NIRPS: Precision Near-IR Exoplanet Spectroscopy

• NIRPS is a high-resolution near-infrared spectrograph with adaptive optics and few-mode fibre design, enabling precise radial-velocity measurements for exoplanet detection.
• It operates in tandem with HARPS to provide simultaneous visible and near-IR observations, enhancing studies of M dwarf systems and exoplanet atmospheres.
• Mitigation of modal noise through AO tip-tilt scanning and mechanical scrambling achieves RV stability near 1 m/s, showcasing advanced instrument performance.

Near InfraRed Planet Searcher (NIRPS) is a fibre-fed, adaptive-optics-assisted, high-resolution near-infrared spectrograph on the ESO 3.6-m telescope at La Silla, developed for precision radial-velocity measurements and high-resolution exoplanet spectroscopy. It was conceived as the near-infrared counterpart to HARPS, with simultaneous operation through a shared front end and dichroic, and it is distinctive in combining single-conjugate adaptive optics with a few-mode fibre architecture. Early design papers describe an operating domain in the YJH bands from about 980 nm to 1800 nm, while later on-sky characterization reports continuous coverage from 972.4 to 1919.6 nm, with two missing orders near 1.38–1.40 μm (Blind et al., 2017, Bouchy et al., 29 Jul 2025).

1. Instrument concept and scientific role

NIRPS is a precision radial-velocity instrument mounted on the ESO 3.6-m telescope at La Silla and designed to operate together with HARPS, which covers the visible domain. In the combined configuration, HARPS receives the visible beam and NIRPS the near-infrared beam, enabling simultaneous observations from the blue visible to the H band. The science case is centered on low-mass planets around M dwarfs, for which the near-infrared offers higher stellar flux and a favorable balance between planet-induced radial-velocity amplitudes and stellar-activity mitigation (Blind et al., 2022, Artigau et al., 2024).

The stated science goals include the detection and characterization of Earth-mass planets in the habitable zones of nearby M dwarfs, radial-velocity confirmation of transiting planets from surveys such as TESS and PLATO, and high-resolution transmission and emission spectroscopy of exoplanet atmospheres. Multiple papers associate the instrument with a target precision of about $1~\mathrm{m\,s^{-1}}$ or better for precision radial velocities, and later on-sky papers report stable radial-velocity performance at the level of $1~\mathrm{m\,s^{-1}}$ over several weeks, with a Proxima measurement reaching 77 cm/s (Blind et al., 2022, Bouchy et al., 29 Jul 2025).

The instrument also occupies a specific geographic and technical niche. It is described as the only planned near-infrared precision radial-velocity spectrograph in the Southern hemisphere, and its architecture differs from other near-infrared radial-velocity facilities by adopting an AO-fed few-mode fibre design rather than a seeing-limited, fully multi-mode approach. This choice was made to preserve both high spectral resolution and a compact cryogenic spectrograph on a 3.6-m telescope (Blind et al., 2017, Blind et al., 2022).

2. Optical architecture, front end, and observing modes

The NIRPS Front-End is installed at the Cassegrain focus and replaces the old telescope adaptor, while also interfacing with the existing HARPS front end. A HARPS/NIR dichroic in the telescope beam transmits visible light to HARPS and reflects the red and near-infrared beam toward NIRPS. Commissioning reported that this dichroic is essentially transparent to HARPS from an RV standpoint, with no measurable RV offset (Blind et al., 2022).

The front end includes a Single-Conjugate Adaptive Optics system optimized not for wide-field imaging but for high ensquared energy into a small on-axis fibre. The wavefront sensor is a $14\times14$ Shack-Hartmann operating from 700 to 950 nm on an OCAM2 EMCCD, and the deformable mirror is an ALPAO DM241. The AO loop is operated at up to $1~\mathrm{kHz}$, with later commissioning papers describing a measured loop delay of 3 frames and control in a Karhunen–Loève basis with up to 100 modes controlled over the 136 active actuators in the pupil (Blind et al., 2022).

Two fibre modes define the principal observing configurations. The High Accuracy Fiber (HAF, or HA mode) is a small AO-fed fibre intended for maximum radial-velocity precision; the High Efficiency Fiber (HEF, or HE mode) is a larger fibre used for fainter targets or poorer conditions. Across the literature, the HA fibre is described as 0.4″ on sky with a 29 μm octagonal core, while the HE mode is 0.9″ on sky with a larger octagonal input and a pupil-slicing rectangular fibre in the spectrograph feed. The HA mode is associated with spectral resolution values ranging from a median $R \approx 88{,}000$ to a provided resolving power of 90,000, while early design discussions describe the HAF concept as $R \approx 100{,}000$. The HE mode is reported at $R \approx 75{,}200$, $R \approx 75{,}000$, or effectively $R \approx 80{,}000$, depending on the paper and the exact definition used (Blind et al., 2017, Bazinet et al., 8 Aug 2025, Bouchy et al., 29 Jul 2025).

This architecture supports a compact cryogenic spectrograph. The few-mode HAF choice is explicitly linked to a compact white-pupil spectrograph and cryostat, smaller than a seeing-limited design would require at the same resolution. A plausible implication is that NIRPS represents a specific compromise between throughput, optical stability, and instrument volume rather than an attempt to maximize any single quantity in isolation.

3. Few-mode fibres, AO coupling, and the underlying formalism

A defining characteristic of NIRPS is its deliberate use of few-mode fibres. For a circular step-index fibre, the normalized frequency is

$V = \frac{2\pi a}{\lambda}\,\mathrm{NA},$

and the approximate number of guided modes is

$1~\mathrm{m\,s^{-1}}$0

For the 29 μm HAF, the number of modes in the near infrared is reported as $1~\mathrm{m\,s^{-1}}$1 at the long-wavelength end and up to $1~\mathrm{m\,s^{-1}}$2 at the short-wavelength end, placing NIRPS in a regime intermediate between single-mode operation and the $1~\mathrm{m\,s^{-1}}$3–$1~\mathrm{m\,s^{-1}}$4 modes typical of classical seeing-limited multi-mode astronomical fibres (Blind et al., 2017).

In this regime, fibre coupling must be treated coherently as overlap with discrete guided modes rather than by purely geometrical encircled-energy arguments. The coupling efficiency into mode $1~\mathrm{m\,s^{-1}}$5 is written as

$1~\mathrm{m\,s^{-1}}$6

and the total coupling is

$1~\mathrm{m\,s^{-1}}$7

To propagate AO residuals, the pupil field is modified as

$1~\mathrm{m\,s^{-1}}$8

These relations are used in the NIRPS coupling analysis, following the formalism cited from Horton et al. (2007) (Blind et al., 2017).

The measured and simulated behavior emphasizes why AO is central even though NIRPS is not single-mode. The 0.4″ HAF is small on sky, and coupling depends on the instantaneous point-spread function structure. The AO system is therefore used to recover light from AO speckles into the fibre core and to maintain a mean coupling of about 50% over YJH for stars as faint as $1~\mathrm{m\,s^{-1}}$9 under median seeing. Later commissioning results report that for bright stars ($14\times14$0), encircled energy in 0.5″ and 0.9″ apertures matched simulations, even though the measured Strehl ratio at 1400 nm plateaued near 35% rather than the $14\times14$1 expected from simulation under the observed seeing. That discrepancy was interpreted as an additional residual wavefront error of approximately 130–170 nm RMS (Blind et al., 2017, Blind et al., 2022).

The few-mode concept is repeatedly presented as an alternative to both single-mode and classical multi-mode architectures. Relative to single-mode injection, it relaxes sensitivity to residual wavefront errors, non-common-path aberrations, and central obscuration, while still permitting a compact high-resolution spectrograph. Relative to a large seeing-limited fibre, it preserves spectral resolution and cryogenic compactness. The principal cost is modal noise.

4. Modal noise and scrambling strategy

Modal noise is the dominant technical drawback of the NIRPS few-mode architecture. In the instrument papers, it is defined as instability of the illumination at the fibre output, in both the near field and far field, producing instability of the spectrograph line-spread function and ultimately radial-velocity noise. The underlying mechanism is the coherent interference of a small number of guided modes with different propagation constants, so that changes in injection conditions, fibre stress, bending, temperature, and wavelength alter the output speckle pattern (Blind et al., 2017, Blind, 2022).

Because only a small subset of modes may be significantly excited under high-Strehl injection, the effective modal diversity can be even smaller than the total guided-mode count. The 2017 study notes that in some conditions only 30–40% of the modal content may be significantly excited, and the 2022 modal-noise paper reports that a raw single HAF fibre without mitigation would generate RV fluctuations of order $14\times14$2 under realistic injection variations. That level is incompatible with the instrument’s precision goals (Blind et al., 2017, Blind, 2022).

The mitigation strategy validated for NIRPS combines AO-based scanning and dedicated fibre scrambling hardware. Several approaches were studied. Mode-selective excitation with the deformable mirror was shown to be possible in principle for specific modes such as LP$14\times14$3, but it reduced total coupling by more than a factor of 2 and was deemed impractical for broadband astronomical spectroscopy. Injection of defocus or higher-order Zernike aberrations also degraded coupling by more than 10% while exciting only a small number of additional modes (Blind et al., 2017).

By contrast, continuous tip-tilt scanning across the fibre core was found to be the most effective AO-based scrambling mechanism. Simulations showed that appropriate tip-tilt amplitudes could excite up to $14\times14$4 of the guided modes, and laboratory tests on a dedicated bench in Geneva with narrow-band filters from 950 to 1650 nm reported a factor 2–3 improvement in line-spread-function stability. The adopted strategy is therefore a continuous scan of the point-spread function across the fibre core during each exposure (Blind et al., 2017).

The second element is mechanical scrambling with fibre stretchers and a double-scrambler-equipped fibre link. The 2022 modal-noise study reports that AO scanning alone and a stretcher alone each yield roughly an order-of-magnitude gain in RV stability, bringing a raw $14\times14$5 instability down to the $14\times14$6 level, while the combined approach reaches stability at the few $14\times14$7 level in the laboratory. The same paper concludes that the envisioned strategies validate stability better than $14\times14$8 for the HAF channel (Blind, 2022).

Later on-sky characterization reports that modal noise can be aptly mitigated by fibre stretchers and AO scanning mode, but it remains the leading limitation at the highest signal-to-noise ratio and affects the reference fibre more strongly than the science fibre. This has operational consequences: simultaneous Fabry–Pérot reference use in fibre B is not always optimal, and sky use of the B fibre is often preferred (Bouchy et al., 29 Jul 2025).

5. On-sky performance, calibration, and radial-velocity precision

Commissioning and early science papers characterize NIRPS as a high-stability cryogenic spectrograph. The later performance paper reports that the thermal control system maintains 1 mK stability over several months, and the spectrograph drift is sufficiently small that stable radial-velocity precision at the level of $14\times14$9 is achieved over several weeks. The same paper reports an overall throughput from the top of the atmosphere to the detector peaking at 13 percent and a Proxima radial-velocity precision of 77 cm/s (Bouchy et al., 29 Jul 2025).

The 2024 first-light paper framed this result as breaking the $1~\mathrm{kHz}$0 barrier at infrared wavelengths. On Proxima, nightly binned LBL radial velocities after modeling planetary and stellar signals reached residuals of 0.86 m/s in that early report, while the later Proxima system paper gives a more extended campaign of 420 spectra over 159 nights, 149 nightly binned RVs with a standard deviation of $1~\mathrm{kHz}$1, a median uncertainty of 55 cm/s, and final residuals of 80 cm/s after a joint model of planets and activity (Artigau et al., 2024, Mascareño et al., 29 Jul 2025).

Operational performance depends on target brightness, seeing, and chosen mode. Later throughput estimates and empirical RV information content suggest photon-noise-limited performance near $1~\mathrm{kHz}$2 in 30 minutes at $1~\mathrm{kHz}$3 for a late M dwarf in HE mode, with poorer precision for earlier or faster-rotating stars. In direct comparison on Proxima, NIRPS delivered more precise RV data than simultaneous HARPS and a more significant detection of the planetary signals, while on other targets such as TOI-406 and TOI-672 it contributed RV precision in the few $1~\mathrm{kHz}$4 regime on faint M dwarfs (Artigau et al., 2024, Osborn et al., 12 Mar 2026).

Calibration combines tungsten lamps, U–Ne hollow-cathode lamps, and a Fabry–Pérot etalon, with later papers also mentioning a laser frequency comb installation. Telluric correction is a major requirement in the near infrared, and two reduction ecosystems are repeatedly discussed: the ESO NIRPS data-reduction software derived from the ESPRESSO DRS, and APERO adapted from SPIRou. Several science papers show that results are robust against reasonable pipeline choices, although telluric residuals remain important in regions of strong terrestrial $1~\mathrm{kHz}$5 absorption (Bazinet et al., 8 Aug 2025, Allart et al., 28 Jul 2025, Mercier et al., 28 Jul 2025).

6. Scientific applications and early results

NIRPS was designed for both precision velocimetry and atmospheric spectroscopy, and early results span both areas. In radial-velocity science, it has been used to refine the masses of known planets and to discover or confirm additional companions around M dwarfs. The Proxima campaign confirmed the existence of Proxima d when combined with simultaneous HARPS and archival data, and refined Proxima b and d to minimum masses of $1~\mathrm{kHz}$6 and $1~\mathrm{kHz}$7, respectively (Mascareño et al., 29 Jul 2025). In the TOI-756 system, NIRPS and HARPS characterized a $1~\mathrm{kHz}$8 transiting sub-Neptune and revealed an additional cold eccentric giant at 149.6 days, while the line-by-line treatment of NIRPS spectra enabled specific masking of OH-contaminated regions in the near infrared (Parc et al., 16 Oct 2025).

In stellar spectroscopy, simultaneous HARPS and NIRPS data have been used to identify activity-sensitive near-infrared lines in M dwarfs. A blind search on Proxima and Gl 581 found hundreds of activity-sensitive lines, with 32% of identified Proxima lines showing rotation-period detections versus 1% for Gl 581, demonstrating that the near infrared is not activity-free but provides a different spectral and diagnostic landscape from the optical. This suggests that NIRPS is valuable not only because activity amplitudes may be reduced in the near infrared, but because activity can be modeled at the line level with simultaneous optical reference information (Silva et al., 28 Jul 2025).

NIRPS is also already established as an atmospheric spectroscopy instrument. In dayside emission from WASP-121 b, it enabled simultaneous detections of $1~\mathrm{kHz}$9 and OH and a retrieved photospheric abundance ratio

$R \approx 88{,}000$0

consistent with thermal dissociation chemistry, while also revealing evidence for Fe and Mg and an apparent velocity shift of

$R \approx 88{,}000$1

in the water signal (Bazinet et al., 8 Aug 2025). In transmission spectroscopy of WASP-69 b, three NIRPS transits resolved the helium triplet with an average excess absorption of $R \approx 88{,}000$2, a delayed post-egress signal lasting about 50 minutes, and a blueshift reaching $R \approx 88{,}000$3, while Rossiter–McLaughlin analysis showed that NIRPS can reach precisions similar to HARPS for spin-orbit studies (Allart et al., 28 Jul 2025).

At the same time, NIRPS has clarified limits imposed by the near-infrared spectral domain itself. On WASP-189 b, the non-detection of Fe in NIRPS despite detection in HARPS was interpreted as the effect of strong $R \approx 88{,}000$4 continuum opacity damping near-infrared line contrast. Joint HARPS+NIRPS retrievals constrained the $R \approx 88{,}000$5 abundance and the free-electron density, illustrating that a near-infrared non-detection can itself be physically diagnostic (Vaulato et al., 28 Jul 2025).

These results are embedded in a large Guaranteed Time Observation program. The instrument papers describe a five-year program of roughly 720–725 nights, divided into blind radial-velocity surveys of nearby low-mass stars, mass measurements of transiting planets around M dwarfs, and high-resolution time-series spectroscopy of exoplanet atmospheres. A plausible implication is that NIRPS is intended not merely as a niche infrared extension to HARPS, but as a survey-scale facility for building statistically meaningful samples in M-dwarf planet demographics, stellar activity characterization, and comparative exoplanet atmospheres (Artigau et al., 2024, Bouchy et al., 29 Jul 2025).