NIRPS: Infrared Spectrograph for Exoplanet Studies

• NIRPS is a high-resolution, AO-fed near-infrared echelle spectrograph that extends HARPS’ capabilities by enabling precision radial-velocity studies of low-mass stars and exoplanets.
• It employs few-mode fibers, adaptive optics, and dual science modes to achieve sub-meter-per-second RV precision and broad spectral coverage from 0.37 to 1.8 μm.
• NIRPS supports diverse science cases—from exoplanet confirmation and atmospheric spectroscopy to stellar characterization—while addressing challenges like modal noise and telluric contamination.

Searching arXiv for recent NIRPS papers to ground the article in published work. Near-InfraRed Planet Searcher (NIRPS) is a high-resolution, high-stability, adaptive-optics-assisted, fiber-fed near-infrared echelle spectrograph on the ESO 3.6 m telescope at La Silla, designed primarily for precision radial-velocity work on low-mass stars and for high-dispersion exoplanet spectroscopy in the near infrared. It operates simultaneously with HARPS through a VIS–NIR dichroic, extending the HARPS wavelength domain into the Y, J, and H bands and providing a combined optical-to-near-infrared baseline for chromatic activity mitigation, dynamical mass measurements, stellar characterization, and transmission or emission spectroscopy (Artigau et al., 2024, Bouchy et al., 29 Jul 2025).

1. Scientific role and programmatic context

NIRPS was developed to address a central limitation of optical Doppler surveys: M dwarfs emit most of their flux in the near infrared, while the reflex amplitudes induced by low-mass planets are larger around low-mass stars. The instrument is therefore explicitly aimed at low-mass exoplanets around M dwarfs, mass measurements of transiting planets, and atmospheric studies in spectral regions containing diagnostics such as water and helium (Artigau et al., 2024, Bouchy et al., 29 Jul 2025).

Its scientific framing is organized around a Guaranteed Time Observation program that began in April 2023. The early-science paper describes a five-year program spanning 720 observing nights, whereas the later on-sky performance paper describes 725 nights over five years; both papers divide the program into a blind radial-velocity search around low-mass stars, mass and density characterization of transiting planets around M dwarfs, and high-resolution time-series spectroscopy of exoplanet atmospheres (Artigau et al., 2024, Bouchy et al., 29 Jul 2025). This suggests that NIRPS was conceived not as a single-purpose velocimeter but as a multi-regime facility spanning Doppler detection, stellar spectroscopy, and atmospheric diagnostics.

The scientific rationale is tightly linked to simultaneous operation with HARPS. The visible beam is sent to HARPS and the near-infrared beam to NIRPS, so the same target can be monitored across a broad wavelength range. In the on-sky performance paper, the combined range is given as 378–1920 nm, while the commissioning paper summarizes the joint system as providing roughly 0.37–1.8 μm coverage (Artigau et al., 2024, Bouchy et al., 29 Jul 2025). For activity mitigation, this chromatic baseline is operationally important: a Keplerian signal should remain achromatic, whereas line-profile distortions induced by magnetic activity often vary with wavelength, line depth, and species.

2. Instrument architecture and the few-mode design choice

NIRPS is built around four subsystems: the front end, calibration unit, fiber link, and cryogenic spectrograph plus detector (Artigau et al., 2024). The front end sits at the telescope Cassegrain interface, includes atmospheric dispersion correction, adaptive optics, fiber injection, and guiding, and uses a dichroic to send visible light to HARPS while directing the infrared to NIRPS. The front-end paper describes a 14×14 Shack–Hartmann wavefront sensor, an ALPAO DM241 deformable mirror with 15×15 actuators, loop frequencies from 250 Hz to 1 kHz, and an AO architecture specifically optimized for coupling into small fibers rather than image quality in the usual seeing-limited sense (Blind et al., 2022).

The instrument operates in two science modes. High-accuracy mode uses a 0.4″ octagonal fiber and delivers resolving power near $R \sim 90{,}000$, while high-efficiency mode uses a 0.9″ octagonal fiber and delivers $R \sim 75{,}000$ (Bouchy et al., 29 Jul 2025). The early-science paper quotes $R \approx 84{,}000$ for HA and $R \approx 72{,}000$ for HE, reflecting earlier summary values, while the on-sky performance characterization reports the measured medians near 88,000 and 75,200, respectively (Artigau et al., 2024, Bouchy et al., 29 Jul 2025). The spectrograph is a white-pupil cross-dispersed echelle instrument in a cryogenic vacuum vessel at about 75 K, using a 4096 × 4096 Teledyne Hawaii-4RG detector (Artigau et al., 2024, Bouchy et al., 29 Jul 2025).

A defining design choice is the use of AO-fed few-mode fibers rather than either a classical highly multimode fiber or a single-mode fiber. The conceptual basis is given by the few-mode relation

$$N_{\mathrm{modes}} \approx \frac{V^2}{4}, \qquad V = \frac{2\pi\,\mathrm{NA}\,a}{\lambda},$$

so smaller core radius, smaller numerical aperture, and longer wavelength reduce the number of guided modes (Blind et al., 2017, Blind, 2022). For the high-accuracy fiber, NIRPS operates in a regime of roughly 10 to 35 modes across its wavelength range (Blind et al., 2017, Blind, 2022).

This choice was motivated by a system-level trade-off. Few-mode injection allows a more compact cryogenic spectrograph while preserving high coupling under realistic seeing and on faint stars. The design papers report coupling $\ge 50\%$ up to magnitude $I=12$ and a gain of 1–2 magnitudes over a single-mode solution, with the benefit that the spectrograph can remain substantially smaller than a seeing-limited near-infrared instrument (Blind et al., 2017). The cost is much stronger modal noise, which becomes a first-order radial-velocity error source.

3. Calibration, data reduction, and modal-noise control

NIRPS calibration is based on tungsten flat fields, uranium-neon lamps for absolute wavelength calibration, and a Fabry–Pérot cavity for dense line sampling and drift monitoring; the commissioning paper also describes a slot for a laser frequency comb (Artigau et al., 2024). The fiber link incorporates scrambling devices, including a fiber stretcher, double scrambler, and slicer elements, because modal noise is more severe in the near infrared and especially acute in a few-mode system (Artigau et al., 2024, Blind, 2022).

Two independent reduction pipelines are used. The Geneva pipeline is based on the ESPRESSO pipeline and is the nominal ESO/DFS path, while APERO was originally developed for SPIRou and natively supports infrared processing and line-by-line radial-velocity extraction (Artigau et al., 2024). The line-by-line method is repeatedly favored in NIRPS science papers because it is more robust than classical cross-correlation against telluric residuals, detector outliers, and line-dependent systematics on M dwarfs (Artigau et al., 2024, Mascareño et al., 29 Jul 2025).

Modal-noise mitigation is integral to the instrument concept. The design studies show that continuous tip-tilt scanning of the fiber core is the most effective AO-based scrambling strategy, because it excites a large fraction of the available modes while preserving throughput better than higher-order aberration injection (Blind et al., 2017). The later modal-noise paper adds a semi-empirical validation using measured near-field and far-field fiber outputs injected into a Zemax spectrograph model; it concludes that raw few-mode photocenter instability at the $\sim 10$ m/s level can be reduced by about a factor of 10 with AO scrambling and by about a factor of 10 with the fiber stretcher, reaching the few 10 cm/s regime in the combined design study (Blind, 2022).

The classical double scrambler is treated more ambivalently. In the few-mode regime, near-field and far-field are strongly coupled, and the 2022 modal-noise study reports no clear winner between a standard feedthrough and a double scrambler, while also noting losses of about 20% in some cases and the possibility that the double scrambler can introduce additional modal noise (Blind, 2022). This is one of the clearer technical controversies in the NIRPS literature: the architecture is viable, but the usual multimode intuition about scrambling does not transfer straightforwardly to few-mode operation.

4. Commissioning status and measured performance

NIRPS had first light on 17 May 2022 and began official operations on 1 April 2023 after staged commissioning, front-end tests, fiber-link installation, cryogenic spectrograph integration, and a significant grating replacement. The original Bach echelle grating was replaced in July 2022 by an etched crystalline silicon grating from IOF-Fraunhofer because of ghosting and throughput issues (Artigau et al., 2024).

The on-sky performance paper reports continuous spectral coverage from 972.4 to 1919.6 nm, a peak total throughput from the top of the atmosphere to the detector near 13%, and thermal control at the 1 mK level over several months (Bouchy et al., 29 Jul 2025). The commissioning paper similarly emphasizes cryogenic stability better than 1 mK RMS, vacuum below $10^{-6}$ mbar, and typical spectrograph drift of about 0.1 m/s/day (Artigau et al., 2024). The later performance characterization refines this to intrinsic drifts of 4.1 cm/day in HA and 3.4 cm/day in HE, with residual dispersions of 72 cm/s and 64 cm/s after removing the long-term linear trend (Bouchy et al., 29 Jul 2025).

Radial-velocity performance is one of the defining achievements of NIRPS. The commissioning paper states that the instrument has broken the 1 m/s RV precision barrier at infrared wavelengths and cites Proxima residuals of 86 cm/s RMS in 2024, with a median absolute deviation of 44 cm/s (Artigau et al., 2024). The Proxima system paper reports 420 NIRPS spectra over 159 nights, 149 nightly binned RVs with a median uncertainty of 55 cm/s, and residual RMS near 80–81 cm/s after full activity-plus-planet modeling, concluding that NIRPS outperforms contemporaneous HARPS on that star (Mascareño et al., 29 Jul 2025). The on-sky performance paper summarizes the Proxima showcase more compactly as an RV precision of 77 cm/s (Bouchy et al., 29 Jul 2025).

AO performance is also central. NIRPS uses AO not merely for image sharpening but to enable efficient injection into 0.4″ and 0.9″ fibers. The on-sky characterization gives effective correction up to about $I \approx 14.5$, typical Strehl of 30–40% at 1350 nm, and encircled energies of about 55% in the 0.4″ fiber and 70% in the 0.9″ fiber for bright stars (Bouchy et al., 29 Jul 2025). The front-end paper also documents a commissioning discrepancy relative to simulations, consistent with an additional unmodeled disturbance of 130–170 nm RMS, and describes ADC wobble and DM thermal behavior as important lessons learned (Blind et al., 2022).

5. Radial-velocity science and dynamical characterization

NIRPS has rapidly become a confirmation and characterization instrument for transiting and non-transiting companions around low-mass stars.

System	NIRPS role	Outcome
TOI-6508b	Decisive spectroscopic follow-up	Brown dwarf validated
TOI-756 b/c	Joint RVs with HARPS	Inner mass, outer giant discovered
GJ 3090 b/c	Multidimensional GP with HARPS	Mass refined, second planet confirmed
TOI-672 b	Dominant RV sensitivity on M dwarf	Hot super-Neptune confirmed

In TOI-6508b, NIRPS was the key spectroscopic follow-up instrument that turned a TESS candidate into a confirmed transiting brown dwarf. Six spectra were obtained over three nights, reduced with the nominal NIRPS pipeline, and cross-correlated with an M4 mask using only the 1.4–1.87 μm orders because of modest SNR. The resulting RVs showed a very large-amplitude signal, and a joint Metropolis–Hastings MCMC fit with TESS, SPECULOOS-South, and LCOGT photometry yielded an eccentric solution with $R \sim 75{,}000$0 days, $R \sim 75{,}000$1, $R \sim 75{,}000$2 km s$R \sim 75{,}000$3, and $R \sim 75{,}000$4 (Barkaoui et al., 27 Feb 2025).

The TOI-756 study is framed as the first result of the NIRPS-GTO “Sub-Neptunes” program and describes TOI-756 c as “the first planet detected with NIRPS.” There NIRPS and HARPS were used simultaneously over 64 NIRPS nights and 57 HARPS spectra. After correction of BERV-crossing systematics with a line-by-line approach that removed $R \sim 75{,}000$5 km s$R \sim 75{,}000$6 windows around the 25% most intense OH lines, the joint solution measured the transiting sub-Neptune TOI-756 b and discovered the eccentric giant TOI-756 c, with a further linear trend suggesting a third component (Parc et al., 16 Oct 2025).

For GJ 3090, NIRPS contributed 84 usable RVs from 98 nights and was analyzed jointly with HARPS RVs, HARPS differential temperatures, and TESS photometry in a Rajpaul-style multidimensional GP framework. This chromatic and multi-observable treatment was essential because the 15.9 d signal of GJ 3090 c lies near the stellar rotation period of $R \sim 75{,}000$7 d. The combined analysis refined the transiting planet GJ 3090 b to $R \sim 75{,}000$8 and confirmed GJ 3090 c as a non-transiting sub-Neptune with a minimum mass of $R \sim 75{,}000$9 (Lamontagne et al., 14 Jan 2026).

The same instrumental combination underpins broader compositional programs. The CMF subprogram used NIRPS both for precise RV masses and for stellar abundances in systems such as GJ 1132 b, GJ 1252 b, and LTT 3780 b, reporting masses of $R \approx 84{,}000$0, $R \approx 84{,}000$1, and $R \approx 84{,}000$2, respectively, and comparing planetary CMFs with host-star refractory abundances derived from NIRPS spectra (Weisserman et al., 8 Apr 2026).

6. Atmospheric spectroscopy, stellar spectroscopy, and broader scientific scope

Although its name emphasizes planet searching, NIRPS is also an atmospheric and stellar-spectroscopy instrument. Several early science papers use it for time-resolved transmission or emission spectroscopy.

In WASP-69b, NIRPS provided three transits in HE mode simultaneously with HARPS and enabled, in the same dataset, both a Rossiter–McLaughlin analysis and helium-triplet transmission spectroscopy. The study reports a slightly misaligned orbit with $R \approx 84{,}000$3 deg in the final combined fit, clear helium excess absorption of $R \approx 84{,}000$4, and a post-egress blueshift reaching $R \approx 84{,}000$5 km s$R \approx 84{,}000$6 about 50 minutes after optical egress; EVE simulations then infer a cometary-like tail and a mass-loss rate of $R \approx 84{,}000$7 g/s (Allart et al., 28 Jul 2025).

In WASP-189b, NIRPS was used simultaneously with HARPS for two transits of an ultra-hot Jupiter. Neutral Fe was detected in HARPS at $R \approx 84{,}000$8 but not in NIRPS, and injection–recovery tests showed that a Fe-only template would have been recovered in the NIRPS data at $R \approx 84{,}000$9, whereas a template including Fe + H$R \approx 72{,}000$0 would not. The joint retrieval therefore interprets the near-infrared non-detection as continuum suppression by H$R \approx 72{,}000$1, not absence of iron (Vaulato et al., 28 Jul 2025). This is a significant correction to a common misconception that non-detection in the near infrared necessarily reflects lower abundance or inadequate sensitivity.

In WASP-121 b, NIRPS dayside emission spectroscopy over six nights yielded simultaneous detections of H$R \approx 72{,}000$2O, OH, Fe, and Mg. The retrieved ratio $R \approx 72{,}000$3 indicates about as much OH as H$R \approx 72{,}000$4O at photospheric pressures, consistent with chemical-equilibrium expectations for thermal water dissociation, while a species-dependent H$R \approx 72{,}000$5O velocity offset of $R \approx 72{,}000$6 km s$R \approx 72{,}000$7 is reported as not reproduced by current global circulation models (Bazinet et al., 8 Aug 2025).

NIRPS is also being used to characterize stellar spectral variability itself. A blind line-by-line search using simultaneous HARPS and NIRPS observations of Proxima and Gl 581 identified 760 activity-sensitive NIR lines in Proxima and 312 in Gl 581, with 17 lines showing rotation-period detections in both stars (Silva et al., 28 Jul 2025). Solar HELIOS observations with NIRPS similarly examined the 10833 Å He triplet and found significant variability on timescales from minutes to days, with telluric contamination and stellar activity as likely short- and long-term drivers; however, injection–recovery tests showed that, for the tested 5% planetary signal, recovered atmospheric parameters remained consistent with input values at the 1$R \approx 72{,}000$8 level (Mercier et al., 28 Jul 2025).

7. Limitations, interpretation, and significance

Several recurrent limitations define the practical use of NIRPS. Telluric absorption and OH emission are major systematics in the near infrared, and the TOI-756 paper documents 100–200 m s$R \approx 72{,}000$9 outliers when $$N_{\mathrm{modes}} \approx \frac{V^2}{4}, \qquad V = \frac{2\pi\,\mathrm{NA}\,a}{\lambda},$$0 falls below about 10 km s$$N_{\mathrm{modes}} \approx \frac{V^2}{4}, \qquad V = \frac{2\pi\,\mathrm{NA}\,a}{\lambda},$$1 (Parc et al., 16 Oct 2025). Modal noise remains intrinsic to the few-mode concept, and the instrument literature treats fiber stretchers and AO scanning as necessary rather than optional components of the precision-RV budget (Blind, 2022, Bouchy et al., 29 Jul 2025). Some applications also remain sample-limited: the TOI-6508 study explicitly notes that additional RV observations are required to better constrain the mass and eccentricity (Barkaoui et al., 27 Feb 2025).

At the same time, NIRPS has altered the practical scope of infrared spectroscopy. The commissioning and performance papers argue that infrared Doppler work has moved from a regime of “good enough” planet confirmation to sub-meter-per-second science (Artigau et al., 2024, Bouchy et al., 29 Jul 2025). The Proxima program demonstrates that this is not merely an engineering metric: NIRPS delivered a more precise time series than simultaneous HARPS for a benchmark active M dwarf and recovered both Proxima b and evidence for Proxima d in the infrared (Mascareño et al., 29 Jul 2025).

A broader implication is that NIRPS occupies a distinct niche rather than simply duplicating HARPS at longer wavelengths. Its simultaneous coupling to HARPS provides chromatic leverage for disentangling activity, its near-infrared throughput makes faint red M dwarfs accessible, and its wavelength range supports high-resolution studies of helium escape, hydride-ion continuum opacity, refractory and volatile chemistry, and stellar abundances (Parc et al., 16 Oct 2025, Lamontagne et al., 14 Jan 2026, Allart et al., 28 Jul 2025, Bazinet et al., 8 Aug 2025). In that sense, the “planet searcher” designation is narrower than the instrument’s demonstrated use: NIRPS functions as a precision near-infrared Doppler spectrograph, a stellar-abundance facility, and a high-dispersion atmospheric spectrograph within a single integrated architecture.