SHELLQs-JWST: Revealing the Spectra of Extended Emission in 12 z > 6 Quasar Host Galaxies using the JWST NIRSpec Fixed Slit

TL;DR

This study leverages JWST/NIRSpec Fixed Slit spectroscopy to disentangle quasar light from extended host emission in 12 z>6 SHELLQs quasars. By modeling the spatial PSF with a two-Gaussian fit derived from a calibration star and applying a multi-round subtraction, the authors reveal host continua and narrow emission lines in most targets, enabling measurements of star formation rates and ionization properties. The extended-line diagnostics place the hosts near or above the high-z AGN/galaxy regimes, yet many hosts align with the z≈6 star-forming main sequence, suggesting limited quasar feedback on 10 Myr timescales. Notably, J1146+0124 exhibits a central outflow in [OIII]5008, while J0217−0208 shows evidence for rotation, underscoring structural diversity in early quasar hosts and highlighting the efficacy and limitations of PSF-based decomposition in fixed-slit JWST data for probing SMBH–galaxy coevolution at cosmic dawn.

Abstract

We present an analysis of the rest frame optical JWST NIRSpec Fixed Slit spectra of extended host galaxy emission in 12 quasars from the Subaru High-z Exploration of Low-Luminosity Quasars (SHELLQs) sample at redshifts 6.0 < z < 6.4. The spatial point spread function is modeled primarily by a sum of two Gaussians as a function of wavelength and is used to fit and subtract the quasar from the 2D spectra, leaving only extended galaxy emission which we analyze. Ten of 12 systems show spatially extended line emission and five of 12 systems show an extended stellar continuum. From the extended [OIII]5008 emission line, we measure a 132 ${\pm}$ 19 km/s ionized outflow in one system and 52 ${\pm}$ 12 km/s rotation, suggesting a coherent disk, in another. From the extended narrow H$α$ emission, which we hypothesize is ionized by star-forming regions rather than the quasar, we measure star formation rates ranging from ${\sim}$ 7 to 111 M${_\odot}$/yr, the majority of which are consistent with the star-forming main sequence at z ${\approx}$ 6. The positions of our host galaxies on the log10[OIII]5008/H$β$ vs. log10[NII]6584/H$α$ (R3N2) Baldwin-Phillips-Terlevich (BPT) diagram indicate ionization rates typical of AGN activity in the low-redshift universe, but are consistent with the placement of similar z ${\approx}$ 6 quasar host galaxies, suggesting that the R3N2 line ratios cannot distinguish AGN and star-formation powered line emission at high redshifts. We conclude from the consistency between our quasar host sample with z ${\sim}$ 6 galaxies that the presence of a low-luminosity AGN causes little significant change in the properties of galaxies at z ${\approx}$ 6 on 10 Myr timescales.

Figures (21)

• Figure 1: Example of the wavelength dependence of the centroid ($\mu$), amplitude ($\rm{A}$), and width ($\sigma$) of a single Gaussian fit $\rm{(PSF_{1G})}$ to the level 3 spatial profile of star 1808347 at each wavelength, where the blue line represents the parameters of the initial fit to the data and the red line is the same data smoothed by a 30 pixel rolling median. The centroid and width increase with wavelength, and both show a distinct quasi-periodic pattern of wiggles, which are discussed in $\S$\ref{['subsec:width variation']}. We fit the level 3 (rectified) rather than the level 2 (unrectified) 2D spectrum here to better illustrate the variation of the centroid with wavelength, since the dynamic range of $\mu$ is too large in level 2 data.
• Figure 2: 2D (left) and 1D (right) level 3 residuals of three models for the PSF along the spatial direction across the full wavelength range of the detector. All 2D spectra are normalized to the maximum value of the calibration star's 2D spectrum and are presented using the same stretch, although different colors, with all values less than zero shown in white. The y-axes for both 2D spectra and 1D residuals are pixels in the cross-dispersion direction of the detector, with the PSF center placed at zero. The 1D residuals are plotted horizontally in order to match the orientation of the 2D spectra. The residual profile averaged over all wavelengths are shown for all four plots, with the profile of interest for each row bolded. First row: The left panel shows the 2D spectrum of the calibration star 1808347 used as a PSF, and the right panel shows the average spatial profile of the calibration star summed over all wavelengths and normalized to a peak of unity. In this and the following panels, red represents the star profile. Second row: Single Gaussian model $\rm{(PSF_{1G})}$ residuals. While the central four spatial pixels are fit well, the right panel (which has been enlarged compared to the right panel of the first row) shows that the wings of the profile have residuals on the order of 4$\%$, highlighted in blue. Third row: Double Gaussian model $\rm{(PSF_{2G})}$ residuals. Residuals are reduced to the order of $\approx1\%$, as shown in green in the right panel. The profiles of the Single Gaussian model (blue) and star (red) are shown at lowered opacity in the background of the right panel for comparison. Fourth row: Final model $\rm{(PSF_{F})}$ residuals, showing that the full PSF model leaves uncorrelated and small 2D residuals, which, when summed over all wavelengths (black), have an amplitude of less than $0.05\%$.
• Figure 3: Example of the double Gaussian model fit to a single wavelength of the 2D spectrum of calibration star 1808347. Only the pixels being fit by the model are shown. The data are shown as a black histogram with gray fill. The green line with green fill shows the shape of the primary Gaussian $(\rm{A_1}, \sigma_1, \mu)$, while the pink line with pink fill shows the shape of the secondary Gaussian $(\rm{A_2}, \sigma_2, \mu)$. The sum of the two Gaussians (i.e. the full double Gaussian model) is shown by the blue line. The residual of the double Gaussian model fit subtracted from the data is shown as a black histogram, which lies very close to zero and is also shown in the zoomed-in panel below the main figure. The residuals across all wavelengths are smoothed in the wavelength direction using a 30-pixel rolling median and are added to the double Gaussian model to create the final PSF model. We emphasize that the primary and secondary Gaussians have no physical meaning and their form was chosen purely to best fit the PSF. It is therefore incorrect to interpret the secondary Gaussian as representing a more spatially extended physical source.
• Figure 4: PSF width variation with wavelength. The smoothed data for each of the 12 individual quasars is shown in gray, while the average of the individual dithers for all twelve quasars is shown in green. Solid lines are combined and rectified level 3 data, and dash-dotted lines are for individual level 2 dithers (one per quasar). The pink lines represent the PSF width of star 1808347 for the level 2 (dash-dot) and level 3 (solid) data. All widths are from the Single Gaussian model $\rm{(PSF_{1G})}$ to illustrate aspects of the PSF which would be more complex to plot in the Double Gaussian model. We highlight in this figure two distinct and important phenomena: the width variation with wavelength (i.e. wiggles), and the broadening of the PSF in the level 3 data. The wiggles are present across all sources, implying this is a systematic rather than a physical effect. The rectified and co-added level 3 PSF width is on average broader by $\sim0.15$ pixels than the individual level 2 dithers. To avoid undue broadening of the PSF, we model and subtract off the PSF in the individual level 2 dithers before re-entering them into the regular Stage 3 jwst pipeline.
• Figure 5: See $\S$\ref{['sec:individual objects']} for discussion of the plot contents. Panels $(a)$ and $(b)$: Show the 2D spectrum before $(a)$ and after $(b)$ PSF subtraction for J0911$+$0152. Note that the aspect ratio of the figure means that the pixels are not square. The y-axes span $1".2$ (12 pixels). The x-axes match that of panel $(c)$. The edges of the central two pixels in which the quasar dominates are marked by horizontal blue lines. Panel $(c)$: 1D quasar + galaxy spectrum (gray) and PSF subtracted spectrum (light blue). The wavelengths of prominent emission lines are marked with vertical dashed lines, and the linear continuum fit to the extended galactic emission is shown as a bold, orange dashed line. The shaded regions show the wavelength ranges over which the continuum and the emission line spatial profiles are averaged to produce the extended continuum (magenta), $\rm{[OIII]}5008$ (green), and $\rm{H}\alpha$ (pink) profiles shown in panel $(e)$. The response function of the F356W filter as a function of wavelength, normalized to half the plot height, is shown in light gray for illustrative purposes. In both inset plots, the best fit emission line model is plotted as a smooth red line, and the emission line fit residual is represented by a black histogram. The individual $\rm{[NII]}6551$, $\rm{[NII]}6585$, and $\rm{H}\alpha$ line fits are plotted in cyan, magenta, and lime green, respectively. Panel $(d)$: $1".78 \times 1".78$ postage stamp cutout of the NIRCam F356W image of the host galaxy with the quasar point source and any nearby galaxies subtracted off, taken from ding2025shellqsjwstunveilshostgalaxies. The position and angular size at the time of observation of the S200A2 slit is overlaid on the image with the central 12 NIRSpec pixels used for data analysis numbered for comparison with panels (a), (b), and (e). The central two pixels where the quasar dominates are shaded gray. The two $0".2$ by $0".5$ regions outlined in white define the modified slit in which we perform our flux calculations. Also overlaid is the ellipse from which we extract the total host galaxy flux used to calculate the star formation rate for the whole galaxy ($\S$\ref{['subsec:star formation rates']}). Panel $(e)$: Average spatial profiles per NIRSpec pixel of the extended continuum (magenta), $\rm{[OIII]}5008$ (green) and $\rm{H}\alpha$ (pink) emission lines, and the photometry convolved with the S200A2 Fixed Slit aperture (blue), each offset by 0.1 pixels to prevent overlap of error bars. The central two pixels in which the quasar dominates and the galaxy is over-subtracted is shaded light gray.