DMFTindex

DMFTindex is a proposed data indexing and storage framework intended to support dynamical mean-field theory (DMFT) calculations. It aims to provide a unified representation for core DMFT quantities, such as Green's functions, self-energies, and hybridization functions, and to enable efficient storage, retrieval, and querying of the large, multi-dimensional data sets produced by DMFT solvers and impurity solvers.

The motivation for DMFTindex is to improve data organization, portability, and reproducibility across different DMFT workflows. By defining standardized data models and interfaces, it seeks to facilitate sharing of results between codes, streamline benchmarking of solvers, and support reproducible computational pipelines from raw solver output to analysis and visualization.

Architecture and data model DMFTindex envisions a hierarchical data model that indexes quantities by frequency, momentum, and orbital indices. Core entities include G(iωn, k, α) for Green's functions, Σ(iωn, k, α) for self-energies, and Δ(iωn, k, α) for hybridization. The indexing scheme supports multiple frequency grids, momentum meshes, and orbital sectors, with extensible metadata describing the calculation context. Backends may include HDF5, Zarr, or memory-mapped formats, and the project anticipates adapters for common DMFT software. A query interface or API would allow retrieval of specific slices, projections, or time/frequency transforms without loading entire data sets.

Features DMFTindex emphasizes efficient I/O and storage, including chunked storage, compression, and parallel access. It prioritizes interoperability, metadata richness, versioning, and extensibility to accommodate new DMFT quantities or solver outputs.

Applications and development DMFTindex is intended to serve research groups performing DMFT studies, enabling standardized data exchange, reproducible workflows, and scalable benchmarking. The concept is under active discussion among the DMFT community, with open-source development and potential alignment with existing tools and standards in computational condensed matter physics.