unifiedkappa

Computes lattice thermal conductivity, including both the diagonal (particle-like, Peierls) and off-diagonal (wave-like, coherence) contributions, with a unified treatment of phonon transport. It supports two kinds of calculation:

• Minimum thermal conductivity (minikappa): needs only harmonic force constants. Every phonon mode gets the maximum scattering rate, Γ = tau_factor · ω / 2π (the default tau_factor=2 gives Γ = ω/π). This is the model of Xia et al., PNAS 120, e2302541120 (2023); it bounds the isotropically averaged high-temperature minimum kappa of crystals to roughly 0.1–2.6 W/m/K. This package supersedes the original Minikappa scripts. It produces the same output format but uses the public phonopy API instead of a patched phonopy.
• Unified kappa: uses the anharmonic phonon scattering rates from a ShengBTE/FourPhonon calculation. Diagonal group velocities are taken from ShengBTE (BTE.v_full) so that they are consistent with its scattering rates, and off-diagonal group velocities are computed with phonopy.

Off-diagonal group velocity tensors, with degeneracy properly treated, are computed by unifiedkappa/full_group_velocity.py.

Installation

pip install unifiedkappa    # once released on PyPI

From source:

git clone git@github.com:dgaines2/unifiedkappa.git
cd unifiedkappa
pip install -e .            # or: uv pip install -e .

For plotting, install the optional matplotlib dependency:

pip install -e ".[plot]"

For development (pytest, ruff, pre-commit, coverage, matplotlib):

pip install -e ".[dev]"
pre-commit install

Usage

Minikappa from harmonic force constants

import numpy as np
from unifiedkappa import UnifiedkappaManager

manager = UnifiedkappaManager.from_parameters(
    poscar_path="POSCAR",
    supercell_matrix=np.eye(3) * 4,
    primitive_matrix=np.eye(3),
    force_constants_filename="FORCE_CONSTANTS",
    mesh=[12, 12, 12],
)
minikappa_results = manager.run_minikappa(temperatures=[300.0, 600.0], tau_factors=[2.0])

Unified kappa from a ShengBTE/FourPhonon run

from unifiedkappa import UnifiedkappaManager

# supercell and q-point mesh are read from CONTROL
manager = UnifiedkappaManager.from_control(
    "path/to/shengbte_dir", poscar_filename="POSCAR"
)
unified_results = manager.run_unified_kappa()
minikappa_results = manager.run_minikappa(temperatures=[300.0])  # also available here

Frequency-resolved kappa

With save_histogram=True, the diagonal and off-diagonal contributions are also written binned by the frequencies of the two modes, which unifiedkappa.plotting can plot:

from unifiedkappa import UnifiedkappaManager
from unifiedkappa.plotting import make_kappa_histogram_plots

manager = UnifiedkappaManager.from_control(
    "path/to/shengbte_dir", poscar_filename="POSCAR", save_histogram=True
)
manager.run_unified_kappa()
# saves path/to/shengbte_dir/T300K/unifiedkappa-{d,od}_xx.{pdf,png}
make_kappa_histogram_plots(
    "path/to/shengbte_dir/T300K/unifiedkappa", fig_formats=("pdf", "png")
)

See examples/ for runnable NaCl examples of all of the above.

• Minikappa outputs (minikappa-{T}-{tau_factor}.dat) are saved in the base ShengBTE (or phonopy) directory as they're based on only harmonic properties
• Unifiedkappa outputs (T{T}K/unifiedkappa.dat) are saved in each ShengBTE temperature directory as they're based on anharmonic phonon scattering rates
• Each output file holds three flattened 3x3 tensors in W/m/K: diagonal, off-diagonal, and total
• Note that in order to read scattering rates produced by ShengBTE, you must explicitly specify onlyharmonic=.FALSE. in the CONTROL file

Testing

pytest                 # full suite (~20 s)
pytest -m "not slow"   # skip the full 25x25x25 mesh regression tests

The regression tests compare against the reference outputs in examples/.

Citation

If you use the minikappa model, please cite:

Y. Xia, D. Gaines II, J. He, K. Pal, Z. Li, M. G. Kanatzidis, V. Ozoliņš, and C. Wolverton, "A unified understanding of minimum lattice thermal conductivity," Proc. Natl. Acad. Sci. U.S.A. 120, e2302541120 (2023). https://doi.org/10.1073/pnas.2302541120

@article{xia_unified_2023,
  title   = {A unified understanding of minimum lattice thermal conductivity},
  author  = {Xia, Yi and Gaines, II, Dale and He, Jiangang and Pal, Koushik and Li, Zhi and
             Kanatzidis, Mercouri G. and Ozoli{\c{n}}{\v{s}}, Vidvuds and Wolverton, Chris},
  journal = {Proceedings of the National Academy of Sciences},
  volume  = {120},
  number  = {26},
  pages   = {e2302541120},
  year    = {2023},
  doi     = {10.1073/pnas.2302541120},
}