Observational Techniques & Instrumentation
Atmospheric Dispersion Corrector: Undoing the Sky's Prism
Point a telescope 60° from the zenith and a star that should be a point becomes a tiny rainbow smear: blue light lands nearly 3 arcseconds higher in the sky than red light, spread across the visible band. An atmospheric dispersion corrector (ADC) is the pair of counter-rotating glass prisms that erases that smear, gluing the colors back into a single sharp image before they reach the detector.
The Earth's atmosphere acts as a weak, wavelength-dependent prism because the refractive index of air varies slightly with color. An ADC introduces an equal and opposite prism — tunable in strength and orientation — so that the net chromatic spread is driven back toward zero. Without it, adaptive-optics imaging, high-resolution spectroscopy, and precise astrometry all degrade the moment you observe away from the zenith.
- RegimeGround-based optical/NIR imaging & spectroscopy off-zenith
- Key number~3 arcsec visible-band spread at 60° zenith angle
- Driven byWavelength dependence of air's refractive index, (n−1) ≈ 2.8×10⁻⁴
- First describedDifferential refraction quantified by A. Filippenko (1982); Amici-prism ADCs from early 20th c.
- Observed withAO imagers (SPHERE, GPI), fiber/slit spectrographs (ESPRESSO, MUSE, DESI)
- Matters forAO PSF quality, spectrophotometry, exoplanet RV, astrometry
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What it is and why it matters
An atmospheric dispersion corrector (ADC) is an optical device — almost always a pair of prisms — placed in a telescope's beam to cancel the chromatic spreading of starlight caused by the atmosphere. The atmosphere refracts (bends) light passing through it, and because air's refractive index depends slightly on wavelength, blue light is bent more strongly than red. Away from the zenith this turns a point source into a short vertical spectrum, with the bluest light displaced toward the zenith and the reddest away from it.
That smear is fatal for modern high-precision work. An adaptive-optics system can deliver a diffraction-limited core of ~40 milliarcseconds on an 8 m telescope, yet uncorrected dispersion at moderate airmass spreads colors by tens of that width. In spectroscopy, only part of the source's light enters a slit or fiber, and which part depends on wavelength — corrupting relative fluxes. The ADC restores the image so that all colors again converge to one place.
The mechanism, step by step
Air's refractive index exceeds unity by a small amount, (n−1) ≈ 2.8×10⁻⁴ at optical wavelengths and standard conditions. Snell's law then bends incoming rays toward the vertical by an angle that, for a plane-parallel atmosphere, scales as R ≈ (n−1)·tan z, where z is the zenith angle. Because n itself falls with increasing wavelength (normal dispersion, dn/dλ < 0), red rays bend less than blue, and the two separate.
An ADC introduces a controllable, opposite prism. The standard design is two identical Amici prisms (each a cemented triplet of crown and flint glasses) that disperse light with almost no net beam deviation. Rotating the two prisms in opposite directions about the optical axis varies the resultant dispersion vector from zero (prisms opposed) to a maximum (prisms aligned), and orients it along the vertical. The observer tips the prisms to produce exactly the magnitude the current airmass demands, pointed opposite the sky's dispersion, so the two cancel.
Characteristic numbers, scales, and the key relation
The governing relation, following Filippenko (1982) and Smart's refraction theory, is the differential refraction between two wavelengths:
ΔR(λ) ≈ 206265″ × [ n(λ) − n(λ₀) ] × tan z
where n(λ) is the (pressure- and temperature-corrected) refractive index of air and λ₀ a reference wavelength. The tan z factor makes dispersion vanish at the zenith and grow rapidly toward the horizon. Numerically, across the visible band (~0.4–0.8 μm) at good sites, ΔR is about 0.9 arcsecond at z = 30°, roughly 1.6 arcseconds near z = 45°, and nearly 3 arcseconds by z = 60°. In the near-infrared the effect is several times smaller (~0.2″ at z = 30°) because air is less dispersive there. Total refraction, by contrast, is ~34 arcminutes at the horizon — the differential piece an ADC corrects is the small chromatic residual riding on top of that large bulk bend.
How it is deployed and detected
ADCs sit inside instrument fore-optics on essentially every major ground-based facility. On the VLT, adaptive-optics imagers such as SPHERE and integral-field spectrographs like MUSE carry ADCs; Gemini's GPI, Subaru, Keck, and the survey spectrograph DESI all use them, as will the ELT-class instruments (e.g., HROS on TMT, MAVIS on the VLT). Two geometries dominate: the linear ADC (two prisms sliding relative to each other) and the more common rotating ADC (counter-rotating Amici prisms).
The signature of a missing or mis-set ADC is unmistakable: a stellar image elongated along the parallactic angle (the sky-vertical), with a blue tip and red tail. Spectrophotometric standards taken through a fixed slit show wavelength-dependent slit losses; fiber-fed radial-velocity instruments show apparent velocity drifts. Modern control loops set the ADC continuously from the telescope's pointing, temperature, and barometric pressure, so residual dispersion is held to a small fraction of the diffraction limit throughout an exposure.
Where it operates and how it differs from related effects
ADCs matter for any ground-based optical or near-infrared observation taken away from the zenith — which is to say almost all of them, since targets spend little time overhead. They are indispensable for diffraction-limited AO (where the corrected PSF is tiny), for slit and fiber spectroscopy (to keep all colors on the aperture), for precision astrometry, and for exoplanet detection by radial velocity and direct imaging.
It is important to separate the effect from its cousins. Total atmospheric refraction is the large achromatic shift of a star's apparent position toward the zenith — corrected by pointing, not an ADC. Atmospheric extinction is wavelength-dependent dimming, a photometric not a geometric effect. Seeing is stochastic blurring from turbulence. Dispersion is the deterministic, color-splitting residual — the only one of these an ADC addresses. Space telescopes, above the atmosphere entirely, need no ADC; their chromatic errors come from optics, not air.
Open questions and significance
The physics of atmospheric dispersion is settled, but engineering it away at the level the next generation demands is not. On 30–40 m Extremely Large Telescopes, the diffraction core shrinks to a few milliarcseconds, so an ADC must hold residual dispersion to milliarcsecond accuracy across ever-wider bandpasses — hard because a two-glass Amici design cannot make n_ADC(λ) perfectly track the atmosphere's real, humidity- and pressure-dependent dispersion curve. Broadband and multi-glass designs (e.g., proposed HROS-TMT and MAVIS correctors) chase this achromatic ideal.
Open practical challenges include real-time knowledge of local water-vapor content (which shifts n at the ~10⁻⁸ level relevant to extreme-precision RV), correcting the near-infrared where prisms of suitable transmission are scarce, and minimizing the polarization and throughput penalties the extra glass introduces. For high-contrast exoplanet imaging and 1 cm s⁻¹ radial-velocity searches, uncorrected chromatic residuals are now a leading systematic — making the humble ADC a quiet enabler of some of astronomy's most demanding measurements.
| Zenith angle z | Total refraction | Visible-band dispersion Δ (blue−red) | Impact |
|---|---|---|---|
| 0° (zenith) | 0″ | ≈ 0″ | No correction needed |
| 30° | ≈ 34″ | ≈ 0.9″ | Marginal for AO; ADC on |
| 45° | ≈ 58″ (≈1′) | ≈ 1.6″ | Exceeds seeing-limited core |
| 60° | ≈ 100″ | ≈ 2.8″ | Severe; ADC essential |
| 70° | ≈ 160″ | ≈ 4.5″ | Beyond most ADC design range |
Frequently asked questions
Why does the atmosphere disperse starlight like a prism?
Air has a refractive index slightly above 1 (about 1.00028 in the optical), and that index depends weakly on wavelength — it is higher for blue light than for red. Snell's law therefore bends blue rays a bit more than red as they enter the atmosphere at an angle. Away from the zenith the two colors emerge along slightly different directions, spreading a point source into a short vertical spectrum, exactly as a glass prism would.
How much dispersion are we talking about — is it really visible?
Yes, at moderate airmass it dwarfs the resolution of a good telescope. Across the visible band the blue-to-red spread is roughly 0.9 arcsec at 30° from the zenith, about 1.6 arcsec near 45°, and nearly 3 arcsec by 60°. Since an 8 m adaptive-optics system delivers a ~0.04 arcsec core, uncorrected dispersion can be tens of times the diffraction limit.
What is an Amici prism and why use two of them?
An Amici prism is a cemented stack of crown and flint glass designed to disperse light while deviating the beam very little (a 'direct-vision' prism). Using two identical Amici prisms lets you tune the correction: counter-rotating them about the optical axis varies the net dispersion continuously from zero (opposed) to a maximum (aligned), and lets you point it along the vertical to oppose the sky's dispersion at any airmass.
Does the zenith angle really control everything?
Almost. Atmospheric refraction — and hence its chromatic part — scales as tan(z), so it is zero straight overhead and climbs steeply toward the horizon. That is why an ADC is idle at the zenith and works hardest at high airmass, and why control loops compute the required prism setting directly from the telescope's current pointing, along with local pressure and temperature.
How is an ADC different from correcting atmospheric refraction generally?
Total refraction is a large, essentially color-independent shift (up to ~34 arcminutes at the horizon) that moves a star's apparent position toward the zenith; telescopes simply point to the refracted position. An ADC ignores that bulk shift and cancels only the small wavelength-dependent residual — the color splitting — which is the part that blurs images and biases spectra.
Why do near-infrared instruments need less dispersion correction?
Air's refractive index changes much more slowly with wavelength in the infrared than in the blue-visible, so the differential refraction is several times smaller — around 0.2 arcsec at 30° zenith angle versus roughly 1 arcsec in the optical. NIR ADCs are still used on ELT-class systems, but the demands are gentler, and finding prism glasses with good IR transmission is the harder constraint.