Pure is not sterile — the microbiology of blending water, and why purified water needs more protection, not less

Key takeaways

• Purity and sterility are different specifications. Reverse osmosis and deionisation remove dissolved ions, not organisms. A water can read 0.5 µS/cm — effectively no minerals at all — and still carry a viable bacterial population.
• Purification hardware can add bacteria as well as remove it. In a pilot ultra-pure water plant the ion-exchange bed removed as little as 31.3% of viable bacteria and a UV steriliser as little as 35.8%, while an in-line 0.2 µm filter increased the count by 60.7% — the filter itself became a growth site (Journal of Industrial Microbiology and Biotechnology).
• Demineralised water is a nutrient desert, and something still lives there. Oligotrophs such as Ralstonia colonise ultrapure pipework, and biofilm survived at 56.2% surface coverage under turbulent flow at a Reynolds number of 26,000 (Duong et al., Desalination 627).
• The risk is flavour before it is safety, and the answer is already codified. Geosmin is detectable at 4–10 ng/L (IWA), and USP Purified Water pairs ≤1.3 µS/cm conductivity with ≤100 CFU/mL and ≤500 ppb TOC on one certificate (MECO). A blending-water spec sheet should do the same.

Demineralising water does not sterilise it. Reverse osmosis and ion exchange are engineered to strip dissolved ions, and they do that superbly — but organisms are not ions, and the same process that removes calcium and chloride also removes the chlorine residual that was keeping bacteria in check. The result is counterintuitive, and anyone specifying water for spirit dilution should understand it: the purest water in the plant is the least protected water in the plant.

Why doesn’t demineralisation sterilise the water?

Because it is a separation process aimed at the wrong target.

A demineralisation train works on charge and size — RO membranes reject dissolved salts, ion exchange swaps residual cations and anions for hydrogen and hydroxide — and neither step is designed as a biocidal barrier. The pilot-plant data show how partial the biological removal is. Measured against a nutrient-rich standard medium, one study recorded reverse osmosis removing 97.4% of viable bacteria, but the downstream ion-exchange bed only 31.3% and the UV steriliser 72.8%. Measured against R2A, a low-nutrient medium better suited to counting organisms adapted to starvation, the same UV unit scored just 35.8% (Journal of Industrial Microbiology and Biotechnology).

That comparison is not a rounding difference. The study found “statistically greater numbers of bacteria were observed when R2A was used as the growth medium” — meaning the industry-standard plate count systematically underestimates what lives in ultrapure water. A certificate showing a low viable count may be reporting the limits of the method as much as the quality of the water.

The second mechanism is the subtraction of a defence. Municipal supplies carry a chlorine residual precisely to stay microbiologically stable in transit — the same residual that must be stripped out before the water goes near a spirit, because chlorine reacting with phenols produces chlorophenols. Activated carbon does that job, and in doing it creates the problem: carbon “is highly effective — but can also create conditions for microbial growth if not properly controlled,” raising the risk of biofouling downstream (Nuvonic). Dechlorination is not optional for blending water. Neither, therefore, is what replaces it.

How do bacteria live in water with almost nothing to eat?

By being adapted to almost nothing, and by living on surfaces rather than in the liquid.

The organisms that matter are oligotrophs — bacteria that grow at carbon concentrations where ordinary environmental species starve. Ultrapure water systems in the semiconductor industry, far purer than anything a distillery needs, are colonised by exactly this group; Ralstonia is the recurring name, and the traces that survive purification are enough to feed it. Urea, a persistent contaminant of ultrapure water, was found to increase bacterial growth by 1.3–1.8 times (Duong et al., Desalination 627, 2026).

Crucially, they do not float. They attach. Biofilm on pipe walls is the reservoir, and the cells that turn up in a sample are largely cells that detached from it — which is why a spot sample can read clean while the system carries a standing population. That study also punctures the intuition that flow keeps a line clean: strongly attached biofilm formed at 56.2% surface coverage under turbulent flow at a Reynolds number of 26,000, and doubling turbulence still did not remove it completely.

Which is why the most instructive number here is the one that runs backwards. After an in-line 0.2 µm nylon filter — a rating conventionally called sterilising grade — the viable count in that pilot plant rose by 60.7%. A membrane rated to retain bacteria retains them where they are: on a wet surface with a trickle of nutrients passing over it. Left in service too long, an absolute filter becomes an inoculum. Carbon beds concentrate organics by design; resin offers enormous wetted surface at ambient temperature; tanks add residence time, and every hour of stagnation is an hour of growth.

So microbiological control of purified water is a matter of system design and maintenance discipline — dead legs, sample valves, tank venting and headspace, storage time, sanitisation frequency (Nuvonic) — far more than of adding another purification stage. It is where the engineering of a demineralisation plant diverges from the marketing of one: two plants can produce identical conductivity and behave completely differently across a production week.

What limits actually apply to the water that dilutes a spirit?

Two frameworks bracket the question, and neither is written for blending water specifically.

In India, the packaged-water route. Water sold as a packaged product falls under IS 14543 and FSSAI’s product standards, which impose microbiological controls alongside the chemical ones (BIS, Product Manual IS 14543). Under FSSAI’s scheme of testing, microbiological parameters must be tested monthly, and the organism list is a pathogen roll-call required to be absent per 250 mL: total coliforms, faecal streptococci, Staphylococcus aureus, Pseudomonas aeruginosa, sulphite-reducing anaerobes, Salmonella and Shigella, Vibrio cholerae and V. parahaemolyticus, and yeasts and moulds (FSSAI scheme of testing, Regulation 2.10.8).

That is a potability floor — a pathogen-absence standard. As covered in our guide to the regulatory standards for water used in spirits, potable is not the same as suitable, and pathogen absence says nothing about the population of harmless organisms that can still spoil a flavour.

In pharmaceuticals, the purified-water route. This framework actually matches the use case, because it governs water used as a process input rather than as a drink. USP Purified Water carries a conductivity limit of ≤1.3 µS/cm at 25 °C, TOC ≤500 ppb and a microbial action level of ≤100 CFU/mL; Water for Injection tightens the count to ≤0.1 CFU/mL and adds an endotoxin limit of ≤0.25 EU/mL (MECO).

The structure is the lesson more than the numbers. A pharmacopoeial specification states conductivity, organic carbon and microbial count on one certificate, because the three are linked: TOC is the food supply, conductivity is the purity, and the count is what the first two produced. A blending-water spec sheet listing only pH and TDS is describing a third of the water.

Does a high plate count mean the water is unsafe?

No — and the distinction is routinely misused in both directions. The WHO position, settled at an expert meeting in Geneva in 2002, is that heterotrophic bacteria in drinking water are not a health concern for the general public, and that heterotrophic plate count (HPC) should not be used as a safety indicator, since it does not identify potential adverse health effects (WHO; expert meeting report). What HPC is good for is process control: a marked increase after disinfection is a signal to investigate.

For blending water that reframing raises the stakes rather than lowering them. Nobody claims a few hundred CFU/mL endangers a drinker about to pour the water into 40% ABV spirit. The point is that the count is a trend line on the integrity of the system, and a system drifting upward is producing water whose organic load — and therefore whose flavour — is changing batch to batch. Consistency is the entire product promise of blending water, and this is one of the few numbers that shows whether it is being kept.

What does microbial activity actually do to a spirit?

It adds a smell, at concentrations that no mineral could match.

Minerals and microbial metabolites operate on completely different scales — the same asymmetry drawn in our review of whether water type changes whisky’s taste. Calcium becomes tasteable in the hundreds of mg/L. Geosmin, produced by actinomycetes and cyanobacteria, is detectable at roughly 4–10 ng/L and 2-methylisoborneol at 15–29 ng/L (IWA Water Practice & Technology) — four to five orders of magnitude lower. A water can be chemically immaculate, inside every pathogen-absence requirement, and still carry a musty note into a glass of single malt.

One point bounds the risk honestly: the finished bottle is not the vulnerable object. A spirit at 40% ABV sits far above any ethanol concentration that supports microbial growth; the documented rise in spoilage risk as strength falls belongs to the low- and no-alcohol category, below roughly 3.5% ABV (Journal of the American Society of Brewing Chemists). The vulnerable object is the water, upstream — in the tank, the line, the opened container. Where the two risks meet is behind the bar: a batched or pre-diluted cocktail at 20% ABV or below, made up from an open jug and held through a service, is the one point in the chain where low strength, ambient temperature and an uncontrolled water source coincide.

How is purified water kept stable after purification?

With a terminal disinfection step, a sealed package, and an acceptance that the clock starts at filling.

The dominant approach for packaged water is ozone, dosed at the end of the process rather than the beginning. Recent kinetic work in purified water put the requirement at a Ct value of 0.832 mg O₃·min/L for a 5-log reduction across ten strains of five species at 21 °C, and records the ceiling on the other side: US FDA rules specify that “residual ozone level at the time of bottling should not exceed 0.4 mg ozone per liter” (PMC). Ozone is chosen because it leaves nothing behind — it decomposes to oxygen, adding no taste of its own, which is the requirement governing every other decision about blending water.

It carries one liability that belongs on a purchase specification. Ozonating water containing bromide forms bromate, with a WHO provisional guideline value of 0.01 mg/L; reported concentrations in ozonated drinking water run from under 2 to 293 µg/L (WHO). Control means limiting formation rather than removing it afterwards, and reviews propose a source-water bromide threshold near 100 µg/L beyond which ozonation risks breaching the standard (Environmental Science & Technology). Water demineralised before ozonation is at a structural advantage: the bromide left with everything else.

One last property argues for a sealed bottle over a bulk vessel. Purified water is a hungry solvent — it absorbs carbon dioxide readily, forming carbonic acid and dropping the pH toward 5.5 or below, which shifts the very conductivity reading the specification rests on, and it can leach ions out of metals on long contact (The Chemistry Blog; Wychwood Water). The water arriving at a reduction tank is not necessarily the water that left the plant.

Does an opened bottle of blending water go off?

Not dramatically — but its microbiological quality is a function of how it is stored, and the temperature effect is large.

One study stored one-gallon HDPE containers for up to sixteen weeks at temperatures from 2 °C to over 49 °C in four locations. Heterotrophic plate counts did not exceed 100 CFU/mL in a refrigerator or 400 CFU/mL in an indoor cabinet, but reached 4 × 10³ CFU/mL on a covered porch and in a car trunk — and 4 × 10⁴ CFU/mL in containers left on an enclosed porch for years (Duranceau, Emerson & Wilder, International Journal of Environmental Health Research, 2012).

None of that is a safety story; all of it is a consistency story. For anyone diluting whisky at home: keep the bottle capped, cool and out of sunlight, and do not decant into an open jug or a carafe that lives on the bar. For a bar or a distillery: treat opened stock as an in-use item with a defined life, and buy a pack size matched to how fast it actually moves.

What should a trade buyer ask for?

Six things, and only two of them are on a typical spec sheet.

1. Conductivity or TDS, with the measurement temperature stated.
2. pH, understood as a figure that drifts acidic on air contact rather than one that is unstable in a sealed bottle.
3. Total organic carbon — the food supply that decides whether a microbial count stays where it started.
4. A microbiological count with the plating medium named, since a low-nutrient medium such as R2A and a nutrient-rich one give materially different answers in ultrapure water (Journal of Industrial Microbiology and Biotechnology).
5. Pathogen absence to the applicable packaged-water standard, with sample volumes stated — the FSSAI/IS 14543 “absent per 250 mL” list, tested monthly.
6. The terminal disinfection method and its by-product control — if ozone, a bromate figure alongside it.

Then the questions about the system rather than the sample: how often the plant is sanitised, what the storage residence time is, how the tank is vented, what shelf life the sealed pack carries. As with the certifications behind a purity claim, those answers are the more informative ones — the numbers describe one batch, the answers describe every batch.

Glossary

• Sterile — free of viable organisms. Distinct from pure, the absence of dissolved and suspended matter. Blending water is specified as pure; it is not, and need not be, sterile.
• Oligotroph — an organism adapted to growth at extremely low nutrient concentrations; the characteristic inhabitant of purified-water systems, Ralstonia being the common example.
• Biofilm — microorganisms attached to a wetted surface in a matrix they secrete. In purified-water systems the biofilm on pipework, not the bulk liquid, is the reservoir; sampled cells are largely detached from it.
• HPC / TVC — heterotrophic plate count, or total viable count: culturable bacteria in a sample. A process-control indicator, not a health indicator, and highly dependent on the growth medium used.
• TOC (total organic carbon) — the nutrient budget that limits microbial regrowth in purified water. USP Purified Water caps it at 500 ppb.
• Bromate — an oxidation by-product formed when ozone meets bromide. WHO provisional guideline value 0.01 mg/L.
• Geosmin / 2-MIB — microbial metabolites behind earthy and musty off-odours, detectable at single-digit to low-tens nanograms per litre.

Frequently asked questions

Does demineralised water contain bacteria? It can. Demineralisation removes dissolved ions, not organisms, and it strips the chlorine residual that would otherwise suppress growth. Purified-water systems are routinely colonised by oligotrophs living in biofilm on pipework and filter media (Journal of Industrial Microbiology and Biotechnology).

Is demineralised blending water sterile? No, and it is not specified to be. The relevant benchmark is USP Purified Water at ≤100 CFU/mL, not Water for Injection at ≤0.1 CFU/mL (MECO).

Can bacteria in water spoil a bottle of whisky? Not by growing in it — 40% ABV is far above any concentration that supports microbial growth. The risk is an off-odour carried in by the water, at thresholds as low as 4–10 ng/L for geosmin (IWA).

Why is a 0.2 µm filter not enough on its own? Because a filter retains organisms rather than destroying them, and a wet membrane with nutrients passing over it can become a growth site. In one pilot plant viable counts rose 60.7% across an in-line 0.2 µm filter (Journal of Industrial Microbiology and Biotechnology).

Does a high plate count mean the water is unsafe to drink? No. WHO’s position is that heterotrophic bacteria in drinking water are not a health concern for the general public and that HPC should not be used as a safety indicator; its value is as a process-control signal (WHO).

Should an opened bottle of blending water be refrigerated? Cool, capped and dark is the practical rule. Plate counts stayed below 100 CFU/mL in a refrigerator and 400 CFU/mL in a cabinet, but reached 4 × 10³ CFU/mL in a car trunk or on a porch (Duranceau et al., 2012).

Blenders International · Water Science series