materials in comparison to the traditional methods that deform the entire structure. However, the lack of accurate theoretical models to predict the strain field and optimization tools to design stressors hinders the understanding of the underlying mechanisms and their widespread applications. In this work, we demonstrate that the strain field can be described by the classic Eshelby’s inclusion theory, where the stressor induces an eigenstrain in the covered 2D materials through interfacial interaction. After accounting for the competition between in-plane and out-of-plane deformation, this eigenstrain acts as an inclusion and influences the strain distribution in the surrounding region, which can be solved using the complex potential method. To validate the theoretical predictions, we developed a full atomistic molecular dynamics (MD) simulation method. In this method, stressors are modeled by multiple layers of graphene atoms with artificially added interlayer bonds and angles, and the eigenstrain is introduced by changing the equilibrium distance in the intralayer bonded interaction. The results show excellent agreement between the theoretical predictions and simulation, validating both the theoretical model and the correctness of the proposed simulation method. Additionally, we developed a deep neural network (DNN)--based optimization workflow to design the layout of stressors for designated displacement, strain, and pseudo-magnetic fields. In the optimization process, the DNN serves as a surrogate model, taking the target fields as input and outputting the optimized stressor layout. The theoretical models act as forward models, converting the stressor layout into generated fields. The loss is calculated between the target fields and generated fields, providing feedback to the discriminator to decide whether to continue the optimization process or accept the current design. As a demonstration, we implemented the optimization workflow to design pseudo-magnetic fields in a graphene monolayer and strain solitons/Moiré patterns in a bilayer graphene system. The results show promising accuracy as well as high efficiency in designing the stressors. This theoretical model, combined with optimization algorithms, provides a quick and highly accurate methodology for programming strainscapes in 2D materials by arranging the distribution of stressors with various shapes, sizes, and types. This work also lays the foundation for controlling sophisticated transport phenomena in 2D materials using stressors and highlights the application of mechanics theory in this emerging area of materials science.

Presenting Author: Qingchang Liu University of Illinois Urbana-Champaign

Presenting Author Biography: Dr. Liu is currently a postdoc research associate working with Dr. Harley Johnson at the I-MRSEC center. His research focuses on the theoretical models, numerical simulations, and machine learning-based property design of 2D materials.

Authors:

Qingchang Liu University of Illinois Urbana-Champaign
Harley T Johnson University of Illinois Urbana-Champaign