The Chemistry of Photonics Manufacturing: How AR Waveguides & Silicon Photonics Are Cleaned, Etched & Dried

In our work supplying optical-device and photonics fabricators, one thing becomes obvious fast: the light-bending magic happens in a cleanroom, but it is built on a bucket of decidedly un-magical wet chemistry. Before a single photon travels through a waveguide, the substrate has been cleaned, coated, patterned, etched, stripped, and dried — and every one of those steps is a chemical process.

This guide walks through the chemicals that actually run a photonics line, what each one does, the grades that matter, and where the real hazards live. It is written for the process engineers, R&D chemists, and procurement teams spec'ing their supply — and, honestly, for anyone who has ever wondered what connects a $50,000 photomask to a bottle of rubbing alcohol.

Why is photonics manufacturing suddenly a high-volume chemical buyer?

Photonics manufacturing is scaling because the AI-datacenter and AR-hardware booms both hit a wall that only light can climb over. As AI clusters push past 100 terabits per second per node, copper electrical links stall on bandwidth-per-watt — so the industry is moving the interconnect into the optical domain, right up against the processor.

In early 2026, Nvidia outlined plans to use light for GPU-to-GPU communication, and analysts now describe co-packaged optics (CPO) — where the optics sit in the same package as the switching ASIC — as potentially mandatory for the largest AI systems. GlobalFoundries launched its CPO platform, SCALE, in May 2026. Meanwhile, AR eyewear runs on diffractive waveguides etched into glass or silicon carbide — a second, consumer-facing photonics wave with its own fabs. Every one of those chips and lenses is wet-processed. That is why a photonics fab looks, from the loading dock, a lot like a chemical distributor's best customer.

If you want the physics of why light is eating the interconnect — and why silicon, which does not naturally emit light, is still the substrate everyone builds on — this deep-dive from an independent semiconductor analyst is the best primer we have found:

What does the photonics wet-chemistry process actually look like?

A photonics wet process is a repeating loop of clean, coat, pattern, etch, strip, and dry. The exact recipe varies by device — a silicon-nitride waveguide is not an AR glass lens — but the chemical skeleton is remarkably consistent across the industry:

1. Piranha clean — strip all organic residue from the bare substrate (sulfuric acid + hydrogen peroxide).
2. Dehydration bake & spin-coat — drive off adsorbed water, then spin on photoresist dissolved in a carrier solvent (PGMEA).
3. Expose & develop — project the pattern, then dissolve away the exposed (or unexposed) resist in a developer.
4. Etch — transfer the pattern into the substrate by dry plasma (RIE) or wet etch; AR waveguides often use a nanoimprint imprint-to-etch route on glass or SiC.
5. Strip & clean — a second Piranha bath removes the hardened resist; an ammonium-hydroxide SC-1 clean lifts remaining particles.
6. Marangoni dry — a final DI-water rinse followed by an IPA-driven, spot-free dry.

The two Piranha cleans consume the bulk of the process-side acid, and sodium hydroxide does the highest-volume work downstream — neutralizing all of that spent acid before the water is discharged. The resist solvent and the final dry are lower-volume but zero-compromise on purity. To see the full flow inside an actual integrated-photonics foundry — wafers, lithography, and all — this short factory walkthrough is worth three minutes:

Let's take the chemistries in order.

Why does photonics fabrication start with Piranha solution?

Photonics fabrication starts with Piranha because nothing else strips organic contamination off a substrate as completely or as cleanly. Piranha solution is a mixture of concentrated sulfuric acid and hydrogen peroxide — typically around three parts acid to one part peroxide — that oxidizes and dissolves organic matter down to the molecular level, leaving a hydroxyl-terminated, hydrophilic surface that the next coating can wet perfectly.

The sulfuric acid is the volume workhorse here: it is the bulk of the bath, it provides the dehydrating, strongly acidic medium, and its heat of dilution with the peroxide is what drives the reaction. In practice, a photonics line pours far more sulfuric acid than any other single chemical, which is exactly why it is the first thing procurement locks down on contract.

If you already run a semiconductor line, this is familiar territory — we cover the same chemistry from the chip side in our guide to high-purity sulfuric, nitric, and ammonium hydroxide in semiconductor fabrication, and the peroxide half of Piranha in our hydrogen peroxide in semiconductor manufacturing guide.

How are AR waveguides different from silicon-photonic chips?

AR waveguides and silicon-photonic chips share the chemistry but differ in the substrate and the patterning route. Silicon photonics is fabricated on silicon or silicon-nitride wafers using the same lithography-and-etch flow as logic chips, because it is designed to be CMOS-compatible and made on existing fab lines. AR waveguides — the transparent lenses in smart glasses that pipe an image into your eye — are instead built on high-index optical glass (such as Borofloat 33) or silicon carbide, and are frequently patterned by nanoimprint lithography, where a master stamp presses the diffractive grating directly into a resist before an imprint-to-etch step transfers it into the glass.

What does not change is the wet chemistry that brackets every one of those steps. Both routes start from a Piranha-clean substrate, both rely on controlled caustic cleaning between steps, and both finish with an IPA-assisted spot-free dry — because a single particle or water spot on a nanometer-scale grating is a visible defect in the final image. The substrate and the stamp are exotic; the buckets of chemistry are the same ones we ship every day.

How is sodium hydroxide actually used in a photonics fab?

Sodium hydroxide's main job in a silicon-photonics or semiconductor fab is not on the wafer at all — it is in the waste-treatment system, neutralizing the spent acid baths to a compliant pH before the water is discharged. A line that runs Piranha and other strong-acid steps produces a large, continuous stream of acidic wastewater, and NaOH (commonly dosed as a 50% solution) is one of the two most common chemicals used to bring that stream back toward neutral. (DAS Environmental — semiconductor wastewater)

This surprises people who expect the strong base to be a surface cleaner — but on the active device, it is the opposite. Sodium is a classic contaminant: sodium ions are small and mobile in silicon dioxide even at room temperature, and they shift MOSFET threshold voltages, so fabs work hard to keep sodium away from the wafer. The particle-removing surface clean instead uses ammonium-hydroxide-based SC-1 (the RCA clean: NH₄OH + H₂O₂ + water), and a second HCl/H₂O₂ step (SC-2) is run specifically to strip alkali ions like sodium back off. (Biolin Scientific, RCA clean) We cover that acid-and-base surface chemistry — which we also supply — in our semiconductor fabrication chemicals guide.

Where NaOH does touch a substrate directly is glass: caustic cleaning and controlled etching of optical glass and labware, where the sodium-sensitivity that rules it out on silicon MOS is not a concern. Our Sodium Hydroxide Etching & Optical Cleaning Guide covers that glass-side use in detail.

Why is isopropyl alcohol used to dry wafers and optics?

Isopropyl alcohol dries optical surfaces without leaving water spots because it exploits the Marangoni effect. Water has a surface tension of about 72 mN/m; IPA is about 23 mN/m. When an IPA-rich vapor meets a wet substrate as it is slowly withdrawn from a rinse, the alcohol adsorbs at the top of the meniscus and drops the surface tension there; the surrounding water is still at 72 mN/m, so a gradient forms and the water film is pulled off the surface as a clean sheet rather than beading and evaporating into spots. (R&D World, surface-tension-gradient drying)

The spots matter because when ultrapure water simply evaporates off a wafer, it does not leave nothing behind: dissolved oxygen and trace silica concentrate at the drying front and leave microscopic watermarks that can mask an etch, block a deposition, or fail inspection. Marangoni drying removes the water before it can evaporate, which is why it is the standard finish on high-value optics and wafers.

On a nanometer-scale waveguide or a mirror-flat lens, a single dried-on water spot is a defect. That is why the final step of almost every photonics wet process is a DI-water rinse followed by an IPA-assisted dry. It is the same physics we detail for chips in our isopropyl alcohol in semiconductor manufacturing guide, and the same reason IPA lives on every electronics repair bench.

What photoresist solvent do photonics fabs use — and does Alliance sell it?

The photoresist solvent most photonics and semiconductor fabs reach for is PGMEA — propylene glycol monomethyl ether acetate, CAS 108-65-6. It is a slow-evaporating solvent carrying both ether and ester groups, which makes it excellent at dissolving resist polymers for spin-coating and at cleanly removing the thick bead of resist that builds up at a wafer's edge (edge-bead removal, or EBR).

Semiconductor-grade PGMEA is usually stabilized with a small amount of BHT (butylated hydroxytoluene) to suppress peroxide formation during storage and shipping — an important spec detail when you buy it. Reference specs: Eastman EastaPure PM Acetate and Sigma-Aldrich PGMEA.

What chemical grade do photonics and optics applications actually need?

The right grade depends entirely on where the chemical touches the device, and it pays to be precise about this rather than over-buying purity you will never use or under-spec'ing a step that matters. Front-end wet steps that contact the active optical surface demand ultrapure, electronic/semiconductor-grade chemistry with tightly controlled particle and metal-ion counts. Tooling cleaning, parts and fixture prep, reclaim, R&D, and prototyping have far more forgiving requirements.

That grade discipline is the same logic we apply to wafer dewaxing in our d-limonene for semiconductor wafer dewaxing guide: match the chemical to the step, not to the marketing.

What do teams get wrong when sourcing photonics chemistry?

The most expensive mistakes in photonics sourcing are rarely about the chemical itself — they are about grade, storage, and substitution. Four recur often enough to be worth naming:

• Over-spec'ing purity for non-critical steps. Buying electronic-grade chemistry to clean fixtures and tooling burns budget on parts-per-billion metal control that never touches the device. Reserve ultrapure for front-end surfaces; use ACS or technical grade everywhere else.
• Under-spec'ing the step that matters. The opposite error — running a technical-grade chemical against the active optical surface — shows up later as yield loss that is far more expensive than the chemical you saved on.
• Sealing used Piranha. It keeps evolving oxygen gas; a capped bottle is a pressure hazard. Used Piranha is vented and neutralized per procedure, never stored tight.
• Substituting one glycol ether for another. PGMEA, PGME, and Glycol Ether EE are chemically distinct with different solvency and safety profiles. They are not drop-in replacements — match the resist and the process, not the shelf.

Underneath all four is the same principle: water and chemical purity are process parameters, not commodities. Treat DI-water quality, grade selection, and storage as part of the recipe, and the chemistry stops being the variable that costs you a wafer.

Key numbers & sources