Prismo

A high-performance Python-based FDTD solver for waveguide photonics.

Prismo implements the Finite-Difference Time-Domain method for solving Maxwell's equations, with features designed for photonic integrated circuits, waveguides, and optical devices.

Features

• 3D vectorial FDTD on Yee grids with automatic time-stepping
• GPU acceleration via CuPy with transparent CPU fallback
• Dispersive materials including Lorentz, Drude, Debye, and Sellmeier models
• Material library with Si, SiO2, Si3N4, Au, Ag, Al, and ITO
• Anisotropic materials supporting full permittivity and permeability tensors
• PML absorbing boundaries using the CPML formulation
• Eigenmode solver for 2D waveguide structures
• Advanced monitors for frequency-domain (DFT), power flux, and mode expansion
• S-parameter extraction with Touchstone export
• Data export to CSV and Parquet formats
• Lumerical compatibility for importing FSP files and material databases
• Parameter sweeps with parallel execution

Installation

pip install pyprismo

For GPU acceleration:

pip install pyprismo[acceleration]

For development:

git clone https://github.com/rithulkamesh/prismo.git
cd prismo
pip install -e ".[all]"

Note: The package name on PyPI is pyprismo, but you still import it as prismo:

import prismo  # Import name stays as 'prismo'

Quick Start

import numpy as np
import prismo

# Select backend
prismo.set_backend('cupy')  # Use GPU, or 'numpy' for CPU

# Create simulation
sim = prismo.Simulation(
    size=(10e-6, 5e-6, 0),  # 10×5 μm, 2D
    resolution=50e6,         # 20 nm grid spacing
    boundary_conditions="pml",
)

# Add source
source = prismo.GaussianBeamSource(
    center=(-4e-6, 0, 0),
    size=(0, 2e-6, 0),
    frequency=193e12,  # 1550 nm
    pulse_width=10e-15
)
sim.add_source(source)

# Add DFT monitor
wavelengths = np.linspace(1.5e-6, 1.6e-6, 11)
dft = prismo.DFTMonitor(
    center=(4e-6, 0, 0),
    size=(0, 2e-6, 0),
    frequencies=(299792458.0 / wavelengths).tolist()
)
sim.add_monitor(dft)

# Run
sim.run(time=50e-15)

# Analyze
spectrum = dft.get_power_spectrum('Ex')

Documentation

• User Guide - Tutorials and how-to guides
• API Reference - Complete API documentation
• Examples - Working example scripts

Materials

# Use pre-defined materials
si = prismo.get_material('Si')
sio2 = prismo.get_material('SiO2')
au = prismo.get_material('Au')

# List all materials
print(prismo.list_materials())

# Create custom dispersive material
material = prismo.LorentzMaterial(
    epsilon_inf=2.0,
    poles=[prismo.LorentzPole(omega_0=2e15, delta_epsilon=1.0, gamma=1e13)]
)

Mode Solver

# Solve for waveguide modes
mode_solver = prismo.ModeSolver(
    wavelength=1.55e-6,
    x=x_coords,
    y=y_coords,
    epsilon=epsilon_profile
)

modes = mode_solver.solve(num_modes=3, mode_type='TE')
fundamental = modes[0]
print(f"Effective index: {fundamental.neff.real:.4f}")

# Use mode as source
mode_source = prismo.ModeSource(mode=fundamental, direction='+x', ...)

S-Parameters

# Extract S-parameters
s_analyzer = prismo.SParameterAnalyzer(
    num_ports=2,
    frequencies=frequencies
)

# Calculate metrics
s21 = s_analyzer.get_s_parameter(1, 0)
insertion_loss = s_analyzer.get_insertion_loss_db(1, 0)

# Export to Touchstone
prismo.export_touchstone("device.s2p", frequencies, s_analyzer.s_matrix)

Data Export

# Parquet (efficient, compressed)
exporter = prismo.ParquetExporter(output_dir="./results")
exporter.export_sparameters(
    filename="device",
    frequencies=frequencies,
    sparameters={'S21': s21}
)

# CSV (universal)
csv_exporter = prismo.CSVExporter(output_dir="./results")
csv_exporter.export_spectrum(...)

Performance

Typical performance on NVIDIA A100:

CPU performance (NumPy): ~1-2% of GPU performance.

Requirements

• Python ≥3.9
• NumPy ≥1.21
• SciPy ≥1.7
• Polars ≥0.20 (for Parquet export)

Optional:

• CuPy ≥12.0 (for GPU acceleration)
• Matplotlib (for visualization)

Testing

pytest tests/

Run specific test categories:

pytest tests/test_backends.py
pytest tests/validation/

License

MIT License. See LICENSE for details.

Documentation

📚 Full Documentation on ReadTheDocs

Quick Links

• Installation Guide - Get set up quickly
• Quick Start Tutorial - Your first simulation in 5 minutes
• Tutorials - Step-by-step guides
  • Basic Simulation
  • Waveguide Coupling
  • S-Parameter Extraction
  • Parameter Optimization
• User Guide - Comprehensive documentation
  • Mode Ports
  • Boundary Conditions
  • Validation
• API Reference - Complete API documentation
• Examples - Sample code and demos

For Developers

• Architecture - Code structure and design
• Contributing - How to contribute
• Testing Guide - Testing practices
• Benchmarks - Performance metrics

Citation

If you use Prismo in academic work, please cite:

@software{prismo2025,
  author = {Kamesh, Rithul},
  title = {Prismo: Python FDTD Solver for Photonics},
  year = {2025},
  url = {https://github.com/rithulkamesh/prismo}
}

References

• Taflove, A., & Hagness, S. C. (2005). Computational Electrodynamics: The Finite-Difference Time-Domain Method. Artech House.
• Yee, K. S. (1966). "Numerical solution of initial boundary value problems involving Maxwell's equations in isotropic media." IEEE Trans. Antennas Propagation, 14(3), 302-307.