Extreme winds on the emerging dayside of an ultrahot Jupiter

High-resolution spectroscopy provides a unique opportunity to directly probe atmospheric dynamics by resolving Doppler shifts of planetary signal as a function of orbital phases. Using the optical spectrometer Keck Planet Finder (KPF), we carry out a pilot study on high-resolution phase curve spectra of the ultra-hot Jupiter KELT-9 b. We spectrally and temporally resolve its dayside emission from post-transit to pre-eclipse (orbital phase phi = 0.1 - 0.45). The signal strength and width increase with orbital phases as the dayside rotates into view. The net Doppler shift varies progressively from -13.4 +/- 0.6 to -0.4 +/- 1.0 km/s, the extent of which exceeds its rotation velocity of 6.4 +/- 0.1 km/s, providing unambiguous evidence of atmospheric winds. We devise a retrieval framework to fit the full time-series spectra, accounting for the variation of line profiles due to the rotation and winds. We retrieve a supersonic day-to-night wind speed up to 11.7 +/- 0.6 km/s on the emerging dayside, representing the most extreme atmospheric winds in hot Jupiters to date. Comparison to 3D circulation models reveals a weak atmospheric drag, consistent with relatively efficient heat recirculation as also supported by space-based phase curve measurements. Additionally, we retrieve the dayside chemistry (including Fe i, Fe ii, Ti i, Ti ii, Ca i, Ca ii, Mg i, and Si i) and temperature structure, and place constraints on the nightside thermal profile. Our high-resolution phase curve spectra and the measured supersonic winds provide excellent benchmarks for extreme physics in circulation models, demonstrating the power of this technique in understanding climates of hot Jupiters.

💡 Research Summary

This paper presents a pioneering high‑resolution optical phase‑curve study of the ultra‑hot Jupiter KELT‑9 b using the Keck Planet Finder (KPF) spectrograph. Observations were carried out on three nights in August 2024, covering orbital phases ϕ ≈ 0.05–0.48 with a total of 269 exposures (each 150 s) to minimize smearing given the planet’s large radial‑velocity semi‑amplitude (Kp ≈ 245 km s⁻¹). The data reduction pipeline performed flat‑fielding, order tracing, optimal extraction, laser‑frequency comb wavelength calibration, blaze correction, and telluric removal with molecfit. Stellar lines were removed by dividing each spectrum by a master stellar template, and residual continuum trends were detrended with low‑order polynomials; only ~1.2 % of wavelength channels were masked due to strong telluric contamination.

Synthetic emission spectra were generated with petitRADTRANS, employing the Guillot (2010) analytic temperature‑pressure profile parameterized by infrared opacity (κ_IR), the optical‑to‑IR opacity ratio (γ), and equilibrium temperature (T_eq). Chemical equilibrium abundances were computed on‑the‑fly using easychem, assuming solar metallicity for refractory species. Line opacities for Fe I/II, Ti I/II, Ca I/II, Mg I, Si I, V I/II, and Na I were taken from the Kurucz line list and processed with pyROX up to 9000 K. To accelerate calculations, the original R = 10⁶ opacity tables were down‑sampled by a factor of three, a practice validated in previous high‑resolution retrievals. Model spectra were convolved to KPF’s resolution (R ≈ 98 000) and scaled by the planet‑to‑star area ratio and the stellar PHOENIX model (T_eff = 10 000 K, v sin i = 115 km s⁻¹). Continuum removal was achieved by subtracting a Gaussian‑smoothed version of the model.

Cross‑correlation functions (CCFs) were computed over a velocity grid of –300 to +300 km s⁻¹ in 1 km s⁻¹ steps, weighting by the observational uncertainties. A Kp‑vs‑v_sys map identified the optimal orbital semi‑amplitude Kp ≈ 245 km s⁻¹ and systemic velocity v_sys ≈ –17.6 km s⁻¹. The planetary signal appears as a bright, sinusoidal trace in the CCF map, shifting from –13.4 ± 0.6 km s⁻¹ at early phases (ϕ ≈ 0.1) to –0.4 ± 1.0 km s⁻¹ near ϕ ≈ 0.45. This progression exceeds the expected rotational broadening (6.4 ± 0.1 km s⁻¹), indicating a dominant day‑to‑night wind component. Bayesian retrieval of the time‑dependent wind speed yields a supersonic east‑west flow of 11.7 ± 0.6 km s⁻¹ on the emerging dayside, decreasing to near zero as the dayside rotates out of view.

The authors compare these measurements to three‑dimensional general circulation models (GCMs) that incorporate radiative heating, drag, and magnetic effects. The observed wind amplitude matches models with weak drag, implying limited magnetic braking in the highly ionized upper atmosphere. This weak drag is consistent with space‑based phase‑curve measurements (Spitzer, TESS, CHEOPS) that suggest a relatively efficient heat redistribution (recirculation efficiency ≈ 0.3). Chemical detection of multiple neutral and ionized metals confirms the expectation that ultra‑hot Jupiters are dominated by atomic species, while the retrieved temperature‑pressure profile shows a strong thermal inversion above 10⁻³–10⁻⁴ bar, reaching temperatures > 4000 K. Constraints on the nightside indicate cooler temperatures (≈ 2500–3000 K) at similar pressures, though the limited phase coverage prevents a detailed night‑side map.

In summary, the study demonstrates that high‑resolution, phase‑resolved emission spectroscopy can directly measure atmospheric dynamics on ultra‑hot Jupiters. KELT‑9 b exhibits the most extreme day‑to‑night wind speeds measured to date (≈ 12 km s⁻¹), providing a stringent benchmark for circulation models operating under extreme irradiation, strong ionization, and potentially magnetic drag. The methodology and results open a new window onto the three‑dimensional climate of the hottest exoplanets, highlighting the power of ground‑based high‑resolution spectroscopy to complement space‑based photometric phase curves.