WiTwin Maxwell

Overview#

WiTwin Maxwell is a differentiable full-wave electromagnetic solver. The FDTD field-update loops run as GPU kernels compiled through Slang, while the entire workflow — scene definition, simulation dispatch, result access, and automatic differentiation — stays inside standard PyTorch.

What It Provides#

• FDTD time-domain solver with monitor extraction and multi-frequency DFT sampling
• Differentiable adjoint backward support for inverse design
• PyTorch-native interface — scenes, sources, monitors, and results are all PyTorch-friendly
• SceneModule pattern for wrapping differentiable scenes as torch.nn.Module
• MaterialRegion for trainable density-based material optimization

Quick Links#

When to Use WiTwin Maxwell#

Maxwell solves the full electromagnetic wave equation using FDTD. Use it when you need:

• Accurate near-field EM simulation
• Differentiable optimization of material or geometry parameters
• Multi-frequency DFT field extraction
• Adjoint-based inverse design

If you need a high-level radio propagation simulator instead, see WiTwin Channel.

Workflow#

Every Maxwell workflow follows the same pattern:

import witwin.maxwell as mw
 
scene = mw.Scene(domain=..., grid=..., boundary=...)
scene.add_structure(...)
scene.add_source(...)
scene.add_monitor(...)
 
sim = mw.Simulation.fdtd(scene, frequencies=[1.0e9], run_time=mw.TimeConfig(time_steps=200))
result = sim.run()