The Coronagraph

The problem: glare, not faintness

The hard part of seeing a planet next to a star is almost never that the planet is faint in absolute terms. A young Jupiter glowing from its formation heat would be an easy target for any large telescope if it were floating alone in the dark. The problem is that it sits a fraction of an arcsecond from a star that is millions to billions of times brighter. You are trying to spot a firefly hovering beside a lighthouse, from across an ocean. The lighthouse does not have to be intrinsically bright to drown the firefly — it just has to be next to it.

A coronagraph is the instrument built to defeat exactly this glare problem. It does not make the planet brighter; it makes the star dimmer in the specific region of the image where you want to look. Conceptually it is an artificial eclipse engineered inside the telescope, plus a second trick — invented by Bernard Lyot — that cleans up the light the eclipse alone fails to stop. Get both right and you can pull a signal out of a region of the focal plane where the star's own light has been suppressed by factors of a million, a billion, or — at the frontier — ten billion.

How a Lyot coronagraph actually works

A classical Lyot coronagraph is a chain of four planes, and the magic is that suppression happens in two of them, not one:

1. Entrance pupil. The full circular telescope aperture of diameter D collects light. A distant star is effectively a point source, so its light arrives as a flat (plane) wavefront across the pupil.
2. First focal plane — the occulting mask. The telescope focuses the star to a tiny diffraction-limited spot, the Airy pattern. A small opaque (or specially shaped) mask covers the bright core. This removes the central peak but, crucially, leaves the diffracted light alone.
3. Re-imaged pupil — the Lyot stop. A relay lens forms an image of the pupil again. Here is Lyot's key discovery: the light diffracted by the mask's hard edge does not spread uniformly — it piles up into a bright ring at the edge of the re-imaged pupil. A slightly undersized aperture, the Lyot stop, masks off that outer ring, throwing away the diffracted starlight while passing most of the planet's off-axis light.
4. Final focal plane — the science camera. The detector now sees a deep, dark region around the (hidden) star, in which an off-axis companion still forms a recognisable image.

The reason the simple "black dot" of folklore fails is that it stops at step 2. The Airy diffraction halo of the full aperture — and the bright ring the occulter itself diffracts to the pupil edge — carries far more starlight than any planet. The Lyot stop in step 3 is what actually delivers the contrast. This two-stage architecture is why every serious coronagraph, no matter how exotic its focal-plane mask, still contains a downstream pupil stop.

The physics: diffraction, λ/D, and contrast

Everything about coronagraph performance is set by diffraction. A telescope of diameter D observing at wavelength λ cannot focus a point source to a point; it forms an Airy pattern whose central disk has an angular radius

θ_Airy = 1.22 λ / D    (radians)

The natural unit of angle for any coronagraph is therefore λ/D. For a 2.4 m space telescope (the Hubble / Roman aperture) at λ = 600 nm:

λ/D = 600e-9 m / 2.4 m = 2.5e-7 rad
    = 2.5e-7 × 206265 arcsec/rad ≈ 0.052 arcsec

So one λ/D is about 52 milliarcseconds, and the Airy disk radius (1.22 λ/D) is about 63 mas. A coronagraph can only work outside the region it must mask, so the smallest separation it reaches — the inner working angle (IWA) — is a few λ/D. The two figures of merit that define a coronagraph are:

• Raw contrast C — the residual starlight intensity in the dark region, relative to the unblocked stellar peak. State of the art ranges from C ≈ 10⁻⁶ on the ground to a goal of 10⁻¹⁰ in space.
• Inner working angle (IWA) — the smallest separation, in λ/D, at which a companion is still transmitted at ≈ 50%. Vortex masks reach ≈ 1 λ/D; classical Lyot masks are nearer 3–4 λ/D.

There is also an outer working angle set by the deformable mirror's actuator count: a mirror with N actuators across the pupil can only control spatial frequencies out to (N/2) λ/D, beyond which the dark region ends. The planet you can detect must lie in the annulus between the IWA and the outer working angle, in a dark hole carved by interference.

The brightness contrasts you must beat

The whole design problem is dictated by how faint your target is relative to its star. These are the real numbers that drive instrument requirements:

Target	Band	Planet/star contrast	Where reachable
Solar corona (inner)	Visible	~10⁻⁶ of disk	Lyot solar coronagraph
Bright debris disk	Near-IR	~10⁻⁴ – 10⁻⁵	HST, ground AO
Young self-luminous gas giant	Near-IR (H, K)	~10⁻⁴ – 10⁻⁶	VLT/SPHERE, Gemini/GPI
Mature Jupiter (reflected)	Visible	~10⁻⁹	Roman Coronagraph (demo)
Earth twin in habitable zone	Visible (0.5 µm)	~10⁻¹⁰	Future flagship (HWO) / starshade

The jump from imaging young giants (already done from the ground) to imaging an Earth twin is a factor of about 10,000 in contrast and is the central technological challenge of the next two decades. It is the reason the field has moved from simple Lyot masks to vortex masks, apodized pupils, and active wavefront control with deformable mirrors.

The main coronagraph designs

Since Lyot, dozens of mask architectures have been invented, each trading inner working angle, throughput, and bandwidth differently. The deeper optics of these designs are covered in our companion article on coronagraph imaging; the essentials are:

Design	How it suppresses light	Typical IWA	Trade-off
Classical Lyot	Opaque focal-plane spot + Lyot stop	~3–4 λ/D	Robust, but throughput lost to the stop
Band-limited Lyot	Graded-transmission mask, mathematically nulls diffraction	~3 λ/D	Insensitive to pointing; hard to fabricate
Apodized-pupil Lyot (APLC)	Shaped pupil transmission + Lyot mask	~3 λ/D	Workhorse of SPHERE; needs a smooth apodizer
Vector vortex	Spiral phase ramp sends on-axis light entirely outside the Lyot stop	~1 λ/D	Tiny IWA, high throughput; chromatic
Four-quadrant phase mask	π phase shift in alternating quadrants destructively interferes the core	~1 λ/D	Small IWA; dead zones along the quadrant lines
External starshade	Free-flying flower-shaped occulter casts a deep shadow before the telescope	set by geometry (tens of mas)	No internal wavefront limit; slow to re-point

The trend is clear: classical Lyot masks were robust but threw away light and could not reach small separations; phase-based designs (vortex, four-quadrant) push the IWA down toward 1 λ/D, letting you photograph planets in tighter orbits, at the price of chromaticity and sensitivity to alignment.

Worked example: imaging an Earth twin at 10 parsecs

Consider an Earth-Sun analogue at a distance of 10 parsecs (about 33 light-years). The Earth orbits the Sun at 1 astronomical unit. The angular separation we must resolve is

θ = a / d = (1 AU) / (10 pc)
  = 1 / 10  arcsec   (since 1 AU at 1 pc = 1 arcsec, by definition of the parsec)
  = 0.1 arcsec = 100 milliarcseconds

Now compare that to the diffraction scale of a 6 m space telescope (the class of a future Habitable Worlds Observatory) at λ = 500 nm:

λ/D = 500e-9 / 6 = 8.3e-8 rad = 8.3e-8 × 206265 ≈ 0.017 arcsec = 17 mas
θ / (λ/D) = 100 mas / 17 mas ≈ 5.9 λ/D

So the planet sits about 6 λ/D from the star — comfortably outside an IWA of 1–4 λ/D. Geometry is not the bottleneck. The bottleneck is contrast:

Reflected-light contrast ≈ A_g × (R_planet / a)²
  with geometric albedo A_g ≈ 0.3, R_⊕ = 6.37e6 m, a = 1 AU = 1.496e11 m
  (R/a)² = (6.37e6 / 1.496e11)² = (4.26e-5)² = 1.8e-9
  C ≈ 0.3 × 1.8e-9 ≈ 5.4e-10

The Earth twin is roughly 5 × 10⁻¹⁰ as bright as its star in reflected visible light — about two billion times fainter. To detect it the coronagraph must suppress the star to below ~10⁻¹⁰ and hold that suppression stable while integrating, since photon-starved exposures take hours. That is why the requirement is not merely "block the star" but "control the wavefront to a few picometres for the duration of the exposure." The mask is the easy part; wavefront stability is the hard part.

Discovery: Lyot, the corona, and the move to the stars

The instrument is named for the corona, the Sun's million-degree outer atmosphere. Before 1930 the corona could only be seen during the few minutes of a total solar eclipse, when the Moon happens to block the photosphere. The French astronomer Bernard Lyot (1897–1952) set out to manufacture an eclipse on demand. Between roughly 1930 and 1939 he built, at the high-altitude Pic du Midi Observatory (2,870 m), an instrument with an occulting cone that blocked the solar disk and — his decisive innovation — a downstream stop that removed the light diffracted by the telescope optics. Working at altitude, where the sky scatters very little blue light, and with obsessively clean lenses to minimise scattered light, Lyot obtained the first images and spectra of the corona outside an eclipse, and discovered coronal emission lines. His 1939 results established solar coronagraphy as a routine science.

The same diffraction-suppression principle was later turned outward onto individual stars. Stellar coronagraphy matured once two enabling technologies arrived: adaptive optics, which corrects the atmospheric blur that would otherwise smear the starlight into uncontrollable speckles, and the space environment, which removes the atmosphere entirely. Milestones include the first ground-based AO coronagraphs in the 1990s; the Hubble Space Telescope's coronagraphic modes (NICMOS, STIS, ACS) imaging debris disks such as Fomalhaut; the 2008 direct image of the HR 8799 multi-planet system; the dedicated extreme-AO coronagraphs SPHERE (VLT, 2014) and GPI (Gemini, 2014); JWST's coronagraphic modes (NIRCam, MIRI) launched in 2021; and the Roman Space Telescope Coronagraph Instrument, a technology demonstrator built to reach ~10⁻⁹ contrast and pave the way for a future flagship capable of imaging Earth-like worlds.

Why a perfect mask is not enough: speckles and dark holes

A coronagraph only suppresses light that arrives as a clean, flat wavefront. Any optical imperfection — a sub-nanometre polishing error, a thermal flexure, a smear of atmospheric turbulence — diffracts a little starlight past the mask into the dark region, where it forms a bright "speckle." Speckles are the bane of high-contrast imaging because a static speckle looks exactly like a planet: a compact point of light at a fixed separation. Distinguishing the two requires either angular differential imaging (letting the field rotate so real planets move while speckles stay fixed to the optics) or spectral differential imaging (planets have molecular absorption bands the star does not).

The frontier technique is the dark hole. One or two deformable mirrors with hundreds to thousands of actuators are commanded — using focal-plane wavefront sensing algorithms such as electric-field conjugation — to introduce a tiny, deliberate pattern that destructively interferes with the residual starlight over a chosen region of the image. The result is a half-moon or annular patch of artificially deepened darkness, the dark hole, where the contrast can be pushed orders of magnitude below the raw mask performance. Maintaining the dark hole demands wavefront stability at the picometre level over the hours of an exposure, which is why a future Earth-imaging mission is fundamentally a stability-engineering problem as much as an optics one.

Solar coronagraphs and space weather

The coronagraph never left its original solar role. Spaceborne solar coronagraphs image the corona continuously to monitor coronal mass ejections — billion-tonne eruptions of magnetised plasma that drive space weather. The LASCO coronagraphs aboard the SOHO spacecraft (launched 1995) have watched the corona for three decades, and as a serendipitous bonus have discovered thousands of sungrazing comets that pass behind the occulting disk. ESA's Proba-3 mission flies two spacecraft in formation 144 m apart so that one casts a precise shadow on the other — an externally occulted solar coronagraph that mimics a starshade and reaches very close to the solar limb. Solar and stellar coronagraphy remain the same instrument applied to wildly different contrast and angular-scale regimes.

Common misconceptions and subtleties

• "The mask alone hides the star." No — the focal-plane mask removes the bright core, but the diffracted halo it leaves is far brighter than any planet. The Lyot stop in the re-imaged pupil is what actually delivers the contrast. A coronagraph is a two-plane device.
• "A bigger mask gives better contrast." A larger occulter does suppress more starlight, but it also raises the inner working angle, hiding the very planets you want closest to the star. Coronagraph design is a tug-of-war between contrast and IWA, not a free lunch.
• "Coronagraphs and adaptive optics are alternatives." They are partners. AO (or, in space, wavefront control with deformable mirrors) delivers the flat wavefront the coronagraph needs; the coronagraph then suppresses that cleaned-up starlight. Neither works alone for exoplanet imaging.
• "The corona is hot because it is close to the surface." The opposite — the corona at over a million kelvin is far hotter than the ~5,800 K photosphere beneath it, the unsolved coronal heating problem. The coronagraph is the instrument that made this temperature inversion measurable outside eclipses.
• "Detecting a point of light is enough." A static speckle perfectly mimics a planet. Confirmation requires showing the source has planet-like behaviour — common proper motion with the star over time, or molecular spectral features — not just a single bright pixel in the dark hole.
• "A starshade is just a bigger mask." A starshade is external and works by geometry, casting a diffraction-engineered shadow before the telescope, so the starlight never enters the optics and the telescope's own wavefront errors no longer set the contrast floor. That is a fundamentally different (and complementary) strategy from an internal coronagraph.