Can chlorine in water taint whisky? Chlorophenols, TCA and the parts-per-trillion problem

Key takeaways

• Chlorine plus phenol makes chlorophenol. Chlorophenols “are present in drinking-water as a result of the chlorination of phenols, as by-products of the reaction of hypochlorite with phenolic acids” (WHO, Guidelines for Drinking-water Quality).
• The taste thresholds are parts per billion. “The taste thresholds in water for 2-chlorophenol, 2,4-dichlorophenol and 2,4,6-trichlorophenol are 0.1, 0.3 and 2 µg/l, respectively,” and microorganisms in distribution systems “may sometimes methylate chlorophenols to produce chlorinated anisoles, for which the odour threshold is considerably lower” (WHO GDWQ, ch. 10).
• In spirit the numbers are nanograms. SWRI work with 217 panellists found “an average detection threshold of 4.40 ng/L TCA in 20% ABV neutral ethanol,” and in whisky “detection thresholds of 2 to 33 ng/L depending on the whisky matrix tested” (Baxter, Réveillé & Conner, J. Inst. Brew. 2024).
• Your lab may not see it. For routine SPME-GC-MS, “instrumental limits of detection are ca.10 ng/L” — above what a good nose can find (Baxter et al. 2024).
• RO does not remove chlorine; carbon does. Membranes must be protected from it: “eventual degradation may occur after approximately 200 – 1,000 hours of exposure to 1 mg/L of free chlorine” (DuPont FilmTec™).

Yes — and at concentrations far below anything you would notice in the water itself. Chlorine reacts with naturally occurring phenols to form chlorophenols, detectable by taste in water at as little as 0.1 µg/l; microbes in distribution pipework then methylate those chlorophenols into chloroanisoles, whose sensory thresholds in spirit are measured in nanograms per litre. This is why free chlorine in blending water is not specified as a low number. It is specified as absent.

What happens when chlorine meets phenol?

A substitution reaction that nobody intends. Municipal disinfection puts hypochlorite into water that already contains natural organic matter, part of which is phenolic. The WHO fact sheet is explicit: chlorophenols “are present in drinking-water as a result of the chlorination of phenols, as by-products of the reaction of hypochlorite with phenolic acids, as biocides or as degradation products of phenoxy herbicides,” and “those most likely to occur in drinking-water as by-products of chlorination are 2-chlorophenol, 2,4-dichlorophenol and 2,4,6-trichlorophenol” (WHO, chemical fact sheets).

The same chemistry is a recognised fault in brewing, where the phenolic partner comes from the process rather than the catchment: “chlorophenols are formed by chlorine and chloramine in the brewing water, which bind to the polyphenol metabolites produced by the yeast during fermentation,” giving off-flavours “described as clove, medicinal, smoky or ‘band-aid’” (Muriqi, Červenka & Sýs, Food Technology and Biotechnology, 2025).

For a spirits producer the lesson generalises uncomfortably, because a phenolic partner is not hard to find — the catchment, a distribution main, a carbon bed that has stopped working, or the spirit itself, since a peated malt is sold on its measured phenol content. Nobody has an interest in discovering empirically which route dominates. The industrial answer is to remove the chlorine before it can meet anything, which is why carbon sits at the front of a distillery water train rather than somewhere in the middle.

How small a concentration can actually be tasted?

Smaller than most water specifications bother to measure. WHO gives the organoleptic numbers directly: “chlorophenols generally have very low taste and odour thresholds. The taste thresholds in water for 2-chlorophenol, 2,4-dichlorophenol and 2,4,6-trichlorophenol are 0.1, 0.3 and 2 µg/l, respectively. Odour thresholds are 10, 40 and 300 µg/l, respectively” (WHO GDWQ, ch. 10).

Read the two rows together and something unusual appears: for 2-chlorophenol the taste threshold is a hundred times lower than the odour threshold. The compound reaches the mouth before it reaches the nose. A water that smells clean can still taste wrong — precisely the failure mode that matters when the water’s entire job is to disappear into a finished spirit.

The regulatory consequence is stated plainly: “the guideline value for 2,4,6-trichlorophenol exceeds its lowest reported taste threshold” — 0.2 mg/l against a 2 µg/l taste threshold — so “if water containing 2,4,6-trichlorophenol is free from taste, it is unlikely to present a significant risk to health” (WHO, chemical fact sheets). Drinking-water regulation protects health first and leaves taste to the consumer. A distiller has no such luxury: the concentration that makes water merely unpleasant is the concentration that makes a bottling run unsellable.

Why is TCA worse than the chlorophenol it came from?

Because a common environmental reaction converts a parts-per-billion problem into a parts-per-trillion one. WHO flags the step as routine in water distribution: “microorganisms in distribution systems may sometimes methylate chlorophenols to produce chlorinated anisoles, for which the odour threshold is considerably lower” (WHO GDWQ, ch. 10). The spirits literature names the same mechanism: “the synthesis of TCA is via the O-methylation of its precursor 2,4,6-trichlorophenol (TCP) by fungi and bacteria” (Baxter, Réveillé & Conner, Journal of the Institute of Brewing, 2024).

2,4,6-trichloroanisole is the compound behind cork taint in wine, and it behaves no better in spirit. “Above the threshold concentration, 2,4,6-trichloroanisole (TCA) contributes musty/stale/damp cardboard off-notes in whisky,” and the Scotch Whisky Research Institute’s threshold work shows how little is needed: 217 panellists gave “an average detection threshold of 4.40 ng/L TCA in 20% ABV neutral ethanol,” with individual thresholds spanning “0.25 to >10 ng/L,” while in real whisky the expert panel found “TCA detection thresholds of 2 to 33 ng/L depending on the whisky matrix tested” (Baxter et al. 2024).

Worse, the damage begins before anyone can name it. Very low levels of TCA “can disrupt olfactory signal transduction, suppressing odour detection in wine… where sub-threshold concentrations of TCA suppressed desirable flavours,” and “practical observations at The Scotch Whisky Research Institute (SWRI) support the ‘flattening’ of the desirable aroma characteristics in whisky spiked with below threshold concentrations of TCA” (Baxter et al. 2024). A tainted batch need not present as musty. It can simply present as a duller version of the whisky it was meant to be — a quality failure with no obvious cause, which is the hardest kind to trace.

The arithmetic is worth doing, with the caveat that it is arithmetic rather than a measured survey. Reducing a cask-strength spirit to 40% ABV means water makes up roughly a third of the bottle, so blending water carrying TCA at 10 ng/L — a level no routine water report would flag, or even look for — would contribute on the order of 3 ng/L to the finished product, inside the 2–33 ng/L band in which SWRI’s panel detected TCA in whisky. The margin between “clean water” and “tainted bottling” is thinner than the resolution of an ordinary certificate of analysis.

Why can’t the laboratory just catch it?

Because on this compound the instrument is less sensitive than the panellist. SWRI states it directly: “when using conventional solid phase microextraction-gas chromatography-mass spectroscopy (SPME-GC-MS) analysis, instrumental limits of detection are ca.10 ng/L. This highlights the importance of sensory panels in identifying TCA tainted spirit and removing it from the production chain” (Baxter et al. 2024).

That inverts the usual relationship between a lab and a nose, with a direct operational consequence: a “not detected” on a routine TCA screen does not mean the spirit is clean. It means the concentration is somewhere below roughly 10 ng/L — above the average panellist’s threshold, and well above the most sensitive.

Hence the SWRI guidelines, worth following exactly as written. The recommendations “to optimise the sensory detection of 2,4,6-trichloroanisole in spirit samples are (i) nose samples at 20% ABV, (ii) as soon as possible after dilution conduct sensory tests and (iii) chill the sample” (Baxter et al. 2024). The first is standard industry practice for nosing anyway; the other two are specific to taint, and the paper’s headspace work explains why — “a marked decline in headspace TCA after three days at room temperature,” so a sample left standing reads cleaner than it is.

If every supply must be chlorinated, where does that leave a distillery?

With a treatment problem it cannot legislate away, because the disinfectant is mandatory by design. India’s drinking-water standard sets a minimum, not a maximum: IS 10500:2012 requires “free residual chlorine, mg/l, Min 0.2,” applicable “only when water is chlorinated,” tested “at consumer end,” and notes that “when protection against viral infection is required, it should be minimum 0.5 mg/l.” The same table caps chloramines (as Cl₂) at 4.0 mg/l with no relaxation (IS 10500:2012).

Now look a few rows further down. Phenolic compounds (as C₆H₅OH) are capped at 0.001 mg/l acceptable, 0.002 mg/l permissible — one part per billion (IS 10500:2012). A single standard therefore requires an oxidant to be present and holds its reaction partner to a thousandth of a milligram per litre. That is not an inconsistency; it is the drinking-water world’s implicit acknowledgement of the reaction above — written for public health at the tap, not for a liquid whose only job is to carry no flavour of its own.

Chlorine is perceptible in its own right long before any of this: “most individuals are able to taste or smell chlorine in drinking-water at concentrations well below 5 mg/l, and some at levels as low as 0.3 mg/l,” while for monochloramine “odour and taste thresholds of 0.65 and 0.48 mg/l have been reported” (WHO GDWQ, ch. 10). A supply fully compliant with national drinking-water rules can therefore sit within a rounding error of the level at which a sensitive person tastes the disinfectant itself — before the taint chemistry has started.

Does reverse osmosis remove chlorine?

No — and this is the commonest misconception in specifying demineralised water. Chlorine is not something an RO membrane removes; it is something the membrane must be protected from. DuPont’s guidance is unambiguous: “the feed must be dechlorinated to prevent oxidation of the membrane,” and “eventual degradation may occur after approximately 200 – 1,000 hours of exposure to 1 mg/L of free chlorine (200 – 1,000 ppm-h tolerance)” (DuPont FilmTec™ chlorination/dechlorination manual).

The actual barrier is carbon, doing two jobs at once. It destroys the residual disinfectant — “residual free chlorine can be reduced to harmless chlorides by activated carbon” (DuPont) — and it adsorbs the organics, taint compounds included: WHO records that “2,4,6-trichlorophenol concentrations can be reduced using GAC” (WHO). That second job matters, because small uncharged organic molecules are exactly what an RO membrane is worst at rejecting.

Sizing the bed is where this goes wrong in practice, since chlorine and chloramine are not the same duty. Published contact-time guidance puts “empty bed contact time for chlorine removal = 2 minutes,” against “at least 10 minutes” for chloramine on standard carbon or “at least 4 minutes” on catalytic carbon, because with chloramine “the rate of the catalytic reduction reaction is much slower” and it “requires the use of more GAC and careful monitoring for breakthrough” (Urbans Aqua, Activated Carbon EBCT). A bed sized for free chlorine that later sees a utility switch to chloramination has quietly stopped doing its job.

And carbon is never free of consequences. A carbon bed is dechlorinated, nutrient-rich, warm and dark — the most biologically active vessel in the plant, and the reason blending water must be handled as microbiologically controlled rather than merely pure. Methylation of chlorophenols to chloroanisoles is, after all, a microbial reaction.

What should a buyer specify?

Five lines, and they are not the usual five:

1. Free chlorine and total chlorine, both reported, both non-detect — not “within limits.” Total chlorine catches chloramine, which a free-chlorine test alone misses.
2. Phenolic compounds, tested to the IS 10500 method and reported against the 0.001 mg/l acceptable limit (IS 10500:2012).
3. A stated limit of detection behind every non-detect. “ND” without an LOD is not a result — and with routine TCA analysis running at “ca.10 ng/L” (Baxter et al. 2024), the LOD carries the information.
4. Sensory sign-off on the water itself, not only on the spirit: the logic behind nosing blending water as a control sample.
5. Carbon change-out records, plus notification if the municipal supply switches disinfectant — the barrier is a consumable with a breakthrough curve, and a move to chloramine changes the contact time it needs several-fold.

Most of this sits naturally on a certificate of analysis and costs nothing to ask for. The genuinely hard one — a real limit of detection on a compound active at single nanograms per litre — is the one worth insisting on, because here the instrument is the weaker detector and the panel is the backstop.

Glossary

• Free residual chlorine — the disinfecting chlorine species remaining after demand is satisfied. Specified as a minimum, not a maximum, in drinking-water standards.
• Chloramine (combined chlorine) — chlorine bound to ammonia, used as a longer-lasting disinfectant. Slower and harder to strip on carbon than free chlorine.
• Chlorophenol — a phenol ring carrying chlorine atoms, formed when chlorine or hypochlorite reacts with phenols. Taste thresholds in water from 0.1 µg/l.
• Chloroanisole (e.g. TCA) — the methyl ether of a chlorophenol, formed by microbial O-methylation, with far lower odour thresholds than its parent.
• TCA — 2,4,6-trichloroanisole, the classic cork-taint compound, giving musty, stale and damp-cardboard notes in spirit.
• Sub-threshold masking — suppression of desirable aromas by a taint present below the level at which it can itself be identified.
• EBCT (empty bed contact time) — how long water spends inside a carbon bed’s volume; the governing design parameter for dechlorination.
• LOD (limit of detection) — the lowest concentration a method can reliably distinguish from zero. A “not detected” result means nothing without it.

Frequently asked questions

Will boiling or standing remove chlorine before I dilute a whisky at home? Both reduce free chlorine, which is volatile. Neither is reliable against chloramine, and neither does anything about chlorophenols or chloroanisoles once formed — those are the taint, not the disinfectant. For diluting whisky at home, a still, low-mineral bottled water avoids the question.

Does a domestic carbon jug filter remove chloramine? Not dependably. The governing variable is contact time, and guidance puts chloramine at “at least 10 minutes” on standard carbon against 2 minutes for free chlorine (Urbans Aqua). A jug that water passes through in seconds is sized for taste, not removal.

My water report says chlorine is compliant — isn’t that enough? Compliant means the residual is high enough to disinfect and low enough to be safe. Blending water needs the opposite of a range: absence. The governing numbers are taint thresholds — parts per billion for chlorophenols, parts per trillion for chloroanisoles — not health thresholds.

Is this the same fault as cork taint in wine? The same compound family, often the same compound. In spirits TCA is “linked to contaminated wood-based materials (casks, cardboard packaging materials, pallets, wooden materials used in the construction of warehouses)” as well as to chlorinated precursors (Baxter et al. 2024). Water is one entry route among several, which is exactly why it is worth closing.

Wouldn’t demineralised water be free of this by definition? It is free of dissolved minerals, which is a different claim. Demineralisation removes ions; chloroanisoles are small, uncharged organic molecules. The protection comes from the carbon stage and the microbiological control around it — one more case where “pure” has to be read as a list of measured parameters rather than a single word.

Blenders International · Water Science series