Direct Imaging of Exoplanets

The dot beside the spotlight

Most exoplanet detection methods are indirect. The transit method looks at the star's brightness; radial velocity looks at the star's spectrum; astrometry at the star's position. The planet itself contributes nothing — you infer its existence from the star's response.

Direct imaging is the exception. You photograph the planet. Photons that left the planet's atmosphere reach your camera. The challenge is straightforward to state and brutally hard to solve: the planet is very faint and very close to a very bright star. For a Jupiter analog at 5 AU around a Sun-like star at 10 parsecs the separation on the sky is 0.5 arcseconds — within the diffraction pattern of any ground-based telescope on Earth, dominated by the star's light. Worse, the planet's brightness is only ~10⁻⁵ that of the star at near-infrared wavelengths, ~10⁻⁹ in visible light. The planet is buried in glare.

The first successful images appeared in 2004 (2M1207 b, a 5-Jupiter-mass companion to a brown dwarf, by Chauvin et al. with VLT/NACO) but the field truly opened in 2008 when Christian Marois and colleagues at Keck and Gemini directly imaged three giant planets around HR 8799 — the first multi-planet system seen this way. That image, with three faint dots beside a masked central star, has become one of the iconic exoplanet images of the century.

The contrast problem in numbers

How much brighter is the star than the planet? Two terms contribute to a planet's apparent flux: thermal emission (blackbody radiation from the planet's own heat) and reflected light (starlight bouncing off the planet's atmosphere or surface).

	Effective T (K)	Peak λ	Contrast at 2 μm	Contrast at 0.5 μm
Young Jupiter (10 Myr)	~1500	~2 μm	~10⁻⁴	~10⁻⁹
Young Jupiter (100 Myr)	~800	~3 μm	~10⁻⁵	~10⁻¹⁰
Mature Jupiter (5 Gyr)	~130	~25 μm	~10⁻⁹	~10⁻¹⁰
Earth analog	~290	~10 μm	~10⁻⁷	~10⁻¹⁰

The strategy that has worked: search for young, hot, wide-separation planets in the near infrared (2 μm). They self-luminate from formation heat (still ~100 Myr away from full cooling), the contrast is "only" 10⁻⁵, and the separation on the sky is large enough that the diffraction pattern of an 8-m telescope can resolve planet from star.

Earth analog imaging is much harder: 10⁻¹⁰ contrast at 0.1 arcsecond separation. No current instrument achieves it. Roman Coronagraph (2027+) demonstrates 10⁻⁸; the proposed Habitable Worlds Observatory (HWO, 2040s) is designed for 10⁻¹⁰. Direct imaging of habitable Earth analogs is still a generation away.

Coronagraphs and extreme adaptive optics

Three pieces of hardware make direct imaging possible:

1. Coronagraph. An optical device — first invented by Bernard Lyot in 1930 to observe the solar corona — that blocks light from a specific small angle. Modern stellar coronagraphs use either a focal-plane mask (a tiny opaque spot that intercepts star light) followed by a pupil-plane stop (Lyot stop) to suppress the diffraction pattern that ringed the mask. Variants include vortex coronagraphs (phase singularities), shaped pupil designs, and pupil-plane apodizers. Goal: rejection of starlight by factors of 10⁵–10¹⁰ within a designed range of working angles.
2. Adaptive optics. A deformable mirror that rapidly (1000+ Hz) corrects atmospheric turbulence using a wavefront sensor. Extreme AO (ExAO) — the regime needed for high-contrast imaging — has 1000+ actuators across an 8-m pupil and Strehl ratios above 80% at 1.6 μm. SPHERE (VLT) and GPI (Gemini) are the leading ground-based ExAO + coronagraph instruments.
3. Post-processing. Even after coronagraph and ExAO, speckles from imperfections in the optical train remain at planet-comparable brightness. Angular differential imaging (ADI; Marois 2006) takes advantage of telescope-pupil rotation relative to the sky during an observation: planets rotate with the sky; quasi-static speckles do not. Subtracting median-combined frames reveals the planet. KLIP and other PSF-subtraction algorithms further suppress speckle noise.

The HR 8799 system — a benchmark

HR 8799 is a young (~30 Myr), A5V star at 39 pc with mass ~1.5 M_☉ and luminosity ~5 L_☉. In 2008 Marois et al. used Keck and Gemini-North adaptive optics + ADI to detect three planetary-mass companions; in 2010 the same team announced a fourth, the innermost. The four planets:

Planet	Semi-major axis	Orbital period	Mass	Effective T
HR 8799 e	14.5 AU	~50 yr	7–10 M_J	~1100 K
HR 8799 d	24 AU	~100 yr	7–10 M_J	~1100 K
HR 8799 c	38 AU	~190 yr	7–10 M_J	~1100 K
HR 8799 b	68 AU	~460 yr	5–7 M_J	~900 K

The four planets are in approximate 1:2:4:8 mean-motion resonance, suggestive of orbital migration during formation. Multi-epoch imaging (2008–present) shows clear orbital motion — the inner planet has moved through a substantial fraction of its 50-year orbit.

JWST observations in 2023–24 produced the first space-based near-IR spectra of each planet (NIRSpec IFU at 1–5 μm). The spectra show molecular features of CO, CH₄, H₂O, and indicate slightly different atmospheric compositions for each planet — suggestive of either formation at different distances or different evolutionary histories.

Worked example: contrast and Strehl ratio

How much contrast does an 8-m ExAO instrument achieve, and is HR 8799 in range?

An ideal diffraction-limited 8-m telescope at 1.6 μm has angular resolution θ ~ 1.22 λ/D ≈ 50 mas. HR 8799 e at 14.5 AU and d = 39 pc has angular separation 14.5/39 = 0.37 arcsec = 370 mas, well beyond the diffraction limit.

The point-spread function (PSF) of a ground-based telescope without AO is dominated by atmospheric seeing (FWHM ~ 1 arcsec at typical sites). With ExAO Strehl S = 0.9, ~ 90% of light goes into a diffraction core ~ 50 mas wide, leaving 10% in a halo. The halo intensity at 370 mas from the core, for a typical AO PSF, is:

I_halo(θ) ≈ (1 - S) × I_peak × (λ/D / θ)²
         ≈ 0.1 × I_peak × (50/370)²
         ≈ 0.1 × I_peak × 0.018
         ≈ 1.8 × 10⁻³ × I_peak

Without coronagraph, this halo brightness at 370 mas is ~10⁻³ of stellar peak — much brighter than the 10⁻⁵ planet. Coronagraph adds a further suppression of ~10² in this halo region, giving residual stellar contamination at ~10⁻⁵. ADI / PSF subtraction in post-processing improves this by another factor of ~10–100, reaching final detection limits of ~10⁻⁶ at 370 mas separation.

HR 8799 e (10⁻⁵ contrast, 370 mas) is comfortably above the detection limit. HR 8799 b (68 AU, 1.7 arcsec separation, ~10⁻⁵ contrast at larger separation where AO halo is fainter) is even easier. All four planets are routinely detected; the technique works.

For an Earth analog at 10 pc: 0.1 arcsec separation, 10⁻¹⁰ contrast. Both numbers are orders of magnitude worse than HR 8799. Hence the need for next-generation missions.

Variants and current frontiers

• JWST coronagraphic imaging. NIRCam and MIRI coronagraphs achieve 10⁻⁷ contrast at 1″ separation in mid-IR — exquisite for cold, wide-separation planets. JWST has imaged HR 8799 (2023), Eps Indi A b (2024), 14 Her c (2024) and dozens of other directly imaged exoplanets.
• Eps Indi A b (2024). JWST-imaged Jupiter analog at 11 K, mass ~6 M_J around a 3.5 Gyr K dwarf at 3.6 pc. The first directly imaged planet around a star of cosmologically relevant age.
• Snowflake binary planetary-mass companions. Some 'planets' detected by direct imaging are massive enough (>13 M_J) to be brown dwarfs by deuterium-burning. The boundary is fuzzy; the 7–13 M_J range produces objects that look planetary but may have formed by gravitational instability rather than core accretion.
• Polarimetric direct imaging. SPHERE-ZIMPOL and IRDAP measure polarisation of scattered light from circumstellar disks and (in principle) planets. Polarised light is easier to discriminate from speckles than total intensity.
• Interferometric direct imaging. GRAVITY+ at the VLTI combines four 8-m telescopes to produce ~milliarcsecond imaging. Pointed at known imaged planets (β Pic b, HR 8799 e), it has measured astrometric positions to better than 10 μas and atmospheric spectra at higher resolution than single-dish.
• Habitable Worlds Observatory. NASA flagship space telescope (planned launch ~2040), 6-m UV/visible/NIR observatory with internal coronagraph designed for 10⁻¹⁰ contrast at 0.1 arcsec. Primary science: image habitable-zone Earth analogs at ~25 nearby stars, take spectra, search for biosignatures (O₃, H₂O, CH₄).

What direct imaging contributes

• Spectra of exoplanet atmospheres. Direct imaging is the only method that yields photons from the planet itself. JWST coronagraph + NIRSpec/MIRI now provides R~1000 spectra of imaged planets — enabling atmospheric retrieval and compositional inventory.
• Orbit determination by orbital motion. Multi-epoch imaging tracks the planet's position over years; combined with astrometric or radial-velocity reflex motion of the host star, gives mass and 3D orbit. HR 8799 e's mass is constrained to 7–10 M_J this way.
• Planet–disk relationships. Many imaged planets reside in circumstellar debris disks. Imaging both planet and disk reveals planet–disk gravitational interactions: planet-carved gaps (β Pic), eccentric rings (Fomalhaut), spirals (HD 100546). This is direct imaging of planet formation in action.
• Rotation rates. Doppler broadening of CO bands in directly imaged planet spectra reveals rotation period. β Pictoris b rotates in 8 hours — faster than any solar-system planet. Such measurements are unique to direct imaging.
• Habitable-world prospect. The long-term vision is to image and characterise Earth-like planets around the nearest stars, looking for biosignatures. HWO is the mission to do it.

Common pitfalls

• Confusing direct imaging with general exoplanet photography. Direct imaging means resolving the planet from its star — a single pixel of brightness alongside a masked stellar core. It is not a pretty picture of the planet's surface (none has been resolved).
• Calling all imaged objects "planets". Objects with mass >13 M_J burn deuterium and are formally brown dwarfs. Some direct-imaging discoveries straddle this boundary; classification depends on accurate mass measurement, often via a combination of luminosity, age, and atmospheric models.
• Forgetting the selection bias. Direct imaging samples a tiny corner of the exoplanet population: young, hot, massive, wide-orbit. The thousands of compact Kepler planets — small, cold, close-in — are entirely invisible to current direct imagers. Hot Jupiters are also invisible (too close in angular separation).
• Treating contrast and angular resolution as independent. Coronagraph designs trade contrast against inner working angle. A coronagraph that reaches 10⁻¹⁰ contrast at 0.5 arcsec might only reach 10⁻⁶ at 0.1 arcsec. Earth-imaging design has to win both.
• Imagining ground-based instruments can image Earth analogs. Atmospheric speckle noise puts a floor at ~10⁻⁷–10⁻⁸ even with ExAO. Earth-analog imaging requires space-based instruments above the atmosphere.