Water for coffee

Carbonate chemistry#

Four equilibria tie carbon dioxide, alkalinity and scale together.¹

1. CO₂ (dissolved) + H₂O ⇌ H₂CO₃ (carbonic acid)
2. H₂CO₃ ⇌ HCO₃⁻ + H⁺
3. HCO₃⁻ ⇌ CO₃²⁻ + H⁺ (only relevant in strongly basic water)
4. Taken together with limestone: CaCO₃ + CO₂ + H₂O ⇌ Ca²⁺ + 2 HCO₃⁻

Equations 1–3 are given in the SCA handbook, which lumps dissolved CO₂ and carbonic acid together, as is customary in water chemistry.¹ Equation 4 is the sum of dissolution and protonation that the handbook describes in words: acidic water dissolves carbonate rock and acquires calcium and hydrogen carbonate.¹ Read forwards, it explains how limestone becomes hardness and alkalinity. Read backwards, it is scale: heating drives off CO₂, and calcium carbonate precipitates.^1,3

Normal rain has a pH of about 5.6 because carbon dioxide from the air dissolves in it;⁵ the SCA handbook explains the slight acidity of pure water in contact with air the same way.¹ Most tap water carries carbon dioxide in excess of that equilibrium, which strongly influences its pH.¹ Water in equilibrium with air holds much less. Our own calculation from Henry’s law gives about 0.7 mg/L: with a solubility constant of 3.4 × 10⁻⁴ mol/(m³·Pa) at 25 °C and a temperature coefficient of 2,300 K,¹⁹ air with 400–420 ppm CO₂ (40.5–42.6 Pa) gives 0.69–0.73 mg/L at 20 °C. The SCA handbook gives a lower value (see Contested). Stirred in an open container for a day, a hard tap water in one of the handbook’s examples lost most of its excess carbon dioxide and its pH rose markedly.¹

Charge must balance. Calcium carbonate removed as scale takes one calcium ion but two hydrogen carbonate ions, because each calcium carries two charges.¹ This is why units based on equivalents, such as ppm as CaCO₃, make hardness, alkalinity and scale directly comparable.¹

Why filter needs less alkalinity than espresso#

Filter coffee uses far more water per gram of coffee than espresso, so the same water brings far more buffer into contact with the same acids.¹ The SCA assumes a ratio of coffee to beverage of about 1:15 for filter and 1:2 for espresso (see Brew ratio and strength for the different ratios).¹ It concludes that espresso tolerates up to about 150 ppm alkalinity before buffering becomes apparent, while filter needs less.¹

Our own arithmetic, from these cited numbers: 1 g of coffee brewed at 1:15 with water of 50 ppm alkalinity meets 0.75 mg of acid-neutralising capacity (as CaCO₃), or about 0.015 mmol of bicarbonate. At 1:2 it meets 0.10 mg – 7.5 times less. Espresso brewed with 150 ppm water still meets only 0.30 mg per gram. This simple calculation uses beverage weight and ignores water held back in the grounds.

Hardness behaves differently. The SCA finds that total hardness has less impact than alkalinity and that even hard water can give a high-quality filter brew.¹

Scale and corrosion#

Scale. Scale forms where calcium meets carbonate and heat, mostly as calcium carbonate; at very high pH magnesium hydroxide can form too.¹ It lowers heat transfer and clogs valves and jets, especially in hot-water sections.¹ Gypsum-rich water forms scale of a different kind.¹ Silica scales regardless of pH and alkalinity, and is hard to remove.² Lower pH means less scale.¹

The SCA estimates maximum scale from a modified Langelier index and turns it into a yearly amount from the number of drinks.¹ Our own arithmetic, following its example: at 100 drinks a day and 0.1 L per drink including flushing, a machine uses about 3,650 L a year, so water able to deposit 30, 100 or 200 mg/L of scale can leave about 110, 365 or 730 g. The estimate holds best when the machine stands hot for hours; at high throughput the water does not stay hot long enough for all of it to form.¹

Corrosion. The SCA lists the conditions that make corrosion likely:¹

• alkalinity below 40 ppm as CaCO₃, which leaves too little buffer;
• pH below 6 or above 8;
• chloride, sulfate or nitrate above about 80% of the alkalinity in equivalents (a Larson–Skold index above 0.8). For gypsum waters this becomes: total hardness no more than 80% higher than alkalinity;
• for salt-influenced waters, electrical conductivity (µS/cm) more than three times the alkalinity (ppm as CaCO₃). The SCA derives this rule from measured waters, among them tap waters near Barcelona.

Chloride is a particular concern. The SCA advises calcium or magnesium sulfate rather than chlorides for hardening water used in espresso machines, because chlorides increase the chance of corrosion; espresso machine makers often advise against water with significant amounts of chloride.¹ Soft water that is not stabilised tends to corrode pipes in general.⁷

Decarboniser limits. A decarboniser turns bicarbonate into carbonic acid. In an open kettle the CO₂ escapes when the water is heated, but in a closed espresso system it stays.¹ Removing a large amount of alkalinity therefore leaves much dissolved CO₂; the water turns acidic, corrosion risk rises, and espresso can foam with large, short-lived bubbles.¹ Our own arithmetic: removing 250 ppm of alkalinity (as CaCO₃) converts 5 mmol/L of hydrogen carbonate into 5 mmol/L of CO₂, about 220 mg/L. The SCA therefore advises setting the bypass so that output alkalinity is 40–70 ppm for waters up to 250 ppm, and so that alkalinity falls by no more than 200 ppm for harder waters.¹

The SCAA handbook adds that corrosion in brewers is rarely caused by natural water: acid flux from soldered copper pipe joints is a more usual culprit.²

Off-flavours and contaminants#

• Chlorine and chloramines used for disinfection give unpleasant flavours and should be removed, most simply with activated carbon.^1,2 Chlorine can react with coffee phenols to form chlorophenols, which have a medicinal taste.²
• Iron turns coffee greenish. In a 1955 study, a difference could be seen side by side at 1 ppm, and above 7 ppm coffee with cream had a clearly unpleasant greenish cast.¹⁴
• Lead is toxic. Only reverse osmosis and deionisation remove it completely.¹
• Taste thresholds. In water, the taste threshold for calcium is 100–300 mg/L, depending on the anion.⁷ In the 1955 study, coffee made with water containing single ions at their thresholds in water could not be told apart from coffee made with distilled water.¹⁴

Measuring water#

Drop titration. Outside a laboratory, hardness and alkalinity are measured by titration, as two separate tests.¹ Hardness is titrated with EDTA; alkalinity with a strong acid such as hydrochloric acid.¹ Reagent is added drop by drop until the colour changes completely, and the drops are counted.¹ The SCA recommends a resolution of at least 20 ppm (1 °dH). Doubling the sample volume halves the value of each drop, which helps with low-mineral treated waters.¹

Conductivity meters. Electrical conductivity depends on temperature: depending on the salts present, it rises by about 1–3% per °C.¹⁰ Measurements are therefore referred to a standard temperature, typically 25 °C, and field meters often correct with a fixed coefficient of 0.019–0.020 per °C.¹⁰ The SCA handbook uses 20 °C as its reference and warns that many cheap meters do not correct at all.¹

Converting conductivity to TDS adds a second uncertainty. In the natural waters studied by Hem of the US Geological Survey, the ratio of dissolved solids (mg/L) to conductivity (µS/cm) ranged from 0.54 to 0.96.¹¹ It is much higher for dissolved calcium carbonate than for table salt.¹ The heritage SCAA standard assumed one fixed factor, but many meters use another or a variable factor.¹ The SCA therefore treats a TDS estimate as rough, with large deviations possible.¹

Daily checks. Once a treatment is set, a conductivity meter can track it. For decarbonisation, conductivity falls roughly in proportion to the alkalinity removed, which the SCA confirmed on Swiss tap waters.¹ A softener barely changes conductivity, so conductivity cannot track it.¹

Making water from salts#

Recipe waters start from almost pure water and add salts in known amounts.¹ Sodium hydrogen carbonate raises alkalinity only; magnesium or calcium sulfate or chloride raises hardness only.¹ As an example, the SCA dissolves 3 g of pure sodium bicarbonate in 1 L of reverse-osmosis water to make a concentrate of 100 °dH alkalinity; 10 g of it in 1 L of water adds 1 °dH.¹ We checked the arithmetic from molar masses: 3 g of NaHCO₃ is about 35.7 mmol, 35.7 mmol/L of hydrogen carbonate is about 100 °dH, and diluting about 10 g in a litre is roughly a hundredfold dilution.

Mixing follows a lever rule. For a target alkalinity, the share of water 1 is (alkalinity of water 2 − target) ÷ (alkalinity of water 2 − alkalinity of water 1); the same formula works for hardness.¹ The SCA’s worked example mixes a water of 5 ppm alkalinity with one of 150 ppm to reach 60 ppm.¹ The handbook prints 64.1% of the soft water, but its own formula gives 62.1% (our own arithmetic): (150 − 60) ÷ (150 − 5) = 90 ÷ 145 = 0.621, and 0.621 × 5 + 0.379 × 150 = 60 ppm.

Reference values#

Sample waters#

The water chart under Check your water offers sample waters. They are worked examples from the SCA handbook: points used to show how mixing and treatment move a water, not measurements of real supplies.¹

Real waters can sit far from the limestone diagonal: gypsum adds hardness without alkalinity, and sea-water intrusion adds alkalinity without hardness.¹

Open questions#

• Calcium versus magnesium. Theory and preliminary experiments suggest magnesium extracts slightly more than calcium, but its effect on sensory properties remains unclear.^1,8 The total-hardness concept treats the two as one.¹
• Brew ratio, temperature and pressure. The SCA expects the optimum to shift with brew ratio and with temperature and pressure, and calls for correction factors or separate zones.¹
• Roast style. Anecdotal evidence suggests that harder, more alkaline water can partly compensate for underdeveloped roasts.¹ No study opened for this entry tested it.
• Local taste. Where tap water is very soft, preferences and roasting styles may have adapted to it.¹
• The evidence behind TDS 150 mg/L. The SCAA target rests partly on a small blind tasting by committee members, which the handbook itself says cannot be considered statistically valid.²

Key studies#