Nothing Is Simple: The $120,000 Gallon of Water

I have a theory I keep coming back to, and it’s this: nothing is as simple as it seems at first glance. Nothing. Pick anything you think is trivially easy, look hard at how it’s actually done well, and the simplicity evaporates.

Hanging a door is my favorite example. It sounds like a Saturday-afternoon job — screw some hinges on, done. But do it right and you’re suddenly reckoning with reveal gaps, hinge-side versus latch-side clearance, whether the frame is even plumb (it isn’t), which way the floor slopes, how the door will move as the seasons change the humidity, and the fact that “level” and “looks right” are frequently not the same thing. The difference between a door a pro hung and a door you hung is a lifetime of that door quietly binding, or swinging open on its own, or scraping the jamb every winter.

So I collect these examples. And I recently found a spectacular one, thanks to an article at Signore Galilei: the freezing and boiling points of water.

The thing every fourth-grader knows

Water freezes at 0°C and boils at 100°C. This is possibly the most universally known fact in all of science. It’s on the first page of the textbook. It is the example of a clean, round, obvious natural constant.

It is also, once you need it to be precise, almost entirely wrong. Or at least so heavily qualified that the round numbers become a polite fiction we agree to teach children.

Here’s how it comes apart.

Not all water is the same water

Start with the molecule. “H₂O” hides an assumption: that a hydrogen atom is a hydrogen atom and an oxygen atom is an oxygen atom. They aren’t. Hydrogen comes in isotopes — ordinary protium, heavier deuterium, radioactive tritium — and oxygen comes as oxygen-16, -17, and -18. The heavy versions are rare, but they’re really there, mixed into every glass of water you’ve ever drunk.

And heavy water behaves differently. Water built from deuterium and oxygen-18 freezes at around 4°C, not 0°C. Even the trace natural variation in isotope ratios is enough to shift the freezing point by about 0.001°C — which sounds like nothing until you realize a decent lab thermometer can see 0.001°C.

Worse, the ratios aren’t even constant around the planet. Heavier isotopes evaporate a hair more reluctantly, so ocean water, rainwater, glacier water, and groundwater all carry slightly different isotopic fingerprints. The water in Vienna is not, at the level that matters, the same substance as the water in the Pacific. So “the freezing point of water” isn’t one number — it’s a small smear of numbers that depends on which water you grabbed.

So somebody had to define “official” water

You can’t calibrate the world’s thermometers against a quantity that changes depending on which faucet you used. So metrologists did the only thing they could: they defined a reference water.

In 1961 Harmon Craig proposed Standard Mean Ocean Water — an idealized average of ocean isotope ratios. After the inevitable squabble over competing standards, the International Atomic Energy Agency settled it in 1966, and the winning recipe, prepared and kept in Vienna, became Vienna Standard Mean Ocean Water, or VSMOW. It is, quite literally, the official definition of what “water” means for the purpose of precision measurement.

And here’s the punchline from the article that pulled me in. You can buy VSMOW. A single ampoule — five milliliters, about a teaspoon — runs about $159. Do the arithmetic out to a gallon and you land at roughly $120,000 per gallon for the most officially, precisely defined water on Earth. A round number every child knows, backstopped by a bottle of water worth more than a car.

The freezing point isn’t even the fixed point anymore

It gets better. Once you’re chasing this level of precision, the freezing point itself turns out to be a bad reference — because it depends on pressure and on how much air is dissolved in the water. (The old 0°C was quietly defined using air-saturated water at exactly one atmosphere. Remove the dissolved air and the number moves.) And pure water is perfectly happy to supercool well below 0°C without freezing at all if you don’t give it a nucleation site.

So the standard the world actually uses isn’t the freezing point. It’s the triple point of water: the single, exquisitely specific combination of temperature and pressure where ice, liquid, and vapor coexist in equilibrium. That happens at 0.01°C and about 611 pascals — a near-vacuum. It’s reproducible to a fraction of a millikelvin, which the sloppy old freezing point never could be. Fill a sealed glass cell with VSMOW-composition water, purge the air, and you have a triple-point cell — the physical object labs calibrate against. Those cells defined the Kelvin itself until the 2019 SI redefinition handed that job to a fixed value of the Boltzmann constant.

That’s why the water has to cost $120,000 a gallon. The whole tower of precision rests on every triple-point cell in every lab containing the same water, down to the isotope.

And boiling? Boiling gave up entirely

The boiling point is, if anything, worse. It’s wildly sensitive to pressure — that’s why your pasta water boils cooler in Denver and why the summit of Everest boils water at around 68°C, nowhere near hot enough to cook an egg. Water will also superheat past its boiling point if it’s clean and undisturbed. So “100°C” was only ever true at one specific atmospheric pressure, on a good day, with a following wind.

When the metrologists built the modern temperature scale — the International Temperature Scale of 1990, or ITS-90 — they looked at the boiling point of water and simply threw it out. The current scale is pinned to seventeen fixed points, and not one of them is a boiling point of water. There’s the triple point of water at 0.01°C, the melting point of gallium at 29.7646°C (handy because it melts in your hand and is stunningly reproducible), and the freezing points of a ladder of metals — indium, tin, zinc, aluminum, silver — reaching up past 960°C.

The delicious irony: on ITS-90, water doesn’t actually boil at 100°C. Measured against the modern scale, the steam point comes out around 99.974°C. The most famous number in science is off by a couple hundredths of a degree, and the official response was to stop using it.

This is the whole point

None of this makes the fourth-grade version wrong in any way that matters for boiling an egg. Zero and a hundred are perfectly good numbers for cooking, weather, and staying alive. The oversimplification is doing real work — it’s a lie we tell because the truth is not useful at the kitchen scale.

But the moment you need it to be true — really, defensibly, calibrate-a-national-standard true — the clean fact fractures into isotope ratios, dissolved gases, atmospheric pressure, supercooling, superheating, triple points, and a teaspoon of Viennese ocean water that costs a hundred and fifty-nine dollars.

That’s the door hinge again. That’s every “simple” thing I’ve ever looked at closely. The simplicity was never in the thing; it was in how far away we were standing. Walk up to it — any of it — and it turns out to be somebody’s entire career, a stack of edge cases, and a surprising bill.

Nothing is simple. It only looks that way from across the room.