Spatially Resolved Outflows in a Seyfert Galaxy at z = 2.39

TL;DR

This study provides the first spatially resolved analysis of outflows in a high-redshift Seyfert galaxy, revealing AGN-driven outflows with minimal impact on star formation, thanks to gravitational lensing and adaptive optics.

Contribution

It offers a detailed spatial examination of AGN outflows at cosmic noon using gravitational lensing and adaptive optics, a novel approach for such distant galaxies.

Findings

AGN-driven outflows are present with broad emission line components.

The outflows are consistent with AGN ionization diagnostics.

The AGN does not significantly suppress star formation in the host galaxy.

Abstract

We present the first spatially resolved analysis of rest-frame optical and UV imaging and spectroscopy for a lensed galaxy at z = 2.39 hosting a Seyfert active galactic nucleus (AGN). Proximity to a natural guide star has enabled high signal-to-noise VLT SINFONI + adaptive optics observations of rest-frame optical diagnostic emission lines, which exhibit an underlying broad component with FWHM ~ 700 km/s in both the Balmer and forbidden lines. Measured line ratios place the outflow robustly in the region of the ionization diagnostic diagrams associated with AGN. This unique opportunity - combining gravitational lensing, AO guiding, redshift, and AGN activity - allows for a magnified view of two main tracers of the physical conditions and structure of the interstellar medium in a star-forming galaxy hosting a weak AGN at cosmic noon. By analyzing the spatial extent and morphology of the Ly-alpha and dust-corrected H-alpha emission, disentangling the effects of star formation and AGN ionization on each tracer, and comparing the AGN induced mass outflow rate to the host star formation rate, we find that the AGN does not significantly impact the star formation within its host galaxy.

Tables

Table 1. Table 1 Central 0.5 " × "\times 0.5 " " " Binned AGN Flux Measurements

Line		Observed	Observed	Dereddened
	FWHM	Image Plane	Source Plane	Source Plane
	(km s-1)	(erg s-1 cm-2)	(erg s-1 cm-2)	(erg s-1 cm-2)
Narrow Component
H$\beta$	190	8.77 $\pm$ 2.68$\times$10-16	2.80 $\pm$ 1.04$\times$10-18	6.94 $\pm$ 2.58$\times$10-18
$[$O III$]\lambda$5007	190	1.70 $\pm$ 0.49$\times$10-15	5.56 $\pm$ 2.45$\times$10-18	1.34 $\pm$ 0.59$\times$10-17
$[$O I$]\lambda$6300	190	1.42 $\pm$ 0.73$\times$10-16	4.52 $\pm$ 2.87$\times$10-19	8.68 $\pm$ 5.51$\times$10-19
H$\alpha$	190	3.49 $\pm$ 0.48$\times$10-15	1.07 $\pm$ 0.15$\times$10-17	1.98 $\pm$ 0.29$\times$10-17
$[$N II$]\lambda$6584	190	1.16 $\pm$ 0.16$\times$10-15	3.46 $\pm$ 0.50$\times$10-18	6.39 $\pm$ 0.92$\times$10-18
$[$S II$]\lambda$6716	190	2.98 $\pm$ 1.06$\times$10-16	8.43 $\pm$ 3.49$\times$10-19	1.53 $\pm$ 0.63$\times$10-18
$[$S II$]\lambda$6731	190	2.48 $\pm$ 0.86$\times$10-16	7.65 $\pm$ 3.17$\times$10-19	1.39 $\pm$ 0.58$\times$10-18
Broad Component
H$\beta$	705	3.34 $\pm$ 1.02$\times$10${}^{-16a}$	8.89 $\pm$ 3.31$\times$10${}^{-19a}$	2.20 $\pm$ 0.82$\times$10${}^{-18a}$
$[$O III$]\lambda$5007	705	1.16 $\pm$ 0.34$\times$10-15	3.87 $\pm$ 1.70$\times$10-18	9.33 $\pm$ 4.11$\times$10-18
$[$O I$]\lambda$6300	725	4.89 $\pm$ 2.52$\times$10-16	1.18 $\pm$ 0.75$\times$10-18	2.56 $\pm$ 1.63$\times$10-18
H$\alpha$	725	1.33 $\pm$ 0.18$\times$10-15	3.40 $\pm$ 0.49$\times$10-18	6.30 $\pm$ 0.91$\times$10-18
$[$N II$]\lambda$6584	725	1.72 $\pm$ 0.24$\times$10-15	4.61 $\pm$ 0.66$\times$10-18	8.50 $\pm$ 1.22$\times$10-18
$[$S II$]\lambda$6716	725	2.76 $\pm$ 0.99$\times$10${}^{-16b}$	8.84 $\pm$ 3.66$\times$10${}^{-19b}$	1.61 $\pm$ 0.67$\times$10${}^{-18b}$
$[$S II$]\lambda$6731	725	6.40 $\pm$ 2.28$\times$10-16	1.72 $\pm$ 0.71$\times$10-18	3.13 $\pm$ 1.30$\times$10-18

Table 2. Table 2 lensing constraints

ID	$\Delta {\alpha }^a$	$\Delta {\delta }^a$	z${}_{}{}^b{}_{spec}{}^{}$	z${}_{}{}^c{}_{model}{}^{}$	rmsd
	h:m:s	d:m:s		(″)
1.1	0:33:41.167	+2:42:21.200	2.39	–	0.15
1.2	00:33:39.977	+02:42:10.4602	2.39	–	0.08
1.3	00:33:41.549	+02:42:16.802	2.39	–	0.20
1.4	00:33:41.586	+02:42:18.223	2.39	–	0.13
1a.1	00:33:41.167	+02:42:21.1589	2.39	–	0.14
1a.2	00:33:39.9765	+02:42:10.4568	2.39	–	0.08
1a.3	00:33:41.5704	+02:42:17.4662	2.39	–	0.09
1a.4	00:33:41.5835	+02:42:17.9503	2.39	–	0.09
2.1	00:33:41.1822	+02:42:21.0352	2.39	–	0.09
2.2	00:33:39.964	+02:42:10.3392	2.39	–	0.15
3.1	00:33:41.2903	+02:42:22.1782	2.096	–	0.35
3.2	00:33:41.4269	+02:42:21.7955	2.096	–	0.05
3.3	00:33:41.4035	+02:42:15.6917	2.096	–	0.46
3.4	00:33:40.048	+02:42:12.4045	2.096	–	0.05
4.1	00:33:41.1767	+02:42:19.7782	2.39	–	0.05
4.2	00:33:39.9723	+02:42:09.8256	2.39	–	0.09
5.1	00:33:41.126	+02:42:19.845	2.39	–	0.11
5.2	00:33:40.0265	+02:42:10.1024	2.39	–	0.30
6.1	00:33:40.634	+02:42:16.441	0.969	–	0.04
6.2	00:33:41.1598	+02:42:16.7427	0.969	–	0.10
7.1	00:33:38.019	+02:43:27.487	–	${5.70}_{-0.54}^{+0.69}$	0.12
7.2	00:33:38.574	+02:43:35.552	–	${5.70}_{-0.54}^{+0.69}$	0.36
7.3	00:33:40.08	+02:43:35.6793	–	${5.70}_{-0.54}^{+0.69}$	0.21
8.1	00:33:39.4425	+02:43:19.8843	–	${5.72}_{-0.09}^{+2.03}$	0.06
8.2	00:33:39.5671	+02:43:18.1456	–	${5.72}_{-0.09}^{+2.03}$	0.05
9.1	00:33:38.137	+02:43:22.043	–	${3.92}_{-0.34}^{+0.32}$	0.09
9.2	00:33:38.648	+02:43:32.52	–	${3.92}_{-0.34}^{+0.32}$	0.11
9.3	00:33:40.3541	+02:43:29.3617	–	${3.92}_{-0.34}^{+0.32}$	0.06

Equations

$$E(B-V)=-\frac{2.5log(\frac{F_o}{F_i})}{R_{\lambda }}=\frac{2.5log(\frac{(H\alpha /H\beta {)}_i}{(H\alpha /H\beta {)}_o})}{R_{H\alpha }-R_{H\beta }}$$

$$F_i=\frac{10^{-0.4R_{\lambda }E(B-V)}}{F_o}$$

$$M_{NLR}=\frac{m_{p}L(H\alpha )}{n_e(j_{H\alpha }/j_{H\beta })a_{H\beta }^{eff}h{\nu }_{H\beta }}$$

$$M_{NLR}=2.523\times 10^5\times L_{42}(H\alpha )M_{\text{astrosun}}$$

$$\frac{dE}{dt}=0.5\frac{M{v}^2}{t}$$

Full text

1. Introduction

Galaxies at the peak of cosmic star formation live in a state of punctuated equilibrium, where continuous accretion of gas from the cosmic web feeds large molecular gas reservoirs, and is balanced by star formation and outflows. Galactic wind feedback is widely acknowledged to play a critical role in the evolution of galaxies by expelling gas from the central regions of galaxies, shutting down their global star formation and regulating their stellar mass and size growth (Davé et al., 2012; Vogelsberger et al., 2013). However, the physical mechanisms involved and the relative importance of active galactic nuclei (AGN) and star formation as the main feedback drivers remain poorly understood. AGN-driven feedback is evident in luminous but rare QSOs and radio galaxies, but observational evidence is lacking for AGN feedback in less extreme, $“$normal$”$ star-forming galaxies (Fabian, 2012). In optical and infrared spectroscopy, evidence of AGN outflows, which can produce feedback, are observed as relatively broad emission-lines, with full width at half maximum (FWHM) $>$ 250 km s*-1*, inside the Narrow Line Region (NLR), a region of relatively low density ionized gas extending from the nuclear torus to distances of hundreds to thousands of parsecs from the nucleus.

Recent studies by Fischer et al. (2017, 2018) find outflows may not be powerful enough in nearby AGN to drive gas out to bulge-radius distances of 2 – 3 kpc. Kinematics within the NLR are largely due to rotation and in situ acceleration of material originating in the host disk. Spatially-resolved outflowing gas in Type 2 Seyferts and nearby (z $<$ 0.12) QSO2s extends to a fraction of radii typical of host galaxy stellar bulges (r $\sim$ 2 – 3 kpc). These findings suggest that outflows at z $\sim$ 0 may not be powerful enough to evacuate gas from their entire bulges. Several other studies have reached similar conclusions (Karouzos et al., 2016; Villar-Martín et al., 2016; Keel et al., 2017; Ramos Almeida et al., 2017).

Förster Schreiber et al. (2014) and Genzel et al. (2014) have reported evidence for likely AGN-driven outflows in the central regions of massive (log(M∗/M*☉)) $\geq$ 10.9) main-sequence star-forming galaxies (SFGs) at high redshifts (z $\sim$ 2) with FWHM $\sim$ 1000 – 1500 km s-1* and elevated [N II]/H$\alpha$ ratios $\geq$ 0.5. The outflows are resolved over the inner 2 – 3 kpc of the galaxies and detected in the forbidden [N II] and [S II] lines as well as in H$\alpha$. Therefore, these broad emission lines can not be due only to a virialized, parsec-scale AGN broad-line region. The mass outflow rates are estimated to be comparable to or exceed the star-formation rate (SFR) of the galaxy, thus creating an important avenue for the quenching of star formation. The next step is to measure the size, geometry, velocity profile and mass loading through high-resolution mapping of an outflow region. However, sensitivity and spatial resolution restrictions currently limit us to barely resolving ionized-gas structures in only a few of the largest and most massive SFGs at z $\sim$ 2.

Our team has recently discovered a bright, lensed galaxy, SGAS J003341.5+024217, henceforth SGAS 0033+02 (Figure 1), as described in the Magellan Evolution of Galaxies Spectroscopic and Ultraviolet Reference Atlas (MegaSaura; Rigby et al. 2018), that offers a unique opportunity to spatially resolve the influence of AGN feedback in a galaxy residing near Cosmic Noon at z $\sim$ 2.4.

SGAS 0033+02 was identified as a candidate lensed system through the Sloan Giant Arcs Survey (Gladders et al. in prep) in which objects with arc-like morphology are identified along lines of sight with photometric evidence for cluster- or group-scale masses, via a direct visual examination of Sloan Digital Sky Survey imaging data. Follow-up $gri$ imaging acquired with the MOSCA imager on the 2.5m Nordic Optical Telescope on UT 15 September 2012 confirmed the arc-like morphology of this system, and a spectroscopic redshift of z=2.378 was obtained with the same telescope using the ALFOSC spectrograph on the Nordic Optical Telescope on UT 01 September 2013.

Fortuitously, a bright (g$\sim$15.4) star appears in projection only 7″ from the main image of the lensed arc SGAS 0033+02. Recognizing this, we obtained laser guide star adaptive optics observations with SINFONI instrument on the VLT

VLT/SINFONI IFU observations of outflows in luminous 1.5¡z¡3 AGN have been resolved in detail in previous studies (Nesvadba et al., 2008, 2011; Perna et al., 2015; Cresci et al., 2015; Carniani et al., 2015; Brusa et al., 2016; Nesvadba et al., 2017; Förster Schreiber et al., 2018). However, through the combination of observations across several observatories, we are able to spatially resolve the size, geometry, and mass loading of AGN outflows on scales of 10s of parsecs for the first time at high redshift.

2. Observations and Data Reduction

2.1. HST WFC3 Imaging Observations

Imaging of SGAS 0033+02 was acquired using the HST Wide Field Camera 3 during two visits on 2016 October 30 and 2016 November 8. In the IR channel, images were taken in the F140W and F105W filters with cumulative exposure times of 459s and 1026s respectively. In the UVIS channel, exposures were taken in the F410M, F814W, and F555W filters with cumulative exposure times of 7256s, 1900s, and 1748s respectively. At the redshift of the source, z = 2.39, these filters provide a wide wavelength coverage, but isolate Ly$\alpha$ emission entirely within the F410M filter.

The HST imaging data were reduced using the software package, DrizzlePac111drizzlepac.stsci.edu. Images were aligned using tweakreg, then drizzled, using astrodrizzle, to a common reference grid with a scale of 0.03 arcseconds/pixel, with a Gaussian kernel and a drop size of 0.8. Three hot pixels in the IR channel near or within the main arc consistently failed to flag in astrodrizzle, resulting in artifacts in the final data products that could easily be mistaken for real substructure within the arc. These hot pixels were flagged manually in the data quality extension of the flat-field calibrated files before creating the final drizzled images, creating final data products free from these artifacts.

Continuum-subtracted Ly$\alpha$ imaging was produced using the F410M medium band filter with the F555W filter providing the continuum flux. Given the high equivalent width of Ly$\alpha$ in the MagE spectrum described below, EW${}_{obs}=$203 Å, and the F410M bandpass of 70 Å, we calculate that Ly$\alpha$ contributes 74$\%$ of the flux in F410M, with the remainder coming from continuum. We then scale the F555W image to match that continuum level, using annular aperture photometry of SGAS0033 in the F410M and F555W HST images, covering the same region as the MagE aperture.

2.2. MagE Magellan Observations

Observations of SGAS 0033+02 were obtained with the MagE instrument on the Magellan Baade telescope in UT 2015 Nov 07 and 10, for a total of 7 hrs of integration. The spectra cover observed wavelengths of 3200 - 8280Å, including Lyman alpha. Description of the observations and data reduction, and the MagE spectra themselves, were published by Rigby et al. (2018). Their Figure 1 shows that over the course of the observations, the 2”x10” MagE slit covered the full extent of the SGAS 0033+02 arc.

2.3. VLT SINFONI and MUSE IFU Observations

Observations of SGAS 0033+02 using VLT/SINFONI+AO were taken across several nights (2015 September 8, October 10, and December 4, 6, 9, and 12) in the H-, and K-bands, with resolving powers of R = $\lambda$/$\delta\lambda$ = 3000 and 4000 and covering spectral regions between 1.45 - 1.85 and 1.95 - 2.45 $\mu$m respectively, with a pixel scale 0.05$"\times$ 0.1$"$ and sampling a field of view of 3.2$"\times$ 3.2$"$. Observations were carried out in observing blocks (OBs) of an OSOOSO pattern, alternating object (O) and sky (S) positions. Each OB was dithered by 0.15*′′* around the central position to mitigate bad pixels and cosmic rays. 8 individual exposures of 600s were obtained in the H-band and 28 individual exposures of 600s were obtained in the K-band, for totals of 1 hr 20 min and 4 hr 40 min of on-source integration, respectively. VLT/SINFONI data were reduced using the software package SPRED developed specifically for SPIFFI (Schreiber et al., 2004; Abuter et al., 2006) following the procedures described in Förster Schreiber et al. (2009). The offsets between individual cubes were determined from the known dither pattern within each OB, and the location of the acquisition star observed before each OB. The final PSF is created by fitting a circularly symmetric 2D Gaussian profile to acquisition star exposures taken prior to each OB of the science target, and results in a FWHM of 0.19$”$ in K and 0.18$”$ in H-band. The PSF FWHMs correspond to the effective resolution of all observations for our target. Early B-type standard stars were observed each night to provide flux calibration and telluric correction.

Observations of SGAS 0033+02 using VLT/MUSE were obtained under the program 098.A-0459(A). The 1 arcmin field-of-view is sampled with 349 $\times$ 352 0.2*′′* wide spaxels. Our setup provided a wavelength range from 4650 to 9300 Å at a resolving power R ranging from 2000 to 4000. Each spectral bin is 1.25 Å  wide. The observations were carried out in ’service mode’ during dark time, with clear-sky conditions, airmass below 1.8, and seeing better than 0.7*′′* on the nights of 2017 September 19, 20. We obtained a total of 12 exposures of 700s on-target each. The exposures were taken within “Observing Blocks” of 4 exposures each. We applied a small dithering and 90 deg rotations between exposures to reject cosmic rays and minimize patterns of the slicers on the processed combined cube. We reduced all the observations using the MUSE pipeline recipe v1.6.4 and ESO reflex v2.8.5. The individual exposures were combined into one final science datacube. The total on-target time was therefore 2.3 hours. The sky subtraction was improved on this cube using the Zurich Atmospheric Purge (ZAP) algorithm v1.0.

3. Image Plane Analysis

3.1. SINFONI Spectroscopic Fitting

Emission-line kinematics and fluxes of H$\alpha$, [N II], [O III], and H$\beta$ were measured in each spaxel of our SINFONI H- and K-band data cubes by fitting Gaussians in an automated routine. Our fitting process, previously discussed in depth in Fischer et al. (2017), uses the Importance Nested Sampling algorithm as implemented in the MultiNest library (Feroz & Hobson, 2008; Feroz et al., 2009, 2013; Buchner et al., 2014) to compute the logarithm of the evidence, $lnZ$, for models containing a continuum plus zero to three Gaussian components per emission line. Gaussians were defined using Gaussian parameters $\mu$ (centroid), $\sigma$ (dispersion), and H (peak height). When comparing two models, i.e. a model with zero Gaussians ($M_{0}$) and a model with one Gaussian ($M_{1}$), the simpler model is chosen unless the more complex model, $M_{1}$, has a significantly better evidence value, $|ln(Z_{1}/Z_{0})|>5$ (99% more likely). Fits of emission lines in individual spaxels used different models for each band. H-band models first measured [O III] $\lambda$5007, simultaneously fitting a second set of components to [O III] $\lambda$4959 in order to properly account for flux contributions from wing emission between both lines, and then tested for the presence of H$\beta$. Gaussian wavelength centroid and disperson parameters of [O III] $\lambda$4959 components were fixed following parameters used in fitting [O III] $\lambda$5007 components, with the flux of [O III] $\lambda$4959 fixed to be 1/3 that of the [O III] $\lambda$5007 flux. Gaussian wavelength centroid and disperson parameters of H$\beta$ components were fixed in the same manner, as we assume that the lines originate from the same emission region, and the H$\beta$ flux was left as an open parameter. K-band models first measured H$\alpha$ and then tested for the presence of [N II] $\lambda\lambda$6548,6584. Gaussian wavelength centroid and disperson parameters of [N II] $\lambda\lambda$6548,6584 were also fixed following parameters used in fitting H$\alpha$, again under the assumption that the lines originate from the same emission region, with the flux of [N II] $\lambda$6548 fixed to be 1/3 that of the [N II] $\lambda$6584 flux, which was left as an open parameter.

Initial input parameters in our models are selected based on physical considerations. The centroid position for each Gaussian was limited to a 40Å range around the wavelength that contained the entirety of the line profiles throughout each data cube. Gaussian standard deviation ranged from the spectral resolution of the H- and K-band gratings, to an artificial FWHM limit of $\sim$800 km s*-1*. Gaussian height was defined to allow for an integrated flux that ranged from a 3$\sigma$ detection to a maximum integrated flux of 3$\sigma\times$ 104.

Fits from the H- and K-band observations are mapped in Figure 3. Observed velocity, FWHM, and integrated fluxes are shown for H$\alpha$ and [O III]$\lambda$5007, with additional integrated fluxes for H$\beta$ and [N II], as their velocity and FWHM measurements are identical to [O III] and H$\alpha$, respectively. Doppler-shifted velocities are given in the rest frame of the galaxy using air rest wavelengths of each line. We found emission lines present in most spaxels to be best fit with a single Gaussian, with H$\alpha$ and [N II] emission lines containing two-component line profiles in spaxels surrounding the K-band continuum peak (shown as a cross in each map of Figure 3) of the lensed galaxy arc. Two-component fits are sorted by FWHM into separate H$\alpha$ / [N II] maps in Figure 3. Component blending due to lower signal-to-noise ratios for the broad component is observed in regions between fits with different numbers of components, as a jump in line dispersion is observed in the narrow-component FWHM plot at the border between single and double component fits.

We find a majority of the emission-line gas fit with single-components, or the narrower of two components, is near systemic velocity or slightly redshifted. Emission-line knots north and south of the continuum peak show symmetric redshifted kinematics. Additional faint filaments observed in H$\alpha$ and [O III] east and west of the continuum peak also show symmetric redshifted velocities. The broad H$\alpha$ and [N II] emission-line components over the continuum peak are typically blueshifted, with an average FWHM of $\sim$540 km/s and maximum and average offsets of $\sim$-140 km s*-1* and $\sim$-40 km s*-1*, respectively. We measure the spatially-resolved maximum extent of the broad-FWHM, blueshifted gas in the image plane by fitting the region with an ellipse of rmaj= 0.35*′′, b/a = 0.4, and PA = 30$\degree$ east of north. At 7” from the guide star, we note a degradation of the reported K-band PSF is expected, with the Strehl Ratio of observations for SGAS 0033+2 decreasing by approximately 20$\%$ per the SINFONI User Manual. Temporal variations of the atmosphere also add uncertainty on the effective PSF during the observations, with individual exposures of PSF stars in similar observations by Förster Schreiber et al. (2018) indicating typical OB-to-OB variations of $\sim$30$\%$ in PSF FWHM. As such, assuming an effective PSF during the observations to be $\sim$ 0.3′′*, the spatial extent of the observed outflows remains well resolved.

3.2. Ionization Source Diagnostics

We compare measured line flux ratios in an ionization diagnostic diagram (i.e. BPT diagram; Baldwin et al. 1981) to spatially resolve the source of ionization throughout the image plane arc and determine whether the observed blueshifted outflows can be attributed to an AGN. Note that measured ratios are not affected by magnification as lensing effects are achromatic. To account for the high redshift of our target, we used a redshift-dependent classification that utilizes the standard optical diagnostic line ratios [O III]/H$\beta$ vs [N II]/H$\alpha$ as detailed in Kewley et al. (2013). Our initial diagnostic diagram, provided in the left plot in Figure 4, compares line ratios using the integrated flux across all components of each line. Grey points in this distribution have single component fits for each emission line, while red points use summed fluxes of H$\alpha$ and [N II]$\lambda$6584 emission lines across both a narrow- and broad-component. Decomposing these two component line emission lines into individual narrow- and broad- components to obtain their individual ratios, as shown in the right plot of Figure 4, we find that the narrow components align with the grey points of the left figure, and that the broad components exhibit an [N II]/H$\alpha$ ratio that suggests AGN ionization. Note that the position of the broad components on the diagram uses the same [O III]/H$\beta$ ratio as their corresponding narrow lines because broad [O III]/H$\beta$ components are not observed in individual spaxels.

In order to detect broad-component signatures of [O III] and H$\beta$, we binned spectra over a $0.5"\times 0.5"$ square surrounding the continuum peak and a majority of the blueshifted outflows (binned region is shown in the broad-component H$\alpha$ and [N II] maps of Figure 3). Fits to the resultant H- and K-band spectra are shown in Figure 5, where we are able to detect a broad emission line for [O III], as well as emission from [O I]$\lambda$6300 and [S II]$\lambda\lambda$ 6716,6731, but remain unable to detect broad H$\beta$. Fit parameters for each emission-line in the image-plane binned spectra are provided in Table 1. To determine a lower limit on the summed broad-component [O III]/H$\beta$ ratio, we estimate the flux of the unobserved H$\beta$ broad-component to scale to its narrow-component in a similar fashion to the observed broad- and narrow-components of H$\alpha$ in the same binned region. The estimated flux of 3.34$\times$10*-16* erg s*-1* cm*-2* is consistent with our measurements, such that the broad H$\beta$ would be likely be undetectable compared to the low signal of the brighter [O III]$\lambda$ 4959 broad-component.

Flux ratios derived from our binned spectrum are plotted as larger, filled points in Figure 4, where the red circle, and green and blue half-circles represent flux ratios using both components, the narrow component, and the broad component, respectively. We find a lower limit on the broad-component [O III]/H$\beta$ ratio to be 0.54, which is elevated relative to the narrow- and summed-component ratios. This suggests that a majority of the broad-components would likely have larger [O III]/H$\beta$ ratios using their true line fluxes instead of estimates and remain in the AGN ionized portion of the diagnostic diagram. In tandem, the measured emission-line flux ratios and kinematics suggest that we are observing outflows from an AGN in the arc image plane of SGAS 0033+02.

3.3. Spatially Resolved Ly$\alpha$ Structure

We compare the image plane morphology of the H$\alpha$ gas from our SINFONI spectral fits to that of the Ly$\alpha$-emitting gas from HST imaging, as shown in Figure 6, to determine if the AGN outflows have some influence on the propagation or escape of Ly$\alpha$ photons. We find that the Ly$\alpha$-emitting gas is most prominent between, rather than cospatial with, the brightest knots of H$\alpha$ that reside over the AGN and likely star-forming regions. This discrepancy between the morphology of Ly$\alpha$ and H$\alpha$ has also been reported in similar studies of local starburst galaxies (Östlin et al., 2009; Hayes et al., 2013) and high redshift (z $<$ 2.5) quasar hosts Bayliss et al. (2017).

We also compare the spectral signatures of Ly$\alpha$ and H$\alpha$ in Figure 6, with Ly$\alpha$ emission obtained from long-slit MagE observations covering the full spatial extent of the arc as detailed in Rigby et al. 2018. Comparable H$\alpha$ emission was obtained by binning SINFONI spaxels that overlap with locations of the strongest Ly$\alpha$ flux knots in the HST imaging (boxes in the flux map of Figure 6). The observed velocity structure of Ly$\alpha$ in comparison to Balmer emission is typical in studies of green pea galaxies (Yang et al., 2017; Orlitova et al., 2018). Although the sampled spectra are immediately adjacent to the detected AGN outflows, fitting Gaussians to the binned H$\alpha$ spectrum, we measure a FWHM of $\sim$200  km s*-1*, which suggests relatively undisturbed kinematics, and do not detect a secondary, outflow component. These observations suggest that the AGN outflows in SGAS 0033+02 are anti-correlated with the observed Ly$\alpha$ structure.

To estimate the intrinsic properties of the AGN in the source plane of SGAS 0033+02, we must apply a gravitational lens model to our observed image plane data. Details on the methods used to convert image plane observations of SGAS 0033+02 into source plane data are detailed in the Appendix. From our model, we find that the main arc of SGAS 0033+02 straddles a lensing critical line, which separate regions of different image multiplicities. As such, the observed structure in this arc is approximately half of the galaxy observed in the counter images.

4. Source Plane Analysis

4.1. Extent of AGN Outflows

Figure 7 shows the source plane reconstruction of the fraction of SGAS 0033+02 observed in the main arc as it would have been seen without the presence of the lens. Orange and green contours represent the source plane extents of the narrow and broad H$\alpha$ emission-line components from SINFONI observations shown in Figure 3, respectively. Measuring the radial extent of the broad-component (i.e. outflows) in the source plane, we report a length of r $\sim$100 pc. This is likely the maximum outflow extent in the observed half of the galaxy, as the location of the outflows is adjacent to the rest-frame optical continuum peak of the galaxy in Figure 7, which suggests that they reside near the galaxy nucleus and AGN. However, this measurement should be treated as a lower limit of the true outflow extent, as we have no kinematic data on the other half of the galaxy which is not observed in the arc. We can measure the distance between the furthest knot of emission in the other half of the galaxy, as seen in the F555W image of Counter Image 1 which traces the extent of the observed H$\alpha$ emission, and its F140W continuum peak to set an upper limit on the maximum possible outflow distance as r $\sim$830 pc.

4.2. Intrinsic Flux Measurement

By reconstructing the source plane image of SGAS 0033+02, we can also determine the magnification any given point in the image plane. Demagnified fluxes for the AGN-ionized outflows in SGAS 0033+02 were obtained by dividing the image plane datacubes by a magnification map at matched pixel scale resolution, as determined from our strong lens model. Spectra in the central $0.5"\times 0.5"$ square were again binned and fit to measure the total demagnified flux. In this second iteration of fitting, line dispersons and centroids were fixed to the values obtained from the fit to the image-plane spectrum, with only the total flux (i.e. the Gaussian amplitude) allowed to vary. Source plane fluxes are provided in Table 1.

Before analyzing our measured fluxes, we applied a reddening correction using a standard Galactic reddening curve (Savage & Mathis, 1979) and color excesses calculated from the observed, source-plane H$\alpha$/H$\beta$ ratio (Osterbrock & Ferland, 2006), assuming an intrinsic recombination value of 2.85. The extinction was calculated using

[TABLE]

where E(B-V) is the color excess, Rλ is the reddening value at a particular wavelength, and Fo and Fi are the observed and intrinsic fluxes, respectively. Galactic reddening values are R${}_{H\alpha}\approx$ 2.5 and R${}_{H\beta}\approx$ 3.7. Corrected line fluxes are then given by

[TABLE]

with dereddened source-plane fluxes listed in Table 1.

4.3. AGN Mass Outflow Rate

We use the dereddened source-plane flux of the broad H$\alpha$ component to estimate the total, observed ionized gas mass in the NLR outflows, for case B recombination (Peterson, 1997; Osterbrock & Ferland, 2006). The total luminosity of H$\beta$, originating from clouds within a total volume Vc, is $L(H\beta)=n_{e}n_{p}a^{eff}_{H\beta}h\nu_{H\beta}V_{c}$, with $a^{eff}_{H\beta}$ and $\nu_{H\beta}$ being the effective recombination coefficient and rest frequency of H$\beta$, and $n_{e}$ and $n_{p}$ being the number densities of electrons and protons. We assume completely ionized hydrogen clouds, therefore $n_{e}\sim n_{p}$. H$\alpha$ and H$\beta$ luminosities are related such that $L(H\alpha)=(j_{H\alpha}/j_{H\beta})L(H\beta)$, where $j_{H\alpha}/j_{H\beta}$ is the intensity of H$\alpha$ relative to H$\beta$. Assuming the same density for all clouds, $n_{p}m_{p}$, with $m_{p}$ being the proton mass, the total ionized gas mass is $M_{NLR}=n_{p}m_{p}\times V_{c}$. From the relations made above:

[TABLE]

[TABLE]

where $L_{42}(H\alpha)$ is in units of 1042 erg s*-1*. Intrinsic $a^{eff}_{H\beta}$ and $j_{H\alpha}/j_{H\beta}$ were taken from Osterbrock & Ferland (2006), assuming a temperture of $T=10^{4}$ K. We derive an electron density (ne cm*-3*) using an estimated [S II] $\lambda\lambda$6716/6731 line ratio for the AGN ionized broad-emission-line component (Allen, 1979; Osterbrock & Ferland, 2006). We measure a dereddened [S II] $\lambda$6731 broad-component flux of 3.13$\times$10*-18* erg s*-1* cm*-2*, do not detect a comparable broad [S II] $\lambda$6716 component, and instead use a flux of 1.61$\times$10*-18* erg s*-1* cm*-2* as a flux upper limit as this represents a dereddened 3$\sigma$ flux detection at this wavelength, assuming a similar line disperson. These fluxes produce a maximum ratio $\sim$ 0.5, from which we assume $n_{e}\sim 10^{4}$ cm*-3*. Using a luminosity distance of $D_{L}$ = 6.071$\times$1028 cm (Wright, 2006), we measure the dereddened L(H$\alpha$) of the outflowing wind to be 2.92$\times 10^{41}$ erg s*-1* in the source plane, and calculate a gas mass of $M_{NLR}=7.37\times 10^{4}M_{\astrosun}$. As this measurement is derived from a 0.5$"\times$0.5$"$ bin containing spectra from both sides of the critical line, the reported value assumes similar fluxes on the side of the lensed system hidden by the lensing critical line. Using this gas mass, we then calculate the mass outflow rate in this region by dividing the total mass $M$ by the time $t$ it takes to travel across the extent which we observe the outflows, where $t=R/v$. We assume a maximum outflow extent of $\sim$ 100 pc, as derived from the strong lens model. Observed radial velocities of the outflows are on the scales of tens of km s*-1*, however, these are likely compromised by projection effects. We instead use the maximum blueshifted velocity defined as 1/2 $\times$ the full width at zero maximum (FWZM), approximately the 3$\sigma$ velocity offset from the centroid of the broad H$\alpha$ component measured in our binned spectrum as our deprojected velocity, $v=924$ km s*-1*. Using these parameters, we measure a mass outflow rate of $\dot{M}=$ 0.67 M*☉* yr*-1*. The power of the outflow $dE/dt$ is then calculated as:

[TABLE]

for a log($\dot{E}$) = 41.33 erg s*-1*. We use the dereddened source-plane flux of the [O III] $\lambda$5007 and [O I] $\lambda$6300 broad components to measure the bolometric luminosity of the AGN, using the method from Netzer 2009, log (Lbol) = 3.8 + 0.25 logL([O III]$\lambda$5007) + 0.75 logL([O I]$\lambda$6300), for a log(Lbol) = 45.02 erg s*-1*. The resulting ratio of outflow power to bolometric luminosity is log($\dot{E}/L_{bol}$) = -3.76, less than the 0.5% threshold typically required to provide a significant impact on the host galaxy (Hopkins & Elvis, 2010).

4.4. Star Formation Rate

We convert the narrow-line, non-AGN ionized H$\alpha$ luminosity not attributed to AGN ionization (i.e. the H$\alpha$ flux measured in the left column of Figure 3) to a star formation rate (SFR), by using the relation from Kennicutt (1998), where SFR $(M_{\astrosun}~{}yr^{-1})=7.9\times 10^{-42}L(H\alpha)$, and adjusting to the initial mass function (IMF) from Chabrier (2003), which reduces the SFR by a factor of 1.7. We measure source-plane H$\alpha$ luminosities north and south of the lensing caustic to be L(H$\alpha$)north = 4.9$\times$1042 erg s*-1* and L(H$\alpha$)south = 1.5$\times$1043 erg s*-1*, which convert to SFRs of 22.8 M*☉* yr*-1* and 70.7 M*☉* yr*-1*, respectively. We note that these rates may be upper limits, as there may be contributions to the HII regions by AGN ionization. We can compare our SFR measurements to those in Livermore et al. (2015), which show a correlation between SFR in star-forming clumps and their sizes, by isolating a lower-limit SFR in the discrete, fully imaged H$\alpha$ knot north of the continuum peak, as shown in Figure 3 and 6. Here, we measure a demagnified F(H$\alpha$) = 1.132$\times$10*-17* erg s*-1* cm*-2*, which converts to a SFR of 2.43 M*☉* yr*-1*, over an area of 0.7 milliarcsecond2 in the source plane (Figure 7) for an approximate radius of 15 milliarcseconds, or 125 pc. Measurements for the global star formation rate and the clump star formation rate both exceed the mass outflow rate of the AGN. Therefore, the central AGN, in its current state, is incapable of displacing enough material to quench star formation in this galaxy.

5. Discussion

Producing this spatially resolved analysis of AGN outflows in a ”normal” star-forming galaxy at z $\sim$ 2, we find it to be similar to weak AGN with strong star formation in the nearby universe. The measured source-plane bolometric luminosity of this object suggests that we are observing a Seyfert-like AGN in SGAS 0033+02. In addition, the observed recombination emission-line dispersons indicate that SGAS 0033+02 is a Type 2 AGN, where the central engine is obscured along our line of sight. With the observed morphology of the bright, outflowing NLR being relatively compact, we find this target to be analogous to the nearby Seyfert 2 NGC 1068 (Crenshaw & Kraemer, 2000; Das et al., 2006). From HST WFPC2 [O III] imaging (Schmitt et al., 2003), the enclosed [O III] flux within a 100 pc radius of the nucleus for this nearby AGN is 9.11 $\times 10^{-12}$ erg s*-1* cm*-2*, which converts to a luminosity of L([O III]) = 1.7$\times$1041 erg s*-1*. The measured flux in this system originates from one half of the NLR, with the other half extinguished below the plane of its host disk. For comparison, we can divide the measured L([O III]) of SGAS 0033+02 of 4.32$\times$1041 erg s*-1* in half and find its luminosity to be on par with NGC 1068 at L([O III]) = 2.16$\times$1041 erg s*-1*. Notably, as a Seyfert AGN with an intrinsic observed F([O III]) = 3.87$\times$10*-18* erg s*-1* cm*-2*, it is unlikely that the broad emission-line component attributed to AGN ionization in SGAS 0033+02 would be detected in a typical field galaxy at z = 2.391. Combined with the effects of star-formation dilution hiding narrow AGN NLR signatures near systemic velocity, these findings suggest many AGN may go undetected in surveys of galaxies residing near Cosmic Noon (Trump et al., 2015).

There is no evidence in our current observations that we are missing broad outflowing emission-line components at greater radii due to lesser amounts of magnification. As shown in Figure 6, binning H$\alpha$ lines exterior to where we detect AGN outflows result in single Gaussian fit without the presence of a second, broad component. However, it remains unclear if we are observing the true extent of the AGN outflows because, as described above, the main arc of SGAS 0033+02 is a partial image that contains roughly half of the galaxy seen in the counter images. Assuming that the outflows originate from the optical continuum peak, we cannot know the extent of the winds in the other half of the system without kinematics measurements for one of the counter images.

Comparing the extent of the AGN-ionized region to the AGN [O III] luminosity of SGAS 0033+02, we find that it has a relatively small extent for its luminosity when compared to the radius vs luminosity correlation for NLRs found in previous studies (Schmitt et al., 2003; Liu et al., 2010; Fischer et al., 2018; Dempsey & Zakamska, 2018). Although we are likely observing the maximum extent of the AGN outflows in our observations, a narrow AGN-ionized emission-line component displaying rotation kinematics that extends to larger distances would be undetected due to dilution by the larger flux contribution of the HII star forming region. Measuring the source plane radial extent of the [O III] emission shown in Figure 3, we find a maximum R${}_{[O~{}III]}$ $\sim$ 800 pc. Assuming an AGN-ionized component exists throughout, a radial extent of 800 pc paired with a log(L[O III]) = 41.6 erg s*-1* places SGAS 0033+02 in line with previous findings from Seyferts and QSOs in the nearby universe.

6. Conclusions

We have analyzed spatially-resolved, rest-frame UV / optical imaging and spectroscopy of a Seyfert AGN at z $\sim$ 2 for the first time. Our major findings are:

1. AGN-ionized outflows extend to a radius of r $\sim$ 100 pc. We calculate a mass outflow rate over this distance of $\dot{M}=0.55\,M_{\astrosun}~{}yr^{-1}$. The corresponding ratio of outflow power to bolometric luminosity is exceedingly low, log($\dot{E}/L_{bol}$) = -3.76, suggesting the AGN does not significantly impact the host galaxy.

2. SGAS 0033+02 also exhibits a star formation rate on the order of tens of solar masses per year, which greatly exceeds the AGN mass outflow rate. As such, the current state of the AGN in SGAS 0033+02 would be unlikely to quench star-formation within the galaxy.

3. The positions of outflowing winds and Ly$\alpha$ emission are anti-correlated. Ly$\alpha$ exists where the outflow is not, therefore the outflow has not destroyed Ly$\alpha$ over the whole arc. Ly$\alpha$ structure in this galaxy is also similar to those in galaxies not hosting AGN.

4. SGAS 0033+02 resembles weak AGN with strong star formation observed in the local universe. Faint emission-line signatures of these low-luminosity AGN make their detection at z$\sim$2 extremely difficult without gravitational lensing. Combining faint AGN emission with line-dilution from strong star formation, it is possible that many AGN are missed in survey work at this redshift.

Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere under ESO programmes 094.A-0746(A) and 098.A-0459(A). Based on observations made with the NASA/ESA Hubble Space Telescope, obtained from the data archive at the Space Telescope Science Institute. STScI is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. This paper includes data gathered with the 6.5 meter Magellan Telescopes located at Las Campanas Observatory, Chile. The authors would like to thank the anonymous referee for their helpful comments. This paper benefited from error analyses by M. Revalski and discussions with D. M. Crenshaw and C. L. Gnilka. TCF was supported by an appointment to the NASA Postdoctoral Program at the NASA Goddard Space Flight Center, administered by Universities Space Research Association under contract with NASA and by NASA through grant number HST-AR-15019.006-A from the Space Telescope Science Institute, which is operated by AURA, Inc., under NASA contract NAS 5-26555. SL was partially funded by UCh/VID project ENL18/18. LFB was supported by Anillo ACT-1417.

Appendix A Gravitational Lens Modeling

A.1. Lensing Mass Models Methodology

Here, we provide a brief summary of the gravitational lensing analysis used in this work and we refer the reader to Kneib et al. (1996), Smith et al. (2005), Verdugo et al. (2011) and Richard et al. (2011) for a more in depth description. We take a parametric approach, using Lenstool (Jullo et al., 2007) to model the cluster mass distribution surrounding our target as a series of dual pseudo-isothermal ellipsoids (dPIEs, Elíasdóttir et al. 2007), which are optimized through a Monte Carlo Markov Chain minimization.

To model the cluster mass distribution, Dark Matter (hereafter DM) dPIE clumps are combined to map the DM at the cluster scale. Galaxy scale DM potentials are used to describe galaxy scale substructure. Considering the large number of galaxies in the cluster, it is not feasible to optimize the parameters of every potential, as the large parameter space will lead to an unconstrained minimization. Moreover, individual galaxies contribute only a small fraction to the total mass budget of the cluster, so their effects on lensing are minimal unless they are in close proximity, in projection, to the lensed galaxies. To reduce the overall parameter space we scale the parameters of each galaxy to a reference value, using a constant mass-luminosity scaling relation (see Limousin et al. 2007).

A.2. Selection of Cluster Members

We used Sextractor in the ”white” image of the MUSE data to detect all the sources and define apertures for PyMuse (https://pypi.org/project/PyMUSE/) to integrate the spaxels and thus to obtain the spectra for each of the galaxies. PyMuse can also run Redmonster (Hutchinson et al. 2016) to determine individual redshifts. All the spectra, and Redmonster best candidates, were visually inspected to assign the redshift for each galaxy.

We then constructed a galaxy cluster catalog using the red sequence technique (Gladders & Yee, 2000), by selecting in a color-magnitude diagram galaxies that show a similar color. Our final catalog contains 80 cluster members.

As the brightest galaxies, or bright cluster galaxies (BCGs), of galaxy clusters tend to not follow the cluster red sequence, we remove the BCG of the South-East sub cluster (Newman et al., 2013a, b). We keep the other BCG in the scaling relation due to the lack of constraints to properly model the lensing potential shape on that side. In addition, we detected several spirals galaxies in the MUSE data cube at z$\sim 1.03$ (Barrientos et al. in prep.) that may influence the lensing configuration of the bright arc of SGAS 0033+02. We include the two closest ones ($\alpha=00^{\rm h}33^{\rm m}41.6576^{\rm s}$, $\delta=+02^{\rm o}42\arcmin 13.7186\arcsec$ and $\alpha=00^{\rm h}33^{\rm m}41.0841^{\rm s}$, $\delta=+02^{\rm o}42\arcmin 05.5126\arcsec$) in our lensing potentials, but model them separately as individual potentials at the cluster redshift.

A.3. Lensing Constraints

We consider a large number of constraints for the bright arc in order to obtain the most accurate source reconstructions. Figure 8 exhibits an HST F555W/F814W/F140W image marking the positions of all constraints used in our model and the resultant critical line. We also provide an enlarged, labeled image of the region near SGAS 0033+02 in Figure 9, with the positions and redshifts of these systems listed in Table 2. From our model, we find the lensing critical line at z = 2.39 lies directly over the center of the arc of SGAS 0033+02, such that the north and south ends of the arc are reflections of one another. This is supported by the symmetries on each side of the arc observed both in imaging and kinematics.

We find that the arc contains an unusual asymmetry that cannot be accounted for by the lensing model, observed in the rest-frame UV continuum F555W image, as shown in Figure 10. As the critical line from the strong lensing model crosses at the flux peak in the F140W image, we observe that the small and faint emission knot just north to the critical curve does not show a symmetric counterpart on the other side of the arc. As such, this emission could be due to a transient in the arc and we do not include this feature in our constraints. In addition, the southern emission knot in the F555W image is significantly brighter than the corresponding knot in the top arc. This knot coincides with the H-alpha knot visible in SINFONI data (see Figure 3), which also show this flux asymmetry. A possible explanation to this discrepancy is that the observed flux of this feature is time variable, however additional observations are required to test such a scenario.