Saros Cycle

The puzzle: why eclipses cluster, then return

A solar eclipse needs three coincidences at once. The Moon must be new, so it sits between us and the Sun. It must also be crossing the plane of Earth's orbit — the ecliptic — because the Moon's own orbit is tilted about 5.14°, and most new moons pass harmlessly above or below the Sun. And to decide whether the eclipse is a fat total one or a thin annular ring, the Moon must be at the right distance, since its apparent size swings by 12% over an orbit. Get all three right and a shadow touches Earth. Get any one wrong and nothing happens.

Each of these conditions runs on its own clock, and the clocks have stubbornly different lengths. The phase clock — new moon to new moon — is the synodic month, 29.53 days. The node clock — the Moon returning to the same crossing point of the ecliptic — is the draconic month, 27.21 days. The distance clock — perigee to perigee — is the anomalistic month, 27.55 days. Because these are incommensurate, eclipses look almost random from year to year. The Saros is the discovery, made by Babylonian astronomers more than two and a half thousand years ago, that the three clocks nearly chime together again after a specific interval: 18 years and about 11 days. After that interval the whole geometry repeats, and so does the eclipse.

The three lunar months and their near-commensurability

The entire phenomenon rests on a numerical accident. Take 223 synodic months and ask how many draconic and anomalistic months fit into the same span of time:

223 synodic months  = 223 × 29.530589 d = 6585.3211 d
242 draconic months = 242 × 27.212221 d = 6585.3575 d
239 anomalistic mo. = 239 × 27.554550 d = 6585.5375 d
19 eclipse years    = 19  × 346.620076 d = 6585.7814 d

All four land within about 0.46 of a day — under eleven hours — of one another over a span of more than eighteen years. That is the engine of the Saros. After 6585.32 days:

• the synodic match (within 0.04 day) guarantees the Moon is again new (or full), so an eclipse of the same kind is possible;
• the draconic match (within 0.04 day) puts the Moon back near the same node, so it again crosses the ecliptic and the shadow reaches Earth;
• the anomalistic match (within 0.22 day) returns the Moon to nearly the same distance, so the eclipse has nearly the same character — total stays total, annular stays annular.

The synodic and draconic agreements are the tight ones, which is why the existence of an eclipse repeats so reliably. The anomalistic agreement is looser, which is why the type of eclipse drifts slowly over a series.

The extra third of a day: a 120° westward jump

The Saros is not a whole number of days. The 0.3211-day remainder on the end of 6585 days is what makes the cycle so visually striking. In that extra third of a day, the Earth keeps turning under the Moon's shadow:

Δλ = 0.3211 d × 360°/d = 115.6° ≈ 120° of longitude

So an eclipse seen over, say, the central Pacific will, one Saros later, be seen roughly 120° of longitude to the west. Three Saroses later the longitude shift has wrapped almost all the way around (3 × 120° ≈ 360°) and the eclipse returns to nearly the same part of the globe. That triple-Saros interval is the exeligmos, 54 years and about 33 days, and it was prized precisely because it brings the eclipse back home. There is also a small latitude drift: each successive eclipse track creeps a little toward one pole, because the node match is not perfect.

The numbers at a glance

Quantity	Value	What it sets
Synodic month	29.530589 d	Lunar phase (new ↔ full)
Draconic (nodical) month	27.212221 d	Position relative to nodes
Anomalistic month	27.554550 d	Earth–Moon distance (total vs annular)
Eclipse (draconic) year	346.620076 d	Node alignment with the Sun
Saros (223 synodic months)	6585.3211 d = 18.03 yr	Eclipse repeats, shifted ~120° W
Inex (358 synodic months)	10571.95 d = 28.95 yr	Eclipse repeats, opposite node
Exeligmos (3 Saros)	19755.96 d = 54.09 yr	Eclipse returns to same longitude
Moon's orbital tilt	5.145°	Why most new moons miss
Node regression period	18.6 yr	Why eclipse seasons drift ~19 d/yr

Note the near-coincidence between the node-regression period (18.61 years) and the Saros (18.03 years). They are not the same number — the Saros is fixed by the synodic-draconic match, not by the nodal cycle directly — but they are close enough that one Saros carries the Moon's geometry through almost exactly one full circuit of the node alignment plus a small remainder.

Eclipse seasons and the regressing nodes

Eclipses do not happen at every new moon because the Moon is usually too far above or below the ecliptic. They are confined to eclipse seasons — windows about 34–37 days long when the Sun lies within roughly 18° of a lunar node. The line of nodes is not fixed in space: gravitational tugs from the Sun make it regress (rotate backward) once every 18.6 years, the famous nodal precession. Because the nodes are sliding to meet the Sun, the eclipse year is only 346.62 days, about 19 days shorter than the 365.26-day sidereal year. That is why eclipse seasons arrive roughly 19 days earlier each calendar year, and why pairs (or trios) of eclipses cluster about every six months.

The Saros stitches these seasons together across decades. One Saros is 38 eclipse-half-year intervals (each ≈ 173.3 days) plus a small surplus, which is exactly why the recurring eclipse lands near the same node, just a touch off — and that small offset is what eventually walks a Saros series off the Earth entirely.

The life of a Saros series

Group every eclipse that is one Saros apart and you get a Saros series, each assigned a number. A series is born when the Sun-node alignment is just barely good enough to produce a grazing partial eclipse near one of Earth's poles. Each subsequent member, 18 years later, finds the alignment a little better; the partial deepens, the shadow track marches across the latitudes, and after a dozen or so cycles the series produces its first central (total or annular) eclipse. The series then delivers a long run of central eclipses near its midpoint before the alignment degrades again and it expires in a string of partials near the opposite pole.

The arithmetic of a typical series:

Members per series:   ~69 to 87 eclipses
Series duration:      ~1226 to 1550 years
Active series at once: ~40 solar Saros series
Central-eclipse run:   tens of consecutive total/annular events

A concrete example is Saros 145, the series that produced the widely observed total solar eclipse of 11 August 1999 over Europe and the 21 August 2017 "Great American Eclipse." Its next member falls on 2 September 2035 over China and Japan, and the one after that on 12 September 2053. Saros 145 began with a partial near the north pole on 4 January 1639, will peak with its longest total eclipse around 2522, and will end on 17 April 3009 — a 1370-year lifetime containing 77 eclipses.

Saros, Inex, and the panorama of eclipse cycles

The Saros is the most famous eclipse cycle but not the only useful one. Its natural partner is the Inex, equal to 358 synodic months (≈ 28.95 years). The Inex is special because it returns an eclipse to almost the same longitude but at the opposite node, and its node match drifts in the opposite sense to the Saros. Plot every eclipse on a grid whose axes are "number of Saroses" and "number of Inexes" and the chaos of eclipse dates collapses into a clean, nearly periodic lattice — the Saros–Inex panorama introduced by George van den Bergh in 1955. Adjacent eclipses in the same Saros series sit one step apart on the Saros axis; the Inex links neighbouring series.

Cycle	Synodic months	Length	Behaviour
Semester	6	177.2 d	Next eclipse season, opposite node
Tzolkinex	88	7.1 yr	Opposite node, used by Maya
Saros	223	18.03 yr	Same node, ~120° W shift
Inex	358	28.95 yr	Opposite node, same longitude
Exeligmos	669	54.09 yr	Triple Saros, same longitude
Saros + Inex panorama	—	millennia	Organises all eclipses on a 2-D grid

Babylon, Antikythera, and Halley

The Saros is one of the oldest quantitative results in astronomy. Babylonian scribes recorded eclipse observations on clay tablets for centuries and recognised, by at least the 7th–6th century BC, that lunar eclipses recur after 223 months. They did not have the modern decomposition into synodic, draconic, and anomalistic months, but they had the period — and they used it to flag dates when an eclipse was possible, the necessary warning for a culture that read eclipses as omens.

The Greek Antikythera mechanism (c. 150–100 BC), recovered from a shipwreck in 1901 and decoded by X-ray tomography in the 2000s, contained a spiral dial of 223 cells on its back face — a mechanical Saros calendar — together with a small subsidiary dial of 3 cells for the exeligmos, used to add 0, 8, or 16 hours so the predicted eclipse time could be corrected to the local clock. It is the earliest known device that physically embodies the Saros. The word "Saros" itself was attached to the eclipse cycle by Edmond Halley in 1691, borrowing (somewhat erroneously) a term from the Greek lexicographer Suidas; the misnomer stuck.

Common misconceptions and edge cases

• "The same eclipse comes back to the same place after 18 years." No — the extra third of a day shifts it ~120° west. You must wait a full exeligmos (54 years, three Saroses) for the track to return near your longitude, and even then it has crept in latitude.
• "The Saros equals the 18.6-year nodal precession." They are close but distinct. The Saros (18.03 yr) is defined by the synodic–draconic match. The 18.6-year period is the time for the line of nodes to regress once around the sky. Their near-equality is a coincidence, not a definition.
• "Every eclipse in a series is the same type." Not necessarily. Because the anomalistic match is loose, the Moon's distance drifts across a series; many series transition from annular to total (or vice versa) as the apparent lunar diameter crosses the Sun's. Each individual eclipse is fixed, but the run can change character.
• "The Saros predicts the exact path." It predicts that an eclipse will occur and roughly where, but the precise centre line requires full numerical ephemerides — the Earth's rotation, lunar distance, and small perturbations all matter at the kilometre level.
• "Lunar and solar Saros series share numbers." They are numbered separately. Solar Saros 145 and lunar Saros 145 are unrelated families. Both kinds of series exist because the same commensurability governs new moons (solar) and full moons (lunar) at a node.