The Saros cycle, and two eclipses that changed physics

Discover named the shadows. Understand measured the geometry. This layer asks the harder questions: why do eclipses repeat in a predictable cycle, how did astronomers sixteen centuries ago compute them without a satellite in sight, and what has a shadow race across a desert ever proved about the universe?

Three threads run through this layer: the Saros cycle, a genuinely ancient pattern that still governs eclipse prediction software today; the history of getting it right, from Babylon through Aryabhata to two expeditions in the twentieth century; and the discipline of testing a belief against evidence, using eclipse superstition as the example.

Understand quoted a separation limit of about 1.475° without building it up piece by piece. Do that properly now, because one of its four terms — parallax — turns out to matter well beyond eclipses.

Parallax is the apparent shift of a nearby object against a distant background when you change your viewpoint. Hold a finger at arm's length, close one eye then the other, and it jumps against the wall behind it. The nearer the finger, the bigger the jump.

The Moon is close enough that its position in the sky, measured from two different places on Earth at once, visibly differs — by up to about 0.95°, called its horizontal parallax. The Sun is so much further away that its own parallax is only about 8.78 arcseconds (0.00244°) — visible only with careful instruments, and historically one of the hardest numbers in astronomy to pin down. Chapter 15 of Extend explains how Venus transits were used to measure exactly this.

Two edge cases round out the geometry from Understand: a solar eclipse that changes character partway along its own path, and the slow, steady drift of the nodes themselves that makes every eclipse season arrive a little earlier each year.

The Moon has more than one kind of "month", because there is more than one thing to return to.

• The synodic month (29.5306 days) is new moon to new moon — the phase cycle.
• The draconic month (27.2122 days) is node to node — how long the Moon takes to return to the same crossing point of its tilted orbit.
• The anomalistic month (27.5546 days) is perigee to perigee — how long it takes to return to the same distance from Earth.

For an eclipse to repeat almost exactly — same phase, same node, same distance, so the same kind of eclipse at nearly the same latitude — you need a stretch of time that is a whole number of all three months at once. Nobody designed this to happen. It very nearly does anyway, and the length of time it takes is called the Saros.

Discover told the story: Aryabhata said eclipses were shadows, not a demon. This chapter asks the harder question — what calculation actually let him say when the next one would happen?

Indian astronomy of this period worked with a geocentric model (Earth at the centre, as almost everyone did before the 16th century) but treated the Sun, Moon and planets' motions with real mathematical rigour. The tools were tables of sines (called jya, the ancestor of the word "sine" itself, via Arabic and Latin translation), mean and corrected longitudes, and the same node geometry used throughout this topic.

Most of history's eclipses were watched, recorded and then left alone. Two were used as instruments — as the only available piece of scientific apparatus capable of testing an idea that could be tested no other way at the time.

18 August 1868, Guntur, India. During totality, the French astronomer Jules Janssen pointed a spectroscope — an instrument that splits light into its component colours — at the pearly loops of gas around the eclipsed Sun (prominences). He found a bright yellow line that did not match sodium or any other known element's signature.

Janssen realised something else: prominences are bright enough in that one colour that you do not need an eclipse to see them at all — you can filter for exactly that wavelength on any clear day. He designed a way to do this the very next day, without waiting for another eclipse. Independently, the English astronomer Norman Lockyer worked out the same trick from London a few months later.

The line sat at 587.6 nanometres, close to but distinctly different from sodium's well-known lines. Nobody could find a matching element on Earth. It was named helium, after helios, the Greek word for the Sun — a genuinely new element, identified in space a full 27 years before it was first isolated on Earth, in 1895, from a uranium mineral.

29 May 1919, Príncipe (West Africa) and Sobral (Brazil). Four years earlier, Einstein's general theory of relativity had predicted that gravity bends the path of light — and specifically, that starlight grazing the Sun's edge should be deflected by about 1.75 arcseconds, almost exactly double the value Newtonian gravity alone would predict for a particle of light, about 0.87 arcseconds.

The only way to test this in 1919 was to photograph stars that appeared very close to the Sun in the sky and see whether their positions shifted compared with their normal positions months later, when the Sun was elsewhere. That measurement is completely impossible in ordinary daylight, because the Sun's glare drowns out every nearby star. During totality, for a few minutes, the sky darkens enough for those faint stars near the Sun's limb to be photographed at all.

Expeditions led by Arthur Eddington (Príncipe) and Andrew Crommelin (Sobral) did exactly this. The measured deflection matched Einstein's predicted 1.75 arcseconds far better than Newton's 0.87 arcseconds. The results, announced in November 1919, made Einstein an overnight global celebrity and gave general relativity its first major experimental support.

Discover mentioned that some traditional eclipse customs are not supported by evidence. This chapter is about how you would actually check that, because the method matters more than the specific answer, and it transfers to every other claim you will ever be asked to evaluate.