Reconciling the Stellar and Nebular Spectra of High Redshift Galaxies

TL;DR

This study demonstrates that reconciling rest-frame far-UV stellar spectra with rest-frame optical/UV nebular emission in high-redshift galaxies requires self-consistent stellar population and photoionization modeling that includes massive-star binaries. The analysis of a representative z ~ 2.4 sample (KBSS-LM1) finds that low stellar metallicity (Z* ≈ 0.1 Z⊙) combined with higher nebular metallicity (Zneb ≈ 0.5 Z⊙) and enhanced O/Fe due to core-collapse supernova yields best explains the observed spectral features, line ratios, and electron temperatures. BPASSv2 binary models reproduce the hardest EUV spectra and the observed He II signatures, while single-star models fail to simultaneously fit all constraints. The inferred gas-phase abundances satisfy local N/O and C/O patterns when placed on the REL scale, and the results imply that high-z galaxies exhibit elevated nebular excitation at fixed O/H due to non-solar abundance patterns and binary-driven stellar evolution, with important implications for interpreting strong-line metallicities and planning future JWST observations.

Abstract

We present a combined analysis of rest-frame far-UV (1000-2000 A) and rest-frame optical (3600-7000 A) composite spectra formed from very deep observations of a sample of 30 star-forming galaxies with z=2.4+/-0.1, selected to be representative of the full KBSS-MOSFIRE spectroscopic survey. Since the same massive stars are responsible for the observed FUV continuum and the excitation of the observed nebular emission, a self-consistent stellar population synthesis model must simultaneously match the details of the far-UV stellar+nebular continuum and-- when inserted as the excitation source in photoionization models-- account for all observed nebular emission line ratios. We find that only models including massive star binaries, having low stellar metallicity (Z_*/Z_{sun} ~ 0.1) but relatively high ionized gas-phase oxygen abundances (Z_{neb}/Z_{sun} ~ 0.5), can successfully match all of the observational constraints. We argue that this apparent discrepancy is naturally explained by highly super-solar O/Fe [4-5 times (O/Fe)_{sun}], expected for gas whose enrichment is dominated by the products of core-collapse supernovae. Once the correct ionizing spectrum is identified, photoionization models reproduce all of the observed strong emission line ratios, the direct T_e measurement of O/H, and allow accurate measurement of the gas-phase abundance ratios of N/O and C/O -- both of which are significantly sub-solar but, as for O/Fe, are in remarkable agreement with abundance patterns observed in Galactic thick disk, bulge, and halo stars with similar O/H. High nebular excitation is the rule at high-z (and rare at low-z) because of systematically shorter enrichment timescales (<<1 Gyr): low Fe/O environments produce harder (and longer-lived) stellar EUV spectra at a given O/H, enhanced by dramatic effects on the evolution of massive star binaries.

Figures (17)

• Figure 1: Portions of the MOSFIRE KBSS-LM1 composite spectra (comprised of the same 30 galaxies included in the composite LRIS spectrum in Figures \ref{['fig:all_deep_full']} in the J, H, and K bands (top, middle, and bottom panels, respectively). Each panel also shows (red) the 1$$ error spectrum for the composite. The locations of various strong emission lines are indicated in each panel; the green curve is the adopted stellar continuum fit. The ordinate in each panel is the average flux density that would be observed if all objects were at the mean redshift, $z=2.396$, while the wavelength scale is shifted to the rest frame.
• Figure 2: Stacked composite rest-frame UV spectrum of 30 galaxies in the initial KBSS-LM1 sample, with $\langle z \rangle = 2.396\pm0.111$ (black histogram). Some prominent emission and absorption features are identified, with color-coded labels: stellar absorption features (red), interstellar absorption features (dark green), nebular emission lines (blue), and excited fine structure emission lines (dark violet). The emission line spectrum is discussed in §\ref{['sec:measurements']}.
• Figure 3: An example illustrating the calculated contribution of nebular continuum emission (blue) to a BPASSv2-z001-100bin stellar population synthesis model assuming continuous star formation over $10^8$ years. The red spectrum is that of the stars only, while the black spectrum is the sum of the stellar and nebular continuum that would be used for comparison to the observed spectrum. The spectra are scaled such that the purely stellar spectrum is normalized to $f_{ } = 1$ at 1500 Å. For this particular model, the contribution of the nebular continuum emission to the total ranges from $\simeq 3.5$% at 1300 Å to $\simeq 6.5$% at 2200 Å.
• Figure 4: As in Fig. \ref{['fig:all_deep_full']}, the KBSS-LM1 composite FUV spectrum is plotted as a black histogram. The observed spectrum has been corrected for the mean intergalactic+circumgalactic attenuation appropriate at $z\simeq 2.40$ (rudie13). Superposed are two of the best-fitting population synthesis models (see Table \ref{['tab:chi2']} and §\ref{['sec:model_details']}). Both model spectra include the predicted contribution of the nebular continuum emission, calculated using the photoionization models described in §\ref{['sec:PSMs']}. The model spectra were reddened assuming the calzetti00 starburst attenuation law, where $E(B-V)$ was adjusted to minimize $^2$ with respect to the observed spectrum after masking spectral regions containing strong interstellar absorption or nebular emission lines (violet).
• Figure 5: Zoom in of the spectral regions near the N5 (left panels) and C4 (right panels) wind lines, comparing various S99 (top) and BPASSv2 (bottom) models with the KBSS-LM1 spectrum. The top panels include three different IMFs for the S99-z002 models to illustrate the IMF dependence of the P-Cygni profiles. As in Figure \ref{['fig:all_deep_full_wmods']}, regions of the spectrum that were excluded from the global fits (see §\ref{['sec:PSMs']}) are shaded in cyan. The color-coding of the line identifications is the same as in Fig. \ref{['fig:all_deep_full']}.