Designing a Zero-Dispersion Point at 4 μm: A Workflow

Supercontinuum in the mid-IR lives or dies on dispersion engineering. Here’s the workflow I converged on during my Ph.D. for pushing a chalcogenide waveguide’s zero-dispersion wavelength (ZDW) toward a 4 μm pump.

The target

For soliton-based broadening you want to pump in the anomalous dispersion regime, close to the ZDW. Bulk $\text{As}_2\text{Se}_3$ has its material ZDW far out; the waveguide geometry has to drag waveguide dispersion back to meet your pump.

The loop

1. Parametrize the geometry. Split-core, inverted rib-core, D-shaped: each has levers (core width, etch depth, residual thickness) that shift waveguide dispersion differently.

2. Solve the modes. FEM (Femwell/COMSOL) or FDTD (Lumerical) gives effective index $n_{\text{eff}}(\lambda)$ across the band.

3. Fit and differentiate. From $n_{\text{eff}}(\lambda)$, compute

   $$\beta_2(\omega) = \frac{d^2}{d\omega^2}\big[\beta_0(\omega)\big]$$

   usually via a Taylor expansion around the pump.

4. Check nonlinearity too. A design with perfect dispersion but weak confinement loses on $\gamma$. The figure of merit is the pair $(\beta_2, \gamma)$, not either alone.

5. Propagate. Feed the fitted $\beta_k$ and $\gamma$ into a GNLSE solver and look at actual spectra: coherence, bandwidth, edge placement.

What I learned the hard way

• Fabrication tolerance beats elegance. A semi-ellipse core that needs ±10 nm precision is worse than a slightly inferior design robust to ±100 nm.
• Residual thickness is a hidden superpower in rib geometries: small changes there move the ZDW dramatically without touching lithography-critical dimensions.
• Always sanity-check with full-vector FDTD at least once per design family; scalar approximations drift badly at high index contrast ($n \approx 2.8$).

The split-core design from this workflow became my Journal of Computational Electronics paper, and the same pipeline now runs as parametric scripts rather than manual GUI clicks.