3-butanedione

fruity ketone Key food odorant

Found in 5 ingredients across the CompKitchen flavor graph.

Look up another compound — search by name, CAS number, PubChem CID, or FEMA #.

Key food odorant (generalist)

Dunkel et al. 2014

Identified in the meta-analysis of 227 food samples as one of 16 generalist odorants — appears in >25% of foods, typically from generic Maillard, lipid-oxidation, or fermentation pathways. Characteristic odor: buttery.

Source: Angew. Chem. Int. Ed. 2014, 53, 7124–7143

Properties

Odor description: buttery, creamy, caramel, sweet
Odor threshold — orthonasal, in clean water: 2.8 ppb – 18.0 ppb median 5.5 ppb · 4 published values · see all 27 with sources
CAS number: 431-03-8

Regulatory · United States

FEMA GRAS — Generally Recognized As Safe for flavoring use

? Partial match

FEMA number: 2370
GRAS publication: GRAS 3
FEMA primary name: DIACETYL-2,3-Butanedione; Biacetyl; Dimethyl

Use-level matrix not extracted for this substance — check the source publication for current ppm guidance.

Source: FEMA GRAS publications 1-33.

How this molecule is described

Sources disagree about odour more often than they disagree about chemistry. Rather than show one answer, here is every description on file with how many ingredient records carry it.

Descriptions on file

Ordered by how many independent descriptor streams give each wording, then by how often it appears. The bar is still replication — one source copied onto many ingredients scores high there — so the two can disagree.

pine FlavorDB 2.0, Flavordb Odor Column 2 sources agree

green, orange FlavorDB 2.0, Flavordb Odor Column 2 sources agree

sweet, caramel, butter, creamy, burnt FlavorDB 2.0, Flavordb Odor Column 2 sources agree

fruity, floral, earthy, butter, creamy, apple, garlic, honey, caramel, rose, pine FlavorDB 2.0

almond, apple, garlic FlavorDB 2.0

buttery, creamy, caramel, sweet shown above Llm Adjudicated

—

Bars are the share of ingredient records carrying each description. High agreement means the sources replicate each other — not necessarily that they are right.

Published thresholds

27 measured values from the literature. Grouped by what was actually measured — a threshold smelled in clean water, tasted in the mouth, or measured in a real food are three different quantities and are not pooled.

Orthonasal, in clean water

2.8 ppb – 18.0 ppb

The reference measurement — smelled in water. OAV is computed against this.

Threshold	Matrix	Source	Paper
2.8 ppb	water/ethanol 6:4 (food: american bourbon whisky)	Literature corpus	30972045
4.4 ppb	water_odor	Leffingwell	—
6.5 ppb SIDA	water	Rychlik 1998 · 3 studies	—
18.0 ppb SIDA	water	Dunkel 2014	—

In a food matrix

0.390 ppb – 44.0 ppb

Measured in a real food, where the matrix shifts detection — not a clean-water value.

Threshold	Matrix	Source	Paper
0.390 ppb	water (food: lime)	Literature corpus	30972045
0.640 ppb	water (food: unknown)	Literature corpus	30972045
0.940 ppb	water (food: longjing green tea infusion)	Literature corpus	30972045
0.960 ppb	water (food: green tea infusion (jingshan cha))	Literature corpus	30972045
1.0 ppb	water (food: reference standard (water))	Literature corpus	30972045
4.0 ppb	water (food: smoked cooked loin (pork))	Literature corpus	30972045
6.5 ppb	water (food: wheat porridge / bread crumb)	Literature corpus	30972045
6.5 ppb	water (food: extruded wheat flour)	Literature corpus	30972045
15.0 ppb	water (food: coffee)	Literature corpus	30972045
44.0 ppb	water (food: fish)	Literature corpus	30972045

Retronasal / in-mouth

5.0 ppb – 32.5 ppb

Tasted rather than smelled — a different sense, not comparable.

Threshold	Matrix	Source	Paper
5.0 ppb	water_flavor	Leffingwell	—
5.0 ppb SIDA	water_retronasal	Rychlik 1998 · 3 studies	—
32.5 ppb SIDA	oil_retronasal	Rychlik 1998 · 2 studies	—

In air

0.050 ppb

Airborne detection — not comparable to a water threshold.

Threshold	Matrix	Source	Paper
0.050 ppb	air	Nagata 2003	—

Other matrix

2.8 ppb – 100 ppb

Measured in a non-aqueous medium (oil, starch, cellulose).

Threshold	Matrix	Source	Paper
2.8 ppb	ethanol/water (food: rum / red wine)	Literature corpus	30972045
2.8 ppb	ethanol/water 40/60 (food: rum (aged))	Literature corpus	30972045
2.8 ppb	ethanol/water 40/60 (food: rum (young))	Literature corpus	30972045
4.5 ppb	oil (sunflower oil) (food: gouda cheese)	Literature corpus	30972045
6.5 ppb SIDA	starch	Rychlik 1998 · 1 studies	—
10.0 ppb SIDA	oil_orthonasal	Rychlik 1998 · 1 studies	—
25.0 ppb SIDA	cellulose	Rychlik 1998 · 1 studies	—
66.0 ppb SIDA	beer_orthonasal	Rychlik 1998 · 1 studies	—
100 ppb	wine matrix (12% ethanol) (food: rum / red wine)	Literature corpus	30972045

Only values we can attribute are listed. Where a compendium's licence does not permit public attribution its figures are omitted here, so every number on this page traces to a citable source.

Where you'd actually smell it

Odor Activity Value = measured concentration ÷ odor threshold (4.45 ppb, the engine's canonical value — see the 27 published values). OAV ≥ 1 = perceptible.

Food	Concentration (ppb)	OAV	Source
whey	30000.0	6742 ⚡	DOI
sausage	7920.0	1780 ⚡	DOI
yogurt	5000.0	1124 ⚡	PMID 40438204
coffee	880.0	198 ⚡	Corpus paper
cocoa	190.0	42.7 ⚡	—
wine	100.0	22.5 ⚡	PMID 30704594
date	3.0	0.67	PMID 35300387

Concentrations from PubMed Central Open Access papers (free literature corpus, MIT-style licensed). Copper = OAV ≥ 10 (strongly perceptible). Green = OAV ≥ 1 (perceptible).

Found in

Measured impact first, then profile share

Every ingredient known to contain 3-butanedione. Ingredients with a measured concentration come first, ranked by odour activity; the rest follow by how large a share of that ingredient's profile this molecule is. The figure beside each is how many different compounds that ingredient has in total — not the concentration of 3-butanedione in it. Full concentrations are in the OAV table above.

421 cmpds

128 cmpds

201 cmpds

407 cmpds