Cement hardening
A Socratic walk-through of cement hardening — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does concrete set perfectly well underwater, when you would expect it to need to dry out?
Everything else we mix with water and let harden — paint, clay, mud — hardens by losing the water, so the intuition that concrete "dries out" is not lazy but the obvious generalisation. And yet bridge piers finish setting under the river, and a slab left to dry fast in the sun comes out weaker than one kept deliberately wet for a week.
Two facts pointing the same way. If drying were the mechanism, keeping the material wet would prevent the very thing you want. So what is the water doing, if not leaving?
Reasoning it through
REASONING #Begin with what is in the bag. Portland cement is not a powdered stone but a manufactured one, made by burning limestone with clay at around 1450 °C. The clinker that comes out is dominated by calcium silicates — chiefly alite, tricalcium silicate, with belite alongside it — plus smaller amounts of aluminate and ferrite phases. These are minerals formed far from the conditions at the bottom of a bucket, which is the whole point: they are thermodynamically unstable in the presence of water. The kiln stored energy in them, and mixing releases it.
So what does the water do? Not dissolve and later evaporate — it reacts. The silicate grains dissolve at their surfaces, the surrounding solution becomes supersaturated with respect to a different set of minerals, and those precipitate: a poorly crystalline, extremely fine gel of calcium silicate hydrate, together with crystals of calcium hydroxide. The water is consumed and locked chemically into the products. Ask the question that settles it — where has the water gone in a set block? A substantial fraction is still in there, bound, and driving it off means heating the concrete hundreds of degrees, at which point the concrete is destroyed.
Now the mechanical question, which is separate. Why does a slurry become a solid? Because the hydrate gel grows outward from every grain into the water-filled spaces, and the growths from neighbouring grains meet and interlock. Hardening is space being filled by crystal growth. That explains a fact which otherwise looks perverse: adding extra water to make a mix easier to place makes the finished concrete weaker, even though water is a reactant. The extra water adds no reaction — there was already enough — but holds open space the gel cannot fill, and those become capillary pores. Roughly, full hydration needs something near 0.4 parts water to one part cement by mass, counting both what is chemically bound and what stays held in the gel; the figure is approximate and depends on definition, but the direction is not in doubt, and mixes are often placed wetter purely for workability.
Time to test the account rather than accept it. If setting really is hydration rather than drying, a cement that sets by a different route should behave in the opposite way — and one exists. Ordinary lime mortar, burnt limestone without the silicate clinker, hardens by reabsorbing carbon dioxide from the air and slowly turning back into calcium carbonate. It cannot set underwater, and it cannot set deep inside a thick wall where no air reaches. That is exactly the contrast the theory demands, and it is why the Romans, wanting harbour works, added volcanic ash to their lime: the reactive silica gave them silicate hydrates and so a mortar that set in seawater.
One further prediction. If the reaction is exothermic, a large enough pour should cook itself — and large dams do have to be built in thin lifts or with cooling pipes cast into them, because the heat of hydration would otherwise crack the structure as it later contracted.
The analogy
THE ANALOGY #Think of a dry-stone wall being built inside a flooded trench, where the mortar is grown rather than applied: each stone slowly sprouts crystal fingers into the water around it, and the fingers from neighbouring stones tangle together. Nothing needs to dry, because the water is not in the way of the joint — it is the medium the joint grows through, and its dissolved contents are the joint's raw material.
in a real paste the growing fingers are also being fed by the dissolution of the stones themselves, so the grains shrink as the joints grow, which no wall does.
Clarifying the model
THE MODEL #Curing is not waiting; it is keeping water available. A slab that dries early stops reacting, and unreacted cement is expensive filler. Ponding, misting and plastic sheeting all keep the reaction supplied, and the conventional test at 28 days is a convention about how much of a very long reaction to count, not a point at which it ends — concrete kept wet goes on gaining strength for years.
Concrete does nevertheless dry, and that matters, but as a separate and later physical process: loss of water from the pores causes drying shrinkage, and shrinkage restrained by anything at all causes cracking. So "concrete does not set by drying" and "drying matters to concrete" are both true, about different things.
Two boundaries are worth marking. The reinforced-concrete explanation in this collection takes hardened concrete as given and asks why it needs steel to carry tension; this one asks how it became hard at all. And the mineral-from-solution mechanism here is the one behind limescale in a kettle, but it settles differently: limescale precipitates because heating drives off dissolved carbon dioxide and lowers the solubility of a mineral already in solution, while cement paste manufactures its own supersaturation by dissolving a metastable mineral the kiln made.
An honest limit: calcium silicate hydrate is poorly understood at the atomic scale. Its composition varies, it is nearly amorphous, and its structure is a live research question — so "a gel that fills the pores and interlocks" is a sound account of what it does and a vaguer one of what it is.
A picture of it
THE PICTURE #How to readRead left to right as elapsed time since the water went in, with each entry naming what is happening chemically above and what a builder would notice below. The second column is the one to dwell on — the dormant period is why concrete can be mixed at a plant and placed an hour later, and it is a genuine pause in the reaction, not merely slowness. Note that nothing anywhere on this line is evaporation, and that the line does not end.
What became clearer
WHAT CLEARED #Concrete does not dry; it grows. The kiln makes minerals that should not exist in the presence of water, and mixing lets them dissolve and re-precipitate as a hydrate gel that fills and bridges the spaces between the grains. Water is the reagent and the medium, not an impurity to be driven off — which is why setting underwater is unremarkable, why premature drying is a defect rather than a completion, and why a mix made wet enough to pour easily is buying workability with permanent porosity.
Where to go next
ONWARD #- Why supplementary materials such as fly ash and slag change both the strength curve and the durability of the result.
- How chloride and sulfate attack exploit the pore network that the mix water leaves behind.
Key terms
TERMS #| Term | What it means |
|---|---|
| Clinker | the fused nodules produced by burning limestone with clay, ground with gypsum to make Portland cement. |
| Hydration | the reactions in which cement minerals dissolve and re-precipitate as hydrated products, consuming water chemically. |
| Calcium silicate hydrate | the nearly amorphous gel that makes up most of the binding phase in hardened cement paste. |
| Water-to-cement ratio | mixing water divided by cement, by mass; the single strongest predictor of hardened strength. |
| Pozzolan | a reactive silica-bearing material, such as volcanic ash or fly ash, that converts lime into further hydrate. |
Every term the collection defines is gathered in the glossary.