The DUVET Survey: Direct T_e-based Metallicity Mapping of Metal-enriched Outflows and Metal-poor Inflows in Markarian 1486

We perform four-moment fits to the stellar continuum with model spectra from the Binary Population and Spectral Synthesis code (BPASS, v2.2.1; Eldridge et al. 2017; Stanway & Eldridge 2018) ¹⁰ for both grating settings of each spaxel with pPXF (Cappellari 2017), adopting the "135_300" initial mass function (IMF; refer to Table 1 in Stanway & Eldridge 2018). Since we detect many faint emission lines in our KCWI data, we first run a continuum fit on the summed global spectrum of Mrk 1486 in which the iterative sigma-clipping approach of Cappellari et al. (2002) ¹¹ is employed to mask out these faint emission features. We apply these masks to individual spaxels during subsequent fitting to minimize the impact of faint emission features on the template fitting.

Emission-line fits are performed on the continuum-subtracted data for each spaxel according to the following procedure: velocity (or redshift, z) and velocity dispersion (σ) is obtained by simultaneously fitting single-component Gaussian profiles to the Hβ, Hγ, and [O iii] λ λ 4959, 5007, emission features in the red setting using a χ² minimization procedure. Treating these values as fixed for that spaxel, fluxes are then fit individually for each emission line, using a 30 Å subinterval of the complete spectrum centered on the expected centroid of the line at the best-fit redshift. We fit the [O ii] λ λ 3726, 3729, [Ne iii] λ3869, Hδ, Hγ, and [O iii] λ4363 lines in the blue grating setting, and the Hγ, [O iii] λ4363, Hβ, [O iii] λ4959, [O iii] λ5007 lines in the red grating setting. ¹² [O ii] λ λ 3726, 3729 doublet lines are fit simultaneously, since they are partially blended at this resolution; no restrictions are imposed on the flux ratio. Additionally, we fit the [O iii] λ5007 and [O iii] λ4959 fluxes independently to verify that the saturation of [O iii] λ5007 has been appropriately corrected for in the data reduction (see Section 2). We corrected for dust extinction based on the Hγ/Hβ flux ratio observed in the red grating setting, assuming an intrinsic Balmer decrement of f_Hγ /f_Hβ = 0.468 (Dopita & Sutherland 2003; Groves et al. 2012) and a Cardelli et al. (1989) extinction law with R_V = 3.1. We found a peak extinction of A_V = 0.96 ± 0.12 (E(B − V) = 0.31 ± 0.04) on the disk. At large separations, A_V ≈ 0 was commonly observed, suggesting little extincting dust is present in the outflows and inflows, perhaps due to the low metallicity of the system.

Line-flux uncertainties are estimated using a bootstrapping method in which the observed spectrum is perturbed at each spectral pixel by a normal distribution with standard deviation derived from the KCWI variance cube. We measure fluxes for 100 such synthetic spectra and the standard deviation of the resulting flux distribution is adopted as the 1σ flux uncertainty. Additionally, uncertainties from the reddening correction were propagated through to the line ratios and derived properties presented throughout the remainder of the paper.

Our Keck/KCWI observations afford detections of the critical [O iii] λ4363 auroral line at S/N > 5 per spaxel over a spatial extent of 99 (6.9 kpc) along the major axis and 58 (4.1 kpc) along the minor axis, well beyond R₉₀ (Figure 1). We derive an electron temperature (T_e ) map for Mrk 1486 from the [O iii] λ4363/λ5007 emission-line ratio according to the relation set out by Nicholls et al. (2020), with resultant temperatures in the range 1.21 ± 0.09 ≤ T_e /10⁴ K ≤ 1.81 ± 0.08 (Figure 2, top-left panel). The edge-on orientation of Mrk 1486 enables us to derive T_e profiles along and above the disk plane with consistent methodology, providing unique direct-method metallicity measurements of inflowing and outflowing gas.

Zoom In Zoom Out Reset image size

The upper-right panel of Figure 2 shows our derived minor-axis T_e profile for a 261 wide strip centered on the peak KCWI white-light flux. T_e is seen to peak on the plane of the disk at T_e /10⁴ K = 1.41 ± 0.01 and decreases in both directions. This decrease is most significant in the northwest (NW; red) direction, dropping by (0.17 ± 0.05) × 10⁴ K to T_e /10⁴ K = 1.24 ± 0.05 at R_minor = −1.6 kpc. In the southeast (SE; blue) direction we instead measure a decrease of (0.10 ± 0.07) × 10⁴ K to T_e /10⁴ K = 1.31 ± 0.07 at R_minor = 1.6 kpc. In both directions, the gradient in T_e reverses in the final resolution elements.

Our derived major-axis T_e profile is shown in the lower panels of Figure 2. The highest temperature values preferentially reside at larger radius, consistent with a negative radial metallicity gradient.

Recent studies have shown that metallicity measurements derived from only a single temperature measurement can result in misleading metallicity trends if the ionization conditions of the emitting H ii regions are not constant (Yates et al. 2020; Cameron et al. 2021). Figure 2, therefore, shows the O32 profile ¹³ along the minor and major axes. This ratio is a commonly used ionization parameter (log U) diagnostic. There is a negative O32 minor-axis gradient, consistent with the fiducial assumption that the ionization parameter will be lower further from the disk. Since our DUVET observations of Mrk 1486 yield an auroral line ratio only for [O iii], in the next sections we use predictions from photoionization modeling to explore the contribution of the ionization parameter to the observed temperature gradient, and ultimately infer metallicity variations.

We compare observed R_O3 = [O iii] λ4363 / λ5007 and O32 line ratios from Mrk 1486 to predictions from Mappings V model grids (A. D. Thomas 2021, private communication). These models were computed with Mappings V (Sutherland & Dopita 2017) for a STARBURST99 continuous star formation model, ensuring a fully populated IMF (Leitherer et al. 2014) over a range of nebular metallicity, ionization parameter, and ISM pressure values.

We find that while the R_O3 ratio is dependent on pressure in high-metallicity systems, the effect is minimal below $12+\mathrm{log}({\rm{O}}/{\rm{H}})\lesssim 8.3$, which our R_O3 measurements strongly favor for Mrk 1486. What little pressure dependence R_O3 exhibits at low metallicity is strongest at high pressure. However, our measurements of the [O ii] λ3729/λ3726 doublet ratio, which can be used as an effective pressure diagnostic (Kewley et al. 2019a), strongly favor low pressure (log (P/k) ≲ 6.6) throughout Mrk 1486. Thus, we conclude that our observed temperature variations are not caused by pressure variations. We measure a median density of n_e (O II) = 100 cm⁻³ and infer a median pressure of log (P/k) = 6.0. For the remainder of the paper, we fix the pressure to log (P/k) = 5.8 in our model grids. ¹⁴

Figure 3 shows how predictions for the R_O3 auroral line ratio (which traces T_e ) and the O32 ratio (which traces log U) compare to the observed values along the minor and major axes of Mrk 1486. For visual clarity, we plot the models only for four values of metallicity, each at five different values of ionization parameter. The left panel of Figure 4 then shows a metallicity map for Mrk 1486 derived from interpolation of these grids, performed with a χ² minimization procedure based on the R_O3 and O32 line ratios. The lowest metallicities we infer have $12+\mathrm{log}({\rm{O}}/{\rm{H}})\lesssim 7.297$, lying beyond the low-metallicity extreme of our model grids. Inferring oxygen abundances in this extreme low-metallicity regime is very susceptible to modeling uncertainties and has extremely few empirical constraints from observations (e.g., Senchyna et al. 2019; Stanway & Eldridge 2019). Direct T_e measurements from a lower ionization zone (e.g., T_e (O ii)) are required to robustly determine the oxygen abundance of most metal-poor spaxels in Mrk 1486. We instead adopt $12+\mathrm{log}({\rm{O}}/{\rm{H}})\lesssim 7.297$ as an upper limit in these cases for remainder of this analysis.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figure 3(a) we show observed minor-axis points from a 261 wide region centered on the peak of the white-light flux, with the color bar indicating the projected offset. Particularly notable are the NW (red) measurements decreasing steeply toward lower R_O3. This suggests that the metallicity increases out to a minor-axis offset of R_minor = −1.6 kpc in this direction.

The increase in minor-axis metallicity is clearly seen in the metallicity map (Figure 4). The highest metallicities we observe in Mrk 1486 are offset from the disk plane along the minor axis, particularly in this NW direction. This is consistent with a scenario in which star formation activity drives metal-enriched outflows (see Figure 4, right panel), which has been widely invoked to explain metallicity-scaling relations (e.g., Tremonti et al. 2004; Mannucci et al. 2010; Sanders et al. 2021). The maximum metallicity measured along the minor axis ($12+\mathrm{log}({\rm{O}}/{\rm{H}})={8.10}_{-0.06}^{+0.07}$ at R_minor = −1.6 kpc) is 1.6 times (0.20 dex) higher than the luminosity-weighted average metallicity of the inner disk.

Chisholm et al. (2018) used UV absorption-line measurements from HST's Cosmic Origins Spectrograph (COS) to measure outflow metallicities for a sample of low-redshift galaxies including Mrk 1486. Their data yield a much higher outflow metallicity for Mrk 1486 of Z_outflow/Z_⊙ = 0.56 ± 0.03 ($12+\mathrm{log}({\rm{O}}/{\rm{H}})=8.44$, assuming a solar value of $12+\mathrm{log}{({\rm{O}}/{\rm{H}})}_{\odot }=8.69$).

Using our measurement of the ISM metallicity, this yields an outflow enrichment factor of Z_outflow/Z_ISM = 3.5 for Mrk 1486. ¹⁵ While this value is substantially higher than ours, we note that different metallicity measurement techniques are prone to systematic offsets and values obtained via different methods are difficult to compare (e.g., Kewley & Ellison 2008; Maiolino & Mannucci 2019). Furthermore, the outflow metallicity measured by Chisholm et al. (2018) is derived from kinematically offset absorption lines observed for the 25 aperture of COS, centered on the plane of the disk. This measurement probes gas that is flowing along the line of sight toward us. Our DUVET observations instead probe outflowing gas along a projected spatial offset. It is difficult to understand how these differences in geometry would affect the metallicity.

The metallicity increase is milder in the SE direction, only reaching a maximum value of $12+\mathrm{log}({\rm{O}}/{\rm{H}})={8.02}_{-0.14}^{+0.08}$ at R_minor = 1.7 kpc. This suggests that the bipolar outflows in Mrk 1486 may be asymmetric, and that the NW outflow is perhaps more enriched at this point in the evolution of Mrk 1486. In both minor-axis directions, we observe that the metallicity decreases for the few spaxels at the largest offsets (see also Figure 2, upper-right panel). This could be an indication of inhomogeneity within an outflow lobe, suggesting a clumpy structure. However, we strongly caution against overinterpreting these few data points. These are our faintest spaxels and the [O iii]λ4363 peak line flux is only ∼10⁻¹⁹ erg s⁻¹ cm⁻² Å⁻¹. At these low fluxes, systematic errors from sky subtraction may be important, which are not currently included in our error bars. Hence we stop short of drawing strong conclusions from these points without deeper data.

In Figure 3(b), all observed spaxels are plotted with a discrete color bar distinguishing points within the major-axis 90% i-band flux radius (black) from those beyond it (orange and green). We find that spaxels nearer to the core are generally observed to have higher metallicity. Almost 95% (17/18) of points consistent with very low-metallicity gas ($12+\mathrm{log}({\rm{O}}/{\rm{H}})\lt 7.695$) are situated outside R₉₀. For an edge-on galaxy, a line-of-sight measurement at a given projected radius along the major axis will sample emission from a range of physical radii from the galaxy's center. At a low projected radius, observed line-of-sight emission will be dominated by gas from the inner disk, which is likely to have been enriched by recent star formation activity. If a disk is experiencing accretion, at the largest projected radii the emission may arise from a mixture of gas in the outer stellar disk, and dilution from metal-poor inflows. These measured metallicities are thus likely to be overestimates of the metallicity of accreting gas.

Near the major axis's NE extremity we observe a number of points with $12+\mathrm{log}({\rm{O}}/{\rm{H}})\lesssim 7.297$. This upper limit is at least 0.6 dex lower than the ISM metallicity (Z_inflow/Z_ISM ≲ 0.25). Our observed major-axis metallicity trend does not stabilize before the coverage cuts off, suggesting that even lower metallicities might be observed at larger radii. This indicates that the sources of the gas inflow into Mrk 1486 are likely below ∼5% Z_⊙. From our measurements we report that outflowing gas from Mrk 1486 is at least 6.3 times (0.8 dex) more metal rich than inflowing gas, noting that this value represents a lower limit. Direct T_e measurements from a lower ionization zone (e.g., T_e (O ii)) and improved spatial coverage from deeper data will be important in better constraining this value.

Metal-enriched outflows and metal-poor inflows are frequently invoked as important mechanisms for explaining a number of galaxy observables including scaling relations (e.g., Dalcanton et al. 2004; Tremonti et al. 2004; Mannucci et al. 2010; Sanders et al. 2021) and metallicity gradients (Sharda et al. 2021b, 2021a). The difference between our Z_outflow/Z_ISM and those observed in Chisholm et al. (2018) are roughly a factor of 2–3. This is significant compared to differences in analytical predictions of what values of Z_outflow/Z_ISM are required to match observed mass–metallicity relations (MZRs), particularly in the mass range of 10⁹–10¹⁰ M_⊙. We note that the difference between these two observations, as well as the asymmetry in Z_outflow/Z_ISM we observe in Mrk 1486, are both reasons emphasizing the dire need for observations like those reported here to understand basic observables like the MZR.

Metal loss from metal-enriched outflows is predicted to be a critical parameter in determining metallicity gradients. The model outlined by Sharda et al. (2021b) quantifies this by the so-called "yield reduction factor", ϕ_y . Larger values of ϕ_y imply that metals are mixed efficiently with the ISM prior to ejection, and hence metal retention is higher. This results in steeper metallicity gradients and a larger possible range of gradients. Adopting mass-loading factors μ ≈ 1–4 (Heckman et al. 2015; Chisholm et al. 2018), our measurement of Z_outflow/Z_ISM from Mrk 1486 yields an estimated value in the range ϕ_y ≈ 0.8–0.95 (refer to Appendix A in Sharda et al. 2021a). This predicts that galaxies with masses ∼10^8.5–10^9.5 M_⊙ have steeper gradients than those observed in large IFU surveys (Figure 2 in Sharda et al. 2021a). On the other hand, models with ϕ_y in our measured range reproduce global metallicity-scaling relations more readily. We note, however, that Mrk 1486 is a multiple sigma outlier below the global MZR, and it remains unclear how appropriate these generalized models are for starbursting galaxies. Finally, we note that our observations probe only the warm (∼10⁴ K) ionized outflowing phase. Thus, our estimates of the outflow metallicity do not include metals in hotter phases, which may be significant (e.g., Lopez et al. 2020). Any metal enrichment in hot-phase outflows would decrease ϕ_y for the same mass-loading factor and, in the absence of hot-phase outflow measurements, the quoted value of ϕ_y should be considered as an upper limit.