Eclipsing Binary

The geometry that makes eclipses possible

A binary star is two stars orbiting their common centre of mass. The orbital plane has some inclination i relative to the observer's line of sight: i = 0 means we look down on the orbit from above, i = 90° means we look at it edge-on. For an eclipse to occur at all, the line of sight must pass through the volume swept by the two stars — equivalently, the impact parameter b = a cos(i) / R₁ (in units of the larger star's radius) must satisfy b < 1 + R₂/R₁. For typical main-sequence binaries with a of a few stellar radii, that requires i within only a few degrees of 90°. The eclipsing geometry is therefore rare: only a small fraction of binaries eclipse from any given vantage point, and that fraction is what makes EBs both special and immediately diagnostic.

Once the geometry is achieved, what we see is a flat or smoothly varying baseline interrupted twice per orbit by a dip. The deeper dip is by convention called the primary minimum; the shallower one is the secondary minimum. The two dips are separated by exactly half the orbital period in circular orbits, and by a phase fraction that depends on eccentricity and longitude of periastron otherwise. From that one diagram — flux versus phase — you can already read off the period, the rough geometry, and the temperature ratio.

Primary and secondary minima — why the depths matter

The intuition that the deeper eclipse is "the bigger star blocking the smaller" is wrong. Eclipses are blockings of flux, not area, and flux per unit area scales as the fourth power of effective temperature. The deeper eclipse is therefore the one in which the higher surface brightness is blocked, regardless of which star is geometrically in front.

For a total eclipse (the smaller star fully behind the larger), the ratio of the two dip depths is simply

ΔF_primary / ΔF_secondary = (T_hot / T_cool)⁴   (Stefan-Boltzmann)

where T_hot and T_cool are the surface temperatures of the two stars. If you measure depths of, say, 0.7 mag and 0.05 mag, that is a flux ratio of about 14, so T_hot/T_cool ≈ 14^(1/4) ≈ 1.93. With one absolute temperature from a spectrum, the other follows for free. This is one of the cleanest measurements in all of astrophysics.

Primary minimum, in canonical Algol-type usage, is when the cooler star eclipses the hotter one — bright surface blocked, deep dip. Secondary minimum is when the hotter star passes in front of the cooler — dimmer surface blocked, shallow dip. (The Wikipedia naming convention follows depth, not which star is in front; depth is what an observer measures.)

Duration of eclipse gives the radii

For circular orbits with relative orbital velocity v_rel = 2π a / P, the eclipse contact times encode the radii directly. Define

• First contact: the moment a limb of the eclipsing star first touches the limb of the eclipsed star.
• Second contact: the moment the eclipsing star is fully inside the eclipsed star (for a total eclipse) — totality begins.
• Third contact: totality ends.
• Fourth contact: limbs separate.

For a central total eclipse (impact parameter b = 0):

t₁₄ − t₁₁ ≈ 2 (R_eclipsed + R_eclipsing) / v_rel    full duration
t₂₃       ≈ 2 (R_eclipsed − R_eclipsing) / v_rel    totality
t₁₂ = t₃₄ ≈ 2 R_eclipsing / v_rel                   ingress / egress

Two equations, two radii — solved. For non-central eclipses (b > 0) the geometry adds an impact-parameter term that the light-curve fit recovers simultaneously. The light-curve solver does not need to "assume" anything about stellar structure: the durations are pure geometry.

The three canonical morphologies

The shape of the light curve outside the eclipses divides eclipsing binaries into three Wikipedia-canonical morphological classes, traditionally named after their prototypes.

Class	Prototype	Roche lobe state	Light curve outside eclipses
Detached (Algol-type)	β Persei (Algol)	Both stars well inside	Flat — two spherical stars
Semi-detached (β Lyrae-type)	β Lyrae	One star fills its lobe	Smooth sinusoidal variation due to tidal distortion + mass stream
Contact / overcontact (W UMa-type)	W Ursae Majoris	Both overfill, shared envelope	Continuous sinusoid; two nearly equal minima

Algol-type (detached) systems are the cleanest case: two well-separated stars, light curve flat between eclipses, deep primary and shallow secondary. β Persei itself has period 2.867 d, primary depth 1.27 mag, secondary depth 0.06 mag. The Algol Paradox — that the less massive star is more evolved — is now understood as the signature of past mass transfer in a system that was once semi-detached.

β Lyrae-type (semi-detached) systems are caught in active mass transfer. One star, having evolved off the main sequence, expanded to fill its Roche lobe and is streaming material onto its companion. The light curve varies continuously even between eclipses because the lobe-filling star is tidally distorted into an egg shape; we see different cross-sections through the orbit. β Lyrae itself has period 12.94 d and the period is lengthening at about 19 seconds per year as the system loses orbital angular momentum to the mass stream.

W UMa-type (contact / overcontact) systems have both stars overfilling their Roche lobes and sharing a common convective envelope. The two stars have nearly the same surface temperature (because they share an envelope) despite very different masses, and the light curve is a continuous sinusoid with two nearly equal minima. Periods are typically below 0.5 days — the shortest period non-degenerate binaries known. The class is thought to evolve toward eventual coalescence; the 2008 transient V1309 Sco is widely interpreted as a W-UMa merger caught in the act.

Why eclipsing binaries are the gold standard

Combine a high-precision light curve (depths, durations, eccentricity, third light) with high-precision double-lined spectroscopy of both components (RV semi-amplitudes K₁, K₂, systemic γ, projected v sin i) and the chain of inference is essentially deterministic:

1. Period. From the photometric data alone — repeated dips.
2. Eccentricity and periastron longitude. From the phase separation of secondary minimum and from the RV curve shape.
3. Mass ratio q = M₂/M₁. From the ratio of RV semi-amplitudes: K₁/K₂ = M₂/M₁ = q.
4. Inclination i. Fixed near 90° by the existence of the eclipse; refined by depth and duration.
5. Sum of radii / a. From the total eclipse duration.
6. Ratio of radii. From the totality-to-ingress ratio.
7. Surface-brightness ratio → temperature ratio. From the depth ratio (Stefan-Boltzmann).
8. Absolute scale (a, M₁, M₂, R₁, R₂). From Kepler's third law with the inclination known: M₁ + M₂ = (4π²/G) a³/P², and a = (K₁ + K₂) P / (2π sin i).
9. Absolute temperatures. From spectral analysis of the disentangled component spectra.
10. Luminosities. L = 4π R² σ T⁴ — geometric, no models invoked.
11. Distance. Apparent flux + bolometric L + interstellar extinction → distance, with no rung above EBs on the ladder.

The best detached EBs (e.g. the Andersen 1991 compilation and its modern successors) deliver masses and radii at 1 % accuracy or better. Nothing else outside the Solar System matches that precision; eclipsing binaries are the calibration anchor for every stellar evolution model that exists.

Worked example: an Algol-clone

Suppose photometry of a target gives:

• Period P = 3.0 days
• Primary minimum depth = 0.80 mag (flux ratio in dip = 0.477)
• Secondary minimum depth = 0.10 mag (flux ratio in dip = 0.912)
• Total duration of primary eclipse = 8.0 hours; flat bottom (total) lasting 1.5 hours

And double-lined spectroscopy gives K₁ = 75 km/s, K₂ = 220 km/s (so q = K₁/K₂ = 0.34 means M₂/M₁ = 0.34, i.e. the spectroscopic primary is the more massive star).

Total-eclipse depth ratio implies (in flux) ΔF_prim/ΔF_sec ≈ 0.523/0.088 ≈ 5.94, so T_hot/T_cool ≈ 5.94^(1/4) ≈ 1.56. If a spectrum tells us T_hot = 9000 K (a B9 star), T_cool ≈ 5770 K (a G2 subgiant).

Kepler's third law with i = 90° (eclipses guarantee this is close):

a = (K₁ + K₂) P / (2π sin i)
  = (75 + 220) km/s × 3.0 × 86400 s / (2π × 1)
  ≈ 1.22 × 10⁷ km
  ≈ 17.5 R☉

And the masses:

M₁ + M₂ = (4π²/G) a³/P²
        ≈ 3.8 M☉
M₁ = (M₁ + M₂) / (1 + q) ≈ 2.85 M☉
M₂ ≈ 0.95 M☉

From the durations, with v_rel = 2π a / P ≈ 295 km/s:

R₁ + R₂ ≈ ½ × t_total × v_rel ≈ 6.1 R☉
R₁ − R₂ ≈ ½ × t_flat  × v_rel ≈ 1.1 R☉
→ R₁ ≈ 3.6 R☉,  R₂ ≈ 2.5 R☉

So the system is a 2.85 M☉ B9 dwarf eclipsed by a 0.95 M☉ G2 subgiant. The less massive star is the more evolved one — the Algol Paradox. Bolometric luminosities follow from L = 4π R² σ T⁴: L₁ ≈ 78 L☉, L₂ ≈ 5 L☉. Distance follows from the apparent magnitude and bolometric correction.

Eclipsing binaries on the distance ladder

The product 4π R² σ T⁴ is a true bolometric luminosity, not a parallax-calibrated or period-calibrated proxy. That makes detached EBs unique on the cosmic distance ladder: they sit on their own rung, anchored on basic geometry and Stefan-Boltzmann, and provide a check against parallax (within the Milky Way) and a distance to galaxies that hosts no Cepheid or TRGB calibration disagreement.

• LMC: 49.59 ± 0.54 kpc. Pietrzyński et al. (2019, Nature) used 20 long-period (~100 d) late-type giant EBs in the Large Magellanic Cloud to deliver the most precise extragalactic distance ever measured — 1.1% total uncertainty. The LMC distance is the foundation of the Cepheid leg of the distance ladder for H₀.
• SMC: 62.44 ± 0.47 kpc. Graczyk et al. (2020) using the same surface-brightness-color relation method.
• M31, M33. Several EBs to about 4% accuracy each.
• Galactic open clusters. NGC 4337, NGC 188, and others have member EBs that deliver ~1% cluster distances — used to anchor isochrone-based age determinations.

The price of this precision is that detached EBs are scarce: each candidate requires multi-year photometry plus repeated high-resolution spectroscopy. The technique scales with survey investment, not with raw star counts.

Modern surveys and the explosion in known EBs

• Kepler / K2. About 3000 EBs in the original Kepler field, with photometry precise enough to detect grazing eclipses and pulsation-eclipse superposition. The KEBC (Kepler Eclipsing Binary Catalog) is the gold standard for high-precision light curves.
• TESS. All-sky photometry at 2-minute and 30-minute cadence; over 15,000 EBs catalogued through the first extended mission, with the count still growing each sector.
• ASAS-SN. The All-Sky Automated Survey for Supernovae has produced V-band light curves of essentially every star brighter than V ≈ 17 over the entire sky; the ASAS-SN variable star catalog includes ~60,000 EBs.
• OGLE. Optical Gravitational Lensing Experiment toward the Magellanic Clouds and the Galactic bulge — tens of thousands of EBs, especially valuable for studying low-metallicity populations.
• Gaia DR3. Has now flagged ~2 million eclipsing-binary candidates by epoch photometry; only a few percent have been confirmed.
• Vera Rubin Observatory (LSST). Will discover an order of magnitude more EBs again, with deep, multi-band, ten-year baselines.

The combination of survey photometry and follow-up spectroscopy — TESS plus APOGEE plus 4MOST/WEAVE — is now producing thousands of new precision EB solutions per year.

Modelling: Wilson-Devinney and PHOEBE

The standard light-curve modelling tool for half a century has been the Wilson-Devinney (WD) code, introduced by Robert Wilson and Edward Devinney in 1971. WD handles the full Roche-geometry distortion of close binary stars, limb darkening, gravity brightening (von Zeipel's theorem), the reflection effect, third light contamination, and starspots. Inputs are the parameter set (masses, radii, T, i, e, ω, limb-darkening coefficients, spot configurations); outputs are synthetic light and RV curves; the inverse problem — fitting the model to data — is run by external optimisers (often differential corrections or modern MCMC).

PHOEBE (PHysics Of Eclipsing BinariEs), maintained by the Villanova group, is the modern WD-compatible re-implementation: same physics, modern statistical inference (emcee, dynesty), proper systematics handling, and a Python API. PHOEBE 2 is now the de facto standard for refereed EB papers.

Where eclipsing binaries show up in astrophysics

• Calibrating stellar evolution. Every precision mass-radius point that constrains the input physics of MESA, PARSEC, BaSTI or any other stellar evolution grid traces back to a detached EB. The famous Andersen 1991 sample of 45 well-measured systems still anchors today's isochrones.
• Mass-luminosity relation. Empirical M-L for main-sequence stars from 0.1 to 30 M☉ comes overwhelmingly from EBs.
• Mass transfer and binary evolution. Algol's paradox, β Lyrae's secular period change, W UMa's common envelope — these are the empirical anchors of binary stellar evolution theory.
• Type Ia supernova progenitors. Some Type Ia SNe trace back to mass-transfer systems with white-dwarf accretors; EB studies of cataclysmic variables and their semi-detached relatives constrain the channel.
• Apsidal motion test of GR / stellar structure. In eccentric eclipsing binaries, the orbit's periastron precesses due to a combination of classical tidal effects and a small GR contribution. The apsidal motion rate constrains the internal density distribution of the stars — a unique test of stellar structure theory.
• Planets via eclipse-timing variations. An unseen third body perturbs the eclipse timings of a close binary, producing periodic O-C variations. Several circumbinary planets (including the Kepler-16, -34, -35 systems) have been or could have been found this way.

Common pitfalls

• Confusing primary star with primary minimum. Primary minimum is the deeper dip; primary star is the more massive (or more luminous) star. These are independent labels. In a typical Algol-like system, primary minimum is the eclipse of the primary star.
• Assuming spherical stars in close binaries. In semi-detached and contact systems, the stars are not spheres but Roche equipotentials. Treating them as spheres badly biases the radii — always use a code that handles Roche geometry (WD, PHOEBE) for periods below a few days.
• Ignoring third light. A nearby unrelated star in the photometric aperture dilutes the eclipse depths, biasing temperatures and radii. Third-light contamination is essential to fit, especially with Gaia DR3 multiplicity catalogues now making third-body identification routine.
• Linear limb darkening on precision data. Linear limb-darkening laws fail at the 10⁻³ photometric precision of Kepler and PLATO. Use quadratic, square-root, or four-parameter Claret laws.
• Treating reflection as negligible. The hotter star irradiates the cooler one, raising its effective surface temperature on the day side. The "reflection effect" raises the secondary minimum by typically a few percent of total flux and shifts derived temperatures if neglected.
• Eclipses ≠ transits. Stellar eclipses block flux of order 10⁻¹ to 10⁰; planetary transits block 10⁻⁴ to 10⁻². They are the same geometry, but the diagnostic and modelling toolkit is different — exoplanet transit fitting routinely ignores parts of the binary toolkit, and EB modelling ignores small-body irradiation.