There Is More to Outshining: 2D Dust Effects on Stellar Mass Estimates at $3 \leq z < 9$ with JWST in the JADES Field

TL;DR

This study examines biases in stellar mass estimates when using galaxy-integrated SED fitting versus spatially resolved (pixel-by-pixel) SED fitting for a large sample of high-redshift galaxies in GOODS-South observed by JWST/JADES and HST. By applying a flexible two-component dust attenuation model and exploring multiple star-formation histories, the authors quantify how dust geometry and age gradients drive differences between the two fitting approaches. They find that resolved masses are consistently higher, with a mass-dependent offset that is largest at low stellar masses and correlates with differences in mass-weighted ages and dust attenuation slopes; resolved analyses also reveal higher A_V and grayer attenuation curves. The results highlight the importance of accounting for 2D dust structure and complex SFHs when deriving stellar masses from integrated light and offer practical guidance for improving mass estimates in large, high-redshift galaxy samples.

Abstract

Dust attenuation modifies the observed spectral energy distribution (SED), leading to biases in the properties inferred from integrated SED fitting. As spatially resolved SED modeling becomes feasible for large high-redshift samples, it is increasingly important to assess how dust attenuation affects resolved mass estimates. We evaluate the impact of dust attenuation on stellar mass estimates derived from integrating spatially resolved SED fitting results. We perform spatially resolved and integrated SED fitting on a sample of 3408 galaxies at $3 \leq z < 9$ from the GOODS South field, combining deep NIRCam from the JWST Advanced Deep Extragalactic Survey (JADES) and HST/ACS imaging from GOODS and CANDELS. We compare galaxy-integrated properties derived from fitting the summed SED with those obtained from spatially resolved SED modeling. Using a two-component dust attenuation model with a variable slope, we investigate how the dust attenuation slope, A(V), and stellar population properties contribute to discrepancies in the resulting stellar mass estimates. Resolved stellar masses are systematically higher than integrated estimates, with a median offset of +0.24 dex. Resolved analyses recover higher dust attenuations ($ΔA(V)\approx +0.08$ mag), lower birth cloud fractions ($Δμ\approx -0.28$), and grayer attenuation curves ($Δδ_{\mathrm{ISM}} = +0.08$), arising from preferential sampling of compact star-forming regions. Integrated fits underestimate stellar ages by $\sim23\%$ at $z < 5$ and 31$\%$ at $z \gtrsim 5$. The stellar mass offset correlates strongly with the age difference and the attenuation slope difference, indicating that age-dependent outshining and spatially varying dust geometry are primary drivers of the discrepancy between resolved and integrated stellar masses.

Figures (7)

• Figure 1: Characterization of the galaxy sample used in this work. Left: redshift distribution of the sample. The distribution at $5 \leq z < 9$ is magnified in the inset for better visualization. Middle: distribution of stellar masses derived from SED fitting based on aperture-photometry fluxes for the sample from Merlin2024. Right: offset from the star-forming main sequence, defined as $\Delta \mathrm{MS} = \log(\mathrm{SFR_{SED}}) - \log(\mathrm{SFR_{MS}}(M_{\star}, z))$, as a function of redshift, computed relative to the evolving main-sequence relation of Speagle2014. Black circles indicate binned medians with 16th–84th percentile ranges. The shaded region denotes the $\pm1\sigma$ intrinsic scatter of the main sequence, and the dashed lines mark the $\pm0.6$ dex thresholds commonly used to separate starburst and quiescent regimes.
• Figure 2: Resolved specific star formation rate (sSFR) in Gyr$^{-1}$ as a function of resolved stellar mass for the galaxy sample used in this work, color-coded by redshift. Each point represents an individual galaxy. The $y$-axis on the right side shows the corresponding star-formation timescale ($1/\mathrm{sSFR}$) in Gyr. The solid black line indicates the star-forming main sequence (MS) from Speagle2014, evaluated at $z \simeq 3.6$, while the dashed lines denote offsets of $\pm 0.6$ dex (a factor of four) relative to the MS. Hexagonal and triangular symbols mark the median sSFR in bins of stellar mass for galaxies in the redshift ranges $3 \leq z < 5$ and $5 \leq z < 9$, respectively. The representative median measurement uncertainty in $\log \mathrm{M}_\star$ and $\log \mathrm{sSFR}$ is shown by the black error bar in the upper-left corner.
• Figure 3: Resolved dust attenuation $A(V)$ versus stellar mass, color-coded by mass-weighted stellar age. Large circles show binned medians with error bars indicating the 16th-84th percentile spread in $A(V)$. Literature comparisons from Arteaga2023 (turquoise triangles) and arteaga2024 (red triangle) are shown for reference. The error bar in the lower-right corner represents median measurement uncertainties.
• Figure 4: Comparison between resolved and integrated stellar masses, corrected for aperture as described in Section \ref{['processing']}, color-coded by the main attenuation parameters derived from the resolved SED fits: total $A_V$ (top left), ISM attenuation slope $\delta_{\mathrm{ISM}}$ (top right), $\mu$ (bottom left; the fraction of total attenuation arising in the diffuse ISM), and the UV continuum slope $\beta$ (bottom right). Each panel shows the relation between $\log M_{\star,\mathrm{resolved}}$ and $\log M_{\star,\mathrm{integrated}}$ for galaxies at $3 \leq z < 5$ (circles) and $5 \leq z < 9$ (squares), with median binned values shown in color and error bars indicating the $1\sigma$ dispersion. Solid, dashed, and dotted lines mark the one-to-one relation and offsets of $\pm 0.5$ and $\pm 1.0$ dex, respectively. The arrow indicates the direction of increasing attenuation parameter. Yellow squares, turquoise triangles, and red triangles mark, respectively, the measurements from Lines2025, Arteaga2023, and arteaga2024.
• Figure 5: Difference between resolved and integrated stellar masses as a function of resolved stellar mass, for different SFH assumptions. Error bars indicate the 16th-84th percentile range in each bin. Results are shown for constant SFHs (CSFH) with minimum stellar ages of 10, 25, 50, and 100 Myr, delayed SFH, and delayed + burst SFH.