Skip to content

Welcome to Goldilocks

Goldilocks helps you generate Quantum Espresso SCF input files with optimal parameters predicted by machine learning models.

Use the menu to navigate the documentation.

Introduction

This application generates input files for Quantum Espresso single point energy calculations for 3D inorganic materials with optimized kmeshes.

What Goldilocks Does

Performing a simple SCF energy calculation requires choosing several parameters: - Functional: Exchange-correlation functional (PBE or PBEsol) - Pseudopotentials: From the SSSP library, with efficiency or precision modes - Cutoffs: Plane wave cutoff values for energy and density - K-mesh: K-point sampling grid - Smearing: Smearing method and corresponding degauss value - Magnetic configuration: For magnetic compounds, appropriate spin configuration and starting magnetizations

These parameters can usually be chosen in multiple ways. Choosing a set that provides a given calculation accuracy typically requires additional calculations to check convergence. Goldilocks predicts good parameter sets using machine learning models, helping you avoid running convergence studies or providing a good starting point for such calculations.

Key Features

  • ML-based k-point prediction: Uses Random Forest or ALIGNN models trained on large dataset of convergence values to predict optimal k-point spacing with confidence intervals
  • Automatic parameter selection: Chooses pseudopotentials and cutoffs based on SSSP recommendations
  • Multiple generation methods: Deterministic (ASE-based) or LLM-assisted generation
  • Database integration: Search for structures in free materials databases (JARVIS, Materials Project, MC3D, OQMD)

Information about data generated for this project can be found Data section.

Contributors

This application is one of the deliverables of the Goldilocks project (EPSRC EP/Z530657/1, Goldilocks convergence tools and best practices for numerical approximations in Density Functional Theory calculations) led by Barbara Monatani (STFC, UK), Gilberto Teobaldi (STFC, UK), Alin Marin Elena (STFC, UK), and Susmita Basak (STFC, UK). Data was generated by Junwen Yin (STFC, UK).

The creator of the application is Elena Patyukova (STFC, UK).