Abstract:The discovery of new superconducting materials, particularly those exhibiting high critical temperature ($T_c$), has been a vibrant area of study within the field of condensed matter physics. Conventional approaches primarily rely on physical intuition to search for potential superconductors within the existing databases. However, the known materials only scratch the surface of the extensive array of possibilities within the realm of materials. Here, we develop InvDesFlow, an AI search engine that integrates deep model pre-training and fine-tuning techniques, diffusion models, and physics-based approaches (e.g., first-principles electronic structure calculation) for the discovery of high-$T_c$ superconductors. Utilizing InvDesFlow, we have obtained 74 dynamically stable materials with critical temperatures predicted by the AI model to be $T_c \geq$ 15 K based on a very small set of samples. Notably, these materials are not contained in any existing dataset. Furthermore, we analyze trends in our dataset and individual materials including B$_4$CN$_3$ (at 5 GPa) and B$_5$CN$_2$ (at ambient pressure) whose $T_c$s are 24.08 K and 15.93 K, respectively. We demonstrate that AI technique can discover a set of new high-$T_c$ superconductors, outline its potential for accelerating discovery of the materials with targeted properties.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to discover new superconductors with high critical temperature ($T_c$) in the field of high - temperature superconducting materials. Traditional methods mainly rely on physical intuition to search for potential superconducting materials, but these methods can only access a small part of the known material database. Therefore, the paper proposes an artificial intelligence search engine named **InvDesFlow**, aiming to explore and discover new high - temperature superconducting materials by integrating deep - model pre - training and fine - tuning techniques, diffusion models, and physics - based methods (such as first - principles electronic structure calculations). ### Specific Goals: 1. **Generate New Crystal Structures**: Use the diffusion model to generate new crystal structures that do not exist in the existing database. 2. **Predict and Screen Superconducting Materials**: Through pre - training and fine - tuning techniques, combined with graph neural networks (GNN) and atomic docking methods, predict whether the generated crystals have superconducting properties and screen out candidate materials with high critical temperatures. 3. **Verify Material Stability**: Use density functional theory (DFT) calculations to verify the stability and superconducting performance of candidate materials. ### Main Achievements: - Using InvDesFlow, 74 dynamically stable materials were generated, and their predicted critical temperature ($T_c$) values are all higher than 15 K. - The trends of these materials were analyzed, and two specific materials, B$_4$CN$_3$ and B$_5$CN$_2$, were studied in detail. Their critical temperatures at different pressures are 24.08 K and 15.93 K respectively. - The electronic structures, phonon properties, and electron - phonon coupling (EPC) of these materials were further verified by DFT calculations, confirming the superconducting properties of B$_5$CN$_2$ and B$_4$CN$_3$. ### Method Innovations: - **Generation Model**: Directly generate crystal structures instead of first generating chemical formulas and then predicting structures, which improves the generation accuracy and efficiency. - **Post - processing Optimization**: Introduce the DPA - 2 model for atomic docking to avoid atomic collisions, optimize bond lengths, and ensure the rationality of the structure. - **Classification Model**: Introduce a superconducting classification model to ensure that the generated materials do indeed have superconducting properties. - **Formation Energy Prediction**: Improve the formation energy prediction model to improve the prediction accuracy. - **Active Learning**: Gradually expand the chemical space of high - critical - temperature superconducting materials through active learning. In conclusion, InvDesFlow significantly enhances the ability to discover new high - temperature superconducting materials by integrating multiple advanced artificial intelligence and physical methods, providing new possibilities for the development of materials science and related technologies.

InvDesFlow: An AI search engine to explore possible high-temperature superconductors

Inverse Design of Next-generation Superconductors Using Data-driven Deep Generative Models

Deep Generative Model for Inverse Design of High-Temperature Superconductor Compositions with Predicted T c > 77 K.

High-Throughput DFT-Based Discovery of Next Generation Two-Dimensional (2D) Superconductors

Deep Generative Model for Inverse Design of High-Temperature Superconductor Compositions with Predicted <i>T</i><sub>c</sub> > 77 K

An Integrated Machine Learning Model for Accurate and Robust Prediction of Superconducting Critical Temperature

Searching High Temperature Superconductors with the assistance of Graph Neural Networks

Materials Informatics Reveals Unexplored Structure Space in Cuprate Superconductors

High-throughput computation and structure prototype analysis for two-dimensional ferromagnetic materials

Predicting superconducting transition temperature through advanced machine learning and innovative feature engineering

Designing High-Tc Superconductors with BCS-inspired Screening, Density Functional Theory and Deep-learning

Deep Learning Model for Finding New Superconductors

Machine-learning approach for discovery of conventional superconductors

Sampling the Whole Materials Space for Conventional Superconducting Materials

Efficient exploration of high-Tc superconductors by a gradient-based composition design

Sampling the Materials Space for Conventional Superconducting Compounds

Predicting new superconductors and their critical temperatures using unsupervised machine learning

Machine Learning Prediction of Superconducting Critical Temperature through the Structural Descriptor

Recent advances in high-throughput superconductivity research

Superconductivity with $T_c$ up to 30.7 K in air-annealed CaFeAsF