You Cannot Correct the Sky

A star that appears to be flickering wildly is not flickering. Its output over the seconds you are watching is, for practical purposes, constant.

The light you are receiving left that star years ago — four for the nearest, thousands for most of what you can see — and travelled the entire distance without incident. It arrived at the top of the atmosphere as a flat, orderly wavefront, still perfectly steady.

Everything that happened to that light happened in the last few microseconds of a journey of years.

The word for the effect is scintillation, and its cause is not in the sky. It is in about a hundred kilometres of air, most of it in the lowest few, which the light has to cross at the very end in order to reach an eye standing at the bottom of it.

This is not a subtle correction. It is the difference between a fact about the star and a fact about the observer’s situation, and the two produce identical experiences. What you see is genuinely unsteady. Nothing you could do to the star would make it steadier, because the star was never the problem.

The atmosphere is not a uniform sheet of air. It is a shifting mess of cells at slightly different temperatures and therefore slightly different densities, and light bends when it crosses between them.

Each cell refracts the wavefront a little. The flat sheet that arrived from space is corrugated by the time it reaches the ground: parts of it converge, parts diverge, and the pattern of light and shadow that lands on your eye slides around as the cells drift and churn.

So the brightness at your particular pupil rises and falls, tens of times a second, while the total light arriving over a wide area stays constant. You are not watching the star change. You are watching a moving pattern cross your position.

The light is being redistributed, not altered. Somebody standing a few metres away is receiving your missing photons at that instant, and their star is bright while yours is dim.

The same mechanism produces the shifting bright bands on the bottom of a swimming pool, and for the same reason: an uneven medium between a steady source and a flat surface.

Here is how we know the atmosphere is doing it and the star is not. On the same night, through the same air, at the same altitude in the sky, planets barely twinkle at all.

Jupiter’s light crosses exactly the same turbulence as the star next to it. If scintillation were a property of the source, that would make no sense. It is a property of angular size.

A star, however enormous, is so far away that it is effectively a point. All its light takes essentially one path through the air, so whatever that path does to it, it does to all of it. A planet is close enough to present a disc — small, but resolvable — and different parts of that disc send their light through different turbulent cells at the same moment.

The disc is wide enough to be having several different atmospheres at once, and they cancel. The flickers average away, and the planet sits there steadily.

The astronomer’s phrasing is exact: planets’ “angular diameters are so large as to quench the scintillations.” Quench, not resist. Nothing about a planet is tougher. It is simply large enough on the sky that no single cell can speak for the whole of it.

Which is the first practical consequence of the whole zine. The flicker did not go away because the light got better. It went away because more of the medium was being sampled at once.

If the twinkling is caused entirely by the medium, then the twinkling is a measurement of the medium. And astronomers use it as one.

Watch how fast a star scintillates, how deeply, and at what spatial scale, and you can work backwards to the turbulence that produced it — how strong it is, and roughly at what altitudes it sits. Observatories run instruments that do nothing but this, so they can tell in advance what kind of night they are going to have.

The unsteadiness that ruins the image is also the only direct report on the thing ruining it.

That reframing is worth holding onto, because it inverts the obvious reading twice over. First the flicker stopped being a fact about the star. Now it stops being noise at all: it is the atmosphere’s own account of itself, written in the only handwriting available.

Nobody reads that report as a complaint about the star. It would not occur to anyone. The star’s behaviour is not what is being reported on, and the instrument pointed at the flickering is pointed, in every sense that matters, at the air.

Take one star. Watch it rise. It will scintillate violently when it is low — often flashing colours — and settle considerably by the time it is overhead.

Nothing about the star changed in those hours. What changed is the length of the path. Straight up, light crosses one thickness of atmosphere. Near the horizon it crosses many times that, at a shallow angle, through the densest and most disturbed air near the ground — so there is simply more medium to be disturbed by.

The identical source, on the identical night, is steady in one part of the sky and frantic in another. The variable is entirely how much air is in the way.

Observers act on this without discussion. You do not photograph something at ten degrees if you can wait until it is at seventy. You do not conclude that the object is unreliable at low altitude. You wait, or you go somewhere the path is shorter, and the shortest paths are why observatories sit on mountains.

Altitude, in other words, is not a property of the star. It is a property of the arrangement, and it is the single largest thing anyone can change about how well a faint thing can be seen.

Put the telescope outside the atmosphere and scintillation stops. Not improves. Stops.

Hubble does not see stars twinkle. There is nothing between it and them to do the twinkling, so the light it collects is the flat, orderly wavefront that has been travelling unmolested for however many years, arriving in exactly the state it left in.

The flicker was never in the light. Remove the medium and it is simply not there, and no adjustment to the star was required to achieve that.

This is the cleanest demonstration available that the problem was situational, and it cost several billion dollars to run. That is worth saying plainly rather than letting the moral arrive too easily: the fix that works perfectly is the most expensive object humans have put in orbit, and it works for one telescope pointed at one thing at a time.

Everyone else is still at the bottom of a hundred kilometres of moving air, and has to do something else.

If you cannot leave the atmosphere, you can measure exactly what it is doing to the wavefront and undo it, hundreds of times a second, in a mirror.

That is adaptive optics, and the loop is unglamorous and exact. A sensor reads the shape of the incoming wavefront and works out how it has been rumpled. A computer converts that into a set of corrections. A thin deformable mirror, pushed by hundreds or thousands of actuators, bends into the opposite shape. The distorted wavefront reflects off it and comes out flat. Then it does it again, before the air has time to move.

The star is not asked to be steadier. The instrument is reshaped, continuously, to match the conditions the light is actually arriving in.

It works spectacularly. Ground telescopes running adaptive optics resolve detail that was flatly impossible from the same site without it, on the same nights, looking at the same objects.

And note the direction the effort runs. Nothing is done to the source, nothing is demanded of the light, and the correction is not a one-off adjustment that stays done. It is continuous, it tracks the conditions, and the moment it stops the blur returns immediately.

Adaptive optics is not a solved problem handed out freely. It has hard constraints, and they are where this zine actually lives.

It needs a reference. The system measures the distortion against something bright enough and close enough on the sky to what you care about. If there is nothing suitable, one has to be manufactured — a laser fired into the upper atmosphere to make an artificial spot.

It works over a small patch. The correction is valid only within a limited angle of that reference, because further away the light came through different air. Outside the patch you are back where you started.

It never finishes. Deformable mirrors, wavefront sensors, real-time control, lasers, and people to run all of it — and the instant the loop opens, the blur returns within milliseconds. It was never a fix. It is a subscription.

You can correct a patch of sky. You cannot correct the sky. And the correction lasts exactly as long as somebody keeps paying for it.

So the three real options are: wait until the light is higher, spend enormously to get above the air, or spend continuously to cancel it in one small region. Every one of them is an intervention in the conditions, and every one of them has a price attached that the star does not pay.

Which is the whole finding, and it is not a comforting one. Knowing exactly where a problem lives does not make it cheap to fix. It only makes it impossible to keep blaming the wrong thing.

Somebody already made this argument about people, and made it properly. The social model of disability relocates the problem: a person is not disabled by their body but by a world built to a narrow specification.

That relocation is exactly the move on spread two. The unsteadiness was never a property of the thing being looked at. It belongs to what lies between. Say it about a staircase and it is the social model; say it about a wavefront and it is atmospheric optics.

And it was necessary. Before it, the problem was located in the person, and every proposed remedy was therefore aimed at the person — correction, therapy, effort, compliance. Moving it out was the precondition of asking any better question.

But relocating a problem is the free part. It costs nothing, it can be done in an afternoon, and it has never by itself produced a single ramp.

Everything after the relocation costs money. A lift, an interpreter, a quiet room, a support worker, a flexible schedule, a lighting scheme — these are the deformable mirror, and they have exactly its properties. They are expensive. They cover a limited area. They need somebody to keep running them. And they stop working the moment the funding stops, not gradually, but at once.

So the honest position is uncomfortable in both directions at the same time. The people who locate the problem in the person are wrong, demonstrably, and it is worth saying so every time. And being right about that has never yet built anything. A correct account of where a difficulty lives is not a remedy for it, and it should not be offered as one.

What it does do is settle the question of whose problem it is — which is not nothing, because it is the argument that has to be won before anybody is asked to pay.

A piece arguing that a correct diagnosis is not a remedy can be turned into several bad sentences without much effort. So, plainly: