ArXiv briefing · diamond photonics · process control · metrology

Plasma Etch Process Optimization for Photonic-Grade Diamond-on-Insulator Substrates and Thickness Evaluation using Colorimetry

arXiv:2606.20412 is a fabrication paper with a practical punchline: thin a directly bonded single-crystal diamond membrane into a photonic-grade diamond-on-insulator substrate without transfers, under-etches, or sacrificial gymnastics — with a colorimetry-based thickness check that tracks WLI at 5 nm resolution.

Paper: arXiv:2606.20412

A dark technical banner in the same treatment as the other briefings: gold headline, schematic center, subtle dot field, and a small footer marker.

The authors report a recipe that preserves diamond bonding and surface quality, reaches a thickness of 300 nm or below over a 0.5 × 0.5 mm² uniform area, and keeps roughness under 0.5 nm. They also show a colorimetry-based thickness estimator that tracks white-light interferometry and gives a useful 5 nm resolution from microscope images.

Why this matters

Diamond photonics has never lacked ambition. What it often lacks is a substrate route that plays nicely with manufacturing. Color centers, waveguides, and chiplets all want a low-loss platform, but the substrate stack has to be thin, uniform, and mechanically credible enough to survive the rest of the flow. That is the real bottleneck here.

This paper is interesting because it attacks the boring part directly: plasma etch process development. The result is not a one-off demo on an exotic corner case. It is a recipe aimed at turning direct-bonded diamond membranes into usable thin-film DOI substrates.

The story is not just “we etched diamond thinner.” It is “we found a process window that still looks manufacturable.”

The process in one sentence

The core move is an ICP-RIE recipe tuned to thin a bonded (100) single-crystal diamond membrane while keeping the interface intact and the surface smooth enough for photonic devices.

Key engineering tradeoff: aggressive enough etching to thin a 10 µm plate all the way down, but controlled enough to avoid destroying bonding quality or turning the surface into a scattering problem.

What the paper shows

Direct-bonded diamond membranes thinner than 50 µm can be converted into large-area thin-film DOI substrates.
The authors demonstrate a free-standing photonic chiplet made from the resulting substrate.
Diamond thickness of ≤300 nm is reached from a 10 µm starting plate.
Surface roughness stays below 0.5 nm, which is the kind of number photonics people actually care about.
The workflow stays within a standard two-step lithography flow, which is a big deal for adoption.

A simple thickness proxy

The colorimetry contribution is the quiet second half of the paper. In spirit, the microscope image color is being used as a proxy for optical thickness, with the spectral response translating into a thickness estimate:

\hat{t} = f(\mathbf{c}, I_{\mathrm{illum}}, S_{\mathrm{ref}})

That is deliberately generic, because the implementation is empirical rather than a closed-form universal law. The point is practical: the color field of the microscope image is mapped to thickness quickly enough that you can evaluate large areas without waiting on metrology bottlenecks, and the result matches white-light interferometry closely enough to be useful.

\Delta t_{\max} = \max_{(x,y)\in A} t(x,y) - \min_{(x,y)\in A} t(x,y)

For a device substrate, the relevant number is not only the mean thickness but the span across the useful area. That is why the reported 0.5 × 0.5 mm² uniform patch matters. In a photonics flow, the uniformity constraint is often what separates a publishable wafer from a usable one.

Why the color check is valuable

In a fabrication line, the expensive part is often not the measurement itself. It is the time cost of measuring every region you care about. A microscope image is already sitting there. If its color variation can be calibrated against thickness, the same image becomes an early-stage screening tool for etch uniformity, yield, and process drift.

That makes the paper more than a materials result. It is a process-control result. The substrate is the outcome, but the metrology path is what makes the workflow scalable.

\sigma_{\mathrm{rms}} < 0.5\,\mathrm{nm} \qquad \text{and} \qquad t \le 300\,\mathrm{nm}

My read

The strongest part of the work is its restraint. It does not promise some magical diamond transfer stack or a fragile ultra-specialized pathway. It builds around a direct-bonded starting point, keeps the process conventional, and reports numbers that are immediately legible to photonics engineers.

If you are working on diamond photonics, this is the kind of result that changes the conversion rate from “interesting demo” to “something we can actually fabricate repeatedly.”

That is the Meridian-style criterion: not novelty for its own sake, but a route that can survive the first manufacturing conversation.

What I’d keep in mind

There are three practical numbers to remember from this paper: ≤300 nm thickness, <0.5 nm roughness, and 5 nm colorimetric resolution. Those numbers are not decorative. They define the process window, the photonic surface quality, and the usefulness of the fast metrology path.

That is also why the paper reads as a manufacturing-enabling result rather than just a characterization exercise. The value is in the full stack: etch recipe, surface quality, bonded starting point, and rapid thickness estimation.

Takeaway

Photonic-grade DOI is hard because the substrate is the device platform, not just a passive carrier. This paper shows a credible route to thin, uniform, smooth diamond-on-insulator material and adds a lightweight colorimetry check that can speed up the whole loop.

That combination is the story: better substrate formation, simpler device flow, and faster thickness validation.

Reference: arXiv:2606.20412 · Plasma Etch Process Optimization for Photonic-Grade Diamond-on-Insulator Substrates and Thickness Evaluation using Colorimetry