祝融说。

The Clouds over Príncipe

May 29, 1919, the island of Príncipe in the Gulf of Guinea off West Africa. A Portuguese island of cocoa plantations, colonial in its slow and humid rhythm; that day it had been raining since early morning, and the cloud lay so thick it seemed about to renege on the eclipse altogether. The English expedition was quartered at the Roca Sundy plantation on the island's northwest: Eddington and Cottingham, a few wooden crates of instruments, a dismantled astrographic telescope, and a coelostat to fold the sunlight into the tube. Eddington's task list for the journey was frighteningly short: totality lasts only minutes, and within those minutes, while the sun was entirely hidden by the moon, plates of the star field near the sun had to be taken. If the clouds did not break, two years of preparation, ten thousand kilometers of sailing, and the waiting of the Royal Society would all be written off.

Such were the stakes on the table of theory. Einstein's general relativity predicted that the sun's gravity would bend starlight passing its edge by 1.75 arcseconds; computed under the Newtonian framework from the corpuscular theory of light, the deflection should be 0.87 arcseconds — exactly half the Einstein value. Neither number is large, but they are two different worlds: if the star images on the plates had shifted by 1.75 arcseconds, what was rewritten would be not one formula but the framework of space and time that humanity had been running for more than two hundred years. The total eclipse was the only occasion for this verdict — only when the moon wholly veils the sun's glare can the stars near the sun be photographed; and the total eclipse of May 29, 1919, had totality lasting several minutes and fell just as the sun stood in front of the Hyades, that dense field of bright stars — a conjunction arranged, one would say, expressly for the reconciling of accounts. The Astronomer Royal Dyson organized two expeditions: Eddington to Príncipe in West Africa, Crommelin and Davidson to Sobral in Brazil.

As totality drew near, the clouds in fact split open a seam. Sixteen plates were exposed in turn; the great majority fogged in the cloud. The hours that followed are the quietest passage of this story: Eddington stood outside the improvised darkroom, waiting until the plates could be developed. A plate was developing in the solution, and upon it — if the clouds allowed — was written one of two numbers. Chapter Three foreshadowed this night; this chapter unfolds it from end to end; for May 1919 is only the last act of the play — the first act was in Cleveland, thirty-two years before.

A Null Result Thirty-Two Years Long

July 1887, Cleveland, in the United States, the Case School of Applied Science. Albert Michelson and Edward Morley had tuned an interferometer to the limit of the engineering precision of the day: a beam of light made its back-and-forth passage along arms some eleven meters long, and the whole instrument floated on a trough of mercury so that it could be turned smoothly around. What was to be tested was the medium in which light travels — the ether. On the picture of the time, light is a wave in the ether, and the Earth, circling the sun, must be plowing through this "sea of ether"; the speed of light downstream and across should differ measurably, and as the instrument was slowly turned, the interference fringes of the two light paths ought to shift accordingly. Reckoned from the speed of the Earth's orbital motion, the expected fringe shift was about 0.4 of a fringe.

The instrument was turned a full circle, and another. The fringes did not move — the measured shift was less than one-twentieth of the expected, "probably less than one-fortieth" in the experimenters' own conclusion, indistinguishable from zero. The drift of the Earth through the ether could not be measured.

On the textbook telling, this blow struck the heart of classical physics directly. The actual history is blunter, and more instructive. Almost no one at the time felt the edifice tottering: ether theory had elasticity enough to digest a null result, and Michelson himself was more inclined to read the experiment as support for the hypothesis of ether drag; the result lay quiet in the literature for years; Morley and Miller later redid it at higher precision, the null result was reproduced again and again, and again and again quietly filed. RC's General Outline says in Section 1.3, discussing space, time, and causality, that observational consensus "lets coherent subjects share a broadly consistent reality, with minor observational divergences ignored". The null result of Michelson–Morley was precisely such an ignored divergence: re-verification could not smooth it away — it was no matter of one laboratory's craftsmanship; the more precise the instrument, the cleaner the zero — and yet the framework was able, for a time, to decline to book it. In RC's reading this is the first mode of existence of an observational divergence: the suspended account. The divergence has occurred, cross-subject re-verification has even nailed it down, but so long as the framework can still issue cash (the cash of the next section), the consensus can go on pretending the books are balanced.

The last chapter said the prediction machine's IOUs come in two kinds: the 43 arcseconds are a debt one can bargain over, while the null result removed even the party to bargain with — the expected 0.4 fringe depended on the question "does the Earth drift through the ether", and the null result amounted to saying: this question cannot be made to yield anything. For thirty-two years the divergence shouted at a debtor who did not exist.

The Patch of Contraction

The debt of the null result was paid for more than a decade in the currency of the old framework. In 1889 the Irish physicist FitzGerald proposed a conjecture; in 1892 Lorentz in Holland independently proposed the same: a body moving through the ether contracts along the direction of motion by a factor exactly canceling the difference in optical path. If Michelson and Morley's instrument would show no shift of fringes as it turned, that was because the arm that had turned into the direction of motion had shortened by just the extra path the light should have traveled. The conjecture later came to be called the Lorentz–FitzGerald contraction — mathematically seamless, in character a saving hypothesis through and through: the amount of contraction is exactly equal to the amount needed to cancel the observation, not a fraction more or less, and is in principle not independently measurable.

The patches then multiplied layer upon layer. To keep Maxwell's electromagnetic equations of the same form in moving systems, Lorentz introduced "local time"; between 1899 and 1904 he gave the whole set of transformations later named for him; Poincaré polished the mathematics finer, and in 1904 pointed out that the theory satisfied a "principle of relativity". On the eve of 1905, the mathematical skeleton of this patched electrodynamics was nearly identical, letter for letter, with the special relativity about to appear — with one thing lacking: the ether was still there. The absolute frame of rest was still on the register; the contraction was still a "real" physical shortening; local time was still auxiliary bookkeeping invented to save the equations.

Of this condition RC's General Outline has, in Section 3.4, discussing dynamic evolution, a sentence that might have been copied out of the history of physics: "In the early stages of divergence, formal stability at the level of surface institutions can usually be maintained, while the deeper rules undergo continuous adaptive adjustment." The ether was the surface institution: the textbooks went on being written, the experiments went on being done, the professors went on being appointed, and no one proposed abolishing it at once; the deep rules — length, the simultaneity of time, the transformation properties of the equations — were being rewritten one by one in the dark. Chapter One said that a framework survives by inertia, and that inertia can be measured: the observational divergence required to rewrite a surface consensus is of an order far greater than the freedom required to rewrite the deep rules. The community accordingly sank the changes below the waterline: mathematics by the new recipe, ontology under the old signboard. This is not hypocrisy; it is the standard posture of any snapshot's self-maintenance — two thousand years before, Ptolemaic astronomy had played it through once: epicycles to the dozenth tier, and the register of the crystal spheres not one word changed.

But the account of the patches had to come into the light. In the notation of the last chapter: the contraction hypothesis was pure paper money — it staked nothing new, and was printed only to square the old books; and every new note quietly diluted the credit of the whole account. By 1904 the sharp-eyed could already see it: the mathematics had changed its skeleton outright, while the ontological register had not moved a word — the gap between surface and depth was stretched to its widest.

Two Rewritings

A page of historical honesty should be entered here in passing: Einstein himself later said that when he proposed special relativity in 1905 he did not know the details of the Michelson–Morley experiment; what inspired him was the assorted asymmetries within electromagnetic theory and mechanics, not the hammer blow of any one experiment. The null result was less the midwife of the new theory than the first witness it generously received after its birth. This only strengthens this chapter's thesis: the suspended account of an observational divergence can wait eighteen years for a theory — the divergence is nailed to the books first, and the framework that re-locks it arrives afterward; order is not always the order of causation.

The man of the liquidation was not in Holland but in the patent office at Bern. In June 1905 Einstein completed "On the Electrodynamics of Moving Bodies". The paper's handling of the ether was not repair but removal from the register: from two postulates — the laws of physics are equivalent in all inertial frames; the speed of light in vacuum is independent of the motion of its source — he derived directly that lengths contract and clocks slow in moving systems, and that simultaneity varies with the frame of reference. The Lorentz transformations were re-derived by him, but their standing changed utterly: in Lorentz they were the effect of contraction; in Einstein they are the grammar of space and time itself. The ether on the register thenceforth held no office: a debt suspended for eighteen years was not repaid but written off — the debtor does not exist; the debt does not exist. The null result passed from anomaly to matter of course: there is no drift because there is no sea to drift in.

The first rewriting concerned motion and clocks. The second and deeper rewriting had to wait another ten years, and concerned gravitation itself. In 1907, in the patent office, Einstein had the thought he himself called the happiest of his life: a person in free fall does not feel his own weight — gravitation and acceleration are locally indistinguishable. Setting out from this, in 1911 he used the equivalence principle to compute a deflection of sunlight grazing the sun of about 0.87 arcseconds — the same order as the Newtonian value. General relativity took shape in Berlin in November 1915, after eight years of mathematical struggle (Part Three told how Grossmann put the dictionary of Riemannian geometry into his hands during those years) and with the curvature of space-time fully reckoned in, and the predicted value doubled to 1.75 arcseconds: gravitation is not a force but the curvature that matter and energy impose on space-time; the planets are not pulled along their courses but walk the "straightest" paths in a curved geometry. This verdict might have been staged five years earlier: in 1914 Freundlich had already led a party to the Crimea to photograph that year's eclipse; war broke out, the party was interned, and the observation was abandoned — the world's politics postponed the accounting by five years. And this made 1919 the first full examination of general relativity in its completed form: the three questions it had itself staked — Mercury, light, and (not yet arrived) the gravitational redshift of spectral lines — and it handed in the second paper with an eclipse five years late.

In the very month the theory took shape, the first old debt was crossed out. On November 18, Einstein reported to the Prussian Academy: the residual precession of Mercury's perihelion of 43 arcseconds per century is delivered by general relativity without the adjustment of any parameter. He wrote afterward in a letter that when he saw the theory agree with that number, which had hung for more than half a century, his heart raced for days on end. The nature of this scene deserves weighing in this book's ledger: the last chapter told of the old framework's predicament, unable to issue cash and printing only paper; general relativity's handling of Mercury is exactly the reverse — it adjusted not one parameter for Mercury's sake; the 43 arcseconds flowed naturally out of a framework built for an altogether different problem. In other words, the first thing the new framework did was to take the most famous bad debt on the old account over into its own name, principal and interest. This is not the victory of a patch; it is the victory of changing the ledger.

On RC's reading, the core verdict of this chapter can now be stood up: general relativity is a re-convergence, and a re-convergence of the deepest layer — what it re-locked was not some law but the space-time snapshot itself. RC's General Outline says in Section 1.3, discussing space, time, and causality: "the continuity of space and time is the result of different consciousnesses processing dynamic existence in coordination, in a specific manner; this continuity is not an intrinsic property of physical reality but a secondary construction of the consciousness's pattern recognition." Chapter One used this sentence to plant a line: the framework of time and space, which looks like the precondition of all experience, is itself the secondary construction first locked in, and therefore least noticed. The Newtonian space-time of two hundred years is the deepest and thickest layer of the snapshot — absolute space, time flowing evenly — an account thick enough that Kant proclaimed the two the a priori forms of reason (Chapter Two). What general relativity did was put this layer up for repricing: space-time is no longer the stage of observation but has entered the inventory of observation — it curves, its curvature is measurable, its geometry and matter are each other's accounts. Re-convergence is not repair — repair keeps the framework and adjusts parameters; re-convergence re-locks the snapshot itself. And when the snapshot of space-time is re-locked, what becomes of the two hundred years of observations piled upon it? Chapter Three has already given the answer: they continue in office as limiting cases. At low velocities and in weak fields the curved geometry returns to the flat, and every stub of the Newtonian framework remains valid as before. What changed is the ledger, not a burning of the deposits.

The Ritual of Re-observation

So back to the darkroom on Príncipe. The usable plates were developed — of the sixteen, only two were decent; the rest had fogged in the clouds. Eddington made preliminary measurements on the island and completed the reduction back in England; at the other end of the same belt of totality, Sobral's two telescopes fared better: the plates of the four-inch telescope were clear, while those of the astrographic telescope, out of focus, were given a suspended sentence. On November 6, 1919, the Royal Society and the Royal Astronomical Society met jointly in London. Dyson reported the two sets of results: Sobral, 1.98 arcseconds with an error of about 0.12; Príncipe, 1.61 arcseconds with an error of about 0.30 — falling close to Einstein's predicted 1.75 and far from Newton's 0.87. The conclusion: starlight is deflected as general relativity requires. The president of the Royal Society, Thomson, chaired the session, and remarked afterward that this was the most important result for the theory of gravitation since Newton. The next day, November 7, 1919, the front page of the London Times carried a three-line headline: "Revolution in Science — New Theory of the Universe — Newtonian Ideas Overthrown". Overnight, Einstein became a name known to the world.

The thesis of this chapter is: that meeting room in November 1919 held a ritual — the re-observation ritual of re-convergence. Why a ritual should be necessary must first be explained. On the naive imagination, the rightness or wrongness of a theory is settled by experiment, so what need for ceremony? But Chapter Twelve said that observation is never the rightness or wrongness of a private event but the bookkeeping of a joint account: one plate says 1.61 arcseconds — by what warrant should that make the physicists of the whole world change their books? Not by the plate itself, but by the community's public re-examination of the whole process of observation — the expedition was officially organized, the apparatus and methods were published beforehand, and the results were brought back to be audited on the spot at a joint session. This is precisely the charter of Chapter Twelve (results public, method reproducible) executed at the highest specification: re-convergence is not the private epiphany of a genius but a public changing of accounts that requires a ritual to complete it — without the ritual of re-observation, the new snapshot cannot be locked. RC's General Outline gave in Section 1.4, discussing the unity of subject and object, the full procedure of re-convergence: the subjects of the divergence "re-intervene in the Ground of Possibility to achieve re-convergence, completing an innovation of observational consensus, which finally brings an innovation of the rule framework". Science has made this procedure into a public ceremony: divergence (the null result and the 43 arcseconds) accumulates for years; the new framework (general relativity) re-locks the snapshot; and then — this is the ritual's crucial link — the community organizes a public, cross-subject re-observation, so that the new consensus closes on the spot. Two expeditions, two continents, two independent sets of instruments; the results were not announced on the spot but brought back and declared at the joint session; the plates, the reductions, the errors, all entered the public record for anyone to re-examine. A divergence thirty-two years in suspense (sixty, if the account is dated from Mercury's 43 arcseconds) completed its changing of accounts in a single meeting. The newspaper headlines read the direction wrong — Chapter Three has already given the reading, and here it need only be reaffirmed: Newton was not overturned. What was rewritten is the path of light in a gravitational field and the geometry of space-time, while every stub of the Newtonian framework at low velocities and weak fields remains valid, continuing in office as the limiting case on the first page of the new ledger. Local surgery, the ledger reopened, the world not falling apart. Incidentally, the ritual itself must also submit to re-observation: later re-examinations noted that the treatment of the 1919 errors has contestable points, and the eclipse expeditions of 1922 and the radio observations that followed nailed the conclusion down ever harder. The closing of a consensus is not one night's ritual but the decades of continuous re-emphasis after the ritual — and this is precisely the meaning of consensus reinforcement (A6): the loop sets no terminus.

The core drama of Part Four can now be fixed as three acts. Act one, divergence: an observational divergence that cross-subject re-verification cannot smooth away hangs on the books — the null result of Cleveland, the 43 arcseconds of Mercury. Act two, the patch: formal stability at the surface, adjustment of the deep rules — Lorentz's contraction, the vacant office on the ether's register. Act three, re-convergence and the ritual: the new framework re-locks the snapshot, the community re-observes in public, and the consensus closes quickly — general relativity and the plates of 1919. The dynamics of General Outline 1.4 turns, from a passage of philosophical prose, into a stretch of history with dates, with names, and with error bars. Chapter Three said that the navigator's chart is useful precisely because it stands ready to be corrected by today's measurements; these three acts are one complete re-charting by the navigator — measure the deviation (act one), patch in pencil for the time being (act two), and finally redraw the chart and convene the whole crew to re-calibrate (act three). Fallibility is not the disgrace of this way of working; it is its motor: precisely because the ledger stands ready to be overruled by the world, thirty-two years of overruling were at last exchanged for one deeper new account. The first rewriting of "necessity" that the introduction promised is hereby paid: what was rewritten is the necessity of space and time. But the same Einstein will lead a second and deeper rewriting still — not in the heavens, but inside the atom; not about space and time, but about the individual itself. Four years before the plates of Príncipe were developed, an experimental phenomenon had already torn a crack in the "necessity of the individual": no one can predict when a single atom will decay, and yet the decay curve of a gram of radium is as smooth as clockwork.