Session: 04-12-02: Design of Materials and Discovery of Constitutive Models Linking Process-Structure-Property-Performance Relationships

Paper Number: 173496

Atomistic Insights Into the Graphite-to-Diamond Phase Transformation via Deep Neural Network Potential Molecular Dynamics 

The transformation of graphite into diamond remains one of the most extensively studied and technologically relevant solid-state phase transitions in materials science. This process involves a fundamental change in bonding character—from the layered, sp²-hybridized structure of graphite to the three-dimensional sp³-bonded framework of diamond—resulting in drastically different mechanical, thermal, optical, and electronic properties. Diamond, in particular, is valued for its unmatched hardness, wide electronic bandgap, exceptional thermal conductivity, chemical inertness, and potential for quantum sensing applications. These properties enable its use in high-performance tools, thermal management systems, radiation-hardened electronics, and next-generation quantum devices.

In this work, we present a comprehensive molecular dynamics (MD) investigation of the graphite-to-diamond phase transformation enabled by a deep neural network (DNN) interatomic potential. This DNN potential was trained on a diverse and extensive set of ab initio molecular dynamics data, allowing it to capture the complex bonding characteristics of carbon with quantum mechanical accuracy. Unlike conventional empirical potentials, this machine-learned model allows us to explore large-scale systems and long time-scale dynamics while preserving the essential physics of bond breaking and formation under extreme conditions.

We performed large-scale MD simulations on graphite with both AB (hexagonal) and AA (simple hexagonal) stacking configurations to explore how varying stress states influence the transformation mechanisms. Two types of loading conditions were systematically studied: hydrostatic pressure and non-hydrostatic (anisotropic) compression. Under hydrostatic pressures exceeding 100 GPa, graphite transforms into polycrystalline cubic diamond, exhibiting well-developed grain boundaries. Notably, thin regions of hexagonal diamond frequently appear at these boundaries, suggesting their role as transient or stabilizing intermediate structures. In contrast, under non-hydrostatic stress—applied by selectively adjusting pressure along specific crystallographic directions—graphite preferentially transforms into pure hexagonal diamond, effectively demonstrating a stress-driven, phase-selective transformation pathway.

The transformation mechanisms also differ significantly between the two polymorphs. The formation of cubic diamond is facilitated by in-plane sliding and rearrangement of graphene layers, promoting a transition into the cubic sp³ lattice. Meanwhile, hexagonal diamond forms via out-of-plane buckling of graphite layers, reflecting a more localized and cooperative structural rearrangement. These insights reveal that stress state, stacking sequence, and deformation mode critically determine the resulting diamond phase and its microstructure.

Our findings highlight the power and flexibility of machine-learned interatomic potentials in capturing complex phase transformation phenomena with atomistic resolution. This study provides both mechanistic insight and a predictive computational framework for guiding the synthesis of diamond and related carbon allotropes under extreme pressure conditions—without the need for catalysts—paving the way for more efficient, tunable, and scalable carbon-based material design.

Presenting Author: Mehrab Lotfpour University of Nevada Reno

Presenting Author Biography: My name is Mehrab Lotfpour, a fourth-year PhD student. I'm working on computational materials science, focusing on molecular dynamics simulation of materials behavior.

Authors:

Mehrab Lotfpour University of Nevada Reno
Haoran Cui University of Nevada Reno
Yan Wang University of Nevada Reno
Lei Cao University of Nevada Reno