Transmission spectrum of WASP-19b based on our FORS2 observations (black, filled dots), compared to values from Huitson et al. (2013, violet filled stars) (obtained spectroscopically), Mancini et al. (2013, red squares), Lendl et al. (2013, green open circle), and Tregloan-Reed et al. (2013, orange open triangle) (from photometry). The vertical bars represent the errors in the fractional radius determination, while the horizontal bars are the FWHM of the passbands used. We note the high spectral resolution of the FORS2 data, compared to what was available until now. The dashed lines represent the weighted mean plus or minus three scale heights.

Examples of the one-dimensional extracted spectra of the target and reference stars. The target star with the transiting planet, WASP-19, is shown in blue, while the target in aperture 7, which we use as comparison star, is shown in black.

Raw light curves for four of the spectroscopic channels. The plots with circles are the light curves for WASP-19 and the triangles represent the reference star in aperture 7. The extinction functions (Eq. (1)), used for the purpose of detrending these raw light curves, are also shown as solid lines. An offset has been added to the light curves for clarity.

Red noise impact, where standard deviation is calculated as a function of bin size, shown for the broadband and 4 spectrophotometric channels (separate panels) for sliding bins. Solid and dashed lines represent the relation, as expected from uncorrelated (white) noise. Deviation from this model, apparent as a straight line on the log−log plots, is evidence of minimal correlated (red) noise in our data. Using wavelets to model the red noise in all the channels ensured that the impact of this correlated noise is accounted for.

Broadband (white) light curve modelled with the GEMC+PRISM code for the purpose of stellar spot characterization, values for which are shown in Table A.2.

Band-integrated spectrophotometric light curves for the transit of WASP-19b. The central wavelength for each channel is indicated on the left-hand side of each plot, where the integration width is mostly 20 nm. The modelled light curves for each channel are shown as solid lines.

Variation in the linear limb darkening coefficient of the quadratic law, with wavelength, for the spectrophotometric channels. The values were allowed to vary between 0 and 1, as dictated by theory, for all the individual channels. The variations and the general trend agree with theoretical values calculated for photometric filters in our chosen range, from Claret & Bloemen (2011).