Dust Extinction of the Stellar Continua in High Redshift Galaxies (2 < z <7): the Ultraviolet and Optical Extinction Law

Dust in galaxies absorbs a substantial fraction —often exceeding half— of the stellar energy emitted at the ultraviolet and optical wavelengths, significantly impacting our interpreta- tion of galaxy evolution. Accurately quantifying dust obscuration, particularly in galaxies at intermediate to high redshifts, remains a major challenge, introducing considerable un- certainties in the derivation of key physical parameters such as star formation rates, stellar ages, luminosity functions, and more. However, recent observations, such as those from JWST, are increasing the number of spectroscopically confirmed high-redshift galaxies, enabling more detailed studies of dust attenuation. The aim of this work is to derive the dust attenuation curve for star-forming galaxies at high redshift. We make use of the JADES Data Release 3 and the ASTRODEEP photometric catalog from JWST. The selected sample consists of ∼ 100 star-forming galaxies with stellar masses in the range 9 ≲ log(M∗/M⊙) ≲ 11, spanning redshifts 2 ≲ z ≲ 7. Following the method of Calzetti et al. (1994), we characterize the dust attenuation using the UV power-law index, β, and the Balmer optical depth, τlB = τHβ − τHα. The UV slope is derived by interpolating photometric bands, while the Balmer optical depth is estimated from fully resolved Hα and Hβ emission lines, obtained using medium- resolution gratings. We investigate the β − τlB relation for the entire sample, and subsequently divide the sample into stellar mass bins, focusing on galaxies with log(M∗/M⊙) ≥ 9 —a range where dust absorption is expected to be significant. The sample was then further divided into bins of τlB, and, using low-dispersion prism spectra —particularly suited for studying continuum features— we constructed average spectral templates for each bin. Finally, we derived a selective attenuation curve, Q(λ), over the wavelength range 1500 Å < λ < 11400 Å. The obtained selective attenuation curve is in agreement with findings at 1 ≲ z ≲ 2 by Reddy et al. (2015), Shivaei et al. (2020) and Battisti et al. (2022), and in tension with the recent claims of strong evolution over 2 < z < 11 by Markov et al. (2024), which are based on modeling assumptions.