Ice point, triple point of water and other fixed points

The ice point

• Shaved or crushed ice from pure water, packed in a Dewar and saturated with cold water so no free water sits at the bottom.
• Probe immersed to adequate depth, allowed to equilibrate; the reading should be stable to within resolution.
• Re-pack as the ice melts; a slush with floating ice is not an ice point.
• Common errors: tap water, large ice cubes, insufficient immersion, probe touching the Dewar.

Triple point of water

A sealed glass cell with pure water, ice and vapour in equilibrium defines 273.16 K exactly in ITS-90. It is used to measure R(0.01 °C) of a reference PRT before and after use, so its drift can be tracked and its coefficients corrected. Preparation requires an ice mantle and a maintenance bath; the reward is a reference check with negligible uncertainty.

Why fixed points are different from a calibrated reference

A calibrated reference thermometer knows what temperature it is at because somebody told it, and the chain of telling is the traceability. A fixed point knows because physics makes it so: a pure substance in equilibrium between two phases sits at a temperature defined by nature and the pressure, and nothing about the apparatus can move it. That is why fixed points anchor the scale rather than sitting on it.

The triple point of water, and what it is for

A sealed glass cell containing pure water with ice, liquid and vapour in equilibrium realises 273.16 K, which ITS-90 fixes exactly. Its practical use in a working laboratory is drift tracking: measure R(0.01 °C) — often written W or R_TPW — for a reference PRT before and after a calibration campaign, and any change is the thermometer moving rather than the standard. That measurement is also the denominator of the ITS-90 ratio from which every other temperature is computed, so it is not an optional extra for reference work.

Cells are not cheap and they are fragile, which is why the ice point remains the workhorse. An ice point at 0.002 °C is more than good enough to detect a PRT drifting by 0.01 °C, and a Dewar full of shaved ice costs nothing.

Making an ice point that is actually at 0 °C

• Shaved or finely crushed ice from pure water. Cubes leave voids that are insulators, and tap water depresses the freezing point.
• Pack it tightly into an insulated Dewar, then add just enough cold pure water to fill the gaps.
• No free water at the bottom. Water at 4 °C is denser and collects there, and a probe sitting in it reads several tenths high — always in the same direction, which makes it look like a stable result.
• Immerse the probe fully, away from the walls, and let it stabilise. The reading should settle within a millikelvin or two.
• Replenish as it melts. A slurry that has become mostly water is no longer a fixed point.

A fixed point realises a temperature; it does not create traceability on its own. It becomes traceability when the realisation is documented with an uncertainty, which is the difference between a check and a calibration.