Unveiling the trends between dust attenuation and galaxy properties at z ~ 2 - 12 with the James Webb Space Telescope

TL;DR

The paper examines how dust attenuation curves, parameterized by the UV slope $S$ and UV bump strength $B$, correlate with fundamental galaxy properties in 173 JWST-detected dusty galaxies across $2.0 \lesssim z \lesssim 11.4$. Using a customized BAGPIPES SED fitting framework with a flexible attenuation law and a nonparametric SFH, the authors find that lower $A_V$ systems tend to have steeper slopes and stronger UV bumps, while $S$ correlates with mass-weighted stellar age and anticorrelates with sSFR. The UV bump shows a strong positive correlation with oxygen abundance $12 + \log(\mathrm{O/H})$ in a Te-based subsample, suggesting metallicity-linked dust properties. They also document redshift-dependent evolution of the attenuation curves and demonstrate that RT effects, radiation-field strength, and dust composition jointly shape the observed trends, though disentangling these requires larger, spatially resolved datasets and dust-evolution modeling in cosmological simulations.

Abstract

A variety of dust attenuation/extinction curves have been observed in high-redshift galaxies, with mixed results regarding their correlations with global galaxy properties. These variations are likely driven by factors such as intrinsic dust properties, total dust content, and the dust-star geometry. In this work, we explore how the shape of dust attenuation curves-quantified by the UV-optical slope (S) and UV bump strength (B)-correlates with galaxy properties. Our goal is to identify the key physical mechanisms shaping attenuation curves through cosmic time. We build on arXiv:2402.05996, analyzing 173 dusty galaxies at z ~ 2-11.5, with attenuation curves inferred via SED fitting of JWST data using a modified version of BAGPIPES (arXiv:2304.11178). We investigate trends between S, B, and properties inferred from SED fitting: AV, SFR, stellar mass (M*), specific SFR (sSFR), mass-weighted stellar age (a*), ionization parameter (U), and metallicity (Z). For a subset, we also consider oxygen abundance (12 + log(O/H)), derived via Te-based methods. We find that lower AV galaxies tend to have steeper slopes and stronger UV bumps, consistent with radiative transfer predictions involving dust geometry and content. S also correlates with a* and sSFR, suggesting that strong radiation fields in young, bursty galaxies may destroy small grains, flattening the slope. B correlates with 12 + log(O/H), possibly due to metallicity-driven dust composition changes. Overall, attenuation curve shapes appear most strongly linked to: (1) redshift (dust evolution), (2) AV (RT effects), (3) a* or sSFR (radiation field), and (4) oxygen abundance (dust composition). Disentangling these effects requires spatially resolved data and theoretical models including dust evolution.

Figures (16)

• Figure 1: Example of the SED fitting results for a galaxy with a UV-bump detection ($1210\_13510$ at $z \approx 5.12$). Left panel: NIRSpec prism spectrum (blue) and best-fit posterior model (orange; top panel). The shaded regions represent the respective $1\sigma$ uncertainties. The vertical gray and red lines indicate the positions of potential emission lines and the UV bump absorption feature, respectively. The residuals of the best-fit model relative to the observed spectrum ($\Delta f_{\lambda}$) are reported in the bottom panel, along with the 1$\sigma$ uncertainties. Right panel: Best-fit dust attenuation curve with $1\sigma$ uncertainties derived from a bootstrap method that generated 5000 attenuation curves by randomly sampling the $c_1-c_4$ parameters (Eq. \ref{['dust_law']}) from the posterior distribution. For comparison, the Calzetti, MW, and SMC empirical curves are shown as solid, dashed, and dotted black lines, respectively.
• Figure 2: Example of SED fitting results for a galaxy with oxygen abundance measurements ($1210\_5173$ at $z \approx 7.99$). The NIRSpec prism spectrum (black) and the best-fit model (red) are depicted with their respective $1\sigma$ uncertainties. The vertical lines indicate the positions of key optical emission lines we used to estimate the oxygen abundance: $\rm{H\gamma}$ and $\rm{H\beta}$ (orange), $[\ion{O}{II}] \lambda3728$ (purple), $[\ion{O}{III}] \lambda4363$ (blue), and $[\ion{O}{III}] \lambda5007$ (green). The vertical dashed gray lines indicate the positions of the remaining emission lines. The olive lines indicate the continuum emission levels.
• Figure 3: Correlation matrix heatmap for the dust attenuation curve properties (the UV-optical slope $S$ and the UV bump amplitude, i.e., $B$ and $c_4$) and fundamental galaxy parameters as inferred from the SED fitting: Redshift ($z$), $V$-band attenuation ($A_V$), SFR, stellar mass ($\log{M}$), sSFR, stellar age (${\langle a \rangle}_*^{\rm{m}}$), ionization parameter ($\log{U}$), and metallicity ($Z$). The color-code represents Pearson correlation coefficient ($r$) values, which quantify the strength and the sign of the correlation. Typically, $|r| = 1$ signifies a perfect correlation, $|r| = 0.7-1$ a very strong correlation, $|r| = 0.5-0.7$ a strong correlation, and $|r| \sim 0.3-0.5$ a moderate correlation (Benesty2009).
• Figure 4: Feature importance resulting from the RFR method. The score assigned to each galaxy property quantifies its importance in predicting the slope ($S$, horizontal sky blue bars) and the UV bump strength ($B$, orange bars).
• Figure 5: Median dust attenuation curves for our full galaxy sample binned in terms of key parameters: (left panel) $V$-band attenuation ($A_V$), and (right panel) mass-weighted stellar age (${\langle a \rangle}_*^{\rm{m}}$). The median attenuation curves are plotted as solid lines, and the shaded regions indicate their $1\sigma$ dispersion. These curves and uncertainties were estimated using a bootstrapping approach, where 5000 synthetic curves were generated and the $16^{th}$, $50^{th}$, and $84^{th}$ percentiles were taken from the resulting distribution. For comparison, the Calzetti, MW, and SMC empirical curves are displayed as solid, dashed, and dotted black lines, respectively. The bin limits were chosen to maintain a roughly equal number of sources per bin ($\sim 40-50$ sources in four bins).