ENEE 684 Introduction to Photonics, Lecture 11 Dielectric Waveguides

Overview

Ergun Simsek introduces dielectric-waveguide design through integrated-photonics examples and numerical mode solving, demonstrating how Tidy3D can model waveguide geometries, materials, boundaries, and TE/TM modes. He then develops the finite-element method (FEM) eigenvalue formulation, explaining triangular meshes, Nédélec edge elements, effective-index estimates, and how mode solutions support dispersion calculations and tapered-waveguide design.

Key takeaways

A practical waveguide mode solver needs geometry, material permittivity, mesh resolution, boundary conditions, a search region, and often an approximate effective index to find the desired modes efficiently.
In a dielectric interface, tangential electromagnetic field components are continuous while normal components may be discontinuous; Nédélec edge elements build the tangential continuity requirement into FEM.
Lithium niobate's anisotropy makes crystal-axis orientation part of the simulation setup, and rib sidewall geometry can materially affect mode confinement and dispersion.
The effective index as a function of wavelength determines β and enables numerical estimates of group velocity and group-velocity dispersion, supporting dispersion-engineered photonic designs.
Tapered waveguides improve coupling by changing cross-section gradually; inverse tapers can also facilitate mode conversion, but high transmission requires sufficiently gradual transitions.
Mesh refinement improves numerical accuracy but increases memory and computation requirements, so waveguide simulations must balance resolution against available hardware.

Chapters

0:00 Integrated Photonics Uses Waveguides to Route and Filter Light

Dielectric waveguides carry light between components on photonic integrated chips and couple chip light into optical fibers.
Examples include straight and bent guides, periodic grating couplers, and ring resonators used for filtering and wavelength-division multiplexing.
Silicon waveguides with silicon-dioxide cladding are common; silicon nitride offers lower loss and is useful for optical frequency-comb applications.

5:00 Tidy3D and Google Colab Provide Cloud-Based Mode-Solver Access

Simsek contrasts analytically solvable dielectric slabs with practical geometries that generally require numerical solvers.
Tidy3D offers a cloud-based workflow, while Google Colab lets users run its Python tools without installing Python locally.
The course mentions 50 Tidy3D credits for basic simulations and distinguishes its FDTD platform from the frequency-domain mode solver used in the examples.

10:00 Define Waveguide Geometry, Mesh, and Dielectric Materials

A 2D simulation specifies the computational-domain dimensions, background permittivity, waveguide geometry, and operating wavelength.
The example uses a rectangular guide about 1.6 μm wide and 700 nm tall, with silicon dioxide as the background.
Mesh resolution can be set by a target cell size or by steps per wavelength; the example uses 80 steps per wavelength.

15:00 Set Lithium Niobate Crystal Orientation and Solve Guided Modes

Tidy3D's material library includes lithium niobate, an anisotropic electro-optic material whose refractive index depends on crystal orientation.
The optical-axis index identifies the crystal orientation; the lecture uses Python's zero-based indexing when selecting an axis.
The mode-solver setup defines a transverse search plane, requests four modes, and supplies an estimated effective index to guide the numerical search.

20:00 Identify TE and TM Modes from Field Components

In the example, the first two modes have effective indices of approximately 1.79 and 1.75; the first is predominantly TE and the second predominantly TM.
The lecture identifies TE with a dominant transverse electric-field component and TM with a different dominant component, while cautioning that solver conventions matter.
Field plots illustrate electromagnetic boundary conditions: tangential field components remain continuous, while normal components can jump across a dielectric interface.

25:00 Build Substrates and Layered Lithium-Niobate Rib Waveguides

A substrate is added as a separate geometry and material; in Tidy3D, later-listed structures overwrite earlier ones where they overlap.
Lithium-niobate rib guides have sloped sidewalls because vertical etching is difficult, making sidewall angle an important design parameter.
The example uses Tidy3D's anisotropic material model and finds several closely spaced modes, including field distributions that indicate poor confinement.

30:00 Engineer Dispersion and Account for Fabrication Variation

Dispersion engineering balances a material's dispersion against geometry-dependent waveguide dispersion by choosing materials and cross-sectional dimensions.
Effective index versus wavelength yields propagation constant β, from which group velocity and group-velocity dispersion can be calculated.
The lecture describes a fabrication-metrology challenge: electron-beam lithography and SEM measurements can disagree by tens of nanometers, affecting waveguide dimensions.

35:00 Tapered Waveguides Bridge Fiber and Chip Modes

The lecture sets three goals: explain FEM mode solving, introduce the approximate mode-expansion method, and study tapered and inverse-tapered structures.
A taper gradually changes waveguide width to improve coupling across the large mode-size mismatch between an optical fiber and a chip guide.
Linear, quadratic, and exponential tapers are compared; sufficiently gradual transitions can achieve high transmission, while inverse tapers also support mode conversion.

40:00 Convert Maxwell’s Equations into a Waveguide Eigenproblem

The starting point is the vector wave equation derived from Maxwell's equations, with spatially varying permittivity and permeability.
For propagation along z, the field is separated into a transverse profile and a propagation factor involving β, reducing the problem to a 2D cross-section.
FEM represents the cross-section with triangular elements and assigns each triangle the permittivity of its material region.

45:00 Why Material Gradients Matter in the FEM Formulation

When material parameters such as permeability vary with position, derivative operations must account for those spatial variations.
The FEM formulation incorporates the material-dependent terms rather than assuming a homogeneous vacuum-like domain.
The resulting field equations are assembled using basis functions, converting the continuous waveguide problem into a numerical system.

50:00 Use Triangular Basis Functions to Approximate Electric Fields

A one-dimensional illustration shows how weighted triangular basis functions interpolate nodal field values to approximate a continuous electric-field profile.
For a 2D FEM mesh, element coordinates and triangle connectivity determine how local basis functions fit together.
Nédélec-type basis functions are selected to represent electromagnetic fields while respecting interface continuity conditions.

55:00 Nédélec Edge Elements Enforce Electromagnetic Continuity

Transverse field components Ex and Ey are represented on triangle edges, while the longitudinal component Ez is represented at nodes.
This placement enforces tangential-field continuity across neighboring elements and permits normal-component jumps at dielectric interfaces.
The edge-element construction helps suppress nonphysical spurious modes that can appear in less constrained formulations.

1:00:00 Mesh Resolution, Boundaries, and Effective-Index Eigenvalues

The lecture recommends a maximum mesh size below roughly λ/10 for basic accuracy, λ/20 for improved accuracy, and much finer meshes for demanding precision.
Fine meshes increase computational cost and memory demand; the examples contrast an 8 GB computer with 32 GB and 512 GB machines.
The FEM eigenvalue gives β²/k₀² = n_eff², so solving for the eigenvalue yields both the mode field and its effective index.

1:05:00 Shifted Eigenvalue Solving and Numerical Dispersion Analysis

A large computational domain can approximate confinement with a PEC boundary when fields are negligible at its edges; PML is another boundary option but is more involved in the FEM implementation.
An effective-index estimate shifts the eigenvalue search toward the desired solution; the lecture compares this numerical strategy with the estimate supplied to Tidy3D.
Calculating effective index over a range of frequencies provides β values for numerical derivatives and dispersion estimates, including a reported anomalous-dispersion example.

1:10:00 MATLAB FEM Examples, Mode Plots, and Next Topics

The MATLAB solver builds a triangular mesh, computes modes, plots electric and magnetic fields, and classifies polarization from the relative Ex and Ey energy.
Material models in the example include silicon nitride, silicon dioxide, lithium niobate, and doped lithium niobate; the code can also calculate waveguide dispersion.
Simsek closes by previewing mode expansion for tapers and inverse tapers, followed by directional couplers and power dividers in the next lecture.

Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Ergun Simsek.