Water Chemistry for Dualboilers: Magnesium, Calcium and the Extraction of Dark Roasts
Water Chemistry for Dualboilers: Magnesium, Calcium and the Extraction of Dark Roasts
Water is more than 98 percent of what stands in your cup. With a dualboiler it's also the factor that decides how long the machine lasts. We've laid the available research, the SCA recommendations and the actual water values from Zurich and Bern side by side.
Magnesium extracts somewhat more efficiently than calcium – that much is documented. That it tastes better is not. The SCA Water Quality Handbook states in so many words that the sensory difference is still unresolved.
The more important factor is alkalinity (bicarbonate). It buffers the coffee acids away and is clearly higher in Zurich and Bern than the SCA recommendation. Hardness isn't the problem here, bicarbonate is.
With a dualboiler, the steam boiler sets the ceiling: that's where water evaporates, the minerals stay behind and build up. Which is why hardness has a double budget – one for the taste, one for the service life.
What magnesium and calcium actually do
Distilled water extracts poorly. That's undisputed, and the reason lies with the dissolved cations. In 2014, Hendon, Colonna-Dashwood and Colonna-Dashwood calculated in the Journal of Agricultural and Food Chemistry how strongly sodium, magnesium and calcium bind to typical coffee molecules – five coffee acids, caffeine and eugenol as a stand-in for an aroma compound. The cations coordinate to nucleophilic sites in the molecule and form soluble complexes. The ranking: magnesium ahead of calcium, sodium far behind.
1 · Brew water: Magnesium and calcium are dissolved in the water. 2 · Coffee bed: The ions travel in and bind aroma molecules there. 3 · Cup: What has been bound ends up in the drink.
Much simplified model. The ranking magnesium before calcium before sodium comes from a quantum-chemical calculation (Hendon et al. 2014) — not from a measurement on the drink and not from a tasting.
Important for context: Hendon et al. (2014) never brewed or tasted a single coffee – the work is a quantum-chemical calculation. Bratthäll, Figueira and Nording (2024) checked it experimentally: at around 100 ppm, i.e. typical of drinking water, they found no significant difference between magnesium and calcium; only at 1000 ppm did it become measurable. We show the model here because it explains the mechanics well – not because it is proven.
Peer reviewThat check came in 2024 from Bratthäll, Figueira and Nording in Heliyon. They added magnesium and calcium chloride before and after brewing and measured citric, malic, lactic and quinic acid by GC-MS and NMR. Their result: at concentrations common in drinking water, the acid content changed only to a limited extent; only at ten times the salt concentration did it become clear. The authors write explicitly that a scientific basis for the coffee industry's water recommendations is long overdue.
The SCA Water Quality Handbook, written by Wellinger, Smrke and Yeretzian at the Coffee Excellence Center of ZHAW Wädenswil, puts it cautiously: preliminary trials pointed to a slightly higher extraction efficiency for magnesium, but the difference in sensory properties remains unclear and is a subject for future research. We find that restraint likeable – and it matches our own experience that with water, the coarse mistakes matter far more than the fine ion balance.
Alkalinity – the underrated lever
Alkalinity, practically synonymous with bicarbonate content, is not a flavour compound. It is a buffer. Bicarbonate neutralises the acids that are dissolved out of the coffee: HCO3− plus H+ becomes carbonic acid and breaks down into water and CO2. The SCA handbook puts it in one short formula: the higher the alkalinity, the lower the perceived acidity.
At very high alkalinity above roughly 100 ppm CaCO3, such large amounts of CO2 are produced that, according to the handbook, extraction time can lengthen and extraction can tip towards over-extraction. Elevated sodium reinforces this – one reason why classic sodium ion exchangers are the wrong technology for espresso machines: they lower the hardness but leave the bicarbonate standing.
Conversely, too little alkalinity is no solution either. Below 40 ppm CaCO3 the handbook warns explicitly about corrosion, because even small amounts of dissolved CO2 can push the water below pH 6. Water that is too soft is more dangerous for an espresso machine than water that is slightly too hard.
The target values and where Zurich and Bern sit
To the right the water gets harder, upwards it is more strongly buffered. Both axes are linear and in the same unit. For the Swiss scale: 100 ppm CaCO₃ equals 10 °fH.
Rectangles = measured ranges of variation over the year, not single values. Data: Stadt Zürich water quality 2024 · Energie Wasser Bern · SCA Water Quality Handbook.
Zurich's total hardness is still within range, Bern's is not. On alkalinity both cities sit above the recommendation – Bern at around two and a half times the upper limit of 75 ppm. Which is exactly why softening alone is not enough here: a sodium ion exchanger takes the hardness away and leaves the alkalinity standing.
| Parameter | SCA recommendation | Zurich | Bern |
|---|---|---|---|
| Total hardness | 50–175 ppm CaCO₃, for espresso machines 50–150 | 139–160 ppm | approx. 210 ppm |
| Alkalinity | 40–75 ppm CaCO₃ | 127–148 ppm | 176–203 ppm |
| pH | 6–8 | 7.6–7.9 | approx. 7.6 |
| Chloride | below 5 mg/l (common rule of thumb, not an SCA standard) | 5.2–8.6 mg/l | 4.4–6.4 mg/l |
| Sodium | keep low | 6.6–9.1 mg/l | 3.4–4.3 mg/l |
A note on hardness for the Swiss scale: the cantonal laboratory of Zurich classifies below 15 °fH as soft and 15 to 25 °fH as moderately hard. Zurich sits at around 14 to 16 °fH, Bern at about 20 °fH. How to put your water together deliberately is something we've described step by step under composing your own coffee water, and in the water chemistry simulator you can play hardness and alkalinity off against each other directly.
Why the steam boiler sets the limit
This is where the real difference between a single-circuit machine and a dualboiler lies. The brew boiler has water flowing through it: water in, water out. The steam boiler is the only place in the machine where water leaves the liquid phase. The steam goes out mineral-free, the ion load stays behind – and is topped up again with every automatic refill. The steam boiler concentrates.
The steam goes out mineral-free, calcium and magnesium stay behind. Scale precipitates the hotter the water gets — calcium carbonate dissolves less well when warm, not better. So the steam boiler scales up first and more than the brew boiler.
Bar lengths drawn to scale, values from the SCA Water Quality Handbook.
For context: the 30 and 12 mg/l are not formation rates but the amounts the SCA handbook assumes as maximum precipitable – and the right-hand boundary of the core zone derives from that. What ends up in the boiler you can project: at 100 shots a day of 0.1 litre each, 30 mg/l equals around 110 grams of scale per year, 100 mg/l already 365 grams, 200 mg/l around 730 grams. That is why, in Bern and Zurich, we talk about water first at every dualboiler consultation and only afterwards about the machine.
This is exactly why the right-hand boundary of the SCA core zone is not set by taste but by the steam boiler. And why dualboilers are also more vulnerable to the consequences of wrong water treatment than single-circuit machines.
ManufacturerLa Marzocco states in its own water specification: TDS 90–150 ppm, total hardness 70–100 ppm, alkalinity 40–80 ppm, pH 6.5–8, chloride 0–30 ppm, free chlorine 0.05 ppm maximum. That the chloride limit sits at 30 ppm there while the rule of thumb common in the scene is 5 mg/l is not a mistake on either side – the SCA standard, incidentally, names no chloride value at all – these are different risk thresholds. A warranty limit on one, best practice on the other. Good evidence that there is no binding standard here: the SCA currently publishes no water standard at all, and the much-quoted table comes from the 2009 SCAA legacy document.
Dark roasts: what's proven and what isn't
Now to the part where we had to correct our own expectation. The industry formula goes: darker roast equals more soluble equals higher yield. In 2024, Lindsey and colleagues showed in Scientific Reports the opposite – though their work primarily investigated caffeine content and yield came along as a side finding. With filter coffee, extraction yields fell from a roast loss of about 12 to 14 percent onwards. A natural-processed Ethiopian reached around 21 percent yield when light, and around 15 percent when dark under identical conditions.
What dark roasts actually do: at given parameters they reach their yield optimum faster, but that optimum sits lower. And they are more brittle, produce more fines at the same grinder setting and thereby change the flow through the puck. That is a process effect, not a solubility effect.
The common recommendation to choose higher alkalinity for dark roasts so that the bitterness gets rounder is plausible, but not proven. We found no publication that systematically tests roast level against water composition. Anyone telling you otherwise should be able to name the study. Our position on it: try it, write it down, decide for yourself.
What is documented for espresso, on the other hand, and is happily overlooked: the SCA handbook notes that for espresso you can leave alkalinity at up to 150 ppm CaCO3 before a clear buffering effect on the acidity becomes noticeable. The reason is the brew ratio: at 1:2, around seven times less buffer capacity arrives per gram of coffee than with a filter at 1:15. So filter coffee needs less alkalinity than espresso, not more.
Frequently asked questions
Is magnesium really better than calcium for coffee?
What is documented is a slightly higher extraction efficiency for magnesium. At the same time, the SCA Water Quality Handbook writes that the sensory difference between calcium and magnesium remains unclear. An experimental check (Bratthäll et al. 2024) found only limited differences at concentrations typical of drinking water. So “magnesium tastes better” is experience, not a measurement.
Which water values are ideal for an espresso machine?
As a guide: total hardness 50–150 ppm CaCO₃, alkalinity 40–75 ppm CaCO₃, pH 6–8, chloride as low as possible – the SCA names no value for it, in practice below 5 mg/l counts as a rule of thumb –, no free chlorine. For espresso, alkalinity may go up to 150 ppm according to the SCA handbook before the buffering effect becomes noticeable. Manufacturers deviate from this in part – La Marzocco, for example, allows up to 30 ppm chloride.
Is a normal water softener enough for my dualboiler?
Usually not. Classic sodium ion exchangers swap calcium and magnesium for sodium but do not lower the alkalinity. In Zurich and Bern it is precisely the alkalinity that is the critical value. Cartridges that reduce alkalinity make more sense, or reverse osmosis with subsequent remineralisation. La Marzocco explicitly advises against sodium-based softeners in its own specification.
Why is water that is too soft a problem?
Below 40 ppm CaCO₃ alkalinity the SCA handbook warns about corrosion: even small amounts of dissolved CO₂ can push the water below pH 6. On top of that, very mineral-poor water extracts less well. Fully demineralised water does not belong in an espresso machine.
Do dark roasts need different water than light ones?
Here is the honest answer: there is no study that crosses roast level and water chemistry experimentally. It is plausible that higher alkalinity dampens the acidity and makes dark roasts feel rounder. It is not proven. We recommend trying it with your own coffee instead of following a rule of thumb.
Roastery tip
If you want to test your water change, take a coffee you know by heart. Cozy Chocolate and Nutty Delight react noticeably to alkalinity – too much bicarbonate takes their contour away. Wild Peach shows you most clearly when the acidity is buffered away, because that is where there is most of it. That is our observation from the roastery, one of many.
