fcdmft

Official Resources

• Homepage: https://github.com/ZhuGroup-Yale/fcdmft
• Documentation: https://github.com/ZhuGroup-Yale/fcdmft/tree/master/examples
• Source Repository: https://github.com/ZhuGroup-Yale/fcdmft
• License: Open Source (check repo for specific license, often GPL or MIT)

Overview

fcDMFT is a Python-based software package designed for ab initio Full Cell Dynamical Mean-Field Theory (DMFT) calculations. Built upon the PySCF quantum chemistry framework, it extends standard embedding theories to treat solid-state systems with high accuracy. It focuses on GW+DMFT and HF+DMFT methodologies, enabling the study of electronic correlations in periodic crystals using quantum chemical solvers without downfolding to a small subspace.

Scientific domain: Condensed matter physics, Quantum chemistry of solids, GW+DMFT Target user community: Researchers in electronic structure theory, quantum chemistry, and correlated materials

Theoretical Methods

• Full Cell DMFT (no downfolding to minimal basis)
• GW+DMFT (G0W0+DMFT)
• Hartree-Fock + DMFT (HF+DMFT)
• Quantum Chemical Impurity Solvers (CCSD-GF, DMRG, FCI)
• Periodic RPA and GW
• CAS-CI impurity treatment

Capabilities (CRITICAL)

• Full Cell Embedding: Treats the entire unit cell in the embedding scheme, avoiding some supercell approximations and basis set truncation errors typical of downfolding.
• GW Integration: Combines GW quasiparticle energies with DMFT correlations for superior spectral properties.
• Quantum Chemistry Solvers: Utilizes solvers like Coupled Cluster Green's Function (CCGF), DMRG (via Block2), and FCI (via CheMPS2).
• PySCF Integration: leveraged PySCF for integrals, mean-field methods, and periodic boundary conditions.
• Parallelization: Mixed MPI and OpenMP parallelism for performance.

Key Features

Quantum Chemistry Solvers:

• Interfaces with Block2 for DMRG calculations.
• Interfaces with CheMPS2 for Full Configuration Interaction (FCI).
• Implements Coupled-Cluster Green's Function (CCSD-GF) solvers naturally.

Full Cell Approach:

• Performs DMFT calculations in the full unit cell basis, capturing non-local correlations more effectively than single-site approximations.

GW+DMFT:

• Advanced integration of GW screening with DMFT local correlations for accurate spectral properties.

Inputs & Outputs

• Input formats:
  • Python scripts using PySCF objects (e.g., run_dmft.py).
  • Geometry and basis set definitions via PySCF gto.Cell.
  • Scripts like si_gw.py for precursor GW steps.
• Output data types:
  • Green's functions (HDF5 or numpy arrays)
  • Self-energies (frequency dependent)
  • Quasiparticle energies
  • Spectral functions

Interfaces & Ecosystem

• PySCF: Core dependency for integral generation, mean-field (DFT/HF), and PBC tools.
• Solvers: Interfaces with Block2 (DMRG) and CheMPS2 (FCI).
• Python: Fully scriptable Python API for custom workflows.

Workflow and Usage

Users write Python scripts (examples in /fcdmft/examples):

1. Define cell and basis set using PySCF (gto.Cell).
2. Perform a mean-field (DFT or HF) or GW calculation (si_gw.py).
3. Derive the impurity Hamiltonian and GW double-counting terms (si_set_ham.py).
4. Construct the impurity Hamiltonian.
5. Invoke the fcDMFT solver loop (run_dmft.py).

Performance Characteristics

• Scalability: Uses MPI/OpenMP to scale on clusters.
• Cost: Computationally intensive due to full cell treatment and quantum chemical solvers, but offering high accuracy.

Comparison with Other Codes

• vs EDMFTF: EDMFTF (Wien2k) assumes locality in real space and focuses on forces; fcDMFT (PySCF) focuses on quantum chemistry solvers and full unit cell embedding.
• vs Standard DMFT: fcDMFT avoids downfolding to a minimal basis, treating the full cell, which is more expensive but potentially more accurate.
• Unique strength: Seamless integration with quantum chemistry solvers (DMRG, CCSD) via PySCF.

Verification & Sources

Primary sources:

1. GitHub Repository: https://github.com/ZhuGroup-Yale/fcdmft
2. Example scripts: /fcdmft/examples/Si in the repository
3. Associated Publications: Research from the Zhu Group (Yale) on Full Cell DMFT methods.

Verification status: ✅ VERIFIED

• Source code: OPEN (GitHub)
• Integration: Deeply integrated with PySCF
• Active development: Research code from Yale group