CoffeariumBeta
  1. Plant
  2. Farm
  3. Process
  4. Trade
  5. Roast
  6. Brew
  7. Cup
Roasting · Chemistry

Caffeine and roast degree

Whether light or dark roasts hold more caffeine – what roasting does to caffeine in the bean, why the answer depends on counting per bean, per gram or per scoop, and what reaches the cup.

Across the chain: Science

Source-checkedby Coffearium verification pass· Last source-checked 14 Oct 2026· Prepared with AI assistance· 9 sources (8 tier A)· 3 contested· Updated 14 Oct 2026
Reading depth
Contents
  • Caffeine is heat-stable. Roasting changes it far less than it changes chlorogenic acids.3,1
  • Per gram, roasted coffee holds as much caffeine as green coffee or slightly more, because the bean loses other matter.1,3
  • Dark beans are larger and lighter, so a scoop of dark roast holds less coffee by mass than a scoop of light roast.6
  • Brewing studies disagree on which roast gives the most caffeine in the cup; dose, brew method and extraction decide it.2,7,8,4
  • The species matters more than the roast; robusta beans hold about twice as much caffeine as arabica beans.3,1

The dark-roast caffeine claim#

Two opposite claims circulate: dark roast has more caffeine, and dark roast has less. A textbook chapter notes the common belief that roasting reduces caffeine.5 Both claims can be true or false, depending on what is counted.2 Caffeine can be counted per bean, per gram of coffee, per scoop or per cup. The sections below take these one at a time.

Claim Verdict Why
Roasting burns off caffeine Mostly false Caffeine is heat-stable; at most a small share may escape late in dark roasts.3,4,2
Dark roast has more caffeine per gram True, slightly The bean loses other matter while caffeine stays.1,3
Dark roast has less caffeine per scoop True for coffee measured by volume Dark beans are less dense, so a scoop holds less coffee; caffeine per scoop is inferred, not measured.6,1
Dark roast gives a stronger cup in caffeine Not established Brewing studies disagree; dose and extraction matter more.2,7,8
The roast matters more than the species False Beans of robusta hold about twice as much caffeine as arabica beans.3

Caffeine survives the roast#

Caffeine is a stable molecule.3 In a study of one Brazilian arabica, caffeine stayed at 12–13 mg per gram from green coffee to a dark roast. Caffeoylquinic acids, the main chlorogenic acids, fell from 41 to 5 mg/g over the same roasts.3

Caffeine can sublime: it turns from solid to vapour without melting. At atmospheric pressure this happens at 178 °C.2 In one study, every roast passed this temperature for part of the time.2 Caffeine crystals are often found as deposits around roasting machines.5

The losses are small. A 2021 review states that a small part of the caffeine may be lost by sublimation.4 A 2024 study found indirect signs of it only in the darkest roasts, and proposed that the pressure inside the bean raises the sublimation temperature.2

Per bean, per gram, per scoop#

Roasting removes water and other matter. The beans lose 15–22% of their weight and grow by about 50% to almost 100% in volume.9 Caffeine stays behind in the smaller mass.5,1 This changes the answer depending on what is compared.

1 g1 gLightDarkA · Per beanB · Per gramC · Per scoop123
  1. Per bean. Caffeine is heat-stable, so the amount in one bean changes little from green to roasted.3,5 No study opened for this entry weighed it per bean.
  2. Per gram. The bean loses 15–22% of its weight9, so each gram of roasted coffee holds the same caffeine or slightly more.1,3
  3. Per scoop. Roasted beans are larger and lighter. In one study, 100 beans went from 0.66 to 0.37 g/mL, so the same scoop holds less coffee from a darker roast.6
The same caffeine counted three ways: in one bean, in one gram and in one scoop, before and after roasting. Schematic, not to scale.

Per bean. Because caffeine is heat-stable, each bean is expected to keep nearly all of it.3,5 A textbook chapter says the total in the bean stays roughly unchanged as the bean gets lighter.5

Per gram. Each gram of roasted coffee holds as much caffeine as green coffee, or a little more.3,1 In a 2012 Brazilian study, caffeine per 100 g of dry matter rose from green coffee to medium roast in both species. The dark roasts did not differ significantly from the medium roasts.1 The authors explain the rise by the weight that the beans lose.1

025050075010001250150017502000mg caffeine per 100 g dry matterFilled dot: arabica (cv. Catuaí Amarelo) · open dot: robusta (cv. Apoatã)Green1,110 · 1,4861 Light (Agtron/SCAA 75–95)1,1921 1,6981 Medium (55–65)1,2551 1,8841 Dark (25–45)1,2221 1,8481
Data table
Pointmg caffeine per 100 g dry matterSource
Green1,11011101
Green1,48614861
Light (Agtron/SCAA 75–95)1,19211921
Light (Agtron/SCAA 75–95)1,69816981
Medium (55–65)1,25512551
Medium (55–65)1,88418841
Dark (25–45)1,22212221
Dark (25–45)1,84818481
Caffeine per 100 g of dry matter in one arabica and one robusta cultivar, green and roasted to three degrees. At each roast degree, the species differ more than the arabica roasts do.1

Per scoop. A scoop measures volume, not mass. Roasted beans swell and lose weight, so they become less dense.6 In one study, 100 beans grew from 19.13 to 28.57 mL and fell from 12.62 to 10.45 g during a dark roast.6 A scoop of dark roast therefore holds less coffee by mass than a scoop of light roast.6 In a 2024 study of ground coffee, the washed coffee, roasted systematically lighter, had consistently higher density.2 A scoop of dark roast probably also holds less caffeine. In one study, caffeine per gram rose by at most 27% (dry matter).1 In another, whole-bean density fell by 44% (our calculation).6 No study opened here measured caffeine per scoop.

What reaches the cup#

A 2021 review found that the coffee-to-water ratio probably has the greatest influence on caffeine in the cup.4 Roast degree acts on both the dose and the extraction. It sets how much coffee a scoop holds6, and it changes how easily the grounds give up their solids.2

Studies that weighed the coffee still disagree:

  • AeroPress, 2024. Brewed with the same mass of coffee and water, dark roasts gave less caffeine than light and medium roasts. At the same extraction yield they gave more.2
  • Espresso and drip, 2021. Caffeine rose with roast degree in both brews.7
  • Cold brew, 2017. The medium roast gave more caffeine than the dark roast.8
  • Review, 2021. The evidence on caffeine losses in roasting is called inconclusive.4

Differences in dose can hide roast effects. In a lab test of double espressos, the dark roast gave the most caffeine (174 mg). It was also brewed with the largest dose: 20.4 g, against 18.1–18.6 g for the other roasts.3 How much caffeine ends up in different brews is covered in Caffeine, together with official intake advice.

Species matters more than roast#

The species makes a larger difference than the roast.1,3 Beans of robusta (Coffea canephora) hold about twice as much caffeine as beans of arabica (Coffea arabica).3 A 2021 review gives 0.9–1.5% of dry weight for green arabica and 1.2–2.4% for green robusta. In brews, robusta mostly gave 1.4–1.8 times as much as arabica.4

In the 2012 study, robusta held 376–629 mg more caffeine per 100 g of dry matter than arabica at each roast degree. Roasting changed arabica by at most 145 mg and robusta by at most 398 mg.1

Decaffeinated coffee is a separate case. Its caffeine is extracted from the wetted green beans before roasting.5 See Decaffeination.

Reference values#

Values as reported in the sources. The basis matters: dry matter (water removed by calculation), as-is (fresh weight, including moisture), per serving or per volume of brew. Roast degree names are not standardised between studies.

Parameter Value reported Source
Arabica, green to dark roast (one Brazilian lot) 12 → 13 mg/g (as-is) Ludwig et al.3
Arabica cv. Catuaí Amarelo, mg/100 g dry matter Green 1,110 · light 1,192 · medium 1,255 · dark 1,222 Tfouni et al.1
Robusta cv. Apoatã, mg/100 g dry matter Green 1,486 · light 1,698 · medium 1,884 · dark 1,848 Tfouni et al.1
Roast degrees in that study (Agtron/SCAA disc) Light 75–95 · medium 55–65 · dark 25–45; set temperature 200 °C for 7, 10 and 12 min Tfouni et al.1
Green coffee, % of dry weight Arabica 0.9–1.5 · robusta 1.2–2.4 Olechno et al.4
Robusta vs arabica About twice as much (beans); 1.4–1.8 times in most brew studies Ludwig et al.3; Olechno et al.4
Caffeine sublimation point (atmospheric pressure) 178 °C (source: 352 °F) Lindsey et al.2
Bean probe temperature above which caffeine in brews fell About 204–216 °C (source: 400–420 °F) Lindsey et al.2
Roast loss 15–22% Münchow et al.9
Volume gain in roasting About +50% to almost +100%; +48% and +50% for 100 beans (40 min at 170 °C) Münchow et al.9; Jokanović et al.6
100 beans, Guaxupe, green → dark (L* 24.45) 12.62 → 10.45 g · 19.13 → 28.57 mL · 0.66 → 0.37 g/mL; weight loss 17.15% Jokanović et al.6
100 beans, Outspan, green → medium (L* 26.77) 14.86 → 12.71 g · 23.03 → 34.10 mL · 0.53 → 0.37 g/mL (from 7 min); weight loss 14.43% Jokanović et al.6
Bulk density of 100 beans at 0, 7, 14, 21, 28, 35 and 40 min (g/mL) Guaxupe 0.66 · 0.59 · 0.56 · 0.52 · 0.47 · 0.45 · 0.37; Outspan – · 0.53 · 0.49 · 0.49 · 0.47 · 0.42 · 0.37 Jokanović et al.6
Double espresso, lab, by roast Light 165 mg (18.6 g) · medium 152 mg (18.1 g) · dark 174 mg (20.4 g); 65–68% of the caffeine extracted Ludwig et al.3
AeroPress, 10 g coffee in 150 g water 134–165 mg (natural) and 145–165 mg (washed) per 8 fl oz cup Lindsey et al.2
Cold brew, 35.0 g in 350 mL, 24 h (mg/L) Medium 1,180 ± 90 and 1,230 ± 60 · dark 1,080 ± 70 and 990 ± 30 (medium and coarse grind) Fuller & Rao8
Hot brew, same coffees (mg/L) Medium 1,040 ± 70 and 970 ± 70 · dark 1,060 ± 70 and 840 ± 10 (medium and coarse grind) Fuller & Rao8
05101520253035400.30.40.50.60.7Roasting time (min)Bulk density (g/mL)Guaxupe (dark)6 Outspan (medium)6
Data table
SeriesRoasting time (min)Bulk density (g/mL)Source
Guaxupe (dark)00.666
Guaxupe (dark)70.596
Guaxupe (dark)140.566
Guaxupe (dark)210.526
Guaxupe (dark)280.476
Guaxupe (dark)350.456
Guaxupe (dark)400.376
Outspan (medium)70.536
Outspan (medium)140.496
Outspan (medium)210.496
Outspan (medium)280.476
Outspan (medium)350.426
Outspan (medium)400.376
Bulk density of 100 beans of two green coffees during 40 min of roasting at 170 °C; the beans swell and lose mass, so the same volume holds less coffee as the roast goes on.6

Bases and what they hide#

A value per gram “as-is” includes the moisture of the sample. Green coffee holds more water than roasted coffee, so as-is values of green coffee are diluted by water.3 In one study, green coffee held 8.4% water and the roasts 3.2–4.9%.3

A value per dry matter removes water by calculation. It still rises with roasting, because roasting also removes dry matter.1 The authors of the 2012 study point out that their dry-basis results are corrected only for the loss of water.1 They expected sublimation to lower caffeine; their values rose instead, which they explain by the loss of weight in roasting.1

Neither basis gives caffeine per bean. That needs the mass of the bean, or the roast loss, measured on the same samples as the caffeine. If a bean kept all its caffeine, a roast loss of 15–22%9 would raise caffeine per gram (as-is) by about 18–28% (our calculation: 1/0.85 and 1/0.78). The as-is rise reported in one study, 12 to 13 mg/g, is smaller; the values were given in whole milligrams.3 That study did not report roast loss3, so the gap cannot be split between caffeine losses, moisture and rounding.

Sublimation in detail#

Pure caffeine sublimes at 178 °C at atmospheric pressure.2 Inside a roasting bean, pressure rises as gas forms, and the bean’s volume and porosity increase.2 The authors of a 2024 study therefore expect the actual sublimation temperature during roasting to be higher than the literature value.2

That study roasted two Ethiopian arabicas to five levels, from first crack to 7 min after it. Caffeine in 10-min brews first rose with roast degree, then peaked and fell. The fall began at bean probe temperatures above about 204–216 °C (source: 400–420 °F).2 The same pattern in both coffees was read as indirect support for caffeine leaving the beans late in the roast.2 Porosity rose with roast degree and probably affected both sublimation and extraction.2

Bean probe readings are specific to the roaster and are not true bean temperatures (see Roasting). The caffeine itself was measured in brews, not in the beans.2

Extraction and roast degree#

Roasting changes how easily solubles leave the grounds. In the 2024 AeroPress study, extraction yield peaked at roast losses of about 12–14% and fell for darker roasts.2 The authors propose two competing effects: roasting makes the bean more porous, which eases extraction, and it volatilises or breaks down compounds.2 At equal extraction yield, dark roasts gave more caffeine.2

Brew method and grind change the picture again. In the 2017 cold-brew study, the medium roast gave more caffeine than the dark roast, but at medium grind the authors judged the difference insignificant because of large variation.8 In the 2021 study of four roast degrees (Agtron 70.17 to 20.50), espresso (7 g to 30 mL) and drip coffee (15 g to 210 mL) both gave more caffeine at darker roasts.7

The studies differ in coffee, roaster, roast degrees, grind and brew water. Their results cannot be pooled into one rule.2,4

Measuring caffeine across roasts#

Studies measure caffeine by high-performance liquid chromatography (HPLC) with UV or diode-array detection. The 2012 study read it at 272 nm1, the 2014 study at 280 nm.3 The 2024 study used a refractometer for extraction yield and HPLC for caffeine.2

Comparing roast degrees needs three values per sample: the basis of each caffeine value, the moisture and the roast loss. The 2014 study reported moisture but not roast loss.3 For bulk density, the method matters: one study weighed 100 beans and read their volume in a 50 mL graduated vessel6; another measured the freely settled density of ground and sieved coffee.2

Open questions#

  • Caffeine per bean. No study opened for this entry measured caffeine and roast loss on the same samples, so the share lost by sublimation is not established.1,2,3
  • Sublimation inside the bean. The temperature at which caffeine leaves a pressurised, porous bean has not been measured directly.2
  • Per-scoop caffeine. No study opened here measured caffeine per scoop; the per-scoop effect follows from density and caffeine per gram measured separately.6,1
  • Brew results. Brewing studies point in different directions, and the review evidence is inconclusive.2,4

Key studies#

Tfouni et al., 2012 – Roasting one arabica and one robusta. Roasted a Catuaí Amarelo arabica and an Apoatã robusta to light, medium and dark and measured caffeine and chlorogenic acids by HPLC on a dry-matter basis. Caffeine rose from green to medium roast in both and did not differ significantly between medium and dark; the species differed more than the roasts.

International Journal of Food Science and Technology 47(2), 406–415 · doi:10.1111/j.1365-2621.2011.02854.x

Lindsey et al., 2024 – Roast degree, extraction yield and caffeine. Roasted two Ethiopian arabicas to five levels after first crack and brewed them in an AeroPress at 15:1. Dark roasts gave less caffeine under identical brewing but more at identical extraction yield, and caffeine in brews fell for the darkest roasts, which the authors read as indirect evidence of sublimation.

Scientific Reports 14, 29126 · doi:10.1038/s41598-024-80385-3

Ludwig et al., 2014 – What are we drinking? Roasted one Brazilian arabica to three roast degrees and made espressos in the lab. Caffeine stayed at 12–13 mg/g while chlorogenic acids fell sharply, so the caffeine-to-chlorogenic-acid ratio tracked the roast degree.

Food & Function 5, 1718–1726 · doi:10.1039/c4fo00290c

Jokanović et al., 2012 – Physical changes during roasting. Followed the mass, volume, bulk density and texture of 100-bean samples of two green coffees every 7 min during 40 min at 170 °C. The beans grew by about half in volume while bulk density fell to 0.37 g/mL.

Acta Periodica Technologica 43, 21–31 · doi:10.2298/APT1243021J

Fuller & Rao, 2017 – Cold brew by roast and grind. Brewed a medium and a dark roast of one arabica cold and hot at two grind sizes and followed caffeine over time. The medium roast gave more caffeine in most comparisons, though not significantly in every case.

Scientific Reports 7, 17979 · doi:10.1038/s41598-017-18247-4

Contested
  • Which roast gives the most caffeine in the cup: in a 2024 AeroPress study, dark roasts gave less caffeine than light and medium roasts under identical brewing, but more at identical extraction yields2; in a 2021 study of espresso and drip coffee, caffeine rose with roast degree7; in a 2017 cold-brew study, the medium roast gave more than the dark roast, although some differences were within the measurement variation.8 A 2021 review calls the evidence on roasting losses inconclusive.4
  • Whether caffeine leaves the bean during roasting: a textbook chapter says the content per bean stays roughly unchanged5, and a 2012 study found caffeine per dry matter rising with roast degree1; a 2021 review states that some caffeine may be lost by sublimation4, and a 2024 brewing study read falling caffeine in brews from the darkest roasts as indirect evidence of sublimation.2 No study opened for this entry weighed caffeine per bean before and after roasting.
  • Caffeine in the bean by roast degree: according to a 2024 literature summary, the result most commonly reported is that roast degree does not affect caffeine content, while other studies report the highest content in light, medium or dark roasts.2 The studies differ in basis (dry matter or as-is), coffee and roasting conditions.1,3
Share or cite this entry

Sources

  1. ATfouni, S. A. V.; Serrate, C. S.; Carreiro, L. B.; Camargo, M. C. R.; Teles, C. R. A.; Cipolli, K. M. V. A. B.; Furlani, R. P. Z. (2012). Effect of roasting on chlorogenic acids, caffeine and polycyclic aromatic hydrocarbons levels in two Coffea cultivars: Coffea arabica cv. Catuaí Amarelo IAC-62 and Coffea canephora cv. Apoatã IAC-2258. International Journal of Food Science and Technology 47(2), 406–415. doi:10.1111/j.1365-2621.2011.02854.x · other copies
  2. ALindsey, Z. R.; Williams, J. R.; Burgess, J. S.; Moore, N. T.; Splichal, P. M. (2024). Caffeine content in filter coffee brews as a function of degree of roast and extraction yield. Scientific Reports 14, 29126. doi:10.1038/s41598-024-80385-3 · other copies
  3. ALudwig, I. A.; Mena, P.; Calani, L.; Cid, C.; Del Rio, D.; Lean, M. E. J.; Crozier, A. (2014). Variations in caffeine and chlorogenic acid contents of coffees: what are we drinking?. Food & Function 5, 1718–1726. doi:10.1039/c4fo00290c · other copies
  4. AOlechno, E.; Puścion-Jakubik, A.; Zujko, M. E.; Socha, K. (2021). Influence of Various Factors on Caffeine Content in Coffee Brews. Foods 10(6), 1208. doi:10.3390/foods10061208 · other copies
  5. BPietsch, A. (2017). Decaffeination – Process and Quality. In: Folmer, B. (ed.), The Craft and Science of Coffee, Academic Press, ch. 10, pp. 225–243. doi:10.1016/B978-0-12-803520-7.00010-4 · other copies
  6. AJokanović, M. R.; Džinić, N. R.; Cvetković, B. R.; Grujić, S.; Odžaković, B. (2012). Changes of physical properties of coffee beans during roasting. Acta Periodica Technologica 43, 21–31. doi:10.2298/APT1243021J · other copies
  7. AKim, I.; Jung, S.; Kim, E.; Lee, J.-W.; Kim, C.-Y.; Ha, J.-H.; Jeong, Y. (2021). Physicochemical characteristics of Ethiopian Coffea arabica cv. Heirloom coffee extracts with various roasting conditions. Food Science and Biotechnology 30(2), 235–244. doi:10.1007/s10068-020-00865-w · other copies
  8. AFuller, M.; Rao, N. Z. (2017). The Effect of Time, Roasting Temperature, and Grind Size on Caffeine and Chlorogenic Acid Concentrations in Cold Brew Coffee. Scientific Reports 7, 17979. doi:10.1038/s41598-017-18247-4 · other copies
  9. AMünchow, M.; Alstrup, J.; Steen, I.; Giacalone, D. (2020). Roasting Conditions and Coffee Flavor: A Multi-Study Empirical Investigation. Beverages 6(2), 29. doi:10.3390/beverages6020029 · other copies

Tier A · authoritative · B · trade and expert · C · general. Supporting quotes for every source are in the data version of this entry. If a publisher’s site refuses access, try the doi link or “other copies”.

Suggest a change

On Caffeine and roast degree. Tell us what you would change and why. A source that supports it helps most.