We Learned to Build the Eclipse

Start with an instrument so simple it is barely an object: a hole.

By the late sixteenth century, European astronomers measured the Sun by letting it shine through a small aperture onto a screen. You cannot look at the Sun; you can look at what gets through a hole. It was the standard method, and it had a fault nobody could explain.

Solar diameters measured this way came out systematically too large. Not randomly — consistently, in one direction, against what the same astronomers got by other means. Tycho Brahe built careful apertures into his instruments and developed a correction, but the correction was empirical: a fudge that worked, with no account of why.

The missing thing was a theory of how an image forms behind a small opening. Kepler supplied it in Ad Vitellionem Paralipomena (1604) — the same book that, a few pages away, needed a new word for the region between full shadow and full light and made one.

Note the order carefully, because the whole chain repeats it. The method came first and worked badly. The understanding came second and fixed it. Nobody theorised the pinhole and then built one; they built one, got wrong answers for decades, and eventually worked out what the hole was doing to the light.

Which is a thing worth knowing about instruments in general: using one is not the same as understanding it, and the gap between those two can be measured in centuries.

The Sun has an outer atmosphere, and for most of the history of astronomy it was almost entirely unavailable for study.

The corona and the prominences are there all the time. They are simply drowned: the photosphere behind them is so overwhelmingly brighter that nothing fainter survives in the same sky. No. 51 gives the arithmetic — the corona is about as bright as a full Moon, and even a sliver of uncovered photosphere beats it by orders of magnitude.

So the only way to study the Sun's atmosphere was to wait for the Moon to cover the Sun exactly. And the terms of that arrangement are brutal:

This is why nineteenth-century solar physics is a story of expeditions — instruments crated across oceans, assembled in a field, aimed at a sky that might not cooperate, for a window shorter than it takes to boil a kettle.

And note what kind of problem this is, because it is not the obvious one. None of it was unpredictable. Eclipse tracks were computed years ahead, to the minute and to the mile — which is exactly why an expedition could be funded, crated and shipped to arrive on the right morning. Nobody was waiting for a surprise.

What the sky imposed was a structural condition: the observation existed only there, only then, and the entire cost of getting a person and their instruments to that point — the ocean, the crates, the years, the one cloud — was charged to whoever wanted the observation.

Hold onto the shape of that, because the rest of the chain is about moving the condition rather than getting luckier. The thing was always there. Reaching it was your problem.

A total eclipse crossed India on 18 August 1868, and a great many people had travelled a great distance to point spectroscopes at it.

A spectroscope spreads light into its colours and shows you the bright lines that particular atoms emit. Aim one at a prominence during totality and you get a handful of lines, each one a signature.

Among them, that day, was a yellow line sitting very close to — but not on — the familiar pair emitted by sodium.

It was seen by several people. The 1868 eclipse detections are credited to Jules Janssen and also to Georges Rayet, Captain C. T. Haig, Norman Pogson and John Herschel. This matters, and the popular version of the story loses it: a lot of instruments caught that line.

What Janssen took from the day, though, was not the yellow line at all. It was something about the brightness.

He had come to use the eclipse. He left having realised he might not need it.

This is the hinge of the whole chain, and it happened within twenty-four hours.

On 19 August 1868, the day after the eclipse, Janssen set up again — in ordinary daylight, with no Moon in the way — and successfully observed the prominence spectrum anyway.

The trick is to use a narrow slit and enough dispersion. Spreading light out dilutes the continuous glare of the sky across the whole spectrum, while an emission line stays concentrated at its own wavelength. Push it far enough and the line climbs out of the background. The eclipse had been doing that job crudely by removing the glare; the instrument could do it precisely by spreading it thin.

Sit with the timing. He waited years for the alignment, travelled to India for it, got his few minutes — and the thing the alignment actually gave him was the knowledge of what to build. He then built it immediately, and never needed the alignment again.

Every subsequent link in this chain is a repetition of that morning at a larger scale.

Meanwhile, in London, someone had been trying to do exactly this for two years without ever leaving the country.

Norman Lockyer had conceived the same idea in 1866 — that a good enough spectroscope should reach the prominences in daylight — and had then spent two years failing to obtain an instrument adequate to the job. He got there on 20 October 1868.

Both men reported to the Académie des Sciences, and the reports arrived within days of one another. The Académie, faced with two independent claims to the same method, struck a medal for the pair of them.

It is a genuinely nice piece of history, and it is nearly always told as a coincidence — two men, same idea, same season, how remarkable.

It is not much of a coincidence. It is what a well-posed problem looks like when the instruments finally catch up. Lockyer had the idea in 1866 and was blocked by hardware. Janssen got the idea from the eclipse and was not blocked at all. The idea was available; the glass was the constraint.

Both were awarded for the method. Which is worth stating precisely, because of what happens on the next spread.

Here is the version everyone knows: in 1868, during a total eclipse in India, an astronomer discovered helium in the Sun.

That sentence is wrong in almost every part, and the collection this zine belongs to exists to say so.

Neither of the reports that reached the Académie singled out the yellow line as important. The medal was for a technique — how to observe prominences without an eclipse — and the anomalous line sat in the data without anyone announcing it as a new substance.

It was Lockyer, later, who labelled it D3 and proposed that it belonged to an element unknown on Earth. And that was a further step, taken by one of the men, after the fact.

Why this is marked contested. The documents are not in dispute — the dates, the medal, the reports are all on the record. What is contested is the received account built on top of them: the tidy story in which a discovery happens at a moment, to a person, in a place.

That version is not a lie anybody told. It is what happens when a messy sequence gets compressed for retelling, and the compression is not neutral — it deletes the years of argument and leaves a moment of genius. Which is a much better anecdote and a much worse description of how the thing is done.

The same flattening is the subject of Too Good to Check, and it is the reason this collection marks its joints at all.

Lockyer and the chemist Edward Frankland proposed a name — helium, from hēlios, the Sun. For roughly a quarter of a century, a great many chemists simply declined to accept it.

The objection was not that the spectroscopy was sloppy. The objection was that nobody could put the gas on a bench.

Nineteenth-century chemistry was a science of substances you could handle: weigh them, dissolve them, heat them, combine them, hand a sample to a rival so they could fail to disprove you. That was what an element was, operationally.

And notice what makes this hard rather than stupid. The sceptics were applying a real standard, for real reasons. Spectra had already produced false alarms — the next spread is entirely about two of them. Demanding a sample was not obstinacy; it was the discipline's best available guard against exactly the failure mode that was, at that very moment, also occurring.

They were rigorous, and they were wrong. Helium was real the entire time.

What was actually being argued about was never the Sun. It was which kinds of evidence get to count — and the answer being enforced was: the kind we can hold. Everything reachable only at a distance, only by its light, only through an instrument, sat outside the category of the real until the instrument caught up.

Twenty-seven years is a long time to be told the thing you can see is not there.

The sceptics had a point, and this spread is it. The same method that found helium also found two elements that do not exist.

Eclipse spectra of the corona showed lines matching nothing known — most famously a strong green one. By the same reasoning that produced helium, these were attributed to a new element: coronium.

Spectra of glowing nebulae showed their own unmatched lines, and got their own new element: nebulium.

One in three. That is the actual hit rate of "an unmatched line means a new element," and it is why the chemists dug their heels in.

This is the collection's whole reason for existing, in miniature. The popular retelling keeps helium — a triumph, a clean discovery, light from the Sun revealing a secret — and quietly drops coronium and nebulium, which are the part that tells you how the method actually behaves.

Keeping only the successes is not a summary of the science. It is a different claim about it.

William Ramsay was not looking for helium. He was chasing argon, and he treated a uranium-bearing mineral called cleveite with acid to see what came off.

What came off, on 26 March 1895, went into a spectroscope and produced the yellow line. D3. The solar element, in a tube, in a laboratory in London, twenty-seven years after it was first seen 150 million kilometres away.

Ramsay sent a sample to Lockyer, so that the man who had named the thing and spent decades being disbelieved about it could watch it glow on a bench.

Two things are worth taking from this, and they pull against each other.

The sceptics' standard was met, and it was a good standard. A sample settled it in a way that no amount of further argument would have. The demand for evidence you can handle is not a bad instinct.

And the standard cost twenty-seven years. Not twenty-seven years of the thing being uncertain — twenty-seven years of it being real, visible, correctly identified, and refused. The evidence was sufficient in 1868 for anyone willing to count that kind of evidence.

The gap between "we cannot detect it the way we prefer" and "it is not there" is where those twenty-seven years were spent.

Coronium and nebulium were not disproved by someone failing to find them. They were dissolved by someone working out what the lines actually were.

In 1927, Ira Bowen showed that the chief nebular lines come from doubly ionised oxygen — ordinary oxygen, making transitions that are so improbable they essentially never happen in a laboratory. They are called forbidden lines, and the name is a bad joke: they are not forbidden, merely so slow that any normal density knocks the atom out of the state long before it can radiate.

In the near-vacuum of a nebula, nothing knocks it out. The atom waits, and eventually emits, and the line appears — a line you cannot make on Earth because Earth is too crowded.

Then Walter Grotrian in 1939 and Bengt Edlén in 1942 did the same for the corona: those lines are forbidden transitions of iron, ionised to a degree nobody had any terrestrial reason to consider.

And that resolution carried a much larger fact with it. Iron ionised thirteen times over means the corona is millions of degrees — vastly hotter than the surface beneath it, which remains one of the genuinely open problems in solar physics.

So the phantom elements were not simply an embarrassment to be cleared away. Chasing down what they really were produced the discovery that the Sun's atmosphere is hotter than the Sun's surface.

The error was productive because somebody took it seriously enough to explain rather than merely retract. That is the part of the method worth defending.

The chain does not run cleanly upward. Fifty years after Janssen stopped needing an eclipse, physics needed one badly.

General relativity predicted that starlight passing close to a massive body is deflected. Testing it meant photographing stars whose light grazes the Sun — and those stars are, by definition, in the daytime sky right next to the Sun.

Spectroscopy's trick does not help here. You are not trying to pick a bright emission line out of a bright background; you are trying to measure the positions of faint point sources. For that you need the sky genuinely dark, and in 1919 there was exactly one way to darken the sky next to the Sun.

So two expeditions went out for the eclipse of 29 May 1919 — Príncipe off West Africa, and Sobral in Brazil — under the direction of Frank Watson Dyson, the Astronomer Royal, with Arthur Eddington at Príncipe.

Worth keeping, because a chain that only ever ratchets forward is a story rather than a history. Capability is specific. Solving the access problem for one kind of observation left it completely unsolved for another, and the people involved had no choice but to go back to waiting for alignments.

What happened to the resulting photographic plates is the most argued-over episode in the history of this experiment.

Three sets of plates came back. One set was excluded from the published result, and that exclusion has been fought over ever since.

The Sobral astrograph plates had gone out of focus. The most likely culprit is the rapid temperature change during totality, which altered the instrument enough that the plate scale could no longer be trusted.

Those plates, taken at face value, sat closer to the Newtonian prediction. They were set aside. The published result rested on the other two sets, and it favoured Einstein.

The charge. In 1980, John Earman and Clark Glymour argued that this was data selection driven by theoretical preference — that the team kept what agreed with the theory they admired and discarded what did not. The accusation stuck hard, and for decades “Eddington saw what he wanted to see” circulated as a settled fact about one of the most celebrated experiments in physics.

Take the objection seriously, because it is not a silly one and this house has an entire piece — No. 39 — about the machinery disciplines build against precisely this failure.

And note what makes it a genuinely contested joint rather than a settled one: two accounts of the same act, both internally coherent. Rejecting a malfunctioning instrument's data is correct practice. Rejecting inconvenient data is misconduct. From outside, and at a distance of sixty years, the two can look identical.

The question is whether there was a reason to distrust those plates other than what they said.

Janssen's method reached the prominences. It did not reach the corona, which is fainter and spread across the sky around the Sun rather than concentrated in bright lines.

For that, somebody had to build the eclipse itself.

Bernard Lyot's coronagraph puts an occulting disc inside the telescope, at the focus, and blocks the Sun's image before it can scatter through the rest of the optics. The hard part is not the disc — it is everything else: stray light. Dust on a lens, imperfections in the glass, the atmosphere itself, all of it smears a little sunlight across the field and drowns the corona exactly as the sky does.

So Lyot chased scattering. Better glass, obsessively clean optics, a series of internal stops to catch light diffracted at each edge, and a site high enough — Pic du Midi, in the Pyrenees — that there was less air overhead to scatter in.

On 12 July 1931 he obtained the first photograph of the corona ever taken outside an eclipse.

Read the arithmetic of that. Before: minutes per eclipse, an eclipse every year and a half, and a boat. After: hours per clear day, from a mountain you can drive to.

Not an incremental improvement in access. A change in what kind of thing solar physics was — from an expedition science that waited on alignments to an observing programme that ran on Tuesdays.

The accusation against Eddington was answered, and the answer took two forms — one technical, one archival.

The technical answer, 1979. The original plates were re-measured with modern astrometric reduction methods. The re-analysis supported the 1919 team's treatment of the data.

The archival answer. Daniel Kennefick's work on the surviving analysis records establishes two things that the popular accusation gets wrong.

First, the decision was not Eddington's. The Sobral plates were reduced at the Royal Greenwich Observatory under Dyson, and the exclusion was his team's call. Attributing the data selection to Eddington is not historically accurate — and it is Eddington's name, not Dyson's, that the accusation has followed for forty years.

Second, there was an instrumental reason, independent of the answer. The analysis sheets show the astrograph had lost focus, and that its data could only be made to agree with Newton if the instrument had also undergone a large change of magnification. In other words: the Newtonian reading required assuming an additional malfunction. Dyson, who was known for being conservative about unquantifiable systematic errors, judged the instrument untrustworthy.

The chain marks the previous joint contested and this one documented, and the pair is doing real work. A serious accusation was made, taken seriously for decades, and then answered on the evidence.

That is not a scandal in the history of physics. It is the correction mechanism running slowly, in public, exactly as it should. The scandal would have been never asking.

The technique did not stay pointed at our own star. It became the standard method for looking at other people's.

A planet beside its star is the same problem as the corona beside the Sun, several orders of magnitude worse: the star outshines the planet enormously, and the two are separated by a hair's breadth of sky. So the answer is the same answer. Block the bright thing.

Space telescopes now carry coronagraphs as instruments in their own right — Lyot's stray-light problem, solved again at a precision he would have found absurd. And the more ambitious version separates the occulter from the telescope entirely: a starshade, a shaped screen flown in formation tens of thousands of kilometres ahead, casting a manufactured shadow onto the instrument behind it.

There is a second, quieter descendant. The transit method — by far the most productive planet-finding technique we have — works by measuring the tiny dimming when a planet passes in front of its star. It does not remove the obstruction; it reads it. The blockage is the entire signal.

So the chain ends holding both halves at once. Sometimes you manufacture the obstruction because it is the only way to see. Sometimes the obstruction is the thing you were trying to detect.

What has changed since 1868 is not that blocking became useful. It is that we stopped waiting to be granted it.

This joint is a leap and is marked as one. Nothing about coronagraphs licenses a claim about people. What crosses is a question, not a mechanism.

The question is: when the thing somebody needs exists only somewhere else, at a moment not of their choosing, and the whole cost of reaching it is charged to them — what do you do next?

Note that this is not a question about luck, and spread 3 is where the chain refuses that framing. The eclipse was predicted to the minute. What made it expensive was never uncertainty; it was that the access condition sat over there, at a fixed instant, with the travel, the freight, the years of planning and the weather risk all billed to the observer.

Which is a pattern most of us could describe without being asked. The accessible venue is across the city. The clinic that takes your insurance has a nine-month list. The building has a lift, round the back, through the kitchen. The conference that captions is in another country. The thing exists. Reaching it is your problem — and the reaching is paid in time, money, energy and paperwork that nobody else is billed for, by the people with the least of all four.

And when somebody does reach it, it demonstrates something true: look what becomes possible. Which is the eclipse's whole function in this chain — it shows you what is there when the glare is removed.

Janssen's response is the whole point. He did not write up the eclipse and begin planning for the next one eighteen months later. He worked out what the alignment had done for him, and built something that did it where he already was.

What he built was the deliberate, permanent, boring relocation of a condition — out of a sky nobody can reschedule, into an instrument on a bench. Not a favour granted when the stars line up. Not a journey somebody has to fund. A thing simply available to whoever looks.

We usually call that an accommodation. The next spread is about why that word is doing us damage.

And the second half, carried from the twenty-seven years. When something is real but only detectable at a distance, by its light, through an instrument — the answer is to build a better instrument, not to rule the thing out of existence. Helium was there in 1868. The chemists were rigorous, and they were wrong, and the cost of their rigour was paid by nobody but the truth.

Then somebody found it in a rock, and everyone agreed it had been there all along.

There are three states here, not two, and collapsing the last two is where most of the damage happens. Notice that none of them is luck — each is a different answer to the same structural question: where does the condition live, and who pays to close the distance?

Lyot did not apply for access to the corona. He did not receive an exemption entitling him to observe it on Tuesdays. He built an instrument in which the corona is available to anyone who looks through it, permanently, without anyone asking permission or explaining why they need it.

The middle row is what we have named accommodation, and Stimpunks has an entire page whose title is the criticism: Accommodations: Individualized Responses to Structural Design Problems. The problem is structural. The response is individual. That mismatch is the whole trap, and it is doing exactly what you would expect: it converts a design fault into a personal request, then rations the requests.

And the alternative is not the abolition of accommodations, which is what every bad-faith reading of that sentence wants it to mean. It is that they stop being exceptions:

Embedded so deeply that they become expected rather than exceptional is a description of a coronagraph. The occulting disc is not a concession granted to astronomers who cannot handle glare. It is where the instrument's precision comes from. Take it out and you do not get a more rigorous telescope; you get one that cannot see.

The clearest built-in condition of the last century is a slope of concrete at the end of a pavement, and its history is Janssen's next morning almost exactly.

In Berkeley in the early 1970s, disabled people needed the kerbs cut and the city had not cut them. So, as Angela Glover Blackwell tells it:

They did not wait for the alignment. They built the condition, at night, out of a bag of cement. In 1972 the city installed its first official curb cut, on Telegraph Avenue.

One correction, because this chain has already spent a spread on smoothed retellings and will not close on one. The famous version has activists smashing kerbs with sledgehammers and dynamite before dawn. That story is disability lore rather than history — repeated for decades, and not what happened. What happened was quieter, more patient, and more like engineering: somebody poured a ramp.

Blackwell's argument about what followed is the part usually quoted. Curb cuts turned out to serve people nobody had been arguing about — anyone pushing a pram, dragging luggage, wheeling a delivery, on crutches, or simply tired. She named that pattern the curb-cut effect, against a habit of mind she states plainly:

And here is the guardrail, because the curb-cut effect is very easy to misuse. It is an observation about what built-in conditions do. It is not the justification, and it must never become the price of entry. The reason to cut the kerb was that disabled people are people and the pavement was refusing them. That reason was sufficient on its own, and it did not need a dividend for everybody else to be worth acting on.

Design people call this the difference between an edge case and a stress case — a reframing from Eric Meyer and Sara Wachter-Boettcher's Design for Real Life. “Edge case” is a way of writing somebody off. A stress case is the condition that shows you where the design actually fails.

The corona is a stress case. It is the faintest thing next to the brightest thing, and an instrument built to handle it is not a specialised tool for an unusual customer. It is the instrument that turned out to work on other stars.

Two readings of the preceding seventeen spreads are available and wrong, and closing them off is worth a spread rather than a footnote.

This is not an argument against evidence. The chemists who refused helium were applying a real standard, and the same standard correctly killed coronium and nebulium. The 1919 team excluded plates from an instrument that had genuinely malfunctioned. Every rigorous act in this chain was rigorous for a reason. The argument is never for less evidence — it is for noticing when a standard has quietly become a rule about which evidence is allowed to count.

And the correction machinery works. It took sixty years for Earman and Glymour's charge to be raised and answered, and both the charge and the answer belong to the practice rather than to some failure of it. This collection exists to reclaim that practice from the things that wear its clothes, not to hand ammunition to them.