Decarboxylation as a Unit Operation

What the operation is, and what it is not

The plant does not make Δ9-THC or CBD in quantity. It makes their carboxylic acids — THCA, CBDA, CBGA, CBCA and the corresponding varin-chain acids — and those acids are not the molecules that bind CB1 with useful affinity, nor are they what a total-THC calculation is based on. Decarboxylation removes the carboxyl group from the acid, releasing it as carbon dioxide and leaving the neutral cannabinoid. It happens slowly at room temperature over months, faster with light and air, and quickly at process temperatures. As an operation it belongs in the processing chain, not in the chemistry section: the question here is not what the reaction is but where in the process to put it, at what temperature and for how long, and how to prove it finished without over-running it. The reaction chemistry and the structural consequences live on cannabinoids/decarboxylation. What follows is the process. One framing note: decarboxylation is a loss of a carboxyl group from a molecule the plant supplied, and the product is the neutral cannabinoid the plant would have made on its own over time. It is not a conversion of one cannabinoid into a different one, and this shelf carries nothing of that second kind.

Sources: Veress T 1990* · Wang M 2016*

Where it sits in the process, and why that is a real choice industry practice, not published data

Decarboxylation can be run on the biomass before extraction, on the crude after extraction, or, for some products, not at all. Decarboxylating the biomass first means a bulkier, slower heat step with more uneven heat transfer through a poorly conducting bed, but it also drives off moisture and avoids putting acid cannabinoids through the extraction; it is common in ethanol and CO2 operations. Decarboxylating the crude is faster, easier to monitor by mass and by assay, and easier to control because a stirred liquid has far more uniform temperature than a bed of flower, and it is the norm ahead of distillation because the still needs a decarboxylated feed regardless: an acid cannabinoid entering a distillation flask decarboxylates there anyway, releasing CO2 that fights the vacuum, foams the charge and carries material into the head. Not decarboxylating at all is the right answer for acid-cannabinoid products — THCA diamonds, CBDA tinctures, raw-cannabinoid formulations — and for anything where the acid form is the specification. The one thing you cannot do is leave it to chance: partial, uncontrolled decarboxylation mid-process is how a batch arrives at the still with an unpredictable CO2 load and a potency number nobody can reconcile.

• On biomass: simpler workflow, drives off moisture, poor heat uniformity.
• On crude: faster, uniform, monitorable, and it is what a distillation feed requires.
• Not at all: correct for acid-form products; a deliberate specification, not an omission.
• Uncontrolled partial decarboxylation upstream is the failure mode to avoid.

Sources: Composite: extraction-equipment vendor documentation 2026* · Wang M 2016*

Kinetics: first order, steeply temperature-dependent

The decarboxylation of cannabinoid acids is generally described as first order in the acid concentration, meaning the rate at any instant is proportional to how much acid is left, so conversion follows an exponential approach to completion: each successive half-life removes half of what remains. That has two practical consequences. The first is that the last few percent take disproportionately long — getting from 90 to 99 percent conversion costs about as much time as getting from 0 to 90 percent did — which is why people who stop at a visually plausible endpoint routinely leave several percent of acid in the product. The second is that the rate constant follows the Arrhenius relationship, rising steeply and non-linearly with temperature, so a modest temperature increase buys a large rate increase: raising the temperature by about 10 °C in this region typically multiplies the rate by a factor of roughly two to three. That is the whole basis of the time/temperature trade. Perrotin-Brunel and colleagues modelled the Δ9-THC case and reported kinetic parameters for it; Veress and colleagues characterised the process in open reactors and observed that the apparent kinetics are not cleanly single-step in a real matrix; Citti and colleagues studied CBDA specifically and found its kinetics differ from THCA, which matters for a hemp processor because a curve derived from THCA is not a CBDA curve. The complication that defeats naive modelling is that the same heat also drives the loss reactions — evaporation of the neutral cannabinoid, and oxidation of Δ9-THC toward CBN — so the measured acid-to-neutral conversion curve and the measured total-cannabinoid curve diverge as the run goes on.

• First order: exponential approach to completion; the last few percent are slow.
• Arrhenius: roughly a two- to three-fold rate increase per 10 °C in this region.
• THCA and CBDA do not share a curve; a hemp processor needs the CBDA data.
• Loss reactions run in parallel, so the optimum is a maximum in yield, not the end of the acid.

Sources: Perrotin-Brunel H 2011* · Veress T 1990* · Citti C 2018* · Wang M 2016*

The time and temperature curve in practice contested

The working bands in the trade are a low-and-slow option of roughly 110-120 °C for 60-90 minutes and a hot-and-fast option of roughly 140-150 °C for 20-40 minutes, with a middle ground near 130 °C for 30-60 minutes. Low and slow preserves terpenes better (though most of the monoterpene fraction is lost at any of these temperatures unless it is being captured, which is the argument for taking a terpene cut before decarboxylating) and gives a gentler colour result; hot and fast is dramatically shorter and therefore lower in total oxidative exposure per batch despite the higher peak, which is the same total-heat-history argument that governs distillation. The honest and important caveat is that published decarboxylation curves differ substantially — not by a few minutes, but by factors — because they were measured on different things. Matrix matters: flower, kief, crude oil and a solvent solution behave differently. Moisture matters: water content changes heat transfer and appears to influence the apparent rate. Mass and vessel geometry matter enormously: an oven programme that fully decarboxylates a 200 g thin layer of crude in a tray will leave the centre of a 5 kg charge in a beaker substantially unconverted, because what the oven controller reads is not what the middle of the charge experiences. Agitation matters for the same reason. The practical conclusion is unambiguous: a processor must establish their own curve for their own matrix, mass and vessel, and verify against assay rather than trusting a number from a paper, a forum or this page.

Contested — caveat. Published decarboxylation time/temperature curves disagree substantially. Reported optima span roughly 100-160 °C and 10-180 minutes across the literature and the trade, and the disagreement is real rather than sloppy: matrix, moisture, charge mass, vessel geometry, agitation and headspace all shift the curve, and THCA and CBDA differ from each other. Any single curve presented as canonical — including the bands in this table — should be treated as a starting point for your own verification by assay, not as a specification.

Sources: Veress T 1990* · Wang M 2016* · Citti C 2018* · Composite: extraction-equipment vendor documentation 2026*

How to know it is finished

There are three levels of evidence and only the third one is proof. The crudest indicator is the visible evolution of CO2: a decarboxylating crude bubbles and foams as gas leaves, and the cessation of visible bubbling means the bulk of the reaction is done. That is a useful operator cue and it is not an endpoint, because the slow exponential tail produces gas too slowly to see while several percent of acid remains. The second level is mass loss. Full decarboxylation of an acid cannabinoid releases CO2 amounting to about 12.3 percent of the acid mass — 44 mass units of CO2 leaving a THCA molecule of about 358, and similarly for CBDA — so for a crude of known acid content the theoretical mass loss is calculable and a run that has reached it has, to first order, converted. The weakness is that mass loss also includes water and volatilised terpenes and cannabinoids, so on a real crude the number is confounded in both directions and it is a control chart rather than a proof. The third level, and the only one that settles it, is a potency assay of a representative sample showing the acid form consumed to whatever residual level your specification allows, with the neutral form present at the expected stoichiometric quantity and CBN not materially elevated. Sample properly: pull from a mixed, agitated charge, not from the top, because an unstirred vessel is stratified in both temperature and conversion. In a production setting the right pattern is to establish the curve once with time-point sampling and assay, then run to the established time and temperature with periodic verification, rather than assaying every batch to completion or assaying none.

• Cessation of visible CO2 evolution: an operator cue, never an endpoint.
• Mass loss against the roughly 12.3 percent theoretical: a good control chart, confounded by water and volatiles.
• Potency assay of the acid form: the only actual proof of conversion.
• Sample from a mixed charge — an unstirred vessel is stratified.

Sources: Wang M 2016* · Veress T 1990* · Composite: extraction-equipment vendor documentation 2026*

The 0.877 factor and total-THC arithmetic

Because the acid loses CO2 when it decarboxylates, one gram of THCA does not become one gram of Δ9-THC. The mass ratio of the neutral cannabinoid to its acid is the ratio of their molecular masses, about 314 to 358, which is approximately 0.877, and the same factor applies to the CBDA-to-CBD pair to a very close approximation. That factor is the basis of the total-THC calculation used in regulation: total THC equals the measured Δ9-THC plus 0.877 times the measured THCA, which answers the question "how much Δ9-THC would this sample contain if it were fully decarboxylated". It is the decarboxylated basis on which the US hemp programme measures compliance, which is why a hemp lot can be compliant as measured and non-compliant as consumed after heating, and why a processor must run their compliance arithmetic on the decarboxylated basis rather than on the as-received Δ9 number. The corollary inside the plant is a mass-balance one: a fully decarboxylated batch weighs less than the acid crude that went in, so a potency percentage that rises after decarboxylation is partly real concentration and partly the denominator shrinking. The full arithmetic, the rounding conventions, the measurement-uncertainty question and the worked examples are on coa/total-thc-math.

• Total THC = Δ9-THC + 0.877 × THCA. Total CBD = CBD + 0.877 × CBDA.
• 0.877 is a molecular-mass ratio, not an efficiency factor — it is not a yield.
• Regulatory compliance in the US hemp programme is on the decarboxylated basis.
• Potency rising across decarboxylation is partly the denominator, not only the numerator.

Sources: United States Department of Agriculture 2021* · Wang M 2016*

Over-decarboxylation: heat past the endpoint makes CBN

The reaction does not stop when the acid is gone, because the heat is still there and Δ9-THC has somewhere else to go. Continued thermal and oxidative exposure converts Δ9-THC to cannabinol by oxidation and aromatisation of the terpene ring, and CBN is the resulting marker. The degradation pathway has been documented since the classical storage-stability work — Fairbairn and colleagues followed the loss of THC and the appearance of CBN in stored cannabis and preparations, and Trofin and colleagues followed the same progression in oils over long-term storage — and the same chemistry runs faster and hotter in a decarboxylation vessel. The practical reading is that a CBN number on a COA is a process-history readout rather than a cultivar property: a sample with CBN materially above what its input carried has been heated, held hot, aerated, or stored badly, and the number tells you which part of your process is running long or hot. This is also why the same reaction is used deliberately: CBN is produced commercially by running exactly this oxidation on purpose, with controlled heat, time and oxygen exposure, on THC-containing material, and products/cbn-production covers it as an intentional operation. The distinction between a CBN product and an over-cooked batch is entirely whether it was specified and controlled. For a processor trying to avoid it, the levers are the familiar ones: stop at the endpoint rather than past it, keep the lowest temperature that reaches the endpoint in an acceptable time, exclude oxygen, and do not hold a finished charge hot while you deal with something else.

• Δ9-THC oxidises to CBN with continued heat, time and oxygen exposure.
• CBN on a COA reports process history, not genetics.
• The same reaction, specified and controlled, is how CBN is made deliberately.
• Stop at the endpoint; exclude air; never hold a finished charge hot.

Sources: Fairbairn JW 1976* · Trofin IG 2012* · Composite: extraction-equipment vendor documentation 2026*

Vacuum decarboxylation, foaming and headspace industry practice, not published data

Running the decarboxylation under reduced pressure changes three things for the better and one for the worse. It excludes oxygen, which directly suppresses the oxidation pathway to CBN and reduces darkening, and that is the main reason to do it. It removes the evolved CO2 continuously instead of letting it sit over the charge, which by Le Chatelier reasoning favours the forward reaction and in practice shortens the run at a given temperature. And it lets the same conversion be achieved at a lower temperature, because the removal of product gas and the exclusion of oxygen both help. What it makes worse is foaming: CO2 evolving from a viscous liquid under reduced pressure generates a far larger and more persistent foam than it does at atmospheric pressure, and that foam will climb out of the vessel and into the vacuum line if it is given the room. Managing that is straightforward mechanically and it is where most of the practical skill sits. Use a vessel with generous headspace — a third full at most, a quarter is better for the early aggressive phase — and prefer a wide, shallow charge to a deep one, because a shallow layer both foams less and transfers heat far more evenly. Ramp the vacuum rather than applying it fully at once, exactly as on a rotovap, so the initial gas burst is controlled. Stir or rotate, because agitation collapses foam, provides nucleation and eliminates the temperature stratification that otherwise leaves the middle of the charge unconverted. Fit a bump trap or a knock-out pot between the vessel and the vacuum line so that foam that does travel is caught before it reaches anything expensive. And account for the fact that vacuum will also pull off the terpene fraction, so if that fraction is wanted, it is captured in a cold trap ahead of the pump — see terpene-recovery — and not simply exhausted.

• Vacuum suppresses oxidation to CBN, removes evolved CO2, and lets the same conversion happen cooler.
• It also foams much harder: generous headspace, shallow charge, ramped vacuum, agitation, knock-out pot.
• Agitation does three jobs at once: foam control, nucleation, and killing temperature stratification.
• Vacuum decarboxylation strips the terpene fraction — capture it in a cold trap or lose it.

Sources: Composite: extraction-equipment vendor documentation 2026* · Armarego WLF 2017* · Green DW 2019*

Safety, as part of the operation

Decarboxylation is a hot operation that generates gas from a viscous liquid, and the hazards follow from exactly that. The evolved CO2 is a real volume: a kilogram of acid cannabinoid releases on the order of a hundred grams of CO2, which at room conditions is a meaningful volume of asphyxiant gas denser than air, so the vessel must never be sealed and the space must be ventilated — a closed vessel decarboxylating is a pressure vessel with no relief. That is the single most important point on this page from a safety standpoint. Foam-over under vacuum is the second: hot, sticky product travelling up a vacuum line is a burn hazard, a contamination event and, if it reaches a pump, an equipment loss, so a knock-out pot is standard rather than optional. Hot oil and hot glass do the actual injuring: charges at 110-150 °C in a beaker, flask or tray cause immediate deep burns, they do not look hot, and a viscous cannabinoid crude sticks to skin and keeps transferring heat, so heat-resistant gloves, a face shield when handling an open hot vessel, and tongs or a lifter rather than hands are required. Oven and mantle controls fail closed and run away, so a temperature limit and an independent over-temperature cutout are worth having on any vessel that will be left for ninety minutes. If the charge still carries solvent, this operation is where it will come out, in bulk, hot: a decarboxylation step is not a place to discover that the feed was not devolatilised, because a flammable vapour at 140 °C near an oven element is the worst version of that mistake. And the vapour that leaves during the run is not nothing — it is terpenes and volatilised cannabinoid, which is both a product loss and a respiratory exposure, so it is exhausted through extraction or condensed, not released into the room.

• Never seal a decarboxylating vessel. CO2 evolution in a closed vessel is an unrelieved pressure build.
• Ventilate: CO2 is an asphyxiant denser than air.
• Knock-out pot on any vacuum decarboxylation; assume foam will travel.
• Heat gloves, face shield, mechanical lifting — hot viscous crude sticks and keeps burning.
• Independent over-temperature cutout on a long unattended hold.
• Devolatilise before decarboxylating; do not discover residual solvent at 140 °C.
• Exhaust or condense the vapour — it is product and it is an exposure.

Sources: National Research Council (US) 2011* · National Fire Protection Association 2024* · Composite: extraction-equipment vendor documentation 2026*