Industries / Food, farming and forestry / Food and beverage processing / Coffee roasting
Decaffeination
Decaffeination is the work of taking most of the caffeine out of green coffee before it is roasted, by soaking the moistened beans in a solvent, in water or in carbon dioxide under pressure, and then drying them.
5
Milestones achieved, from 1906 to 1985.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1906
Caffeine washed out with benzene
Ludwig Roselius, a Bremen coffee merchant, patented a way to take caffeine out of whole green beans and founded Kaffee HAG to sell the result, by the Max Planck Society's account, which dates the patent to 1905 and the company to 1906. The US patent he filed with Johann Friedrich Meyer and Karl Wimmer in May 1906 swells the beans with steam, frees the caffeine with ammonia or an acid gas, and extracts it with benzene, which it prefers because the extract left after evaporation is almost pure caffeine; a second steaming drives off the leftover solvent.
Credit
Ludwig Roselius, Johann Friedrich Meyer and Karl Wimmer (Kaffee-Handels-Aktiengesellschaft)
1941
Solvent kept off the beans
Norton Berry and Roy Walters of General Foods patented a continuous process in which green beans pass through a battery of extractors against a hot water extract that already holds about 15% coffee solids other than caffeine, so the beans lose their caffeine but keep their other solubles. The solvent, trichloroethylene, touches only the extract, in a separate tower, and the caffeine-free extract goes back to the beans. A Nestlé patent filed in 1976 cites it as the known form of this recirculating water method.
Credit
Norton E. Berry and Roy H. Walters, General Foods
1970
Carbon dioxide as solvent
Under enough heat and pressure, carbon dioxide becomes supercritical, a state between gas and liquid in which it dissolves some substances. Kurt Zosel of the Max Planck Institute for Coal Research noticed in 1967 that caffeine dissolves in it and filed his decaffeination patent in 1970. In the process, moistened beans sit in a pressure tank while carbon dioxide at about 70°C and 162 bar carries the caffeine out to an activated carbon filter, and the gas is used again. The Max Planck Society said in 2014 that the process was used worldwide on about 100,000 tonnes of raw material a year.
Credit
Kurt Zosel, Max Planck Institute for Coal Research (patents held by Studiengesellschaft Kohle)
1979
Carbon filters out the caffeine
Coffex AG of Switzerland filed a European patent in 1979, from a Swiss application of 1978, for decaffeination with water alone: a water extract of green coffee carries the caffeine out of the beans, and activated carbon takes it out of the extract. The carbon is first loaded with sugars so that it holds little but caffeine. The patent's example cut caffeine from 1.3% to 0.04%. The Swiss Water Decaffeinated Coffee Company says a water method found in Switzerland in the early 1930s was revived about 40 years later, and that in 1989 a plant using its process, then the only one in the world, was built near Vancouver, Canada; Kraft General Foods owned it at first.
Credit
Arthur Fischer and Peter Kummer (Coffex AG); Swiss Water Decaffeinated Coffee Company
1985
Methylene chloride kept for decaf
A 1976 National Cancer Institute study found that trichloroethylene caused liver cancer in mice, and the US Food and Drug Administration (FDA) proposed in 1977 to bar it from food. Methylene chloride, the solvent of what the trade calls the European method, also caused cancer in mice in a government inhalation study. In 1985 FDA still allowed it in decaf, up to 10 parts per million in the roasted or instant product, on the grounds that the risk was too small to matter. In 1987 a federal appeals court, in Public Citizen v. Young, held that the law's Delaney Clause, which bars additives found to cause cancer in animals, allows no such exception for color additives. Trichloroethylene's tolerance for decaf, 25 parts per million in ground coffee, is also still in the US rules.
Lessoneach replacement solvent was later found to cause cancer in animals, and the permission for the one still in use rests on a small-risk argument that a court rejected two years later for color additives, which has left it open to challenge since.
Credit
US Food and Drug Administration; National Cancer Institute
Watch for
Solvent petition at the FDA
Environmental Defense Fund and four other petitioners asked FDA in December 2023 to remove methylene chloride, trichloroethylene, benzene and ethylene dichloride from the US food additive rules, arguing that the Delaney Clause bars them because they cause cancer in animals. Methylene chloride is the one that matters for coffee: the National Coffee Association says most US decaf is made by the European method and that a ban "would defy science". In May 2026 FDA reopened comments and asked what manufacturers would need to phase out the uses if it grants the petition in whole or in part. On 10 November 2026, ruling on a companion petition about color additives, FDA struck out methylene chloride in fruit-marking inks under the Delaney Clause but kept the solvents for three spice and color extracts, holding that their residues there are impurities the clause does not cover.
Achieved whenFDA publishes its decision on the food additive petition in the Federal Register.
Claimed by
Environmental Defense Fund, Breast Cancer Prevention Partners, Center for Environmental Health, Environmental Working Group and Lisa Lefferts; comment period reopened by FDA, May 2026
Open
01
Naming the decaf process
A laboratory test that tells, from a sample of roasted decaf, which process made it: methylene chloride, ethyl acetate, carbon dioxide, or water with carbon. Train it on samples whose process is known and publish the data and model.
The briefCredit · unclaimed02
Decaf capacity by method
A public table of the world's decaffeination plants, with owner, location, process and stated capacity, each with a source, summed by process and set against world trade in decaf, to show how much capacity outside methylene chloride exists.
The briefCredit · unclaimed
Open 01 · the brief
Naming the decaf process
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A laboratory test that tells, from a sample of roasted decaf, which process made it: methylene chloride, ethyl acetate, carbon dioxide, or water with carbon. Train it on samples whose process is known and publish the data and model.
Why it is still open
Zou and colleagues (2022) told decaf from regular coffee in eight pairs by their aroma compounds, measured by headspace gas chromatography with mass spectrometry, and found compounds derived from furan marked the decaf; they did not compare processes. Residue tests find only one of the solvents: in 2020 the Clean Label Project had 24 leading decaf brands tested by gas chromatography with mass spectrometry and reported methylene chloride at 50 to more than 90 parts per billion in some sold as "pure". Searches of Crossref and the web in November 2026 found no published method that tells the ethyl acetate, carbon dioxide and water processes apart. Buyers rely on the label and the decaffeinator's paperwork, and the question would matter more if FDA grants the solvent petition.
Built with what exists
Headspace gas chromatography with mass spectrometry (GC-MS) booked at a university or contract laboratory, or a benchtop near-infrared spectrometer for a cheaper version; decaf green coffees bought from importers that state the process for each lot; a roaster and a roast color meter so every sample is roasted to the same color; scikit-learn for a random-forest classifier. No solvents are handled.
First step
Buy 2 kg each of eight decaf green coffees, two per process, from importers' lists, roast them to one color and send 20 samples for headspace GC-MS.
Done looks like
On at least 80 blind samples, at least 20 per process, from more than one decaffeinator wherever more than one uses that process, the test names the process correctly at least 90% of the time, and the spectra and model are public.
Open 02 · the brief
Decaf capacity by method
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of the world's decaffeination plants, with owner, location, process and stated capacity, each with a source, summed by process and set against world trade in decaf, to show how much capacity outside methylene chloride exists.
Why it is still open
In May 2026 FDA asked for the practical considerations of phasing out the solvent uses, and the National Coffee Association says most US decaf uses the European method. The only worldwide figure found for one process is the Max Planck Society's 2014 estimate of about 100,000 tonnes a year decaffeinated with carbon dioxide. Searches in November 2026 found no public tally of decaffeination capacity by process.
Built with what exists
UN Comtrade trade data for decaf green coffee (HS code 0901.12) and roasted decaf (0901.22); company annual reports, environmental permits and trade press; an AI model to pull capacity figures and processes out of documents in several languages, such as Claude Haiku 5.5 or Meta's Muse Ember (see the Note "The jobs agents repeat"), with every figure checked by hand against its source.
First step
Download 2023 to 2025 exports under HS 0901.12 from UN Comtrade, list the ten largest exporting countries and name the decaffeinators in each.
Done looks like
At least 25 plants in at least 10 countries, each with a sourced process and either a sourced capacity or a stated estimate method, totals by process, and a comparison with world exports of decaf green coffee, published as a CSV file.