Density of States (DOS) and Bandgap calculations with UPET

Single-point evaluation of a bulk silicon cell with PET-MAD-DOS. The two silicon atoms are displaced off their equilibrium positions and the DOS and bandgap is computed using the PETMADDOSCalculator. Additionally, we showcase the performance of the denoising algorithm. The DOS is plotted with and without denoising, and the predicted bandgap is printed for both cases. The denoised DOS is less oscillatory and non-negative compared to the raw DOS.

Downloading PET-MAD-DOS model version: 1.0
/home/runner/work/upet/upet/.tox/docs/lib/python3.13/site-packages/metatrain/utils/io.py:228: UserWarning: trying to upgrade an old model checkpoint with unknown version, this might fail and require manual modifications
  warnings.warn(
/home/runner/work/upet/upet/.tox/docs/lib/python3.13/site-packages/metatomic_ase/_neighbors.py:78: UserWarning: `compute_requested_neighbors_from_options` is deprecated and will be removed in a future version. Please use `neighbor_lists_for_model` to get the calculators and call them directly.
  vesin.metatomic.compute_requested_neighbors_from_options(
The keys in results is: dict_keys(['dos_raw', 'fermi_level', 'bandgap', 'dos_denoised'])
Predicted bandgap: 0.5021 eV
True Bandgap: 0.4667 eV (Under the same DFT paramters)

import matplotlib.pyplot as plt
import numpy as np
import torch
from ase.build import bulk

from upet.calculator import PETMADDOSCalculator


atoms = bulk("Si", cubic=True, a=5.43, crystalstructure="diamond")
calculator = PETMADDOSCalculator(version="latest", device="cpu")
with torch.no_grad():
    results = calculator.calculate(atoms)

print(f"The keys in results is: {results.keys()}")
print(f"Predicted bandgap: {results['bandgap']:.4f} eV")
print("True Bandgap: 0.4667 eV (Under the same DFT paramters)")

energy_grid = np.arange(len(results["dos_raw"])) * calculator.energy_interval

plt.plot(energy_grid, results["dos_raw"], label="Raw DOS")
plt.plot(energy_grid, results["dos_denoised"], label="Denoised DOS")
plt.xlabel("Energy (eV)")
plt.ylabel("Density of States (States/eV)")
plt.title("Density of States for Bulk Silicon")
plt.legend()
plt.show()