MAISE

Official Resources

• Homepage: https://github.com/tamercan/MAISE
• Documentation: https://github.com/tamercan/MAISE/wiki
• Source Repository: https://github.com/tamercan/MAISE
• License: Open-source

Overview

MAISE is a package for evolutionary structure prediction that emphasizes the use of neural network potentials to accelerate the search. By training interatomic potentials on-the-fly during the evolutionary search, MAISE reduces the number of expensive ab-initio calculations required to find the global minimum.

Scientific domain: Evolutionary structure prediction, machine learning potentials, materials discovery
Target user community: Computational materials scientists

Theoretical Methods

• Evolutionary Algorithm
• Neural Network Potentials (NNP)
• On-the-fly machine learning
• Density Functional Theory (verification)
• Structure fingerprinting

Capabilities (CRITICAL)

• Accelerated structure prediction using NNPs
• Automated training set generation and potential fitting
• Crystal structure evolution
• Interface with VASP for data generation and final verification
• Dimensionality support (clusters, crystals)

Sources: MAISE documentation, Phys. Rev. Lett. 110, 245501 (2013) (Reference to method)

Inputs & Outputs

• Input formats: Configuration files, VASP inputs
• Output data types: Predicted structures, trained potentials

Interfaces & Ecosystem

• VASP: Primary DFT engine
• Neural Networks: Internal implementation or interface

Performance Characteristics

• Significantly reduces DFT calls compared to standard EA
• ML potential overhead is small compared to DFT

Application Areas

• Complex unit cells
• Systems where DFT is too expensive for direct global search
• Metastable phase search

Community and Support

• Open-source
• Developed by Tamer Can / Stony Brook University

Verification & Sources

Primary sources:

1. GitHub: https://github.com/tamercan/MAISE

Confidence: VERIFIED

Verification status: ✅ VERIFIED

• Repository: ACCESSIBLE
• Method: Evolutionary algorithm + Neural Networks
• Applications: Accelerated structure prediction, ML potentials