Research - Zhu Research Group

Research

Research Overview

Our research is dedicated to the development of advanced computational methods and their application in the design of new materials with exceptional properties. By integrating high-performance computing, machine learning, and theoretical physics, our work aims to accelerate the discovery and optimization of materials for a wide range of technological applications.

Computational Methods and Software Development

libfp: A High-Performance Crystal Fingerprint Library

libfp is a state-of-the-art crystal fingerprint library that combines high-performance C implementation with an intuitive Python interface. The library excels at characterizing local atomic environments in crystal structures and provides precise quantification of structural similarities and differences. Engineered for exceptional computational efficiency, libfp can process large-scale crystallographic datasets with remarkable speed, making it an essential tool for materials science research and computational chemistry investigations. The library's proven reliability is demonstrated through its integration into several materials software packages, including EOSnet, ReformPy, and PALLAS. Experience libfp's capabilities firsthand by accessing our open-source code repository at GitHub.

EOSnet: Machine Learning Models with Graph Neural Networks

We are developing advanced machine learning methods, EOSnet, based on crystal graph neural networks (CGNNs) to address the challenges of predicting and optimizing material properties. EOSnet integrates crystal fingerprints (based on the Gaussian Overlap Matrix), to effectively account for many-body interactions within materials. This approach enables a more comprehensive representation of complex structural information, leading to significant advancements in prediction accuracy. Our models have been successfully applied to predict structural and functional properties, accelerating the discovery and design of materials with tailored characteristics. By leveraging these advanced methods, we aim to push the boundaries of computational materials science.

ReformPy: Rational Exploration of Fingerprint-Oriented Relaxation Methodology

ReformPy is a Python package for Rational Exploration of Fingerprint-Oriented Relaxation Methodology. Structural optimization is a crucial component in computational materials research, particularly in structure prediction. ReformPy introduces a novel method that enhances the efficiency of local optimization by integrating an extra fingerprint space into the optimization process. Our approach utilizes a mixed energy concept on the hyper potential energy surface (PES), combining real energy with a newly introduced fingerprint energy derived from the symmetry of the local atomic environment.

This method strategically guides the optimization process toward high-symmetry, low-energy structures by leveraging the intrinsic symmetry of atomic configurations. The effectiveness of our approach was demonstrated through structural optimizations of silicon, silicon carbide, and Lennard-Jones cluster systems. The results show that the fingerprint space biasing technique significantly enhances the performance and probability of discovering energetically favorable, high-symmetry structures compared to conventional optimizations. ReformPy is anticipated to streamline the search for new materials and facilitate the discovery of novel energetically favorable configurations.

PALLAS: Transition Pathway sampling via swarm intelligence and graph theory

The PALLAS method is a novel computational technique for predicting phase transition pathways between different crystal structures without requiring prior knowledge of the mechanisms involved. This method integrates:

  • Solid-State Dimer Method: Explores transition states between minima on the potential energy surface.
  • Crystal Fingerprint Method: Accurately quantifies the dissimilarity between crystal structures.
  • Swarm-Intelligence Algorithms: Facilitates efficient structure searching to identify possible transition paths.
  • Graph Theory: Automatically analyzes and identifies the most favorable transition pathways.

By combining these approaches, the PALLAS method establishes low-energy transition pathways, providing insights into the mechanisms driving phase transformations in materials.

Materials Design and Discovery

Leveraging our computational methods, we study and design new materials across a range of applications:

  • Superconductors: We investigate novel superconducting materials with the potential for high-temperature superconductivity. Our computational approaches allow us to predict and optimize superconducting properties, guiding experimental efforts in material synthesis.
  • Ferroelectrics: Our research includes the design of ferroelectric materials with enhanced polarization and transition temperatures. By understanding the structural factors influencing ferroelectricity, we aim to develop materials for applications in sensors, actuators, and memory devices.
  • Superhard Materials: We explore the properties of superhard materials for cutting, drilling, and wear-resistant applications. Our methods help identify compounds with exceptional hardness and mechanical strength.
  • Correlated Materials: We study materials with strong electronic correlations, such as transition metal oxides, to uncover novel electronic and magnetic phenomena. These materials have potential applications in quantum computing and spintronics.
  • Materials Under Extreme Conditions: We examine the behavior of materials under extreme pressures and temperatures to understand their structural transformations and stability. This research has implications for geophysics, planetary science, and the development of materials for extreme environments.