The band moves and the orbit does not

The first rung of this anchor treated the habitable zone the way every survey paper treats it: as a band of stellar flux, therefore a band of distance, fixed by the star’s luminosity and scaling as its square root. That is correct at an instant, and an instant is not what habitability is about. A star is not a fixed lamp. It brightens as it burns, by tens of per cent over billions of years, and the band brightens outward with it.

The consequence is that the zone sweeps across the orbits rather than containing them. A planet is not in the habitable zone; it is in the habitable zone for a while, and the while is what matters.

That last number is the one to argue with, and this essay does argue with it. It is worth stating first, because a figure whose most striking claim is also its least reliable one should say so before it says anything else.

Why the band moves at all #

The brightening is not a subtlety and it is not optional. A main-sequence star is in hydrostatic equilibrium with its own nuclear furnace, and the furnace converts four hydrogen nuclei into one helium nucleus, which removes three particles from the core for every helium made. Fewer particles at the same pressure means the core must be hotter and denser, and both raise the reaction rate. In homology terms the luminosity carries the fourth power of the mean molecular weight, and the mean molecular weight of a fully ionised gas rises from 0.62 to 1.33 as hydrogen is exhausted.

The Sun’s own increase is about forty per cent since it arrived on the main sequence, and it will roughly double again by the time it leaves. The figures here use the standard solar law, written in fractional main-sequence age so that it can be applied to any star: luminosity rises by a factor of 1.9 from arrival to departure, and both zone edges follow the square root of that, which is 1.63.

Applying the solar shape to every star is the one assumption in these figures that is not derived, and it deserves naming. What is solidly established is that a star brightens by roughly a factor of two across its main sequence, for the reason above, and that the main sequence is enormously longer for a small star. The detailed shape of the curve does differ — a fully convective M dwarf mixes fresh hydrogen into its core and therefore brightens more slowly for longer, while a star with a convective core brightens faster near the end — and neither departure changes anything this essay concludes, because every conclusion here is a comparison of durations rather than of luminosities.

The durations themselves come from the crudest possible argument and are right to within a factor. A star’s fuel is proportional to its mass and its consumption is its luminosity, so the main-sequence lifetime goes as $M/L$; and since mass decides everything by a power of about four over the relevant range, the lifetime goes as $M^{-3}$. Ten billion years for the Sun, four and a half for a star half again as heavy, a hundred and forty for one at four tenths of a solar mass. Three orders of magnitude in lifetime across a factor of four in mass is what makes the rest of this essay possible.

The practical form of that asymmetry is a piece of advice a catalogue cannot give. Among two planets equally “in the habitable zone” of the same star, the outer one has by far the longer remaining tenure, and the difference is measured in billions of years rather than in per cent.

A small star hardly moves its band #

The sweep factor is the same 1.63 for every star, because the brightening law is the same. What differs, by three orders of magnitude, is how long the sweep takes.

This is the strongest argument in the subject for looking at small stars, and it is not the argument usually given. The usual one is observational: an M dwarf is small and dim, so a planet crossing it blocks more of its light and tugs it harder, and its habitable zone is close in, so the orbit is short and the survey does not have to wait. Those are reasons the planets are easy to find. This is a reason they are worth finding — a rocky planet around a red dwarf has a habitable tenure longer than the current age of the universe, and no planet around a solar-type star has anything like it.

It also sits directly against the difficulty the first rung ended on. Below about half a solar mass the zone lies inside the tidal-locking radius, so the same planets that enjoy a nearly permanent zone have one hemisphere in permanent day. The two arguments point in opposite directions, apply to exactly the same objects, and are both about time scales — one about how slowly the star evolves, the other about how quickly the planet’s spin is dissipated.

And a large one sweeps it past everything #

The collapse is not gradual. A star fifty per cent heavier than the Sun burns through its hydrogen more than five times faster, because luminosity rises as roughly the fourth power of mass while the fuel supply rises only as the first. The zone still sweeps by 1.63; it simply does it in four billion years rather than ten, and the band that survives the sweep is the sliver of overlap left over.

The requirement, as a function of mass #

Putting the two extremes on one axis gives the figure this essay exists for.

A cut-off in stellar mass is a much stronger statement than a narrowing of a band, and it comes out of nothing but the ratio of two timescales. Four billion years is roughly what the Earth took to produce an atmosphere anything could see from outside, which is the only reason to pick it; nothing in the physics prefers it.

The idea has a history worth knowing, because its first version was far more pessimistic than this one and was wrong for an interesting reason. Michael Hart’s calculations at the end of the 1970s coupled a climate model to the brightening Sun and concluded that the continuously habitable zone was about 0.95 to 1.01 AU — six per cent wide, with the Earth barely inside it and any planet a few per cent further out permanently glaciated and any a few per cent closer in a runaway greenhouse. That result stood for a decade and is the origin of the idea that habitable planets are vanishingly rare. What it lacked was the carbonate–silicate thermostat, which had not yet been described: a planet that starts to freeze accumulates volcanic carbon dioxide and warms itself back, and adding that feedback widened the outer edge by more than a factor of two at a stroke. The lesson is not that Hart was careless. It is that the width of this band is set by feedbacks in a planet’s atmosphere rather than by anything about the star, and that the star is the easy half of the calculation.

How long is long enough #

Since that duration is a choice rather than a measurement, the honest thing is to show what the choice does.

The terrestrial evidence for that shorter number is real and is not decisive: isotopically light carbon in Greenland sediments and putative microfossils in Australian cherts put life on Earth within a few hundred million years of the end of the heavy bombardment. One planet is one planet, and it may have been fast or slow.

That is the useful separation. The cut-off near 1.35 solar masses is fixed by the main-sequence lifetime and moves by one per cent between a one-billion-year and an eight-billion-year requirement. The fraction of the band retained by a solar-type star moves from 81 per cent to 36 per cent over the same range, and any statement about “how many habitable planets there are” that depends on that fraction is a statement about the requirement chosen.

The two edges are model outputs #

Everything above has treated the inner and outer edges as though they were known. They are outputs of one-dimensional climate models, and they disagree with each other by more than the effects this essay has been computing.

Both edges are thresholds on a runaway rather than temperatures. The inner one is the point at which water vapour’s own greenhouse becomes self-amplifying: a warmer surface evaporates more water, water vapour is a strong absorber in the infrared, and the extra absorption warms the surface further. Below a critical flux the loop converges on a warmer but finite temperature; above it there is no equilibrium at all until the entire ocean is in the atmosphere and the surface is at more than a thousand kelvin. What fixes the threshold is that a saturated atmosphere’s outgoing infrared flux approaches a ceiling — about 280 watts per square metre — no matter how hot the surface gets, because the emitting layer stays at the top of the water cloud. When the absorbed sunlight exceeds that ceiling, nothing can balance it. That ceiling is a real and well-founded number; where it lands in orbital distance is not, because the absorbed sunlight depends on the albedo and the albedo depends on the clouds.

The inner edge is the weak one. Its conservative value, 0.99 AU for the Sun, comes from a model in which the atmosphere is a column, clouds are prescribed rather than computed, and the planet does not rotate. Three-dimensional models with clouds that form where the circulation puts them give 0.95 AU or closer, because a cloud deck on the day side reflects sunlight and delays the runaway. The difference between 0.99 and 0.95 AU is four per cent in distance — and it is the difference between the Earth leaving the habitable zone in 150 million years and leaving it in about a billion.

A figure whose headline number changes by a factor of six under a four per cent change in an input is reporting the model’s precision, not the Earth’s. That is why the first figure’s crossing was flagged in the opening paragraphs rather than defended. What survives the ambiguity is everything the essay does with ratios: the sweep factor, the mass cut-off, the comparison between stars. Those depend on the edges moving together, and they do.

The faint young Sun, and the asymmetry it exposes #

There is a check available at the other end of the Sun’s history, and the model fails it in an instructive way.

Four billion years ago the Sun was about 75 per cent as bright as it is now. Run the Earth’s present atmosphere at that flux and the oceans freeze, and stay frozen until roughly two billion years ago. The geological record has liquid water, and sedimentary rocks that require it, from at least 3.8 billion years ago onwards. The discrepancy is old — it is Sagan and Mullen’s, from 1972 — and the resolution is that the atmosphere was not the present one: more carbon dioxide, probably some methane, and a greenhouse strong enough to make up the difference.

It is worth noticing how large the discrepancy is in the units this essay has been using. Seventy-five per cent of the present luminosity puts the inner edge at 0.86 AU and the outer at 1.48, so the Earth at 1 AU was comfortably inside the band four billion years ago by the model’s own reckoning — and the model nonetheless freezes it, because the band’s outer half is defined by a greenhouse the Earth is not given. The failure is not in the zone’s edges; it is that a planet’s position in the band says nothing about its temperature without an atmosphere attached.

That resolution is exactly what the outer edge of the habitable zone already assumes. The maximum-greenhouse limit is the distance at which a planet can no longer keep itself warm even with as much CO₂ as the carbonate–silicate cycle can supply, and the cycle is a thermostat: a colder planet weathers rock more slowly, so volcanic CO₂ accumulates and the greenhouse strengthens. The outer edge is defined by giving the planet full credit for that feedback.

The inner edge gives it none. It is computed for a planet whose atmosphere is fixed, and a real planet approaching a runaway greenhouse would first draw down its CO₂ almost completely, which the same thermostat does automatically. The two edges of one band are computed under opposite assumptions about the planet’s ability to regulate itself, and the asymmetry is not accidental — the outer edge has a well-understood feedback available and the inner one has an ill-understood cloud response. It does mean the band is not one object, and comparing its width across stars, as every figure above does, quietly assumes the asymmetry is the same for all of them.

The generalisation #

The structure worth extracting is that a condition evaluated at an instant and the same condition required to persist are different constraints, and the second is not a small correction to the first.

Here the instantaneous zone is 0.72 AU wide for the Sun and the four-billion-year zone is 0.48 AU wide, and above a certain stellar mass the second is empty while the first is at its widest. The same shape recurs whenever a system drifts through a criterion. A resonance that captures does so because a slow drift crosses it in the right direction, not because a body happens to be at the resonant period. A misalignment only cool stars forget is a statement about a tidal timescale beating a stellar lifetime rather than about an angle.

There is a direct consequence for the one number the field most wants, which is the fraction of stars with a rocky planet in the habitable zone. That fraction is estimated by dividing detections by a computed completeness, and the habitable-zone boundary used in the numerator is invariably the instantaneous one — the band evaluated at the star’s present luminosity, because that is what the observation constrains. The part of such a rate that is a definition is larger here than almost anywhere else in the subject: two groups adopting the conservative and the optimistic edges differ by a factor of two before any planet is counted, and neither is applying a persistence requirement at all.

The corollary is a warning about catalogues. A list of habitable-zone planets is a list evaluated at one instant — the instant of observation — and the quantity anybody actually cares about is an integral over time that the list does not contain and cannot be corrected for without knowing each star’s age. Stellar ages are the least well determined property of a field star, routinely uncertain by a factor of two, which is the same factor the answer here turns on.

Where the ladder goes next #

The next rung takes the outer edge seriously as a piece of physics rather than as a boundary. The carbonate–silicate thermostat has a response time of half a million years, a gain that depends on how much land there is to weather, and a limit set by the CO₂ pressure at which the gas begins to condense — and each of those is a number, computable, that fixes how far out the edge can be pushed.

Further rungs on this anchor: the habitable zone of a binary, which is two overlapping bands and a stability criterion; the effect of eccentricity, where the orbit-averaged flux is not the flux at the mean distance and a planet can be inside the zone on average and outside it every perihelion; the pre-main-sequence phase, in which an M dwarf is a hundred times brighter than it will settle at and a planet in its eventual zone spends its first hundred million years boiling; and the observational question of what could actually be measured about such a planet, which is a question about the transit depth of an ozone line.