(455g) Machine-Learning Accelerated Simulation: Toward Design and Synthesis of Nanocarbon Materials

Carbon nanomaterials are of tremendous technological interest due to the manifold of properties they can exhibit. For example, nanodiamond is renowned its hardness and biological inertness and is found in applications spanning industrial lubricants to biological implants and drug delivery vehicles; graphitic nanoparticles including quantum dots, nanotubes, and graphitic nanoonions exhibit tunable electronic and optoelectronic properties that are being explored for quantum computing, energy harvesting, and electronic devices. However, exploration in this materials space remains nascent due to challenges associated with establishing efficient, scalable synthesis strategies, and navigating the massive design space, where both of these challenges stem from a limited understanding of phase behavior and phase transformations in carbon-rich materials. To overcome these challenges, we deploy quantum-accurate simulations using a new T/P transferable carbon model developed through our ChIMES physics-informed machine learning framework to enable quantum-accurate simulation on far greater spatiotemporal scales. Using this model, we provide predictions of the carbon melt line up to 100 GPa and explore kinetic effects during phase transformation. We also provide preliminary insights on how these findings translate to nano-scale materials. Ultimately, this simulation capability is helping to establish an accelerated pipeline for design, discovery, and synthesis of next-generation nanocarbon materials.