Consistent gas-phase C/O abundances from UV and optical emission lines: a robust scale for chemical evolution across cosmic time

TL;DR

The paper tackles systematic biases in nebular C/O abundances introduced by the abundance discrepancy factor (ADF) by directly comparing C^{2+}/O^{2+} derived from UV collisionally excited lines with optical recombination lines. Using UV spectra from HST/COS for six H II regions (four new UV measurements and two with archival data) and existing RL measurements, the authors quantify the carbon ADF and compare UV CEL- and RL-based C/O values. They find a non-zero ADF for C^{2+} of about $0.40\pm0.02$ dex, closely matching the O^{2+} ADF, and show that log(C^{2+}/O^{2+}) obtained from UV CELs agrees with optical RLs within $\sim0.05$ dex, across several temperature-structure scenarios. This establishes that gas-phase C/O abundances inferred from UV CELs and optical RLs are directly comparable over cosmic time, supporting robust chemical-evolution inferences, including rest-UV measurements at high redshift with JWST.

Abstract

The carbon to oxygen (C/O) abundance ratio is a valuable tracer of star formation history, as C and O enrichment occurs on different timescales. However, measurements based on ultraviolet (UV) collisionally excited lines and those based on optical recombination lines may be subject to biases from the abundance discrepancy factor (ADF), which is well established for oxygen but uncertain for carbon. We present precise UV-based measurements of gas-phase C$^{2+}$/O$^{2+}$ ionic abundance in four H II regions which have prior optical-based measurements, combined with archival UV data for two additional H II regions, in order to establish a reliable abundance scale and to investigate biases between the two methods. We find a clear ADF for the C$^{2+}$ ion which is consistent with that of O$^{2+}$, assuming a similar temperature structure in the zones of the nebula which these ions occupy. The C/O abundance derived from UV collisional lines and optical recombination lines is therefore also consistent to within $<0.1$ dex, with an offset of $0.05\pm0.03$ dex in C$^{2+}$/O$^{2+}$ for the standard T$_e$ method. While the absolute C/H and O/H abundances are subject to large uncertainty from the ADF, our results establish that C/O abundances measured from these different methods can be reliably compared. Thus we confirm the robustness of gas-phase C/O measurements for studying galaxy evolution and star formation timescales, including from rest-UV observations of high redshift galaxies with JWST.

Figures (5)

• Figure 1: Rest frame HST/COS spectra of C$\;$] $\lambda\lambda$1907,1909 and O$\;$] $\lambda\lambda$1660,1666 emission lines. The spectra are binned by 6 pixels and scaled arbitrarily for display purposes. Dashed blue lines show the expected wavelengths of each transition, and best fit Gaussian profiles to each emission line are shown in red. NGC 5408 shows a complex C$\;$] profile, that could not be fit well with a single Gaussian.
• Figure 2: C$^{2+}$/O$^{2+}$ abundance derived for N66A at varying extraction widths of the G185M spectrum, which covers the C$\;$] emission. The star indicates the chosen BOXCAR pixel height of 95 pixels. This was chosen based on the spatial profile of emission seen in the 2-D spectra, as described in the text. Uncertainty due to the extraction width was determined using C$^{2+}$/O$^{2+}$ at heights from 57 to 95. This corresponds to an uncertainty of $\pm0.01$ dex, which is small compared to the statistical uncertainty (pink shaded region). In all targets we find that uncertainty due to the extraction aperture is small compared to other sources of uncertainty.
• Figure 3: log(C$^{2+}$/O$^{2+}$) derived from RLs (from Esteban2014, Esteban2009, and Toribio2017) versus the equivalent UV CEL-based measurements described in Section \ref{['sec:abundances']}. For the UV CEL-based abundances, we used $\rm T_e$([O$\;$]) for both C$^{2+}$ and O$^{2+}$. The UV CEL and optical RL values are in good agreement, all falling near the 1:1 line shown in black. The square symbols represent sources from the literature (Mrk 71 and N88A) which are included in our analysis. We note that N66A and N81 overlap due to their nearly identical values.
• Figure 4: The ADF(O$^{2+}$) versus ADF(C$^{2+}$) for the nearby H$\;$ regions in our sample (teal squares) and three objects from Toribio2017 (gray squares). Values for the ADF(O$^{2+}$) originate from Esteban2014 and Toribio2017. In all cases the ADF(C$^{2+}$) and ADF(O$^{2+}$) are consistent, illustrated by the 1:1 line in black.
• Figure 5: $C^{2+}$/O$^{2+}_{CELs}$ versus $C^{2+}$/O$^{2+}_{RLs}$ for the four cases outlined in Section \ref{['sec:cases']} and summarized in each panel. t$^2$, $\rm T_e$([O$\;$]), $C^{2+}$/O$^{2+}_{RLs}$, and $\rm T_e$([S$\;$]) are from the iterature (see Table \ref{['tab:rls']}). The black points represent the four H$\;$ regions presented and the open points are from the literature. A 1:1 relationship is indicated in purple with all cases showing good agreement. The average offset between the ionic abundances obtained from RLs and CELs is displayed in the bottom right of each panel. The teal line is shifted by the average ADF($O^{2+}$) of our H$\;$ regions, showing the trend we would expect if Carbon did not exhibit an ADF.