Quantum

The liquid that boils itself away

Put freshly liquefied hydrogen in a perfectly insulated tank, one that lets in no heat at all, and a fifth of it is gone in a day and two-fifths in four days. The heat that boils it comes from inside the molecules. A hydrogen molecule's two protons are identical fermions, so its rotation is tied to the way their spins are paired: three-quarters of room-temperature hydrogen is locked into odd rotational states that the cold liquid wants to leave. The molecules leave them slowly, one nuclear spin flip at a time, and each one that does releases more heat than it takes to evaporate a molecule. A liquid made the obvious way is a liquid that boils itself.

Assumes: Four states, and one of them is odd · The entropy that is still there at zero

When James Dewar first made liquid hydrogen in 1898, he made it in a vacuum flask of his own invention, and kept it as long as a vacuum flask would allow. Later workers, storing the liquid in better and better containers, noticed something odd: however good the insulation, freshly made liquid hydrogen boiled away much faster than the heat leaking in could account for, and the rate slowed over days, as if the liquid were warming itself and gradually running out of fuel. It was. The fuel was in the molecules’ rotation, held there by the exclusion principle acting not on electrons but on the protons.

No two in the same state introduced the principle as a rule for electrons that gives matter its volume. Its general form is a rule about identical particles: when two identical fermions are exchanged, the state of the whole system must change sign. A hydrogen molecule contains two identical fermions besides its electrons — its two protons — and the rule binds them too. What it binds them to is the way the molecule is allowed to rotate.

Two kinds of hydrogen

Exchanging the two protons of a hydrogen molecule does two things at once: it swaps their spins, and it swaps their positions, which for a rotating molecule is the same as turning it end for end. The spin part of the state either keeps its sign under the swap or changes it. Two spin-halves combine into three symmetric states, the triplet, and one antisymmetric state, the singlet — the count four states, and one of them is odd set out. The rotational part keeps its sign under turning end for end when the rotational quantum number JJ is even and changes it when JJ is odd. The product must change sign. So the singlet, para-hydrogen, can only rotate with even JJ, including the non-rotating J=0J = 0; the triplet, ortho-hydrogen, can only rotate with odd JJ, and its lowest state is J=1J = 1, which is still turning.

The rotational energies are kθJ(J+1)kθ J(J+1), with θ = 85.4 K for hydrogen — a large number, because hydrogen is the lightest molecule and its moment of inertia the smallest. At room temperature many rotational states are occupied, each spin arrangement takes its share in proportion to its number of spin states, and hydrogen is three parts ortho to one part para. As it cools, the molecules want to fall into the lowest state, J=0J = 0, which only para can occupy.

Cold hydrogen wants to be para; warm hydrogen is fixed at three to one. The equilibrium share of para-hydrogen — molecules whose two proton spins are paired and which may only rotate with even angular momentum — against temperature. At room temperature the triplet's threefold spin weight dominates and the share is 25.1%, the composition of ordinary hydrogen. As the gas cools the molecules want to fall into the J = 0 ground state, which only para can occupy: 51% at liquid-nitrogen temperature, 99.8% at the boiling point. Without a catalyst the gas keeps its room-temperature 25% however cold it gets.
Fig. 1 The equilibrium share of para-hydrogen against temperature. At room temperature the triplet’s threefold spin weight dominates and the share is 25.1%, the composition of ordinary hydrogen. As the gas cools the molecules want to fall into the J = 0 ground state, which only para can occupy: 51% at liquid-nitrogen temperature, 99.8% at the boiling point. Without a catalyst the gas keeps its room-temperature 25% however cold it gets.

The three-to-one is exact only when hot. At room temperature the molecule occupies rotational states up to JJ of about five, and the even and odd states hold almost equal shares of the rotational population, so the spin weights alone decide: one singlet state against three triplet states, a quarter para. As the temperature falls the even and odd sums stop being equal, because the lowest even state, J=0J = 0, lies 170 K below the lowest odd one. By liquid-nitrogen temperature the two forms are nearly equal in equilibrium; below about 30 K the even sum is all ground state and the odd sum all but empty, and the Boltzmann factor for the step, e−170/Te^{-170/T}, does the rest.

At its boiling point, 20.3 K, hydrogen in equilibrium is 99.8 per cent para. But reaching equilibrium means turning an ortho molecule into a para one, which means flipping one proton’s spin relative to the other’s, and almost nothing in a hydrogen molecule can do that. The protons’ magnetic moments are tiny, their interaction with each other weak and symmetric, and a collision between two molecules turns them about without touching their nuclear spins. Cooled without help, hydrogen keeps its room-temperature 3 : 1 composition — normal hydrogen — for days, weeks or longer. The entropy that is still there at zero traced how this frozen mixture confused the measurement of hydrogen’s heat capacity until David Dennison realised in 1927 that the two forms were not converting during the experiment, and mixing what is already mixed counted the two forms as genuinely different substances for the purposes of entropy.

Why the flip is so slow

The obstacle is a matter of symmetry before it is a matter of strength. Turning an ortho molecule into a para one means changing the symmetry of the pair of nuclear spins, from the symmetric triplet to the antisymmetric singlet. Anything that acts on both protons identically cannot do that: a uniform magnetic field, however strong, turns both spins alike and leaves their combined symmetry untouched, and so does any collision that treats the two ends of the molecule the same. What is needed is a magnetic field that differs between the two protons — a field gradient across the 0.74 ångströms that separate them — and the fields that vary appreciably over such a distance are the fields of other magnetic moments close by.

In the gas at room temperature, molecules collide with each other billions of times a second and almost never convert; a sample of pure para-hydrogen kept in a glass bulb at room temperature takes weeks or more to drift back towards three to one. That slowness is what made para-hydrogen a laboratory substance in the first place. Karl Bonhoeffer and Paul Harteck in 1929 cooled hydrogen over activated charcoal at liquid-air temperature and then at the temperature of liquid hydrogen, found the gas came out with a different thermal conductivity — the heat capacity of para-hydrogen differs from ortho’s at low temperature, and with it the rate at which the gas conducts heat — and showed that it kept that difference at room temperature for days once the charcoal was removed. The charcoal’s surface had supplied the field gradients; without it, the conversion simply stopped.

The heat the spins are holding

An ortho molecule at the boiling point is in J=1J = 1; its para equivalent is in J=0J = 0. When it finally converts, the difference, 2kθ2kθ = 170.8 K worth of energy per molecule, is released as heat into the liquid around it.

The heat the spins give up is more than it takes to boil the liquid. The heat released when hydrogen of the room-temperature composition, three parts ortho to one para, relaxes to the equilibrium composition at a given temperature, per gram, compared with the latent heat of vaporisation of the liquid, 446 J/g. At the boiling point each ortho molecule drops from J = 1 to J = 0 and gives up 2θk, and the total is 527 J/g — 118% of the heat that would boil the whole liquid. Above about 40 K the equilibrium shift is smaller and the heat falls below the latent heat; the trouble is a property of the cold.
Fig. 2 The heat released when hydrogen of the room-temperature composition relaxes to the equilibrium composition at a given temperature, per gram, compared with the latent heat of vaporisation of the liquid, 446 J/g. At the boiling point each ortho molecule drops from J = 1 to J = 0 and gives up 2θk, and the total is 527 J/g — 118% of the heat that would boil the whole liquid. Above about 40 K the equilibrium shift is smaller and the heat falls below the latent heat.

For normal hydrogen, three-quarters ortho, the total is 527 joules per gram. The latent heat of vaporisation of liquid hydrogen — the energy needed to evaporate a gram of it, the heat that changes no temperature — is 446 joules per gram. The conversion heat is larger. If all of a tank of normal liquid hydrogen converted at once, the heat released would boil the whole tank and have some left over.

That is unusual. Most of the heat a substance can release on cooling is released before it liquefies, and the liquid holds only its latent heat and a little sensible heat. Hydrogen’s conversion heat is a reservoir of energy that survives liquefaction intact, because the slow spin flip has postponed it, and it is large because θ is large: a molecule whose lowest rotational step is 170 K deep in temperature units is carrying, at 20 K, energy worth eight times its thermal energy.

A tank that boils itself

The conversion is slow but not infinitely slow. In the liquid, molecules are close together, and each ortho molecule — which carries a small rotational magnetic moment and whose nuclear spins add to one — exerts a weak, varying magnetic field on its neighbours, enough to flip a neighbour’s nuclear spin occasionally. The rate of conversion is therefore proportional to the number of ortho molecules converting and to the number catalysing them: the ortho fraction obeys dx/dt=−kx2dx/dt = -kx^2, with kk measured at about 0.0114 per hour. The heat goes into evaporating liquid, and each mole converted evaporates 1.58 moles of liquid.

A perfect tank of ordinary liquid hydrogen still boils away. The fraction of liquid hydrogen left in a tank that admits no heat at all, against time, for liquid filled with different ortho fractions. The ortho molecules convert slowly, catalysed by each other's magnetic moments at a rate proportional to the square of their fraction, and each conversion releases more heat than evaporating a molecule needs. Normal hydrogen, 75% ortho, loses 18% in the first day and 42% in a hundred hours, and would eventually lose 69% with no heat leak whatever. Liquid made 95% para loses a few per cent over the same time.
Fig. 3 The fraction of liquid hydrogen left in a tank that admits no heat at all, against time, for liquid filled with different ortho fractions. Each conversion releases more heat than evaporating a molecule needs. Normal hydrogen, 75% ortho, loses 18% in the first day and 42% in a hundred hours, and would eventually lose 69% with no heat leak whatever. Liquid made 95% para loses a few per cent over the same time.

The result is a loss of eighteen per cent of the liquid in the first day, forty-two per cent in four days, slowing as the ortho molecules that drive it become scarcer, heading for sixty-nine per cent lost in the long run — all with no heat entering the tank at all. The two numbers in the arithmetic do not depend on each other, and it is a coincidence of hydrogen’s properties that they are so close. The latent heat is set by the weak attraction between hydrogen molecules in the liquid; the conversion heat by the molecule’s rotational constant, which is set by the proton’s mass and the bond length. Had hydrogen been a little heavier, or its molecules a little stickier, the conversion heat would have been smaller than the latent heat and the tank would only partly boil itself. As it is, each gram of ortho-hydrogen that converts evaporates 1.58 grams of liquid.

The square in the rate law is what makes the loss front-loaded: with three-quarters of the molecules ortho, each finds a catalyst at almost every encounter; as they dwindle, they convert one another more and more rarely.

It is worth putting the conversion heat beside the heat that leaks in through a tank’s insulation, because that is how it showed itself. At the start, with three-quarters of the molecules ortho, the conversion releases about 4.5 joules per gram per hour: 1.25 watts for every kilogram of liquid. A well-insulated storage dewar holding a thousand litres, seventy-one kilograms of liquid hydrogen, might lose half a per cent a day to its heat leak, about two watts. The same tank filled with freshly made normal liquid generates nearly ninety watts inside itself. For the first days the insulation is almost irrelevant to the loss, and a better tank does not help.

This sets a clear rule for anyone who wants to keep liquid hydrogen for more than a few hours: convert it before it goes into the tank.

Why liquefiers convert the hydrogen before they let it into the tank. The share of a tank of liquid hydrogen lost to ortho–para conversion alone, against the ortho fraction it was filled with: over thirty days, and in the long run. Filled as normal hydrogen, a tank loses 64% in a month from this cause, before any heat leaks in through its insulation. Liquid that leaves the plant 95% para loses 2.3% in a month — small enough to bear, which is why commercial liquid hydrogen is specified at 95% para or better and why every liquefier carries a catalyst bed.
Fig. 4 The share of a tank of liquid hydrogen lost to ortho–para conversion alone, against the ortho fraction it was filled with: over thirty days, and in the long run. Filled as normal hydrogen, a tank loses 64% in a month from this cause, before any heat leaks in through its insulation. Liquid that leaves the plant 95% para loses 2.3% in a month — small enough to bear.

A tank filled with normal liquid loses nearly two-thirds of it in a month from conversion alone. A tank filled with 95 per cent para loses 2.3 per cent in the same month, comparable to or smaller than what a good insulated tank loses to heat leaks, and the commercial specification for liquid hydrogen, and for the hydrogen in the tanks of rockets that burned it in the 1960s and since, is 95 per cent para or better. Every hydrogen liquefier therefore carries a catalyst: a bed of a paramagnetic material, typically a hydrous iron oxide, whose strongly varying magnetic fields at its surface flip nuclear spins in molecules that touch it in seconds rather than days.

Where the heat is taken out decides the cost

Converting the hydrogen does not make the conversion heat go away. It only decides where it is released: in the plant, where a refrigerator can carry it away, rather than in the tank. And the cost of carrying heat away depends on the temperature at which it is picked up. A perfect refrigerator lifting heat QQ from a temperature TT to a room at 300 K needs work Q(300/T−1)Q(300/T - 1), the inverse of the ceiling on every engine, and at 20 K that is fourteen joules of work for every joule lifted.

The same heat costs least where it is warmest. The minimum work a perfect refrigerator would need to carry the ortho–para conversion heat of a gram of hydrogen out to a 300 K room, for three ways of doing the conversion. Released all at 20 K the heat costs 7.26 kJ/g, because a refrigerator at 20 K pays fourteen joules of work for every joule it lifts. Converting at liquid-nitrogen temperature first, then finishing at 20 K, costs 5.33. Converting continuously as the gas cools, so that each part of the heat is removed at the warmest temperature at which it appears, costs 2.12. Liquefiers spread catalyst through their heat exchangers for this reason.
Fig. 5 The minimum work a perfect refrigerator would need to carry the ortho–para conversion heat of a gram of hydrogen out to a 300 K room, for three ways of doing the conversion. Released all at 20 K the heat costs 7.26 kJ/g. Converting at liquid-nitrogen temperature first, then finishing at 20 K, costs 5.33. Converting continuously as the gas cools, so that each part of the heat is removed at the warmest temperature at which it appears, costs 2.12.

Releasing all the conversion heat at 20 K would cost, in an ideal refrigerator, 7.26 kilojoules of work per gram of hydrogen. Part of the conversion can happen at higher temperatures, because the equilibrium para fraction is already 51 per cent at 77 K: converting there first, in a catalyst bed cooled by liquid nitrogen, and finishing at 20 K brings the cost to 5.33. The cheapest scheme lets the hydrogen convert continuously as it cools, so that each share of the heat is removed at the warmest temperature at which equilibrium releases it, and costs 2.12 kilojoules per gram — less than a third of the worst case. Modern liquefiers come close to this by packing catalyst into the heat exchangers themselves. The same bookkeeping, in which heat is cheapest to remove where it is warmest, is why the gas that cools by being let go is used in stages rather than all at the cold end.

The same rule, in other places

The coupling between nuclear spin and rotation is not peculiar to hydrogen; it is the exclusion principle’s general form, and it acts on any molecule with identical nuclei. Deuterium, whose nuclei are bosons with spin 1, has the opposite sign rule and its own ortho and para forms with weights of six to three. In oxygen-16, whose nuclei have spin zero, the rule removes half the rotational states altogether: the molecule has no states of odd JJ at all, and its spectrum is missing every other line, the same symmetry that the label three identical quarks forced found requiring a new quantum number when three identical quarks seemed to sit in a symmetric state. The order locked into para-hydrogen can also be spent deliberately. A para-hydrogen molecule’s two nuclear spins are in a pure singlet, a state of perfect spin order that no thermal population of a magnet at room temperature comes near. Add para-hydrogen across a double bond in a chemical reaction, so that its two protons end up in a molecule where they are no longer equivalent, and that order is converted into a nuclear polarisation thousands of times larger than a magnetic resonance spectrometer’s magnet produces by itself. Parahydrogen-induced polarisation, demonstrated by Clifford Bowers and Daniel Weitekamp in the 1980s, is now used to amplify magnetic resonance signals in chemistry and in medical imaging research — the exclusion principle’s sign rule, stored at 20 K and released as signal.

And the mixture that stays mixed at absolute zero found the slow heat release from hydrogen trapped in copper — converting from ortho to para over weeks — setting the base temperature of the most sensitive refrigerators: the same postponed energy, leaking out at a rate measured in nanowatts.

What makes hydrogen the case that matters in practice is the size of θ. Heavier molecules have smaller rotational steps, their ortho and para forms differ by a small energy, and their equilibrium mixture at any accessible temperature is close to the high-temperature one. Only hydrogen, as the lightest, has a rotational step large enough to hold more energy than its own latent heat, and only hydrogen is liquefied in quantities large enough for the difference to be measured in tonnes.

What the rate law leaves out

The boil-off curves assume a perfectly insulated tank, a liquid at a uniform 20.3 K, and the single second-order rate constant of 0.0114 per hour measured for the pure liquid; real tanks leak heat as well, which adds to the loss, and conversion is faster on the walls if they are paramagnetic and in the presence of dissolved oxygen, which is a powerful catalyst. The conversion heat is computed for an isolated rigid rotor with θ = 85.4 K, neglecting the small effects of the liquid’s environment on the rotational levels. The refrigeration costs are ideal Carnot work and real liquefiers need several times as much; their ratio, not their size, is what the figure shows. And “eventually” in the long-run losses means many months, over which any real tank’s heat leak would dominate.

Still open: hydrogen for a hydrogen economy

The physics of conversion is settled; its engineering at scale is not. Proposals to ship liquid hydrogen in tankers, as natural gas is shipped, need liquefaction to cost a small fraction of the hydrogen’s energy, and today it costs about a third. Better catalysts, active at lower temperatures and with faster kinetics, and liquefier designs that convert continuously with less exchanger area are active research, as is the opposite use: warming para-hydrogen while catalysing it back towards the room-temperature mixture absorbs the conversion heat, a cold reserve that could cool superconducting cables or boil-off gas for free. How close a large plant can be brought to the continuous-conversion limit, and at what capital cost, is the question on which several national hydrogen strategies turn.

The two protons of a hydrogen molecule are identical fermions, so the molecule’s state must change sign when they swap: paired spins allow only even rotation and parallel spins only odd, and hydrogen cooled without a catalyst stays three parts ortho to one para, locked out of the J = 0 ground state that 99.8% of it should occupy at 20.3 K. Each ortho molecule that finally converts releases 2θk, 527 J/g for normal hydrogen, 118% of the latent heat, so a perfectly insulated tank of it boils away — 18% in a day, 42% in a hundred hours — which is why liquefiers convert the hydrogen first, and why doing it as the gas cools rather than at 20 K cuts the ideal work from 7.26 to 2.12 kJ/g.

Part 8 of 8

This essay is one argument about Exclusion. The others:

The objects named here

The third axis, after the field and the reading path: the things themselves, and every essay that touches each one.

The Boltzmann factorCarnot efficiencyLatent heatNuclear spinOrtho paraSpin statistics