WFC3/UVIS G280 Grism: UV-Optical Exoplanet Spectroscopy

• WFC3/UVIS G280 grism is a slitless spectroscopic mode on HST covering 0.20–0.80 µm, ideal for simultaneous UV and optical observations.
• It disperses stellar point-spread functions into both +1 and –1 orders, achieving resolutions from ~70 to 160 for detailed transit and eclipse analyses.
• Tailored calibration and extraction techniques address curved spectral traces and systematics, enabling ppm-level precision in exoplanet atmospheric retrievals.

The WFC3/UVIS G280 grism is a slitless, cross-dispersed spectroscopic mode implemented on the UVIS channel of the Hubble Space Telescope's Wide Field Camera 3 (HST/WFC3). Optimized for the 0.20–0.80 µm band, it enables simultaneous ultraviolet (UV) and optical spectroscopy—primarily of bright transiting exoplanet systems—by dispersing stellar point-spread functions directly across the detector without the use of a slit, thus maximizing spectrophotometric throughput at short wavelengths and capturing both the +1 and –1 spectral orders in a single configuration. This mode has become an important tool for characterizing exoplanet atmospheres, probing aerosols, and measuring reflected light, particularly given its enhanced UV efficiency compared to other HST spectrographs. Its calibration, reduction, and science application frameworks are highly specialized, reflecting the unique characteristics and challenges of UVIS slitless spectroscopy.

1. Instrumental Configuration and Spectral Characteristics

The G280 grism is installed in filter wheel 4 of the UVIS channel, enabling coverage from 200 nm to 800 nm in a single exposure for both first orders (+1 and –1). The instrumental configuration is entirely slitless, with the dispersed starlight extending along curved traces on the detector. The typical setup utilizes subarrays—e.g., 2250 × 650 pixels or 2100 × 800 pixels on UVIS chip 2—with a plate scale of ≃0.0395″ per pixel. POSTARG Y offsets are used (e.g., –50″) to center the spectrum for optimal order capture.

Spectral resolution is both wavelength- and PSF-dependent, scaling approximately as $R(λ) = λ/Δλ$: $R ≃ 70$ at 0.25 µm, rising to $R ≃ 160$ at 0.75 µm. Detector-limited pixel sampling determines the effective resolution, and A plausible implication is that at blue wavelengths (200–300 nm), G280’s throughput is 2–6× higher than that of STIS/G430L, making it particularly valuable for Rayleigh slope and near-UV absorber studies (Boehm et al., 22 Oct 2024, Radica et al., 7 Mar 2025).

The pixel-to-wavelength solution is computed using the GRISMCONF package. For the ±1st orders, the relation is typically given by:

$λ(p) = a₀ + a₁·p + a₂·p² + a₃·p³$

where $p$ is the pixel coordinate along the trace, and $a_i$ are coefficients calibrated for each chip and configuration. For more general applications, the full 2D mapping $λ(x, y)$ may be expressed as:

$λ(x, y) = a₀ + a₁·Δx + a₂·Δx² + a₃·Δy + …$

through coefficients determined from direct on-sky calibrations (Radica et al., 7 Mar 2025).

System throughput is a product of telescope optics, grism, filter, and detector quantum efficiency. Peak efficiency approaches 25–30 % near 250 nm, with steep decline to ≃ 5–10 % by 800 nm. The detector quantum efficiency (QE) itself rises from ≃ 20 % at 0.2 µm to ≃ 90 % at 0.45 µm, then falls back to ≃ 20 % at 0.8 µm.

Wavelength (µm)	Throughput η(λ) (%)	Resolution R(λ)
0.20	15	70
0.40	35	110
0.60	20	130
0.75	10	160
0.80	5	160

2. Observational Strategies and Cadence

Time-series applications for exoplanet transit and eclipse spectroscopy utilize consecutive HST orbits with frequent short exposures (e.g., 78 s), stacking exposures across multiple orbits to optimize phase coverage. Boehm et al. (2024) employed 79 exposures across five orbits (15 in the first, 16 in each subsequent) for a single WASP-127 b transit, acquiring both pre-transit direct F300X images for wavelength zero-pointing and spectroscopic time series for transit depth extraction (Boehm et al., 22 Oct 2024). The “stare” mode (no spatial scans, no dithering) is recommended to prevent trace curvature mismatches between spectral frames.

For reflected light or eclipse studies, multiple visits can be coadded. For example, LTT 9779b observations acquired four eclipses, stacking to reduce the white-light detection threshold to 113 ppm at 3$\sigma$ (Radica et al., 7 Mar 2025).

Subarrays are centered to capture both ±1 orders. POSTARG offsets are used to optimize placement; practical advice is to always capture both ±1 orders for a $\sqrt{2}$ SNR improvement (Radica et al., 7 Mar 2025).

3. Calibration and Data Reduction Techniques

Data reduction with G280 requires a tailored workflow due to the slitless, curved-trace, and order-overlap characteristics:

1. Starting files: Calibrated “flt” files from calwfc3 (bias, dark, flat-field corrections applied).
2. Cosmic-ray and hot-pixel cleaning: Both temporal (median filtering across time per pixel, typically 4–5$\sigma$ clipping) and spatial (local median or Laplacian edge detection [van Dokkum 2001]) algorithms are used.
3. Background subtraction: Median-stacked G280 sky frames are preferred; alternatives include polynomial fits to off-trace pixels, and estimates from detector corners or non-illuminated pixels.
4. Subarray embedding: Header keywords (NAXIS, LTV1, LTV2) map subarray data back into the full frame, where needed.
5. Spectral trace fitting: Trace is parameterized as $y(x) = c_0 + c_1x + c_2x^2$ per exposure, typically using least-squares polynomial fits to centroids.
6. Spectral extraction: Aperture extraction centered on trace, with width $w$ (e.g., ±10–12 pixels); either boxcar or optimal extraction (Horne 1986).
7. Wavelength calibration: Apply GRISMCONF or Pirzkal et al. (2017, 2020) solution, anchoring with pre-transit direct images.
8. Flux calibration: Use PHOTFLAM, GAIN, and exposure time:

$$f_\lambda = \frac{(\mathrm{counts} - B)\,\mathrm{PHOTFLAM}}{\mathrm{GAIN} \times t_{\rm exp}}$$

For physically calibrated $f_\lambda$ in $\mathrm{erg\ s^{-1} cm^{-2} \AA^{-1}}$ (Radica et al., 7 Mar 2025).

1. Spectral binning: Spectra are divided into uniform bins (e.g., 59 × 10 nm bins for WASP-127 b), combining ±1st orders where possible.
2. Light curve generation: White-light curves (integrated over full band) and spectroscopic light curves (per bin) are constructed, normalized to out-of-transit baseline.

4. Transit and Eclipse Analysis Frameworks

For exoplanet transmission studies, light curves are detrended and fit using parametric systematics models in tandem with analytic transit models. Typical systematics decorrelation includes terms for thermal ramps, HST orbital phase, and trace offsets, as in:

$$f(t) = F_0 \cdot T(t, \theta) \cdot \left[1 + c_1\phi_{\rm orbit} + c_2 x_{\rm shift} + c_3 y_{\rm shift} + \ldots \right]$$

where $T(t, \theta)$ is the transit model (e.g., Mandel & Agol 2002, implemented in batman), and systematics coefficients are marginalized over (e.g., ExoTiC-ISM library with 51 parametric models (Boehm et al., 22 Oct 2024)). Detrending may also employ linear decorrelation vectors derived from PSF, spectral trace, or telescope pointing data.

The extraction of spectroscopic transmission spectra proceeds by binning light curves spectrally and fitting for the planet-to-star radius ratio $R_{\rm p}/R_\star$ per bin, with shared orbital parameters (e.g., $a/R_\star$, $i$, $T_0$) across bins. Limb-darkening parameters are fixed or Gaussian-constrained using stellar atmosphere models.

In eclipse detection or upper limit settings, injection–recovery methodologies are employed: simulated eclipse models are injected into the data at varying depths, and the significance of recovery is assessed over multiple noise realizations, with $3\sigma$ detection thresholds defined via Bayesian evidence (Δln Z ≈ 3) (Radica et al., 7 Mar 2025).

5. Achieved Precision, Throughput, and Comparative Performance

Empirical noise and sensitivity results for G280 mode have been demonstrated in both transit and eclipse contexts:

• WASP-127 b (transit): Achieved broadband white-light precision of 91 ppm; median precision per 10 nm spectral bin is 240 ppm (single transit). Near-UV precision (290–370 nm) is 66 ppm in one transit (vs. ∼400 ppm across two transits for STIS/G430L) (Boehm et al., 22 Oct 2024).
• LTT 9779b (eclipse): On-sky per-visit white-light precision is 200 ppm; stacking four visits yields ≃70 ppm. A 3σ upper limit on eclipse depth is set at 113 ppm for the 0.2–0.8 µm white-light band (Radica et al., 7 Mar 2025).

G280 provides complete UV–visible coverage overlapping combined STIS/G430L+G750L in a single exposure and matches WFC3/IR grisms for precision and bin size at their overlap region (1100–1700 nm). The superior UV throughput of G280 enables Rayleigh slope constraints and high-altitude haze detection at <1 transit, whereas STIS typically requires two transits at greater noise.

6. Science Applications and Atmospheric Retrieval

G280 enables the derivation of transmission and reflectance spectra for exoplanets with high sensitivity to aerosols, alkali metals, and cloud properties:

• Transmission spectroscopy (e.g., WASP-127 b): Integrated retrievals (POSEIDON) detect sodium at 4.1σ (with joint STIS and IR data) and constrain patchy, high-altitude silicate clouds with cover fraction $\phi_{\rm cloud} = 0.31 \pm 0.08$, particle sedimentation efficiency $f_{\rm sed} ∼ 0.3$–1, and eddy diffusion $K_{\rm zz} = 10^9$–$10^{11}\ \mathrm{cm}^2\,\mathrm{s}^{-1}$. Mean particle sizes are sub-micron, with optical depth $\tau_{\rm cloud}(\lambda)$ parameterized as:

$$\tau_{\rm cloud}(\lambda) = \int Q_{\rm ext}(a, \lambda) \pi a^2 n(a) dz$$

where $Q_{\rm ext}$ is from Mie theory, $n(a)$ from Virga, and cloud top pressures are in the millibar–microbar regime.

• Reflected light (e.g., LTT 9779b): Forward grids (VIRGA/PICASO) compared to eclipse upper limits indicate silicate condensates and submicron, reflective cloud particles, though vertical mixing constraints remain weak (Radica et al., 7 Mar 2025).

A plausible implication is that, as the community's primary HST-accessible mode for 200–800 nm spectroscopy, G280 will remain central for UV exoplanet studies, at least until future facilities such as the Habitable Worlds Observatory.

7. Limitations, Challenges, and Best Practices

Operationally, G280 observations face several unique challenges: order overlap (especially beyond 0.55 µm), PSF smearing, sky background variability (notably at orbit start due to Earth limb), and systematics linked to HST's orbital temperature, jitter, and pointing. Detector artifacts (hot pixels, cosmic rays) require robust cleaning. SNR and detection thresholds are ultimately photon and read-noise limited in the red, and background-limited in the blue.

Order overlap can cause up to 2 % contamination for λ > 450 nm but is generally negligible for exoplanet transit analysis in ±1st orders. The use of both orders is encouraged for SNR gain and robust trend correction. Spectral trace fitting and wavelength registration to direct images are critical for sub-nm solution accuracy.

Key practical guidelines include:

• Always extract both ±1st orders and combine.
• Use a median estimator for background over mode for robustness.
• Regularly cross-correlate for zeropoint shifts (≃ 35–40 Å).
• Incorporate jitter, drift, and state vectors into systematics models, selecting by Bayesian evidence.
• Plan exposures to capture ingress/egress, which are valuable for precise orbital parameter constraints.

Conclusion

The WFC3/UVIS G280 grism enables high-throughput, UV–optical slitless spectroscopy across 0.20–0.80 µm, facilitating transmission and eclipse observations of exoplanet atmospheres at ppm-level precision. Its performance—white-light precision approaching 91 ppm in a single transit, persistence of high UV throughput, and ability to constrain atmospheric structure and composition—renders it a leading mode for pre-JWST and post-JWST UV/optical exoplanet characterization. The available data and reduction pipelines provide a reproducible pathway from detector frames to science-ready transmission spectra, with best practices robust against instrument systematics and astrophysical challenges (Boehm et al., 22 Oct 2024, Alam et al., 12 Nov 2025, Radica et al., 7 Mar 2025).