SDSS-IV MaNGA: Data-Model Discrepancy in Temperature-sensitive Line Ratios for Star-forming Galaxies

TL;DR

The paper investigates the discrepancy between Te-based direct-method metallicities and photoionization-model–calibrated strong-line metallicities using MaNGA data. By binning spaxels in metallicity and ionization parameter, stacking to detect faint auroral lines, and deriving Te for five ions, the authors compare direct-method abundances to model predictions and find a median offset of about $-0.09$ dex, smaller than previous calibrations but still dependent on $\log U$. They introduce a data-driven dust attenuation correction for low-ionization auroral lines, revealing that standard F99 corrections overcorrect these lines, and apply median-slope corrections to improve consistency. The study finds significant data–model discrepancies, especially for $[O\,II]$ and $[S\,II]$ at high metallicity, indicating limitations of one-dimensional photoionization models to capture the complexity of real H II regions and the need for three-dimensional, density-structured modeling and higher-resolution observations. Overall, the work advances metallicity measurements in spatially resolved galaxies and highlights the ongoing challenges in reconciling Te-based abundances with model-based calibrations, particularly across varying ionization conditions.

Abstract

Gas-phase metallicity is a fundamental parameter that helps constrain the star-forming history and chemical evolution of a galaxy. Measuring electron temperature through auroral-to-strong line ratios is a direct approach to deriving metallicity. However, there is a longstanding discrepancy between metallicity measured through the direct method and that based on the photoionization models. This paper aims to verify and understand the discrepancies. We bin ~ 1.5 million spaxels from SDSS-IV MaNGA according to metallicity and ionization parameters derived from theoretical strong-line calibrations. We stack the spectra of spaxels within each bin and measure the flux of strong lines and faint auroral lines. Auroral lines for [OII], [SII], [OIII], and [SIII] are detected in the stacked spectra of most bins, and the [NII] auroral line is detected in fewer bins. We apply an empirical method to correct dust attenuation, which makes more realistic corrections for low ionization lines. We derive electron temperatures for these five ionic species and measure the oxygen and sulfur abundances using the direct method. We present the resulting abundance measurements and compare them with those model-calibrated strong-line abundances. The chemical abundances measured with the direct method are lower than those derived from the photoionization model, with a median of 0.09 dex. This discrepancy is smaller compared to the results based on other metallicity calibrations previously reported. However, we notice that the direct method could not account for the variation in ionization parameters, indicating that the precise calibration of metallicity using the direct method has yet to be fully realized. We report significant discrepancies between data and the photoionization model, which illustrates that the one-dimensional photoionization model is incapable of representing the complexity of real situations.

Figures (12)

• Figure 1: Density distribution of the data (grey) and the selected samples (colored) in refined optical diagnostic (P$_1$--P$_2$) diagram. The blue grids are the photoionization models for star-forming regions, and the grey grids are the photoionization models for AGNs; each cross corresponds to a unique pair of metallicity and ionization parameter. Both grids are applied a cut so that only the parts that within the middle 98% of the total data along the hidden P$_3$ axis are shown. The definitions of P$_1$ and P$_2$ are in the text. The demarcation line corresponds to $f_{SF}=0.90$, which means the maximum contamination to H$\alpha$ from AGN-ionized regions is 10 %.
• Figure 2: Location of the bins in the metallicity - ionization parameter diagram. Each square represents a bin, and squares are color-coded by the number of spaxels in the bin. Bins with fewer than 50 spaxels are not shown in this diagram.
• Figure 3: Spectra of [S ii] auroral lines (top left), [O ii] auroral lines (top right), [S iii] auroral line (mid left), [O iii] auroral line (mid right), and [N ii] auroral line (bottom left) from a sample bin (Z$^{-0.10}_{-0.15}$, log(U)$^{-2.80}_{-2.85}$) with 2909 spaxels. In each panel, from top to bottom, the three rows correspond to a single spectrum, the stacked spectrum (blue), as well as the best-fit stellar continuum spectrum(red), and the stacked spectrum after removing the stellar continuum, where the auroral lines can be clearly detected. The shaded regions of the top row represent the uncertainties of flux, taken from MaNGA maps files. The shaded regions of the stacked spectra are errors propagated from individual spaxels. In the [O iii] panel, the dashed grey lines represent the [Fe ii]$\lambda$ 4359 (right) and [Fe ii] $\lambda$ 4288 (left).
• Figure 4: Auroral-to-strong line ratio of [S ii] (left) and [O ii] (right) versus normalized Balmer decrement (H$\alpha$ / H$\beta$ / 2.86) in one randomly-chosen metallicity-ionization parameter bin, both in logarithm space. Each data point represents the line ratio of a sub-bin that is constructed by spaxels with similar Balmer decrements. The error bars describe the uncertainties of line ratios. The green line is the F99 correction relation with $R_v$ = 3.1, which is generally steeper than the data distribution. The red line is our fitting for these data points, which considers both the x and y axes uncertainties simultaneously.
• Figure 5: Slopes of relative attenuation distribution in the blue histograms of the linear fitting for [S ii]$\lambda \lambda$ 4069,4076/ [S ii]$\lambda \lambda$ 6716,6731 (left) and [O ii]$\lambda \lambda$ 7320,7330/ [O ii]$\lambda \lambda$ 3726,3729 (middle), and [S iii]$\lambda$ 6312/ [S iii]$\lambda \lambda$ 9068,9531 (right) of all the sub-bins. The green lines represent the slopes of relative attenuation from F99, and the red lines represent the median slopes of 168 bins. The values of median slopes are presented in Table \ref{['table:3_2']}.