Apodization

Apodization is the smooth tailoring of a periodic structure's coupling strength along its length to produce a desired spectral response with reduced sidelobes. The term originates from optical pupil engineering (Greek: "without feet"), where soft pupil weighting removes the diffractive "feet" — sidelobes — of an Airy pattern. The same principle applies to gratings, filters, and any periodic interaction structure.

Why apodization is needed. A uniform-strength grating produces a $\text{sinc}^2$ spectral response — the Fourier transform of the rectangular grating envelope. The $\text{sinc}^2$ function has substantial sidelobes (13% peak in the first sidelobe), which cause:

• Filter crosstalk: a WDM channel filter has leak-through to adjacent channels through its sidelobes
• Reflectivity ripple: a DBR mirror has ripple in its reflectivity vs wavelength, producing non-flat laser output
• Grating-coupler nonuniform response: input grating responses have spectral ripples that complicate calibration

Apodization tapers the grating strength at the ends to remove the abrupt step in the structure's envelope; the corresponding spectral response has smoothly-falling skirts and substantially lower sidelobes.

Standard apodization profiles.

Profile	Sidelobe suppression	Bandwidth penalty	Application
Rectangular (no apodization)	None (sinc²)	Minimum bandwidth	Default; widest passband
Gaussian	$> 40$ dB sidelobes	$\sim 30$% bandwidth increase	High-performance filters, FBG sensors
Hamming	40 – 45 dB	$\sim 25$% bandwidth increase	Telecom filters; standard for ITU WDM
Blackman	$> 50$ dB	$\sim 45$% bandwidth increase	Narrowband notch filters
Tukey (raised cosine)	adjustable	adjustable	General-purpose engineered response
Kaiser	adjustable	adjustable	Optimizes mainlobe-sidelobe tradeoff
Bartlett (triangular)	$\sim 25$ dB	$\sim 50$% bandwidth increase	Simple cases

Implementation methods.

Method	What varies	Constraint
Period chirp	Grating period $\Lambda(z)$	Mainly for bandwidth shaping, not sidelobe suppression
Filling factor apodization	$f(z) = w_\text{tooth}(z)/\Lambda$	Most common in silicon photonics; uses standard lithography
Etch depth modulation	Tooth depth varies along length	Requires multi-step etch; rarely used
Coupled-grating apodization	Two parallel waveguides with varying coupling	Used in resonator-filter designs
Phase-mask apodization	Periodic-mask amplitude varies along length	Standard in FBG (fiber Bragg grating) fabrication

Example: silicon photonic grating coupler apodization.

A uniform-period grating coupler has a Gaussian-shaped output field but with a sharp leading edge. Apodizing the grating filling factor to ramp gradually from $f = 0.3$ at the chip edge to $f = 0.5$ at the center produces a smoother field profile matching a Gaussian fiber mode and improves coupling efficiency by 1 – 2 dB.

Example: fiber Bragg grating apodization.

FBG sensors require very narrow-linewidth reflection peaks for high sensor sensitivity. Gaussian-apodized FBG profiles produce $-3$ dB linewidths of $\sim 0.1$ nm with $> 30$ dB sidelobe suppression, compared to 0.05 nm linewidth and $-13$ dB sidelobes for uniform gratings. The sensor's resolution and discrimination improve in proportion.

Computational design. Modern apodized gratings are designed by inverse-design or transfer-matrix optimization:

1. Specify the desired spectral response (e.g., flat-top passband with $-50$ dB sidelobes)
2. Compute the required reflectivity profile $r(\lambda)$
3. Inverse-Fourier transform to get the spatial coupling profile $\kappa(z)$
4. Map $\kappa(z)$ onto a physically realizable structure (filling factor variation, period variation, or both)
5. Forward-simulate to verify response and iterate

Practical limits.

• Lithographic minimum feature size: Apodization requires varying tooth widths; some apodized designs would require sub-100 nm features at the grating ends, beyond the lithographic capability of standard DUV processes.
• Process tolerance: Apodized designs are sensitive to fabrication variation. A grating with $f$ varying from 0.3 to 0.5 has tighter relative tolerance than a uniform $f = 0.5$ grating.
• Insertion loss tradeoff: Apodization typically reduces the peak coupling/reflectivity slightly (by 0.2 – 0.5 dB for typical Gaussian apodization) in exchange for sidelobe suppression.

References: Othonos & Kalli, Fiber Bragg Gratings: Fundamentals and Applications in Telecommunications and Sensing (Artech House, 1999), Ch. 4 for FBG apodization; Chrostowski & Hochberg, Silicon Photonics Design, Ch. 4 for silicon photonic grating apodization.