Coffee roasting chemistry: what breaks down and what is made

A model of the order things happen in

CoffeeHQ explanation

A model for holding the rest of the page in mind. The stages overlap, each runs at a different rate in the outside of a bean and in its middle, and the sources give no temperature at which one hands over to the next.

Maillard, Strecker, caramelisation, pyrolysis: what each word claims

Widely documented

The Maillard reaction is not one reaction. It is the name for the network that starts when a sugar meets an amino group, on an amino acid or a protein, and that runs on through many intermediates to brown polymers at one end and small smelly molecules at the other. A study that followed the aroma of roasting coffee with a high-resolution instrument describes the Strecker degradation as part of the later stages of that network: the reaction of compounds made earlier in it with amino acids, giving aldehydes, pyrazines and other products. Strecker is a branch of Maillard, not a rival to it.

Caramelisation, strictly, is what happens to a sugar heated alone. Inside a coffee seed a sugar is never alone: it is surrounded by amino acids, proteins and chlorogenic acids. The sources read here list caramelisation among the reactions of roasting without saying how much of the seed's sugar takes that route and how much the Maillard one, and the thesis that reviewed the field in 2000 says plainly that the reactions had not been completely worked out, because what happens inside a bean is hard to reproduce outside one. So when a roaster says a coffee has caramelised, the word is describing a stage or a taste, not a measured reaction; and a tasting note of caramel is not evidence of caramelisation.

Pyrolysis means breaking molecules apart by heat alone. It is the word used for the far end of a roast, where material is being destroyed faster than anything new is being made. No source read here gives a point at which it begins.

The practical use of getting the words right is to distrust a certain kind of sentence: one that says the Maillard reactions make the nutty notes, the Strecker reactions the fruity ones and caramelisation the sweetness. No source opened for this page assigns flavours to reactions in that way.

Water first

Sourced

A doctoral thesis on roasting, summarising the literature before it, puts the weight lost in a roast at generally 14 to 20 per cent, most of it water. A further part is dry matter turned to gas, some five to eight per cent for a medium roast. The reactions themselves make water as well, which leaves as steam; so the water that leaves a roast is more than the water that was in the green coffee.

Water also sets the pace of the chemistry. In that thesis's laboratory roasts, aroma formed fastest once the bean had dried below about five grams of water per hundred. Before that, heat goes into evaporation and the wet interior stays cooler than the surface.

Sucrose and the cell walls

Sourced

Sucrose is the seed's main free sugar, several per cent of its dry weight. A review of coffee carbohydrates describes it as well documented that the low-molecular-weight sugars are almost totally degraded in roasting, with minimal amounts left in the bean. Sucrose is used up, not preserved; whatever sweetness a roasted coffee has is not sucrose that survived.

The bulk of the seed is cell-wall polysaccharide, and roasting changes that too. The same review reports that up to 40 per cent of the polysaccharides were degraded after a long roast. The arabinogalactans are the most easily damaged: after even a light roast their molecules were at least ten times smaller. The galactomannans shorten as well. With the long chains cut and less entangled with one another, the review says, they dissolve more easily in water.

That is part of why roasted coffee gives up a fifth of its weight to hot water when green coffee would not. It is also a caution about a common shorthand. Roasting does make more of the bean's structure soluble; but brewing studies do not show extraction simply rising with roast degree, and in one of them the darkest roasts extracted least.

Chlorogenic acids: rearranged, then lost

Sourced

The chlorogenic acids are the seed's most abundant small molecules after sucrose, and the most studied in roasting. A review of coffee phenolics describes three things happening to them. At the start of a roast they isomerise: the most common form falls sharply while two rarer forms rise, in some cases to almost double. Part is turned into lactones, by losing a molecule of water and closing into a ring, which the review says happens after six to seven per cent of weight loss and accounts for about 7 per cent of the chlorogenic acids in arabica and 5.5 per cent in robusta. And part is split and broken down into smaller compounds.

The overall loss is steep. The review gives 8 to 10 per cent of the chlorogenic acids lost for every 1 per cent of dry matter lost, and up to 95 per cent under drastic conditions. A study of Ethiopian arabica measured a fall of a little over half in its own roast, from 4.22 to 1.94 grams per 100 grams. The acids also end up inside the brown polymers that roasting builds.

The review attributes a rise in bitterness during roasting to the release of caffeic acid and the formation of lactones and other phenol derivatives. That is the reviewers' statement, made on the authority of earlier work, and the same review calls the relationship between chlorogenic acids and cup quality still unclear. It is the nearest this page comes to a compound and a taste, and it is reported as a review's attribution.

Trigonelline, caffeine and the oil

Sourced

Trigonelline breaks down in roasting, by an amount that depends on the roast. A review, citing earlier work, reports that it decomposes readily above 180 degrees Celsius, giving nicotinic acid and a second non-volatile product, and volatile pyrroles and pyridines. In the Ethiopian study's roast it fell by under a tenth.

Caffeine is the stable one, with a qualification. The older literature calls it thermally quite stable. A study that roasted two coffees to many degrees and brewed them identically found less caffeine in the cup from the darkest roasts, and its authors read that as indirect evidence that some is lost late in a roast.

The oil changes least. A review of coffee lipids reports only small changes in the fatty acid composition during roasting and says the sterols are hardly affected. The compounds dissolved in the oil are less stable: the esters of the diterpenes cafestol and kahweol fall as roasting temperature rises, and breakdown products of both, which increase with temperature, have been identified in roasted coffee. One diterpene found only in robusta passes through roasting unchanged, which is why it can be used to detect robusta in a roasted blend. Tocopherols fall to between about four fifths and all of their starting level.

A study of aroma during roasting lists the oxidation of lipids among the sources of volatile compounds, so the oil is not inert. But the picture of oil as the part of the bean that roasting mostly leaves alone is the reviews' own.

What is made: gas, brown polymers and aroma

Sourced

Carbon dioxide is the main gas. A doctoral thesis measured how much stays in the bean: about 6 to 7 milligrams per gram in a light roast, about 11 in a medium and about 15 to 16 in a dark, with no further increase beyond that. In the thesis's experiments the chlorogenic acids were one source of it but not the main one; sucrose heated alone gave almost none, while sucrose heated with an amino acid produced a great deal, and about half came from the seed's smaller molecules. The gas, on that account, is largely a by-product of browning.

Melanoidins are the brown polymers. A review describes them as built on a backbone of the seed's own polysaccharides, with protein and chlorogenic acid fragments bound in. They are a large part of what dissolves in a cup of coffee and most of its colour.

The aroma is made of much smaller molecules, hundreds of them. A study that roasted one coffee to four degrees and had assessors smell each compound as it left the instrument found 25 that had a detectable smell, 22 of them present at every degree of roast. Some of the largest peaks in its measurements had no detectable smell at all. How much of a compound there is does not say how much it contributes.

Nor is the chemistry uniform between beans. When single beans from 25 lots were roasted one at a time under identical conditions, the spread between beans of the same lot was, for some volatile compounds, larger than the average amount itself. A roast is a statistical event: a batch is thousands of small reactors that did not all do the same thing.

The dark end: when breakdown overtakes formation

Sourced

Several measurements point the same way at the far end of a roast. In the thesis's roasts, some aroma compounds were already declining in the final stage while others still rose. In a study on a drum roaster, the acidity of the coffee peaked at first crack in every profile and was back to the level of the green coffee by second crack. The gas held in the bean stops rising between dark and very dark, which the thesis that measured it puts down to the starting materials running out and to gas escaping at second crack. And in the aroma study, the very dark roast was marked by compounds its assessors described as burnt and sulphurous.

One intermediate shows the pattern in miniature. Hydroxymethylfurfural, formed from sugars, was still rising after 40 minutes in the coolest of three laboratory ovens; in the two hotter ones it peaked and then fell, almost to nothing at the hottest. A compound can be a product of roasting at one stage and a casualty at the next.

This is the chemistry behind the observation that dark roasts taste more alike than light ones. It is a reasonable reading of these measurements and it is CoffeeHQ's reading: none of the studies set out to test it.

What goes, what stays and what arrives

Sourced

A summary of direction. Each row names whose statement it is; the studies used different coffees and different ways of heating them.

The reviews gather analyses made over several decades by different methods, and three of them date from 2006. A direction is safer to take from this table than a figure.

Three things often said, and where they stand

Accounts differ

The sugars caramelise, so a longer roast is sweeter

Unsupported by anything read. The seed's sugars are almost entirely destroyed in any roast, and the one set of tasting studies read points, weakly, the other way: pooled over seven studies, sweetness fell very slightly with longer times after first crack, by so little that the authors call the effect more likely to be spurious, and in a comparison of four roasts taken to one colour the fastest finish was rated sweetest.

Lighter roasts keep more acid

Supported for what was measured, with a qualification. Total chlorogenic acids fall steadily with roasting. But the acidity a laboratory measures by titration rose to a peak at first crack before falling, so the lightest possible roast is not the most acidic one.

Each reaction makes its own family of flavours

Unsupported by anything read. The sources describe which compounds form and which of them can be smelled; none assigns nutty, fruity or sweet to Maillard, Strecker or caramelisation.

What this account of roasting chemistry leaves out

CoffeeHQ explanation

It gives no temperature or time for any reaction and none for first or second crack. It assigns no flavour to a compound or a reaction. It says nothing about acrylamide, furan or any other compound of interest to health; a review of those exists and was not read for this page. It does not cover the amino acids and proteins in any detail, or the minerals at all, because nothing was read on them.

The three reviews that carry most of it date from 2006 and gather older analyses; methods have changed and a newer survey might move several figures. The thesis on carbon dioxide used one coffee in a laboratory roaster. Nothing here was measured by CoffeeHQ.

What to do next

• What happens when coffee roasts — The physical side: weight, volume, colour, the two cracks and how heat moves through a bean.
• Green coffee chemistry — The starting materials, with the review each figure comes from.
• Coffee chemistry: seed, roast and cup — The shorter account that follows the same compounds on into the brew.
• Why coffee goes stale — What becomes of the gas and the aroma once the roast is over.
• Reading a roast profile — How the path of a roast is recorded, and what has been tested about its shape.