Abstract:For a very long time, computational approaches to the design of new materials have relied on an iterative process of finding a candidate material and modeling its properties. AI has played a crucial role in this regard, helping to accelerate the discovery and optimization of crystal properties and structures through advanced computational methodologies and data-driven approaches. To address the problem of new materials design and fasten the process of new materials search, we have applied latest generative approaches to the problem of crystal structure design, trying to solve the inverse problem: by given properties generate a structure that satisfies them without utilizing supercomputer powers. In our work we propose two approaches: 1) conditional structure modification: optimization of the stability of an arbitrary atomic configuration, using the energy difference between the most energetically favorable structure and all its less stable polymorphs and 2) conditional structure generation. We used a representation for materials that includes the following information: lattice, atom coordinates, atom types, chemical features, space group and formation energy of the structure. The loss function was optimized to take into account the periodic boundary conditions of crystal structures. We have applied Diffusion models approach, Flow matching, usual Autoencoder (AE) and compared the results of the models and approaches. As a metric for the study, physical PyMatGen matcher was employed: we compare target structure with generated one using default tolerances. So far, our modifier and generator produce structures with needed properties with accuracy 41% and 82% respectively. To prove the offered methodology efficiency, inference have been carried out, resulting in several potentially new structures with formation energy below the AFLOW-derived convex hulls.

What problem does this paper attempt to address?

This paper aims to solve the key problems in new material design, especially how to quickly and effectively discover new materials with specific properties. Traditional methods usually rely on time - consuming and costly trial - and - error experiments and computational methods, such as density functional theory (DFT) - based methods, which require a large amount of computational resources. To solve these problems, the author proposes a new conditional generation method, using advanced machine - learning techniques to generate and optimize crystal structures. ### Main contributions of the paper 1. **Conditional structure modification**: - By optimizing the stability of existing structures, more stable polymorphs are generated. Specifically, this method is achieved by calculating the energy difference between the most favorable structure and all less stable polymorphs. 2. **Conditional structure generation**: - Completely new structures are generated, which meet the specific properties defined by the user. To this end, the author uses a variety of generation models, including diffusion models (Diffusion models), flow matching (Flow matching) and autoencoders (Autoencoder). ### Methods and techniques - **Data representation**: - The representation of materials includes information such as lattices, atomic coordinates, atomic types, chemical features, space groups and formation energies. This information is organized in matrix form for easy model processing. - **Loss function**: - The loss function takes into account the periodic boundary conditions of the crystal structure to ensure the periodic accuracy of the generated structure. Specifically, the loss function includes the L1 loss of atomic coordinates, the L1 loss of lattices and the periodic boundary condition loss. - **Generation models**: - A variety of generation models are used, including diffusion models (DDPM and DDIM), conditional flow - matching models (CFM), etc. These models can generate crystal structures that meet specific conditions from noise. ### Experimental results - **Generation task**: - The generation models perform well on the validation set, among which the DDIM model performs the best, with an accuracy rate of 82%. - **Modification task**: - The modification models perform relatively weakly on the validation set, but are still able to generate structures with the required properties, with an accuracy rate of 41%. - **Inference experiment**: - Through inference experiments on materials with specific chemical compositions, the author successfully generated multiple potential new structures. The formation energies of these structures are lower than the convex hull energies in the AFLOW database, indicating that these structures may have thermodynamic stability. ### Conclusion This research shows the great potential of the conditional generation method in new material design. Although there are some limitations, such as the limitations of data representation and the insufficient ability to process large - scale structures, the experimental results show that these methods can generate new materials with specific properties in a short time. This provides a new way to accelerate technological innovation in the field of materials science and is expected to bring breakthroughs in fields such as electronics, pharmaceuticals and energy storage.

Unleashing the power of novel conditional generative approaches for new materials discovery

Generative Design of Crystal Structures by Point Cloud Representations and Diffusion Model

Generative Hierarchical Materials Search

Scalable Diffusion for Materials Generation

Physics guided deep learning for generative design of crystal materials with symmetry constraints

Generative Inverse Design of Crystal Structures via Diffusion Models with Transformers

CrysText: A Generative AI Approach for Text-Conditioned Crystal Structure Generation using LLM

3-D Inorganic Crystal Structure Generation and Property Prediction via Representation Learning

A Machine-Learning-Assisted Crystalline Structure Prediction Framework To Accelerate Materials Discovery

MatterGen: a generative model for inorganic materials design

A machine learning potential-based generative algorithm for on-lattice crystal structure prediction

Data-Driven Score-Based Models for Generating Stable Structures with Adaptive Crystal Cells

Coarse-grained crystal graph neural networks for reticular materials design

High‐Throughput Discovery of Novel Cubic Crystal Materials Using Deep Generative Neural Networks

Functional materials discovery using energy–structure–function maps

Atomic structure generation from reconstructing structural fingerprints

Optimality guarantees for crystal structure prediction

Theory-Guided Discovery of Novel Materials

VQCrystal: Leveraging Vector Quantization for Discovery of Stable Crystal Structures

Structural Constraint Integration in Generative Model for Discovery of Quantum Material Candidates