Electron temperature relations and the direct N, O, Ne, S and Ar abundances of 49959 star-forming galaxies in DESI Data Release 2

Abstract

We present the largest direct-method abundance catalogue of galaxies to date, containing measurements of 49$\,$959 star-forming galaxies at $z < 0.96$ from DESI Data Release 2. By directly measuring electron temperatures across multiple ionization zones, we provide constraints on a number of electron temperature relations finding good consistency with previous literature relations. Using these temperature measurements, we derive reliable abundances for N, O, Ne, S and Ar and measure the evolution of abundances and abundance ratios of as a function of metallicity and other galaxy properties. Our measurements include direct oxygen abundances for 49$\,$766 galaxies, leading to the discovery of the two most metal-poor galaxies in the nearby Universe, with oxygen abundances of $\rm 12+\log(O/H) = 6.77_{-0.03}^{+0.03}~\rm dex$ (1.2\% $\rm Z_{\odot}$) and $\rm 12+\log(O/H) = 6.81_{-0.04}^{+0.04}~\rm dex$ (1.3\% $\rm Z_{\odot}$). We identify a rare outlier population of 139 galaxies with high N/O ratios at low metallicity, reminiscent of galaxy abundances observed in the early Universe. We find these high N/O galaxies are more massive than typical galaxies at the same metallicity. We find the Ne/O ratio is constant at low metallicity but increases significantly at $\rm 12+log(O/H) > 8.105\pm0.004$ dex. We show that the S/O and Ar/O abundance ratios are strongly correlated, consistent with the expected additional Type Ia enrichment channel for S and Ar. In this work we present an initial survey of the key properties of the sample, with this dataset serving as a foundation for extensive future work on galaxy abundances at low redshift.

Figures (12)

• Figure 1: Figure of an example DESI spectrum of DESI J031.5272+08.5595 with the fitted FastSpecFitmoustakas2023 model. The spectrum is shown in black with uncertainties in the flux measurements shown by the grey band. The fitted model is shown in red. We highlight the fitted emission lines in each panel with pink vertical bands and the names of the fitted lines. In the bottom panel we show five inset figures showing the fitted emission lines of the [Sii]$\lambda\lambda$4068,4076, [Oiii]$\lambda$4363, [Nii]$\lambda$5755, [Siii]$\lambda$6312 and [Oii]$\lambda\lambda$7320,7330 auroral lines. This spectrum was specifically chosen as all five auroral lines/doublets are detected. The inset image in the top panel shows the Legacy Survey imaging of this target with the 1.5" aperture of the DESI optical fiber at the pointing location of the observation as well as a 10" aperture for scale.
• Figure 2: The redshift distribution of the galaxies with temperature measurements in our sample (grey). The vertical lines show the maximum redshift where the $T_{\rm e}$ can be determined based on the listed ionic species. The coloured histogram outlines in matching colours show the measurement distribution of electron temperatures of each ion.
• Figure 3: The histograms of the electron temperature measurements derived from the auroral emission lines of different ions. The red dashed lines show the median electron temperature measured using each auroral line. The [Oiii]$\lambda$4363 auroral line (corresponding to $T_{\rm e}\rm (O^{++})$) is brightest at high temperatures, the [Siii]$\lambda$6312 emission line (correcponding to $T_{\rm e}\rm (S^{++})$) is brightest at intermediate temperatures and $T_{\rm e}\rm (O^{+})$, $T_{\rm e}\rm (N^{+})$ and $T_{\rm e}\rm (S^{+})$ are brightest at low temperatures. The number of objects with temperature constraints using each ion are indicated in the top-right corner of each panel.
• Figure 4: The electron temperature relation between the high-ionisation zone, $T_{\rm e}(\mathrm{O}^{++})$, and the intermediate-ionisation zone, $T_{\rm e}(\mathrm{S}^{++})$. The red line shows the best fit relation derived from our observations. The shaded region shows the 1$\sigma$ uncertainty in the relation. The total scatter in the relation is $\sigma_{\rm tot}(\mathrm{S}^{++}) =1700~\mathrm{K}$ and the intrinsic scatter is $\sigma_{\rm int}(\mathrm{S}^{++})=900~\mathrm{K}$. The total scatter in $T_{\rm e}(\mathrm{O}^{++})$ is $\sigma_{\rm tot}(\mathrm{O}^{++}) = 1600~\mathrm{K}$ and with intrinsic scatter $\sigma_{\rm int}(\mathrm{O}^{++})=900~\mathrm{K}$. The literature relations by izotov2006, croxall2016 and rogers2021 and mendez-delgado2023 are shown for comparison as indicated in the legend. All data points are coloured by the Gaussian kernel density of the plotted distribution.
• Figure 5: The electron temperature relation between the high-ionisation zone, $T_{\rm e}(\mathrm{O}^{++})$, and the low-ionisation zone, $T_{\rm e}(\mathrm{O}^{+})$. The red line shows the best fit relation derived from our observations. The shaded region shows the 1$\sigma$ uncertainty in the relation. The total scatter in the relation is $\sigma_{\rm tot}(\mathrm{O}^{+}) =2100~\mathrm{K}$ and the intrinsic scatter is $\sigma_{\rm int}(\mathrm{O}^{+})=1700~\mathrm{K}$. The total scatter in $T_{\rm e}(\mathrm{O}^{++})$ is smaller at $\sigma_{\rm tot}(\mathrm{O}^{++}) = 1400~\mathrm{K}$ and with intrinsic scatter $\sigma_{\rm int}(\mathrm{O}^{++})=1100~\mathrm{K}$. The literature relations by campbell1986, izotov2006, andrews2013 and cataldi2025 are shown for comparison as indicated in the legend. All data points are coloured by the Gaussian kernel density of the plotted distribution.