CAREER: Atomistic Investigation of Phase Transition in Nanostructured Silicon--Towards Convergent Understanding with Mechanics-Informed Machine Learning Potential

This Faculty Early Career Development (CAREER) award will support fundamental research on phase transition and plasticity in silicon nanostructures. Nanostructured silicon, such as nanoparticles, nanopillars, and nanowires have been widely used for integrated circuits, micro and nano-electromechanical systems, and photovoltaics due to the unique mechanical, optical, and electrical properties.

The knowledge generated in this research will be applied to enhance the mechanical reliability of silicon nanodevices, to improve the machining of silicon materials, and to explore novel approaches for silicon phase engineering, thereby advancing national health, prosperity, and welfare. In addition, the machine learning approach will provide an innovative computational framework to develop mechanics-inspired interatomic potentials to study the deformation of silicon and other materials at the atomic scale with high fidelity.

The research will be integrated into undergraduate and graduate education through two activities. A set of computation modules called “The Atomic View of Materials” will be designed and integrated into mechanics courses in the mechanical engineering curriculum. These modules will help students visualize the mechanical behavior of materials and get trained on advanced modeling and simulation skills.

In addition, this project will boost the participation of underrepresented minorities in research through internships.

Recent nanomechanical experiments have shown many unique and intriguing features of phase transitions in nanostructured silicon that are different from bulk silicon in many aspects. The kinetic mechanism of phase nucleation and propagation at atomic scale remains unknown, however. The research objective of this project is to quantitatively determine the roles of stress field, grain/phase boundaries, and plastic deformation on phase transition in silicon nanostructures.

It is noted that there is no reliable interatomic potential to describe silicon phase transition, thus a novel mechanics-informed machine learning potential based on deep neural network will be developed to fill the gap. The potential will be trained with the dataset containing stress-dependent phase transition minimum energy paths, which will provide sufficient resolution to sample the energy landscape under high stress.

The phase transition will be investigated with a combination of molecular dynamics simulations and finite deformation nudged elastic band method. In addition, a new finite deformation Dimer method will be developed based on the conventional Dimer method to probe phase nucleation under high stress and finite deformation without having to specify a final state structure.

The results generated in this project will be used to understand experimental observations from literature and collaborators.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.