Tadka chemistry: why it is the same extraction as perfume

A spoon of hot oil over dal and a glass plate of lard under jasmine petals run on the same law of chemistry. The difference is 150 degrees, and it decides what you smell.

Close-up of dried cumin seeds, each about 5 mm long
Photo: Sanjay Acharya, CC BY-SA 4.0, via Wikimedia Commons

A tadka is a fat extraction, and so is classical perfume: in both, an oil or fat pulls aroma molecules out of plant tissue because those molecules are far more soluble in fat than in water, up to about 24,000 times more for limonene. Perfumers did it cold, on lard, over days, to keep the delicate notes. A cook does it at about 180°C in under a minute, which makes extraction roughly 60 times faster and also boils off part of the lightest aromas.

Key takeaways
  • Nernst's distribution law runs both: aroma divides between fat and water in a fixed ratio.
  • log P 4.38 for limonene means about 24,000 parts in oil for every 1 in water.
  • 1.6–2.9% of a cumin seed is essential oil, and 36–50% of that is cuminaldehyde.
  • About 60× faster diffusion in oil at 180°C than at 25°C, by the Stokes–Einstein relation (my estimate).
  • 0.11–0.16% was the oil yield from narcissus petals by cold enfleurage in a 2025 study.
  • Verdict: the link holds; heat is the only real difference, and it costs top notes.

Put the two scenes side by side. In a Bengaluru kitchen, a spoon of mustard seeds and cumin crackles in hot oil and gets poured over dal. In 18th-century Grasse, workers pressed jasmine petals onto glass plates smeared with purified lard, and changed the petals every day or two until the fat smelled like a garden.

One is dinner and one is a luxury good. The chemistry is the same: solvent extraction with a fat. Here I solve both with the same equations, then show exactly where they part ways.

What does tadka chemistry actually extract?

A tadka extracts a spice's essential oil, the small fraction of volatile, fat-loving molecules that carry its smell. In cumin that fraction is about 2% of the seed by weight, and a single molecule, cuminaldehyde, is close to half of it.

The best recent numbers come from a 2020 study of cumin seed from four countries. It also shows that most of the seed's weight is not aroma at all but ordinary vegetable fat.

Biochemical Composition of Cumin Seeds, and Biorefining StudyMerah O, Sayed-Ahmad B, Talou T, Saad Z, Cerny M, Grivot S, Evon P, Hijazi A · Biomolecules 10(7):1054 · 2020

The team analysed cumin seed from Lebanon, France, Algeria and Syria. They hydrodistilled 200 g samples for 3 hours to get the essential oil (1.6 to 2.9% of dry matter) and Soxhlet-extracted the fixed oil (13.4 to 29.1%). Cuminaldehyde made up 36.2 to 49.5% of the essential oil, followed by p-cymene, β-pinene, γ-terpinene and two menthadienals. The limit: hydrodistillation boils seed in water for hours, so it measures what the seed contains, not what a 40-second tadka releases.

Read the paper (free)
Main aroma molecules in cumin essential oil, by origin
See the data as a table
Lebanon (%)France (%)Algeria (%)Syria (%)
Cuminaldehyde43.8943.9349.5336.19
p-Cymene13.7710.0410.37.77
β-Pinene11.4912.6110.568.21
γ-Terpinene10.66.448.6713.14
1,4-p-Menthadien-7-al9.8217.2217.0717.47

Source: Merah et al., Biomolecules 2020 · Share of the hydrodistilled essential oil, GC-MS.

Here is what that means for one spoonful:

  • Spice: 2 g of whole cumin, about one level teaspoon (my estimate of a typical home dose).
  • Aroma available: 2 g × about 2% = roughly 40 mg of essential oil.
  • The star molecule: about 40% of that, so roughly 16 to 20 mg of cuminaldehyde.
  • Solvent: 2 tablespoons of oil or ghee, about 30 ml or 27 g.

Forty milligrams of oil perfumes a whole pot of dal. That tiny dose is the first thing a tadka shares with perfume, where an absolute is measured in grams per tonne of flowers.

How does enfleurage, the old perfume method, work?

Enfleurage is extraction with a solid fat at room temperature: flowers sit on lard or tallow for one to three days, the spent petals are swapped for fresh ones, and the scented fat, called a pomade, is washed with ethanol to give an absolute. It was developed around Grasse in southern France and was the only way to capture flowers like jasmine and tuberose that heat would ruin.

A 2025 lab study ran the method exactly and measured what the fat caught.

Comparison of volatile oil compositions of five Narcissus (Narcissus tazetta L.) varieties extracted by enfleurage and HS-SPME methodsMahen MY, Mahmoodi Sourestani M, Motamedi H, Seyyed Nejad SM · International Journal of Secondary Metabolite 12(3):593–603 · 2025

The authors spread 200 g of palm lard on 50 × 50 cm glass plates, laid on 300 g of fresh narcissus petals for about 48 hours, and repeated with fresh petals four times. The pomade was soaked in ethanol for six days, filtered and evaporated at 45°C. Oil yield was 0.11 to 0.16% across five varieties. In the fat, benzyl acetate dominated (31.5 to 50.4%); in the air above live flowers, measured by headspace sampling, the lighter trans-β-ocimene dominated instead. The limit: one region, one fat, and the evaporation step itself can lose the lightest molecules.

Read the paper (free)

That last finding is the hinge of this post. Fat does not copy a smell; it filters it, keeping heavier, oilier molecules and losing the lightest ones. A tadka does the same thing, only harder, because it is 150 degrees hotter.

White Jasminum sambac flowers among green leaves
Jasminum sambac, native to India. Jasmine was one of the flowers too delicate for steam, so perfumers pressed it into fat instead.Photo: Mokkie, CC BY-SA 3.0, via Wikimedia Commons
The story

India has its own version, and it sits halfway between perfume and tadka. In Kannauj, in Uttar Pradesh, attar makers load flowers and water into copper stills called degs, set over wood or charcoal fires. The steam travels through bamboo pipes into long-necked copper receivers called bhapkas, which stand in water tanks to cool. Inside each bhapka waits a charge of sandalwood oil. The fragrant vapour condenses straight into that oil, which captures and fixes the scent. The method is hydrodistillation, but the trap is fat: the flower's aroma ends in an oil, exactly as cumin's ends in your ghee. Kannauj's attar received a Geographical Indication tag in 2014, and a BBC report cited by the article counted over 700 distilleries two decades earlier and fewer than a hundred now.

Live History India: Kannauj, Perfume Capital of India (2020)

Why do aroma molecules prefer fat to water?

Aroma molecules are mostly small hydrocarbons with few polar groups, so water squeezes them out and oil lets them in; the ratio is fixed by the partition coefficient. This is Nernst's distribution law: at equilibrium, a solute divides between two immiscible liquids in a constant ratio at a given temperature.

\[ K = \frac{C_{oil}}{C_{water}}, \qquad \log P = \log_{10} K_{octanol/water} \]

Chemists measure it with octanol standing in for fat. A 2020 study of aroma release from emulsions tabulated it for common flavour molecules, and the spread is enormous.

Effect of Oil–water Surface Area on the Aroma Release Behavior of Mono-dispersed Oil-in-water EmulsionsTamaru S, Noda T, Igura N, Shimoda M · Food Science and Technology Research 26(2):293–298 · 2020

The Kyushu University team made coconut-oil-in-water emulsions (20% oil) with uniform droplets of 0.4, 2 and 20 µm, added seven aroma compounds at 100 ppm each, and measured release by dynamic headspace GC. Log P (octanol/water) ranged from 0.21 for 2-methylpyrazine to 4.38 for limonene. The finest emulsion released aroma about 1.4 times faster than the coarsest. Compounds with an oil/air log P above 5 released slowest. The limit: an emulsion at room temperature, and partition values were calculated with software rather than measured.

Read the paper

Worked example: where does the aroma go when the tadka hits the dal?

Pour 30 g of scented oil onto 500 g of watery dal and let the two phases settle. Treat oil like octanol, which is an approximation, and use mass ratios as volume ratios.

  1. Mass balance. The fraction of a molecule that stays in the oil is
    \[ f_{oil} = \frac{K\,V_{oil}}{K\,V_{oil} + V_{water}} \]
  2. A fat-loving molecule (limonene, log P 4.38). \( K = 10^{4.38} \approx 24{,}000 \).
    \[ f_{oil} = \frac{24{,}000 \times 30}{24{,}000 \times 30 + 500} = \frac{720{,}000}{720{,}500} \approx 99.93\% \]
  3. A water-friendly molecule (2-methylpyrazine, log P 0.21). \( K = 10^{0.21} \approx 1.6 \).
    \[ f_{oil} = \frac{1.6 \times 30}{1.6 \times 30 + 500} = \frac{48}{548} \approx 8.8\% \]

So the terpene-type aromas of cumin, mustard and curry leaf ride in the oil film on top of the dal, which is why you smell a tadka the moment the bowl arrives. Roasty pyrazines, which are more water-soluble, spread into the body of the dish.

oil / ghee (30 g)water in the dal (500 g)limonene, log P 4.38: about 99.9% stays in oil2-methylpyrazine, log P 0.21: only about 9% in oil
Nernst's distribution law in a bowl of dal. The same law decides what a perfumer's lard holds on to.

What does 180°C change compared with cold fat?

Heat changes two things at once: it makes molecules diffuse out of the spice dozens of times faster, and it raises their vapour pressure several hundredfold, so some of the aroma escapes into the kitchen. Cold enfleurage gives up speed to keep the top notes; tadka gives up top notes to get speed.

Speed: the Stokes–Einstein relation

A molecule moving through a liquid diffuses at a rate set by temperature and the liquid's viscosity. The Stokes–Einstein relation says

\[ D = \frac{k_B T}{6\pi \eta r} \quad\Rightarrow\quad \frac{D_{180}}{D_{25}} = \frac{T_{180}}{T_{25}} \cdot \frac{\eta_{25}}{\eta_{180}} \]

Oil thins sharply when hot. Its viscosity follows an Arrhenius law, \( \eta = \eta_\infty e^{E_a/RT} \), and a 2014 frying study measured the activation energy.

Rheological characteristics of vegetable oils as affected by deep frying of French friesHassanien MFR, Sharoba AM · Journal of Food Measurement and Characterization 8:171–179 · 2014

The authors fried French fries in sunflower oil, cottonseed oil and palm olein for up to 16 hours and measured flow behaviour, fatty acids and polar compounds. Fresh oils were Newtonian. Fitting viscosity from 5 to 100°C to an Arrhenius model gave activation energies of 27.5 kJ/mol for fresh sunflower oil, 39.0 for cottonseed and 47.1 for palm olein. Viscosity rose with frying time and correlated with oleic and linoleic acid content. The limit: the fits stop at 100°C, so using them at 180°C is an extrapolation.

Read the paper
  1. Viscosity ratio for sunflower oil, with \( E_a \) = 27,450 J/mol, 298 K and 453 K:
    \[ \frac{\eta_{25}}{\eta_{180}} = \exp\!\left[\frac{27{,}450}{8.314}\left(\frac{1}{298} - \frac{1}{453}\right)\right] = e^{3.79} \approx 44 \]
  2. Diffusion ratio.
    \[ \frac{D_{180}}{D_{25}} = \frac{453}{298} \times 44 \approx 67 \]
  3. Sanity check. The fit was made between 5 and 100°C, and oils usually thin more slowly at high temperature than a straight Arrhenius line predicts, so treat 44 as an upper bound. My honest range for the speed-up is about 20 to 70 times.

Diffusion time scales as \( t \approx L^2/D \). A cumin seed is about 5 mm long and about 1.5 mm across, so the escape path is under a millimetre. A 20 to 70 times faster \(D\) turns a job of hours into one of minutes. The rest of the gap to a 40-second tadka comes from what cold fat cannot do: seeds split and pop as their water flashes to steam, and stirring keeps fresh oil at the surface.

Loss: the Clausius–Clapeyron equation

Hot oil also pushes aroma into the air. The Clausius–Clapeyron equation gives a liquid's vapour pressure at any temperature from its boiling point, and Trouton's rule supplies the missing heat of vaporisation: for most liquids, \( \Delta H_{vap} \approx 88\ \text{J/mol·K} \times T_b \).

\[ P_{sat}(T) \approx 1\ \text{atm} \times \exp\!\left[-\frac{88\,T_b}{R}\left(\frac{1}{T} - \frac{1}{T_b}\right)\right] \]

For cuminaldehyde, \( T_b \) = 235.5°C = 508.7 K, so \( 88 T_b / R \approx 5{,}384 \) K. At a 180°C tadka:

\[ P_{sat}(453\,\text{K}) = \exp[-5{,}384 \times (0.0022075 - 0.0019659)] = e^{-1.30} \approx 0.27\ \text{atm} \]

At a 25°C enfleurage table the same sum gives about 0.00056 atm. That is a ratio of almost 500. These are estimates, since Trouton's rule is rough for real molecules, but the order of magnitude is the point.

MoleculeFound inBoiling pointPsat at 25°C (est.)Psat at 180°C (est.)
LimoneneCitrus peel, many spices176°C0.005 atm1.1 atm
LinaloolCoriander seed, jasmine198–199°C0.002 atm0.65 atm
Benzyl acetateJasmine, narcissus212°C0.001 atm0.47 atm
CuminaldehydeCumin235.5°C0.0006 atm0.27 atm
EugenolClove254°C0.0003 atm0.18 atm

Boiling points are from the compounds' standard data; vapour pressures are my Clausius–Clapeyron and Trouton estimates, so read them as orders of magnitude, not lab values.

Limonene is above its boiling point in a 180°C pan. It does not all vanish, because it is dissolved in oil, and Raoult's law cuts its partial pressure by its mole fraction. For cuminaldehyde: 18 mg is \(1.2\times10^{-4}\) mol, 27 g of oil is about 0.031 mol of triglyceride, so \( x \approx 0.004 \) and the partial pressure is about \( 0.004 \times 0.27 \approx 0.001 \) atm, roughly 100 Pa. Small enough to survive 40 seconds; large enough that ten minutes of overheated oil would strip it.

Three routes, one trap: fatCold enfleurageKannauj attarTadkapetals on lard25°C, 1–3 daysflowers + waterboiled in a degseeds in oil160–200°Crepeat withfresh petalssteam throughbamboo pipeseeds pop,oil penetratespomadewashed in ethanolcondenses insandalwood oil30–60 s, thenpour onto dalabsolutekeeps heavy notesattaroil-based perfumescented oil filmloses lightest notesSources: Mahen et al. 2025; Live History India 2020; Merah et al. 2020
Grasse, Kannauj and your kitchen all end with aroma dissolved in fat. They differ in temperature and time, which decides what survives.

How strong is a tadka compared with a perfume pomade?

A narcissus pomade from the 2025 study is roughly five to seven times more concentrated in aroma than a typical tadka oil, because the fat was recharged with fresh flowers four times. Run the sums for both:

  1. Pomade. 4 × 300 g of petals at 0.11 to 0.16% yield gives 1.3 to 1.9 g of oil in 200 g of lard:
    \[ c_{pomade} = \frac{1.3\ \text{to}\ 1.9\ \text{g}}{200\ \text{g}} \approx 0.65\text{–}0.95\% \]
  2. Tadka, assuming all 40 mg of cumin oil transfers to 27 g of oil:
    \[ c_{tadka} = \frac{0.040\ \text{g}}{27\ \text{g}} \approx 0.15\% \]
  3. Ratio. \( 0.65/0.15 \approx 4.4 \) to \( 0.95/0.15 \approx 6.4 \).

In practice a tadka transfers well under 100%, since seeds stay mostly whole. Even so, a home tempering is the same order of strength as a perfumer's pomade, which is a striking thing to know about your kitchen.

The old Chiris perfume factory building in Grasse, France
The former Chiris perfumery in Grasse, one of the town's historic houses. Grasse built its trade on extracting flowers into fat and solvent.Photo: Patrick Rouzet, CC BY-SA 4.0, via Wikimedia Commons

Is the sensitivity of the answer to temperature large?

Yes: within the normal 160 to 200°C tadka range, the volatility of cuminaldehyde roughly triples, while extraction speed changes by under a factor of two. That is why cooks watch the pan so closely.

Oil temperatureCuminaldehyde Psat (est.)Limonene Psat (est.)What happens
25°C0.0006 atm0.005 atmEnfleurage: slow, keeps top notes
160°C0.16 atm0.68 atmGentle tadka: seeds sizzle slowly
180°C0.27 atm1.1 atmTypical tadka: seeds crackle in seconds
200°C0.45 atm1.7 atmHot tadka: fast pop, real risk of scorching

Three assumptions could move the result, so here is how each one shifts it:

  • Oil type. Palm olein's activation energy (47.1 kJ/mol) is higher than sunflower's, so its speed-up from 25 to 180°C is even larger.
  • Octanol as a stand-in for oil. Real triglycerides are a little less welcoming than octanol to slightly polar molecules, so log P overstates the oil share for mid-range compounds. For terpenes with log P above 3, the conclusion holds either way.
  • Whole versus ground spice. Grinding shortens the diffusion path \(L\) from about 0.75 mm to tens of microns, and since \( t \propto L^2 \) that cuts extraction time several hundredfold. It is why powdered spices burn so easily in a tadka.

Use this to check what you just read with your own spice and oil.

Tadka vs perfume: where does the aroma go?

Where does the analogy break down?

The link holds for the extraction step and breaks at three points: a tadka also cooks, it has no second solvent, and it is eaten within the hour.

  • Tadka makes new molecules. At 160 to 200°C, sugars and amino acids in the seeds start Maillard browning and make pyrazines and furans that were never in the raw spice. Cold enfleurage makes nothing new; it only moves molecules.
  • Perfume has a second extraction. The pomade is washed with ethanol, a solvent in the middle of the polarity scale, to separate scent from fat. A tadka's second solvent is the dal itself, and by the partition sum above, the fat-loving molecules mostly refuse to move.
  • Time scale. Perfume must last months in a bottle, so it needs fixatives such as sandalwood oil in Kannauj. A tadka has to last ten minutes on a plate, which is why it goes on last.
A tadka is hot enfleurage with a deadline: the same fat-trap, run 150 degrees hotter and thousands of times faster.From the working above
Verdict: the link holds

Tadka and classical perfume extraction are the same physical process: an oil captures volatile, fat-loving aroma molecules because Nernst's distribution law favours them by factors up to about 24,000 to 1. Temperature is the one real difference. At 180°C, diffusion in oil runs about 20 to 70 times faster than at 25°C, which is why a tadka takes seconds instead of days, while the vapour pressure of a molecule like cuminaldehyde rises almost 500-fold, which is why a tadka loses the lightest top notes that cold enfleurage was invented to keep.

Traditional camel-skin bottles used to age attar in Kannauj
Camel-skin bottles from Kannauj, used to age attar. The skin lets water escape while the oil holds the fragrance.Photo: David Prasad, CC BY-SA 2.0, via Wikimedia Commons

What can you do with this in the kitchen?

Treat your tadka like a perfumer would: choose the solvent, control the temperature, and protect the volatile notes. Four habits follow straight from the equations:

  1. Whole seeds first, powders last. Whole seeds need the heat to break open; powders have a tiny diffusion path and scorch within seconds, so add them off the flame.
  2. Stop at the crackle. Once mustard seeds pop and cumin darkens a shade, extraction is mostly done. Every extra 20 seconds mainly feeds the extractor fan.
  3. Pour it on top, at the end. The aroma lives in the oil film. Stirring it deep into the pot and boiling again sends it into the steam.
  4. Match the fat to the job. Ghee and mustard oil carry their own aromas, which become part of the extraction, much as Kannauj's sandalwood oil becomes part of every attar.

The order matters as much as the temperature, and for the reasons behind that, see why spice order in a curry matters. For the hot-oil chemistry of individual spices, from asafoetida to curry leaves, read what happens to spices in hot oil. The browning that cold fat never produces is covered in the Maillard reaction in Indian cooking, and the bigger question of why Indian cooks use so many aromatics at all is in why Indian food uses so many spices. The whole Taste Lab is on the Rasa world page.

What if next?

Questions people ask

Why are spices fried in oil instead of boiled in water?

Most spice aroma molecules are fat-soluble. Limonene, for example, has a log P of 4.38, so it is about 24,000 times more soluble in an oil-like solvent than in water. Hot oil dissolves and carries those molecules; boiling water extracts them poorly and drives many off with the steam.

What temperature should oil be for tadka?

Roughly 160 to 190°C, hot enough for whole seeds to sizzle at once and for mustard seeds to pop within seconds, but well below smoking. Above about 200°C the volatility of aroma molecules climbs steeply and powders scorch.

What is enfleurage in perfume?

It is a method that captures flower scent in fat. Petals sit on odourless lard or tallow for one to three days and are replaced until the fat is saturated; the scented fat, or pomade, is then washed with ethanol to give an absolute. It is now largely replaced by solvent and CO2 extraction.

Is attar the same as enfleurage?

Not quite. Kannauj attar is made by hydrodistillation in copper degs, with the vapour condensing into sandalwood oil in a bhapka. Like enfleurage, it ends with aroma trapped in an oil, but the flowers are heated, as in a tadka.

Sources

  1. Merah O et al.: Biochemical Composition of Cumin Seeds, and Biorefining Study, Biomolecules (2020) (essential oil 1.6–2.9%, cuminaldehyde 36–50%, chart data)
  2. Mahen MY et al.: Comparison of volatile oil compositions of five Narcissus varieties, International Journal of Secondary Metabolite (2025) (enfleurage method, 0.11–0.16% yield, benzyl acetate vs ocimene)
  3. Tamaru S et al.: Effect of oil–water surface area on aroma release from O/W emulsions, Food Science and Technology Research (2020) (log P values for limonene and 2-methylpyrazine)
  4. Hassanien MFR, Sharoba AM: Rheological characteristics of vegetable oils as affected by deep frying, Journal of Food Measurement and Characterization (2014) (Arrhenius activation energies of oil viscosity)
  5. Live History India: Kannauj, Perfume Capital of India (2020) (deg-bhapka process, sandalwood oil base, GI tag, distillery numbers)
  6. Wikipedia: Enfleurage (cold and hot enfleurage, 1–3 day charges, pomade and absolute, Grasse origin)
  7. Wikipedia: Cuminaldehyde (boiling point 235.5°C)
  8. Wikipedia: Limonene (boiling point 176°C)
  9. Wikipedia: Linalool (boiling point 198–199°C)
  10. Wikipedia: Eugenol (boiling point 254°C)
  11. Wikipedia: Benzyl acetate (boiling point 212°C)
  12. Wikipedia: Trouton's rule (heat of vaporisation about 88 J/mol·K times boiling point)
  13. Wikipedia: Partition coefficient (Nernst distribution law and log P)
tadkaenfleuragepartition coefficientNernst distribution lawcuminaldehydeClausius-ClapeyronKannauj attarStokes-Einstein
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Foodblog.

Keep reading in Foodblog

From the other blogs