Supercritical CO2 extraction pushes carbon dioxide above its critical point (304.18 K, about 31 °C, and 73.80 bar) through ground cured vanilla beans, then releases the pressure so the CO2 turns back into gas and leaves the extract behind. Published laboratory studies on Vanilla planifolia report oleoresin yields of 1.5% to 7.77% of bean weight with pure CO2 in one study and 37.2% with hot ethanol in another, and vanillin contents that depend strongly on pressure and temperature. In the EU, carbon dioxide is listed among the extraction solvents allowed for all uses under good manufacturing practice.
This article covers the CO2 route only. For ethanol extraction, fold and oleoresin in general, read how vanilla extract is made. Our own vanilla oleoresin and the oleoresin range have their own pages; as explained below, our sheet describes a solvent extraction and does not describe it as a CO2 extract.
Key facts
- CO2 critical point: 304.18 K and 73.80 bar (NIST).
- Lab study, Mexican beans ground to 16 to 30 µm: maximum yield 7.77% at 50 °C and 476.19 bar after 40 minutes; highest vanillin concentration reported (97.35% w/w) at 408 bar and 40 °C, with a yield of 5.82% (Castillo-Ruz et al., 2011).
- Hot ethanol (Soxhlet, 12 h) gave 37.2 wt.% oleoresin with 5.9 wt.% vanillin; CO2 fractionation of that oleoresin recovered about 84% of its vanillin in an extract at about 42 wt.% vanillin (Romero de la Vega et al., 2016).
- EU: carbon dioxide appears in Annex I, Part I of Directive 2009/32/EC, "Extraction solvents to be used in compliance with good manufacturing practice for all uses" (EUR-Lex).
- US: carbon dioxide is GRAS as a processing aid, among other uses, with no limitation other than good manufacturing practice (21 CFR 184.1240).
How supercritical CO2 extraction works
CO2 is called supercritical when it is held above both its critical temperature and its critical pressure. NIST lists the critical temperature of carbon dioxide at 304.18 K and its critical pressure at 73.80 bar (NIST).
The laboratory protocol of Castillo-Ruz and colleagues shows the sequence (Castillo-Ruz et al., 2011):
- Grinding. Beans were ground to 16 to 30 µm for most runs; 23 g went into a 0.1 L extraction cell.
- Pressurising. Liquid CO2 was pumped in up to the set pressure and temperature, then left 15 minutes without flow.
- Dynamic extraction. Supercritical CO2 then flowed upwards through the ground beans for 40 minutes, carrying the oleoresin out of the particles.
- Separation. In the collection vessel the CO2 was depressurised "down to atmospheric conditions" and left as a gas, through an iced trap that caught any remaining extract.
Romero de la Vega and colleagues measured the solubility of vanilla oleoresin in CO2 at 140 and 180 bar between 310.5 K and 348.3 K (about 37 to 75 °C) and found "a significant change in vanilla oleoresin solubility (2 to 10 g oleoresin/kg CO2) and selectivity towards the extraction of vanillin" with the operating conditions (Romero de la Vega et al., 2016).
What about ethanol as a co-solvent?
An instrument maker's application note explains that vanillin is "a relatively polar compound" that "dissolves more in polar solvents such as methanol"; for quality testing of beans it used 300 bar, 40 °C and 5% methanol as co-solvent, while noting that "ethanol is the preferred solvent for commercial vanilla extract" (Waters, 2012). In an analytical study, "the modifier ratio, split ratio, and the extraction temperature and pressure were the major parameters" of the extraction; the abstract does not name the modifier (Liang et al., 2018).
The two oleoresin studies we read used pure CO2, and we found no open paper giving yield and conditions for vanilla extracted with CO2 plus ethanol, so this article gives no co-solvent figure. If a supplier uses one, ask which and at what ratio; in the EU, ethanol sits in the same "all uses" list as CO2.
What published studies on Vanilla planifolia report
These are laboratory figures, each tied to one raw material, grind and analytical method: compare them within a study, not across studies.
| Study | Material and set-up | Conditions | Reported result |
|---|---|---|---|
| Castillo-Ruz et al., 2011 | Papantla (Veracruz, Mexico) beans; pure CO2; 23 g per run | 272 bar, 50 °C, 5 mm pieces vs 16 to 30 µm powder | Yield 1.5% (5 mm) vs 4.95% (16 to 30 µm) |
| Same | Same, 16 to 30 µm | 40 to 50 °C, 272 to 476 bar, 40 min | Maximum yield 7.77% at 50 °C and 476.19 bar |
| Same | Same | 408 bar, 40 °C, 40 min | Vanillin 97.35% w/w in the extract (the paper's figure), yield 5.82% |
| Romero de la Vega et al., 2016 | Ethanol oleoresin used as feed | Hot ethanol (95.6 wt.%), Soxhlet, 12 h | Oleoresin 37.2 wt.% of beans, 5.9 wt.% vanillin |
| Same | CO2 fractionation of that oleoresin | 310.5 K (about 37 °C), 140 bar | About 84% of vanillin recovered, extract at about 42 wt.% vanillin; about 25 g CO2 per g oleoresin for nearly complete recovery, extract at about 35 wt.% vanillin |
| Nguyen et al., 1991, as reported by Romero de la Vega et al. | Pure CO2 | 110 bar, 36 °C | Yields "as high as 10.5 wt.%", comparable to ethanol/water extraction, but 160 h for a complete extraction |
What the studies say about trade-offs:
- Grind matters first. At the same pressure and temperature, the fine powder gave more than three times the yield of 5 mm pieces (Castillo-Ruz et al., 2011).
- More pressure and heat raise yield, not necessarily quality. "Extreme conditions of pressure and temperature decrease the quality of vanillin possibly because of the degradation of some compounds"; intermediate pressures of 340 to 408 bar and 40 to 45 °C favoured vanillin concentration (same study).
- Time. From 30 to 40 minutes the yield gain was about double the gain from 40 to 60 minutes, but 40 minutes "leaves solute in the solid matrix", and the authors call for an economic analysis to set the optimum time (same study).
- Selectivity drops as you push recovery. In fractionation, the vanillin content of the extract fell as extraction went on, and going beyond about 25 g CO2 per g produced "an almost free-vanillin oleoresin product" (Romero de la Vega et al., 2016).
The vanillin figures differ widely between sources: 97.35% w/w in one study, about 35 to 42 wt.% in another, 11 to 13% in a commercial CO2 extract (FLAVEX). Romero de la Vega and colleagues note that such differences "could be related with selectivity modifications according to the operating conditions". For a buyer, a "CO2 vanilla extract" is not one product: its specification must state the vanillin content.
Castillo-Ruz and colleagues used Mexican beans; the passages of the Romero de la Vega paper we read do not give the origin of theirs. Neither study is about Madagascar beans. For the choice of raw material, see why extractors use Red US and Red Europe beans.
Is CO2 allowed as an extraction solvent?
Yes in both jurisdictions, under different legal mechanisms.
| Point | European Union | United States |
|---|---|---|
| Status of CO2 | Listed in Annex I, Part I of Directive 2009/32/EC, "Extraction solvents to be used in compliance with good manufacturing practice for all uses", with propane, butane, ethyl acetate, ethanol, acetone and nitrous oxide | 21 CFR 184.1240: GRAS, "used in food with no limitations other than current good manufacturing practice", as a leavening agent, "a processing aid", propellant, aerating agent and gas |
| What an extraction solvent is | A solvent used in extraction "which is removed but which may result in the unintentional, but technically unavoidable, presence of residues" (Art. 1) | We found no CFR section that sets a residue tolerance for CO2; Part 173 sets tolerances for other solvents in spice oleoresins |
| Other solvents, for comparison | Part III, flavourings from natural flavouring materials: hexane 1 mg/kg and dichloromethane 0.02 mg/kg in the foodstuff | Spice oleoresins: acetone 30 ppm, hexane 25 ppm, methylene chloride 30 ppm (total chlorinated solvents) |
| Source | Directive 2009/32/EC, consolidated 16/02/2023 | 21 CFR 184.1240; 21 CFR 173.210, 173.255, 173.270 |
Informative summary established from the texts on 14/10/2026, not legal advice.
On residues, Castillo-Ruz and colleagues write that with supercritical CO2 "no residual solvent is present in the final product", and one maker states "no solvent residues" on its specification (FLAVEX). A co-solvent, if used, is a separate question.
What a CO2 vanilla extract is called on a label
European Union. A product obtained from food or plant material "by appropriate physical, enzymatic or microbiological processes" is a "flavouring preparation" (Regulation (EC) No 1334/2008, Art. 3(2)(d)). The list of traditional processes in Annex II includes "Extraction, incl. solvent extraction in accordance with Directive 88/344/EEC", and Directive 2009/32/EC states that references to that repealed directive "shall be construed as references to this Directive". The word "natural" then follows Article 16:
- "natural" alone may only be used "if the flavouring component comprises only flavouring preparations and/or natural flavouring substances" (Art. 16(2));
- a reference to vanilla, as in "natural vanilla flavouring", requires the flavouring component to come "exclusively or by at least 95 % by w/w" from vanilla (Art. 16(4)).
(Regulation 1334/2008, Art. 3, Art. 16, Annex II). One German supplier declares its vanilla CO2 extract in food as "flavouring preparation or vanilla extract" (FLAVEX). The full EU and US comparison is in our guide to natural vanilla flavouring labelling.
United States. "Vanilla extract" is a standardised food: "the solution in aqueous ethyl alcohol of the sapid and odorous principles extractable from vanilla beans", with at least 35% ethyl alcohol by volume (21 CFR 169.175); below 35% it is "vanilla flavoring" (21 CFR 169.177). The oleoresin appears in that standard as an ingredient form: the vanilla constituent "may be added in the form of ... vanilla flavoring concentrated to the semisolid form called vanilla oleo-resin", and an extract made by diluting it must say "Made from vanilla oleoresin" or "Made in part from". The section does not say how the oleo-resin itself is extracted.
Two other provisions apply to an oleoresin used as an ingredient. The definition of natural flavor covers "the essential oil, oleoresin, essence or extractive" derived from a spice or other plant material (21 CFR 101.22(a)(3)), and vanilla (Vanilla planifolia Andr. or Vanilla tahitensis) is listed under "Essential oils, oleoresins (solvent-free), and natural extractives" generally recognized as safe (21 CFR 182.20).
CO2, ethanol and other solvent oleoresins compared
| Supercritical CO2 | Ethanol | Other solvents (acetone, hexane, dichloromethane) | |
|---|---|---|---|
| Process in the sources | CO2 above 304.18 K and 73.80 bar; lab runs at 272 to 476 bar and 40 to 50 °C for 40 min (Castillo-Ruz) | Percolation with aqueous ethanol at 30 to 40 vol.%, 48 to 72 h for a complete extraction (Romero de la Vega); for oleoresin, multistage leaching at 80 °C with ethyl alcohol (Castillo-Ruz) | No vanilla-specific process source read |
| Solvent removal | Depressurisation to atmospheric conditions | Evaporation under vacuum (Romero de la Vega) | Removed; residues capped |
| Residue rules | EU: GMP, all uses. US: GRAS (184.1240) | EU: GMP, all uses | EU: acetone GMP all uses; hexane 1 mg/kg and dichloromethane 0.02 mg/kg for flavourings. US spice oleoresins: acetone 30 ppm, hexane 25 ppm, methylene chloride 30 ppm |
| Yield and vanillin reported | 1.5 to 7.77% yield (lab, pure CO2); commercial extract at 11 to 13% vanillin, 7 to 9 kg beans per kg (FLAVEX) | 37.2 wt.% oleoresin with 5.9 wt.% vanillin (hot ethanol, lab) | No vanilla-specific figure read |
| Uses named in the sources | Confectionery, ice cream, cakes, desserts, beverages, liqueurs; perfuming in cosmetics (FLAVEX) | US: base for standardised vanilla extract "Made from vanilla oleoresin" (169.175) | Not documented for vanilla in the sources read |
Residue rules: Directive 2009/32/EC, 21 CFR 173.210, 173.255, 173.270, 21 CFR 184.1240. Informative summary established from the texts on 14/10/2026, not legal advice.
The yields in the table are not the same material: in Romero de la Vega and colleagues' work, the hot-ethanol oleoresin was itself the feed that CO2 then fractionated.
Our vanilla oleoresin: what the sheet says about the process
Our technical data sheet (VF-TDS-20) describes our vanilla oleoresin as "a dark, viscous, oil-soluble concentrate obtained by solvent extraction of Bourbon Madagascar vanilla beans (Vanilla planifolia)". The composition line reads "Concentrated vanilla Oleoresin extracted from Bourbon Madagascar vanilla beans".
The sheet does not name the solvent and does not say CO2. Our oleoresin is therefore not presented as a supercritical CO2 extract, and nothing in this article should be read as saying it is one. If your formula or label requires a CO2 extract, raise it before sampling: the sheet states nothing more on the method than the words quoted above.
What the sheet does state:
- concentrations of 20x, 30x and 40x;
- oil-soluble, listed as alcohol-free;
- a dose of 0.02 to 0.1% depending on the application;
- uses in industrial flavourings, fragrances and cosmetics;
- 12 to 24 months at +12 °C to +19 °C, in the original packaging, properly closed;
- bottles of 1, 2 and 10 L and drums as packaging formats. Orders start at 50 kg, the minimum for every grade and format.
How to choose and specify a CO2 vanilla extract
Put these questions to any supplier, CO2 or not.
- Which solvent, exactly? CO2 alone, CO2 with a co-solvent (which one, at what ratio), or ethanol. "Solvent extraction" alone does not tell you.
- Total extract or fraction? Pressure and temperature shift both yield and vanillin selectivity (studies above); a supplier may keep its conditions confidential but can say what kind of product it sells.
- What vanillin content, measured how? The figures in this article run from 5.9% (an ethanol oleoresin) to 97.35% (one CO2 study), depending on the product and the method. Ask for the value and the analytical method on the specification.
- How many kilograms of beans per kilogram of product? One maker publishes 7.0 to 9.0 kg per kg (FLAVEX). This ratio, not the word "fold", tells you what you pay for in bean material.
- Which beans? Species, origin and grade. The studies above did not use Madagascar beans.
- Residues and carrier. Residue against the EU and US limits above for other solvents; pure or on a carrier; oil-soluble or not.
- Regulatory wording. Flavouring preparation, natural vanilla flavouring or oleoresin, and on what basis.
- Handling and storage. One CO2 extract has to be melted and shaken before use (FLAVEX); check the storage conditions and shelf life on the sheet.
FAQ
Is a CO2 vanilla extract natural? In the EU, a product obtained by appropriate physical processes, including solvent extraction under Directive 2009/32/EC, is a flavouring preparation, and "natural vanilla flavouring" requires at least 95% w/w from vanilla (Reg. 1334/2008, Art. 3 and 16). The final wording depends on the whole flavouring, not only the extraction method.
Does supercritical CO2 leave solvent in the extract? The CO2 returns to gas when the pressure is released; one study states that "no residual solvent is present in the final product" (Castillo-Ruz et al., 2011). A co-solvent, if used, is a separate question.
Can a CO2 extract be sold as "vanilla extract" in the US? The US standard defines vanilla extract as a solution in aqueous ethyl alcohol with at least 35% alcohol by volume (21 CFR 169.175). A CO2 extract that is not a solution in aqueous ethyl alcohol at 35% or more does not meet that definition as written; an extract made by diluting an oleoresin must say "Made from vanilla oleoresin".
Is Vanilla Fair's oleoresin a CO2 extract? Our technical data sheet says "solvent extraction" and does not say CO2. We do not present it as a CO2 extract.
Get the oleoresin data sheet or a sample
Tell us your application and the dose range you work with. We send the technical data sheet of our oleoresin and samples on request, so you can test it in your own matrix. Orders start at 50 kg, the minimum for every grade and format.
Request the technical data sheet · Request an oleoresin sample · Browse the oleoresin range
Sources
- Castillo-Ruz M.C., Guillermo-Alcocer C.G., Bojórquez-Gamboa R.R., Rocha-Uribe J.A. (2011), Extraction of vanilla oleoresin (Vanilla planifolia Andrews) with supercritical CO2, Tecnología, Ciencia, Educación 26(2):80-84 (full text, Redalyc).
- Romero de la Vega G., Salgado Cervantes M.A., García Alvarado M.A., Romero Martínez A., Hegel P.E., Fractionation of vanilla oleoresin by supercritical CO2 technology, Journal of Supercritical Fluids 108 (2016) 79-88, doi 10.1016/j.supflu.2015.10.022 (open copy, CONICET).
- Liang Y. et al. (2018), Determination of major aromatic constituents in vanilla using an on-line supercritical fluid extraction coupled with supercritical fluid chromatography, Journal of Separation Science 41(7) (abstract).
- NIST Chemistry WebBook, Carbon dioxide, phase change data.
- EU law: Directive 2009/32/EC (consolidated 16/02/2023); Regulation (EC) No 1334/2008, Article 3, Article 16, Annex II (text as adopted).
- US law (Cornell LII, consulted 14/10/2026; text checked against the eCFR edition of 09/10/2026): 21 CFR 169.175, 169.177, 101.22, 173.210, 173.255, 173.270, 182.20, 184.1240.
- Trade and instrument sources, quoted as such: FLAVEX, General Specification, Vanilla Planifolia CO2-to extract (organic) (17/09/2024); Waters Corporation, Quick Quality Assessment of Vanilla Beans Using SFE (2012).
- Vanilla Fair: technical data sheet Oleoresin, VF-TDS-20.
Data verified on 14/10/2026. This article is informational: for a given lot, the technical data sheet and the label prevail.




