Is Pyrolysis an Incinerator? EPA Buried the Question in an Air Curtain Rule.

You have to read a long way into a federal proposal about air curtain incinerators to find the sentence that decides whether a pyrolysis reactor needs a Title V permit. On 20 March 2026, EPA published a rule package whose title is entirely about burning wood waste, yard waste and clean lumber. Tucked inside it, the agency asked whether it should strike the words "pyrolysis/combustion units" out of the Other Solid Waste Incineration rules. Comments closed on 4 May. No final rule has issued. So the whole pyrolysis incinerator classification question, Section 129 or Section 111, MACT-grade limits or not, continuous monitors or not, is sitting open in docket EPA-HQ-OAR-2025-0068, riding on a housekeeping proposal about air curtain burners.

Then on 2 July, EPA proposed the machinery to enforce the rules that word still sits in. The federal plan would impose the revised OSWI emission guidelines on existing units in states and tribal areas that haven't submitted approvable plans of their own. By EPA's count it reaches about sixty units across twelve states. It does not contain the word pyrolysis once.

I build classifiers for a living, so let me say what this looks like from where I sit. It isn't a legal problem yet. It's a labeling problem, and it's the same bug I've shipped myself.

The Definition Was Never Operationalized

The definition of "municipal waste combustion unit" is the hook. It's the thing that pulls a small unit into Clean Air Act Section 129, and the list of unit types it covers includes "pyrolysis/combustion units." The rules never define that phrase. Not in 2005, when the standards were written. Not since.

EPA has admitted this in writing. In its 2021 advance notice on pyrolysis and gasification, the agency said the OSWI rule uses the term but does not actually define it, and conceded there is "considerable confusion in the regulated community regarding the applicability of CAA section 129 to pyrolysis and gasification units." The closest thing to a definition anyone has offered is a description of what such a unit does:

Units that produce gases, liquids, or solids through the heating of MSW, and the gases, liquids, or solids produced are combusted and emissions vented to the atmosphere. (EPA, Advance Notice of Proposed Rulemaking, 86 FR 50296, September 2021)

We built a computer-vision training set for a sorting line in 2023: 38k labeled frames, contaminated against recoverable. We wrote the annotation guide in an afternoon, which was the mistake. Nobody had written down where contaminated ends and recoverable begins. A lightly greasy pizza box. A fibre tray somewhere around twenty percent moisture (annotators are people, and people guess). Round one came back with the boundary frames splitting about two to one between labelers, and the frames the model got most confidently wrong turned out to be precisely the ones the humans had argued over. That 2023 guide cost us six weeks of rework: we threw out the first labeling round and started over. Two annotation rounds, and not one line of model code changed.

The lesson stuck: label quality is the ceiling; the architecture is only the floor. A label with no operational test behind it isn't a label, it's a vibe with a name on it, and everything downstream inherits the ambiguity. EPA has been running the pyrolysis question on a vibe since 2005.

And the agency keeps trying to fix it, then stopping. It proposed removing the pyrolysis reference in 2020, issued the advance notice in 2021, then proposed to withdraw its own 2020 change in June 2023. Now it's back, folded into the air curtain package. Third attempt in six years, and the phrase is still on the books.

Meanwhile the plants get built. EPA's own 2021 table of commercial and pilot-scale pyrolysis and gasification systems operating or near operational ran to roughly forty facilities, and the agency flagged its list as possibly incomplete. Every one permitted under somebody's reading of a word EPA won't define.

Ten Tons a Day Decides More Than Section 129 Does

Now the part almost nobody in the pyrolysis incinerator classification fight talks about, because it isn't in the fight. Inside the OSWI rules there's a throughput cliff, and it moves more money than the section number does.

Under the limits EPA's July proposal would apply to existing units, a unit rated at ten tons a day or less draws a carbon monoxide limit of 220 ppmvd and a particulate limit of 280 mg/dscm. Rate that same unit above ten tons a day and CO tightens to 40 while PM drops to 30, roughly five times stricter on carbon monoxide and close to an order of magnitude on particulate. And above ten tons a day you must install, calibrate, maintain and operate continuous emission monitors for carbon monoxide and oxygen.

That's the line that rewrites your capital plan. Below it you're stack-testing on a schedule. Above it you're running an analyzer bench, an ABB or Emerson NDIR on the carbon monoxide channel, a paramagnetic or zirconia cell for oxygen (the zirconia cells drift less but hate reducing atmospheres, which is a funny thing to say about a pyrolysis plant), plus a data acquisition system, calibration gas, daily drift checks and an annual relative accuracy test audit. That isn't a line item on the capex sheet so much as a permanent headcount. Operators already running waste intelligence software usually find their CEMS data in a separate silo, answering to a separate auditor.

Most pyrolysis demonstration plants land right on top of that cliff. Ten to twenty-five tons a day is the commercial-demo sweet spot, big enough to prove the mass balance, small enough to finance. So the nameplate figure your process engineer picked to make the pro forma close is, quietly, the figure that decides whether you own a CEMS for twenty years. Did anyone on your development team know that when they sized the retort?

One precision, because it changes who is actually exposed. The July federal plan reaches units that commenced construction on or before 31 August 2020. A pyrolysis line commissioned in 2024 isn't in that net at all. It's in the new-source net instead, the performance standards at 40 CFR Part 60 Subpart EEEE, which is the same rulebook through a different door. The definitional strike-out proposed in the March docket would pull the hook out of both doors at once.

The clock, such as it is: states and tribes owe EPA their plans by 30 June 2027, and units caught by the federal plan must comply by 1 July 2030. Comments on that plan close 17 August 2026. Comments on whether pyrolysis belongs in the category at all closed back in May, and the answer still hasn't come. So developers are being asked to comment on the enforcement of a rule whose applicability to them is undecided.

And gasification rides along in the same net, on the same undefined phrase. Which is strange, given two processes that differ enormously in oxygen, in temperature, and in what comes out the other end. The Clean Air Act doesn't care about any of that. It cares whether you burn the products and vent them.

Your CEMS Log Is Evidence in Your Own Classification Fight

Go back to EPA's physics. In that 2021 notice the agency wrote that unlike combustion, pyrolysis is endothermic and doesn't require added oxygen, and that the partial pressure of oxygen through the process "is maintained close to zero."

That is not a legal test. It's an instrument reading. Oxygen partial pressure near zero is a number an analyzer produces, and the OSWI rules oblige you to buy one. So the rule that classifies you requires you to install the very sensor that measures whether the classification is true. I've never seen that anywhere else in air regulation.

Actually, that's too neat, and I want to correct it before somebody builds a permitting strategy on it. Your compliance oxygen analyzer sits in the stack. The stack is downstream of the thermal oxidizer, or the flare, or whatever burns your non-condensable gas. It never sees the inside of the retort. It reads the exhaust of the thing you burned, where oxygen is nowhere near zero, because combustion needs air.

Which turns out to be the entire point. EPA's trigger was never the reactor. Read the description again: the gases produced "are combusted and emissions vented to the atmosphere." The retort is innocent. The oxidizer is the defendant.

So the classification turns on a decision your process engineer already made, on thermodynamic grounds, years before a lawyer looked at it. Do you burn your own non-condensable off-gas to close the heat balance? Almost everyone does, because that gas is free and the alternative is buying natural gas to hold the retort at temperature, and margins on pyrolysis oil are thin enough that the fuel bill decides projects. But burn it, and you're operating a combustion device with a vented stack, bolted to a unit that heats municipal solid waste. Export your oil, buy gas for process heat, and your argument that you're a chemical manufacturer gets considerably stronger while your opex gets worse.

Nobody has priced that trade honestly. Classification isn't a posture you adopt once the permit writer calls, it's an energy-balance decision made at FEED, and the record of it is your stack.

Where this doesn't hold, and it doesn't hold everywhere. If your unit is genuinely closed-loop, product gas piped into a synthesis train, nothing vented, no combustion device on the boundary, the argument gets much easier, and a handful of plants are built that way. Below ten tons a day the monitoring exposure largely evaporates even where the classification doesn't. And roughly two dozen states have passed advanced recycling laws treating pyrolysis as manufacturing rather than waste management, which is where I've watched developers walk into a trap. Those are state solid-waste statutes. A state manufacturing designation does not lift a federal Section 129 obligation. Both can sit on the same facility, and on the current wording they do. If EPA moves pyrolysis to Section 111, states stay free to be stricter, and your Title V exposure won't vanish because a federal category changed. I'm an engineer, not your counsel. But the state law you're leaning on may not reach the rule you should be worried about.

Environmental groups reading the March docket argue that Section 129 is the only Clean Air Act authority currently reaching plastic pyrolysis at all, and that deleting the phrase leaves these plants federally unregulated for air. Industry reads the same docket as the correction of a category error, a rule written in the 1990s for two-chamber starved-air incinerators and catching modern reactors by wording rather than design. Both readings can't be right. EPA has spent six years declining to choose, which leaves pyrolysis incinerator classification where it stood in 2005 and anyone building zero-waste-to-landfill solutions around a thermal conversion unit underwriting an advanced recycling air permit path that doesn't exist yet.

If I were sitting in a pyrolysis control room this quarter, I'd do one cheap thing. Log the stack oxygen and carbon monoxide channels at full resolution, timestamped, and archive the raw traces, not the rolled fifteen-minute averages your data system keeps by default. Nobody has asked you for that record. Somebody will, around 2030, and the version you assemble after the request arrives is worth precisely nothing.

Sources & Notes

The definitional question, the "pyrolysis/combustion units" language and the May comment close come out of EPA's March 2026 air curtain incinerator proposal (docket EPA-HQ-OAR-2025-0068), which is where the agency parked it.

Emission limits, the ten-ton-a-day CEMS trigger, the sixty-unit inventory and the 2027 and 2030 deadlines are from the federal plan EPA proposed on 2 July 2026, docket EPA-HQ-OAR-2025-3028. Comments there close 17 August.

EPA's admission that it never defined the phrase, the near-zero-oxygen description and the roughly forty-facility table sit in the 2021 advance notice at 86 FR 50296. Broader rule history, including the 2025 revision the federal plan implements, sits on EPA's OSWI page.

The 2023 annotation figures are ours. The count of states with advanced recycling statutes is approximate and still moving; check your own.