Device Geometry
Represent electromagnetic response on semiconductor, dielectric, and photonic structure volume.
Integrated design platform
Model, mesh, simulate, and optimize 3D devices in one workspace.
Planck Labs is building the end-to-end design platform for photonic, semiconductor, metasurface, and RF engineering — bringing geometry, meshing, multiphysics simulation, and design optimization into a single tool.
Method
Unknowns live where the material lives.
Radiating Kernels
Use integral operators for open-domain fields and scattering behavior.
Design Optimization
Move geometry and material parameters through optimization loops.
Why Planck
A production high-order EM solver for high-frequency, high-contrast 3D devices.
Built for inhomogeneous, high-contrast, high-frequency problems that strain mainstream methods.
(k₀h)^p error scaling
Error stays bounded per wavelength regardless of electrical size.
Material as design variable
ε(r) and μ(r) optimize alongside shape.
True jumps, unaligned interfaces
Sharp material discontinuities. No staircase, no conformal meshing.
GPU acceleration
Runs on GPU.
Method comparison
| FDTD | FEM | Surface MoM | Planck | |
|---|---|---|---|---|
| Convergence order (3D) | p = 2 (Yee) | p = 2 default | p ≈ 1.5 (RWG) | p ≥ 6 (configurable) |
| DOF increase for 10× lower error | ≈ 32× | ≈ 32× | ≈ 100× (RWG) | ≤ 3× |
| Asymptotic compute (3D) | O(N^4/3) | O(N^3/2) | O(N log N) (MLFMM) | O(N log N) |
| Mesh empty space | Yes | Yes | No | No |
| Open-domain radiation | Needs PML | Needs PML | Analytical | Analytical |
| Inhomogeneous media | Yes | Yes | Hybrid only | Native |
| Sharp material jumps | Staircase | Mesh-conformal | n/a (surface) | Unaligned, true |
| High-frequency dispersion | Grows with size | Pollution effect | Pollution-free | (k₀h)^p, pollution-free |
| GPU acceleration | Yes | Yes (recent) | Limited | Yes |
Method-class characterizations reflect well-known mathematical properties of FDTD, FEM, and MoM under default solver settings. DOF cost shown for one-decade error reduction in 3D, where N scales as F^(3/p). Asymptotic compute scaling assumes appropriate acceleration techniques (MLFMM / FMM / FFT) and well-conditioned formulations with mesh-independent iteration counts. Per-vendor higher-order or specialized options exist but are rarely run in production. Specific products and capabilities cited as of 2026; capabilities of named products may evolve. Comparison shown for the underlying solver class, not for any specific commercial implementation.
Applications
Built for small features, high frequencies, and coupled fields.
Silicon Photonics
Waveguides, bends, couplers, resonators.
Metasurfaces
Periodic media, phase masks, metalenses.
Semiconductor RF
Packages, interconnects, high-frequency devices.
Inverse Design
Geometry and material optimization across device parameters.
Workflow
From device description to optimized response.
01
Geometry
02
Materials
03
Solve
04
Optimize
Contact
Bring a representative device problem.
Geometry, material stack, frequency or wavelength band, and target response are enough to start.