Decaf process language: what a label can prove

Every commercial route works on green coffee

Caffeine is already in the seed before the roast, and roasting does not remove it — which is why decaffeination is done at the green stage, on dried beans, before they go anywhere near a roaster. That single fact fixes the starting load: arabica green coffee carries roughly 1.0 to 1.5 per cent caffeine by dry weight, and robusta about 1.6 to 2.5 per cent, with the species average sitting near 1.2 and 2.2 per cent respectively.

So the job description is precise. A plant has to take a bean holding the equivalent of 12 grams of caffeine per kilogram of dry matter and drive that down to a fraction of a gram, without destroying the cell structure that the roast will later need to expand.

Solvent routes: direct and indirect

Two solvents dominate this family: methylene chloride, also called dichloromethane, and ethyl acetate. In the direct method the beans are steamed first, which opens the cell walls, and then washed with the solvent so that caffeine migrates out of the bean and into the solvent phase; the beans are steamed again to strip the remaining solvent, and re-dried.

The indirect method inverts the order. Beans are soaked in water until the caffeine dissolves into the water, the caffeine-loaded water is then treated with solvent to extract the caffeine from it, and the cleaned water — already saturated with coffee solubles — is returned to the next batch. The beans themselves never contact the solvent in this version, which is why it is sometimes marketed separately.

Residues are regulated rather than assumed away. The US food-additive rules permit methylene chloride for coffee decaffeination and cap the residue in the finished roasted coffee at 10 parts per million; the European Union regulates extraction solvents under Directive 2009/32/EC, which sets residue limits for the solvents permitted in food production.

Water routes: separation by adsorption, not by solvent

Water-process decaffeination keeps the same first step as the indirect solvent method — soak the green beans in water until caffeine leaves the bean — and changes what happens to that water. Instead of solvent, the caffeine is captured on activated carbon or a resin bed, and the water is returned to the beans.

The trick is saturation. Fresh water would strip flavour compounds along with the caffeine, so the process water is loaded with coffee solubles first, which leaves the caffeine as the component that is still undersaturated and therefore the one that keeps migrating out. Commercial versions of this route, including the trademarked Swiss Water Process, run as repeated soak-and-filter cycles rather than a single pass, which is why they take longer than a solvent plant's contact stage.

Supercritical carbon dioxide

Carbon dioxide becomes supercritical above 31.1 °C and 73.8 bar, the point at which it stops behaving like a gas and a liquid separately. In that state it penetrates the green bean like a gas and dissolves caffeine like a liquid, with a selectivity that leaves most of the coffee's non-volatile matrix alone, and it leaves no solvent residue because the solvent is carbon dioxide.

Industrial decaffeination runs well above the critical point rather than at it, in the hundreds of bar, because caffeine's solubility in supercritical CO₂ rises steeply with pressure. Caffeine is recovered from the CO₂ by washing with water or adsorbing it on carbon, and the carbon dioxide is compressed and recycled into the next batch, which is the main reason the route is capital-heavy and water-light.

What "decaffeinated" is allowed to mean

Two different ceilings apply, and they are not the same number. Within the European Union, Directive 1999/4/EC sets the caffeine content of decaffeinated coffee extract at no more than 0.3 per cent of the dry matter — a rule written for soluble coffee, where the dry matter is the powder rather than the bean. For roasted coffee, the working convention across European and North American markets is 0.1 per cent caffeine on a dry basis.

Run that convention against the starting load and the arithmetic of the process becomes visible. Driving arabica from about 1.2 per cent to 0.1 per cent is a reduction of more than 90 per cent, and the process specifications quoted by commercial decaffeinators are written to remove 97 per cent or more. A decaf bag therefore describes a large, measured reduction to a defined ceiling, not the absence of the molecule.

What the label does not prove

Three things sit outside the process word. First, trademarks are not routes: a protected name identifies the plant that ran the process, while "water processed" and "CO₂" name the chemistry, so two bags can carry different words and the same underlying separation step. Second, decaffeination removes more than caffeine — some water-soluble chlorogenic acids and other low-molecular compounds leave with it, and the size of that loss tracks contact time and the number of cycles rather than the label. Third, caffeine per serving is still set at the brewing stage: a 25 mL decaf espresso and a 250 mL decaf filter mug are different doses of a low-caffeine drink, and the published ratio between them follows serving volume rather than the decaf process.

What the label does establish is narrower and firmer: which solvent family, if any, was permitted to contact the coffee, and therefore which residue limit a regulator has to enforce against the bag. That is enough to choose between the routes on grounds that can be checked.