2.1 Emission Line Mapping

Emission line mapping from narrow-band imaging requires careful attention due to the potential contamination from undesired emission lines in a filter band-pass and the continuum emission. For NGC 5189, we used the available narrow and wide filters to remove these contaminants. Our process relied on the available narrow-filter observations (F673N, F657N, and F502N) to produce spatially-resolved flux-density maps of the [OIII]$\lambda$5007Å, H$\alpha$$\lambda$6563Å and [SII] $\lambda \lambda$6716,6731Å emission lines, and the wide-filter observations (F606W and F814W) for continuum subtraction. The images were flux-calibrated using the PHOTFLAM descriptor value, which converts count numbers (electrons sec$^{-1}$) into physical flux units of ergcm$^{-2}$s$^{-1}$Å$^{-1}$arcsec$^{-2}$ (pixel size is $0.0396^{2}$ arcsec$^{2}$). The interpolated and extrapolated F606W and F814W flux-density maps were used to estimate the F657N and F502N continuum fluxes. These were then used to subtract from the F657N and F502N flux-density maps, while the stellar contaminated pixels were excluded from the F673N flux-density map using a flux limit according to the stellar mean flux. The flux-density images of the narrow-filter bands were then transformed to mean flux maps by using the RMS bandwidths (PHOTBW header keyword).

The F657N filter bandpass also includes the [NII] $\lambda \lambda$6548,6584 emission lines, so the H$\alpha$$\lambda$6563 emission flux is a fraction of the flux maps derived from the F657N band. Since both the N$^{+}$ and S$^{+}$ ions have roughly similar ionization energies, the [NII] stratification layer should follow the [SII] morphology (see e.g. Fig. 4 in Danehkar et al., 2014). The ratio [NII]/[SII] is measured to be about $\approx 6$ from a $1 \hbox{$^{\prime\prime}$}\times 5 \hbox{$^{\prime\prime}$}$ slit (García-Rojas et al., 2012) that was taken from a bright knot with the MIKE spectrograph (exposure times of 120 and 1800 sec and airmass of 1.25), so the slit covers a tiny fraction of the nebula. Our current analysis suggests that this value is too large for deriving an [NII] distribution from the F673N band, since the resulting H$\alpha$ flux is inconsistent with the [OIII]/H$\alpha$ flux ratio of 4.05 reported by García-Rojas et al. (2012) and 3.95 measured by Kingsburgh & Barlow (1994). The [OIII]$\lambda$5007/H$\alpha$ surface brightness shows little variation across the nebula (see Fig. 5), so a constraint on the [OIII]/H$\alpha$ flux ratio ($\approx 4$) can be employed to estimate a correct value of the [NII]/[SII] ratio. Through an iterative process we found that the ratio [NII] $=$ $4 \times$ [SII] yields a mean value of the dereddened [OIII] $\lambda$5007 flux over the [NII]-corrected H$\alpha$ flux of 3.98 for the $120\hbox {$^{\prime \prime }$}\times 90\hbox {$^{\prime \prime }$}$ region (shown in Fig. 1), which agrees with both Kingsburgh & Barlow (1994) and García-Rojas et al. (2012). Hence, we adopted the [NII]/[SII] ratio of 4 to produce a map of the [NII] emission. We used this synthesized [NII] map to remove the [NII] contamination from the F657N band, leaving a pure H$\alpha$ line-emission image. Note that the [NII]/[SII] ratio will not always be constant when both shock-ionization and photo-ionization are present due to the shock-excitation dependence of this ratio. Additionally, the HST WFC3 images of the Ring Nebula (NGC6720), calibrated using ground-based spectra showed that the F673N filter has an uncertainty around 10 percent while the F502N and F658N fliters do not require any corrections (O'Dell et al., 2013). Unfortunately, the calibration corrections derived by O'Dell et al. (2013) require the F547M filter that could not be performed using the available filter set.

Following the method by Zeidler et al. (2015), we estimated the continuum flux from the F606W and F814W fluxes. We estimated the continua of the H$\alpha$ and [OIII] emission lines by interpolating and extrapolating the fluxes of the F606W and F814W images at the central wavelengths of the F657N and F502N bands, respectively. Although the F606W filter bandpass includes some nebular emission lines such as H$\alpha$ and [OIII], the short exposure time (300sec in F606W when compared to 3900sec in F657N and 8400sec in F502N) and the F606W-F814W interpolation (also used by Zeidler et al., 2015) prevent any large contributions from these nebular lines. For example, see the faint nebular continuum in Figure 1 (bottom-right panel). The estimated continuum of each image was then subtracted for a better removal of the stellar contamination from the narrow-band images in order to get the final, pure line-emission image. The removal worked well for the image in the brightness range between unsaturated and brighter than 1-$\sigma$ sky, but it has problems with saturated objects. The continuum subtraction might fail for extremely faint emission ($< 1\sigma$ sky). As a result, the continuum reduction is not reliable for obtaining a pure line-emission [SII] image from the F673N, since the [SII] $\lambda \lambda$6716,6731 doublet is extremely weak. Instead, we used a flux limit for the F673N band which excludes those bright pixels associated with the stellar contamination. The flux ratio map [SII]/H$\alpha$ is a key diagnostic in our analysis, but since those pixels associated with stellar contamination were already excluded by the continuum-subtraction in the H$\alpha$ flux map, the ratio map is free of stellar contamination. The PHOTBW header keyword ( $\Delta\lambda$ in Table 1), which describes the RMS bandwidth (in Å), was used to calculate the mean flux maps (in ergcm$^{-2}$s$^{-1}$arcsec$^{-2}$) from each calibrated, continuum-subtracted flux-density images (in ergcm${}^{-2}$s${}^{-1}$arcsec${}^{-2}$Å${}^{-1}$).

The continuum-subtracted flux maps were dereddened using the logarithmic extinction of $c({\rm H}\beta)=1.451 \times E(B-V)=0.47$ (García-Rojas et al., 2012). To correct flux images for the interstellar extinction, we utilized the standard Galactic extinction law with $R_V \equiv A(V)/E(B-V)=3.1$ (Howarth, 1983; Seaton, 1979; Cardelli et al., 1989). We also derive the H$\beta$ map from the H$\alpha$ map by adopting the reddened flux ratio $F({\rm H}\alpha)/F({\rm H}\beta)=4.187$ from García-Rojas et al. (2012). For the dereddening process, we assume that the extinction distribution is uniform, however, it could be inhomogeneous due to contributions from dust grains embedded inside the nebula.

The final dereddened, continuum-subtracted flux maps are presented in Figure 1. These images are sky-subtracted, and pixels with values below 1$\sigma$ are masked to only include statistically significant pixels. Figure 1 shows a pair of dense bright shells in the H$\alpha$ and [OIII] flux maps, which are refereed as “envelopes” throughout this paper: one bright shell is extended from the central star toward the northeast of the nebula ( ${\rm PA} \approx 60^\circ$), while another smaller bright shell is extended from the central star toward the southwest of the nebula ( ${\rm PA} \approx 240^\circ$). The [OIII] flux map almost looks similar to the H$\alpha$ flux map, except for some small low emission regions in both sides of the dense bright envelopes in H$\alpha$ emission. The [SII] flux map contains the same dense envelopes, however, the whole nebula is much fainter and contains several bright filamentary and knotty structures. The continuum density-flux presented in Figure 1 (bottom-right panel) is the continuum contamination distribution estimated at the F657N central wavelength using interpolation between the F606W and F814W density-flux maps, and is used to subtract from the H$\alpha$ flux map. A similar continuum density-flux was produced for [OIII] as well.