London Dry · Cane base · Made for Passover, no kitniyot

Gin de Cana

The third bottle of the kosher Negroni. Built on the nine botanicals Beefeater declares, at the strength bartenders say it takes to cut through the Campari, and without coriander — which is the recipe's knot.

Yield
1 L
Strength
44 % ABV
Botanicals
13
Juniper
10 g/L
Start to glass
9 days

Ingredients · Makes 1 litre at 44%

Fraction 1 · 7 days

  • 10 gjuniper berries, dried and broken
  • 2.5 gangelica root
  • 2 gsweet almond, flaked
  • 0.8 gdried Surinam cherry leaf
  • 0.4 gpriprioca rhizome

Fraction 2 · 48 hours

  • 3 gdried bitter orange peel
  • 2 gdried lemon peel
  • 0.8 gpink pepper optional
  • 0.5 gblack peppercorns
  • 0.3 gbay leaf, dried
  • 0.2 gstar anise optional

Fraction 3 · 12 hours

  • 2.5 gdried orange blossom
  • 0.3 gculinary Lavandula angustifolia

The charge · mix first

  • 458 mlneutral cane alcohol at 96%
  • 542 mlfiltered water, 422 ml in the charge and 120 ml to finish

Method ·

  1. Before · shoppingMix the charge. Mix 458 ml of cane alcohol at 96% with 422 ml of water: that gives 880 ml of charge at 50%. Weigh the botanicals on a 0.01 g scale and crack the juniper berries with the bottom of a glass, without grinding them.
  2. Day 0 · 7 daysFraction 1: juniper, roots and almond. Set aside two thirds of the charge. Into it go 10 g of cracked juniper, 2.5 g of chopped angelica root, 2 g of flaked almond, 0.4 g of priprioca and 0.8 g of Surinam cherry leaf. Dark glass, filled to the neck, out of the light, shaken once a day. The remaining third goes sealed into the fridge.
  3. Day 5 · 48 hoursFraction 2: peels and spices. Into half of the reserved third go 3 g of dried bitter orange peel, 2 g of dried lemon peel, 0.5 g of black peppercorns, 0.3 g of bay and, if you use them, 0.8 g of pink pepper and a point of star anise. Forty-eight hours, not one more: after that the peel gives up the bitterness of the pith.
  4. Day 6 · 12 hoursFraction 3: the flowers. Into the other half go 2.5 g of dried orange blossom and 0.3 g of lavender. Twelve hours only, at room temperature. Flowers steeped longer taste of old tea.
  5. Day 7 · filterFilter and set the reserve aside. Strain all three fractions through a sieve and then through paper, without pressing. Set aside 20% of fraction 1 in a jar: that is the reserve for adjusting the juniper by taste. Combine the rest and top up with water to one litre.
  6. Day 7 · tasteAdjust in a Negroni, not neat. Test Negroni: 20 ml of this gin, 20 ml of Bitter Rosso, 20 ml of Vermouth Rosso and 15 ml of cold water, no ice. If the juniper disappears, drip the reserve into 100 ml of gin, five drops at a time, and taste again. For the batch, multiply the total by ten.
  7. Day 7 to 9 · 2 daysRest and bottle. Two days in the fridge drop whatever is still going to precipitate. Filter through paper while still cold and bottle in dark glass, filled to the neck.

Where to buy each thing

Supermarket
anywhere
flaked almond, black peppercorns, bay, star anise, one lemon and one orange
Herb shop
whole, single-ingredient
juniper, angelica, bitter orange, orange blossom, Surinam cherry leaf, priprioca, pink pepper, lavender
The hard one
not sold in small packs with a Passover hechsher
neutral cane alcohol at 96% · Suppliers
The coriander swap. No coriander, which is kitniyot for the cRc and the Star-K. If you follow the OU, drop the orange blossom and the lavender and put 6 g of coriander seed into fraction 1.
01

What is known and what is reconstruction

For gin, unlike the bitter and the vermouth, there is plenty of published science. What there is not is an analysis that names Beefeater.

What Beefeater declares. Nine botanicals, named, in English and in Portuguese: juniper, Seville orange peel, lemon peel, coriander seed, angelica root, angelica seed, almond, liquorice and orris root. The list has not changed across site captures from 2020 to 2026. The house also declares the process: 24 hours of maceration inside the still itself, and the official blog contrasts this with the vapour basket, to make clear which of the two it uses. Nine Victorian copper stills in Kennington. The proportions have never been disclosed, and the house does not say what grain the alcohol comes from.

What Tanqueray declares. Four: juniper, coriander seed, angelica root and liquorice. The American label says "100% Grain Neutral Spirits". Nothing on process: "distilled four times" and the "Old Tom" still only appear in third-party sources.

What science has measured. A great deal, but never with the brand named. Six gins identified by code in 2005 [1], two Bavarian gins in 2020 [4], sixteen gins in 2023, plus work on how each compound behaves through distillation [5][8]. It is this literature, not the label, that says how much of each botanical to use.

What is reconstruction. Every dose. The choice of strength. And the four plants that took the place of coriander, liquorice, orris and cardamom — each chosen for the compound it delivers, not for its description.

Tags

Confirmed Beefeater declares it. Inferred deduced from tasting notes or from the chemistry. Substitution this project's choice, for kashrut or for availability. Doses are always an estimate.

No off-the-shelf gin works for Passover

Star-K lists Beefeater and Tanqueray as "Approved", but the agency's own article explains that unflavoured gin needs no hechsher for everyday use: it is not certification requested by the manufacturer, and it does not hold for Passover. The cRc's 2026 Passover liquor list carries four gins — Dalton, Dead Drop, Matar Pink Lady and "Spirits Gin" — and neither of these two. Tanqueray once carried a London KF seal, withdrawn around January 2016. Grain gin is chametz, without exception.

02

Flavour profile

Twenty-four independent sources on Beefeater, in English, German, Italian and Spanish. No independent review in Portuguese: the Brazilian ones are shop copy.

DescriptorSourcesLikely compoundIn the clone
Juniper21 of 24
and producer
α-pinene, 27.7–49% of the berry oil [10]; the oil is 0.4–1.9% of the berry, or 0.2–3.42% by another review [3]10 g/L, above the standard. It is what the bitter and the vermouth lack.
Orangepeel, bitter11 of 24
and producer
Limonene (90.6% in bitter orange [14])Dried bitter-orange peel. A direct bridge to the bitter.
Liquorice, anise11 of 24
and producer
Volatiles from the gin itself, not glycyrrhizin [2]Out. See the note below: this descriptor does not come from the liquorice.
Lemon10 of 24
and producer
Limonene 55%, neral 10.4% [16]Dried peel of limão-siciliano or limão-cravo.
Coriander10 of 24
and producer
Linalool, 49–82% of the seed oil [11]; the highest-impact compound in gin [4]Out of the main recipe, for kitniyot. Orange blossom and lavender deliver the linalool.
Earthy, root8 of 24
and producer
β-phellandrene and α-pinene from angelica [18]; musk lactones 1.3% in the root [17]Angelica root, plus priprioca in place of orris.
Pepper7 of 24Piperine and monoterpenes; a carry-over from coriander and angelicaWhole black peppercorns. Without the coriander, it has to be made explicit.
Sweet7 of 24An impression, not sugar: the label declares 0 g of carbohydrateComes from the almond and the limonene. Nothing is added.
Alcoholic heat7 of 24EthanolExpected at 44%. It vanishes in the Negroni.
Almond, marzipan5 of 24
and producer
Trace benzaldehyde; oils that give bodyFlaked sweet almond.
Pine, resin6 of 24α-pinene; threshold of 1650 µg/L [5]Juniper plus Surinam cherry leaf and pink pepper.
Floral, violet4 of 24
(violet 1)
Irones from the orrisNot reproduced: see the note on orris in section 05.
Dry2 of 24—Only two sources call Beefeater dry, against seven for "sweet". It was the count's biggest surprise.
Liquorice does not do what it seems to

The 2008 normalised sensory panel [2] settled on five descriptors for gin — juniper, citrus, anise, spice and liquorice — and liquorice scored the same across every gin analysed, including the ones with no liquorice in them. Glycyrrhizin is a glycoside and does not carry through distillation; in macerated liqueurs it turns up at 1 to 16 mg/L [23]. In other words: the "liquorice" note belongs to gin itself, not to the root. Dropping it costs less than the 11 of 24 votes suggest.

Critical reading

Shop copy in Italian and in Portuguese was discarded: it repeats the distributor. One Portuguese text on Beefeater reads like a translation of an English shop note and was not counted. The "short finish" readings are all from the 47% version, and the "long finish" ones from the weaker versions — what separates the sources is the bottle each one tasted, not the gin.

03

The types of gin

"London Dry" is a method, not a place. What separates the types is what you may add after distillation.

TypeThe ruleThe methodProfile and strength
London DryLondon ginEU cat. 22 [30]: methanol up to 5 g/hL, distilled to ≥70%, no colour, sweetener up to 0.1 g/L, and only alcohol, botanicals and waterEverything distilled together. Nothing is added afterwards, no colour, no aroma.Juniper up front, dry. Minimum 37.5%; in practice 40 to 47%
Distilled ginEU cat. 21 [30]: redistillation of ≥96% agricultural alcohol with juniperAllows flavouring afterwards and distillates made separatelyVariable. Tanqueray No. Ten is one of these, because its citrus heart is distilled separately
Compound gin"bathtub"EU cat. 20 [30]: juniper predominant, with no distillation requiredMaceration or extract in neutral alcoholHas the colour and texture of an infusion. This project's route
Navy StrengthEstablished usage, not a legal categoryDistilled, bottled strong57%, the strength of the gunpowder test
Old TomNo EU category; in Brazil, until 2025, 6 to 15 g/L of sugarSweetened ginLightly sweet. 40 to 43%
PlymouthWas a geographical indication until 2015, when the brand let it lapseDistilledMore earthy, softer juniper. 41.2%
GeneverEU geographical indication, category 19 [30]Moutwijn of rye, maize and malted barley, redistilled with juniperMalty, close to whisky. Chametz
ContemporaryNew WesternA 2006 term, with no legal standingAny methodJuniper shares the stage. A panel of bartenders called it "unsuitable" for Negroni
Sloe ginEU cat. 35: it is a liqueur, minimum 25%Sloe berries macerated in gin, with sugarFruity, sweet
Brazilian law changed in 2025

Decree 6.871/2009, which defined four types of gin and a cap of 15 g/L of sugar, was repealed by Decree 12.709, of 31/10/2025 [31]. The new article keeps 35 to 54% and requires potable ethyl alcohol of agricultural origin redistilled "with juniper or juniper extract", with juniper predominant — but the types and the sugar cap dropped out of the text. Neutral cane alcohol meets the agricultural-origin requirement.

Two consequences for us: a macerated gin, without redistillation, is neither "gin" under the law nor "London Dry" under the European rule. It is a drink for personal use, and this page says so instead of pretending otherwise.

04

What wins the Negroni

Beefeater is the recipe reference. But what wins the blind tastings is Tanqueray, and the reason is not the botanical list.

AspectBeefeaterTanqueray
Declared9, named4, named
Strength40% in the UK, the EU and Brazil. In the US: 47% → 44% (2020) → 40% (2023)47.3% in the US and Brazil; 43.1% in Scandinavia; 41.3% in the UK since 2023
ProfileJuniper 21/24, orange 11, liquorice 11, lemon 10, coriander 10, pepper 7Juniper 19/21, citrus 15 ("subtle"), coriander 12, liquorice 10, pine 8, orange 1
Body"Velvety", oily at 47%; "mild" at 40%Creamy and round, in 3 of 4 comparisons
In the NegroniOne win in head-to-head. Recommended with no comparison by three sources, for its "thick seam of bitter-orange peel and coriander"Four wins. In the 2015 double-blind, 6 to 3. In a panel of five bartenders, 86 against 83
AgainstAt 44%, "the Campari covers it". In one panel, 6th place, "juniper in the background""The juniper steals the complexity"
What actually changes

Today's Beefeater is the same recipe as in 2019 and seven points weaker in the United States. The reviews saying it vanishes in the Negroni are all of the 40% and 44% versions; the ones calling it "truly great" are of the 47%. A survey of ten bartenders shows the same: eight of ten chose gins of 43% or more, with a median around 46%. That is why this clone is 44% and not 40%: in the Negroni the gin is a third of the glass and still has to face the vermouth's sugar and the bitter's bitterness.

What the award-winners teach Brazil, Japan and Europe

GinBaseStrengthDeclared botanicalsWhat we take from it
EtsuHokkaido, JapanCane molasses43%Yuzu, juniper, liquorice, angelica, coriander, bitter orangeProof that cane wins gold. Macerates 22 h in 67% cane alcohol, distils once in copper, no cold filtering. IWSC 2024 gold, 96 points
Ki No BiKyotoRice kitniyot45.7%11 botanicals in 6 groups: juniper/orris/hinoki; yuzu/lemon; gyokuro; sansho/kinome; ginger; shiso/bambooDistilling each group apart and blending after. IWSC 2018 Trophy
BEGCampinas, SPGrain chametz44 and 54%Juniper, coriander, angelica, Surinam cherry leaf, elderflower, cardamom, cinnamon, lemongrass, limão-taiti, tangerine, laranja-limaSurinam cherry leaf as a Brazilian resin note. 18 h cold maceration. IWSC 2024, 98 points on the Navy
Hayman'sLondonWheat chametz41.2%All ten, with origin: juniper, coriander, lemon, orange, angelica, cinnamon, cassia, liquorice, nutmeg, orrisThe backbone of classic London Dry, and 24 h of infusion inside the still. "Best Gin" at San Francisco 2014
Gin MareCataloniaBarley chametz42.7%Arbequina olive, rosemary, basil, thyme, orange, lemon, coriander, juniper, cardamomSeparate distillation per botanical. But Gin Foundry warns it pulls the Negroni toward savoury: the herbs belong to the bitter
No. 209 KFPSan FranciscoCane, OU-P46%The classics, without cardamom: swapped for bayThe only documented kosher-for-Passover gin. The direct precedent for this recipe
Monkey 47Black ForestMolasses47%47, of which the producer reveals partMolasses in Germany's most-awarded gin. Rests 100 days in clay pots

Also looked at and left out: Amázzoni and Arapuru (declared grain base), Roku, Matsui Hakuto and Sakurao (undeclared base), Nikka Coffey (maize and barley), Vitória Régia and Virga (cane, but with awards I could only confirm through a secondary source). None of the twelve producers disclose proportions.

05

Component analysis

Doses per litre of finished gin. The maceration charge is smaller than the final litre: the calculator does the sums.

Beefeater's nine, one by one

ComponentDoseWhat happens to it hereActive principleBasis
JuniperJuniperus communis, dried berries10 gStays, and goes up. It is the legal requirement and what 21 of 24 critics name first. The literature uses 8 to 12 g per litre of charge; we go high, because the Negroni dilutes.α-pinene 27.7–49% of the oil [10]; threshold 1650 µg/L [5]Confirmed
Bitter orangeCitrus aurantium, dried peel3 gStays. It is the bridge to the bitter and to the vermouth: the same limonene in all three bottles is what makes the Negroni read as one thing.Limonene 90.6% [14]Confirmed
LemonCitrus limon or C. × limonia, dried peel2 gStays. Brazilian limão-cravo works: it is the same limonene with more neral.Limonene 55%, neral 10.4% [16]Confirmed
Angelica, rootAngelica archangelica2.5 gStays. It is the root that gives the earthy base and the musk that anchors the rest. See the furanocoumarin cap below.β-phellandrene 6.7–17.2%, α-pinene 7.8–17.2% [18]; musk lactones 1.3% [17]Confirmed
AlmondPrunus dulcis, sweet, flaked2 gStays. Five critics note marzipan. It gives body more than aroma. Sweet almond, never bitter.Oils; trace benzaldehydeConfirmed
Angelica, seed—Out for lack of a source. Angelica seed is not to be found in Brazil. The seed oil is 74.7% β-phellandrene [17], the same main compound as the root: I compensated by raising the root.β-phellandrene 74.7% [17]Substitution
Coriander, seedCoriandrum sativum—Out for kitniyot in the main recipe, and back in the section 06 version for those who follow the OU. It is this page's biggest loss, and the kashrut section explains why.Linalool 49–82% of the oil [11]; the highest OAV measured in gin [4]Confirmed
LiquoriceGlycyrrhiza glabra, root—Out twice over. It is a legume, so kitniyot under the most common custom; and the normalised panel found the "liquorice" note the same across every gin [2], including the ones that carry none.Glycyrrhizin 0.18–0.69% of the root; does not distil [23]Confirmed
OrrisIris germanica or pallida, root—Out for three reasons. Irones only appear after 3 to 5 years of curing the root [22]; on distillation cis-α-irone fell from ~45% to 24.7% [21]; and I found no evidence at all for the "fixative" effect attributed to it. There is no cured root for sale in Brazil.IronesConfirmed

What took their place chosen for the compound, not the description

ComponentDoseWhy it is hereActive principleBasis
Dried orange blossomCitrus aurantium, blossom — azahar2.5 gIn place of coriander. It is the only source of orange-tree linalool that is not kitniyot: the linalool is in the blossom, not in the peel — the peel is limonene.Linalool 28.5–34.4%, linalyl acetate 11.3–19.6%, nerolidol 9.1% [13][15]Substitution
LavenderLavandula angustifolia, culinary grade0.3 gIn place of coriander, second half. Completes the linalool and brings the linalyl acetate. Keep the dose low: above this it turns to soap.Linalool 11.4–57%, linalyl acetate 7.4–44% [12]Substitution
Black pepperPiper nigrum, whole peppercorns0.5 gReplacing the pepper. Seven critics note pepper in Beefeater, and it came from the coriander and the angelica. Without coriander, it needs to be made explicit. Not kitniyot: it is a berry, not a legume seed.Piperine; β-caryophyllene, limoneneInferred
BayLaurus nobilis, dried leaf0.3 gDirect precedent: the No. 209 Kosher for Passover dropped the cardamom and added bay. Cardamom is kitniyot under the OU list, and the bitter already brings it to the Negroni.1,8-cineole, linalool, α-terpinyl acetateSubstitution
Surinam cherry leafEugenia uniflora, dried0.8 gA Brazilian resin note. It is BEG's botanical, which took 98 points at IWSC 2024. It reinforces the juniper's pine by another route.Furanoid sesquiterpenes (curzerene, selina-dienes)Substitution
PripriocaCyperus articulatus, rhizome0.4 gIn place of orris — only for the earthy, woody part. The page does not pretend otherwise: priprioca has no irones and does not reproduce the violet.Mustakone 11.6–20.2%, cyclocolorenone 10.3%, α-pinene 8.3% [26]Substitution
Pink pepper, optionalSchinus terebinthifolius, fruit — pink peppercorn0.8 gReinforcing the juniper, not the coriander. See the allergy note below. It has 6.0–6.5% oil in the fruit, 3 to 15 times more than juniper by weight: that is why the dose is low.δ-3-carene 28–30%, limonene 17–22%, α-phellandrene 12.6–17% [24]Substitution
Star anise, optionalIllicium verum0.2 gIf you miss the anise note that 11 critics attribute to the liquorice. It is not fennel (Pimpinella), which is kitniyot; but I found no clear position on it.AnetholeInferred
Seven things the research overturned

1. Pink pepper has no linalool: it is a plant of pinene, carene and phellandrene, closer kin to juniper than to coriander. It does not substitute for coriander, contrary to what the impression suggests. 2. Orange peel is limonene; the orange tree's linalool is in the blossom [13]. 3. Linalool comes off late in the distillation, not early: it first drops and then rises fast at the end of the heart [5]; and it yields 1.7 times more in pot maceration than in the vapour basket [6]. 4. Liquorice does not sweeten a distilled gin [2][23]. 5. Orris's "fixative" role has no support [21]. 6. Cutting the heads does not control methanol: in the measured distillations, methanol is a heart compound and stays nearly constant across the middle fractions [27]. 7. Rosewood is out: an endangered species, and its essential oil has been on CITES Appendix II since 2010 [32].

Safety caps, not flavour caps

Pink pepper. In one study, 16 of 21 children allergic to cashew or pistachio reacted to pink pepper on a skin test [25]. It is in the same family as cashew. If anyone drinking it has that allergy, leave it out — and tell whoever gets the bottle.

Angelica. The root carries phototoxic furanocoumarins [19]. The human threshold sits around 15 mg of 8-MOP equivalent per person [20]. On the most loaded estimate, 2.5 g/L give about 0.02 mg per 45 ml dose — a margin of hundreds of times. Do not go over 4 g/L.

Basil. Left out: some cultivars run up to 82% estragole [29], and there is no way to know the chemotype of a market bunch.

Methanol. With a dried botanical in neutral alcohol there is no fermentation, so the methanol comes only from the base alcohol. Buy alcohol of declared origin; cutting the heads does not fix this [27].

06

Kashrut control

In the gin, the critical point is not the alcohol: it is the coriander.

Coriander is a decision, not a fact

Coriander seed is the second botanical in almost every gin, the largest source of linalool, and the highest-impact compound measured in gins [4]. And the agencies disagree: the cRc and Star-K list coriander as kitniyot; the OU says it is not, provided it is well checked, following the Rema. There is no single answer, and this page will not invent one.

This page's main recipe is without coriander, which is this family's standard. Anyone who follows the OU makes the swap: drop the orange blossom (2.5 g) and the lavender (0.3 g) and put in 6 g of coriander seed per litre in fraction 1. It comes closer to Beefeater, with a linalool that is easier to pin down — the lavender's varies from 11 to 57% from one batch to another [12]. Ask your rav first.

Base alcohol
Grain gin is chametz, without exception. And there is no neutral cane alcohol with hechsher lePessach in small packaging.
Use neutral cane alcohol bought as a group under supervision, Vodka Polak (BDK's 2021 Passover list; check the current year's), or your own cane distillate. See where to buy.
Coriander
Kitniyot for the cRc and for Star-K; not for the OU, with checking.
Out of the main recipe. The swap is in the note above, and the decision is your rav's.
Liquorice
It is a legume: kitniyot under the most common custom.
Out. And the science says it does not tell one gin from another [2].
Cardamom
Kitniyot under the OU list. Star-K does not list it.
Out, with bay in its place, as No. 209 KFP did. The Bitter Rosso already brings it to the Negroni.
Star anise
It is not fennel (Pimpinella, which is kitniyot), but I found no clear position.
Optional. Ask your rav, or leave it out.
Juniper, angelica, priprioca, Surinam cherry leaf
None carry a seal, and juniper is always imported.
Buy them whole, single-ingredient, and ask your rav whether unhechshered is acceptable.
Almond
A tree nut is not kitniyot, but the processing line may be shared, and flaked almond sometimes carries an anti-caking agent.
Use almond with hechsher, whole, and flake it yourself.
Black pepper
It is a berry, not kitniyot. But ground pepper is milled on a shared line.
Use whole, never ground. Companhia das Ervas was on the 2021 list.
Orange blossom and lavender
Sold as tea or as flavouring; the flavouring may carry a carrier.
Use whole dried blossom, single-ingredient. Never "essence" nor flower water with preservative.
Citrus peel
Industrial dried peel may be treated.
Dry it yourself, from washed fruit, only the coloured part.
Water
—
Use filtered or mineral water. Hard water leaves the gin cloudy on dilution.
Equipment
Glass, funnel and still that touched chametz.
Use new or kashered glass; on Passover, equipment dedicated to this alone.
Limit

Controlling ingredients is not certification. The final word belongs to your rav. The BDK list cited is from 2021: check the current year's.

07

Method

The main route is by maceration, because that is what you do without a still. Anyone with a small still finds the cuts at the end of this section.

A macerated gin is not a London Dry

And this page will not call it that. Maceration also extracts colour, tannin and bitterness, so the liquid comes out somewhere between straw and amber, not colourless [30]. It also clouds more on dilution than a distillate does [28]. In exchange, linalool yields 1.7 times more in maceration than in the vapour basket [6], which helps precisely the part we are reconstructing. It is a drink for personal use.

Beforeshopping

Assemble the charge

The calculator says how much alcohol and how much water to mix. The default: 458 ml of 96% cane alcohol plus 422 ml of water, which give 880 ml of charge at 50%, to yield 1 L of gin at 44%. The literature macerates between 45 and 60% [6].

Weigh the botanicals on a 0.01 g scale: the lavender and the bay are tenths of a gram. Break the juniper berries with the bottom of a glass, without grinding.

Day 07 days

Fraction 1 · juniper, roots and almond

Set aside two-thirds of the charge. Into it: 10 g of broken juniper, 2.5 g of chopped angelica root, 2 g of flaked almond, 0.4 g of priprioca and 0.8 g of Surinam cherry leaf. Dark, full jar, away from light; shake once a day.

The remaining third goes sealed to the fridge, for fractions 2 and 3.

Juniper's monoterpenes oxidise with heat, light and oxygen, and turn into a defective flavour [5]: amber glass, full to the neck, in a cupboard.

Day 548 hours

Fraction 2 · peels and spice

In half of the reserved third: 3 g of dried bitter-orange peel, 2 g of dried lemon peel, 0.5 g of whole black peppercorns, 0.3 g of bay and, if using, 0.8 g of pink pepper and a pinch of star anise. Forty-eight hours, not one more: after that the peel releases pith bitterness.

Day 612 hours

Fraction 3 · the flowers

In the other half: 2.5 g of dried orange blossom and 0.3 g of lavender. Only 12 hours, at room temperature. Over-macerated blossom takes on a stale-tea taste and loses what is best in it.

In the coriander version, this fraction does not exist: the 6 g of coriander seed go into fraction 1, on day 0.

Day 7filter

Filter and set aside the reserve

Strain the three fractions through a sieve and then through a paper filter, without pressing. Set aside 20% of fraction 1 in a jar: it is the reserve for adjusting the juniper by taste. Combine the rest and top up with water to the final volume (the default is 120 ml).

If less is left than the sums say, that is what the botanicals absorbed. Top up with water to the volume; the strength drops a little and that is fine.

Day 7taste

Adjust in the Negroni, not neat

Test Negroni: 20 ml of this gin, 20 ml of Bitter Rosso, 20 ml of Vermouth Rosso and 15 ml of chilled water, no ice. That is how you judge it, not in the glass alone.

If the juniper disappears, drip the reserve into 100 ml of gin, 5 drops at a time (2.5 ml per litre), and redo the test. In the batch, multiply the total by 10. Discard whatever is left of the reserve.

Day 7–92 days

Rest and bottle

Two days in the fridge bring down whatever is still going to precipitate. Filter through paper while still cold and bottle in dark glass, full to the neck. A macerated gin has no short shelf life, but the juniper loses its pine within a few months: make batches you will drink.

08

The neutral cane base

The hard question is not distilling the gin: it is arriving at a base that does not taste of cachaça. And the good news is that there is a number to aim at.

A congener, in cachaça, is not a defect: it is an identity requirement. Brazilian regulation requires a minimum. In gin and in vodka, it requires a maximum. Both are measured in the same unit, and it is the distance between them that defines the work.

DrinkCongener coefficientmg/100 ml of anhydrous alcoholWhere it is written
Cachaça and cane spiritminimum 200
maximum 650
MAPA Ordinance 539/2022 [34]. Also: esters up to 200, aldehydes up to 30, higher alcohols up to 360, furfural + HMF up to 5, copper up to 5 mg/L
Gin and vodkamaximum 50MAPA Normative Instruction 29/2012, art. 9 [35]. Methanol up to 20 mg/100 ml AA, copper up to 5 mg/L
Potable ethyl alcohol of agricultural originBrazilno limitDecree 12.709/2025, art. 86 [31]: requires only ≥95% vol and "distillo-rectification". No analytical limit
Ethyl alcohol of agricultural originEuropean Unionesters 1.3
aldehydes 0.5
higher alcohols 0.5
Reg. (EU) 2019/787, article 5 [33] — not Annex I. In g/hL of 100% alcohol, which is the same thing as mg/100 ml AA. Plus methanol 30 and furfural not detectable, at ≥96% vol
Brazilian regulation gives you no target

Brazil defines neutral alcohol only by strength: 95% and "distillo-rectification", with not a single analytical number. Anyone wanting a target has to use the European one. And there the maximum aldehyde is 0.5 against cachaça's 30: sixty times.

What "tasting of cachaça" exactly is

A 2026 Brazilian study measured 15 fractions of a cane distillation, calculated the odour threshold of each compound in 40% solution, and divided one by the other [36]. The result is the OAV: how many times the compound sits above the threshold at which it is perceived. An OAV below 1 means nobody notices.

CompoundAroma it givesIn the heartmg/100 ml AAOAVHow much it needs to fall
Ethyl acetateFruity, nail-polish remover, ethereal28.328.328×. It is the bottleneck: the first to give away the base
3-methyl-1-butanolisoamylMalty, peppery, fusel oil189.513.513.5×. It is the dominant congener by mass
2-methyl-1-propanolisobutanolMalty, alcoholic, chemical30.54.34.3×
1-propanolFermented, bread, musty23.42.32.3×
AcetaldehydeGreen, pungent, bruised apple6.11.31.3×: already barely perceptible
Acetic acidVinegar22.01.21.2×
Ethyl lactateFruity, green1.40.57nothing. Already below the threshold
MethanolAlcoholic0.80.15nothing
FurfuralCaramel, almond, woody0.360.03nothing. It is the "wood" axis, not the "cane" axis
Two targets, with very different difficulties

For the liquid to stop smelling of cachaça, a reduction on the order of 30× is enough — ethyl acetate is the bottleneck, isoamyl comes next at 13.5×. For it to be neutral alcohol in the European legal sense, the higher alcohols need to fall from 243 to 0.5, that is about 490×.

These are two problems of different orders, and for a gin what matters is the first one — all the more because redistilling with the botanicals adds one more separation stage. The factors here are our own calculation, worked from the data and thresholds in [36] and article 5 [33]. And the legal specification is far stricter than the human nose: the 0.5 mg/100 ml AA limit for higher alcohols sits about 28 times below the isoamyl threshold.

Why redistilling does not solve it the discouraging number, and it needs saying

The same cane wine distilled three ways, with the higher alcohols measured [36], in mg/100 ml AA:

SystemOutput strengthHigher alcoholsEthyl acetateCongenersCarbamateµg/L
Simple distillation45.0%325.339.9414.465.6
Three-body pot still58.7%243.428.3300.238.5
Double distillation69.1%224.523.8265.95.9
Double distillation knocks out 31% of the fusel. You need 93%

From 325 to 224 mg/100 ml AA is a drop of 31%. Reaching the sensory target (13.5× on isoamyl) needs a 93% cut. One more pass is not the answer, and the reason is physical: the per-stage separation factor sits very close to 1 in the range where a pot still operates.

The reason shows up in the fraction-by-fraction measurement. In a pot-still cane distillation at 60–67%, acetaldehyde and ethyl acetate collapse and vanish in 35 to 45 minutes. But isoamyl starts at 282.7 and, after 75 minutes, with the still already at 50%, is still at 111.3 — 39% of the starting value [36]. No cut removes fusel from a still operating above 40%. The way out is different, and it is in the rectification subsection.

As a bonus, note the last column: double distillation took ethyl carbamate from 65.6 to 5.9 µg/L. Rectifying solves that pass-through problem.

Methanol: the congener you do not need to fight

Here the answer is reassuring, and it needs saying with numbers, because Brazil went through a methanol crisis in October 2025 [62] and the confusion between fermenting and adulterating hung in the air.

Cane is low in pectin by biology. Grasses have type-II cell walls, in which pectin is only 2 to 10% of the wall [64], against 7 to 10% pectin in citrus. And it is pectin, hydrolysed by pectin methylesterase, that releases methanol. What is more: cachaça yeast (S. cerevisiae) does not produce methanol; the ones that do are contaminant yeasts with pectinolytic activity.

What gets measured. In a survey of 268 cane spirits and cachaças, only 2.6% went over the limit, and methanol was the least-violated parameter of all [55]. In 12 aged cachaças from Minas, methanol was not detected in any of them — the contaminant that blew past the limit was furfural, in 83% of samples [54]. In the 2026 measured heart: 0.8 mg/100 ml AA [36].

Methanol limitmg/100 ml AAComparison
London gin, EU, cat. 22 [33]5The strictest in the regulation
Vodka, EU, cat. 15 [33]10Stricter than neutral alcohol itself
Ethyl alcohol of agricultural origin, EU, art. 530—
Fruit spirit, EU1000 to 135033 to 45 times that of neutral alcohol
Cachaça heart, measured [36]0.837× below the neutral-alcohol limit and 6× below that of London gin
A well-made cachaça already meets a neutral alcohol's methanol specification

With a margin of about 37 times, and London gin's with a margin of 6 (our calculation over [36] and [33]). Methanol simply is not your problem. Esters and fusel are.

And the heads cut does not control methanol — now with six sources

Methanol's relative volatility to ethanol sits between 0.7 and 1.5, as measured [41]. Close to 1 means separation is impossible: it comes out in every fraction and still remains in the residue. In print: "this confirms that separating the heads fraction does not remove methanol, but aldehydes and ethyl acetate" [40].

Worse for the intuition: methanol is a heart compound. In mezcal, the tails carry 248 mg/100 ml AA against 135 in the heads. In pisco, across 188 industrial samples, the body has more methanol than the heads (28.0 against 22.9). And because ethanol falls faster than methanol through the run, expressed per anhydrous alcohol — which is how the law measures it — methanol rises as you run further. Anyone who needs to lower methanol cuts the tails earlier, not the heads.

Discarding the heads is still right. Just not because of methanol: it is for the acetaldehyde and the ethyl acetate, which really do collapse there.

Fermenting is not adulterating: three orders of magnitude

A cachaça at the legal cap carries 8 mg of methanol per 100 ml of drink, alongside 31.6 g of ethanol — a ratio of about 3,950 to 1 in favour of the very antidote, because ethanol is a competitive inhibitor of the enzyme that turns methanol into formate (our calculation). Counterfeit samples seized by police in São Paulo carried 14 g per 100 ml [63]: 1,750 times the legal cap. No cane fermentation crosses that distance. Only the deliberate addition of industrial methanol does — and that, not home fermentation, is what caused the 2025 health emergency [62].

Three routes to the base

RouteEffortWhen it makes sense
Buy neutral alcoholENA, 95–96%NoneThe right answer for almost everyone. Industry does not make neutral with one column and one cut: it uses five to seven columns, among them a hydroselection column and a dedicated demethylizer [42]. No home column comes close. See where to buy
Rectify a cane distillate you already haveMediumIf you have young white cachaça or your own distillate, and a column with reflux. It is the route the calculator below sizes
Ferment and rectify from scratchHighOnly if you already make cachaça. The advantage is fermenting with neutral in mind, which is the opposite of fermenting for cachaça — and that changes the result more than anything you do afterwards, because the fusel comes from the yeast, not from the distillation

Fermenting with neutral in mind

Fusel is not born in the distillation: it is born in the yeast. Because the subject has its own numbers — and because switching strain changes the congeners by almost three times, for free — it earned a whole section: fermenting the wash, with cane juice and molasses side by side.

Rectifying the only part that solves the fusel

Dilute the charge. This is the step that changes everything, and it has an exact explanation. In a miscible mixture the components do not each boil at their own boiling point: where each compound lands in the distillate depends on the ethanol concentration, not on the pure boiling point.

The number: isoamyl's relative volatility to ethanol is greater than 1 while ethanol stays below about 40% v/v, and falls below 1 above that [39]. In plain terms:

  • Dilute charge, up to ~30%: the fusel is more volatile than the ethanol, it comes out in the heads and can be cut.
  • Concentrated charge, 60–70% — the case of a cachaça redistilled without dilution: the fusel does not come out in the heads; it spreads through the whole heart. This is exactly what the fraction-by-fraction measurement shows.

It is the same principle behind the industrial hydroselection column, which exists precisely to dilute the ethanol with water and so drag along esters, acetaldehyde and half the higher alcohols [42]. An older literature compilation puts the crossover at 32% rather than 40%; either way, 25 to 30% is safe. The technical literature reaches the same number by another route: above 30% heads and tails bleed into the heart, below 25% there are not enough congeners to cut with precision [66].

And one operating detail worth gold: higher heat favours the less-volatile compounds, because stronger heating makes them distil earlier and at higher concentration in the first fractions [44]. Distilling slowly is part of the separation, not a fussy habit.

Where the fusel hides in a column
It sits trapped in the band where ethanol is between 40 and 47% by mass, with peak concentration around 20% by mass [39]. The band is defined by composition, not by height. Industry draws it off through a side draw at exactly that plate, purging 2.0 to 6.5% of the fed ethanol. In a batch pot still you have no such draw — the fusel circulates and returns to the product. That is the structural difference between your column and the plant's.
How many plates, or how much packing
Technical literature: 20 plates or more is the threshold for vodka and neutrals [66]; a cachaça column has 15 to 20 plates and delivers 65–70% at the start [43]; the physical limit is the azeotrope, 97.2%.
Community: 8 plates make a gin base but not a neutral; 10 get there with a lot of reflux; 16 is the practical minimum. In a 2" packed column, 1200–1500 mm of stainless or copper mesh gives 94–95%, with a height-to-diameter ratio of 12:1 to 24:1 and little gain above 20:1.
Reflux
Literature: the concentration of higher alcohols in the distillate rises with low distillate flow and low reflux ratio [39] — that is, high reflux lowers them. Columns perform better at low flows and larger temperature gradients [43].
Community: 30 to 45 minutes at total reflux, with the take-off closed, before starting. I found no such figure in peer-reviewed literature — it is a convergence of practice.
Take-off rate
Community: 0.6 to 1.2 L/h in a 2" column with 1600–2000 W; the physical ceiling for that column at the azeotrope is ~1.5 L/h. Manage the rate, not the temperature — that is the diagnosis practice gives for almost every neutral that came out badly. High flow causes entrainment and flooding, and both can happen at once [43].
The cuts
Cut by nose, not by volume. For neutral the cuts are much tighter than for cachaça — and I found no study that publishes the optimal heads/hearts/tails split for batch neutrality: it is a genuine gap in the literature. Keep heads and tails and add them to the next charge: the ethanol comes back, and the cut's loss stops being a loss.
How many passes
There is no published number. The distilleries that disclose it speak of three to four. Be wary of a "five distillations" label claim: one American distillery explains that this figure, in marketing, usually refers to a column's number of plates, not to passes.
The golden rule, and it comes from the community

Never run the neutral together with the botanicals basket. These are processes with opposite aims: one separates, the other extracts. And the base's approval test is this: acceptable to the nose and the palate both at 40% and diluted to 10%. At 10% the ethanol stops suppressing the aroma and whatever got through shows up.

Activated carbon the right tool for the right problem

Compound or classAdsorptionMeasured numbers
Estersethyl acetateVery good−53.5% at 12 g/L and −81.3% at 26 g/L, in 10 minutes. A rate of 1.47 mg of ester per gram of carbon [45]
Higher alcoholsGood−25% to −44%. Adsorbs in the order isoamyl > isobutanol > propanol: it favours the longer, less polar chain [45][46]
Volatile acidityGood−28% at 12 g/L for 1 h; −75% at 26 g/L for 24 h [45]
FurfuralGood in a bed−77% to −92% in a single column pass [47]; but only 5–13% in dilute batch. Geometry matters
AcetaldehydeVery poor−12.5% with 26 g/L for 24 hours. And +15.9% with 12 g/L for 1 h, and +15.1% in another trial: the value rises [45][46]
MethanolVery poor"The lowest adsorption of all the compounds tested" [48]. In an industrial vodka column: 10.5 → 10.0 mg/L, within the margin of error [49]
EthanolNegligible~60 mg of ethanol per gram of carbon, or ~0.5% at 26 g/L. Measured: 38.0 °GL before and after, with no dose or time effect [45]
Colour, wood aroma and flavourDestroyedColour −80% to −88%, wood aroma −66% to −84%, flavour −64% to −85%, in a single pass through a commercial filter [47]
CopperPoor12 g/L for 1 h took 9.0 down to only 5.34 mg/L, still above the legal limit [46]
Why this is good news

Notice the coincidence: the compound that most gives away the base is ethyl acetate (28×), and it is exactly the one carbon adsorbs best (−81% in a single pass at 26 g/L). Two passes at that dose would give about 28× (our arithmetic over [45]; not a published result). Carbon is poor on acetaldehyde and on methanol — but both are already at OAV 1.3 and 0.15, nearly inaudible. Carbon attacks the bottleneck and is weak exactly where it does not matter.

The studies that cap carbon at 1.5 g/L for up to 1 hour are protecting the aroma of a distillate that is meant to have aroma. In a neutral base, erasing the aroma is the goal: there the high dose is justified.

Four things that decide whether carbon works

1 · Treat at high strength, never diluted. At 40% v/v the reduction in acetaldehyde and methanol is statistically null; at 70–80% v/v it exists [48], because at 40% the water wins the competition for polar sites. Filtering carbon at drinking strength is nearly useless.

2 · Dose matters, time barely does. The esters come out in 10 minutes. The purity plateau sits at 10 g/L for the volatiles in general [48]. And carbon saturates: in a column, removal falls from 96% on the first pass to 54% on the third and 13% on the fifth [50].

3 · Cold, and without soaking. The industrial vodka column works at 18–20 °C, because above that the carbon releases aldehydes. Three days of contact made the acetaldehyde rise from <0.5 to 1.0 mg/L [49].

4 · Never on an aged distillate. One pass erased the ageing of a three-year cachaça in oak [47]. If you made Cana & Carvalho, do not run carbon through it.

Copper: what it solves and what it creates

Copper knocks down the cabbage smell, and that is real. In the controlled experiment — the same cane wine in four identical columns, changing only the packing [37] — dimethyl sulphide in the copper distillate was 27% of the stainless-steel value, and tasters detected sulphide odour in the stainless-steel and porcelain samples, not the copper ones. Aluminium gave the lowest DMS of all and was rejected on flavour: it is not an alternative. The mechanism is copper oxidising DMS to sulphate, with methanol as a by-product — which is why copper distillates carry more methanol and more sulphate. Another study shows that copper is more effective in some parts of the still than in others.

A unit caveat: the paper publishes DMS in mg/L, and 592 mg/L is implausible — the DMS sensory threshold is in the tens of µg/L, and a paper from the same group gives the defect threshold at 4.3 to 5.2 mg/L. There is likely a unit error. The ratios between materials hold either way, and it is on them that the decision depends.

But copper creates two problems, with different solutions.

Copper in the liquid
Between distillations verdigris forms on the walls, and the acidic alcohol vapour dissolves and carries it along. Across 94 commercial cachaças: median 1.72 and maximum 12.25 mg/L, against the limit of 5 [38]; other surveys find 7% to 74% of samples out of spec. The most effective measure is washing the still — a prior run with water, or water with lemon, whose vapour carries off the verdigris. For what remains, cationic resin is much better than carbon: 9.0 → 0.07 mg/L in one pass, taking very little of the organics with it [46].
Ethyl carbamate
Dissolved Cu(II) oxidises cyanide to cyanate, and the cyanate complexed to copper is attacked by ethanol [52]. Just 0.7–0.8 mg/L of copper is enough to saturate the catalysis; to minimise carbamate the distillate needs to stay around 0.15 mg/L [43]. Columns with no copper part at all give carbamate below 10 µg/L, and a still with three stainless-steel plates and reflux control at the dephlegmator gave not detected [43].
The contradiction, resolved
Some studies show copper reducing carbamate, others causing it. Both are true, and the difference is the copper's state. Solid metallic copper in the vapour path sequesters hydrocyanic acid: with copper turnings, cyanide migration into the distillate fell from 88.3% to zero, and transfer of carbamate already formed fell from 73% to 9.6% [53]. Dissolved ions in the distillate catalyse its formation.
The arrangement the literature recommends
Plenty of clean metallic copper in the ascending part, stainless steel in the descending part and at condensation [51], or a packed section with copper rings outside the main body. In a survey of 33 brands, the distilleries below 40 µg/L of carbamate did exactly that. Also: high reflux rates go together with low carbamate [43]. A legal detail: a still built this way loses the right to the name "alambique cachaça", which requires copper throughout [34] — irrelevant here, because what we want is neutral.

How to know you have arrived

The two-strength test
Taste the base at 40% and diluted to 10%. If you notice nail polish, cane sweetness, cabbage or solvent, it is not yet neutral. The nail-polish note is ethyl acetate, the first to give it away.
The blind test against vodka
Diluted to the same strength, your neutral against a reference vodka, in matching glasses, poured by someone else. It is the only controlled sensory test you can run at home.
What you cannot verify
None of this section's limits can be verified without chromatography. Esters, aldehydes, higher alcohols and methanol call for GC-FID; carbamate calls for HPLC with derivatisation. An alcoholmeter measures strength, and nothing else. If you need a number, a food laboratory can run the congener coefficient by the official method.
Copper, that one you can
Colorimetric copper test kits for water exist, and labs run it by atomic absorption. It is the only parameter on this list with an accessible test.

Rectification calculator from cane distillate to gin

Water to dilute
333ml
96% neutral
292ml
Gin at 44%
636ml
Total charge
1333ml
Pure alcohol
400ml
Where the recovery figure comes from

Supplier literature gives 85–92% ethanol recovery in one pot-still pass and 90–96% in a reflux column, warning that "neutrals use tighter cuts". The 70% default here is our own estimate for two passes with neutral-style cuts, without recirculating heads and tails. If you keep the heads and tails and add them to the next charge, raise it to 82%. I found no published figure for rectification loss for neutral.

09

Fermenting the wash

If the route is from scratch. Fusel is not born in the column: it is born here — and the yeast choice weighs more than everything you do afterwards.

The number that decides the recipe

A Brazilian study compared distillery yeast strains in the same wash, under the same conditions, and measured the distillate's congeners [69], in mg/100 ml of anhydrous alcohol:

StrainHigher alcoholsin the heartTotal congenersWhat this means
BG-1146.7225.7The cleanest of those tested. The choice for neutral.
CA-11—< 400with BG-1Isolated at a Catanduva plant by UFLA, it is the most used in Brazilian cachaça. The study pointed to it as the one with the best chemical and sensory composition. Fast start, near-complete sucrose hydrolysis, few acids and aldehydes
PE-2 and SA-1above 360—The strains plants use to open the crushing season, chosen for resisting stress and dominating contaminants — not for profile. They blow past the legal limit of 360 for higher alcohols
Y-904616.0654.6The heaviest. Almost three times BG-1, in the same wash, and also outside the legal limit
Read the first and last row together

Switching strain changes the congeners by almost 3×, for free, before you even light the fire. Nothing you do at the column gives that gain for that effort. If the route is from scratch and the target is neutral, start here: BG-1 or CA-11. Of these, CA-11 is the easy one to find in Brazil, because it is cachaça's own strain.

Abroad, manufacturers sell strains selected by substrate: DistilaMax SR for molasses, RM for cane juice, and HT, which the manufacturer describes as low in production of higher alcohols, aldehydes and esters [81]. EC-1118, a sparkling-wine strain, tolerates ~18% alcohol, ferments from 10 to 30 °C and produces little foam, little volatile acidity and little H₂S. This is manufacturer information, not a comparative study: the only measured comparison between strains I found is the one in the table above.

Cane juice

ParameterValueNote
°Brix of the wash14 to 16The cachaça range [76]. Sources spread from 12 to 20, and do not converge on one number. Dilution: Va = Vc × (Bi − Bf) / Bf
Nitrogen~582 ppmnaturalThe juice already comes nitrogen-rich: 582 ppm of total nitrogen, against a recommendation for the wash of 500 to 600 ppm [75]. In other words, with juice you barely need to supplement — a real advantage over molasses
Ammonium sulphate0.10 to 0.11 g/LThe dose that maximised fermentation efficiency at 15 °Brix [76]. Only if the juice is poor
Temperature29 to 32 °CRecommended in a study that tested 26 to 36 °C. At 36 °C the final strength was highest, and the authors rejected that temperature for sitting outside the literature range. For neutral, stay at the low end: less heat, less fusel
pH4.5 to 5.0Range favourable to the yeast [76]
Duration14 to 36 hIn simple batch. With yeast recycling, it falls to 6 to 10 h [80] — what governs it is the starting cell density, not the substrate
Final strength7 to 12%—
Juice sours in hours, not days

In a trial at 27 °C, 93% of the sucrose loss in the first 14 hours was microbial in origin — against 5.7% enzymatic and 1.3% chemical [70]. Raw juice starts fermenting on its own, with whatever is in the air and on the equipment, within a few hours.

Plan before you mill: have the fermenter clean, the yeast hydrated and everything weighed before the cane goes into the mill. I found no published figure for "inoculate within X hours" — this looks like shop-floor practice nobody wrote down. What the measurement shows is the direction: the faster, the better.

Molasses

ParameterValueNote
°Brix of the wash18 to 25Sources give 18, 21 and 25. For neutral, stay at the low end: a more concentrated wash stresses the yeast and generates more fusel
Nitrogen0.17 g/LMolasses is low in nitrogen, contrary to what its look suggests: molasses diluted to 3% gave 0.17 g/L against 3.07 g/L for a synthetic control medium. Supplementing is not optional here
Calciumbelow
0.18% w/w
Calcium is 10 to 20% of the ash in molasses [71]. At 0.18% the yield already dips a little; at 0.72% it falls 14 to 25%; at 2.16% sucrose fermentation nearly stops. Severity runs Ca²⁺ > K⁺ > Mg²⁺ [72]
Treating the molassessulphuric acid
0.3%
Reduces calcium and heavy metals (Cd, Cu, Fe, Ni, Pb, Zn) and improves fermentation [73]. Settle it and leave the ash at the bottom before inoculating
Temperature and pH35 °C and pH 4.5The yield optimum in one study, at 81% [77]. But for neutral, yield is not the only target: at 20 °C, with a tolerant strain, another study got over 90% yield and 92% viability after three cycles. Cold fermentation is workable, and it is what you want
Final strength8 to 12%—

The surprise: molasses gives less fusel than juice

This runs against intuition, and it is the comparison's strongest finding. One study analysed 25 runs of cane juice against 26 of molasses by GC×GC-TOF-MS and separated the two groups with 100% accuracy [78]. There were twelve marker compounds — and all twelve were higher in the juice distillates, among them 2-methylbutanol and 3-methylbutanol, which are exactly the fusel.

In other words: more processing does not mean more congener in the distillate. If the goal is neutral, molasses starts ahead. Juice has the advantage of its own nitrogen and of needing no calcium correction; molasses has the advantage of starting with less that you will have to remove later. Choose by what you have on hand, knowing both costs.

Nitrogen is the cheapest lever after the strain

The relationship is measured and strong: 200 mg/L of ammonium chloride knocked total higher alcohols down by 39% — isobutanol −47%, isoamyl −28%, phenylethanol −44% [74]. The work is in huangjiu, not cane, but the mechanism is the same: nitrogen-starved yeast synthesises amino acids, and the Ehrlich pathway, which does this, is the pathway that produces the fusel.

In high-density musts, urea plus ammonium sulphate at 300 + 75 mg/L acted in synergy, and doubling to 600 + 150 improved things further still, reducing unwanted by-products. That data point is from grape must.

And here kashrut fights the technique

The complex nutrients the literature recommends carry autolysed yeast hull, yeast extract and thiamine — and the best-known line's own manufacturer certificate says, in print, "strictly KOSHER & PARVE (excluding for Passover)". Kosher year-round, not for Passover.

The way out is the simple mineral nutrient: DAP and ammonium sulphate, which are synthetic salts. Technically it is less complete — it lacks the survival factor a yeast derivative gives. You trade performance for kashrut, on purpose, and this page will not pretend otherwise.

pH: the sources contradict each other, and we know why

The working range is 4.5 to 5.0 [76]. pH's effect on fusel, though, has no single direction, and now it is possible to explain the contradiction rather than just record it:

  • Within the same study, both directions. In molasses at 25 °Brix, pH 3.5 gave the highest isoamyl/isobutanol ratio (2.26); pH 5.0, with supplementation and no cooling, gave the highest absolute concentrations of isoamyl (0.437 g/L) and isobutanol (0.267 g/L) [68]. It depends which number you look at and what else is combined with it.
  • In the mechanism, genes running opposite ways. In the Ehrlich pathway, ARO8 and ARO9 rise in alkaline medium and ARO10 falls [79]. There is no "high pH raises fusel": it depends which enzyme is limiting the reaction at that moment. (The work is in Candida albicans, not Saccharomyces.)

In practice, then: stay at 4.5–5.0 because that is what the yeast likes, and do not try to fix the fusel through pH — fix it through the strain, the nitrogen and the temperature, where the direction is known.

On the acid treatment of the yeast starter, which the Brazilian industry does: the recovered yeast cream goes to pH 2.5–3.0 for 2 to 3 hours with sulphuric acid, as a bactericide, and returns to the fermentation range by dilution in the fresh wash — not by adding more acid.

Kashrut control of the fermentation

The yeast is this recipe's blind spot

The problem is not the yeast: it is the medium it was propagated in. Star-K says that "laboratory yeasts require approval for Pessach, to ensure there was no chametz in the cultivation". The OU goes further and warns not to trust a manufacturer's claim that it "uses only molasses", because the earlier propagation stages may have carried grain — and concludes that ordinary baker's or brewer's yeast should not be consumed on Passover [82].

And distilling does not solve it. R. Zushe Yosef Blech's article, published by the BDK, mentions the distinction of zeia bealma — the distillate as mere condensed vapour — but makes clear that it is not used to permit consumption, only for selling it before the festival [86].

Yeast
Needs hechsher year-round for the cultivation medium, and a specific hechsher lePessach for Passover. I found no yeast with a lePessach seal for sale in Brazil: the "yeasts" category does not even appear on the BDK's Passover list.
Option with a real certificate: Laffort's Zymaflore line has a certificate valid until 31/05/2027, "including for Passover" [83] — but issued by a Modiin Illit body that is none of the major agencies, and I have not confirmed it is sold in Brazil. Or: the spontaneous-fermentation route, below.
The yeast starter, without grain
The traditional country starter carries corn meal, rice bran or soy bran (kitniyot) and sometimes wheat bran (chametz) [84].
Use a yeast starter of only cane juice and wild yeast, acidified with lemon or orange: it is a documented method [85] and carries no grain at all. The cost: you lose control of the strain, and the strain is the biggest lever on fusel. Cassava is neither kitniyot nor chametz, if you want a solid medium.
Nutrient
Nutrients with a yeast derivative are certified "excluding for Passover" by the manufacturer itself.
Use DAP or ammonium sulphate, minerals, with a certificate of reliable origin. Accept the lower performance.
Molasses
Plain, unflavoured molasses needs no certification outside Passover (Star-K). For Passover I found no declaration from any agency saying it is automatic.
Prefer molasses with a lePessach certificate. Note: brown sugar and rapadura need hechsher lePessach, because they are sugar with molasses applied back.
Fermenter
If it has already fermented a wash with grain, there is real disagreement between agencies.
Star-K: plastic cannot be kashered, full stop. OU: can be kashered where there is need. Stainless steel: hagalah. Your rav's decision; changing the vessel is the safer path.
Sulphuric acid for the yeast starter
It is an industrial reagent, and not every grade is food-grade.
Use food-grade, with a certificate of analysis. Or skip the acid treatment and change the yeast every batch.

Yield, and what it does not tell you

About 80 litres of ethanol per tonne of cane is the average Brazilian sources give, with a range of 70 to 90. By that arithmetic, some 12.5 kg of cane per litre of 96% alcohol — our own calculation on that figure, and valid only at industrial efficiency.

Cachaça's rule of thumb, (°Brix − 2) = litres of distillate per 100 L of wine, is calibrated for pot-still distillate at ~54 °GL, not for 96% neutral. Do not use it to size a gin batch: use the rectification calculator from the previous section, which starts from what you already have distilled.

What I did not find: litres of ethanol per tonne of molasses in a reliable Brazilian source; a study linking starting °Brix to final mg/L of fusel in cane; and a measured comparison between baker's yeast, wine yeast and distillery yeast under the same conditions — what circulates about "baker's yeast gives more fusel" is community practice, with no controlled study that I found.

10

Distilling the gin

With the base ready, this part is short and well documented. What is long is the list of risks, and it comes with numbers.

Now this really is what Beefeater declares: reduce the neutral to 50% with water, put the botanicals inside the still and macerate 24 hours before distilling — pot maceration, not the vapour basket. A safety rule from the technical literature: put the water into the still before the alcohol [66], so as not to displace alcohol vapour. And heat slowly: a high flame multiplies by about ten the monoterpenes that generate pungency [9], besides pushing the less-volatile compounds into the first fractions [44].

The charge range is 40–60%, with 8 to 25 g of botanicals per litre and juniper and coriander together making up 70 to 80% of the mass [66]. Botanicals with water-soluble compounds, like the flowers, extract better between 30 and 45%: an elegant alternative is macerating the orange blossom in water alone and using that tea to lower the charge's strength.

The cuts: discard about 5% of the charge volume as heads and collect the heart down to about 55% [7]. For the Negroni it is worth stretching the heart into the peppery, woody zone, where the linalool, the β-caryophyllene and angelica's musk lactones come out [5][17]. But stop before the end: the run's profile finishes in green tea and then used tea [67] — that is the taste of having run too deep.

Afterwards: cut with deionised or distilled water, because hard water flocculates from the calcium and the magnesium. Rest for a few weeks. Cloudiness that appears during dilution and clears in 15 to 20 minutes is normal; cloudiness that stays means the cuts were too wide.

Fire and pressure: the numbers

The real risk is physical, not toxicological. And the numbers are uncomfortable:

Flash point. Measured in a closed vessel, a liquid at 40% ABV has a flash point of 26–28 °C; at 70%, 20–21.5 °C; alcohol at 96%, 13.5 °C [56]. Confirmed with real drinks: rum at 40% gave 28 °C [57]. In other words, a strong distillate at room temperature is already giving off flammable vapour. And there is no strength below which the mixture stops being combustible: at 5% vol the flash point still exists, at 68 °C.

The vapour sinks. Ethanol vapour's relative density is 1.6 [58]: it flows along the floor, toward the burner, and pools under benches and in cellars, far from where it leaked. Flammability range: 3.3% to 19% in air. NFPA 30 counts ventilation as adequate when it keeps the space below a quarter of the lower limit, that is 8,250 ppm [59].

In a closed vessel it is worse than the flash point suggests. The lower explosion point sits up to 30 °C below the flash point in mixtures of differing volatilities [56].

Pressure. In the only distillery accident with an engineering report that I found, a 1,135 L still had a factory test pressure of 3 psi, welds that cracked at 5 psi, a relief valve set at 150 psi — rated only for water heaters — and no pressure gauge at all. The probable cause found was the vapour path blocked by glass marbles used as column packing. The pot was thrown 15 to 29 metres; one worker died [61]. A margin of 30 times in favour of rupture.

What to do, from a labour body [60]: never leave the still unattended; ventilate for real; charge the pot below 40%; keep the receiver at the lowest level possible and inside a larger, non-flammable container that holds any overflow. I would add what the literature above demands: a pressure gauge and a relief valve sized to the vessel's real pressure, indirect heating instead of a flame under the pot, and no marbles or packing that could clog the column.

A correction worth making

It gets repeated that home distilling is banned "because it explodes". That is not the regulator's argument. The American agency that bans it does so for tax and criminal reasons, and its own page on the subject carries no fire warning at all. The risks above are real — but they are the reason to do it properly, not the legal basis for the ban.

Brazilian law, article by article

Registration is mandatory: Law 8.918/1994, art. 1, and Decree 12.709/2025, art. 114 [65], require registering the establishment and the product with MAPA.

An exemption exists for personal consumption: MAPA Normative Instruction 72/2018, art. 32, III, says that instruction "does not apply to production intended for personal consumption, with no commercial purpose" [65] — and it is not a beer-only exemption: it applies to the whole IN, which covers drinks in general. An honest caveat: this IN was issued under Decree 6.871/2009, now repealed, and I found no act saying it survives the repeal. This is a genuinely uncertain point, and the exemption is from registration, not a general licence.

Selling is another matter: operating without registration is a most-serious offence and selling an unregistered product is a serious offence, carrying a fine of R$ 100 to R$ 150 thousand, suspension or revocation, and mandatory referral to the Public Prosecutor's Office. On top of that, the Federal Revenue's special registration, which exists only for a legal entity with industrial facilities and prior MAPA registration. A useful curiosity: under that regime bulk ethyl alcohol falls outside the control seal; gin, vodka and cachaça fall inside it.

The still itself: I found no registration or sealing of the still as such in any of the federal rules checked — Law 8.918/1994, Decree 12.709/2025, MAPA Normative Instruction 72/2018, the RIPI and the special-registration instruction. What exists is inspection of equipment and facilities inside a registered establishment. We did not research state, municipal, health or environmental regulation.

11

Batch calculator

Enter your alcohol's strength and the strength you want at the end. It works out the maceration charge and scales every botanical.

Alcohol
458ml
Water in the charge
422ml
Charge
880ml
Water at the end
120ml
Juniper
10g
BotanicalDoseFractionTime
Why the charge is smaller than the litre

All of the batch's alcohol goes into the maceration charge; it is the water that splits between the charge and the final adjustment. Macerating in a stronger liquid extracts more from the botanicals [6], which is why it is diluted afterwards, not before.

With alcohol at 36% you cannot reach 44%: the calculator warns you and shows the maximum strength possible. In that case, raise the botanicals by about 20% and put less water in the Negroni.

12

In the Negroni

Three bottles, and each brings what the other two do not have.

PartnerWhat it hasThe gin reinforcesThe gin avoids
Bitter Rosso de CanaOrange, wormwood, gentian, cardamom, rosemary, marjoram, thyme; 250 g/LBitter orange: the same limonene across all three bottlesLeaf herbs and cardamom, which belong to the bitter
Vermouth Rosso de CanaCacao, orange, cinnamon, nutmeg, rosemary, caramel; 180 g/LRosemary and juniper share the α-pineneCinnamon, cacao and vanilla, which belong to the vermouth
The glassIce, orange peelLimonene and octanal from the peel twisted on top—
Negroni

30 ml of gin, 30 ml of Bitter Rosso and 30 ml of Vermouth Rosso. Stir with ice for 20 to 30 seconds, strain over a large ice cube and twist an orange peel over the top. Dilution is not loss: it reduces the suppression of aroma by ethanol, and the volatiles become more perceptible [28].

Martini and gin and tonic

Martini: 60 ml of gin and 10 ml of dry vermouth, stirred, with a lemon twist. Gin and tonic: 50 ml with chilled tonic and an orange peel — the peel's limonene links the gin to the quinine.

13

Troubleshooting

By what you notice.

Vanishes in the Negroni
Almost always it is strength, not recipe: that is what happened to Beefeater at 40%. Check the final strength; if it is below 43%, redo the batch stronger before touching the botanicals.
Little juniper, too much blossom
Drip in the fraction 1 reserve. Next batch, 12 g of juniper and 2 g of orange blossom.
Soapy or perfumed taste
Too much lavender, or lavandin instead of angustifolia. It does not come out. Next batch, 0.15 g, and check the species on the label.
Bitterness that dries the mouth
Citrus peel with pith, or fraction 2 left past 48 h, or fractions pressed at the filter. Prevention only.
Stale-tea taste
Blossom macerated past 12 hours. Prevention only.
Colour too dark
Normal in a macerated gin, and the reason it is not a London Dry. If it bothers you, cut the priprioca and the Surinam cherry leaf by half: they stain the most.
Clouds on chilling or diluting
Citrus and juniper oils coming out of solution. Expected below 45%. One night in the fridge, then filter through paper while still cold.
Flat, no pine
Old juniper. The monoterpenes oxidise with light, heat and air; a berry that does not smell strong when broken has already lost what matters.
Solvent smell
On the distilled route, not enough heads cut. On the macerated one, it does not happen: it only comes from the base alcohol, and then the problem is what you bought.
Pulls toward savoury
Too many leaf herbs. This was the criticism made of Gin Mare in the Negroni. Rosemary, thyme and marjoram belong to the bitter; the gin does not carry them.