Kiloparsec-scale outflows are prevalent among luminous AGN: outflows and   feedback in the context of the overall AGN population

C. M. Harrison (Durham University); D. M. Alexander; J. R. Mullaney,; A. M. Swinbank

arXiv:1403.3086·astro-ph.GA·June 19, 2015

Kiloparsec-scale outflows are prevalent among luminous AGN: outflows and feedback in the context of the overall AGN population

C. M. Harrison (Durham University), D. M. Alexander, J. R. Mullaney,, A. M. Swinbank

PDF

TL;DR

This study shows that kiloparsec-scale ionized outflows are common in luminous AGN at low redshift, with implications for galaxy evolution and feedback mechanisms, driven by either star formation or AGN activity.

Contribution

It provides the first comprehensive analysis of the prevalence, extent, and energetics of ionized outflows in a representative sample of luminous, low-redshift type 2 AGN.

Findings

01

Ionized outflows are present in >=70% of the sample.

02

Outflows extend over 6-16 kpc in all targets.

03

Outflow velocities reach up to 1700 km/s.

Abstract

We present integral field unit (IFU) observations covering the [O III]4959,5007 and H-Beta emission lines of sixteen z<0.2 type 2 active galactic nuclei (AGN). Our targets are selected from a well-constrained parent sample of 24,000 AGN so that we can place our observations into the context of the overall AGN population. Our targets are radio-quiet with star formation rates (<~[10-100] Msol/yr) that are consistent with normal star-forming galaxies. We decouple the kinematics of galaxy dynamics and mergers from outflows. We find high-velocity ionised gas (velocity widths of 600-1500 km/s and maximum velocities of <=1700 km/s) with observed spatial extents of >~(6-16) kpc in all targets and observe signatures of spherical outflows and bi-polar superbubbles. We show that our targets are representative of z<0.2, luminous (i.e., L([O III]) > 5x10^41 erg/s) type 2 AGN and that ionised…

Tables4

Table 1. Table 1: Notes: Details of the targets that we observed with the Gemini-GMOS IFU. (1) Object name; (2)–(3) optical RA and DEC positions from SDSS (DR7); (4) the Gemini-GMOS run when we observed this object (see Section LABEL:Sec:QSOs:Observations ); (5) systemic redshifts derived from this work (see Section LABEL:Sec:QSOs:FittingProcedure ); (6) FWHM (in km s -1 ) of the broadest, luminous (contributes at least 30% of the total flux) [O iii ] λ 𝜆 \lambda 5007 emission-line component from Mullaney et al. ( 2013 ) .

Name	RA (J2000)	Dec (J2000)	Run	$z_{sys}$	FWHM[B]
(1)	(2)	(3)	(4)	(5)	(6)
J0945+1737	09:45:21.33	+17:37:53.2	10	0.1283	1027
J0958+1439	09:58:16.88	+14:39:23.7	10	0.1092	878
J1000+1242	10:00:13.14	+12:42:26.2	10	0.1480	815
J1010+1413	10:10:22.95	+14:13:00.9	10	0.1992	1711
J1010+0612	10:10:43.36	+06:12:01.4	10	0.0984	1743
J1100+0846	11:00:12.38	+08:46:16.3	10	0.1005	1203
J1125+1239	11:25:46.35	+12:39:22.6	12	0.1669	1028
J1130+1301	11:30:28.86	+13:01:19.6	12	0.1353	857
J1216+1417	12:16:22.73	+14:17:53.0	12	0.0818	1404
J1316+1753	13:16:42.90	+17:53:32.5	10	0.1505	1357
J1338+1503	13:38:06.53	+15:03:56.1	12	0.1857	901
J1339+1425	13:39:56.48	+14:25:29.6	12	0.1390	830
J1355+1300	13:55:45.66	+13:00:51.0	12	0.1519	1058
J1356+1026	13:56:46.10	+10:26:09.0	10	0.1238	783
J1430+1339	14:30:29.88	+13:39:12.0	10	0.0855	901
J1504+0151	15:04:20.90	+01:51:59.4	12	0.1827	978

Table 2. Table 2: Notes: Details of our IFU targets that we observed with Gemini-GMOS. (1) Object name; (2)-(3) Total [O iii ] λ 𝜆 \lambda 5007 and H β 𝛽 \beta luminosities derived in this work, uncertainties are typically 15%; (4) 1.4 GHz flux densities obtained from the FIRST survey ( Becker et al. 1995 ) and uncertainties that are defined as 3 × \times the RMS noise of the radio map at the source position; (5) Rest-frame radio luminosities using a spectral index of α = 0.7 𝛼 0.7 \alpha=0.7 and assuming S ν ∝ ν − α proportional-to subscript 𝑆 𝜈 superscript 𝜈 𝛼 S_{\nu}\propto\nu^{-\alpha} (we note that a range of α = 0.2 𝛼 0.2 \alpha=0.2 –1.5 introduces an spread of ≲ ± less-than-or-similar-to absent plus-or-minus \lesssim\pm 0.1 dex on the radio luminosity); (6) Radio morphology paramater, where sources with θ > 1.06 𝜃 1.06 \theta>1.06 are classified as extended in the 1.4 GHz FIRST data (see Section 2.4 ); (7) The deconvolved PA and major-axis radius from the FIRST survey for the radio-extended sources; (8)-(10) Infrared luminosties from star-formation, SFRs and bolometric AGN luminosities derived from our SED analyses (see Section 2.4 ); (11) The q IR subscript 𝑞 IR q_{\rm{IR}} (“radio excess”) parameter for each source, where sources with q IR ≤ 1.8 subscript 𝑞 IR 1.8 q_{\rm{IR}}\leq 1.8 are classified as “radio excess” (see Section 2.4 ). ⋆ The 1.4 GHz flux density for SDSS 1355+1300 is taken from the NVSS survey ( Condon et al. 1998 ). † The uncertainty on θ 𝜃 \theta is high and therefore these sources may not be truly extended in their FIRST images.

Name	$\log [L_{[O III]}]$	$\log [L_{H β}]$	$S_{1.4}$	$\log [L_{1.4}]$	$θ_{FIRST}$	PA_1.4/ $R_{1.4}$	$\log [L_{IR, SF}]$	SFR	$\log [L_{AGN}]$	$q_{IR}$
	(erg s^-1)	(erg s^-1)	(mJy)	(W Hz^-1)		^∘/^′′	(erg s^-1)	(M_☉ yr^-1)	(erg s^-1)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
J0945+1737	42.83	41.80	44.5[4]	24.3	1.072[3]	112^∘/2.9	45.5 ${}_{- 0.1}^{+ 0.1}$	81 ${}_{- 25}^{+ 7}$	45.5 ${}_{- 0.1}^{+ 0.2}$	1.8 ${}_{- 0.1}^{+ 0.1}$
J0958+1439	42.60	41.49	10.4[4]	23.5	1.006[9]	-	45.1 ${}_{- 0.1}^{+ 0.1}$	36 ${}_{- 11}^{+ 12}$	45.0 ${}_{- 0.3}^{+ 0.3}$	2.2 ${}_{- 0.1}^{+ 0.1}$
J1000+1242	42.80	41.84	31.8[4]	24.2	1.111[4]	158^∘/3.3	45.0 ${}_{- 0.2}^{+ 0.3}$	29 ${}_{- 16}^{+ 18}$	45.7 ${}_{- 0.1}^{+ 0.1}$	1.3 ${}_{- 0.2}^{+ 0.2}$
J1010+1413	43.21	42.15	8.8[5]	24.0	1.044[13]	-	45.7 ${}_{- 0.1}^{+ 0.2}$	122 ${}_{- 32}^{+ 60}$	46.0 ${}_{- 0.1}^{+ 0.2}$	2.3 ${}_{- 0.1}^{+ 0.2}$
J1010+0612	42.30	41.48	99.3[1.3]	24.4	1.038[3]	-	44.9 ${}_{- 0.1}^{+ 0.1}$	23 ${}_{- 7}^{+ 4}$	45.6 ${}_{- 0.1}^{+ 0.1}$	1.1 ${}_{- 0.1}^{+ 0.1}$
J1100+0846	42.82	41.77	61.3[3]	24.2	1.0230[12]	-	$<$ 46.1	$<$ 365	46.0 ${}_{- 0.1}^{+ 0.1}$	$<$ 2.5
J1125+1239	42.06	40.94	1.7[5]	23.1	0.96[6]	-	$<$ 45.4	$<$ 66	45.2 ${}_{- 0.2}^{+ 0.1}$	$<$ 2.9
J1130+1301	42.03	40.88	1.7[4]	22.9	1.07[7]	122^∘/5.1^†	44.8 ${}_{- 0.2}^{+ 0.2}$	18 ${}_{- 7}^{+ 7}$	45.1 ${}_{- 0.1}^{+ 0.1}$	2.5 ${}_{- 0.2}^{+ 0.2}$
J1216+1417	41.93	-	5.1[4]	22.9	1.07[2]	97^∘/2.7^†	$<$ 44.6	$<$ 10	45.3 ${}_{- 0.1}^{+ 0.1}$	$<$ 2.1
J1316+1753	42.87	41.81	11.4[4]	23.8	1.04[1]	-	$<$ 45.5	$<$ 77	45.4 ${}_{- 0.1}^{+ 0.1}$	$<$ 2.2
J1338+1503	42.64	-	2.4[4]	23.3	1.06[5]	-	45.5 ${}_{- 0.1}^{+ 0.1}$	81 ${}_{- 13}^{+ 13}$	45.2 ${}_{- 0.2}^{+ 0.3}$	2.7 ${}_{- 0.1}^{+ 0.2}$
J1339+1425	41.95	40.84	$<$ 3.0	$<$ 23.2	-	-	$<$ 44.9	$<$ 23	44.3 ${}_{- 0.1}^{+ 0.1}$	-
J1355+1300	42.09	41.13	3.0[5]^⋆	23.2	-	-	$<$ 45.4	$<$ 62	45.7 ${}_{- 0.1}^{+ 0.1}$	$<$ 2.7
J1356+1026	42.95	41.99	59.6[4]	24.4	1.014[2]	-	45.4 ${}_{- 0.1}^{+ 0.1}$	63 ${}_{- 17}^{+ 7}$	45.1 ${}_{- 0.2}^{+ 0.3}$	1.6 ${}_{- 0.1}^{+ 0.1}$
J1430+1339	42.72	41.88	26.4[4]	23.7	1.400[9]	77^∘/7.8	44.4 ${}_{- 0.2}^{+ 0.2}$	7 ${}_{- 3}^{+ 3}$	45.3 ${}_{- 0.1}^{+ 0.1}$	1.3 ${}_{- 0.2}^{+ 0.2}$
J1504+0151	42.20	-	3.0[4]	23.4	1.06[4]	-	45.2 ${}_{- 0.1}^{+ 0.1}$	44 ${}_{- 10}^{+ 10}$	45.2 ${}_{- 0.2}^{+ 0.1}$	2.4 ${}_{- 0.1}^{+ 0.2}$

Table 3. Table 3: Notes: The mid-infrared–far-infrared photometric flux densities, uncertainties and upper limits for the targets in our sample (in units of mJy) which are used in our SED analysis (these photometric data are shown in Fig. 3 ). These were obtained from WISE and IRAS all-sky catalogues ( Moshir et al. 1992 ; Wright et al. 2010 ; see Section 2.4 for details). The first WISE band (W1) falls outside of the wavelength range of our templates, therefore these flux densities are not included in the SED fitting procedure. † The IRAS fluxes for J0958+1439 are from Scanpi ( http://irsa.ipac.caltech.edu/applications/Scanpi ); †† J1100+0846 was not covered by IRAS observations; ⋆ J1130+1301 did not have an entry in the IRAS faint source catalogue; however, inspection of the images indicates that it is detected and therefore we obtained the fluxes from the IRAS reject catalogue.

Name	W1 (3.4 $μ$ m)	W2 (4.6 $μ$ m)	W3 (12 $μ$ m)	W4 (22 $μ$ m)	IRAS 12 $μ$ m	IRAS 25 $μ$ m	IRAS 60 $μ$ m	IRAS 100 $μ$ m
J0945+1737	1.9 $\pm$ 0.6	6.0 $\pm$ 1.8	53 $\pm$ 16	310 $\pm$ 90	$<$ 127	440 $\pm$ 70	890 $\pm$ 60	570 $\pm$ 130
J0958+1439^†	2.0 $\pm$ 0.6	3.8 $\pm$ 1.1	30 $\pm$ 9	130 $\pm$ 40	$<$ 250	$<$ 400	470 $\pm$ 140	$<$ 3000
J1000+1242	3.3 $\pm$ 1.0	7 $\pm$ 2	49 $\pm$ 15	220 $\pm$ 70	$<$ 118	$<$ 372	280 $\pm$ 70	$<$ 428
J1010+1413	1.8 $\pm$ 0.6	5.7 $\pm$ 1.7	56 $\pm$ 17	250 $\pm$ 70	$<$ 122	270 $\pm$ 50	470 $\pm$ 40	$<$ 625
J1010+0612	11 $\pm$ 3	17 $\pm$ 5	64 $\pm$ 19	290 $\pm$ 90	110 $\pm$ 30	290 $\pm$ 60	520 $\pm$ 50	590 $\pm$ 130
J1100+0846^††	15 $\pm$ 5	34 $\pm$ 10	180 $\pm$ 50	540 $\pm$ 160	-	-	-	-
J1125+1239	1.0 $\pm$ 0.3	1.4 $\pm$ 0.4	8 $\pm$ 2	29 $\pm$ 9	$<$ 250	$<$ 400	$<$ 300	$<$ 3000
J1130+1301^⋆	2.8 $\pm$ 0.8	3.6 $\pm$ 1.1	12 $\pm$ 4	45 $\pm$ 14	$<$ 114	$<$ 186	180 $\pm$ 50	$<$ 619
J1216+1417	5.9 $\pm$ 1.8	13 $\pm$ 4	42 $\pm$ 13	150 $\pm$ 50	$<$ 105	$<$ 400	$<$ 300	$<$ 620
J1316+1753	2.0 $\pm$ 0.6	3.2 $\pm$ 1.0	19 $\pm$ 6	90 $\pm$ 30	$<$ 250	$<$ 400	$<$ 300	$<$ 3000
J1338+1503	1.2 $\pm$ 0.4	2.5 $\pm$ 0.8	17 $\pm$ 5	80 $\pm$ 30	$<$ 134	$<$ 126	340 $\pm$ 50	$<$ 583
J1339+1425	0.5 $\pm$ 0.2	0.5 $\pm$ 0.1	1.1 $\pm$ 0.4	6 $\pm$ 2	$<$ 250	$<$ 400	$<$ 300	$<$ 3000
J1355+1300	7 $\pm$ 2	15 $\pm$ 4	40 $\pm$ 13	80 $\pm$ 20	$<$ 250	$<$ 400	$<$ 300	$<$ 3000
J1356+1026	2.2 $\pm$ 0.7	4.6 $\pm$ 1.4	27 $\pm$ 8	190 $\pm$ 60	$<$ 58	210 $\pm$ 50	800 $\pm$ 90	600 $\pm$ 150
J1430+1339	7 $\pm$ 2	11 $\pm$ 3	40 $\pm$ 12	180 $\pm$ 50	$<$ 105	190 $\pm$ 30	260 $\pm$ 40	$<$ 605
J1504+0151	2.0 $\pm$ 0.6	2.8 $\pm$ 0.8	10 $\pm$ 3	30 $\pm$ 9	$<$ 108	$<$ 93	210 $\pm$ 40	$<$ 508

Table 4. Table 4: Notes: Velocity values and sizes derived from the galaxy-integrated spectra and velocity maps shown in Fig. 5 and Fig. A1-A15 (see Section 3.1 and Fig. 4 for velocity definitions). (1) Object name; (2) value of the [O iii ] emission-line width, W 80 subscript 𝑊 80 W_{80} , derived from the galaxy-integrated spectra; (3) maximum value of W 80 subscript 𝑊 80 W_{80} from the spatially resolved maps (with a 95% clipping threshold to remove spurious pixels); (4) velocity offset, Δ v Δ 𝑣 \Delta v , derived from the galaxy-integrated spectrum; (5) absolute value of the maximum Δ v Δ 𝑣 \Delta v in the spatially resolved maps (with 95% clipping); (6) the value of maximum [O iii ] velocity, v 02 subscript 𝑣 02 v_{02} , derived from the integrated spectrum; (7) the maximum velocity shear of the median velocity, v med subscript 𝑣 med v_{\rm{med}} , from the velocity maps (with 95% clipping); (8) qualitative description of the peak velocity maps: R = = rotation-like velocity fields; R/I = = red-blue velocity gradient but irregular velocity field; I = = irregular velocity field; F = = no velocity gradients (see Section 4.2.1 ); (9) observed semi-major axis of the [O iii ] emission-line region from fitting an ellipse with uncertainties that are the typical seeing (i.e., 0. ′′ 7); (10) total projected extent of observed broad emission-line profiles (i.e., with W 80 > 600 subscript 𝑊 80 600 W_{80}>600 km s -1 ) with conservative uncertainties of 2 × 2\times the seeing. Measurement uncertainties on the galaxy-integrated velocities are typically < 10 absent 10 <10 % and therefore the dominant source of uncertainty is the physical interpretation of these values (see Section 4.2 ).

Name	$⟨$ $W_{80}$ $⟩$	$W_{80, \max}$	$⟨$ $Δ v$ $⟩$	$\| Δ v_{\max} \|$	$⟨$ $v_{02}$ $⟩$	$Δ v_{med, \max}$	Gal. Kinematics	$R_{[O III]}$	$D_{600}$
	(km s^-1)	(km s^-1)	(km s^-1)	(km s^-1)	(km s^-1)	(km s^-1)		(kpc)	(kpc)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
J0945+1737	1009	1284	$-$ 273	284	$-$ 1511	138	I	2.7 $\pm$ 1.6	$\geq$ 12
J0958+1439	815	904	$-$ 46	107	$-$ 866	190	R	2.6 $\pm$ 1.4	$\geq$ 10
J1000+1242	795	873	$-$ 58	171	$-$ 761	311	R/I	4.3 $\pm$ 1.8	14
J1010+1413	1449	1525	$-$ 229	350	$-$ 1523	299	R/I	3.9 $\pm$ 2.3	$\geq$ 16
J1010+0612	1280	1468	$-$ 95	216	$-$ 1267	138	R	1.6 $\pm$ 1.3	$\geq$ 6
J1100+0846	1066	1367	$-$ 30	148	$-$ 1192	55	R	1.9 $\pm$ 1.3	$\geq$ 10
J1125+1239	1285	1574	$-$ 265	424	$-$ 1547	73	F	2.9 $\pm$ 2.0	$\geq$ 9
J1130+1301	778	849	149	173	$-$ 616	140	R	2.8 $\pm$ 1.7	8
J1216+1417	1228	1456	124	230	$-$ 1115	40	F	1.5 $\pm$ 1.1	$\geq$ 6
J1316+1753	1127	1169	$-$ 191	326	$-$ 1216	500	R/I	3.1 $\pm$ 1.8	$\geq$ 14
J1338+1503	890	1085	124	182	$-$ 813	145	R/I	3.5 $\pm$ 2.2	$\geq$ 13
J1339+1425	672	724	147	178	$-$ 505	84	R	2.5 $\pm$ 1.7	6
J1355+1300	667	953	$-$ 184	277	$-$ 797	71	R	3.5 $\pm$ 1.8	$\geq$ 6
J1356+1026	900	964	$-$ 215	544	$-$ 1049	523	I	3.1 $\pm$ 1.5	$\geq$ 11
J1430+1339	822	1042	$-$ 152	268	$-$ 999	439	R/I	1.8 $\pm$ 1.1	$\geq$ 9
J1504+0151	1184	1181	$-$ 495	520	$-$ 1739	149	R	2.9 $\pm$ 2.1	$\geq$ 7

Equations12

q_{I R} = lo g [\frac{S _{IR} /3.75 \times 1 0 ^{12} W m ^{- 2}}{S _{1.4} / W m ^{- 2} Hz ^{- 1}}]

q_{I R} = lo g [\frac{S _{IR} /3.75 \times 1 0 ^{12} W m ^{- 2}}{S _{1.4} / W m ^{- 2} Hz ^{- 1}}]

\frac{M _{gas}}{2.82 \times 1 0 ^{9} M _{⊙}} = (\frac{L _{H β}}{1 0 ^{43} erg s ^{- 1}}) (\frac{n _{e}}{100 cm ^{- 3}})^{- 1}

\frac{M _{gas}}{2.82 \times 1 0 ^{9} M _{⊙}} = (\frac{L _{H β}}{1 0 ^{43} erg s ^{- 1}}) (\frac{n _{e}}{100 cm ^{- 3}})^{- 1}

E_{kin} = \frac{1}{2} M_{gas} v_{gas}^{2},

E_{kin} = \frac{1}{2} M_{gas} v_{gas}^{2},

\dot{M}_{out, M1} = \frac{3 M _{gas} v _{out}}{r}

\dot{M}_{out, M1} = \frac{3 M _{gas} v _{out}}{r}

\dot{E}_{out, M1} = \frac{M ˙}{2} (v_{out}^{2} + 3 σ^{2})

\dot{E}_{out, M1} = \frac{M ˙}{2} (v_{out}^{2} + 3 σ^{2})

\dot{E}_{out, M2} \approx 1.5 \times 1 0^{46} r_{10}^{2} v_{1000}^{3} n_{0.5} erg s^{- 1},

\dot{E}_{out, M2} \approx 1.5 \times 1 0^{46} r_{10}^{2} v_{1000}^{3} n_{0.5} erg s^{- 1},

Peer Reviews

No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.

Videos

No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.

Full text

Kiloparsec-scale outflows are prevalent among luminous AGN: outflows

and feedback in the context of the overall AGN population

C. M. Harrison, ${}^{\!1\,*}$ D. M. Alexander, ${}^{\!1}$ J. R. Mullaney ${}^{\!1,2}$ and A. M. Swinbank ${}^{\!3}$

1Department of Physics, Durham University, South Road, Durham, DH1 3LE, UK

2Department of Physics & Astronomy, University of Sheffield, Sheffield, S3 7RH, UK

3Institute for Computational Cosmology, Durham University, South Road, Durham, DH1 3LE, UK

∗Email: [email protected]

Abstract

We present integral field unit (IFU) observations covering the [O iii] $\lambda\lambda 4959,5007$ and H $\beta$ emission lines of sixteen $z<0.2$ type 2 active galactic nuclei (AGN). Our targets are selected from a well-constrained parent sample of $\approx 24,000$ AGN so that we can place our observations into the context of the overall AGN population. Our targets are radio-quiet with star formation rates ( $\lesssim$ [10–100] M*⊙* yr*-1*) that are consistent with normal star-forming galaxies. We decouple the kinematics of galaxy dynamics and mergers from outflows. We find high-velocity ionised gas (velocity widths $\approx 600$ –1500 km s*-1*; maximum velocities $\leq 1700$ km s*-1*) with observed spatial extents of $\gtrsim$ (6–16) kpc in all targets and observe signatures of spherical outflows and bi-polar superbubbles. We show that our targets are representative of $z<0.2$ , luminous (i.e., $L_{{\rm[O~{}III]}}>10^{41.7}$ erg s*-1*) type 2 AGN and that ionised outflows are not only common but also in $\geq$ 70% (3 $\sigma$ confidence) of cases, they are extended over kiloparsec scales. Our study demonstrates that galaxy-wide energetic outflows are not confined to the most extreme star-forming galaxies or radio-luminous AGN; however, there may be a higher incidence of the most extreme outflow velocities in quasars hosted in ultra-luminous infrared galaxies. Both star formation and AGN activity appear to be energetically viable to drive the outflows and we find no definitive evidence that favours one process over the other. Although highly uncertain, we derive mass outflow rates (typically $\approx$ 10 $\times$ the SFRs), kinetic energies ( $\approx 0.5$ –10% of $L_{{\rm AGN}}$ ) and momentum rates (typically $\gtrsim 10$ – $20\times L_{{\rm AGN}}/c$ ) consistent with theoretical models that predict AGN-driven outflows play a significant role in shaping the evolution of galaxies.

keywords:

galaxies: active; — galaxies: kinematics and dynamics; — quasars: emission lines; — galaxies: evolution

1 Introduction

One of the most remarkable discoveries in modern astronomy is that all massive galaxies host supermassive black holes (BHs) with masses that are proportional to that of their host galaxy spheroid (e.g., Kormendy & Richstone 1995; Magorrian et al. 1998; Tremaine et al. 2002; Gültekin et al. 2009; see Kormendy & Ho 2013 for a review). These BHs primarily grow through mass accretion that is visible as active galactic nuclei (AGN) in the centre of galaxies. Theoretical models of galaxy formation have found it necessary to implement AGN “feedback” processes, during which AGN activity injects energy into the gas in the larger-scale environment, in order to reproduce the properties of local massive galaxies, intracluster gas and the intergalactic medium (e.g., BH-mass–spheroid mass relationship; the sharp cut-off in the galaxy luminosity function; colour bi-modality; metal enchrichment; X-ray temperature–luminosity relationship; e.g., Silk & Rees 1998; Churazov et al. 2005; Bower et al. 2006; Hopkins et al. 2006; McCarthy et al. 2010; Gaspari et al. 2011). Placing observational constraints on how AGN activity couples to the gas in galaxies and halos, and where these processes are most prevalent, is an important area of ongoing research (for reviews see: Cattaneo et al. 2009; Alexander & Hickox 2012; Fabian 2012; McNamara & Nulsen 2012).

Several of the most successful galaxy formation models invoke a dramatic form of energy injection (sometimes called the “quasar mode” or “starburst mode”) where AGN drive galaxy-wide (i.e., $\gtrsim 0.1$ –10 kpc) energetic outflows that expel gas from their host galaxies and consequently this shuts down future BH growth and star formation and/or enriches the larger-scale environment with metals (e.g., Silk & Rees 1998; Fabian 1999; Benson et al. 2003; King 2003; Granato et al. 2004; Di Matteo et al. 2005; Springel et al. 2005; Hopkins et al. 2006, 2008; Booth & Schaye 2010; Debuhr et al. 2012). This is in contrast to the “maintenance mode” (or “hot-halo”) feedback where radio jets, launched by AGN, control the level of cooling of the hot gas in the most massive halos (see Bower et al. 2012 and Harrison 2014 for a discussion on the two modes). While there is little doubt that star formation processes (e.g., stellar winds and supernovae) drive galaxy-wide outflows (e.g., Heckman et al. 1990; Lehnert & Heckman 1996; Swinbank et al. 2009; Genzel et al. 2011; Newman et al. 2012; Bradshaw et al. 2013; see review in Veilleux et al. 2005) and are an integral part of galaxy evolution (e.g., Dalla Vecchia & Schaye 2008; Hopkins et al. 2013a), it is believed that AGN activity is required to drive the highest velocity outflows and are particularly important for the evolution of the most massive galaxies (e.g., Benson et al. 2003; McCarthy et al. 2011; Hopkins et al. 2013b; Zubovas & King 2014).

X-ray and ultraviolet spectroscopy has shown that a high-fraction, and potentially all, of high-accretion rate AGN drive high-velocity outflows ( $v\approx 0.1c$ ) close to their accretion disks (i.e., on sub-parsec scales; e.g., Blustin et al. 2003; Reeves et al. 2003; Ganguly & Brotherton 2008; Tombesi et al. 2010; Gofford et al. 2011). However, are AGN capable of driving outflows over galaxy scales as is required by galaxy formation models? A diagnostic that is commonly used to identify outflowing gas over large scales is broad (i.e., exceeding that expected from galaxy dynamics), asymmetric and high-velocity [O iii] $\lambda$ 5007 emission-line profiles. This emission line traces the kinematics of the ionised gas; however, we briefly note that outflowing gas has been observed in multiple gas phases in some AGN (e.g., Rupke et al. 2005; Martin 2005; Fischer et al. 2010; Feruglio et al. 2010; Alatalo et al. 2011; Veilleux et al. 2013; Cimatti et al. 2013; Rupke & Veilleux 2013). As a forbidden transition the [O iii] $\lambda$ 5007 emission line cannot be produced in the high-density sub-parsec scales of the AGN broad-line region (BLR) making it a good tracer of the kinematics in the narrow-line region (NLR) and can be observed over parsecs to tens of kiloparsecs (e.g., Wampler et al. 1975; Wilson & Heckman 1985; Boroson et al. 1985; Stockton & MacKenty 1987; Osterbrock 1989). The [O iii] $\lambda$ 5007 emission line has long been used to identify outflowing ionised gas in small samples of local and low redshift AGN (e.g., Weedman 1970; Stockton 1976; Veron 1981; Heckman et al. 1981; Heckman et al. 1984; Feldman et al. 1982; Vrtilek 1985; Whittle 1985; Whittle et al. 1988; Veilleux 1991; Veilleux et al. 1995; Boroson & Green 1992; Nelson & Whittle 1996); however, the small sample sizes makes it difficult to know how representative these observations are. More recently, large systemic spectroscopic surveys (e.g., the Sloan Digital Sky Survey [SDSS]; York et al. 2000) have enabled the study of NLR kinematics in hundreds to tens of thousands of AGN (e.g., Boroson 2005; Greene & Ho 2005; Komossa et al. 2008; Zhang et al. 2011; Wang et al. 2011; Mullaney et al. 2013) that can constrain both the ubiquity of these outflow features and study them as a function of key AGN properties.

Mullaney et al. (2013) used the SDSS spectroscopic database to study the one-dimensional kinematic properties of [O iii] $\lambda$ 5007 by performing multi-component fitting to the optical emission-line profiles of $\approx 24,000$ , $z<0.4$ optically selected AGN. They showed that $\approx 17$ % of the AGN have emission-line profiles that indicate their ionised gas kinematics are dominated by outflows and a considerably larger fraction are likely to host ionised outflows at lower levels. The fraction of AGN with ionised gas kinematics dominated by outflows increases to $\gtrsim 40$ % for the more radio-luminous AGN (i.e., those with $L_{\rm{1.4GHz}}>10^{23}$ W Hz*-1*), in contrast, when taking into account intrinsic correlations, this fraction shows little dependence on [O iii] luminosity or Eddington ratio (Mullaney et al. 2013). In agreement with smaller studies (e.g., Heckman et al. 1981; Whittle 1985, 1992; Gelderman & Whittle 1994; Nelson & Whittle 1996; Nesvadba et al. 2011; Kim et al. 2013; see also Greene & Ho 2005), this result shows that ionised outflows are most common in AGN that have moderate-to-high radio luminosities. However, while insightful, the origin of the radio emission is often unknown, particularly at the moderate radio luminosities (i.e., $L_{\rm{1.4GHz}}\approx 10^{23}$ –1024 W Hz*-1*) where AGN cores, radio jets, shocks and high-levels of star formation could all contribute (e.g., Del Moro et al. 2013; Condon et al. 2013; Zakamska & Greene 2014). It is therefore vital to measure star-formation rates (SFRs) and properly investigate the origin of the radio emission in the sources that host these outflows to properly interpret these results.

The one-dimensional spectra discussed above provide no insight into the spatial extent or structure of the outflows, for this, we must appeal to spatially resolved spectroscopy. Both longslit and integral-field unit (IFU) observations of AGN, over a large range of redshifts, have identified disturbed and high-velocity ionised gas over kiloparsec scales (e.g., McCarthy et al. 1996; Villar-Martín et al. 1999; Colina et al. 1999; Holt et al. 2008; Nesvadba et al. 2007, 2008; Lipari et al. 2009a,b; Fu & Stockton 2009; Alexander et al. 2010; Humphrey et al. 2010; Greene et al. 2011; Rupke & Veilleux 2011, 2013; Harrison et al. 2012; Westmoquette et al. 2012; Cano-Díaz et al. 2012; Husemann et al. 2013; Liu et al. 2013a, b; Förster Schreiber et al. 2013). Several of these studies have revealed considerable masses of outflowing ionised gas with velocities higher than the galaxy escape velocities, in apparent agreement with basic predictions from galaxy formation models. However, a key limitation of these studies is that it is often difficult to place these observations into the context of the overall AGN and galaxy populations as the samples are small, inhomogeneous and/or only represent the most extreme AGN or star-forming systems in the Universe.

In this paper we are interested in measuring the prevalence, properties, and the potential impact of galaxy-wide energetic outflows. We present Gemini (South) Multi-Object Spectrograph (GMOS; Allington-Smith et al. 2002) IFU observations of sixteen $0.08<z<0.2$ type 2 AGN drawn from the parent sample of Mullaney et al. (2013). Importantly, this means that we can place our IFU observations into the context of the overall AGN population. So that we can properly interpret our results we perform SED fitting and analyse the available radio data to measure SFRs, AGN luminosities and to search for evidence of radio jets in all of our targets. In Section 2, we give details of the IFU observations, data reduction and SED fitting. In Section 3 we provide details of our analysis of the ionised gas kinematics and in Section 4 we present our results. In Section 5 we discuss our results and their implication for understanding galaxy evolution and in Section 6 we give our conclusions. We provide background information and a discussion of the results on individual sources in Appendix A. We have adopted $H_{0}=71$ km s*-1* Mpc*-1*, $\Omega_{\rm{M}}=0.27$ and $\Omega_{\Lambda}=0.73$ throughout; in this cosmology, 1 arcsecond corresponds to $\approx$ 1.5–3.3 kpc for the redshift range of our sample ( $z=0.08$ –0.2).

2 Targets, observations and data reduction

2.1 Target selection

We are interested in studying the spatially resolved ionised gas kinematics in luminous AGN, with the aim of understanding the prevalence and properties of galaxy-wide ionised outflows. For this study, we selected sixteen $z<0.2$ type 2 AGN from the parent catalogue of Mullaney et al. (2013) to observe with the Gemini-GMOS (South) IFU. We describe our selection criteria below (also illustrated in Figure 1) and we provide a list of the targets, along with their positions and redshifts, in Table LABEL:Tab:tab1.

Our parent catalogue (Mullaney et al. 2013) contains fits to the emission-line profiles of 24 264, $z<0.4$ , optical AGN (identified using a combination of “BPT” diagnostics, Baldwin et al. 1981, and emission-line widths) from the SDSS data release Data Release 7 (DR7; Abazajian et al. 2009). Mullaney et al. (2013) decompose the [O iii] $\lambda$ 5007 emission-line profiles into two components (“narrow” and “broad”) to identify emission-line profiles that are broad and asymmetric, indicative of outflow kinematics. A primary aim of this work is to establish if the broad and asymmetric emission-line features observed in one-dimensional spectra are spatially extended and we therefore selected sources that have a luminous broad [O iii] $\lambda$ 5007 emission-line component. Our definition of a luminous broad component is one that contributes at least 30% of the total flux and has FWHM $>$ 700 km s*-1* (see Figure 1). Additionally, we only selected sources that are classified as type 2 AGN by Mullaney et al. (2013), i.e., the permitted and forbidden lines are of similar width, implying that we cannot directly see the broad-line region in these objects (e.g., Antonucci 1993). From these, we selected the sources that have total observed [O iii] luminosities of $L_{\rm{[O~{}III]}}>5\times 10^{41}$ erg s*-1* and have a redshift $z<0.2$ to ensure sufficient emission-line flux for measuring the spatially resolved kinematics with reasonable exposure times for our IFU observations. Finally, we chose sixteen targets with celestial co-ordinates that are accessible with Gemini-South. Our final target list is given in Table LABEL:Tab:tab1 and we give notes in Appendix A on the relevant multi-wavelength observations of these targets that can be found in the literature. The majority of our targets have received little or no attention previously in the literature, with the exceptions of J0945+1737, J1316+1753 and J1356+1026 that have been studied in some detail (see Appendix A).

It is important to address where our targets fit into the overall AGN population if we are to draw global conclusions from our observations. Firstly we note that our targets are all type 2 AGN, which are thought to constitute at least half of the overall AGN population (e.g., Lawrence & Elvis 2010), and that they have observed [O iii] luminosities in the range $L_{{\rm[O~{}III]}}=10^{41.9}$ – $10^{43.2}$ erg s*-1* (see Table 2) that fit the luminosity criteria for type 2 (“obscured”) quasars (i.e., $L_{{\rm[O~{}III]}}\gtrsim 10^{41.9}$ erg s*-1*; Reyes et al. 2008). The absorption corrected [O iii] luminosities of our targets have the range $L_{{\rm[O~{}III]}}=10^{42.4}$ – $10^{43.2}$ erg s*-1* (calculated using the H $\alpha$ /H $\beta$ emission-line flux ratios from Mullaney et al. 2013 and following Calzetti et al. 2000). Additionally, we emphasise that all of our targets are classified as “radio-quiet” in the $\nu L_{\nu}$ (1.4 GHz)- $L_{{\rm[O~{}III]}}$ plane (Xu et al. 1999; Zakamska et al. 2004) which make up the majority of the luminous type 2 AGN population (i.e., $\approx$ 90% based on the sample in Zakamska et al. 2004). Furthermore, considering all galaxies at low redshift (i.e., $0.0<z<0.5$ ), the radio luminosities of our targets [i.e., $L_{{\rm 1.4\,GHz}}=(<0.8-25)\times 10^{23}$ W Hz*-1*] are known to be common with a space density of $\approx$ 10*-4* Mpc*-3* and $\approx$ 10*-5* Mpc*-3* for sources with $L_{{\rm 1.4\,GHz}}=10^{23}$ W Hz*-1* and $L_{{\rm 1.4\,GHz}}=10^{24}$ W Hz*-1* respectively (Simpson et al. 2012).

The key advantage of selecting our sources from a large parent sample of AGN is that we can quantitatively define how representative our targets are. In Figure 2 we show histograms for the FWHM of the broad [O iii] emission-line components for both the parent sample and our targets. Considering all of the type 2 AGN from Mullaney et al. (2013) with $z<0.2$ and with an [O iii] luminosity, $L_{{\rm[O~{}III]}}>5\times 10^{41}$ erg s*-1* (i.e., to match our selection criteria), we find that 45 $\pm$ 3% (246 out of 546 sources and assuming $\sqrt{N}$ errors) have a luminous [O iii] component with FWHM[B] $>$ 700 km s*-1* (Figure 2). This indicates that roughly half of luminous type 2 AGN have a luminous broad component to their [O iii] emission-line profiles and demonstrates that our targets do not represent a rare sub-population. Additionally, this fraction can be considered a lower limit, as the fraction of sources that host weaker broad components will be considerably higher. In Figure 2 we also show histograms of the radio luminosities of our targets and the type 2 AGN in the parent sample that have luminous and broad [O iii] emission-line components. Our targets cover a similar range in radio luminosities to those in the parent sample further demonstrating that our targets are representative of the parent population.

2.2 Gemini-South GMOS observations and data reduction

The observations of our targets were performed using Gemini-South GMOS (Allington-Smith et al. 2002) in IFU mode. The GMOS IFU uses a lenslet array to slice the focal plane into several small components which are each coupled to a fibre. We made use of the one-slit mode with the B1200 grating to obtain the required wavelength coverage to observe the [O iii] $\lambda\lambda$ 4959,4007 and H $\beta$ emission-lines for all our targets. In this mode the IFU consists of 25 $\times$ 20 lenslets sampling a $5^{\prime\prime}\times 3.^{\prime\prime}5$ field-of-view. The GMOS IFU also has a dedicated set of 250 lenslets to simultaneously observe the sky $\approx$ 1*′* away from the target field-of-view. We determined the instrumental dispersion for the observations of each source ( $\Delta v\approx 50$ –60 km s*-1*, with uncertainties $\lesssim$ 5 km s*-1*) by measuring the widths of several skylines close to the observed wavelengths of the emission lines. The emission-line profiles we measure are typically non-Gaussian and the non-parametric line-widths we quote (see Section 3.1) are very large ( $\gtrsim$ 500 km s*-1*); therefore, we do not correct for this instrumental broadening. This will result in the quoted line-width measurements being broadened by $\lesssim$ 2%.

The observations were undertaken during two observing programmes: GS-2010A-Q-8 (PI: Mullaney) taken between 2010 February 09 and 2010 March 19, and GS-2012A-Q-21 (PI: Harrison) taken between 2012 February 16 and 2012 June 26. For the observations in the 2012A run, that covered the fainter targets, the on-source exposures consisted of 6 $\times$ 1500s while for the 2010A run, that covered the brighter targets, the exposure times consisted of 4 $\times$ 1300s. Table LABEL:Tab:tab1 indicates during which run each target was observed. The observations were performed with $V$ -band seeing of $\lesssim 0.^{\prime\prime}8$ (typically $\approx 0.^{\prime\prime}7$ , from measurements of the acquisition images taken before each IFU observation).

We reduced the data using the standard Gemini iraf gmos pipeline to perform wavelength calibration and flat fielding and we used a dedicated set of IDL routines to perform sky subtraction, cosmic ray removal and cube construction. We did not resample the pixels and therefore the pixel scale of the final cubes is 0.*′′17 $\times$ 0.′′*2. To find spatial centroids for the individual data cubes, emission-line free continuum images from each exposure were produced that were then spatially aligned and co-added to create the final mosaics, using a median stack with a 3 $\sigma$ clipping threshold to remove the remaining cosmetic defects and cosmic rays. Flux calibration for each target was performed in two stages. Stage one involved reducing standard star observations in an identical manner to the targets and consequently obtaining the instrumental response using iraf routines from the onedspec package. Stage two used the SDSS spectra to find the absolute flux calibration for each of the objects. This was achieved by fitting a low order polynomial to the emission-line free SDSS spectrum and also an emission-line free spectrum extracted from our IFU data cubes over the same spatial region as covered by the SDSS fibers (i.e., $\approx 3^{\prime\prime}$ diameter). We verified that the equivalent widths of the emission lines were consistent between these two spectra and then used the ratio between these fits to apply the flux calibration to each of the data cubes. We estimate a conservative uncertainty of 15% on the absolute flux calibration using this process (see Abazajian et al. 2009 and Liu et al. 2013a)

2.3 Aligning the IFU data to SDSS images

As part of our analyses we compare the location of the kinematic features we observe in our IFU data, to the host galaxy morphologies based upon three-colour ( $g$ , $r$ , $i$ ) SDSS images (Section 4.2). We aligned the SDSS three-colour images to our IFU data by matching the surface brightness peak in the $g$ -band SDSS images to the surface brightness peak of images constructed from our IFU data cubes, collapsed around $\approx$ 5000 Å. We then rotated the SDSS images to match the orientation of our IFU data cubes. We note that we only use these images for qualitative descriptions of the galaxy morphology with respect to our IFU data and therefore we do not require a more accurate alignment process.

2.4 SFRs, AGN luminosities and the origin of the radio emission

To thoroughly interpret our results we investigate the properties of the AGN and host galaxies of our targets in more detail (i.e., we measure SFRs and AGN luminosities and explore the origin of the radio emission). To obtain the bolometric AGN luminosities and SFRs of our targets (see Table 2) we performed mid-infrared to far-infrared SED fitting to archival photometric data that is available for our targets (described below). Our procedure decomposes the emission from AGN activity and star-formation, so that we do not need to make any assumptions about the relative contribution from these two processes. At the end of this sub-section, we also briefly investigate the possible presence of radio AGN activity (i.e., radio cores or radio jets) in our targets by combining the results of our SED fitting with archival radio data.

For our SED fitting, we obtained mid-infrared to far-infrared photometric flux densities, uncertainties and upper limits (over 3 $\mu$ m–100 $\mu$ m; see Table 3) from (1) the Wide-field Infrared Survey Explorer all-sky survey (WISE; Wright et al. 2010) using the closest source to the SDSS position (all are $<0.^{\prime\prime}6$ ) and (2) the Infrared Astronomical Satellite (IRAS; Neugebauer et al. 1984) faint source catalogue (Moshir et al. 1992) with a matching radius of $\leq$ 30*′′*111We verified these matches by following the log likelihood method of matching IRAS to WISE counterparts outlined in Wang et al. (2014).. We added in quadrature 30% of the measured flux densities to the quoted WISE uncertainties to account for calibration uncertainties in the WISE data (Wright et al. 2010) and the discrete nature of the SED templates used in our analysis (see below; also see Del Moro et al. 2013). For the sources that we were not able to obtain IRAS flux density measurements (see Table 3) we estimated conservative maximum upper limits of 0.25 mJy, 0.4 mJy, 0.3 mJy and 3 mJy for the 12 $\mu$ m , 25 $\mu$ m , 60 $\mu$ m , and 100 $\mu$ m bands respectively. One source (SDSS J1100+0846) falls into an area of the sky that was not observed by IRAS observations and we therefore cannot place photometric constraints on the IRAS bands for this source. The photometric data we used are shown in Figure 3.

To perform the SED fitting we followed the procedure described in Mullaney et al. (2011) that simultaneously fits an empirically derived AGN model (paramaterised in an analytical form) and empirically derived host-galaxy templates to the infrared photometric data (Table 3). All of the sources in our sample are known to host an AGN (from optical spectroscopy) and additionally all but one of the sources are classified as an infrared AGN based upon the WISE colours (W2 $-$ W1 and W3 $-$ W2) by falling inside the “AGN-wedge” defined in Mateos et al. (2012).222SDSS J1339+1425 does not have WISE colours that place it inside the “AGN-wedge”; however, we still found that including an AGN template provided a better fit than any of the star-forming templates alone. We therefore simultaneously fit these data with the five star-forming galaxy templates (“SB1”–“SB5”) and the mean AGN model, originally defined in Mullaney et al. (2011) and extended to cover the wavelength range 3–105 $\mu$ m, by Del Moro et al. (2013).333We tested how much of an affect our limited range of templates has on our quoted results (Table 2) by: (1) re-fitting including an Arp 220 model SED and (2) allowing the second power-law slope of the AGN model to be a free parameter (i.e., $\alpha_{2}$ in Mullaney et al. 2011). The first of these changes allows for a more extreme ULIRG-like SED that is not covered by our main SB templates. The second change allows for a very large range in instrinsic AGN SEDs, to account for the variations seen in AGN of different luminosities (see Fig. 7 in Mullaney et al. 2011). When performing the SED fitting with these two additions we found that all the results were consistent within the errors of our quoted results (Table 2) with only two minor exceptions: (1) The SFR for J1430+1339 had an upper limit of $<2$ M*⊙* yr*-1and (2) the AGN luminosity of J1504+0151 was $\log[L_{{\rm AGN}}]=45.4_{-0.3}^{+0.1}$ erg s-1*. We therefore conclude our primary choice of templates are sufficient for this work. To derive the AGN luminosities and SFRs (Table 2) we use the star-forming template plus AGN template combination that gives the lowest overall $\chi^{2}$ value, rejecting all fits that lie above the photometric upper limits (given in Table 3; see Fig. 3). For the sources where we have no detections in the far-infrared bands (i.e., the IRAS bands), we are unable to measure reliable SFRs and we therefore calculate conservative upper limits by increasing the normalisation of the star-forming templates until either the photometric upper limits were reached or the solution was 3 $\sigma$ above one of the photometric data points.

We calculated the SFRs for each source (Table 2) by measuring the infrared luminosities of the star-formation components from the best-fitting SED solutions ( $L_{\rm{IR,SF}}$ ; integrated over 8–1000 $\mu$ m) and using the relationship of Kennicutt (1998) (corrected to a Chabrier IMF by dividing by a factor of 1.7; Chabrier 2003). To determine the bolometric AGN luminosities for each source (Table 2), we first calculated the AGN continuum luminosity at 6 $\mu$ m from the best fitting SED solution and converted this to an AGN bolometric luminosity using a correction factor of $\times$ 8 (Richards et al. 2006; see Table 2). The conservative upper and lower bounds on both the SFRs and AGN luminosities are derived from summing in quadrature the range in solutions from the different star-formation templates (only those which did not exceed the upper limits) and the formal uncertainties from the best-fit solution. We emphasise that our method decomposes the emission due to an AGN and star formation, therefore removing the need to assume the relative contributions from these two processes. Furthermore, this procedure produces consistent SFRs and AGN luminosities with those derived from mid-infrared spectra where available (see Appendix A and Footnote 14; also see Del Moro et al. 2013).

The bolometric AGN luminosities for our targets span $L_{{\rm AGN}}=(0.2-10)\times 10^{45}$ erg s*-1* (with a median $L_{{\rm AGN}}=2\times 10^{45}$ erg s*-1*) and therefore cover the classical “quasar” threshold of $L_{{\rm AGN}}=10^{45}$ erg s*-1*. We note that these mid-infrared derived AGN luminosities are a factor of $\approx$ 1–20 lower than those that would be predicted using the [O iii] luminosity and the relationship of Heckman et al. (2004) (i.e., $L_{{\rm AGN}}=3500L_{{\rm[O~{}III]}}$ ). However, this is consistent with other studies of luminous ( $L_{{\rm[O~{}III]}}\gtrsim 10^{42}$ erg s*-1*) type 2 AGN, at similar redshifts, that show that $L_{{\rm[O~{}III]}}$ is not linearly correlated with the nuclear luminosity and can over predict the AGN luminosity by an order of magnitude or more (Schirmer et al. 2013; Hainline et al. 2013).

Of the sixteen sources in our sample, eight have far-infrared luminosities that classify them as luminous infrared galaxies (LIRGs; $L_{{\rm IR,SF}}=10^{11}$ –10 ${}^{12}\,\hbox{L$ {}{\odot} $}$ ) and one as an ultraluminous infrared galaxy (ULIRG: $L_{{\rm IR,SF}}>10^{12}\,\hbox{L$ {}{\odot} $}$ ). Of the remaining seven sources, six of them have upper-limits on their far-infrared luminosities which are less than or consistent with the luminosity of LIRGs (see Table 2).444The remaining source, SDSS J1100+0846, which is not covered by IRAS and therefore has poor FIR constraints, has an upper limit consistent with a ULIRG. These infrared luminosities correspond to SFRs of $\lesssim$ [7–120] M*⊙* yr*-1*and are typical for AGN of these luminosities at these redshifts (e.g., Zakamska et al. 2008; Mullaney et al. 2010; Rosario et al. 2012).

Finally, we investigate the origin of the radio emission from our targets. We plot the 1.4 GHz radio flux densities (given in Table 2) on the SEDs shown in Figure 3. We identify if there is radio emission significantly above that expected from their ongoing star-formation (e.g., Helou et al. 1985; Condon et al. 1995), that may indicate radio emission due to an AGN either in form of a radio core or radio jets (e.g., see Del Moro et al. 2013 and references there-in) or due to shocks (e.g., Zakamska & Greene 2014). To do this we use the definition given by Ivison et al. (2010) (see also Helou et al. 1985), calculating the ratio between the far-infrared and radio emission ( $q_{{\rm IR}}$ ) as

[TABLE]

where $S_{{\rm IR}}$ is the rest-frame far-infrared (8–1000 $\mu$ m) flux and $S_{1.4}$ is the rest-frame 1.4 GHz flux density assuming, $S_{\nu}\propto\nu^{-\alpha}$ , with a radio spectral index of $\alpha=0.7$ . We give the value of $q_{{\rm IR}}$ for each source in our sample in Table 2. The quoted uncertainties combine the uncertainties on $S_{1.4}$ , $S_{{\rm IR}}$ and a range in the unknown radio spectral indices (we use $\alpha=0.2$ –1.5). Star-forming galaxies have $q_{{\rm IR}}\approx 2.4$ and we define “radio excess” sources as those with $q_{{\rm IR}}\leq 1.8$ , i.e., $\gtrsim$ 2.5 $\sigma$ away from the peak seen in the star-forming galaxies (Ivison et al. 2010; see also Del Moro et al. 2013 who use a similar definition). Using this definition we identify five of our targets as being radio excess and five sources that are classified as radio normal (see Table 2 and Figure 3); however, we cannot rule out low-level radio emission above that expected from star formation in these targets as it can be difficult to identify using this method, especially in systems with high SFRs (e.g., Del Moro et al. 2013). The remaining six sources have upper limits on their $S_{{\rm IR}}$ values which results in upper limits on their $q_{{\rm IR}}$ values that are consistent with them being either radio normal or radio excess sources.

Based on FIRST and NVSS images only one of our sources (J1430+1339) shows clear evidence for luminous extended radio structures on scales $>$ 5*′′* (see Table 2 and Appendix A). Therefore, we follow the method of Kimball & Ivezić (2008) to search for evidence of extended radio structures (on $\approx 2$ –5*′′* scales) that compares the peak and integrated 1.4 GHz FIRST flux densities ( $F_{{\rm peak}}$ and $F_{{\rm int}}$ respectively). They define the parameter $\theta=(F_{{\rm int}}/F_{{\rm peak}})^{0.5}$ and any source with $\theta>1.06$ is classed as spatially resolved. We provide $\theta$ for all of our targets with FIRST detections in Table 2. For the five sources with $\theta>1.06$ , i.e., those that pass the criteria to be “resolved”, we also provide the position angle and radius of the de-convolved major axis provided by FIRST. Of these five sources, two have very low 1.4 GHz flux densities and therefore the uncertainty on $\theta$ is high. The other three have stronger evidence for being spatially resolved and are also “radio excess” sources; therefore, they are the strongest candidates for having radio jets on $\approx 3$ – $8^{\prime\prime}$ scales (i.e., $\approx 7$ –12 kpc; Table 2), although shocks due to outflows are another possible explanation for this excess radio emission (Zakamska & Greene 2014). High-resolution radio imaging will be instrumental in determining the origin and morphology of the radio emission in the whole sample.

3 Velocity definitions and spatially resolved kinematics

3.1 Non-parametric velocity definitions

We are interested in the spatially-resolved kinematics of the [O iii] $\lambda\lambda$ 4959,5007 and H $\beta$ emission lines for the sixteen type 2 AGN we observed with the Gemini-GMOS IFU. The combination of these emission-line species allows us to measure emission-line region sizes (Section 4.1), spatially resolve the ionised gas kinematics (Section 4.2) and calculate ionised gas masses (Section 4.3) for our targets. The emission-line profiles in type 2 AGN are often found to be complex, consisting of multiple narrow and broad components (e.g., Veilleux 1991; Greene et al. 2011; Villar-Martín et al. 2011b), which is also the case for our sample (Fig. 5; Fig. A1–A15). Due the complexity of the emission-line profiles we opted to use non-parametric definitions to characterise their widths and velocities within each pixel of the IFU data. We used an approach that: (1) is consistent for all of the targets in the sample; (2) allows us to de-couple galaxy dynamics, merger remnants and outflow features and (3) provides results that are directly comparable to observations in the literature. The definitions that we have used throughout, for describing both the galaxy-integrated emission-line profiles and the spatially resolved kinematics, are described below and are illustrated in Figure 4:

•

The peak velocity ( $v_{p}$ ) is the velocity of the peak flux density in the overall [O iii] $\lambda$ 5007 emission-line profile. For our targets, this traces the brightest narrow emission-line component (i.e., those with FWHM $\lesssim$ 500 km s*-1*).

•

The line width, $W_{80}$ , is the velocity width of the line that contains 80% of the emission-line flux such that $W_{80}=v_{90}-v_{10}$ , where $v_{10}$ and $v_{90}$ are the velocities at the 10th and 90th percentiles, respectively (Fig. 4). For a single Gaussian profile this is approximately the FWHM.

•

To measure the velocity offset of the broad underlying wings we define the velocity offset, $\Delta v$ , that is measured as $\Delta v=(v_{05}+v_{95})/2$ , where $v_{05}$ and $v_{95}$ are the velocities at the 5th and 95th percentiles of the overall emission-line profiles respectively (Fig. 4). For a profile that is well characterised by a single narrow Gaussian component and a luminous broad Gaussian component, $\Delta v$ is the velocity offset of the broad component. For a single Gaussian component $\Delta v=0$ .

•

To enable us to compare to previous work we also measured $v_{02}$ and $v_{98}$ , which are the 2nd and 98th percentiles of the flux contained in the overall emission-line profiles (as used in Rupke & Veilleux 2013)555We note that Rupke & Veilleux (2013) fit a broad and narrow component to the emission lines and then define $v_{98}$ for the broad component only. As a result their values are likely to be higher than our values that we defined from the overall emission-line profile so as not to make any assumptions on the kinematic structure of the emission lines (i.e., the number of distinct velocity components). and can be considered the “maximum” projected velocity of the gas.

•

Finally, we measured the median velocity ( $v_{\rm{med}}$ ; Fig. 4) of the overall emission-line profile so that we can compare to the type 2 AGN observed in Liu et al. (2013b). We find that $v_{\rm{med}}$ is likely to be dominated by galaxy kinematics in many cases (i.e., galaxy rotation or mergers; see Section 4.2.2) and should be used with care if interpreting outflow kinematics.

We emphasise that these definitions are used to enable us to consistently compare all of the targets in our sample as well as compare to observations in the literature; however, for a few individual cases some care should be taken when interpreting these measurements due to the uniqueness and complexity of the emission-line profiles. For this reason we discuss the details of kinematics of individual sources in Appendix A.

3.2 Emission-line profile fitting procedure

As several of the velocity definitions outlined above and illustrated in Fig. 4 are measured from the broad wings of the emission-line profiles they are subject to the influence of noise in the spectra, especially in the lower surface brightness regions of our IFU observations. To circumvent this issue, we characterise each emission-line profile using a combination of multiple Gaussian components. We note that we do not draw any physical meaning from the individual Gaussian components, they are only used to characterise the emission-line profiles. Below, we describe the emission-line profile fitting procedure that we use throughout for fitting the galaxy-integrated spectra and also the spectra used to produce the spatially resolved maps and velocity profiles described in Section 3.3.

Initially, we isolated the [O iii] $\lambda\lambda$ 4959,5007 emission-line doublet from the underlying continuum by linearly interpolating each spectrum between line-free regions close to the emission lines, a commonly used method for type 2 AGN (e.g., Husemann et al. 2013; Liu et al. 2013b). The emission-line doublet was then fit with a minimising- $\chi^{2}$ method, using the IDL routine mpfit (Markwardt 2009) with Gaussian components. For every Gaussian component we simultaneously fit the [O iii] $\lambda$ 5007 and [O iii] $\lambda$ 4959 emission lines using the same line-width, a fixed wavelength separation and the intensity ratio was fixed at [O iii] $\lambda 4959$ /[O iii] $\lambda 5007=0.33$ (e.g., Dimitrijević et al. 2007). Initially we fit a single Gaussian component to the emission-line profile and consequently additional Gaussian components were added to the overall fit (up to a maximum of six components) until the important values for our study (i.e., $\Delta v$ , $W_{80}$ and $v_{\rm{p}}$ ; see Fig. 4) became stable within $\leq$ 20%, i.e., the addition or removal of a Gaussian component does not change these values significantly. We only fitted more than one Gaussian component if the emission line was detected above a S/N threshold of $\approx$ 30, with this threshold being motivated by visually inspecting the fits and the reduced $\chi^{2}$ values (see below). For spectra below this S/N threshold, $W_{80}$ cannot be well characterised and we are also unable to define $\Delta v$ . We included appropriate constraints to the line widths to avoid fitting excessively broad components to the continuum or narrow components to noise spikes (i.e., limited by the spectral resolution). We verified that our fitting procedure was effective across the whole field-of-view by visually inspecting the spectra and their best fitting solutions extracted from 25 spatial regions for each object individually (Fig. 5; Fig. A1-A15). An additional verification of the success of our fitting procedure was that, in all but one of the targets,666The source with a higher median reduced $\chi^{2}$ =3.5 is SDSS J1356+1026 that we attribute to the highly complex emission-line profile and exceptionally low noise (Fig. 25). the median reduced $\chi^{2}$ values of the fits to the individual pixels in the data cubes are between 1.0–1.8 with $\geq$ 60% of the pixels having reduced $\chi^{2}$ values between 0.8–2.5.

Using the overall fits to the emission-line profiles, we measure the non-parametric velocity definitions defined in Section 3.1 (see Fig. 4). Random errors on these quantities were calculated by taking the 1 $\sigma$ spread of values calculated from a 1000 random spectra that were generated using the final models plus Gaussian random noise (based on the standard deviation of the continuum). To obtain the final quoted uncertainties, we added these random errors in quadrature with an estimated systematic error that we calculated from the range in the velocity values derived from adding or removing a Gaussian component to the final emission-line models. We note that these uncertainties are extremely small ( $\lesssim$ 10%) in the high signal-to-noise ratio spectra and therefore the uncertainties will be dominated by the physical interpretation of these values (see Section 4).

To enable us to measuring ionised gas masses (Section 4.3) we also measured the H $\beta$ emission-line profiles and fluxes from our IFU data. To do this, we extract the H $\beta$ emission-line profiles from the IFU data cubes in the same manner to that for [O iii] $\lambda$ 4959,5007 and show them in Figure 5 and Figure A1–A15. We find that the overall shapes of the H $\beta$ emission-line profiles are in excellent agreement with the [O iii] emission-line profiles for all but three of our sources (the exceptions are SDSS J1216+1417; SDSS J1338+1503; and SDSS 1504+0151). This indicates that for the majority of the sources, the H $\beta$ emission lines are not significantly affected by stellar absorption features (which is true for many AGN with luminous emission lines; Soto & Martin 2010; Harrison et al. 2012; Liu et al. 2013b) and this emission line traces the same kinematic components as the [O iii] emission line (see also Rodríguez Zaurín et al. 2013). For these thirteen sources we measure the H $\beta$ emission-line fluxes by fitting the [O iii] emission-line models (described above) to the data at the observed wavelength of the H $\beta$ emission lines, allowing only the overall normalisation to vary. The remaining three sources, which have different profiles for these two emission-line species, may have varying ionization states for different kinematic components and/or have higher levels of H $\beta$ stellar absorption features that is certainly plausible given the stellar continua and absorption features observed in their optical spectra (see Appendix A). For this work, we wish to avoid using a variety of approaches for our targets; therefore, we choose to not use the H $\beta$ emission lines in the analyses of these sources.

Due to the lack of clearly defined stellar absorption lines in the spectra across our targets, we define the systemic redshift from the velocity of the peak flux density in the [O iii] $\lambda$ 5007 emission line (see Fig. 5). This choice of systemic redshift effectively traces the mean velocity of the narrow components in the emission-line profiles, which are often attributed to galaxy kinematics (e.g., Greene & Ho 2005; Harrison et al. 2012; Rupke & Veilleux 2013; also see Section 4.2.1); however, this might not always be the case and therefore places an uncertainty on our systemic redshift. For SDSS J0958+1439 and SDSS J1316+1753 where the overall emission line profiles lack a clear single peak (see Fig. 14 and Fig. 21) we take the flux weighted mean velocity of the narrow components (i.e., those with FWHM $<$ 500 km s*-1*) used in the fit to the galaxy-integrated spectrum (i.e., similar to the procedure used in Rodríguez Zaurín et al. 2013). Across the whole sample, the mean velocity offset between our systemic redshift and the SDSS redshift is $-$ 18 km s*-1* with a scatter of 54 km s*-1*. We note that any velocity differences that are measured (such as $W_{80}$ ) are unaffected by the choice in systemic redshift.

3.3 Velocity maps and velocity-distance profiles

We significantly detect [O iii] emission to the edges of the IFU field-of-view in all sixteen sources (i.e., over $\geq$ 10–20 kpc; see discussion in Section 4.1). These extended emission-line regions enable us to trace the ionised gas kinematics over these scales. In Figure 5 we show example velocity maps of $v_{\rm{p}}$ , $W_{80}$ and $\Delta v$ (for the target J1010+1413) that were created by measuring these quantities across the field-of-view following the procedure outlined in Section 3.2 (see Fig. A1–A15 for these maps for the other targets). The spectra used to calculate these values are from averaging over 3 $\times$ 3 pixels (i.e., $\approx 0.^{\prime\prime}6$ ; comparable to the seeing of the observations) at the position of every pixel. At each position we measured the velocities from the fits to the [O iii] emission-line profiles but only if it was detected in the spectra with a signal-to-noise ratio $\geq$ 7. In Figure 5 we also show the peak signal-to-noise ratio [(S/N)peak] map which is the ratio of the peak flux density of the fitted [O iii] emission-line profile to the noise in the spectrum at each pixel position. These maps are discussed in Section 4.

To further aid our analyses we also produced velocity profiles along two axes through the IFU data cubes for each source (shown in Fig. 6 and Fig. A1–A15). We firstly defined a “major axis” for each source based upon the velocity peak maps ( $v_{\rm{p}}$ ; e.g., Fig. 5). The positional angle (PA) of the “major axis” for each source was chosen such that the velocity shear of the $v_{\rm{p}}$ map was the maximum. For a source where $v_{\rm{p}}$ traces galactic rotation this “major axis” corresponds to the kinematic major axis of the galaxy (for a discussion see Section 4.2.1). The second axis was taken perpendicular to the “major axis” (see e.g., Fig. 6). Using the PAs defined above we produced pseudo-long slit spectra with a slit-width of five pixels ( $\approx 0.8^{\prime\prime}$ –1*′′*). We obtained the spectra along each slit (from averaging over a length of 3 pixels) and calculated the velocity measurements outlined in Section 3.1. In these velocity profiles we also plot the ratio of the [O iii] to H $\beta$ emission line fluxes, plotting the 3 $\sigma$ lower-limit on this ratio when H $\beta$ is not detected above 3 $\sigma$ .

4 Results

Using Gemini-GMOS IFU data we study the spatially resolved properties of the [O iii] $\lambda\lambda$ 4959,5007 and H $\beta$ emission lines of sixteen luminous type 2 AGN (Table 2) chosen from our well constrained parent sample (described in Section 2.1; see Fig. 1 and Fig. 2). In Figure 5 and Figure A1–A15 we show the spatially-integrated [O iii] $\lambda$ 5007 and H $\beta$ emission-line profiles, the velocity maps and the peak [O iii] signal-to-noise ratio maps for all of the sources in our sample. In Figure 6 and Figure A1–A15 we show SDSS images and velocity-distance profiles and in Table 4 we tabulate the values that we derive from the galaxy-integrated spectra, the emission-line images and the velocity maps. In the following sub-sections we present our results of the sample as a whole, firstly presenting the sizes and morphologies of the emission-line regions (Section 4.1) before presenting the spatially resolved kinematics (Section 4.2). We give notes on the results from our IFU data for the individual sources in Appendix A.

4.1 Extended emission-line regions: sizes and morphologies

Although the ionised gas kinematics are the focus of this study, it is useful to make some initial comments on the emission-line region sizes and morphologies to compare to previous studies and to aid our later discussion. We detect [O iii] emission over physical spatial extents of $\geq$ 10–20 kpc (i.e., in some cases the [O iii] emission is extended beyond the IFU field-of-view). This is consistent with other studies that have shown that luminous AGN can have emission-line regions up to tens of kiloparsecs in size (e.g., Bennert et al. 2002; Humphrey et al. 2010; Greene et al. 2011; Schirmer et al. 2013; Harrison et al. 2012; Husemann et al. 2013; Liu et al. 2013a; Hainline et al. 2013). In Figure 7 we show the peak [O iii] signal-to-noise maps, for the whole sample, overlaid with contours showing the continuum-free [O iii] $\lambda$ 5007 images (i.e., wavelength collapsed images around the emission line). We fit single ellipses to the [O iii] emission-line images to calculate the semi-major axis ( $R_{{\rm[O~{}III]}}$ ; Table 4). The range of sizes we get from this method are $R_{{\rm[O~{}III]}}=1.5$ –4.3 kpc, similar to, but generally higher than, the ULIRG-AGN composites observed in Rodríguez Zaurín et al. (2013). We note that if we subtract the typical seeing of the observations (see Table 4), in quadrature, from these size measurements they are reduced by $\approx$ 20% with a worst-case scenario of $\approx$ 50% if the seeing reached $0.^{\prime\prime}9$ .777The one source that has a $R_{{\rm[OIII]}}$ value smaller than $0.^{\prime\prime}9$ (i.e., our most pessimistic value of the seeing across the whole sample) was taken under good conditions of $\approx 0.^{\prime\prime}6$ and is therefore spatially resolved. Based on a comparison to high-resolution imaging Rodríguez Zaurín et al. (2013) suggest that $R_{{\rm[O~{}III]}}$ is an over estimate of the emission-line region size for their sample; however, using the kinematics of our IFU data we commonly find spatially resolved emission over much larger extents (see Section 4.2 and Table 4) and therefore find that $R_{{\rm[O~{}III]}}$ is probably an underestimate of the size of the emission-line regions in our sources. Although we do not use identical methods, our observed sizes are comparable to, but towards the lower end of, those seen in the more luminous type 2 quasars in Liu et al. (2013a) as is expected by the [O iii] size-luminosity relationship (e.g., Bennert et al. 2002; Hainline et al. 2013).

In agreement with previous studies of radio-quiet AGN (e.g., Husemann et al. 2013; Liu et al. 2013a; Liu et al. 2014) we find that the emission-line regions in our sample are predominantly round or moderately elliptical (Fig. 7). However, we find tentative evidence that radio excess sources (i.e., those with the strongest evidence of radio emission above that expected from star formation) have a higher incidence of extended and irregular morphologies. It has previously been shown that radio-loud quasars have a higher incidence of irregular morphologies than their radio-quiet counterparts (Liu et al. 2013a and references there-in). Although the radio data presented here are of insufficient spatial resolution to establish a direct (or lack there-of) connection, our results may imply that the presence of radio-AGN activity, even at modest luminosities, is connected to the morphologies of their emission-line regions (see also Husemann et al. 2013 and Appendix A). Alternatively, the “radio excess” in these sources could be due to shocks that are also responsible for the irregular emission line regions that we observe in Figure 7 (see Zakamska & Greene 2014). Multi-frequency high-resolution radio imaging is required to determine the origin and morphology of the radio emission in these objects.

4.2 Spatially resolved ionised gas kinematics

Multiple kinematic components (i.e., multiple narrow and broad components) are seen in the galaxy-integrated emission-line profiles of the sources in our sample (Fig. 5; Fig. A1–A15). There are several possible origins for these kinematic components such as merging nuclei, merger remnants or merger debris, halo gas, rotating gas discs and outflowing or inflowing gas. Indeed, previous spatially-resolved spectroscopy of type 2 AGN has shown that multiple kinematic processes are likely to contribute to the overall emission-line profiles of ionised gas (e.g., Holt et al. 2008; Fischer et al. 2011; Villar-Martín et al. 2011a,b; Rupke & Veilleux 2013). By studying the spatially resolved velocity maps, velocity profiles and peak signal-to-noise maps from our IFU data (Fig. 5; Fig. 6; Fig. A1–A15) we can distinguish between these scenarios and characterise individual kinematic components. Although the focus of our study is to identify and characterise galaxy-wide ionised outflows, in order to isolate the kinematics due to outflows we must also investigate these other kinematic features. In Section 4.2.1 we give a qualitative description of the observed kinematics due to galaxy rotation and merger features, including candidate dual AGN (summarised in Table 4), and in Section 4.2.2 we describe the outflow kinematics. The ionised gas kinematics for individual sources are described in Appendix A. We note here that most of the analysis is described on the basis of inspecting our velocity maps and velocity-distance profiles; however, we checked that these maps are a reasonable representation of the data by inspecting the spectra extracted from several regions across the IFUs (these spectra are shown in Fig. 5 and Fig. A1–A15).

4.2.1 Ionised gas kinematics: galaxy rotation, mergers and dual AGN

Rotating gas discs will produce regular velocity fields, with a gradient of blue to red and relatively narrow emission-line profiles (i.e., $W_{80}<$ 600 km s*-1*). As discussed in Section 3.1 the emission-line peak velocity ( $v_{\rm{p}}$ ; see Fig. 4) predominantly traces the luminous narrow emission-line components in our data. We find regular $v_{p}$ velocity fields and velocity profiles, indicative of tracing galaxy rotation (e.g., Courteau 1997), for seven of the sixteen objects in the sample. We give these a classification of “R” in Table 4 and give a justification for this, for individual sources, in Appendix A. This interpretation of rotating discs is often supported by the kinematic “major axis” (defined in Section 3.3) being aligned with the morphological major axis in the disc-like galaxies revealed in the SDSS images and/or the optical continuum seen in our IFU data. Indeed, it has previously been observed that the narrow-components (or “cores”; i.e., with the broad-wings removed) of ionised emission-line profiles, can be good tracers of stellar dynamics (i.e., traces galactic rotation and/or stellar velocity dispersion) or molecular gas disks (e.g., Greene & Ho 2005; Barth et al. 2008; Barbosa et al. 2009; Rupke & Veilleux 2013; but see Zakamska & Greene 2014). Further confirmation that the $v_{p}$ velocity fields are tracing rotation would require detailed kinematic analysis (e.g., Courteau 1997; Krajnović et al. 2006) that is beyond the scope of this paper. In a further five sources we see a velocity gradient from blue to red in the $v_{p}$ velocity maps and profiles; however, the velocity fields are more irregular and therefore they may not be dominated by galaxy rotation but instead may trace kinematics due other processes such as merging components or outflows (given a classification “R/I” in Table 4; see Appendix A for discussion of individual cases).

We next consider kinematics due to mergers. Mergers are known to create extended emission-line regions due to companion galaxies or merger debris and have been previously been observed in spatially resolved spectroscopy of AGN (e.g., Villar-Martín et al. 2011a,b; Harrison et al. 2012). Furthermore, they may be expected to be common in our sample as $\gtrsim$ 60% of type 2 AGN at moderate redshifts $z\approx 0.3$ –0.4 show signs of morphological disturbance at some level in $HST$ imaging (Villar-Martín et al. 2012).888We note that Villar-Martín et al. 2012 refer to low-level morphological disturbances of several types based on the classifications of Ramos Almeida et al. 2011. The number of major mergers observed in AGN is much lower (e.g., Cisternas et al. 2011). The [O iii] emission-line peak signal-to-noise ratio maps (Fig. 7) allows us to search for faint spatially-distinct emission-line regions which could indicate companion/merging galaxies, merger debris, tidal features or halo gas. Based on these maps, five sources show spatially distinct emission-line regions with distinct velocities to the rest of the host galaxy. One of these has clear signs of major-merger activity based on the SDSS image and our IFU data (J1356+1026; Fig. 25), three are likely to be distinct emission-line regions (i.e., merger debris or halo gas; J0945+1737; Fig. 13; J1010+1413; Fig. 5 and J1000+1242; Fig. 15) and one could have several interpretations (J1316+1753; Fig. 21; these objects are discussed in detail in Appendix A).

Major mergers of galaxies naturally leads to the possibility of “dual” AGN; i.e., two nuclei which are gravitationally bound and are both actively accreting material. Systematic spectroscopic searches for AGN with luminous double peaked [O iii] emission-line profiles has become a popular method to identify candidate dual AGN (e.g., Wang et al. 2009; Smith et al. 2010; Liu et al. 2010; Ge et al. 2012; Comerford et al. 2013). However, these double peaks could also be due to other processes such as rotating gas disks, merging components or outflows, all illuminated by a single AGN (e.g., Fu et al. 2011; Fu et al. 2012; see Comerford et al. 2012 for a discussion). Two of our sources show luminous double peaks (J1316+1753 and J1356+1026) in their galaxy-integrated [O iii] emission-line profiles. Based on our data, the strongest case for a dual AGN is J1356+1026 (see also Greene et al. 2012), where we observe two distinct emission-line and continuum regions (see Appendix A for further discussion of these two sources).

4.2.2 Ionised gas kinematics: galaxy-wide outflows

In the IFU data of all sixteen targets we find broad [O iii] $\lambda\lambda$ 4959,5007 emission-line profiles across the field-of-view (i.e., $W_{{\rm 80,max}}\approx 720$ –1600 km s*-1*; Table 4) with a range of maximum projected velocity offsets ( $|\Delta v_{\rm{max}}|=110$ –540 km s*-1*; Table 4). Such broad (i.e., $W_{80}\gtrsim 600$ km s*-1*) and asymmetric [O iii] emission-line profiles, are difficult to explain with mergers or galaxy rotation, even for the most massive galaxies, and are attributed to high-velocity outflowing gas (e.g. Heckman et al. 1981; Veilleux et al. 1995; Greene & Ho 2005; Barth et al. 2008; Nesvadba et al. 2008; Greene et al. 2011; Westmoquette et al. 2012; Harrison et al. 2012; Liu et al. 2013b; Rupke & Veilleux 2013). As with all of these studies we favour outflow over inflow due to the exceptionally high velocities and line-widths that we observe. In the previous sub-section we isolated kinematic components due to other features (e.g., mergers and galaxy rotation) and we find that these have little effect on the observed kinematic structure of the line-width, $W_{80}$ and also in most cases that of the velocity offset, $\Delta v$ (the exceptions are J1356+1026 and J1316+1753 for which $\Delta v$ is affected, see Appendix A for details); i.e., these maps are dominated by outflow kinematics. We therefore use these maps to describe the outflow kinematic structures and morphologies. We note that Liu et al. (2013b) use the velocity structure of the median velocity ( $v_{\rm{med}}$ ; Fig 4), in which they predominantly see a blue to red velocity gradient, as part of their outflow analysis of their IFU data of $z=0.3$ –0.6 radio-quiet luminous type 2 AGN. We observe a very similar range in maximum line-widths and $v_{\rm{med}}$ maximum velocity gradients (i.e., $\Delta v_{{\rm med,max}}\approx 55$ –520 km s*-1*; Table 4) to the objects in Liu et al. (2013b). However, by spatially resolving $v_{\rm{med}}$ we find that this quantity is predominantly dominated by the kinematics of the $v_{p}$ velocity field (Fig. 6; Fig A1–A15) and consequently $v_{\rm{med}}$ can often be dominated by kinematics due to galaxy rotation and mergers (see Section 4.2.1). Therefore, we favour $\Delta v$ over $v_{\rm{med}}$ to describe the projected outflow velocity offset for the rest of our analysis.

For an outflow consisting of ionised clouds (observable in [O iii] emission) embedded in a bulk outflow (e.g., Heckman et al. 1990; Crenshaw & Kraemer 2000) the projected velocity offsets will be sensitive to inclination effects while the emission-line widths (i.e., $W_{80}$ ) are more likely to reflect typical bulk motions and be less sensitive to inclination (also see discussion in Harrison et al. 2012). We therefore concentrate on the $W_{80}$ maps to investigate the sizes and morphologies of the outflows (Fig. 8), comparing them with the $\Delta v$ maps when appropriate (discussed in Appendix A for individual objects). The observed outflow velocities and morphologies will also depend on other factors. For example, an outflow with the far side obscured by dust can explain the excess of blue wings in [O iii] emission-line profiles (e.g., Heckman et al. 1981; Vrtilek 1985; Veilleux 1991) while, outflowing redshifted components can be explained if the outflow is highly inclined and/or is extended beyond the likely obscuring material (e.g., Barth et al. 2008; Crenshaw et al. 2010; Greene et al. 2012; Rupke & Veilleux 2013). We observe both red and blue velocity offsets in our sample (Table 4). A range of structures of outflows have been observed in galaxies both with and without powerful AGN, including spherically-symmetric outflows and bi-polar superbubbles (e.g., Veilleux et al. 1994; Greene et al. 2012; Liu et al. 2013b; Rupke & Veilleux 2013). Detailed modelling of the outflow kinematics of individual objects is beyond the scope of this paper; however, we briefly describe the morphology and kinematics of the outflows that we observe in our sample below (we describe individual objects in more detail in Appendix A).

We define the quantity $D_{600}$ (values in Table 4) which is the maximum projected spatial extent of the broad [O iii] emission-line profiles (with $W_{80}>600$ km s*-1*; i.e., where the kinematics are likely to be dominated by outflows) based upon the velocity profiles (Fig. 6 and Fig. A1–A15). The outflow features are located over large extents in all sources ( $D_{600}\gtrsim$ [6–16] kpc; Table 4) with the spatial extents being mostly lower limits due to the limited field-of-view of our observations. We note that in most cases the broad [O iii] emission is clearly spatially extended, due to the morphological or kinematic structure that we observe (see below) and in all cases it is extended beyond the seeing; however, a conservative uncertainty on these $D_{600}$ measurements would be $\approx 2\times$ seeing (see Table 4). Although our sample was initially selected to have luminous and broad [O iii] emission-line components (see Section 2.1), these results demonstrate that when these components are seen in the one-dimensional spectra of luminous type 2 AGN (as is the case for at least 45 $\pm$ 3% of objects; see Section 2.1 and Fig. 2), they are always found over galaxy-wide scales when observed with spatially resolved spectra (i.e., over at least a few kiloparsecs). To assess how confident we can be of the fraction of extended ionised outflows within our selection criteria we ran Monte Carlo simulations. To do this, we took 246 mock galaxies (i.e., the number from the parent catalogue that met our selection criteria; Section 2.1), assigned them as either “extended” or “not-extended” (in various ratios) and randomly drew 16 objects from this sample 104 times. We found that, at most, 30% of the sources can be “not-extended” in the parent population for the probability of selecting 16 “extended” sources by random chance to be $\leq$ 0.3% (i.e., the 3 $\sigma$ confidence level). In other words, when broad and luminous emission-line components are observed in one-dimensional spectra (i.e., when $\gtrsim$ 30% of the ionised gas appears to be outflowing; Section 2.1) of luminous ( $L_{{\rm[O~{}III]}}>5\times 10^{41}$ erg s*-1*) $z<0.2$ type 2 AGN, we can be confident, with 3 $\sigma$ significance, that in at least 70% of instances these features are extended on a few kiloparsec scales or greater.

In Figure 8 we show the $W_{80}$ maps for the whole sample to illustrate the morphology of the outflows. From this figure we can see a large diversity in the morphology. Several of the sources show a preferential axis, which is orientated away from the galaxy “major axis” (albeit not always perpendicular); however, in some cases we require a stronger constraint on the major axis of the host galaxies to confirm the relative orientations. This is consistent with the collimated winds emerging along the minor axis seen in star-forming galaxies (e.g., Heckman et al. 1990; Rupke & Veilleux 2013). The most impressive outflows that we observe are those which have the kinematic features of $\approx$ 10–15 kpc scale super-bubbles; i.e., high-velocity bi-polar emission-line regions each side of the nucleus. Combining our IFU data with archival imaging and spectroscopic data (see Appendix A) we identify three super-bubble candidates (one was previously known: J1356+1026; Greene et al. 2012 and, as far as we are aware, two are new: J1000+1242 and J1430+1339; see Appendix A). In two cases (J1356+1026 and J1430+1339) we only observe the base of these bubbles and therefore require IFU observations with a larger field-of-view to fully characterise the kinematics of the outflow structure. In the third case (J1000+1242) we see a clear blue and red sided outflow. Similar outflows have been seen in galaxies both with and without luminous AGN (e.g., Veilleux et al. 1994; Greene et al. 2012; Rupke & Veilleux 2013).

In four of our targets the morphology of the outflows are circular (Fig. 8) with reasonably constant $W_{80}$ and $\Delta v$ values seen across the extent of the emission-line regions. This kinematic structure is seen in the majority of the more luminous type 2 AGN in Liu et al. (2013b) who argue that these structures are characteristic of spherically symmetric, or wide-angle bi-conal outflows, and a power-law [O iii] surface-brightness profile (see Liu et al. 2014 for similar observations of type 1 AGN). We note that circular morphologies could (at least partially) be an artifact of the fitting routine since in areas of low surface brightness only single Gaussian components are fit and we are not able to fully characterise the emission-line profile (see Section 3.2); however, this not likely to be the full explanation for the observed morphologies (see Appendix A; also see Fig. 5 of Liu et al. 2013b). Finally, in one source (J1355+1300) we observe a spatially distinct high-velocity region, offset from the nucleus, that has a few possible interpretations including a small-scale outflowing bubble (see Appendix A for a discussion).

4.3 Outflow properties: estimates of mass, energy and momentum

The mass, energy and momentum being carried by galaxy-wide outflows are important quantities to constrain in order to understand the role that outflows play in galaxy formation (discussed in Section 5.2) and to also help determine the physical mechanisms that are driving them (discussed in Section 5.1). Outflows are likely to be entraining gas in multiple phases (i.e., ionised, molecular and neutral) and multiple gas phase observations are required to fully characterise the outflow properties (e.g., Shih & Rupke 2010; Mahony et al. 2013; Rupke & Veilleux 2013). However, the warm ionised gas, which we observe in our observations, provides initial information on the outflows in this sample and could represent a large fraction of the overall mass and energy of the total outflows, as has been shown for some AGN (e.g., Rupke & Veilleux 2013). The kinematic structures of the outflows across our sample are diverse and complex (see Section 4.2.2) which makes accurate characterisation of the outflows challenging without doing detailed kinematic modelling of individual systems. However, as is often invoked throughout the literature, we can apply some simple outflow models to the whole sample to provide first order constraints on the mass, energy and momentum involved in the outflows and to enable a direct comparison to other studies. As most of the calculations outlined below require measurements of the H $\beta$ luminosity we do not include the three sources where we did not make this measurement (see Section 3.2).

An important quantity for measuring ionised gas masses, and hence mass outflow rates, is the electron density ( $n_{e}$ ). This quantity is often measured from the emission-line ratio [S ii] $\lambda$ 6716/ $\lambda$ 6731 which is sensitive to electron density (e.g., Osterbrock & Ferland 2006). The [S ii] doublet is not covered in our IFU observations; however, using single-component Gaussian fitting to the SDSS spectra we find that the emission-line ratios for our sources are in the range [S ii] $\lambda$ 6716/ $\lambda$ 6731 $=$ 0.8–1.2, corresponding to a range in $n_{e}\approx 200$ –1000 cm*-3* with a median of $n_{e}=500$ cm*-3* (Osterbrock & Ferland 2006; assuming $T_{e}=10^{4}$ K). This range of $n_{e}$ is in agreement with that previously found for type 2 AGN (e.g., Greene et al. 2011). We note that these are average electron densities within the SDSS fibre and the values for the outflowing kinematic components are unknown and have the potential to be outside of the electron density range that the [S ii] $\lambda$$\lambda$ 6716,6731 doublet is sensitive to (i.e., $n_{e}\gtrsim 10^{4}$ cm*-3*; Osterbrock & Ferland 2006; see discussion in Holt et al. 2011 and Rodríguez Zaurín et al. 2013), providing a source of uncertainty in the mass and energy outflow rates in all studies that use this method. However, we note that Greene et al. (2011) provide convincing arguments that the emission line regions in their $z<0.5$ radio-quiet type 2 AGN are likely to be clumpy and that standard mass estimates (such as those we derive below) are likely to be under-estimated.

The observed ionised gas mass (i.e., the gas which is emitting in H ${\beta}$ ) can be estimated assuming purely photoionized gas with “Case B” recombination (with an intrinsic line ratio of H $\alpha$ /H $\beta=$ 2.9) and an electron temperature of $T=10^{4}$ K, following Nesvadba et al. (2011) and Osterbrock & Ferland (2006), using the relation:999To be consistent with Liu et al. (2013b), Equation (2) is given in the exact form that is provided in Nesvadba et al. (2011). However, we note that if you follow section 4.2 in Osterbrock & Ferland (2006) with the appropriate value for the recombination coefficient of H $\beta$ for the conditions described (i.e., $\alpha_{{\rm H}\beta}^{{\rm eff}}=3.03\times 10^{-14}$ cm3 s*-1*) the normalisation constant in Equation 2 is $6.78\times 10^{8}\,\hbox{M$ {}_{\odot} $}$ . This factor of four difference to the quoted normalisation constant has no significant bearing on our order of magnitude estimates of the gas mass.

[TABLE]

This relationship between the observed line-emitting gas mass, $n_{e}$ and $L_{{\rm H}\beta}$ (or an equivalent using $L_{{\rm H}\alpha}$ ) is commonly adopted in studies of outflows (e.g., Holt et al. 2006; Genzel et al. 2011; Rodríguez Zaurín et al. 2013; Liu et al. 2013b; Schnorr-Müller et al. 2014) with normalisation factors that vary within a factor of a few, depending on the exact assumptions. Using our extinction un-corrected $L_{{\rm H}\beta}$ values (Table 2), we obtain total observed ionised gas masses of $M_{{\rm gas}}=$ (2–40) $\times 10^{7}$ M*⊙, assuming the commonly adopted $n_{e}=100$ cm-3* (e.g., Liu et al. 2013b), or $M_{{\rm gas}}=$ (0.4–8) $\times 10^{7}$ M*⊙* assuming our median value of $n_{e}=500$ cm*-3*. We note that the ionised gas phase will make up only a fraction of the total gas content. For example, taking a typical SFR for our targets (40 M*⊙* yr*-1*) and assuming a star forming region of $\approx 1$ kpc would imply molecular gas masses of $\approx 10^{10}$ M*⊙* (Genzel et al. 2011)

The simplest method that we can use to estimate the total kinetic energy in the outflows is the following,

[TABLE]

Following Liu et al. (2013b) we assume $n_{e}=100$ cm*-3* and that all of the observed ionised gas is involved in the outflow101010As discussed in Section 4.2.1 not all of the ionised gas appears to be involved in an outflow. The broad kinematic components (FWHM $>$ 700 km s*-1*) in the [O iii] and H $\beta$ emission lines typically contribute 30–80% of the total flux (based on Mullaney et al. 2013) which may give an estimate of the fraction of gas involved in the outflows. However, the emission-line ratios of up to H $\alpha$ /H $\beta=8$ (from Mullaney et al. 2013) implying that the intrinsic $L_{{\rm H}\beta}$ values (and hence total calculated masses) may be up to a factor of 10 times higher than the observed values (following Calzetti et al. 2000). Therefore, these two effects will cancel each other out to some level and this assumption is sufficient for our order-of-magnitude estimates with a bulk-outflow velocity $v_{{\rm gas}}=W_{80}/1.3$ (suitable for the spherically symmetric or wide-angle bi-cone outflow models outlined Liu et al. 2013b; see also Harrison et al. 2012 for different arguments on using line-widths to estimate the bulk velocities). Across the sample this leads to total kinetic energies of $E_{{\rm kin}}=(0.5$ – $50)\times 10^{56}$ erg. Assuming the typical maximum radial extent of the outflows is 6 kpc (see Table 4) and a continuous outflow, this implies outflow lifetimes of $t_{{\rm out}}\approx 6$ kpc $/v_{{\rm gas}}=$ (5–11) Myrs and consequently outflow kinetic energy rates of $\dot{E}_{{\rm kin}}=(0.1$ – $30)\times$ 1042 erg s*-1*.

An alternative approach is to assume a spherical volume of outflowing ionised gas (following e.g., Rodríguez Zaurín et al. 2013; see also Holt et al. 2006), which gives mass outflow rates of

[TABLE]

and outflow kinetic energy rates of

[TABLE]

where $\sigma$ is the velocity dispersion and $v_{{\rm out}}$ is the outflow velocity. To match Rodríguez Zaurín et al. (2013) as closely as possible111111We note that Rodríguez Zaurín et al. (2013) use a different definition of $v_{{\rm out}}$ to $\Delta v$ ; however, the two methods agree within a factor of $\approx$ (1–3) with a median of 1.1. Likewise, their definition for $\sigma$ agrees with $\sigma=W_{80}/2.355$ within a factor of $\approx$ (1–2) and median of 1.2. we take the outflow velocity to be $v_{{\rm out}}=\Delta v$ , the velocity dispersion to be $\sigma=W_{80}/2.355$ , the outflow radius to be $r=R_{{\rm[O~{}III]}}$ (Table 4) and again use $n_{e}=100$ cm*-3*. Using this approach (we will call “Method 1”) we obtain $\dot{M}_{{\rm out,M1}}=3$ – $70$ M*⊙* yr*-1and $\dot{E}_{{\rm out,M1}}=$ (0.3–30) $\times$ 1042* erg s*-1*. This range in outflow kinetic energy rate is in excellent agreement with the range of values using the previous method, which we will now ignore, and the two methods agree within a factor $\leq 3$ for all sources. Both of these methods provide strict lower limits, as we are only observing the line-emitting gas and hence the total mass involved in the outflow could be much higher (see also e.g., Greene et al. 2011; Liu et al. 2013b).

We also consider the mass and energy injection rates assuming an energy conserving bubble in a uniform medium (e.g., Heckman et al. 1990; Veilleux et al. 2005; Nesvadba et al. 2006 and references there-in), which gives the relation:

[TABLE]

where $r_{10}$ is taken to be half the extent of the observed broad [O iii] emission (in units of 10 kpc; we assume $r_{10}=0.6$ for all sources; Table 4), $v_{1000}$ are the velocity offsets in units of 1000 km s*-1* (we assume $v_{1000}=W_{80}/1300$ ; Table 4) and $n_{0.5}\approx 1$ , is the ambient density (ahead of the expanding bubble in units of 0.5 cm*-3*). Using this method (we will call “Method 2”) we obtain values of $\dot{E}_{{\rm out,M2}}\approx(0.7$ – $7)\times 10^{45}$ erg s*-1*. The mass outflow rates are then given by $\dot{M}_{{\rm out,M2}}=2\dot{E}_{{\rm out,M2}}/v_{c}^{2}\approx(9$ – $20)\times 10^{3}$ M*⊙* yr*-1*.

We consider “Method 1” and “Method 2” to be lower and upper bounds of the mass outflow rates and kinetic energy rates (see also discussion in Greene et al. 2012) and the fiducial values we will use throughout are the mean (in log space) of these two values which are $\dot{E}_{{\rm out}}=(0.15$ – $45)\times 10^{44}$ erg s*-1* and $\dot{M}_{{\rm out}}=170$ – $1200$ M*⊙* yr*-1*. Comparing these values to those found using the approach followed in Liu et al. (2013a,b) provides confirmation that using these fiducial values are reasonable. They argue that the NLR in luminous type 2 AGN can be well described by relatively dense clouds embedded in a hot, low-density, volume-filling wind. A spatial profile with a constant [O iii]/H $\beta$ emission-line ratio followed by a sharp decline in this value indicates the break where the clouds become optically thin to ionizing radiation (see full description in Liu et al. 2013b). In four of our sources (these are J0958+1439; J1100+0846; J1316+1753; J1339+1425; see Appendix A) we see the same sharp decline in the [O iii]/H $\beta$ ratio (in most of the other sources this transition may happen beyond our field-of-view). For these four sources, the kinetic energies and mass outflow rates using their method (see Appendix A) are all within a factor of $\approx$ 1–3 of the fiducial values we chose to adopt here.121212Although here we use a different method to Liu et al. 2013b we verified our results using a similar method to these authors using iraf ellipse.

Finally, in preparation for our discussion in Section 5.1, we estimate the mean and upper and lower bounds of outflow momentum rates by taking the mass outflow rates calculated above and assuming ${\dot{P}}_{{\rm out}}={\dot{M}}_{{\rm out}}v_{{\rm gas}}$ .

5 Discussion

We have presented our IFU data for a sample of sixteen luminous type 2 AGN which traces the ionised gas kinematics (via [O iii] and H $\beta$ emission lines) over kiloparsec scales. We have decoupled the kinematics of outflows from other kinematic components, such as galaxy dynamics and mergers, and show that energetic, galaxy-wide ionised outflows are ubiquitous in our sample. Our sample is selected from a well constrained parent sample so that we can place our observations into the context of the overall AGN population. In this section we use our observations, combined with previous observations from the literature, to discuss the likely driving mechanisms of the outflows (Section 5.1) and the role that these outflow may play in galaxy evolution (Section 5.2).

5.1 What drives the outflows?

The dominant processes that drive galaxy-wide outflows in massive galaxies and the efficiency to which they are able couple to the gas are currently sources of uncertainty in galaxy formation models. Several possible mechanisms have been suggested to drive galaxy-wide outflows, for example: the mechanical input from stellar winds and supernovae (e.g., Leitherer et al. 1999; Heckman et al. 1990); direct radiation pressure from the luminosity of an AGN or star formation (e.g., Murray et al. 2005); the interaction of radio jets (launched by the AGN) with a clumpy and multiphase ISM (e.g., Sutherland & Bicknell 2007; Wagner et al. 2012); and AGN winds, initially launched at the accretion disk, that propagate through the galaxy and sweep up the ISM (e.g., King et al. 2011; Faucher-Giguère & Quataert 2012). In this subsection we will investigate which of these processes could be responsible for driving the outflows observed in our targets.

5.1.1 Morphology and structure

A possible method to distinguish between an AGN-driven and star-formation driven outflow could be the morphology and structure of the outflow. Galactic-scale star-formation driven outflows are known to propagate along low density regions, perpendicular to galaxy discs (e.g., Heckman et al. 1987; Ohyama et al. 2002; Veilleux et al. 2005). In contrast, outflows driven by an AGN on small scales have no relation between the orientation of the outflow and the disk in the host galaxy (e.g., Fischer et al. 2013). However, if AGN-driven outflows propagate to galaxy-wide scales they may also be forced to propagate away from the galaxy disk along a path of least resistance (e.g., Faucher-Giguère & Quataert 2012; Gabor & Bournaud 2014) which could make galactic-scale outflows driven from a nuclear star forming region or an AGN indistinguishable (see also Hopkins & Elvis 2010; Heckman et al. 2011 and Diamond-Stanic et al. 2012 for other reasons why it could be difficult to distinguish between AGN and star formation driven outflows). In some observations compact radio jets ( $\lesssim$ few kiloparsec in size) are inferred to play a role, where the combination of high resolution radio data and spatially resolved spectroscopy has shown that high velocity gas (in multiple phases) is spatially co-incident with radio emission (Baldwin et al. 1980; Whittle et al. 1988; Capetti et al. 1999; Whittle & Wilson 2004; Morganti et al. 2005, 2013; Emonts et al. 2005; Leipski & Bennert 2006; Barbosa et al. 2009; Fu & Stockton 2009; Storchi-Bergmann et al. 2010; Shih et al. 2013). However, even when radio jets are present it appears that they may not always be fully responsible for the observed outflows and other processes are also required (e.g., Villar-Martín et al. 1999; Rupke & Veilleux 2011; Riffel et al. 2014).

The morphologies of the outflows in our targets are illustrated in Figure 8. In our sample we observe no obvious relationship between the morphology of the observed outflows and the SFRs, AGN luminosities or radio excess parameters of the host galaxies; i.e., irrespective of the host galaxy or AGN properties we see a range in outflow morphology including those which are escaping away from the galaxy discs. This is in broad agreement with Rupke & Veilleux (2013) who find no structural difference between the outflows in the ULIRGs with and without luminous AGN. Interestingly, all of superbubble candidates are radio excess sources (see Section 4.2.2 and Appendix A). However, deeper and higher-resolution radio data are required to search for small and low luminosity radio jets, which we cannot identify in the FIRST data, and to ultimately determine the origin of the radio emission (see Section 2.4).

5.1.2 Outflow velocities compared to AGN and galaxy properties

Another possible method to determine the driving mechanism of outflows is to search for trends between outflow properties and properties of the host galaxy or AGN. Positive correlations have been claimed to exist between outflow properties, albeit using different observations, for each of SFR (e.g., Rupke et al. 2005; Martin 2005; Bradshaw et al. 2013), AGN luminosity (e.g., Cicone et al. 2014) and radio luminosity (e.g., Nesvadba et al. 2011). Of course these analyses are complicated by the fact that AGN luminosity, SFR and radio luminosity are closely related (e.g., Tadhunter et al. 1998; Ivison et al. 2010; Mullaney et al. 2012; Chen et al. 2013) and that small duty cycles may mean that an AGN-driven outflow may be observed when the AGN itself is no longer observable (e.g., King et al. 2011; Gabor & Bournaud 2014).

In Fig. 9 we show that the line widths ( $W_{80}$ ) and maximum velocities ( $v_{02}$ ) of our sample are comparable to those found in more luminous type 2 AGN (Liu et al. 2013b) and ULIRG-AGN composite galaxies (Rodríguez Zaurín et al. 2013).131313Although Rodríguez Zaurín et al. (2013) do not quote the values of $W_{80}$ and $v_{02}$ directly, we reconstructed the emission-line profiles (from the Gaussian components given in their paper) and then measure these values directly. We observe no clear trends between $v_{02}$ or $W_{80}$ and $L_{\rm{[O~{}III]}}$ , despite four orders of magnitude in [O iii] luminosity. Additionally, although limited in dynamic range, we do not see any obvious correlation in $W_{80}$ or $v_{02}$ with radio luminosity (Fig. 9). These conclusions are consistent with Mullaney et al. (2013) who demonstrate that the broadest emission lines are prevalent for AGN with radio luminosities of $L_{1.4\rm{GHz}}\gtrsim 10^{23}$ W Hz*-1* (i.e., the range that is covered in Fig. 9), but found no clear trends with [O iii] luminosity when taking radio-luminosity matched samples of AGN. We note that Figure 1 is not radio-luminosity matched so gives an apparent positive trend between [O iii] luminosity and line width; see Mullaney et al. (2013) for details.

In Figure 10 we compare the galaxy-integrated line-widths ( $W_{80}$ ) and maximum velocities ( $v_{02}$ ) to the infrared luminosities from star formation ( $L_{{\rm IR,SF}}$ ; a proxy for SFR), bolometric AGN luminosities ( $L_{{\rm AGN}}$ ) and the radio excess parameter ( $q_{IR}$ ) derived in Section 2.4. We also compare to the ULIRG-AGN composite galaxies from Rodríguez Zaurín et al. (2013)141414We fit the SEDs to the IRAS and WISE data for these sources in a consistent manner to that described in Section 2.4; however, we only accept AGN components if the AGN template contributes $\geq$ 50% of the flux at 19 $\mu$ m due to the difficulty in reliably identifying low-luminosity AGN in ULIRGs using this method (see Del Moro et al. 2013). Eight of these ULIRGs also have AGN luminosities derived from Spitzer infrared spectra in Veilleux et al. (2009) that we use in favour of our AGN luminosities. We found that the AGN luminosities and IR luminosities are well matched (within a factor of $\lesssim$ 2) between our method and that in Veilleux et al. (2009). and the $z<0.06$ ULIRGs from Rupke & Veilleux (2013) (who make these velocity measurements from various ionised emission-line species). In Figure 10, the only possible trend we observe is between the maximum velocity ( $v_{02}$ ) and AGN luminosity, although we are limited in measurements below $L_{{\rm AGN}}=10^{45}$ erg s*-1*. Based on these samples, although high-velocity outflow features are observed across the samples, the most extreme velocities (i.e., $v_{02}\gtrsim 2000$ km s*-1*) and line-widths (i.e., $W_{80}\gtrsim$ 1500 km s*-1*) appear to be predominantly found in the quasar-ULIRG composite galaxies (Fig. 10). This is in broad agreement with other studies of the outflows in ULIRGs and quasars, in multiple different gas phases, that find the most extreme outflows in the quasar-ULIRG composites (e.g., Harrison et al. 2012; Westmoquette et al. 2012; Veilleux et al. 2013; Cicone et al. 2014; Hill & Zakamska 2014). Further evidence that the incidence of extreme ionised outflows in quasar-ULIRG composite galaxies is high, compared to the overall population, is that 16/17 of the objects in Rodríguez Zaurín et al. (2013) have extreme ionised gas kinematics (i.e., significant broad [O iii] emission-line components) which is much larger than we find for the overall optical AGN population (see Section 2.1), and considerably higher than expected given their moderate [O iii] luminosities (i.e., $L_{{\rm[O~{}III]}}\lesssim 10^{42}$ erg s*-1*; Fig. 1). Furthermore, the flux-weighted average line-widths (FWHM ${}_{{\rm avg}}\gtrsim 800$ km s*-1*) of these quasar-ULIRG composites are representative of the top $\approx 2$ –3% most extreme objects in Mullaney et al. (2013). This result is not sufficient to establish which process is driving the observed outflows, but instead may demonstrate that the most extreme outflows are most prevalent during an active star formation and AGN phase as is predicted by some evolutionary models (e.g., Sanders et al. 1988; Hopkins et al. 2008).

Finally, we note that Mullaney et al. (2013) show that the fraction of sources with their ionised gas dominated by outflows, was considerably higher for sources with moderate-to-high radio luminosities ( $L_{{\rm 1.4\,GHz}}\gtrsim 10^{23}$ W Hz*-1*). Based on the analysis of the radio emission in our sources and the comparison samples (Section 2.4; Figure 10), star formation could contribute considerably to the radio emission. Additionally, as noted by Zakamska & Greene (2014), radio emission could be produced by shocks due to outflows that were driven by e.g., quasar winds. Further analysis of the parent population is required to establish the physical process that is driving the trend between radio luminosity and ionised outflows.

5.1.3 Coupling efficiencies and momentum fluxes

A popular method to investigate the likely drivers of galactic-scale outflows is to compare the energy and momentum available from each potential driver (i.e., supernovae and stellar winds, radiation pressure [from AGN or star formation], AGN winds, or radio jets) to the kinetic energy and momentum in the outflows. Although we are limited by a large uncertainty in the outflow kinetic energies (see Section 4.3), we have measured SFRs and AGN luminosities and have identified the sources that are most likely to host radio jets (see Section 2.4 and Table 2). In Figure 11 we compare the ratio of our outflow kinetic energy rates ( ${\dot{E}}_{{\rm out}}$ ) with (1) the bolometric AGN luminosities; (2) the infrared star-formation luminosities151515We note that for infrared luminous galaxies (i.e., $L_{{\rm IR}}\geq 10^{11}$ L*⊙*), the infrared luminosity makes up the bulk of the bolometric luminosity; e.g., Sanders & Mirabel 1996; Veilleux et al. 2005. and (3) the mechanical radio jet power (estimated using the 1.4 GHz luminosity and the relationship in Cavagnolo et al. 2010). In Figure 12 we compare the momentum rates of the outflows ( ${\dot{P}}_{{\rm out}}$ ) with the momentum flux output radiatively by (1) star formation (i.e., $L_{{\rm IR,SF}}/c$ ) and (2) the AGN (i.e., $L_{{\rm AGN}}/c$ ). Using these results we now explore the plausibility of different potential driving mechanisms and compare to theoretical predictions.

Figure 11 shows that a similar outflow coupling efficiency is required to the radiative output of both the AGN (i.e., $\approx 0.5$ –10%) and star formation (i.e., $\gtrsim 0.5$ –40%), indicating that either of these processes could power the outflows if they are able to couple their energy to the gas. One way for star formation to drive an outflow is from stellar winds and supernovae. An estimate of the maximum energy injection from stellar winds and supernovae, which is used throughout the literature, is $\approx 7\times 10^{41}$ (SFR/ M*⊙* yr*-1*) erg s*-1* (applicable for stellar ages $\gtrsim$ 40 Myrs and following Leitherer et al. 1999; Veilleux et al. 2005).161616If we were to follow instead Dalla Vecchia & Schaye 2008 to estimate this energy input, the values would be a factor of $\approx$ 2 lower. This corresponds to a maximum ${\dot{E}}_{{\rm out}}/L_{{\rm IR,SF}}\approx 2$ % (following Kennicutt 1998 but correcting to a Chabrier IMF) that we indicate Figure 11. Based on these assumptions, stellar winds and supernovae are unlikely to be fully responsible for the observed outflows. If we instead consider a momentum driven wind with momentum deposition from the radiation pressure of stars, we expect ${\dot{P}}_{{\rm out}}=\tau L_{{\rm SF,IR}}/c$ (following Murray et al. 2005) that based on our calculations would require unrealistically high optical depths (i.e., $\tau\gg 1$ ; see Figure 12). Further investigation of the possibility that star formation drives the outflows could be achieved using high resolution optical-continuum imaging to measure the SFR surface densities in our targets (e.g., Murray et al. 2011; Heckman et al. 2011; Diamond-Stanic et al. 2012; Förster Schreiber et al. 2013).

The estimated coupling efficiencies between the bolometric AGN luminosity and the outflows (i.e., $\approx 0.5$ –10%) are consistent with those predicted by various models (e.g., Di Matteo et al. 2005; Hopkins & Elvis 2010; Zubovas & King 2012; Debuhr et al. 2012). However, we note that these coupling efficiencies do not tell us how the AGN couples to the gas. It is thought that direct radiation pressure from an AGN is unlikely to drive momentum-driven galaxy-wide outflows (e.g., King et al. 2011; Debuhr et al. 2012) and this is supported by the high outflow momentum rates we observe (Figure 12). However, it has been predicted that for high-accretion rate AGN, an accretion-disc wind could sweep up material and become a galaxy-wide energy-driven outflow (e.g., King et al. 2011; Faucher-Giguère & Quataert 2012; Debuhr et al. 2012; Zubovas & King 2012). These models predict momentum rates of ${\dot{P}}_{{\rm out}}\approx 10$ – $40L_{{\rm AGN}}/c$ on $\approx$ 1–10 kpc scales that is in good agreement with our observations (Figure 12) as well as multi-gas-phase outflow observations of other AGN (e.g., Cicone et al. 2014; Zubovas & King 2012 and references there-in). An AGN could also couple to the gas in their host galaxies though the interaction of a radio jet with the ISM. In Figure 11 we show that the coupling efficiencies between the radio jet power and the outflows is estimated to be $\approx$ 20–160% (calculated for the sources with the strongest evidence for radio AGN activity only; see Section 2.4). These basic calculations may rule out radio jets being solely responsible for all of the outflows observed (Fig. 11; but see Appendix A for a discussion on the radio jets in individual objects); however these numbers are uncertain and reasonably high coupling efficiencies are predicted for outflows driven by jet-ISM interactions (e.g., Wagner et al. (2012) predict $\approx 10$ –40%).

In summary, based on our analyses we find no definitive evidence that the outflows we observe are universally driven by the AGN (either launched by accretion-disk winds or radio jets) or by star formation. However, using the arguments that we have presented we find that supernovae, stellar winds, radio jets or radiation pressure are unlikely to be solely responsible for all of the outflows that we observe. Although uncertain, we have shown that the outflows we observe have properties that are broadly consistent with predictions of energy-conserving outflows that are initially driven by accretion-disc winds. It is certainly viable that multiple processes contribute to driving the observed outflows as has been seen in local systems, where star-formation winds, nuclear winds (potentially AGN-driven) and radio jets all are playing a role (e.g., Cecil et al. 2001; Storchi-Bergmann et al. 2010; Rupke & Veilleux 2011; also see discussion on our individual targets in Appendix A). As has been highlighted throughout this section, multi-phase (i.e., ionised, atomic and molecular) observations of the outflows, combined with high resolution multi-wavelength imaging will provide the required information to determine the relative role of these different outflow driving mechanisms in our targets.

5.2 What role do these outflows play in galaxy evolution?

Galaxy-wide outflows are required in galaxy formation models, driven by both star formation and AGN activity, to reproduce the observable properties of galaxies and the intergalactic medium (e.g., Silk & Rees 1998; Di Matteo et al. 2005; Hopkins et al. 2008; Dalla Vecchia & Schaye 2008 Booth & Schaye 2010; Debuhr et al. 2012; Hopkins et al. 2013a). In these models some fraction of the gas escapes from the potential of the host galaxy, regulating future star formation and black hole growth and enriching the larger-scale environment with metals. In this section we will assess what impact the outflows that we observe may have on the evolution of their host galaxies.

If we consider the estimated mass outflow rates (derived in Section 4.3) we find that they are $\approx$ 6–20 times greater than the SFRs of our targets (with one exception at 100 times [J1430+1339] and excluding upper limits). These “mass loading” values are typically higher than those observed in star-forming galaxies but similar to those seen in luminous AGN, derived from a variety of observations of outflows in different gas phases (e.g., Martin 1999; Heckman et al. 2000; Newman et al. 2012; Cicone et al. 2014). We do not know the multi-phase outflowing gas properties of our targets; however, these observations imply that the star forming material may be rapidly depleted, as has also been suggested from molecular gas observations of a few nearby ULIRG-AGN composite galaxies (Cicone et al. 2014).

We can assess if the outflows will be able to escape the potential of their host galaxies. The escape velocities of typical type 2 AGN within the luminosity range of our sample are $\approx 500$ –1000 km s*-1* (assuming an isothermal sphere of mass out to 100 kpc; Greene et al. 2011). These escape velocities correspond to circular velocities of $\approx 200$ –400 km s*-1*, which are likely to be representative of the intrinsic values for our galaxies, based on those in which we identify rotation (see Section 4.2.1). If we assume the intrinsic bulk velocities of the outflows are $v_{{\rm out}}=W_{80}/1.3$ (see Section 4.3) then we have a range of outflow velocities $v_{{\rm out}}=510$ – $1100$ km s*-1* (with a median of 780 km s*-1*) which are comparable to the escape velocities. The maximum projected velocities are even higher (up to $v_{02}\approx-1700$ km s*-1*; Table 4). If we instead consider a $\approx$ 1011 M*⊙* galaxy inside a $\approx$ 1013 M*⊙* dark matter halo with Navarro-Frenk-White (Navarro et al. 1996) density profile, the gas is unlikely to escape unless it is travelling at $\gtrsim$ 1000 km s*-1* (see the calculation in Section 5.2 of Harrison et al. 2012). Indeed, some models have shown that even massive outflows may stall in the galaxy halo, re-collapse and cool at later times (along with new fuel supplies), resulting in re-ignition of star formation and black hole growth (e.g., Lagos et al. 2008; McCarthy et al. 2011; Gabor et al. 2011; Rosas-Guevara et al. 2013; Hopkins et al. 2013b; see also Gabor & Bournaud 2014).

In summary, while it is not possible to constrain the ultimate fate of outflowing gas from observations, we have observed outflows with extremely high velocities and the estimated outflow kinetic energy rates (i.e., $\approx 0.5$ –10% of $L_{{\rm AGN}}$ ) and momentum rates ( $\gtrsim 10$ – $20\times L_{{\rm AGN}}/c$ ) are in broad agreement with the requirements of models that invoke AGN-driven outflows to regulate star formation and black hole growth in luminous AGN (e.g., Di Matteo et al. 2005; Hopkins & Elvis 2010; Debuhr et al. 2012). Even if the outflows do not escape the galaxy halos, they may be the required pre-cursor to the postulated “maintenance mode” of feedback that appears to be necessary to control the level of cooling in massive halos at later times (e.g., Churazov et al. 2005; McCarthy et al. 2011; Gabor et al. 2011; Bower et al. 2012).

6 Conclusions

We have presented optical IFU observations covering the [O iii] $\lambda\lambda$ 4959,5007 and H $\beta$ emission lines of sixteen $z=0.08$ –0.2 type 2 AGN. Our targets were selected from a parent sample of $\approx 24,000$ $z<0.4$ AGN (Mullaney et al. 2013) and we demonstrate that they are representative of the $45\pm$ 3% of luminous ( $L_{\rm[O~{}III]}\geq 5\times 10^{41}$ erg s*-1*) $z<0.2$ type 2 AGN that have a significant fraction ( $\gtrsim$ 30%) of their ionised gas outflowing (Section 2.1). The fraction of AGN with ionised outflows at lower levels will be much higher. We use infrared SED decomposition on our targets to derive SFRs ( $\lesssim[10$ –100] M*⊙* yr*-1*) and AGN luminosities ( $L_{{\rm AGN}}=10^{44.3}$ –1046.0 erg s*-1*) for our targets. We also show that the radio emission in our targets (luminosity range: $L_{{\rm 1.4\,GHz}}\leq 10^{24.4}$ W Hz*-1*) is due to a combination of star formation and other processes, with five targets showing a “radio excess” above that expected from star formation and at least three of these showing spatially resolved radio emission (on scales $R_{1.4}\gtrsim 7$ kpc) that could be due to radio jets or shocks (Section 2.4). In summary, our targets are taken from a well constrained parent sample and are not extreme star forming systems or radio loud AGN that have been the focus of many similar studies.

The main results from our analysis are:

•

We find extended [O iii] emission-line regions over the full-field-of-view of our IFU observations (i.e., total extents of $\geq$ 10–20 kpc). In most of the sources these emission-line regions are circular or moderately elliptical; however, we observe some irregular morphologies, particularly in the radio excess sources (see Section 4.1).

•

By tracing narrow [O iii] emission-line components across the field-of-view, we identify a range of kinematic features associated with galaxy dynamics including galaxy rotation, merger debris and double nuclei. We isolate these kinematic components from any kinematics due to outflows (see Section 4.2).

•

We find high-velocity and disturbed ionised gas (velocity widths of $W_{80}\approx 600$ –1500 km s*-1*) extended over $\gtrsim(6$ –16) kpc in all of our targets. With our knowledge of the parent sample we conclude that, $\geq$ 70% (3 $\sigma$ confidence level) of the outflow features observed in the [O iii] emission-line profiles of $z<0.2$ luminous type 2 AGN are extended on kiloparsec scales (see Section 4.2).

•

The bulk outflow velocities across the sample are $v_{{\rm out}}\approx 510$ – $1100$ km s*-1*, which are comparable to the galaxy escape velocities. The maximum projected gas velocities reach up to $v_{02}\approx$ 1700 km s*-1* (see Section 4.2). These velocities indicate that ionised gas is currently being unbound from their host galaxies but it is not clear if it will permanently escape their host galaxy halos (see Section 5.2).

•

We observe a range of kinematic structure in the outflows including signatures of spherical outflows and bi-polar superbubbles. In several cases the outflows are preferentially oriented away from the plane of the host galaxy. We observe no obvious relationship between the outflow kinematic structures as a function of AGN luminosity, SFR or the presence or absence of radio AGN activity (see Section 4.2 and Section 5.1). However, we do find that all of our three superbubble candidates are radio excess sources.

•

Based on our analyses we find that both star formation and AGN activity may contribute to driving the outflows that we observe, with no definitive evidence that favours one over the other in the sample as a whole; however, we discuss individual objects in more detail (see Section 5.1). By combining our observations with those from the literature, we show that kiloparsec-scale ionised outflows are not confined to the most luminous AGN, extreme star-forming galaxies or radio-loud AGN; however, we find that the most extreme ionised gas velocities (with maximum velocities: $v_{02}\gtrsim$ 2000 kms*-1* and line-widths $W_{80}\gtrsim$ 1500 km s*-1*) are preferentially found in quasar-ULIRG composite galaxies (see Section 5.1).

•

Although based on simplifying assumptions, we estimate kinetic energy outflow rates ( ${\dot{E}}_{{\rm out}}\approx(0.15$ – $45)\times 10^{44}$ erg s*-1*) and mass outflow rates (typical $\approx 10\times$ the SFRs) that imply that considerable mass and energy are being carried in the observed outflows (see Section 4.3).

•

It is not possible to provide direct evidence of the long-term impact of individual outflows; however, although uncertain, we find that the mass outflow rates, the outflow kinetic energy rates ( $\approx 0.5$ –10% of $L_{{\rm AGN}}$ ) and outflow momentum rates (typically $\gtrsim$ 10–20 $\times L_{{\rm AGN}}/c$ ) are in broad agreement with theoretical predictions of AGN-driven outflows that are postulated to play significant role in shaping the evolution of galaxies (see Section 5.2).

In this paper we have investigated the prevalence, properties and potential impact of galaxy-wide ionised outflows. By selecting targets for detailed observations from our well constrained parent sample we have been able to place our observations into the context of the overall AGN population. We have established that galaxy-wide ionised outflows are prevalent in AGN. Our investigation was based upon outflows of ionised gas since it is currently is the only suitable gas phase for performing large statistical studies of outflows; however, it is imperative that we now obtain multi gas-phase observations of outflows from representative samples of objects, such as our targets, to fully characterise the properties and impact of galaxy-wide outflows in the global population.

Acknowledgments

We thank Michael Hogan for useful discussions on interpreting the FIRST radio data. We thank the referee, Jenny Greene, for her prompt and constructive comments. We gratefully acknowledge support from the Science and Technology Facilities Council (CMH through grant ST/I505656/1; DMA through grant ST/I001573/1 and AMS through grant ST/H005234/1) and the Leverhulme Trust (DMA; JRM). JRM also acknowledges support from the University of Sheffield via its Vice-Chancellor Fellowship scheme. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. The data used for this work are available through the Gemini Science Archive under the programme IDs: GS-2010A-Q-8 and GS-2012A-Q-21.

Appendix A Notes on individual sources

In this appendix we provide some notes on individual sources by highlighting relevant previous multi-wavelength observations from the literature and describing some of the key features in our IFU observations. The majority of these sources have received little or no attention previously in the literature, with the exceptions of J0945+1737, J1316+1753 and J1356+1026 that have been studied in some detail. The IFU data for all of our sources are displayed in this Appendix (Fig. A1-A15), except for J1010+1413 that can be found in Figure 5 and Figure 6 as it is used as an example in the main text.

J0945+1737

In agreement with our classifications (see Table 2) the source J0945+1737 has previously been identified as a LIRG with a type 2 AGN nucleus (Kim et al. 1995; Veilleux et al. 1995) and as a type 2 (“obscured”) quasar by Reyes et al. (2008). It has been shown to have a single nucleus based upon HST observations (Cui et al. 2001). The infrared luminosity that we measure (Table 2) is consistent with previous measurements (Cui et al. 2001) and our AGN luminosity measurement is consistent with that derived from Spitzer mid-infrared spectroscopy (Sargsyan et al. 2008). This source has evidence for a galaxy-scale radio jet due to its radio excess, above that expected from star formation, and the extended radio emission observed in FIRST ( $R_{1.4}\approx$ 6.6 kpc; see Section 2.4 and Table 2).

Our IFU data for this source are shown in Figure 13. The [O iii] peak velocity ( $v_{p}$ ) field is irregular (Fig. 13) and therefore, based on these data alone, we do not attribute the $v_{p}$ velocity field to galaxy rotation; however, we do note that the continuum image from our IFU data is elongated roughly in alignment with our kinematic “major axis”. The high-velocity broad underlying wings in the [O iii] emission-line profile are located over the full field-of-view ( $\approx 8\times 11$ kpc) but dominate in the central regions (with $W_{80}\approx 1100$ km s*-1* and $\Delta v\approx-260$ km s*-1*), preferentially located along a N–S axis (i.e., with a PA $\approx 0$ ). This is not parallel to the radio axis based on the FIRST data (PA $\approx$ 110*∘*) that may disfavour a connection between a radio jet and the outflow; however, high resolution radio data are required to accurately measure the morphology and origin of the radio emission.

J0958+1439

The IFU data for J0958+1439 are shown in Figure 14. The data reveal that the “flat-top” emission-line profile seen in the galaxy-integrated spectrum is primarily due to the sum of two luminous narrow kinematic components separated in velocity by $\approx$ 400 km s*-1*, with the red component dominating in the south and the blue in the north. At the limit of our resolution ( $\approx$ 1.4 kpc) we do not see two separate continuum or emission-line regions. Additionally, the similarity of the H $\beta$ and [O iii] emission-line profiles indicate similar [O iii]/H $\beta$ ratios for each kinematic component (also seen in the velocity profile; Fig. 14) suggesting that the two kinematic components are illuminated by a single AGN. The kinematic “major axis” we identify from the peak-velocity ( $v_{p}$ ) map is oriented parallel to the morphological major axis observed in the SDSS image and our continuum data (Fig. 14). Overall, we favour the interpretation that the $v_{p}$ velocity map is tracing galaxy rotation, although we cannot rule out a late stage merger without higher spatial resolution imaging or spectroscopic data.

In addition to the narrow [O iii] emission-line components there are underlying broad wings in the emission-line profile over the full field-of-view ( $\approx$ 7 $\times$ 10 kpc; Fig. 14). These broad emission-line profiles (with $W_{80}\approx 850$ km s*-1* and $\Delta v\approx-50$ km s*-1*) are more dominant perpendicular to the kinematic “major axis”. This potentially indicates that the outflow is escaping away from a galactic disk. The emission-line flux ratio $\log($ [O iii]/H $\beta)\approx 1.1$ remains constant over the central $\approx$ 6 kpc and then abruptly declines, in a similar manner to that seen in the type 2 quasars of Liu et al. (2013a) (see Section 4.3). Therefore, to measure the outflow kinetic energy following Liu et al. (2013b) we adopt a value of $R_{{\rm br}}=3$ kpc and measure a H $\beta$ surface brightness of $\Sigma_{{\rm H}\beta}=1.4\times 10^{-15}$ erg s*-1* arcsec*-2* in an annulus at this radius (corrected for cosmological dimming). Using this method we obtain an outflow kinetic energy of ${\dot{E}}\approx 2.8\times 10^{43}$ erg s*-1*, which is in excellent agreement with the value that we derive in Section 4.3 of ${\dot{E}}_{{\rm out}}\approx 2.7\times 10^{43}$ erg s*-1*.

J1000+1242

Reyes et al. (2008) previously identified J1000+1242 as a type 2 (“obscured”) quasar. This source has excess radio emission (above that expected from star formation; Table 2) that is marginally resolved based on the FIRST data ( $R_{1.4}\approx 8.4$ kpc with a PA $\approx 160^{\circ}$ ; see Section 2.4 and Table 2). Although fairly irregular, our IFU data reveal a velocity gradient from blue to red in the peak of the [O iii] emission-line profile ( $\Delta v_{p}\approx 200$ km s*-1*) at a PA of $\approx$ 240*∘* (Fig. 15). The peak signal-to-noise ratio map, combined with the velocity maps reveal a distinct luminous emission-line region $\approx$ 3.5 kpc north-east of the nucleus that has a reasonably narrow emission-line profile ( $W_{80}\lesssim 600$ km s*-1*). This feature may be merger debris, halo gas or an outflow remnant (see Section 4.2.1).

Roughly perpendicular to the $v_{p}$ velocity gradient (with a PA $\approx 160$ –170*∘) we observe regions with broad underlying wings (with $W_{80}\approx 850$ km s-1*) in the [O iii] emission-line profile. From the spectra in boxes 12, 17 and 18 in Figure 15 it can be seen that this broad emission is built up from multiple narrow components. This emission shows a velocity shift from blue (with $\Delta v\approx-100$ km s*-1*) to red (with $\Delta v\approx 100$ km s*-1*) over $\approx 10$ –15 kpc. We also see an irregular morphology of the emission-line region. All of the above indicates that we are observing an outflowing bi-polar superbubble, with a strikingly similar kinematic structure to previously identified superbubbles (i.e., J1356+1026 from Greene et al. 2012; J0319-0019 from Liu et al. 2013b; and Mrk 273 from Rupke & Veilleux 2013). The PA of the superbubbles we observe in J1000+1242 is parallel to the axis of the extended radio emission and found over a similar spatial extent, potentially demonstrating a direct connection between a radio jet and the outflow, although shock-induced radio emission is another possibility. The SDSS image of this source (Fig. 15) is extended beyond the field-of-view of our observations and therefore IFU observations over larger regions are required to trace the full kinematic structure of this target.

J1010+1413

We show our IFU data for J1010+1413 in Figure 5 and Figure 6. This source has the broadest [O iii] emission-line width of the entire sample ( $W_{80}=1450$ km s*-1*) and is kinematically complex. In the central regions we observe a smooth gradient from red to blue in the [O iii] peak velocity ( $v_{p}$ ; Fig. 6), although the overall velocity field is irregular (Fig. 5). There are luminous narrow [O iii] emission-line components located to the north and south of the nucleus, separated by $\approx$ 16 kpc in projected distance and $\approx 350$ km s*-1* in projected velocity, that appear to be associated with emission-line regions that are apparent in the SDSS image (Fig. 6). These features may be merger debris, halo gas or outflow remnants (see Section 4.2.1). The extremely broad emission-line profiles, with velocity offsets reaching up to $\Delta v\approx-350$ km s*-1*, is located out to the very edge of the field-of-view (i.e., over $\gtrsim 16$ kpc).

J1010+0612

Reyes et al. (2008) previously identified J1010+0612 as a type 2 (“obscured”) quasar. This source has a clear radio excess above that expected from star formation alone (Table 2). Based on the FIRST data we do not spatially resolve any radio emission on scales $\gtrsim 2^{\prime\prime}$ (see Section 2.4 and Table 2).

We show our IFU data for this object in Figure 16. We identify a regular blue to red velocity field, indicative of galactic rotation, in the [O iii] peak velocity (with $\Delta v_{p}\approx 300$ km s*-1*). Broad underlying wings lead to emission-line widths of $W_{80}\approx 1100-1300$ km s*-1* and velocity offsets of $\Delta v\approx-100$ km s*-1* and are located over the full extent of the field-of-view ( $\approx 6\times 9$ kpc) with a roughly circular morphology.

J1100+0846

Reyes et al. (2008) previously identified J1100+0846 as a type 2 (“obscured”) quasar and we show our IFU data for this object in Figure 17. Although the peak velocity ( $v_{p}$ ) map shows some irregular features we see a smooth velocity gradient ( $\Delta v_{p}\approx 100$ km s*-1*) along a PA $\approx 250^{\circ}$ . Additionally, this kinematic “major axis” is parallel to the morphological major axis seen in the SDSS image (with hints of the same axis in our continuum image), providing evidence that kinematics due to galaxy rotation are being observed. Broad wings (leading to $W_{80}\approx$ 700–1100 km s*-1*) are observed in the emission-line profiles over the full field-of-view ( $\approx 9\times 6$ kpc). The uncertainties on the velocity offset of the broad emission (i.e., $\Delta v$ ) are large; however, the velocity seems to be highest in the southern regions with velocities between $\Delta v\approx-100$ km s*-1* and $-80$ km s*-1*. These blue-shifted velocities could indicate the near-side of an outflow that is moving away from the almost face on galaxy disc. In the velocity profile (Fig. 17), we observe that the emission-line flux ratio of $\log($ [O iii]/H $\beta)\approx 1.0$ –1.1 is constant over the central regions then declines at a break radius of $R_{{\rm br}}=2.5$ kpc (see Section 4.3). We therefore measure a H $\beta$ surface brightness of $\Sigma_{{\rm H}\beta}=1.7\times 10^{-15}$ erg s*-1* arcsec*-2* in an annulus at this break radius (corrected for cosmological dimming) and calculate the outflow kinetic energy following Liu et al. (2013b). The kinetic outflow energy we obtain using this method, i.e., ${\dot{E}}\approx 6.3\times 10^{43}$ erg s*-1*, is in excellent agreement with the value that we calculate in Section 4.3, i.e., ${\dot{E}}_{{\rm out}}\approx 6.9\times 10^{43}$ erg s*-1*.

J1125+1239

We show our IFU observations for J1125+1239 in Figure 18. We observe an emission-line region that is extended right to the edges of the field-of-view (over $\gtrsim$ 13 kpc). We do not observe any velocity gradients in the peak of the [O iii] emission-line profile (i.e., $v_{p}$ ; Fig. 18). There is faint, high velocity ( $\Delta v\approx-250$ to $-300$ km s*-1*) and extremely broad [O iii] emission ( $W\approx 1200$ –1500 km s*-1*), out to the very edges of the field-of-view. Due to the low surface brightness the morphology of the broad emission is uncertain; however, the SDSS image shows faint extended emission beyond the IFU field-of-view, parallel to the axis where we predominantly observe this broad emission (i.e., PA $\approx 250^{\circ}$ ; Fig. 18).

J1130+1301

We show our IFU observations for J1130+1301 in Figure 19. The peak velocity map ( $v_{p}$ ) is indicative of galactic rotation showing a smooth velocity gradient from red to blue with $\Delta v_{p}\approx 300$ km s*-1*. This interpretation is strengthened by the kinematic major axis being aligned with the morphological major axis observed in the SDSS image and the continuum observed in our data. The broad emission-line wings (with $W_{80}>600$ km s*-1*) are found perpendicular to the galaxy major axis over $\approx$ 8 kpc. Although both blue and red wings are observed in the emission-line profile the red wing is more luminous resulting in an overall velocity offset that is positive (i.e., $\Delta v\approx$ 150 km s*-1*).

J1216+1417

Spectroscopy of J1216+1417 has previously identified Wolf-Rayet stars are located in this source and consequently shows that a recent star formation episode has occurred (i.e., $\approx 2$ –5 Myrs; Brinchmann et al. 2008). Furthermore, a subtle 4000Å break and stellar absorption features are seen in the SDSS spectrum and may explain why the H $\beta$ emission-line profile is different to that seen in [O iii] (Fig. 20). We show our IFU data in Figure 20. We see little velocity structure in the $v_{p}$ velocity map. The [O iii] emission line profile is irregular, with both a blue wing (out to $v\approx$ 1000 km s*-1*) and a large red wing out to $v\approx$ 2000 km s*-1*. This is seen in both lines of the [O iii] $\lambda\lambda$ 4959,5007 emission-line doublet indicating that it is a true [O iii] feature associated with this source. Further kinematic analysis, beyond the scope of this work, is required to fully identify the origin of these high-velocity wings. The blue and red wings observed in the [O iii] emission line are predominantly found in a circular region across the central $\gtrsim$ 6 kpc; however, the low surface brightness of the emission in the outer regions puts some uncertainty on the spatial extent and morphology.

J1316+1753

The target J1316+1753 has a double peaked [O iii] emission-line profile, with components separated by $\approx$ 400–500 km s*-1* (Fig. 21). As a result of this, this source has previously been identified as a dual AGN candidate (Xu & Komossa 2009; also see Section 4.2.1). In the peak signal-to-noise ratio map we observe two spatial regions separated by $\approx 2$ –3 kpc and orientated with a PA $\approx 130^{\circ}$ associated with these kinematics components; however, at the resolution of our observations ( $\approx$ 2 kpc) we do not observe two continuum peaks (Fig. 21). The similarity of the emission-line ratios in both systems, and the lack of a double continuum source, suggests that these kinematic components may be illuminated by a single ionising source. Also, the kinematic components have similar fluxes and the velocities are independent of projected distance from the centre; therefore, the double-peaked profile may by the result of gas kinematics, jet interactions and/or AGN-driven outflows (Fig. 21; also see Xu & Komossa 2009; Smith et al. 2010; Barrows et al. 2013). The lack of evidence for extended luminous radio emission on $\gtrsim 2^{\prime\prime}$ scales from the FIRST data (Table 2) could argue against a jet interaction (as seen in other sources e.g., Rosario et al. 2010); however, deep and high-resolution radio data would be required to confirm this. Additionally the axis of the two components is aligned with the morphological major axis seen in the SDSS image and the continuum emission seen in our data (Fig. 21). Overall, all of these observations favour the scenarios of a rotating gas disk or two merging components (separated by $\lesssim 2$ kpc) illuminated by a single AGN to explain the double peaked [O iii] emission-line profile in this source (also see Smith et al. 2012 and Comerford et al. 2012). Combining these IFU data with deep high-spatial resolution optical, X-ray and radio images would provide the data required to robustly distinguish between all of the possible scenarios.

In addition to the narrow components our IFU data for J1316+1753 (Fig. 21) reveal a very broad [O iii] emission-line profile ( $W_{80}\approx$ 1100–1200 km s*-1*), preferentially located perpendicular to the kinematic “major axis” defined from the $v_{p}$ map. This [O iii] emission has a velocity offset of $\Delta v\approx-200$ km s*-1* and is extended right to the edge of the field-of-view (i.e., over $\gtrsim 14$ kpc). In the velocity profile, we observe an emission-line flux ratio $\log($ [O iii $]/{\rm H}\beta)\approx 1.0$ –1.1 over the central regions and that declines at a break radius of $R_{{\rm br}}=5$ kpc (see Section 4.3). Therefore we measure a H $\beta$ surface brightness of $\Sigma_{{\rm H}\beta}=6.2\times 10^{-16}$ erg s*-1* arcsec*-2* in an annulus at this break radius (corrected for cosmological dimming) and measure the outflow kinetic energy following Liu et al. (2013b). Following this method we obtain an outflow kinetic energy of ${\dot{E}}\approx 5.4\times 10^{43}$ erg s*-1* that is within a factor of 3 of the value we calculated in Section 4.3, i.e., ${\dot{E}}_{{\rm out}}\approx 1.7\times 10^{44}$ erg s*-1*.

J1338+1503

We show our IFU data for J1338+1503 in Figure 22. We note that hints of stellar continua are seen in the SDSS spectrum and contamination from stellar absorption may cause the difference in the H $\beta$ and [O iii] emission line profiles (Fig. 22). The peak velocity map ( $v_{p}$ ) for J1338+1503 is irregular; however, we do see a general shift from blue to red. This source has a predominantly red wing, in addition to a weaker blue wing leading to an overall velocity offset of $\Delta v\approx 120$ km s*-1* across the central few kiloparsecs. The broad emission ( $W_{80}\approx 900$ km s*-1*) is preferentially found perpendicular to the velocity gradient (i.e., the kinematic “major axis”) that we observe in $v_{p}$ (Fig. 22).

J1339+1425

The source J1339+1425 is the least luminous AGN in our sample (in both $L_{{\rm AGN}}$ and $L_{{\rm[O~{}III]}}$ ) and is the only source not detected at 1.4 GHz by FIRST or NVSS (Table 2). We show our IFU data for this source in Figure 23 where we observed [O iii] emission right to the edge of the field-of-view of our observations (i.e., over $\gtrsim$ 12 kpc). This emission-line region is dominated by narrow kinematic components (i.e., with $W_{80}<600$ km s*-1*) that trace out a small velocity gradient ( $\Delta v_{p}\approx 100$ km s*-1*) from blue to red. We find that the kinematic major axis is broadly consistent with the continuum morphological axis observed in our data and the SDSS image (Fig. 23). We therefore appear to be tracing galactic rotation in the $v_{p}$ map.

Across the central $\approx$ 6 kpc we observe a blue- and red-wing in the [O iii] emission-line profile, with an overall velocity offset of $\Delta v\approx 150$ km s*-1*, that is preferentially located at $\approx 50^{\circ}$ –70*∘* away from this kinematic “major axis”. In the velocity profile, we observe that the emission-line flux ratio $\log($ [O iii]/H $\beta)\approx 1.1$ is constant over the central regions and then declines at a break radius of $R_{{\rm br}}=2$ kpc (see Section 4.3). We therefore measure a H $\beta$ surface brightness of $\Sigma_{{\rm H}\beta}=7.5\times 10^{-16}$ erg s*-1* arcsec*-2* in an annulus at this break radius (corrected for cosmological dimming) and calculate the outflow kinetic energy following Liu et al. (2013b). Using this method we obtain an outflow kinetic energy of ${\dot{E}}\approx 5.5\times 10^{42}$ erg s*-1* that is within a factor of $\approx 3$ of the value we calculated in Section 4.3, i.e., ${\dot{E}}_{{\rm out}}\approx 1.5\times 10^{43}$ erg s*-1*.

J1355+1300

We show our IFU data for J1355+1300 in Figure 24. The majority of the emission-line region is dominated by narrow [O iii] emission-line profiles ( $W_{80}\approx$ 300–400 km s*-1*) and we observe a small velocity gradient (with $\Delta v_{p}\approx 100$ km s*-1*) from blue-to-red in a north-south direction in the peak velocity map ( $v_{p}$ ) indicative of galactic rotation. Located $\approx 3$ kpc to the south-east of the nucleus there is a distinct kinematic component with a velocity between $\approx-500$ km s*-1* and $-600$ km s*-1* which has a high-velocity blue-wing in the emission-line profile (see grids of spectra in Fig. 24). The similarity of the galaxy-integrated [O iii] and H $\beta$ emission-line ratios for this kinematic component (see Fig. 24) could imply that the ionising source is the same as that for the rest of the emission-line region and is therefore likely to be illuminated by the central source. At the depth of our observations, we observe no spatially distinct continuum source or emission-line region coincident with this kinematic component (Fig. 24). The exact origin of this high-velocity feature is difficult to determine with these observations alone; however, the highly-disturbed nature and lack of associated continuum strongly suggests an outflow. This features is similar to that seen in J1430+1339 and therefore could indicate a smaller-scale outflowing bubble and/or could be due to a jet-ISM interaction at this position (as is seen in other sources, e.g., Emonts et al. 2005). Due to the faintness of the radio emission, and the lack of IRAS detections we are unable to constrain the radio excess parameter for this source (see Section 2.4 and Table 2); therefore, deep and high-resolution radio imaging is required to determine if radio jets are present and if they are aligned with this high-velocity feature.

J1356+1026

As a well known type 2 (“obscured”) quasar (Reyes et al. 2008) this source has already received a lot of attention in the literature. Two merging galactic nuclei with a projected separation of $\approx 2.5$ kpc are seen in the optical and near-infrared (Shen et al. 2011; Fu et al. 2011; Greene et al. 2012). In agreement with this, our IFU data reveal two continuum sources and, moderately offset from this, corresponding emission-line regions with distinct velocities (Fig. 25). We note that Fu et al. (2012) did not confidently associate an emission-line region with the southern nucleus in their IFU data, although our data are unambiguous (Fig. 7 and Fig. 25). Combined with the observation that these kinematic components have slightly different [O iii]/H $\beta$ emission-line flux ratios implies two distinct type 2 quasars (Fig. 25; also see Greene et al. 2012). This source has a clear radio excess above that expected from star formation alone (see Section 2.4 and Table 2). However, there is no evidence for extended radio emission at the few mJy level on scales of $\gtrsim 2^{\prime\prime}$ based on the FIRST data (Section 2.4; Table 2).

Using longslit observations, Greene et al. (2012) revealed [O iii] emission extended over 10s of kiloparsecs, reaching beyond the field-of-view of our IFU observations. In particular they reveal a “bubble” of [O iii] emission with a spatial extent of 12 kpc to the south and high-velocity “clumps” to the north. These “bubble” and “clumps” reach a projected velocity of $\approx 250$ km s*-1*. Greene et al. (2012) postulate a quasi-spherical outflow that extends from the south to the north, where the outflow is forced into this bi-polar shape due to high density regions of gas and dust in the central galaxies (e.g., Faucher-Giguère & Quataert 2012). Our IFU observations (Fig. 25) cover the base of these features. It is worth noting that despite our lack of spatial coverage and different approaches, our outflow energy injection rate estimate ( $\dot{E}_{{\rm out}}\approx 1.3\times 10^{44}$ erg s*-1*; Section 4.3) is consistent with the fiducial range $=10^{44}$ – $10^{45}$ erg s*-1* quoted in Greene et al. (2012). Based on our SED fitting (Section 2.4) the bolometric luminosity of the AGN in this source is $\approx 1\times 10^{45}$ erg s*-1* while the infrared-luminosity from star formation is $\approx 2\times$ 1045 erg s*-1* (SFR $\approx 60\,\hbox{M$ {}_{\odot} $\,yr$ {}^{-1} $}$ ; Table 2). Therefore, in contrast to the conclusions of Greene et al. (2012) we find there is a comparable amount of energy available from the AGN and star formation to power the outflow. The observed radio emission in FIRST is on much smaller scales than the $\approx 10$ kpc outflowing bubble that may argue against a jet-driven outflows; however, deep and high-resolution radio imaging is required to determine the origin and morphology of the radio emission in this source.

J1430+1339

Reyes et al. (2008) previously identified J1430+1339 as a type 2 (“obscured”) quasar. Additionally, this source was identified by Galaxy Zoo (Keel et al. 2012) as having an extended emission-line region, due to the arc shaped “purple haze” to the north east of the SDSS image (Fig. 26). This galaxy has consequently received the nickname the “Teacup”. Follow-up HST imaging171717http://blog.galaxyzoo.org/2012/06/14/hubble-spies-the-teacup-and-i-spy-hubble has revealed that this is actually a 5–10 kpc emission-line “loop” accompanied by a smaller emission-line “arc” on the opposite side of the nucleus, potentially analogous to the bi-polar outflow observed in J1356+1026 (Fig. 25; Greene et al. 2012). The exact origin of these emission-line regions are unknown; however, our IFU data (Fig. 26) cover the base of the north-east loop and reveal it has a velocity of $\approx-900$ km s*-1* from the systemic with a high velocity tail out to $\approx-1400$ km s*-1*. The [O iii] peak signal-to-noise map show faint extended features similar to those seen in outflowing superbubbles (e.g., Liu et al. 2013b). Additionally these features looks remarkably similar to the bi-conal shells of ionised gas seen in the high-redshift radio galaxy MRC 0406–244 that appear to be due to gas that has been expelled from the host galaxy by radio jets and/or star formation (Hatch et al. 2013; see also Nesvadba et al. 2008). Indeed, the PA of these high velocity ionised gas features in J1430+1339 are roughly aligned with, and found over a similar scale to, the extended radio emission seen in the FIRST data (see Section 3 and Table 2). High-resolution radio imaging of this source will play a crucial role in determining the true origin of the radio emission (e.g., radio jets or shocks). IFU observations covering the full extent of the emission-line region will reveal the full velocity structure in these regions. The peak velocity ( $v_{p}$ ) map shows a large velocity gradient from blue to red ( $\Delta v_{p}\approx 600$ km s*-1*) along an axis with PA $\approx 30^{\circ}$ (Fig. 26).

J1504+0151

Reyes et al. (2008) previously identified this source as being a type 2 (“obscured”) quasar. We note that there are stellar continuum and absorption features visible in the SDSS spectrum of this source and therefore contamination from stellar absorption may result in the different H $\beta$ and [O iii] emission line profiles (Fig. 27). This source is one of the faintest in our sample; however, we still observe an [O iii] emission-line region out to the edges of the field-of-view (i.e., over $\gtrsim 13$ kpc; Fig. 27). The peak velocity map ( $v_{p}$ ) shows a smooth velocity gradient from blue to red (with $\Delta v_{p}\approx 150$ km s*-1*) indicative of galaxy rotation with a PA $\approx 140^{\circ}$ . Due to the faint emission lines, it is difficult to establish the true extent and morphology of the broad, high-velocity emission (which has $W_{80}\approx 1000$ km s*-1* and a velocity offset between $\Delta v\approx-300$ and $-500$ km s*-1*); however, it appears to dominate in the central few kiloparsecs (Fig. 27).

Bibliography221

The reference list from the paper itself. Each links out to its DOI / PubMed record.

1Abazajian et al. (2009) Abazajian K. N. et al., 2009, APJS, 182, 543
2Alatalo et al. (2011) Alatalo K. et al., 2011, Ap J, 735, 88
3Alexander & Hickox (2012) Alexander D. M., Hickox R. C., 2012, New Astron. Rev., 56, 93
4Alexander et al. (2010) Alexander D. M., Swinbank A. M., Smail I., Mc Dermid R., Nesvadba N. P. H., 2010, MNRAS, 402, 2211
5Allington-Smith et al. (2002) Allington-Smith J. et al., 2002, PASP, 114, 892
6Antonucci (1993) Antonucci R., 1993, ARAA, 31, 473
7Baldwin et al. (1980) Baldwin J. A., Carswell R. F., Wampler E. J., Boksenberg A., Smith H. E., Burbidge E. M., 1980, Ap J, 236, 388
8Baldwin et al. (1981) Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP, 93, 5