Transit spectrophotometry of the exoplanet HD189733b. I. Searching for water but finding haze with HST NICMOS

We present Hubble Space Telescope near-infrared transit photometry of the nearby hot-Jupiter HD189733b. The observations were taken with the NICMOS instrument during five transits, with three transits executed with a narrowband filter at 1.87 microns and two performed with a narrowband filter at 1.66 microns. Our observing strategy using narrowband filters is insensitive to the usual HST intra-orbit and orbit-to-orbit measurement of systematic errors, allowing us to accurately and robustly measure the near-IR wavelength dependance of the planetary radius. Our measurements fail to reproduce the Swain et al. absorption signature of atmospheric water below 2 microns at a 5-sigma confidence level. We measure a planet-to-star radius contrast of 0.15498+/-0.00035 at 1.66 microns and a contrast of 0.15517+/-0.00019 at 1.87 microns. Both of our near-IR planetary radii values are in excellent agreement with the levels expected from Rayleigh scattering by sub-micron haze particles, observed at optical wavelengths, indicating that upper-atmospheric haze still dominates the near-IR transmission spectra over the absorption from gaseous molecular species at least below 2 microns.

💡 Research Summary

This paper presents a focused near‑infrared (NIR) transit study of the hot‑Jupiter HD 189733b using the Hubble Space Telescope’s NICMOS instrument. The authors observed five separate transits: three with a narrow‑band filter centered at 1.87 µm (NIC1 F187N) and two with a filter at 1.66 µm (NIC2 F166N). By employing narrowband imaging rather than low‑resolution spectroscopy, they dramatically reduced the usual HST intra‑orbit and orbit‑to‑orbit systematics (thermal “breathing”, pointing drift, detector non‑linearity), enabling a robust measurement of the wavelength‑dependent planetary radius at the 10⁻⁴ level.

Data reduction followed a standard NICMOS pipeline (dark subtraction, flat‑fielding, non‑linearity correction) and then applied a “divide‑out‑of‑transit” (divide‑oot) technique to normalize each orbit’s light curve by its out‑of‑transit baseline. Residual pointing variations were tracked via centroid positions and removed, while limb‑darkening was modeled using a four‑parameter non‑linear law derived from PHOENIX stellar atmospheres. The transit model of Mandel & Agol (2002) was fitted using a Markov Chain Monte Carlo (MCMC) approach, yielding planet‑to‑star radius ratios of 0.15498 ± 0.00035 at 1.66 µm and 0.15517 ± 0.00019 at 1.87 µm. The difference between the two bands (ΔR = 0.00019 ± 0.00038) is statistically indistinguishable from zero, indicating a nearly flat transmission spectrum across this NIR interval.

These results directly contradict the water‑absorption feature reported by Swain et al. (2008) below 2 µm, a discrepancy that reaches the 5‑σ level. Instead, the measured radii align closely with the Rayleigh‑scattering slope inferred from optical observations (≈0.3–0.8 µm), which has been attributed to a high‑altitude haze composed of sub‑micron particles. The authors demonstrate that a simple Rayleigh scattering model, with an effective particle size well below the observed wavelengths, reproduces both the optical and the new NIR radius values. In this scenario, the haze resides at pressures of order 0.1–1 mbar, where it creates an optical depth large enough to mask molecular absorption bands from water, methane, or carbon dioxide. Consequently, the NIR transmission spectrum is dominated by scattering rather than by gaseous opacity, at least up to ~2 µm.

The paper also discusses the broader implications for exoplanet atmospheric characterization. First, it highlights that any attempt to detect molecular signatures in transmission must account for, and ideally correct, the contribution of high‑altitude aerosols. Second, it validates the use of narrowband photometry as a powerful tool for mitigating instrument systematics, achieving radius precision comparable to or better than that of low‑resolution spectroscopic modes. Third, the authors outline how future facilities—particularly JWST’s NIRSpec and NIRISS—can extend these measurements to longer wavelengths (3–5 µm) where water’s strong bands become less susceptible to Rayleigh scattering, offering a clearer window on the planet’s bulk composition.

In summary, the study provides compelling evidence that HD 189733b’s upper atmosphere is dominated by a haze of sub‑micron particles that produce Rayleigh‑like scattering across the optical and near‑infrared. This haze effectively obscures the water absorption signature that had been previously claimed, reshaping our understanding of the planet’s atmospheric chemistry and cloud physics. The work underscores the necessity of high‑precision, systematics‑resilient observations and sophisticated aerosol modeling for reliable atmospheric retrievals of hot‑Jupiter exoplanets.