Coronagraph Imaging

The contrast problem nobody else has to solve

A Sun-like star at 10 parsecs has apparent magnitude V ≈ 5 — bright. The Jupiter that orbits it at 5 AU sits 0.5 arc-second away and reflects roughly 10^-9 of the host's optical flux. A true Earth twin at 1 AU sits 100 mas away and reflects 10^-10. Two challenges sit in series. Resolution: the planet must be separable from the star. Contrast: the planet must rise above the locally-scattered starlight where it lies. Adaptive optics (or being in space) addresses the resolution problem. The coronagraph attacks the contrast problem.

For a circular, unobscured telescope the on-axis stellar PSF is the familiar Airy pattern, with central peak and rings whose envelope falls off only as 1/θ³. At θ = 5 lambda/D the ring intensity is ~10^-4 of the peak — six orders of magnitude brighter than a Jupiter-analog planet sitting there. No amount of integration time helps; you are photon-shot-noise limited by the starlight you cannot photon-count out of. The coronagraph is the optical preconditioner that knocks down those rings before they reach the detector.

The Lyot design — building block of everything else

Bernard Lyot's 1930 instrument had four ingredients still recognizable today. (1) A primary objective forms a focal-plane image of the Sun. (2) An opaque disk in that focal plane occults the solar disk itself, passing only the corona-and-diffraction light around it. (3) A relay lens re-images the system pupil. (4) An undersized "Lyot stop" in that re-imaged pupil blocks the bright ring of diffracted light that the focal-plane disk scatters into the pupil edge. Detector beyond.

Telescope → Focal plane mask (opaque disk, ~3-4 lambda/D)
                  ↓
                Relay lens
                  ↓
                Re-imaged pupil + Lyot stop (undersized aperture)
                  ↓
                Detector

Stellar light:  attenuated by mask × ring rejected by Lyot stop ≈ 10⁻³-10⁻⁳
Planet light:   off-axis → misses mask → passes through Lyot stop ≈ 0.3-0.7×

This is the workhorse design. SPHERE, GPI, and many older systems are evolved Lyot variants. The achievable contrast is set by how aggressively you can undersize the Lyot stop (which costs throughput and resolution) and how well you can apodize the pupil to soften the diffraction pattern reaching the mask.

The vortex coronagraph — phase, not amplitude

A vector vortex mask is a transparent element in the focal plane that rotates the polarization (or equivalently the phase) by exp(i l phi) where l is an even integer and phi is the azimuthal angle. The mask itself is fully transmissive — no light is absorbed. What happens is mathematically dramatic: when you Fourier-propagate a circularly symmetric on-axis stellar Airy pattern multiplied by exp(i l phi) back to the pupil plane, the resulting field is identically zero inside the original pupil aperture (for l = 2, 4, ...). All on-axis light is diverted to the outside of the geometric pupil, where a Lyot stop blocks it. An off-axis planet hits the singularity off-centre, picks up no net winding phase, and the diversion does not happen — its light reaches the detector with ~99% throughput.

Why we care: inner working angle 2 λ/D for the charge-2 vortex (versus 3-4 for Lyot), full off-axis throughput, no chromatic effects to first order with achromatic vortex implementations. Charge-4 vortex sacrifices a little IWA for robustness to low-order aberrations.

Worked example — can SPHERE image a Jupiter analog at 5 AU around alpha Cen?

alpha Centauri A is the obvious target. Distance d = 1.34 pc. A Jupiter analog at 5 AU sits at angular separation 5 AU / 1.34 pc = 3.73 arc-seconds.

Telescope diameter (VLT)              D = 8.2 m
Wavelength (H-band)                   λ = 1.65 μm
Diffraction unit                      λ/D = 1.65e-6 / 8.2 = 2.01e-7 rad
                                            = 0.0415 arc-sec = 41.5 mas
Separation in λ/D                    3.73" / 0.0415" ≈ 90 λ/D

SPHERE Lyot IWA                       ~3 λ/D = 0.12"   (easily clears 3.73")
Contrast at 90 λ/D after ADI         ~10⁻⁳ (post-processed)
Required contrast for Jupiter analog  ~10⁻⁹ (reflected light, 5 AU)
→ Limited by 4 orders of magnitude   — not detectable in reflected light.

So SPHERE cannot see a true Jupiter analog in reflected light around alpha Cen A. What it can see is a Jupiter-analog if the planet is young and self-luminous in the IR (10^-6 contrast achievable at hundred-Myr ages), and that's what the existing 25-planet catalogue contains. The Roman Coronagraph (~10^-9 raw contrast in space) closes one order of magnitude. HWO (~10^-10) closes the rest and reaches Earth-twin reflected-light contrasts at last.

Major coronagraph design families compared

Design	Inner working angle	Off-axis throughput	Demonstrated contrast	Operational instrument	Strength / weakness
Classical Lyot	~3-4 λ/D	~0.5	10-5 raw, 10-7 post-ADI	SPHERE, GPI, CHARIS	Simplest — mid-IWA limit
Apodized Pupil Lyot (APLC)	~3 λ/D	~0.45	10-8 lab, 10-6 on-sky	SPHERE-IRDIS, Roman CGI	Best contrast at moderate IWA
Vector Vortex (l=2)	~2 λ/D	~0.7	10-8 at HCIT (NASA JPL)	SCExAO, Keck/NIRC2, KPIC	Small IWA, sensitive to tip-tilt
Vector Vortex (l=4)	~3 λ/D	~0.7	10-9 in lab	Subaru SCExAO	Robust to low-order aberrations
Four-Quadrant Phase Mask	~1.5 λ/D	~0.5	10-6 on-sky	SPHERE-IRDIS (commissioning)	Smallest IWA, chromatic
PIAA-CMC	<1 λ/D	~0.6	10-9 lab	SCExAO; HWO study	Best IWA at cost of complexity
Starshade (external)	~60 mas absolute	~1.0	10-10 simulated	(none flown)	No internal optics — formation flying

How coronagraphy got from sun to exoplanets

• 1930. Bernard Lyot uses a custom occulting disk to observe the solar corona without an eclipse from Pic du Midi. The "Lyot stop" terminology dates to this paper.
• 1995. First stellar coronagraph used at Calar Alto detects faint brown dwarf companion Gliese 229 B — the first cool sub-stellar object directly imaged.
• 2004. Roddier-Roddier and four-quadrant phase mask concepts developed; the case for small IWA designs becomes urgent.
• 2005. Mawet, Riaud, Surdej, Habraken propose the vector vortex coronagraph using liquid crystal polymer optics. First on-sky use at Hale-200" in 2010.
• 2008-2010. Christian Marois and Bruce Macintosh image HR 8799 b/c/d/e with Keck NIRC2 + adaptive optics + coronagraph — four planets in one system.
• 2014-2015. SPHERE (VLT) and GPI (Gemini) come online: dedicated extreme-AO instruments with apodized pupil Lyot coronagraphs, achieving 10^-6 contrast at 0.5".
• 2018. Keppler, Mueller, Vigan resolve PDS 70 b/c inside a circumstellar disk gap — the first planets caught actively forming.
• 2027. Nancy Grace Roman Space Telescope launches with the Coronagraph Instrument (CGI), an apodized pupil Lyot expected to demonstrate 10^-9 raw contrast on stellar targets.
• 2040s. Habitable Worlds Observatory (HWO, formerly LUVOIR/HabEx) launches a ~6-m optical/UV telescope with a vortex or PIAA coronagraph targeting 10^-10 at 60 mas — the regime where Earth-twins in reflected light become photon-shot-noise limited.

Why coronagraphy matters for the next decade of astrophysics

• Habitable-zone characterization. Reflected-light coronagraphy is the only way to take spectra (O₂, H₂O, CH₄) of mature rocky planets around Sun-like stars — transit spectroscopy needs the planet to cross the disk.
• Disk imaging. Circumstellar disks at 10-100 AU are 10^-4-10^-7 fainter than their host stars — comfortably inside SPHERE-class capabilities — and reveal planet-induced gaps, rings, and warps.
• Brown dwarf demographics. Cool companions (T < 1000 K) too faint for radial velocity at wide separations are detected directly.
• Multiplicity of star formation. Coronagraphs let us resolve binaries and triples down to a few AU separation at the distances of nearby star-forming regions.
• Active galactic nuclei. Suppressing the AGN core reveals host-galaxy structure and circum-nuclear disks.
• Telescope demonstration. Each generation of mirror stability, deformable mirror count, and AO control is forced to higher precision by coronagraph requirements — benefitting all imaging science.

Common misconceptions

• "A coronagraph is just an occulting mask in front of the telescope." No — the mask sits inside the telescope at a focal plane, paired with a Lyot stop in a re-imaged pupil. External occulters are starshades, a different (and complementary) approach.
• "Coronagraphs are like camera black levels." Contrast levels of 10^-9 are nine decades below the source. There is no electronics-domain analog — the suppression is done in the optical Fourier transform itself.
• "Bigger telescopes solve the problem." Larger D shrinks lambda/D, putting IWA closer to the star — but larger D also amplifies wavefront-error scattering inside the dark hole. Stability matters as much as size.
• "Coronagraphs only work on stars." The same principle suppresses the AGN core, the lunar limb in earthshine studies, and the cosmic-microwave dipole in some balloon experiments.
• "Vortex coronagraphs lose the planet's polarization." Modern vector vortex implementations have demonstrated <1% polarization-dependent throughput in the planet beam.
• "Coronagraphs work at any wavelength." Achromatic designs exist, but most coronagraph contrast figures-of-merit are quoted at a specific wavelength — ground-based work mostly in H/K band, space (Roman, HWO) in optical 500-900 nm.

Open questions

• Picometer wavefront stability. HWO needs the mirror to drift less than 10 pm over an hour. Current state of the art is closer to 1 nm. The 100× gap is driven by thermal, mechanical, and control loop noise; closing it is the dominant cost driver for HWO.
• Detection floor versus characterization floor. Detecting a planet at 10^-10 needs ~10⁴ photons; getting a low-resolution spectrum that distinguishes O₂ from O₃ needs 10⁶-10⁷. Each integration time goes up two orders of magnitude.
• Speckle calibration limits. Even with perfect optics, residual atmospheric and quasi-static speckles create the floor. New algorithms (KLIP, regime-switching ADI, RDI with reference libraries) keep improving but no analytic limit is known.
• Comparison with starshades. Internal vs. external is still actively debated for HWO. Starshades win on contrast in principle but lose on observing efficiency in practice.