RISTRETTO: High-dispersion Coronagraphy Instrument

• RISTRETTO is a visible-light high-dispersion coronagraphy instrument that integrates extreme adaptive optics, a coronagraphic IFU, and a diffraction-limited spectrograph to detect exoplanets.
• It targets exoplanets in reflected starlight at very small angular separations, achieving contrasts of order 10⁻⁷ by spatially suppressing stellar leakage before spectral separation.
• The system serves as a pathfinder for future ELT instruments and enables detailed characterization of exoplanets, with Proxima b as a key science driver.

RISTRETTO is a proposed ESO Very Large Telescope visitor instrument for visible-light high-dispersion coronagraphy, designed to detect and characterize nearby exoplanets in reflected starlight by combining extreme adaptive optics, coronagraphy, single-mode-fiber integral-field sampling, and a diffraction-limited high-resolution spectrograph. In the instrument literature it is presented as an evolution of the original idea of coupling SPHERE and ESPRESSO, but reworked as an independent, AO-fed system optimized for planets at about $2\lambda/D$ from their host stars, with Proxima Centauri b as the main sizing case and scientific driver (Chazelas et al., 2020, Blind et al., 2024).

1. Scientific objective and observational regime

RISTRETTO was conceived for a regime that is difficult for both classical direct imaging and transit spectroscopy: nearby planets observed in reflected visible light at very small angular separations. For Proxima Cen b, the design papers describe the relevant separation as about 35 mas, i.e. $2\lambda/D$ at $\lambda=750$ nm, while the coronagraph paper gives a maximum separation of 37 mas, corresponding to about $2\lambda/D$ at $\lambda \approx 730$ nm on a VLT-class telescope (Blind et al., 2022, Blind et al., 24 Mar 2025). In that regime the planet/star contrast is of order $10^{-7}$, so the instrument must suppress stellar leakage spatially before using spectral information to separate planet and star.

The central measurement principle is high-dispersion coronagraphy. RISTRETTO combines front-end starlight suppression with a high-resolution spectrograph that uses Doppler shifts and line structure to disentangle the faint planetary spectrum from residual stellar leakage. Multiple design papers state that this spectral-analysis stage can provide an additional effective contrast gain of order $\sim 1000$ (Blind et al., 2022, Blind et al., 2024). The reflected-light contrast itself is written as

$$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$

or equivalently in later simulations as

$$\frac{F_p(\lambda)}{F_s(\lambda)} = \frac{R_p^2}{a^2}\,A_g(\lambda)\,g(\alpha),$$

linking detectability to planet radius, orbital distance, geometric albedo, and phase function (Lovis et al., 2022, Bugatti et al., 9 Sep 2025).

From the Proxima b use case, the instrument literature derives a consistent set of requirements: an inner working angle at or below about $2\lambda/D$, stellar coupling or raw contrast below $2\lambda/D$0, planet coupling above 50%, visible-band coverage over roughly 620–840 nm with $2\lambda/D$1, and spectral resolution above $2\lambda/D$2 (Lovis et al., 2024, Lovis et al., 2022). A recurrent design point is that the front end need not by itself reach planet-level contrast: the coronagraphic chain is required to deliver around $2\lambda/D$3, while the final separation of planetary and stellar spectra is completed by the high-resolution spectrograph.

2. System architecture and operating concept

RISTRETTO is consistently described as a three-part system: an extreme adaptive optics front end, a coronagraphic integral field unit, and a diffraction-limited visible spectrograph (Blind et al., 2022, Lovis et al., 2024). The front end is responsible for Strehl and low-order stability; the IFU provides spatial sampling and single-mode-fiber injection; the spectrograph performs the high-resolution analysis needed for Doppler separation.

Subsystem	Function	Representative characteristics
XAO front end	Visible diffraction-limited correction and low-order control	dual WFS concept, $2\lambda/D$4 kHz loop target
Coronagraphic IFU	Starlight suppression and spatial sampling	7 spaxels, single-mode fibers, $2\lambda/D$5 working region
Spectrograph	High-resolution visible spectroscopy	620–840 nm, early design $2\lambda/D$6, final manufacturing paper $2\lambda/D$7

The IFU geometry is central to the concept. Early design papers describe a 7-spaxel layout with 1 central lenslet centered on the star and 6 lenslets arranged in a ring at $2\lambda/D$8, feeding single-mode fibers (Blind et al., 2022). Later prototyping and manufacturing papers retain the 7-fiber architecture and emphasize that the single-mode feed is essential because the instrument operates at the diffraction limit (Restori et al., 2024, Chazelas et al., 2024). Because the full $2\lambda/D$9 ring is not covered uniformly in one orientation, the observing strategy includes two exposures with the IFU rotated by $\lambda=750$0; this yields an average $\lambda=750$1 geometrical transmission in the early coronagraph study, while the prototyping paper notes that the doubled integration time is equivalent to a 50% loss (Blind et al., 2022, Restori et al., 2024).

The system was originally framed as a pathfinder for reflected-light spectroscopy of nearby exoplanets on 8 m telescopes and, later, for similar architectures on the ELT. That pathfinder role is explicitly connected to ELT-ANDES and ELT-PCS in multiple papers (Lovis et al., 2022, Lovis et al., 2024).

3. Coronagraphic IFU and the PIAA Nuller

The distinctive front-end technology of RISTRETTO is the PIAA Nuller, or PIAAN, a coronagraphic integral field unit designed to feed a diffraction-limited spectrograph through single-mode fibers (Blind et al., 24 Mar 2025). It combines pupil remapping based on a phase-induced amplitude apodizer, moderate apodization, and a single-mode-fiber IFU so that on-axis stellar light is nulled in the fiber overlap integral while off-axis companion light remains efficiently coupled.

The basic coupling formalism is written as

$\lambda=750$2

where $\lambda=750$3 is the focal-plane electric field and $\lambda=750$4 is the single-mode-fiber fundamental mode (Blind et al., 24 Mar 2025). The instrument papers repeatedly emphasize that this is not a conventional opaque coronagraph: the PIAA reshapes the stellar PSF so that its diffraction rings become fainter and higher frequency, allowing destructive interference in the external fibers. In the concept papers, the resulting stellar null is described as $\lambda=750$5, while planet throughput remains high because the apodization is deliberately moderate (Blind et al., 2022, Blind et al., 24 Mar 2025).

The required performance for the Proxima b case is explicitly stated as contrast $\lambda=750$6 near $\lambda=750$7, over about 30% bandwidth, with companion transmission $\lambda=750$8 (Blind et al., 24 Mar 2025). The optimized PIAAN concept reaches a theoretical contrast of $\lambda=750$9 and transmission of 72% at $2\lambda/D$0 over 30% bandwidth, while a laboratory prototype reached $2\lambda/D$1 over the full bandwidth; the same paper also reports a refined prototype result around $2\lambda/D$2 after wavefront optimization (Blind et al., 24 Mar 2025). A prototyping paper from the following year summarizes the expected operating point as contrast down to $2\lambda/D$3, bandwidth $2\lambda/D$4, and transmission $2\lambda/D$5 at $2\lambda/D$6 (Restori et al., 2024).

Sensitivity to low-order aberrations is a recurring design constraint. The PIAAN study gives a tip-tilt tolerance of about $2\lambda/D$7, around $2\lambda/D$8 mas at 750 nm on a VLT, and a coma tolerance of about $2\lambda/D$9 RMS (Blind et al., 24 Mar 2025). The same paper attributes the practical limit to XAO residuals rather than to the coronagraph alone, with a required residual low-frequency wavefront error of approximately

$\lambda \approx 730$0

or roughly 8–10 nm RMS at 750 nm. This point is important because it corrects a common simplification: in RISTRETTO the coronagraph is not treated as an isolated contrast engine, but as a component whose performance is tightly coupled to visible-light XAO quality.

Prototype development has focused on hardware close to the intended final architecture. The IFU consists of 7 standard single-mode fibers in a hexagonal array with a 3D-printed microlens array fabricated directly on top; second-generation bundles manufactured in 2024 used a 125 $\lambda \approx 730$1m pitch and an “octopus” fiber breakout, while the first PIAA optics were manufactured by Nutek as two surfaces engraved into a single 20 mm thick CaF$\lambda \approx 730$2 rod, with measured $\lambda \approx 730$3 nm RMS (Restori et al., 2024).

4. Extreme adaptive optics and wavefront sensing

The XAO system is the other enabling half of the instrument. The visible high-contrast requirement is repeatedly stated as Strehl ratio above 70% at 750 nm, corresponding to total WFE below about 70 nm RMS, together with low-order WFE within 3 cycles at or below 10 nm RMS (Blind et al., 2024, Shinde et al., 2024). These figures are driven directly by the small working angle and the sensitivity of single-mode-fiber nulling to low-order residuals, especially pupil discontinuities and low wind effect.

A major design result is the “two-sensors-one-DM” architecture. In the detailed XAO paper, the Red WFS operates in the near-IR around 1500–1700 nm and closes the main loop for high Strehl, while the Blue WFS operates around 850–1000 nm and is dedicated to improving contrast by controlling low-order and pupil-fragmentation effects (Blind et al., 2024). The baseline pairing is an unmodulated pyramid WFS for the red channel and a Zernike WFS for the blue channel. The control law is written as

$\lambda \approx 730$4

with the blue sensor taking control of the first 100 low-order modes after the red loop is closed (Blind et al., 2024).

The preference for unmodulated sensing arises from both sensitivity and speed. The XAO architecture paper states that an unmodulated pyramid in the red channel is roughly 10 times more sensitive to low orders than the modulated case and provides about 2 magnitudes of margin for bandwidths greater than 100 nm in the NIR (Blind et al., 2024). A dedicated comparison of the 3-sided unmodulated PWFS, 4-sided unmodulated PWFS, and Zernike WFS found that dynamic range increases by more than a factor of 2 between 600 nm and 1600 nm for all three sensors, that the 3-sided pyramid has about 20 nm more dynamic range than the 4-sided version in the simulations, and that the ZWFS offers the least cross-talk but the smallest dynamic range and nearly zero open-loop optical gain at 600 nm (Shinde et al., 2024). This comparative result supports a design philosophy in which longer-wavelength pyramid sensing is preferred for the main loop, while the ZWFS is attractive for residual-phase fidelity and petal-mode control.

Low wind effect is treated as a first-order issue rather than an afterthought. The XAO architecture paper describes it as phase discontinuities produced by radiative cooling of the spiders and requires control to better than 10 nm PTT (Blind et al., 2024). In the same study the unmodulated pyramid is described as sensitive to pupil discontinuities but poor at controlling low wind effect in the tested cases, whereas the zWFS is described as “perfectly capable” of correcting those modes to the required level. Reduced phase shifts such as $\lambda \approx 730$5 are favored over the standard $\lambda \approx 730$6 because of better stability in turbulence and non-common-path conditions; the reported simulation failures were 5 out of 60 for the $\lambda \approx 730$7-zWFS, reduced to 3 out of 60 when only 50 modes were controlled, versus 16 out of 60 for the $\lambda \approx 730$8-zWFS (Blind et al., 2024).

Detector characterization for the wavefront sensors has also been reported. The C-RED ONE cameras purchased for RISTRETTO use SAPHIRA HgCdTe eAPD arrays with 320 × 256 pixels, 24 $\lambda \approx 730$9m pitch, and sensitivity from 0.8–2.5 $10^{-7}$0m (Shinde et al., 2024). In CDS mode, the measured maximum frame rate was 1738 Hz full frame and 44,667 Hz for a 32 × 32 subwindow, while readout noise converted to electrons at eAPD gain 50 was 0.92 $10^{-7}$1 for one camera and 0.79 $10^{-7}$2 for the other (Shinde et al., 2024). Those measurements are directly relevant to RISTRETTO’s requirement for kHz-class, low-noise near-IR wavefront sensing.

5. Spectrograph design, stability philosophy, and manufacturing

The back end of RISTRETTO is a single-mode visible high-resolution spectrograph. Early design papers describe it as a classical echelle or double-pass white-pupil instrument derived from the heritage of HARPS and ESPRESSO, but adapted for diffraction-limited operation with seven input channels rather than one (Chazelas et al., 2022). Later manufacturing papers retain the same stability philosophy while describing a more compact, self-contained realization intended for installation on a Nasmyth platform and possible transport to different telescopes (Chazelas et al., 2024).

The main spectrograph parameters have remained stable across the design sequence, with some evolution in the nominal resolving power. The 2020 and 2022 studies specify $10^{-7}$3 to $10^{-7}$4 over 620–840 nm, with a goal of $10^{-7}$5 and at least seven fibers (Chazelas et al., 2020, Chazelas et al., 2022). The 2024 manufacturing paper reports the final spectrograph resolution as 130000 in the 620–840 nm band, with seven diffraction-limited spaxels and a spectral format spanning 34 echelle orders, each carrying seven fibers (Chazelas et al., 2024). A 4k detector is required for the full order set, and the selected detector is a deeply depleted Teledyne e2v CCD231-84-x-E74 with astra-multi-2 coating (Chazelas et al., 2024).

A central geometrical solution is the slit format. The seven standard single-mode fibers are arranged into a long tilted slit, or in the final manufacturing paper as seven fibers on a nearly horizontal slit, so that the projected separation in cross-dispersion is sufficient to avoid order overlap on the detector (Chazelas et al., 2022, Chazelas et al., 2024). The trade-off is that the different fiber spectra are shifted relative to one another in wavelength, so the red and blue cutoffs are not identical for all channels. The design papers note, however, that the same geometry also increases detector separation between equal wavelengths from different fibers, reducing coherent interference effects (Chazelas et al., 2022).

The spectrograph is warm, thermally controlled, and under vacuum rather than cryogenic. Because the single-mode fibers have very small étendue, the full spectrograph fits in a vacuum tank only 85 cm in diameter (Chazelas et al., 2024). The manufacturing paper states that the instrument is designed for velocity stability at the 10 m s$10^{-7}$6 level, with vacuum better than $10^{-7}$7 mbar and thermal regulation better than 20 mK (Chazelas et al., 2024). Earlier optical requirements include line-position stability to 1/100 pixel over 24 h, line-spread-function stability over 24 h, and throughput above 40% (Chazelas et al., 2022).

Thermal engineering occupies a large fraction of the mature design. The enclosure uses a water-glycol loop between a control rack and the instrument enclosure, stabilized by a Peltier-based chiller with 200 W of cooling capacity and 600 W of heating capacity; inside the enclosure, a water/air heat exchanger, forced-air circulation, and 100 W heaters provide fine control (Chazelas et al., 2024). The insulating box is built from PUR with 0.8 mm steel skins; vacuum-tank feet are isolated from the spectrograph frame with G10 sheet and additionally regulated with flexible polyimide heaters. The control hardware uses Lakeshore 240 temperature input modules connected to a Beckhoff PLC via PROFIBUS, with heater power modulated by PWM through a solid-state relay using a 1 s cycle and 10 ms resolution (Chazelas et al., 2024).

The reported thermal performance is already close to operational requirements. In spring 2024 tests outside the Geneva Observatory, the enclosure achieved bottom-of-tank stability better than $10^{-7}$8 K and upper-tank stability of $10^{-7}$9 K, comfortably better than the $\sim 1000$0 K requirement (Chazelas et al., 2024). The same tests met the ESO Nasmyth thermal-load constraints of no more than 150 W to outside air and a skin-to-ambient difference of $\sim 1000$1C under the stated convective condition; the paper gives a derived thermal power limit of 45 W and reports that measurements stayed within the allowed envelope (Chazelas et al., 2024).

Manufacturing details also reflect the fact that the resolution is diffraction-limited. The main lenses were specified to an average quality of 2/20 rms, while the last lens, which also serves as the detector cryostat window and field lens, was polished with emphasis on a smooth $\sim 1000$2 decline in defect power to minimize order distortion (Chazelas et al., 2024). The grating is an off-the-shelf MKS/Grating Labs component and is described as the weakest optic in terms of quality, with a reported 2/4 PV wavefront error. At the time of that report, the main lenses had already been manufactured, whereas the cross-disperser prism and a test fiber link were still awaited (Chazelas et al., 2024).

The detector head is a differential vacuum cryostat, rigidly mounted to the optical table and softly connected to the tank by a flexible bellow, with cooling from a Thales LPT9310 cryocooler with active vibration reduction (Chazelas et al., 2024). Earlier spectrograph papers also describe an exposure meter, a supercontinuum flat-field source, uranium-neon wavelength calibration, and a single-mode Fabry–Perot etalon under development, all chosen to match the single-mode illumination path as closely as possible (Chazelas et al., 2022).

6. Simulations, projected science return, and broader significance

The principal science case is reflected-light spectroscopy of nearby exoplanets, especially Proxima b, but the projected return has broadened as the simulations and science studies have matured. Early science papers state that RISTRETTO should be able to characterize at least 10 exoplanets around nearby stars, including gas giants, warm super-Earths, and temperate rocky worlds, and also highlight accreting protoplanets, Solar System targets, and resolved stellar surfaces as important secondary cases (Lovis et al., 2022, Lovis et al., 2024). The reflected-light papers emphasize measurements of broadband and chromatic albedo, orbital inclination, and true mass, with molecular absorption features such as $\sim 1000$3, $\sim 1000$4, and $\sim 1000$5 as longer-term targets (Lovis et al., 2024).

Published detectability estimates are strongly assumption-dependent and have evolved with the realism of the simulations. An early performance estimate based on Proxima b gave $\sim 1000$6 in 40 nights under assumptions including about 5% total transmission (Blind et al., 2022). A later overview paper states that if Proxima b has an Earth-like atmosphere, it may be detectable in about 5 nights at its brightest accessible orbital phase (Lovis et al., 2024). End-to-end simulation papers then move from night-based estimates to exposure-hour budgets. One simulator study, based on 11 nights and 99 hours, reports $\sim 1000$7 for a planet-bearing model and recovers the injected inclination as $\sim 1000$8 for a true $\sim 1000$9 case (Bugatti et al., 2024). A later analysis using realistic epochs, AO/coronagraph coupling maps, PyEchelle, and nested sampling reports reflected-light detection of Proxima b in about 55 hours of observing time and states that $$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$0 and $$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$1 can be detected in about 85 hours under an Earth-like atmosphere assumption (Bugatti et al., 9 Sep 2025).

The simulation methodology itself has become part of the project’s significance. The RISTRETTO simulator paper uses realistic host-star spectra from Expecto, a 3D Global Climate Model plus PICASO for high-resolution reflected-light spectra, barycentric correction with Barycorrpy, and PyEchelle to generate synthetic 2D echelle frames (Bugatti et al., 2024). The later Proxima-b study adds a direct forward-modeling analysis of long- and short-exposure ratios and a model hierarchy ranging from no planet to constant albedo, chromatic albedo, a full albedo spectrum, and simple molecular absorption (Bugatti et al., 9 Sep 2025). This suggests a transition from proof-of-detectability studies toward observation-specific inference strategies.

Beyond reflected light, RISTRETTO has become a serious H$$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$2 instrument for forming planets. A 2026 study on accreting protoplanets uses the built-in combination of XAO, coronagraphy, seven-spaxel IFU sampling, and $$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$3 spectroscopy to derive 5$$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$4 line-flux detection limits of $$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$5 in a 1 hour exposure for analogs of PDS70b, PDS70c, WISPIT2b, and 2MJ1612b (Blackman et al., 12 Jun 2026). That paper argues that RISTRETTO can not only spatially isolate planetary H$$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$6 from stellar contamination but also resolve the line profile well enough to constrain pre-shock density $$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$7, inflow velocity $$C = \left(\frac{R_p}{a}\right)^2 A_g(\lambda)\, g(\alpha),$$8, accretion rate, and flow geometry.

The instrument’s broader role is explicitly pathfinding. Several papers frame RISTRETTO as a demonstrator for future ELT instrumentation, particularly the SCAO-IFU mode of ELT-ANDES and the future ELT-PCS (Lovis et al., 2022, Lovis et al., 2024). A 2026 review places it in the same lineage as HARPS, ESPRESSO, NIRPS, and ANDES, while distinguishing it by its direct reflected-light objective rather than precision radial-velocity or transmission-spectroscopy use (Bouchy et al., 10 Apr 2026). In that sense, RISTRETTO is not merely a compact high-resolution spectrograph: it is a coordinated attempt to make reflected-light spectroscopy of nearby non-transiting planets practical from the ground, at the diffraction limit of the VLT.