Abstract:The demand for multifunctional materials has motivated the move from near-equilibrium materials to metastable i.e. out-of-equilibrium phases that can meet several desired target properties. The search for such metastable phases with exotic properties is non-trivial and often serendipitous. Inverse design approaches based on evolutionary search have been powerful tools, but such traditional searches have focused on identifying primarily stable and metastable materials with the lowest enthalpy. The inverse design of materials, with a focus on a desired property such as, for example, hardness is a challenging task because of the expensive computational cost involved in sampling multiple structures. The recent advances in machine learning have brought new powerful AI techniques to the forefront which can potentially revolutionize the inverse design and discovery of materials, especially metastable phases capable of meeting multifunctionality. In this work, we develop and apply an automated reinforcement learning workflow for inverse design that integrates first principles physics and atomistic simulations with machine learning (ML), and high-performance computing to allow rapid exploration of the superhard and ultrahard metastable phases of Carbon. We demonstrate an automatic machine learning based inverse design workflow to map new undiscovered metastable states ranging from near equilibrium to those far-from-equilibrium that satisfy multiple property objectives, specifically bulk moduli, shear moduli and hardness. We create a comprehensive library of carbon stable and metastable phases with varying hardness and subsequently shortlist 10 top performing candidate carbon structures, including two newly reported phases, based on their hardness and characterize their temperature dependent mechanical properties. A neural network model is built using featurization of allotropes of carbon to predict the quasi-harmonic Gibbs free energies. The Gibbs free energies of the top performing phases are analyzed to get an estimate of the experimental synthesizability of these superhard and ultrahard carbon phases. In general, we show using machine learning based inverse design approaches how hitherto inaccessible metastable states can be identified and potentially synthesized to meet the demand for multifunctional materials.

Informatics-Driven Design of Superhard B-C-O Compounds

A Possible Superhard Orthorhombic Carbon

Research progress of high hardness B-C-O compounds

Superhard solid solutions of diamond and cubic boron nitride

Synthesis of Novel Superhard Materials under Ultrahigh Pressure

Machine learning and evolutionary prediction of superhard B-C-N compounds

Superhard Orthorhombic Phase of B2CO Compound

Machine learning enabled discovery of superhard and ultrahard carbon polymorphs

Thermodynamic aspects of materials' hardness: prediction of novel superhard high-pressure phases

Design and Theoretical Study of Novel Multifunctional 3D-BC2N Polymorphs

Superhard B2CO phases derived from carbon allotropes

Superhard Materials with Low Elastic Moduli: Three-dimensional Covalent Bonding As the Origin of Superhardness Inb6o

Crystal Structures and Mechanical Properties of Superhardbc2nandbc

First-Principles Study of O-Bn: A Sp(3)-Bonding Boron Nitride Allotrope

Superior toughness and hardness in graphite-diamond hybrid induced by coherent interfaces

Superhard sp<SUP>2</SUP>-sp<SUP>3</SUP> hybridized B<sub>2</sub>C<sub>3</sub>N<sub>2</sub> with 2D metallicity

First-principles structural design of superhard materials

Structural and Electronic Properties of C-Bc2n Supper Hard Material: an Ab-Initio Study

Design of a Series of Metallic BxNx+1 with Tunable Mechanical Properties

Superhard Phases of Simple Substances and Binary Compounds of the B-C-N-O System: from Diamond to the Latest Results (a Review)

Predicting Superhard Materials via a Machine Learning Informed Evolutionary Structure Search