XMFlib 0.1.7

A library for pair probability prediction using machine learning.

XMFlib

English | 中文

XMFlib is a machine learning-based library for predicting pair-site probabilities, designed for surface science and materials simulation. Leveraging pre-trained neural network models, it can quickly predict various types of pair probabilities based on input interaction energy, temperature, and coverage.

---

Features

• Supports multiple surface types (e.g., 100, 111 facets)
• Built-in multi-layer perceptron (MLP) models for efficient inference
• Simple and user-friendly API, easy to integrate into research and engineering projects
• Compatible with PyTorch, making it easy to extend and customize models

---

Installation

pip install XMFlib

---

Virtual Environment Setup (Recommended)

conda create --name <env_name> python=3.9
conda activate <env_name>
pip install XMFlib

---

Usage Example

from XMFlib.PairProbML import PairProbPredictor

predictor = PairProbPredictor()
result = predictor.predict(
    facet=100,                  # Facet type, options: '100' or '111'
    interaction_energy=0.3,     # Interaction energy (eV)
    temperature=400,            # Temperature (K)
    main_coverage=0.7           # Main species coverage (0~1)
)
print("Predicted probabilities:", result)

Example output:

Predicted probabilities: [0.002484329044818878, 0.38522598147392273, 0.5955939292907715]

The list corresponds to:

• Pee: probability of a vacancy-vacancy pair (empty-empty site)
• Paa: probability of a specie-specie pair (specie-specie)
• Pae: probability of a specie-vacancy pair (specie-empty site)