vibroglass

A Python toolkit for computing lattice dynamics and hydrodynamic properties of glasses and amorphous materials.

vibroglass provides an efficient, parallelised pipeline for computing vibrational spectra using the Haydock-Lanczos algorithm -- without the need for full matrix diagonalisation.

Features

• Vibrational Dynamic Structure Factor S(Q, omega) via the Haydock-Lanczos recursion
• Stochastic VDOS with element-resolved projections
• IR spectra from Born effective charges
• Hydrodynamic thermal transport analysis (disorder widths, heat capacity, kappa)
• kaldo interoperability for importing phonon data
• Supports LAMMPS, NumPy, and kaldo input formats

Installation

pip install -e .

Dependencies

Core: numpy, scipy, opt_einsum, ase, pandas

Optional: kaldo (phonon interoperability), matplotlib (plotting)

Quick start

Load a system and compute VDOS

from vibroglass import VibrationalSystem, VibrationalSpectra, LanczosOptions
import numpy as np

# Load dynamical matrix and atoms from a directory containing
# dynmat.npz and replicated_atoms.xyz
system = VibrationalSystem(root="path/to/data")

# Configure the Lanczos computation
options = LanczosOptions(hl_steps=300, eta=1.0)
spectra = VibrationalSpectra(system, options=options)

# Compute stochastic VDOS with 10 random vectors
vdos = spectra.compute_stochastic_vdos(nstoc=10)

# Average spectra
omega = spectra.options.omega_array
mean_spectrum = np.mean([omega * vdos[i]["S"] for i in range(10)], axis=0)

Compute S(Q, omega)

spectrum = spectra.compute_vdsf(nq=8)

# Access longitudinal and transverse components
Q = spectrum["Q"]
omega = spectrum["omega"]
S_L = spectrum[0]["L"]["S"]  # first Q-point, longitudinal
S_T = spectrum[0]["T"]["S"]  # first Q-point, transverse

Compute IR spectrum

ir = spectra.compute_IR_spectrum(
    polarizations=[0, 1, 2],
    charges_dict={"Si": 3.2, "O": -1.6},
)

Load from different formats

from vibroglass import load_dynmat_and_atoms

# From sparse .npz (recommended)
dynmat, atoms = load_dynmat_and_atoms("path/to/data")

# Also supports kaldo's second.npy and LAMMPS Dyn.form

Package structure

src/vibroglass/
  __init__.py               # Public API
  io/
    loaders.py              # Dynamical matrix I/O (npz, npy, LAMMPS)
    kaldo.py                # kaldo phonon adapters
  core/
    system.py               # VibrationalSystem
    lanczos.py              # Haydock-Lanczos algorithm
    kernels.py              # Delta-function kernels (Lorentzian, Gaussian, DHO)
  analysis/
    spectra.py              # VibrationalSpectra, LanczosOptions
    structure_factor.py     # S(Q, omega), phi(Q), IR vectors
    hydrodynamic.py         # Thermal transport analysis

Examples

See the examples/ directory for complete workflows:

Example	Description
0_aSi_DSF/	Compute the VDSF for amorphous silicon (1728 atoms)
1_IR_and_stoc_VDOS/	Element-resolved VDOS and IR spectra for amorphous SiO2

Running tests

tox

This runs both linting (ruff) and the test suite (pytest, 68 tests).

Key references

Reference	Description
A. Fiorentino, P. Pegolo, and S. Baroni, Npj Comput Mater 9, (2023).	Hydrodynamic extrapolation of lattice thermal conductivity in glasses
A. Fiorentino, E. Drigo, S. Baroni, and P. Pegolo, Phys. Rev. B 109, (2024).	Haydock-Lanczos algorithm for the VDSF; importance of anharmonicity
A. Fiorentino, P. Pegolo, S. Baroni, and D. Donadio, Phys. Rev. B 111, (2025).	Mass-disorder scattering in SiGe alloys; VDSF and Raman spectra