Nulling Interferometry

The problem: a star a billion times too bright

Picture a Sun-like star ten parsecs away and an Earth-like planet orbiting it at one astronomical unit. On the sky the planet sits about 0.1 arc-second from the star — close, but not impossibly so. The killer is not the angle; it is the brightness ratio. In reflected visible light the planet is roughly 10⁻¹⁰ as bright as the star: ten billion times fainter, drowned in the diffraction halo and scattered glare of its host. A normal telescope can resolve the separation and still see nothing, because the faint pixel you want is buried under a torrent of starlight spilling out of the bright pixel next door.

Coronagraphs attack this by masking the star inside a single telescope. Nulling interferometry attacks it differently and arguably more elegantly: it uses the wave nature of light itself to make the star cancel out. Take two separate telescopes, bring their beams together, but first delay one beam by exactly half a wavelength. The two wavefronts coming straight down the line of sight to the star now arrive in perfect antiphase and annihilate each other — a deep central null. The planet, sitting slightly off to the side, traveled a different path length to the two telescopes; with the right baseline its light arrives in phase and stacks up on a bright fringe. That is nulling interferometry: the star erases itself, and the planet emerges.

The Bracewell nuller and its transmission map

Ronald Bracewell proposed the idea in a 1978 Nature letter, "Detecting nonsolar planets by spinning infrared interferometer." His scheme is the two-element nuller. Two telescopes separated by a baseline vector B each collect light; one beam passes through an achromatic π phase shifter; the beams are then summed onto a single detector.

For a source at small angle θ off the array axis (measured along the baseline), the geometric path difference between the two telescopes is B·θ, which corresponds to a phase difference of 2πBθ/λ. Add the deliberate π shift and combine. The transmitted intensity as a function of position on the sky is

T(θ) = sin²(π B θ / λ)

Read that map carefully. On axis, θ = 0, so T = 0 — the star sits exactly in the null and is cancelled. The transmission then climbs, reaching its first maximum of T = 1 where the argument equals π/2, i.e.

θ_peak = λ / (2 B)     (first bright fringe)

The transmission pattern is a set of light and dark fringes laid across the sky, like the bars of a comb, with angular spacing λ/B. The trick of designing a nuller is to choose the baseline B so that the planet's expected angular separation lands on or near a bright fringe while the star sits in the dark central null. For a 10 µm observation and a 0.1 arc-second (≈ 5 × 10⁻⁷ rad) planet, the baseline that puts the planet on the first peak is B = λ/(2θ) ≈ 10 m — comfortably achievable on the ground or with formation-flying spacecraft.

Why the mid-infrared changes the game

Nulling does not make the contrast problem disappear; it suppresses the star, but you still need the planet to be detectably bright relative to whatever leaks through. This is why every serious nuller targets the thermal infrared rather than visible light.

The reason is the planet's own glow. A temperate planet at ~255 K (Earth's effective temperature) radiates as a blackbody peaking near 11 µm by Wien's law, λ_max ≈ 2898/T µm·K. Meanwhile the star's Planck curve, which towers over the planet in the optical, falls onto its Rayleigh-Jeans tail in the mid-IR. The net effect is that the planet-to-star flux ratio improves dramatically:

Band	Wavelength	Earth/Sun contrast	Why
Visible	0.5 µm	~2 × 10⁻¹⁰	Reflected sunlight, tiny planet
Near-IR	1–2 µm	~10⁻⁹	Mostly still reflected light
Mid-IR	10 µm	~7 × 10⁻⁷	Planet thermal emission peaks; star on R-J tail
Far-IR	20 µm	~10⁻⁶	Planet still bright; lower spatial resolution

So the mid-IR buys roughly three to four orders of magnitude of contrast for free. It also happens to be where the molecular fingerprints of a biosphere live: the 9.6 µm ozone band, the 15 µm CO₂ band, and the 6.3 µm water band. A mid-IR nuller spectrum of a temperate planet would in principle reveal whether its atmosphere is in chemical disequilibrium — the classic Lovelock biosignature argument. That triple payoff (easier contrast, achievable baselines, biosignature lines) is why Bracewell, ESA's Darwin, and NASA's TPF-I all settled on the thermal infrared.

Stellar leakage and the depth of the null

If the star were a true point source on axis, the null would be infinitely deep. It is not — it is a disk of angular radius θ*. Its limb sits slightly off-axis and therefore leaks a little light through the bottom of the null. Expanding the transmission map for a small uniform stellar disk gives the geometric stellar leakage:

L_geom ≈ (1/4) (π B θ* / λ)²

For the Sun seen from 10 pc, θ* ≈ 2.3 × 10⁻⁹ rad. With B = 10 m and λ = 10 µm, the argument πBθ*/λ ≈ 7 × 10⁻³, so L_geom ≈ 1 × 10⁻⁵. That residual is comparable to the planet signal you are hunting, which is why deeper nulls demand either shorter baselines (smaller leakage but worse resolution) or careful subtraction of a known, modelled leakage term.

Three other effects fight you. The π phase shift must be achromatic — exactly π across the whole 8–13 µm band — because a phase error δφ fills the null to a floor of order (δφ)²/4; a 1 % phase error already limits the null to ~2.5 × 10⁻⁵. Amplitude mismatch between the two beams contributes (δA/2A)². And on the ground, residual atmospheric piston — random optical-path jitter between the two telescopes — must be measured and corrected to a small fraction of a wavelength in real time. Stacking these, ground-based instruments reach raw nulls of about 10⁻³ to 10⁻⁴; the Darwin specification called for a stabilised null of 10⁻⁵ to 10⁻⁶ in the benign environment of space.

Telling a planet from a dust disk

Even a perfect null leaves a serious confusion problem: the exozodiacal dust disk. Most stars are surrounded by a faint cloud of warm debris dust — our own Solar System's zodiacal cloud is the local example — and that dust is bright in exactly the mid-IR a nuller observes. A dust disk is roughly axisymmetric about the star, so it leaks through the symmetric transmission map in a smooth, orientation-independent way. A planet is a single off-axis point. The decisive difference is symmetry, and you exploit it with modulation.

In Bracewell's original "spinning" scheme, you rotate the whole array on the sky. As the baseline turns, the comb of fringes sweeps over the planet, so the planet's transmitted flux oscillates at the rotation frequency, while the symmetric dust signal stays constant. Synchronous (lock-in) detection at the spin frequency isolates the planet. More advanced four-telescope architectures — the Angel cross and the Mariotti/Darwin configurations — add internal modulation: by switching the relative phases of the four beams, the instrument chops between two transmission maps whose difference is antisymmetric. A point planet produces a signal in the chopped difference; a symmetric dust cloud cancels out completely. This is what let mission studies claim sensitivity to an Earth-analogue embedded in an exozodi several times brighter than our own.

Nulling interferometer vs coronagraph

Nulling and coronagraphy are the two great families of high-contrast starlight suppression. They are complementary, not competing — they win in different regimes.

Property	Nulling interferometer	Coronagraph
Suppression mechanism	Destructive interference between apertures	Focal-plane mask + pupil stop in one aperture
Apertures	Two or more separated telescopes	Single filled aperture
Inner working angle set by	Baseline B (can be tens of m)	Telescope diameter D (a few λ/D)
Native band	Mid-infrared (8–20 µm)	Visible / near-IR
Planet/star contrast targeted	~10⁻⁵ to 10⁻⁷ (thermal)	~10⁻⁸ to 10⁻¹⁰ (reflected)
Dust rejection	Modulation / array rotation	Spatial — disk resolved directly
Image vs measurement	Fringe response, model-fit position	Direct image of the field
Biosignature lines	O₃ 9.6 µm, CO₂ 15 µm, H₂O 6.3 µm	O₂ 0.76 µm, H₂O near-IR

The headline distinction is the inner working angle. A coronagraph's smallest resolvable separation is a few λ/D, tied to the diameter of a single mirror — pushing it down means building an ever-larger monolithic telescope. A nuller's resolution is set by the baseline B between telescopes, which can be made much larger than any single mirror, so two modest telescopes flown tens of metres apart resolve separations a single instrument never could.

Real instruments and hard numbers

• Keck Interferometer Nuller (KIN). Combined the two 10 m Keck telescopes on Mauna Kea across an 85 m baseline in the N band (8–13 µm), operating from roughly 2004 to 2012. Its main science was measuring warm exozodiacal dust around nearby stars to levels of a few hundred "zodis," informing the design budgets for future planet-finders.
• Large Binocular Telescope Interferometer (LBTI). Nulls the two 8.4 m LBT mirrors on Mount Graham across a 14.4 m centre-to-centre baseline in the N band. Its HOSTS survey (Hunt for Observable Signatures of Terrestrial Systems) observed dozens of nearby stars and placed the tightest limits to date on habitable-zone dust, with a median sensitivity of a few zodis — crucial because too much dust would swamp any future Earth-imaging mission.
• Palomar Fiber Nuller and GENIE/PIONIER-class testbeds. Single-aperture and ground-based demonstrators that proved achromatic phase shifting and deep-null stabilisation in the lab and on sky.
• Darwin (ESA) and TPF-I (NASA). The flagship free-flying nuller concepts of the 2000s. Both envisioned three to four 1.5–4 m telescopes formation-flying tens to hundreds of metres apart at the Sun–Earth L2 point, combining their beams to null nearby Sun-like stars and detect Earth-analogues in thermal emission, with spectroscopy of O₃, CO₂ and H₂O. Both were studied to advanced design maturity and cancelled around 2007 on cost and technology-readiness grounds. The Large Interferometer For Exoplanets (LIFE) is the modern revival of the same architecture.

A representative space-nuller budget: four telescopes, baselines tunable from ~10 to ~200 m, achromatic π shift good to a null of 10⁻⁵, observing 10–20 µm, integrating for hours per target to pull a 10⁻⁷-contrast planet out of the residual leakage plus exozodi. The path-length between spacecraft must be held stable to a few nanometres — a fraction of a wavelength — which is precisely the formation-flying metrology challenge that, more than anything, kept these missions on the drawing board.

Common misconceptions and edge cases

• "The null is a single dark spot." It is not a spot but a dark fringe — a line (for two telescopes) or a more complex pattern (for four) of zero transmission that extends across the field, flanked by bright fringes spaced λ/B. The star is parked on the central dark fringe; the planet must be coaxed onto a bright one.
• "A nuller takes a picture of the planet." A two-element nuller produces no image — it delivers a single number, the flux passing the transmission map, as a function of array orientation and baseline. The planet's position and brightness are recovered by fitting that modulated signal. Imaging requires many baselines (aperture synthesis) or many telescopes.
• "More baseline is always better." Longer B sharpens resolution and pushes the first bright fringe inward, but it also increases geometric stellar leakage as B² and can place higher-order bright fringes on top of the dust disk, raising the background. The baseline is a tuned compromise, not a free parameter to maximise.
• "The phase shift is just a mirror tilt." Achieving a π shift that is identical across an octave of wavelength is genuinely hard. Simple geometric delays are chromatic. Real achromatic phase shifters use field-reversal optics, dispersive-plate pairs, or through-focus designs, and getting them right to better than ~1 % is what sets the null floor.
• "Exozodiacal dust is a minor nuisance." It is the dominant astrophysical noise source for a nuller. A cloud just a few times denser than our own zodiacal light can raise the mid-IR background enough to bury an Earth-twin, which is exactly why the LBTI HOSTS survey to measure that dust was considered a prerequisite for any future mission.