NIRSpec/G395H: High-Res JWST Spectroscopy

• NIRSpec/G395H is a high-resolution fixed-slit spectroscopic mode on JWST that covers a 2.8–5.2 μm window for detailed atmospheric studies.
• It employs a combination of a grating with a long-pass filter and dual HgCdTe detectors, using advanced techniques like PCA for systematics removal in time-series data.
• The mode achieves transit-depth precisions from ≲20 ppm to ≈200 ppm, enabling precise measurements of molecular species and atmospheric properties in exoplanets and brown dwarfs.

The NIRSpec/G395H mode is the high-resolution, fixed-slit spectroscopic configuration of the James Webb Space Telescope’s Near-Infrared Spectrograph, covering the 2.8–5.2 μm window. This mode is optimized for exoplanet atmosphere characterization via transit and emission spectroscopy, as well as for studies of brown dwarfs and planetary atmospheres. By combining high resolving power (R ≈ 2700), broad spectral coverage (including key molecular bands of H₂O, CO₂, CH₄, CO, SO₂, SiO, and others), and finely sampled, stable time-series data, NIRSpec/G395H enables the detection and quantification of atmospheric constituents at high fidelity. Recent studies have demonstrated the mode’s capability to probe mean molecular weight, bulk metallicity, and isotopologue ratios, with demonstrated median transit-depth precisions ranging from ≲20 ppm for bright super-Earth hosts to ≈200 ppm for fainter or more challenging sources. Systematic effects, particularly those correlated with detector properties and instrument configuration, are an active area of study, with pipelines employing pixel-level decorrelation, group-level destriping, and multi-component systematics modeling. The G395H mode has been used extensively in the JWST Cycle 1–3 large programs, including COMPASS and DREAMS, and is shaping the frontier of comparative planetology for close-in small-planet atmospheres.

1. Instrumental Overview and Configuration

NIRSpec/G395H deploys the G395H grating in combination with the F290LP long-pass filter, dispersing incident flux onto two Teledyne HAWAII-2RG HgCdTe detectors: NRS1 spans 2.87–3.72 μm, and NRS2 covers 3.82–5.14 μm, with a small gap (∼3.72–3.82 μm) due to the detector join (Teske et al., 27 Feb 2025, Alderson et al., 2024, Wallack et al., 2024). The native resolving power is R ≈ 2700, with two pixels per FWHM of the instrumental line-spread function (FWHM ≃ λ/R). The spectral sampling is ∼0.7 Å/pix at 3 μm, increasing to ∼2 Å/pix at 5 μm. In time-series Bright Object Time Series (BOTS) applications, the SUB2048 subarray (2048 × 32 or 2048 × 2048 pixels) and NRSRAPID/GRPSAMP readout patterns are typical (Alderson et al., 2024, Gordon et al., 22 Nov 2025).

Typical integration times per exposure and group numbers are selected to balance dynamic range and detector systematics: lower groups-per-integration (N=3–4) favor high-cadence, but exhibit increased systematic residuals (Gordon et al., 22 Nov 2025), whereas higher group numbers (N ≳ 7) suppress red noise but risk saturation on bright targets (Teske et al., 27 Feb 2025, Alderson et al., 2024). Target acquisition and tracking are performed via the S1600A1 slit (1.6″ × 1.6″) or S200A1 slit (0.2″ × 3.3″) for fixed-slit work. No micro-shutter array is utilized in standard G395H time-series operation (May et al., 2023).

2. Data Acquisition and Calibration Workflow

The NIRSpec/G395H calibration chain follows standard JWST processing, with specific G395H adaptations for time-series stability (Alderson et al., 2024, Alderson et al., 2022, Wallack et al., 2024):

• Stage 1 (detector-level): reference pixel correction and superbias subtraction to remove 1/f noise, linearity correction, dark subtraction, and cosmic-ray "jump" detection (often using elevated thresholds, e.g. 15σ, to avoid false positives in bright sources) (Wallack et al., 2024, Sarkar et al., 2024).
• Stage 2: ramp fitting, flat-fielding, wavelength calibration (assign_wcs, extract_2d), and production of slope images ("rateints").
• Stage 3 (extraction): group-level background subtraction (column or frame medians), bad-pixel masking, optimal/intrapixel/broad box aperture extraction along traced spectra (aperture widths ∼4–10 pixels), and correction for trace position and shape drift (Alderson et al., 2024, Teske et al., 27 Feb 2025).
• Systematics modeling: major sources are trace x,y shifts, focus/rotation mode changes, residual 1/f noise, and detector-specific correlated noise, especially pronounced in NRS1 at short wavelengths and in low group-number sequences (Gordon et al., 22 Nov 2025, Luque et al., 2024). Principal component analysis (PCA) on pixel-level time series, as well as multi-component polynomial and linear models incorporating trace position and time, are regularly employed (Gordon et al., 22 Nov 2025).

3. Spectroscopic Performance and Sensitivity

Achievable precision in transit or phase-curve spectroscopy depends on source magnitude, binning strategy, and systematics control:

• Bin sizes: Binning spectra to 30–60 pixels per channel (Δλ ≈ 0.02 μm, R∼200) is routine, yielding per-bin transit depth uncertainties of ~18–34 ppm for high-S/N super-Earth/sub-Neptune targets across NRS1 and NRS2 (Teske et al., 27 Feb 2025, Sikora et al., 2024, Alderson et al., 2024). For faint targets or higher spectral resolution (native R ∼ 2700), per-bin precision can be ~100–300 ppm (Lew et al., 2024), though photon noise is often approached in the absence of severe systematics.
• Noise scaling: Actual uncertainties often exceed PandExo (Pandeia) predictions by factors of 1.1–1.3 (NRS1) and 1.1–1.4 (NRS2) depending on the groups-per-integration and systematics treatment (Gordon et al., 22 Nov 2025, Alderson et al., 2024). Residual red noise is most pronounced for N < 5 groups/integration or for observations in the 2.8–3.5 μm region on NRS1, where time-variable astigmatism, defocus, and alternating column noise ("popcorn mesa") can dominate (Gordon et al., 22 Nov 2025, Luque et al., 2024).
• Detector offsets: Systematic inter-detector offsets (∼40–100 ppm) are frequently observed between NRS1 and NRS2 and are modeled as free parameters in light-curve fitting (Sarkar et al., 2024, Redai et al., 9 Jul 2025, Gordon et al., 22 Nov 2025, Wallack et al., 2024).

4. Atmospheric Retrievals and Molecular Feature Detection

G395H enables detection and quantification of key atmospheric molecular species in both exoplanet and brown dwarf observations via:

• Direct molecular features: Within the 2.8–5.2 μm window, robust detections of H₂O (3–3.7 μm), CO₂ (4.2–4.5 μm), SO₂ (4.05 μm), SiO (4.0–4.3 μm), and CH₄ (3.3 μm), as well as the CO P/R-branch structure (4.6–5 μm), have all been reported (Alderson et al., 2022, Gapp et al., 2 Jun 2025, Gressier et al., 19 Sep 2025, Grant et al., 2023, Lew et al., 2024). For example, H₂O and CO₂ are detected at significance >20σ and >28σ in WASP-39b at R∼600, and SiO at 5.2σ in WASP-121b at native resolution.
• Cross-correlation techniques: Standard and optimal cross-correlation analyses using high-resolution line-by-line models (e.g. from petitRADTRANS or CHIMERA) have yielded 6.6–7.5σ CO detections (including isotopologues) in WASP-39b, and similar approaches for H₂O and CO₂ (Esparza-Borges et al., 2023, Esparza-Borges et al., 29 Sep 2025, Lew et al., 2024).
• Retrieval architectures: Forward models (e.g. PICASO, CHIMERA, NEMESIS, POSEIDON, petitRADTRANS) are convolved to instrumental resolution (via Gaussian LSF matching R=2700) before comparison to observed spectra, with atmospheric metallicity, mean molecular weight (μ), and absorber abundances as fit parameters. Best-practice model selection propagates both photon and systematics noise in the likelihood, and model-dependent definitions of μ and feature amplitude (Δδ ∼ 5H/R_★, H = k_B T/μg) remain central for small-planet regimes (Teske et al., 27 Feb 2025, Wallack et al., 2024).
• Limits and upper bounds: Featureless or nearly flat spectra (within ≤20–50 ppm) can robustly exclude atmospheres with μ ≲ 6–8 g mol⁻¹ or metallicities ≲200–300× Solar for cloud-top pressures ≳10⁻³ bar (Teske et al., 27 Feb 2025, Grado-White et al., 2024, Redai et al., 9 Jul 2025, Gordon et al., 22 Nov 2025, Alderson et al., 2024, Wallack et al., 2024).

5. Systematic Noise, Best Practices, and Lessons from Large Surveys

Comprehensive analyses from the COMPASS program and cross-instrument comparison studies have defined best practices and remaining challenges:

Noise Source / Issue	Typical Impact	Correction / Mitigation
1/f and column-correlated noise	10–100 ppm/bins	Group-level median subtraction, reference-pixel use
Inter-detector offset	40–100 ppm	Free offset in modeling, fit NRS1/NRS2 independently
Group-number (low) systematics	×1.1–1.4 noise	Use ≥4 groups, avoid saturation tradeoffs
Trace morphology changes	20–200 ppm	Pixel-level PCA, multi-vector decorrelation
Instrument red noise (2–5 h)	100–200 ppm	Prayer-bead/bootstrapping error propagation

Feature extraction and atmospheric characterization require repeated, multi-visit confirmation to distinguish real planetary signatures from noise, especially for feature amplitudes ≲30 ppm (May et al., 2023, Alderson et al., 24 Jan 2025). Co-adding spectra across visits or even across targets has not yet revealed robust features below these levels, indicating that degeneracies between high μ, clouds/hazes, and noise floors remain a limiting factor (Gordon et al., 22 Nov 2025). Red noise due to detector or instrument cycles is persistent for long time series and requires careful uncertainty propagation (e.g., prayer-bead) (Luque et al., 2024).

Key recommendations for future G395H programs include:

• Prefer ≥4 (ideally 7+) groups per integration;
• Employ PCA-based systematics removal, with at least 6 principal vectors;
• Inflate PandExo predicted noise estimates by 10–20%;
• Treat NRS1 and NRS2 as separate data sets and model baseline offsets;
• Plan for multiple (2–10) transits for robust molecular feature confirmation.

6. Scientific Impact and Applications

NIRSpec/G395H has had immediate scientific impact across a range of exoplanet and substellar atmosphere studies:

• Super-Earths and Sub-Neptunes: Systematic G395H surveys (e.g., COMPASS) have confirmed that many 1–2.5 R⊕ planets are either airless or possess high-μ, cloud- or haze-masked atmospheres, with metallicity floors ≳175–300× Solar (Teske et al., 27 Feb 2025, Alderson et al., 2024, Redai et al., 9 Jul 2025, Wallack et al., 2024).
• Giant Planets and Hot Jupiters: Precise, simultaneous detections and abundance constraints for multiple molecules offer new insights into C/O ratios, atmospheric metallicities, and formation/migration histories (Alderson et al., 2022, Gapp et al., 2 Jun 2025, Ahrer et al., 16 May 2025).
• Disequilibrium Chemistry: Detection of SO₂ in HAT-P-26b (ln B=13.5), SiO in WASP-121b (5.2σ), and simultaneous H₂S+SO₂ (∼2σ each) in smaller planets demonstrates access to disequilibrium species and photochemical tracers (Gressier et al., 19 Sep 2025, Gapp et al., 2 Jun 2025, Banerjee et al., 2024).
• Brown Dwarfs: WISE 1828 studies recover six major molecular carriers and deliver percent-level C/O and metallicity, and isotopologue limits (e.g., ^12CO /^13CO > 40) at R ≈ 2700 (Lew et al., 2024).
• Phase curves and atmospheric dynamics: High-cadence, continuous phase curves (e.g., HD 80606 b, TOI-1685 b) with G395H provide time-resolved mapping of thermal and compositional evolution across planetary orbits (Sikora et al., 2024, Luque et al., 2024).

7. Future Directions and Limitations

The G395H mode is rapidly defining the sensitivity floor for small planet atmospheric retrievals and the delineation of the “Cosmic Shoreline”—the boundary between airless and envelope-retaining worlds (Luque et al., 2024). Further improvements to instrument modeling (including for red noise and detector systematics), extended multi-visit datasets, and joint use of complementary JWST modes (NIRISS, MIRI) are poised to break current degeneracies in μ, metallicity, and cloud properties (Wallack et al., 2024, Redai et al., 9 Jul 2025).

While the mode achieves near-photon-limited performance for some bright targets, residual systematics (especially near 3.2 μm and in long time series) currently set a practical ∼20 ppm lower bound per-bin and ∼30–70 ppm for challenging cases (Luque et al., 2024, Gordon et al., 22 Nov 2025). Upper limits on potentially volcanic sulfur species, precise isotopologue ratios, and atmospheric escape signatures remain compelling targets as calibration and systematics control continue to improve (Banerjee et al., 2024, Lew et al., 2024).

The robust methodologies and best practices developed in recent cycles inform future strategy, ensuring that NIRSpec/G395H will continue to be an essential tool for high-precision, comparative exoplanetology and atmospheric chemistry in the JWST era.