Science
Coffee chemistry: from seed to cup
What a green coffee seed contains, what roasting breaks down and builds, what dissolves into the cup, and why knowing the compounds does not yet explain the taste.
In short
A coffee seed arrives at the roaster as a dense store of carbohydrate, with smaller amounts of sugars, acids, oil and nitrogen compounds. Roasting breaks down some of these, leaves others almost untouched and builds new ones that were never in the seed. Brewing then dissolves about a fifth of what is left.
This page follows those three steps. It reports what the studies read for it measured, and it is careful about one thing in particular: knowing that a compound is present, or that it falls during roasting, is not the same as knowing what it does to the taste of a cup.
What is in the green seed
Sourced
A 2025 study that profiled green beans of three species describes more than 700 compounds in unroasted coffee as having been considered to influence flavour, and names the groups that matter most: the alkaloids caffeine and trigonelline, sugars, the chlorogenic acids and other organic acids. The first compounds it identifies in its own chromatograms are trigonelline, sucrose, quinic, malic and citric acids, three caffeoylquinic acids and caffeine.
Sucrose is the main sugar, and that study treats it as both a contributor to taste and a starting material for what roasting makes. The chlorogenic acids are a family, not one substance: caffeic or ferulic acid joined to quinic acid in several arrangements, of which the 5-caffeoylquinic form is the most abundant.
The species differ before anything is done to them. The same study found robusta significantly higher in caffeine than arabica, lower in sucrose and higher in one of the chlorogenic acids, and carrying more of a group of amino-acid derivatives. Among arabicas the spread is real too: across nine districts of south-west Ethiopia, green coffee averaged from 1.05 to 1.52 grams of caffeine and from about 3.7 to 4.5 grams of chlorogenic acids in every 100 grams of dry matter. The paper's text and its table differ slightly on the highest chlorogenic acid figure, 4.45 against 4.51.
Those chlorogenic acid figures are lower than the ones on CoffeeHQ's page about green coffee chemistry, and the difference is one of method, not a contradiction. That page quotes a 2006 review of chromatographic analyses, which puts the total of all the chlorogenic acids at 4 to 8.4 per cent of dry matter in arabica and 7 to 14.4 per cent in robusta. The Ethiopian study quantified its coffees against a single chlorogenic acid standard, and its individual samples ran from 2.8 to 5.4 grams per 100 grams, which its authors describe as generally within the ranges reported before. A figure for chlorogenic acids means little without knowing which members of the family were counted and how.
Most of the seed is none of these. The bulk is structural carbohydrate in the cell walls, with protein and oil making up much of the rest. The proportions, and the reviews each comes from, are on the page about green coffee chemistry; a review of coffee lipids, for one, puts the oil at about 15 per cent of green arabica and about 10 per cent of robusta.
What roasting does to each group
Sourced
A summary of direction, drawn from the studies named in each row. It is a reading aid and not a set of measurements: the studies used different coffees and different ways of heating them.
| Variable | Typical range | What moving it does |
|---|---|---|
| Caffeine | Survives; little is lost | A study of Ethiopian arabica found caffeine unchanged by roasting where trigonelline and chlorogenic acids fell. A second measured slightly more caffeine per gram after roasting, which its authors attribute to the bean losing other mass. A third, brewing roasted coffee under fixed conditions, found less caffeine in the cup from the darkest roasts and read that as some loss late in a roast. |
| Chlorogenic acids | Fall steeply, more with heat and time | In the Ethiopian study chlorogenic acids fell by a little over half on roasting, from 4.22 to 1.94 grams per 100 grams. In beans heated in a laboratory oven, chlorogenic acid fell from about 34 to under 3 milligrams per gram at the harshest setting, following simple first-order decay. A thesis on roasting cites losses of up to four fifths in dark roasts, and the 2006 review of coffee phenolics says drastic roasting may destroy up to 95 per cent. |
| Trigonelline | Breaks down, by an amount that depends on the roast | A review, citing earlier work, reports that it decomposes readily at roasting temperatures above 180 °C, giving nicotinic acid and N-methylpyridinium, which are not volatile, and volatile pyrroles and pyridines. How much goes depends on the roast: in the Ethiopian study's roast it fell by under a tenth, from 0.91 to 0.84 grams per 100 grams. |
| Sucrose | Consumed, by the sources' account | The species study, citing earlier work, describes it as a precursor of smaller acids such as formic, acetic and lactic, and as a participant in caramelisation. No measurement of its loss was read for this page. |
| Amino acids and sugars together | React to make colour and aroma | The Maillard reaction and Strecker degradation are the names given to these reactions. Browning measured by light absorbance rose steadily with roasting time and temperature in the oven study. |
| Carbon dioxide | Formed, and partly trapped | Released for weeks afterwards. The guide to staling covers what was measured. |
The oven study heated beans on a tray for up to 40 minutes, which is not how coffee is roasted. Its directions agree with the other sources; its rates should not be read as those of a roasting machine.
What roasting makes that was not there
Sourced
Roasted coffee owes its brown colour to melanoidins: large molecules built during roasting on a backbone of the seed's own polysaccharides, with protein and chlorogenic acid fragments bound in. A review describes them as part of the non-volatile fraction of the drink, quotes a concentration of 2 to 4 milligrams per millilitre in a cup of coffee from an earlier source, and notes that they have been related, together with caffeine, trigonelline and chlorogenic acids, to bitterness and astringency.
The aroma is made of far smaller molecules. A study that roasted one coffee to four degrees and had assessors smell each compound as it left the instrument identified 25 aroma-active compounds, 22 of them present at every roast degree. Pyrazines, a furanone and 2,3-pentanedione stood out at the medium roast, described by the assessors as sweet, toasted hazelnut and caramel; the very dark roast was marked by compounds described as burnt and sulphurous.
The same study makes a point that is easy to miss. Several of the largest peaks in its measurements, acetic acid and pyridine among them, had no detectable smell at all in the assessment, while compounds present in small amounts did. How much of a compound there is does not tell you how much it contributes.
One intermediate shows why roast chemistry resists simple rules. Hydroxymethylfurfural, formed from sugars, was still rising after 40 minutes in the coolest oven; in the two hotter ones it peaked at 20 minutes and then fell as it was itself broken down, almost to nothing at the hottest. A compound can be a product of roasting at one stage and a casualty of it at the next.
What reaches the cup
Sourced
Water takes only part of the roasted coffee. In studies of immersion brewing the share dissolved settles at about a fifth of the dry weight. What dissolves is a mixture in which the well-known compounds are a minority.
One espresso study measured three of them alongside the total. From a 20-gram dose brewed to 40 grams of drink, the cup held on average about 183 milligrams of caffeine, 118 of the main chlorogenic acid and 98 of trigonelline, against 3.9 grams of dissolved solids in all. The three together are roughly a tenth of what was dissolved. The other nine tenths are melanoidins, other acids, carbohydrate fragments, minerals and much else that nobody measured in that study.
The same study found that the compounds do not respond alike to how a shot is made. Flow rate had the strongest influence on how much of each ended up in the cup, and caffeine was the least affected of the three.
Coffee oil carries its own compounds, among them the diterpenes cafestol and kahweol. Whether they reach the cup depends mostly on the filter. The review of coffee lipids reports that less than 0.2 per cent of the coffee's lipids pass into filtered coffee made in a household machine, against 1 to 2 per cent into an espresso, and the diterpenes travel with the oil; a laboratory that measured the diterpene esters in brews made several ways found the same ordering, with boiled coffee and French press above both. The guide to extraction covers it. And the order in which compounds leave the grounds is less tidy than the usual account: in the one study of drip coffee collected in fractions, the earliest liquid was both more sour and more bitter than what followed.
What chemistry cannot yet say about taste
Accounts differ
It is tempting to finish a page like this with a table matching compounds to flavours. The sources read here do not support one. The species study calls caffeine linked to bitterness and says chlorogenic acids are thought to contribute to bitterness, acidity and astringency; it also says that the sucrose content of green beans is, by itself, considered a poor predictor of quality.
Where correlation has been tested directly the results are patchy. In the Ethiopian study, the trigonelline and chlorogenic acid contents of roasted coffee correlated with some cup scores, body and flavour among them, and not with others, and caffeine correlated with the scores for bitterness and astringency and with nothing else. A correlation in one set of coffees from one region is a lead, not a mechanism.
So the honest summary has three layers. Which compounds are in coffee, and roughly how roasting moves them, is well measured. Which of them have been associated with a taste or a smell is partly known. How a particular coffee's chemistry produces the particular cup in front of you is not something any study read for this page claims to predict.
What this page does not state
CoffeeHQ explanation
It gives no composition table for the whole seed: that is on the page about green coffee chemistry, with the review behind each row. It says little about lipids, proteins and the polysaccharides themselves, which that page and the guide to what happens when coffee roasts cover, and nothing about minerals, because nothing was read on them.
It assigns no flavour to any compound, gives no roast temperature or time as correct, and makes no claim about health. The measurements quoted come from a handful of coffees; a different coffee would give different numbers.
What to do next
- Green coffee chemistry — What the raw seed contains before roasting: cell-wall polysaccharides, oil and diterpenes, protein and chlorogenic acids, with the review each figure comes from.
- What happens when coffee roasts — The physical side of the same process: weight, volume, colour and the two cracks.
- Coffee extraction — How much dissolves, why strength and extraction are different, and what a filter holds back.
- Caffeine in coffee — What was measured in real servings and per gram of coffee used.
- Coffee flavour and tasting — The other half of the question: how taste and smell are described, since chemistry does not settle it.