Aerosol3D

Aerosol3D

Aerosol3D is a Python toolkit for modeling the 3D geometry and optical properties of atmospheric aerosol particles. It provides a unified pipeline from geometric construction to Discrete Dipole Approximation (DDA) optical computation.

Features

• 3D Geometry Modeling — Build spheres, ellipsoids, cubes, and import fractal aggregates
• Coating Algorithms — Apply distance-based, potential-based, CCM (Closed-Cell Model), and CAM (Coated-Aggregate Model) coatings
• Optical Computation — Solve optical properties via DDA with LDR polarizability (Draine & Goodman 1993) and 4 precision levels, or Mie theory (PyMieScatt, near-instant for spheres)
• Parallel Orientation Averaging — Distribute DDA orientation averaging across CPU cores via joblib, with tqdm progress bars and error-tolerant execution
• Effective Medium Approximation — Three EMA methods (volume-weighted, Maxwell-Garnett, Bruggeman) for computing effective refractive indices of mixed-composition particles
• Core-Shell Mie Solver — Exact Mie solution for coated spheres using PyMieScatt's layered-sphere API
• Optical Property Export — Multi-wavelength AerosolOpticsData container with auto-computed Legendre moments and NetCDF I/O
• Optical Visualization — Spectral properties, phase function comparison, Legendre convergence diagnostics, and comparison summary plots
• 3D Visualization — Generate screenshots and rotation videos using PyVista
• Material Database — Built-in refractive index data for common aerosol materials
• Flexible I/O — Export to VTP and voxel formats

Installation

Requires Python >= 3.10.

pip install Aerosol3D

For development:

git clone https://github.com/openEarthModelling/Aerosol3D.git
cd Aerosol3D
pip install -e ".[dev]"

Optional Dependencies

• Julia backend (required for DDA optical computation): Install Julia and run pip install pyjulia
• GPU acceleration: Requires CUDA-capable GPU and Julia CUDA packages

Quick Start

from Aerosol3D import (
    AerosolParticle, create_sphere, MixingState,
    preset_material, save_screenshot, solve_optics, SimulationConfig
)

# Create a black carbon sphere
soot = preset_material("black_carbon")
particle = AerosolParticle(
    name="bc_sphere",
    mixing_state=MixingState.INTERNAL,
    unit="nm",
)
particle.add_mesh("core", create_sphere((0, 0, 0), 50.0), soot)

# 3D visualization
save_screenshot(particle, "sphere.png", colors={"core": "black"})

# Optical computation (Mie or DDA)
config = SimulationConfig(wavelength=550.0, source="solar")
result = solve_optics(particle, config, solver="MIE")
print(f"Extinction efficiency: {result.qext}")

Export multi-wavelength results and visualize:

from Aerosol3D.optics import from_optical_results

results = [solve_optics(particle, SimulationConfig(wavelength=w), solver="MIE")
           for w in [450, 550, 650]]
data = from_optical_results(results, n_legendre=32)
data.to_netcdf("optics.nc")

See the examples/ directory for complete workflows including fractal aggregates, coated particles, and DDA-Mie comparison pipelines.

Examples

API Overview

Core Classes

• AerosolParticle — Particle container with multiple meshes/materials
• MixingState — Internal / external mixing state
• Material — Refractive index and density
• FractalAggregate — Fractal aggregate geometry

Geometry

• create_sphere(center, radius)
• create_ellipsoid(center, radii)
• create_cube(center, size)

Coating

• apply_distance_coating(particle, thickness, material)
• apply_potential_coating(particle, thickness, material)
• apply_ccm_coating(particle, thickness, material)
• apply_cam_coating(particle, thickness, material)

Optics

• solve_optics(particle, config, solver="DDA"|"MIE"|"MIE_CORESHELL", orientational_average=False, n_dirs=50, n_jobs=32, show_progress=True) — Optical solver dispatch with LDR polarizability (DDA) and optional parallel orientation averaging
• SimulationConfig(wavelength, source, precision="medium") — Simulation parameters; precision controls DDA dipole spacing (low/medium/high/ultra)
• AerosolOpticsData / from_optical_results(results, n_legendre) — Multi-wavelength optical property container
• compute_legendre_moments(phase_function, theta) — Legendre expansion of scattering phase function

Effective Medium Approximation

• volume_weighted(volumes, refractive_indices) — Linear volume-weighted mixing
• maxwell_garnett(volumes, refractive_indices) — Maxwell-Garnett mixing rule
• bruggeman(volumes, refractive_indices) — Bruggeman symmetric mixing rule

Core-Shell Optics

• solve_mie_coreshell(particle, config) — Core-shell Mie solver for coated spheres
• Particle.coreshell_geometry — Compute core-shell geometry from particle composition

Optical Visualization

• plot_spectral_properties(data) — Extinction/scattering/absorption spectra
• plot_phase_function(data) — Phase function P11 vs scattering angle
• plot_optical_comparison(data1, data2) — Side-by-side comparison of two datasets
• plot_phase_function_comparison(data1, data2) — Phase function comparison with relative difference
• plot_legendre_convergence(data) — Legendre moment convergence diagnostics

I/O & Visualization

• save_screenshot(particle, path)
• save_rotation_video(particle, path)
• save_vtp(particle, path) / save_voxel(particle, path)

Development

# Run tests
pytest

# Run tests with coverage
pytest --cov=Aerosol3D --cov-report=term-missing

License

MIT License. See LICENSE for details.

Acknowledgments

DDA optical computation is powered by CEMD.jl via a Python-Julia bridge.