Home

PAIPAI Wiki

Welcome to the PAIPAI wiki.

PAIPAI, the Package for Alloy Interstitial Predictions using Artificial Intelligence, is a Monte Carlo framework for exploring chemically complex metallic structures with interstitial solutes, defects, grain boundaries, and surfaces using machine-learning interatomic potentials.

This wiki is organized as a user-facing guide. For a deeper technical derivation, see the LaTeX manual in the main repository:

docs/paipai_v2_manual.tex

Start Here

• Installation
• Quick-Start
  • Input Structures and findinter
  • Running Search Mode
  • Running FiniteT Mode
  • Prefast Mode
  • Output Files and Diagnostics
  • Examples and Run Scripts

Reference

• Commands
  • paipai
  • findinter
  • packmc
  • warrencowley
• Troubleshooting and FAQ
• Version Guide

Version Summary

Branch	Purpose
main	Original stable implementation for conservative reproduction of older workflows.
v2.0-dev	New PAIPAI workflow with search and finiteT, worker pools, reference-state tracking, resume/continue support, findinter, and analysis tools.
v2.1-dev	v2.0 plus optional prefast warmup, online learning, and trial ranking before fast-worker screening.

Core Idea

PAIPAI separates two things:

1. The Monte Carlo state, defined by site occupations on a fixed reference structure.
2. The relaxed physical structure, whose energy is evaluated by an MLIP after geometry relaxation.

This separation is central to v2.0 and v2.1. The file SAVE stores the discrete reference state, while CONTCAR stores relaxed coordinates.