THE INNER KILOPARSEC OF Mrk 273 WITH KECK ADAPTIVE OPTICS

Vivian U; Anne Medling; David Sanders; Claire Max; Lee Armus; Kazushi Iwasawa; Aaron Evans; Lisa Kewley; Giovanni Fazio

doi:10.1088/0004-637X/775/2/115

1. INTRODUCTION

When gas-rich galaxies collide and merge, gaseous material can lose angular momentum and funnel toward the center of the merger system (e.g., Barnes & Hernquist 1992). There it can feed starburst and active galactic nuclei (AGNs) activities which power luminous and ultraluminous infrared galaxies ((U)LIRGs). The detailed processes of how merger dynamics contribute to AGN activity (Sanders et al. 1988; Hopkins et al. 2006; Comerford et al. 2009), how mass builds up in the central supermassive black hole (Hopkins et al. 2010; Treister et al. 2010; Rosario et al. 2011), how much energy is deposited into the system by the ionizing sources (Younger et al. 2009; Sanders & Mirabel 1996), and when the active nuclei are triggered (Donley et al. 2010; Hopkins 2012; Farrah et al. 2009) are areas of active research. Because AGN have been observed in mergers with double nuclei (Komossa et al. 2003; Ballo et al. 2004; Koss et al. 2012; Fu et al. 2011; Liu et al. 2011; McGurk et al. 2011; Gerke et al. 2007; Mazzarella et al. 2012), it is known that an AGN can "turn on" before the final coalescence of the nuclei. However, many mergers do not host dual AGN (e.g., Teng et al. 2012). It is inconclusive from the models when an AGN is triggered and what dictates whether or not synchronous ignition occurs (Van Wassenhove et al. 2012).

These questions can be addressed with high-resolution observations of the nuclei in nearby LIRG mergers. Previous integral-field spectroscopic work on large samples of (U)LIRGs have mapped the large-scale distribution and kinematics of ionized gas, compared them to that of stellar components, determined metallicity gradients, and found evidence for galactic outflows and shocks (e.g., Westmoquette et al. 2012; Rich et al. 2011, 2012; García-Marín et al. 2009). Much of the gas is concentrated within the dusty cores of these (U)LIRGs, so longer wavelengths must be probed to understand the nuclear regions. Piqueras López et al. (2012) has presented an atlas of the two-dimensional molecular and ionized gas structure in the southern local (U)LIRGs as probed in the near-infrared, showing the extent and kinematics of the gaseous components at ∼0.2 and 0.9 kpc spatial resolution. These seeing-limited observations with an average angular resolution of ∼0 farcs 63 FWHM set the framework of the large kiloparsec-scale conditions of the gas and prompt detailed high-resolution investigations of the nuclear regions.

Due to the advent of laser guide star adaptive optics (LGS AO; Wizinowich et al. 2006; van Dam et al. 2006), black holes in local (U)LIRGs can now be examined at the subarcsecond level (Max et al. 2007; Melbourne et al. 2008; Engel et al. 2010; Davies et al. 2010; Medling et al. 2011). In particular, the OH-Suppressing Infra-Red Imaging Spectrograph (OSIRIS; Larkin et al. 2006), an integral-field unit (IFU) with near-infrared wavelength coverage (1 μm <λ < 2.4 μm) on the Keck II Telescope, offers imaging and kinematic information at <0 farcs 1 spatial resolution with LGS AO (e.g., Wright et al. 2009; Law et al. 2009; Do et al. 2009; McConnell et al. 2011; Walsh et al. 2012). NIRC2 (PIs: Keith Matthews and Tom Soifer), a diffraction-limited infrared camera also on the Keck II Telescope, offers scales of ∼0 farcs 01 pixel⁻¹ in the 1–5 μm range in its narrow camera mode. They both probe the near-infrared regime where dust extinction does not pose as severe a problem.

To understand the mechanisms responsible for powering the extreme infrared luminosities in (U)LIRGs, we are conducting a survey of nearby galaxy mergers using OSIRIS to probe the gas kinematics and energetics of the nuclear regions in these systems. Observed as part of the larger campaign, Mrk 273 (= UGC 08696) is the focus of this paper as an advanced merger with infrared luminosity L_IR = 10^12.21 L_☉. At a distance of z = 0.038 (systemic velocity cz = 11, 400 km s⁻¹ and physical scale = 0.754 kpc/''),¹⁰ it exhibits a prominent optical tidal tail (∼40 kpc) extending to the south (Figure 1). With its high infrared luminosity, this ULIRG has a heated dusty core that obscures the central ionizing sources in optical light. As one of the brightest and closest ULIRGs, Mrk 273 has been well studied across the full wavelength spectrum, including Spitzer, Hubble, Chandra, and GALEX observations from the Great Observatories All-sky LIRG Survey (Armus et al. 2009; Howell et al. 2010; Iwasawa et al. 2011b). However, the details of its power engine remain controversial.

**Figure 1.** Left: *HST* ACS B(F435W) + I(F814W) composite image of Mrk 273. In these and subsequent images, north is up and east is to the left. The box indicates the zoomed-in region in the gray-scale panels. The middle and right panels show the zoomed-in B- and I-band images, respectively. The three crosses mark the N, SE, and SW sources of interest identified in radio observations (Carilli & Taylor 2000; Bondi et al. 2005) whose near-infrared characteristics will be presented in this paper.
Download figure:
Standard image High-resolution image

There is substantial evidence in the literature suggesting that at least one AGN resides within Mrk 273: Armus et al. (2007) detected the [Ne v] line in Spitzer's mid-infrared IRS spectrum, and measured hot gas with T > 300 K in the continuum. Polycyclic aromatic hydrocarbon strength diagnostics from mid-infrared Infrared Space Observatory spectroscopy classify the galaxy as an AGN (Lutz et al. 1999; Rigopoulou et al. 1999). Hard X-ray emission and the Fe k line (e.g., Xia et al. 2002; Iwasawa et al. 2011a; Ptak et al. 2003) have also been detected in support of the presence of a hard X-ray AGN, though most of the total power is likely to be from circumnuclear starbursts (Iwasawa et al. 2011a).

The status of Mrk 273 as a ULIRG and a late-stage pre-coalescence merger makes it an important galaxy for understanding the role of mergers in galaxy evolution. The specific goals of our high-resolution IFU study are to locate the nuclei of the progenitor galaxies, determine the characteristics of any supermassive black holes, verify their active or dormant nature, and to investigate the relative effect of AGN activity and star formation on the surrounding nuclear regions.

Multiwavelength studies have decomposed the nuclear region into multiple components. Three main components have been identified from the detected radio continuum: north, southeast, and southwest (N, SE, SW; Condon et al. 1991; Smith et al. 1998) as indicated in Figure 1. These components show dissimilar characteristics in most bands; it is not clear which of them may be remnant nuclei. We combine our data with existing evidence to determine the likely sources of each. The SE component is only weakly detected in the near-infrared (Scoville et al. 2000) and weak or undetectable in the X-ray given the limited resolution of Chandra (Iwasawa et al. 2011a), but the N and SW X-ray emission coincide with peaks in the near-infrared continuum flux. OH maser emission along with steep radio spectral index suggest the presence of a low-luminosity AGN with a binding mass of 1.39 ± 0.16 × 10⁹ M_☉ in the N component (Klöckner & Baan 2004). The N radio-1.4 GHz emission is further resolved into individual compact sources thought to be clustered supernovae (Smith et al. 1998; Carilli & Taylor 2000; Bondi et al. 2005). More recently, Iwasawa et al. (2011a) found, in Chandra images with improved astrometry, that the hard X-ray flux is superposed on the SW compact source, with a 6–7 keV excess extending toward the north. They determined that the known hard X-ray AGN is located at the SW component, and attributed the 6–7 keV excess to enhanced Fe k emission either from the SW source or from a heavily absorbed Compton-thick AGN at the N component.

In this paper, we present high-resolution infrared integral-field spectroscopy and imaging observations of Mrk 273 taken with the OSIRIS and NIRC2 instruments on the Keck II Telescope, probing down to the ⩽0 farcs 1 scale. These near-infrared high-resolution maps provide great detail for identifying the location of the hardest photons from AGN and shocks as well as the dynamics of the gas. Observations and data processing are described in Section 2. Resulting spectra and images are analyzed and presented in Sections 3 and 4; our interpretations are detailed in Section 5. Our conclusions are summarized in Section 6. We adopt H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_m = 0.28, and Ω_Λ = 0.72 throughout the paper.

2. OBSERVATIONS AND DATA REDUCTION

2.1. OSIRIS Data Cubes

The Keck OSIRIS observations were obtained on UT 2011 May 22 under clear, dry, and stable conditions with the LGS AO system. Two filter-scale combinations were used to observe Mrk 273: the broad K-band filter (λ = 1965–2381 nm, hereafter Kcb) at the 100 mas pixel⁻¹ scale (∼75 pc for Mrk 273), and a narrow H-band filter (λ = 1652–1737 nm; hereafter Hn4) at the 35 mas pixel⁻¹ scale (∼26 pc for Mrk 273). The tip-tilt star employed for the AO corrections had R magnitude m_R = 16.1 at 33'' away from the target, well within the LGS requirement constraints. We adopted a position angle (P.A.) of 30° east of north and an observing scheme of object–sky–object with 10 minute integration per frame, totaling 50 minute and 60 minute on-source times for the Kcb-100 mas and Hn4-35 mas observations, respectively. The A0V star 81 UMa was observed at two different airmasses in the Kcb filter (three in Hn4) for telluric corrections.

The data sets were processed using the OSIRIS pipeline version 2.3 (Krabbe et al. 2004). The full Astronomical Reduction Pipeline template (ARP_SPEC) incorporates dark-frame subtraction, channel level adjustment, crosstalk removal, glitch identification, cosmic ray cleaning, spectra extraction for data cube assembly, dispersion correction, scaled sky subtraction for enhanced OH-line suppression, and telluric correction. The extracted spectra were subsequently cleaned to remove bad pixels. The cleaned frames were registered using the SW peak position and combined using the pipeline's clipped mean algorithm. The final data cubes encompass signal at each pixel in the x, y, and wavelength dimensions. The FWHM of the tip-tilt star was 0 farcs 05 in the Kcb setting and 0 farcs 06 in the Hn4 setting.

The final OSIRIS Kcb and Hn4 cubes have been flux-calibrated to the respective K'- and H-band NIRC2 flux-calibrated images (see Section 2.2) in the following manner. Spatially coincident regions were first defined within resolution uncertainties in the OSIRIS and NIRC2 images of the matching bands. Spectroscopically, the bandwidth differences between the OSIRIS data cubes and the NIRC2 images have been taken into account while integrating over the Kcb and Hn4 wavelength ranges, respectively. Conversion factors thus determined offer a calibration to absolute fluxes in physical units that allow for line ratio computations across the two bands.

It is important to understand the point-spread function (PSF) intrinsic to our OSIRIS observations before comparing them to models. We estimate our PSF by fitting a Moffat profile to images of the tip-tilt star taken right before our galaxy observation sets. However, this is a best-case limit for the PSF, because the AO correction will degrade somewhat when moving off-axis to the galaxy. The tip-tilt star provides low-order corrections, which vary on scales of the isokinetic angle. (This angle is much larger than the isoplanatic angle, which governs how wide a field-of-view an LGS can correct.) At Mauna Kea, the isokinetic angle is approximately 75'' (van Dam et al. 2006); this represents the tip-tilt star distance at which Strehl will be degraded by 1/e. Because our tip-tilt star is at a separation of ∼33'' from Mrk 273, we expect our Strehl to be degraded by a factor of ∼0.6 from the tip-tilt star profile. To model this, we convolve our tip-tilt star PSF model by a Gaussian of FWHM 0 farcs 05 to estimate our true PSF.

2.2. NIRC2 Images

High-resolution images were also obtained with NIRC2, using the narrow camera mode (0 farcs 009942 pixel⁻¹; 8 pc pixel⁻¹) and the LGS AO System, on UT 2012 May 20. Images were taken in the H (λ_c = 1633 nm) and K' (λ_c = 2124 nm) bands, with exposure times of nine minutes each. Individual frames were exposed for 30 s for each of two co-adds; frames were dithered using the standard "bxy3" dither pattern that avoids the noisy lower-left quadrant of the chip. The same tip-tilt star used in the OSIRIS observations was used here. We reduced our images using the CfAO Treasury Survey pipeline (Glassman et al. 2002; Melbourne et al. 2005). A sky frame was produced by combining sky regions of each dithered frame, then subtracted from each frame. A flat field was created by normalizing the constructed sky frame. This flat field was then divided out of each frame in order to correct for varying sensitivities across the detector. The sky-subtracted-and-flattened images were subsequently aligned and median-combined to obtain the final reduced image, which were then flux-calibrated using images taken of the UKIRT Faint Standard star FS 133 (Hawarden et al. 2001). Standard star exposures were 5 s frame⁻¹, also dithered with the "bxy3" script; four frames were combined for the H-band standard, three for the K'-band standard. A FWHM of ∼0 farcs 06 has been measured from a compact source south of the SE component in both the H- and K'-band images.

3. EMISSION LINE FITTING AND ERROR ANALYSIS

From each of the reduced OSIRIS Kcb and Hn4 data cubes, various emission lines were mapped as follows. First we extracted the continuum with power-law fits and subtracted it from the spectrum. We measured the signal-to-noise ratio (S/N) of the lines of interest based on the line flux and root-mean-square residuals of the continuum. We fit the lines with a Gaussian profile to determine the flux, velocity (shift from the systemic velocity, 11,400 km s⁻¹), and velocity dispersion (sigma of the Gaussian). In order to gauge the accuracy of the line-fit parameters, synthetic data cubes were then generated using the line parameters fitted to each pixel. The S/N ratio measured above determined the random noise level added to the synthetic cube. We then refit the line parameters of the synthetic cube 500 times with random noise added. The final uncertainty in each parameter's original measurement was obtained from the distribution of that of the resulting models.

We adopted this procedure for each line species. Different lines are treated separately from each other, but all the H₂ transitions or both the Brγ and Brδ lines are fitted simultaneously. An example of the simultaneous fit of the H₂ transitions for a typical pixel in the Kcb cube is shown in Figure 2. Visual inspection ensured that the line fits were sensible with appropriate continuum levels determined within each spectral window. We noted that in many of the pixels with [Si vi] detection, the fit of a single Gaussian profile resulted in large residuals. In such cases we fit a broad and a narrow Gaussian component to these lines; this significantly improved the residuals (Figure 3). The pixels associated with the two components appeared to be spatially segregated. The physical implication of this double-component fit is further discussed in Section 4.2.4.

**Figure 2.** Simultaneous fitting of the different H₂ transitions for an arbitrarily selected pixel as detected in the Kcb data cube. The thin grey line represents the data spectrum; the thick black lines represent the model fits. For each spectrum, all H₂ lines are required to have the same velocity and velocity dispersion; flux is fitted separately for each line. A similar fitting routine is used for Brγ and Brδ.
Download figure:
Standard image High-resolution image

**Figure 3.** A double-Gaussian component fit to the [Si vi] line profile in three regions of the integrated SE component. The top panel includes the entire SE component; while the middle and bottom panels split the spectrum into the northwest edge and remaining pixels. (These regions are identified in the middle panel of Figure 10.) In each panel, the black line represents the data spectrum, the green dashed and red dotted lines represent the two components that contribute toward the total model fit in solid blue. In the total spectrum, the [Si vi] line clearly shows two velocity components: one with a narrow velocity dispersion and one with a broader velocity dispersion. We find by splitting the spectrum up spatially that the broader velocity component comes mainly from the northwest edge of the SE component, while the narrow portion of the line appears more heavily in the remaining section. Parameters for each fitted component are listed in each panel.
Download figure:
Standard image High-resolution image

In order to increase the S/N ratio for regions with inadequate flux, we adopted optimal Voronoi binning for the investigation of the kinematics using the Cappellari & Copin (2003) code. This algorithm bins pixels together in a region until the emission line in that region achieves sufficient S/N, which we chose here to be 3. This allows us to explore the spatial resolution in areas with high S/N and to examine those regions with low S/N in meaningful spatial regions.

4. MAPS AND DATA CUBES

4.1. NIRC2 Near-infrared Images

Taken at the 0 farcs 01 pixel⁻¹ narrow-camera mode, the H- and K'-band NIRC2 images provide an unprecedented high-resolution view of the nuclear components in these wavelengths (Figure 4). All three components, as denoted by the crosses corresponding to the N, SW, and SE, are distinct in both bands.

**Figure 4.** Keck NIRC2 H- (left) and K'-band (right) images taken with the narrow camera (001 pixel scale). The two images are displayed in logarithmic scale in units of erg s⁻¹ cm⁻² Å⁻¹ with N pointing up. These images reveal clumpy structure and wispy gas in the nuclear regions of Mrk 273. Crosses denote the three components from Figure 1, and the arrows point to three different clumps resolved in the N component. See the text for a detailed discussion of each.
Download figure:
Standard image High-resolution image

Exhibiting a disky morphology different from that seen in the optical bands, the N component is both extended and clumpy in the near-infrared. We used GALFIT (Peng et al. 2002, 2010) to quantitatively characterize its morphology and find that the underlying structure is best fit by a central point source (which we call N1) plus a Sérsic profile of index n = 1.63 and effective radius r_eff = 0 farcs 69 ± 0.01, or 520 ± 8 pc, in the H-band continuum. In order to avoid the dusty region between the nuclei, we performed the GALFIT analysis on the eastern half of the disk only (see Figure 5). The residual map shows some underlying substructure in the disk, but otherwise a Sérsic disk plus central point source model (Figure 5, top) were a better fit for the N component than one without the central point source (Figure 5, bottom). In Figure 4, approximately 0 farcs 2 to the northeast of N1 is N2, which looks like a faint, compact clump. It is brighter in the K' band than it is in the H; the difference may be due to dust extinction. It appears distinct and could plausibly be the site of a stellar cluster. It closely coincides with the N2 as designated by Carilli & Taylor (2000), at the location of a hypothesized clump of supernovae. To the southeast of N1 at a distance of ∼0 farcs 15 is N3, which is extended and elongated in the E–W direction. It may be diffuse emission associated with N1 given its proximity, or it may be a separate clump. There is a larger difference in the brightness of N1 and N3 in the K' band relative to that in the H band, indicating that N1 is redder than N3.

**Figure 5.** Example of fitting results of GALFIT: here, the N (top row), SW (second row), SE (third row), and N (without central PSF in model; bottom row) components in the NIRC2 H-band image have been fitted with a Sérsic, Sérsic, Gaussian, and Sérsic profile, respectively. The leftmost column shows the input data that were fed into GALFIT, the middle column demonstrates the best-fit model parameters, the rightmost column indicates the residual maps. In the case of the N and SW components, we performed GALFIT analysis on a mirrored version of the eastern and southern halves, respectively, in order to avoid the dusty region between the nuclei.
Download figure:
Standard image High-resolution image

Lying ∼1 farcs 0, or 750 pc, to the southwest from N1, the SW component is also resolved at these scales with a north–south extension (r_eff = 68 ± 8 pc in the H band and 38 ± 8 pc in the K' band). The high-resolution 0 farcs 01 pixel⁻¹ data further constrain the size of this compact emission. Detected more notably in the H band, the north–south extended feature appears asymmetric and elongated toward the south. The northern extension resembles a tidally disrupted feature, which suggests that it is material that has not yet settled during the course of the merger. The SW source appears the reddest in the frame, most likely due to dust. This extinction obscures it entirely in the Hubble Space Telescope (HST) B-band image (Figure 1).

Two distinct sources can be seen in the southeast region relative to the N disk. The southernmost one is compact and appears to be a star cluster. The one denoted with a cross (see Figure 4) is more diffuse and extended; it marks the site of the SE source seen in radio images (Carilli & Taylor 2000; Bondi et al. 2005). Its spatial extent is larger than the compact source to the south from which the PSF is confirmed. This suggests that the SE component may not simply be a stellar cluster as was suggested by the NICMOS images (Scoville et al. 2000). Given its proximity to the N component and the morphology of the low surface brightness feature between it and the N component, the SE source may be associated with the N component.

4.2. K Broadband Integrated OSIRIS Spectra

In the K broadband spectra (0 farcs 1 pixel⁻¹ resolution), only two spatially segregated components appear in the continuum of Kcb: the N extended emission and the SW compact source. The SE diffuse clump is very weak in the integrated image. Figure 6 shows the integrated spectra of the N, SW, and SE components, respectively. These spectra exhibit very different features. The SW spectrum is dominated by the continuum and has strong [Si vi] 1.964 μm and H₂ emission; other emission lines are present but weak. The N spectrum features multiple extremely strong H₂ transitions along with Brγ and He i. The SE component is weak in the continuum but exhibits emission in the H₂, [Si vi], and Brackett lines as well.

**Figure 6.** OSIRIS integrated spectra of the N component (top), the SW component (middle), and the SE component (bottom) from 50 minutes of on-source exposure, taken in K band with the 01 pixel⁻¹ lenslet. Lines of interest are labeled. The telluric-corrected continuum in the SW still shows a steeper gradient than in the N and SE, indicating the presence of hot dust. The SE spectrum shows comparatively little continuum emission but clear evidence of line emission. Most emission lines in the SW are relatively weak. On the other hand, [Si vi] appears strongly in the SW and the SE but not in the N component.
Download figure:
Standard image High-resolution image

Line parameters of the various regions such as the effective radii for the fitted Sérsic component or FWHM of the fitted Gaussian component, axis ratios, integrated flux ratios, mean velocities, and velocity dispersions are listed in Table 1. The integral-field capability of OSIRIS allows the examination of the spatial distributions and kinematics of these emission lines individually; we present these results in the following section.

Table 1. Parameters of the Different Regions

Region	Line	r_eff	r_eff	Axis Ratio	Flux Ratio	$\bar{v}$	σ
Region	Line	(arcsec)	(pc)	Axis Ratio	Flux Ratio	(km s⁻¹)	(km s⁻¹)
N	H_2c	0.27 ± 0.01	204 ± 8	0.75 ± 0.01	1.15 ± 0.01	6 ± 2	225 ± 2
	Brγ	0.30 ± 0.06	226 ± 45	0.75 ± 0.01	0.48 ± 0.02	21 ± 7	228 ± 7
	He i	0.27 ± 0.01	204 ± 8	0.75 ± 0.01	0.22 ± 0.02	45 ± 15	230 ± 20
	[Si vi]	..	..	..	..	..	..
	[Fe ii]	0.40 ± 0.04	302 ± 30	0.78 ± 0.01	0.13 ± 0.01	6 ± 4	213 ± 9
	H-band continuum	0.69 ± 0.01	520 ± 8	0.74 ± 0.01	..	..	..
	K'-band continuum	0.80 ± 0.01	603 ± 8	0.88 ± 0.01	..	..	..

SW	H_2c	..	..	..	0.29 ± 0.03	50 ± 15	180 ± 15
	Brγ	..	..	..	..	..	..
	He i	..	..	..	..	..	..
	[Si vi]	0.14 ± 0.02^⋆	106 ± 15	0.08 ± 0.2	0.37 ± 0.06	−290 ± 35	220 ± 50
	[Fe ii]	..	..	..	0.04 ± 0.04	7 ± 30	155 ± 45
	H-band continuum	0.09 ± 0.01	68 ± 8	0.41 ± 0.01	..	..	..
	K'-band continuum	0.05 ± 0.01	38 ± 8	0.47 ± 0.01	..	..	..

SE	H_2c	0.14 ± 0.03^⋆	106 ± 24	0.56 ± 0.05	0.59 ± 0.02	−117 ± 4	173 ± 4
	Brγ	0.07 ± 0.04^⋆	53 ± 30	0.42 ± 0.05	0.30 ± 0.03	45 ± 20	240 ± 20
	He i	..	..	..	..	..	..
	[Si vi]	0.16 ± 0.02^⋆	121 ± 15	0.66 ± 0.09	0.29 ± 0.03	−125 ± 20	150 ± 15
	[Fe ii]	0.06 ± 0.01^⋆	45 ± 8	0.23 ± 0.05	0.06 ± 0.04	−10 ± 40	205 ± 60
	H-band continuum	0.06 ± 0.01^⋆	45 ± 8	0.61 ± 0.04	..	..	..
	K'-band continuum	0.13 ± 0.01^⋆	98 ± 8	0.51 ± 0.01	..	..	..

Bridge	H_2c	0.35 ± 0.03^⋆	264 ± 23	0.68 ± 0.04	0.73 ± 0.02	−125 ± 4	175 ± 5
	Brγ	0.38 ± 0.14^⋆	287 ± 106	0.16 ± 0.06	..	..	..
	He i	..	..	..	..	..	..
	[Si vi]	0.13 ± 0.03^⋆	98 ± 23	0.17 ± 0.06	..	..	..
	[Fe ii]	..	..	..	0.10 ± 0.01	−110 ± 15	240 ± 20
	H-band continuum	..	..	..	..	..	..
	K'-band continuum	..	..	..	..	..	..

Notes. r_eff is the effective radius for the fitted Sérsic component; in the ^⋆ cases, this is the FWHM/2 of the fitted Gaussian component. The symbol ".." indicates weak or no detection.

Download table as: ASCII Typeset image

4.2.1. H₂

Five vibrational transitions of molecular hydrogen lie within the Kcb coverage of OSIRIS: 1–0 S(3) (λ_rest = 1.9576 μm), 1–0 S(2) (λ_rest = 2.0338 μm), 1–0 S(1) (λ_rest = 2.1218 μm), 1–0 S(0) (λ_rest = 2.2235 μm), and 2–1 S(1) (λ_rest = 2.2477 μm) (hereafter H_2a, H_2b, H_2c, H_2d, and H_2e, respectively). These vibrationally excited emission lines trace the warm molecular gas, and result from a mixture of thermal and nonthermal components, where the excitation mechanisms may involve shock-heating, ultraviolet (UV) photoioniziation by O and B stars, or X-ray ionization (Mouri 1994). The intensity ratios among these lines can be used to distinguish between the thermal and non-thermal excitation mechanisms due to the difference in their efficiencies in populating the various vibrational levels.

Shock-excited H₂ emission has been observed in many galaxies; a notable object among these observations is the prototypical late-stage merger NGC 6240 with dual AGN at L_IR = 10^11.93 L_☉, for which the thermal H₂ temperature was determined to be 2000 K (Mouri 1994). Draine & Woods (1990) further proposed that most of the H₂ line emission in NGC 6240 is attributed to X-ray irradiation from either supernova explosions or high-velocity shock waves associated with the merger. Like NGC 6240, Mrk 273 displays prominent emission in H_2c. Figure 7 shows the unbinned flux, Voronoi-binned velocity and velocity dispersion maps of H_2c along with the distribution of the K-band continuum in contours. The H_2c flux is extended in the N component. In addition, the flux bridges toward the SE component, which is the site of radio continuum emission (∼1'' southeast of the N peak).

A velocity gradient typical of an inclined rotating disk can be seen in the velocity map of H_2c (middle panel in Figure 7). The range of the velocity is [−206, 220] km s⁻¹; the rotational axis of the rotating disk has a P.A. of 330°. The velocity map spans a larger region beyond the extent of the N disk. Of particular note are the negative velocities extending toward the SE component along the bridge in the emission. Along the major axis of the disk, we measure a velocity dispersion of ∼140 km s⁻¹; along the minor axis and toward the SE, this increases to ∼200 km s⁻¹ (right panel in Figure 7). This lends evidence to the presence of biconical collimated turbulent outflows from the center of the N source; the significance of this outflow is further discussed in Section 5.

4.2.2. Excitation Mechanisms of H₂

Multiple H₂ transitions are prominent in the Kcb data, particularly within the N disk. Using equations from Reunanen et al. (2002) and Rodríguez-Ardila et al. (2004, 2005) that make use of the ratios among the various H₂ transitions (e.g., H_2e/H_2c and H_2b/H_2d), we can gain insight into the temperature conditions within Mrk 273's nuclei. The rotational temperature does not vary significantly across the nuclear regions. Integrating over the regions defined by the different components, we derive for the N disk a vibrational temperature T_vib ≃ 2590 ± 80 K and a rotational temperature T_rot ≃ 1605 ± 115 K. The measured T_vib is consistent with that measured for a sample of Seyferts (T_vib ≲ 2600 K) from Reunanen et al. (2002). In addition, the fact that the values for T_vib and T_rot differ appreciably suggests that the thermal excitation may be due to fluorescence (Rodríguez-Ardila et al. 2004). The computed H₂ line ratios along with derived temperatures for the various components are listed in Table 2.

Table 2. H₂ Line Ratios

Line Ratios	N	SW	SE	Bridge	Total (Mean)	Total (Median)
H_2a/H_2c	1.19 ± 0.02	0.31 ± 0.07	0.55 ± 0.03	0.75 ± 0.04	0.70	0.75
H_2b/H_2c	0.45 ± 0.01	0.11 ± 0.05	0.27 ± 0.03	0.26 ± 0.03	0.27	0.27
H_2d/H_2c	0.29 ± 0.01	0.15 ± 0.05	0.17 ± 0.02	0.20 ± 0.03	0.20	0.20
H_2e/H_2c	0.18 ± 0.45	0.04 ± 0.12	...	...	0.11	0.18
T_vib (K)	2590 ± 80	2535 ± 740	2240 ± 230	1925 ± 305	2325	2535
T_rot (K)	1605 ± 115	740 ± 150	1645 ± 355	1230 ± 205	1305	1605

Download table as: ASCII Typeset image

H₂ gas in galaxies can be excited by a number of mechanisms in the interstellar medium: UV fluorescence, shocks, and X-ray heating (Shull & Beckwith 1982, and others). A diagnostic involving the ratios of different H₂ transitions can be used to distinguish among the various means of exciting the gas (Rodríguez-Ardila et al. 2004). The H_2e/H_2c and H_2a/H_2c ratios in these nuclear regions are consistent with slow shocks (v_s ∼ 10–14 km s⁻¹) being driven into very dense gas (n = 10⁵ cm⁻³; Shull & Hollenbach 1978).

4.2.3. Atomic Hydrogen and Helium

After the various transitions of H₂, the Brγ line (λ_rest = 2.166 μm) and the He i line (λ_rest = 2.059 μm) are the next most prominent emission lines in the N spectrum. Brγ is an indicator of the ionizing radiation field and stellar activity, and has been detected in many Seyfert galaxies (e.g., Riffel et al. 2006). The ionization energy of atomic helium is 24.6 eV; therefore, He i line emission is expected to originate from regions with the most massive and youngest stars (Böker et al. 2008).

Figures 8 and 9 show the flux, velocity, and velocity dispersion of the Brγ and He i lines, respectively. The Brγ flux, and similarly the He i flux, is detected in the N disk and appears clumpy around the center; this may be due to clumpy star formation in the circumnuclear regions. The depression in flux of these atomic gases at the nucleus may suggest heavy obscuration by dust or ionization of the gas by a strong source like an AGN. The bulk of the Brγ gas coincides with the N continuum emission. There is very little Brγ emission at the SW component, but there is Brγ emission extending toward the SE source, albeit much weaker than in the H₂ lines. There is also faint He i emission in the vicinity of the SE source.

**Figure 8.** Same as Figure 7, but for Brγ (N points up). Brγ traces the same rotation in the N component as H_2c, but is much weaker along the bridge to the SE component.
Download figure:
Standard image High-resolution image

**Figure 9.** Same as Figure 7, but for He i (N points up). He i traces the same rotation in the N component as H_2c and Brγ, but is much weaker along the bridge to the SE component. As with Brγ, He i flux does not peak at the center of the N component, perhaps due to extinction or a change in the ionizing radiation.
Download figure:
Standard image High-resolution image

Brγ and He i also display similar kinematics. The velocity structure of a strongly rotating disk similar to that of H₂ is found for Brγ and He i. The centers of the rotating gas disks coincide with the peak of the continuum flux. The velocity dispersion at the disk is ∼100 km s⁻¹ but increases to ∼170 km s⁻¹ at the center. We see no evidence of a broad-line region.

4.2.4. [Si vi]

[Si vi] (λ_rest = 1.964 μm), a near-infrared tracer of AGN activity due to its high excitation potential (= 167 eV; Rodríguez-Ardila et al. 2004), has a very different spatial distribution than the other species, as seen in Figure 10. A fit to the line in the SE region with a Gaussian profile returns poor residuals. A visual inspection prompted us to add a second velocity component which significantly improved the fit (Figure 3). Two kinematically segregated components in this region are traced by [Si vi]: a narrow component with velocity v_n = −162 ± 6 km s⁻¹ and velocity dispersion σ_n = 99 ± 8 km s⁻¹; and a broad component with v_b = 65 ± 35 km s⁻¹ and σ_b = 325 ± 25 km s⁻¹.

**Figure 10.** Same as Figure 7, but for [Si vi] (N points up). [Si vi] is undetected in the N component, but appears strongly in the SE and SW components. In the SW, [Si vi] is broad, but see Figure 3 for a detailed analysis of the SE [Si vi] spectrum, which shows two kinematic components. The two kinematically and spatially segregated regions in Figure 3 are marked in the center panel: the NW edge is outlined in solid black, the remaining SE pixels are outlined in dotted black line. The NW edge shows broader kinematics than the dotted region.
Download figure:
Standard image High-resolution image

The bulk of the [Si vi] emission is found not in the N disk, likely due to heavy extinction, but at the bridge and the SE component. In the unbinned flux map (Figure 10), it extends N–S and spans a 0 farcs 189 × 0 farcs 163 (143 pc × 123 pc) region. Its velocity and velocity dispersion are highest near the bridge, though its relatively low S/N ratio limits the precision with which the location of the emission can be determined. [Si vi] is also strongly detected at the SW component, the site of the hard X-ray AGN (Iwasawa et al. 2011a). The peak position of the [Si vi] flux roughly coincides with that of the continuum, with v = −290 ± 30 km s⁻¹ and σ = 222 ± 50 km s⁻¹.

4.3. H-band Integrated Spectra

At a scale of 0 farcs 035 pixel⁻¹, the Hn4 continuum emission maps the resolved, clumpy N disk and compact SW component. Figure 11 shows the integrated spectra of the N, SW, and SE components, respectively, highlighting the [Fe ii] line found in all three regions. The [Fe ii] line is typically broad (σ ∼ 200 km s⁻¹) and prominent in the N and SE components, and dominated by the continuum in the SW. The SE component is weak in the continuum. Line parameters of the regions for the [Fe ii] line emission can be found in Table 1.

4.3.1. [Fe ii] Emission

In the Hn4 band, the [Fe ii] line at [λ_rest = 1.644 μm] shows strong emission in the center of Mrk 273. Because Fe is strongly depleted in the interstellar medium, gas-phase [Fe ii] emission (with ionization potential of 16.2 eV; Mouri et al. 1993) in galaxies has been used as a sensitive indicator of shocks (Graham et al. 1987). The potential origins for the [Fe ii] emission include regions of partially ionized gas within supernova remnants in starburst nuclei and within narrow-line regions in AGN (Mouri et al. 1993). The contribution from photodissociation regions near O and B stars is negligible. Extracting slices associated with the [Fe ii] line from the Hn4 data cube, we present the flux, velocity, and velocity dispersion maps in Figure 12 with continuum contours for comparison. Like the continuum, the line data are resolved into multiple clumps in the N disk, but the emission is relatively weak at the N1 site, dominant at N2, and detected at N3. The differences in the line-to-continuum flux ratios between N1 and the other clumps (N1: 0.016, N2: 0.022, N3: 0.022) may suggest that they have different origins (see Section 5). The [Fe ii] line flux is negligible in the SW nucleus but very bright in the SE, with a slightly N–S elongated morphology. The N disk contains more total integrated [Fe ii] flux, but the SE peak dominates the surface brightness in the entire inner kpc region. The velocity structure of the [Fe ii] line also indicates strong rotation (±240 km s⁻¹ centered on the systemic velocity), with high velocity dispersion near the center of the N disk and the strongest velocity dispersion at the SE component.

**Figure 12.** Same as Figure 7 but for [Fe ii] (λ_rest = 1.644 μm) from the Hn4 band (N points up). The [Fe ii] velocity map shows strong rotation about the N component, with a small increase in velocity dispersion at the center. [Fe ii] flux is clumpy in the N component, resolving N1, N2, and N3 due to high resolution as in the NIRC2 images (Figure 4). However, the fact that N1 is weaker than N2 and N3 could be due to extinction or to differing ionizing radiation fields. The strongest and broadest [Fe ii] appears in the SE component.
Download figure:
Standard image High-resolution image

5. DISCUSSION

The mechanisms that power the enormous luminosities in Mrk 273 are likely to be associated with the three nuclear components situated within the dusty core. Here we discuss the nature of these regions.

5.1. The Northern Disk

The nature of the N component has been controversial due to different interpretations of the multiwavelength data: clustered radio supernovae (Smith et al. 1998), Seyfert 2 AGN (Xia et al. 2002), and a starburst with the possibility of an additional Compton-thick AGN from excess 6–7 keV emission that might be associated with a high equivalent-width Fe k emission (Iwasawa et al. 2011a). The Fe k line as indicative of a heavily obscured AGN is also present in a dual-AGN LIRG merger Mrk 266, where the LINER nucleus exhibits enhanced shock-excited H₂ emission and diminished coronal line compared to the Seyfert nucleus (Mazzarella et al. 2012). Here, our OSIRIS data cube provides evidence of a strongly rotating gas disk through the distribution and kinematics of the near-infrared emission line gas. The strong H_2c flux suggests localized shocks and fits well with the models of clustered radio supernovae (Smith et al. 1998), lending support to their hypothesis that a starburst provides much of the power. Carilli & Taylor (2000) have resolved individual compact sources that may be supernova remnants; they suggest that the diffuse emission may be a result of excitation from the remnants. Our OSIRIS data resolve the individual clumps and provide spatial information on the physical conditions of the ionizing source in the N disk. We consider the evidence in this section.

5.1.1. Strong Rotation in Emission Lines

The velocity maps for the emission lines (H₂, Brγ, He i, and [Fe ii]) detected in the N disk show structure consistent with a strongly rotating disk. The deprojected rotational velocity is ∼ ±240 km s⁻¹, and its structure is indicative of Keplerian rotation about a large mass (see Section 5.1.2). We suggest that the rotating disk in the N is the remnant nucleus of a progenitor galaxy in this merger system, with its gas rotating about a central supermassive black hole or a very massive central star cluster. In the next section we use dynamical modeling to determine the enclosed mass.

5.1.2. Black Hole Mass

From the strong rotation inferred from the emission lines, we measure the kinematics to understand the underlying mass profile. Though rotation in the N disk is seen in every emission line, we focus on the [Fe ii] kinematics because these data have significantly higher spatial resolution, by a factor of three, compared to the other emission lines. To measure the dynamical mass of the central source, it is crucial that we resolve within the black hole's sphere of influence. The radius of the sphere of influence is given by

$\begin{equation} r_{\rm BH} = \frac{{\it GM}_{\rm BH}}{\sigma ^2}, \end{equation} \tag{ 1 }$

where G is the Gravitational constant, M_BH is the mass of the black hole, and σ is the velocity dispersion. We assume a conservative estimate for the mass of the black hole to be M_BH ∼ 10⁹ M_☉, given that Klöckner & Baan (2004) had measured a mass of 1.39 ± 0.16 × 10⁹ M_☉ from radio interferometry observations of OH masers. The circumnuclear velocity dispersion we measure from the [Fe ii] line is ∼215 km s⁻¹. Therefore, the radius of the sphere of influence is at least r_BH ⩾ 95 pc. Given our resolution element of 0 farcs 08, or 60 pc, there are 23 spaxels for which the enclosed mass is dominated by the central point mass (see model description below). These spaxels are sufficient to provide a lever arm on black hole mass and, when combined with the surrounding spaxels, enable a reliable dynamical mass measurement from the [Fe ii] line kinematics.

We follow the Keplerian disk fitting method outlined in Medling et al. (2011) to compare the measured kinematics to those of simple models. First, we create velocity maps using $v = \sqrt{G M /r}$ and assuming a central point mass; we also solve for the inclination of the disk and the position of the central point mass. We use the Levenberg–Marquardt least squares fitting routine MPFITFUN (Markwardt 2009; Moré 1978) to select the parameters that best fit the [Fe ii] velocity map. We fix the central mass position for generalization to a more complete mass profile. In our second set of models, we estimate the stellar mass profile by fitting a disk to the continuum emission. We subtract off the central point source from the center of the continuum light before fitting. We fit this using the simple radially varying surface density profile Σ(r) = Σ₀r^−γ, making the mass in a ring 2πΣ₀r^{1 − γ}dr. We fit the inclination, γ, and Σ₀, and find that the fitted inclination is consistent with the inclination from the first dynamical model (55°); we fit a γ of 1.0. In our final set of models we adopt $v = \sqrt{{\it G M}(r) / r}$ , with the mass profile containing a single point mass plus a radially varying disk mass of surface density Σ(r) = Σ₀r^−γ. We fix the inclination and the exponent γ from our previous two models, and vary only the point mass and the disk mass density normalization Σ₀. (Note that we allow the last to vary in order to avoid making assumptions about mass-to-light ratio (M/L). On such small scales, dark matter should be negligible, but estimating a M/L ratio for a stellar population of unknown age with unknown extinctions would introduce more uncertainties than allowing it to be a free parameter.) In each set of models, we convolve our light-weighted model with the PSF (see discussion in Section 2) before comparing it to our observations. Errors on parameters have been estimated using a Monte Carlo approach, refitting the black hole mass from a model with 100 iterations of statistical errors and using the resulting distribution.

A comparison between the measured [Fe ii] velocity map and the kinematic models is found in Figure 13. The results illustrate that a disk model with a central point mass turns out a better fit than one without. We find that the best-fit central point mass is 1.04 ± 0.1 × 10⁹ M_☉. This mass is consistent with that measured from radio interferometry of an OH maser, 1.39 ± 0.16 × 10⁹ M_☉ (Klöckner & Baan 2004), and is similar to that of the southern black hole of NGC 6240 (8.7 × 10⁸ M_☉ < M_BH < 2.0 × 10⁹ M_☉), another late-stage galaxy merger (see Medling et al. 2011). The mass resemblance to the NGC 6240 black hole suggests that this central point mass may likely be a supermassive black hole. From the kinematics, we cannot determine whether Mrk 273's N disk hosts a quiescent black hole or a Compton-thick AGN as suggested by Iwasawa et al. (2011a). In either case, the presence of the disk and supermassive black hole confirms that this is the remnant nucleus of one progenitor galaxy.

Though we have shown from our GALFIT analysis that the infrared continuum light profile is best fit with a model of Sérsic disk along with a central point source, we note the possibility that the central mass may be an unresolved massive star cluster instead of a supermassive black hole. Given that massive star clusters found in NGC 6240 have photometric masses of 7 × 10⁵–4 × 10⁷ M_☉ (Pollack et al. 2007) and those found in the Milky Way have upper mass limits of 10⁸ M_☉ (Murray 2009), ⩾10–30 of the most massive of these clusters must reside within Keck's resolution limit to feature the observed kinematics. The two most massive clusters described in Murray (2009) have masses of 10⁸ M_☉ and measured radii of ∼100 pc, suggesting that a cluster 10 times as massive and in a much smaller volume would be improbable. Since it is implausible that 10⁹ M_☉ of stars would be dynamically stable within this compact region, we conclude that the central mass must be a supermassive black hole.

5.1.3. Spatial Distribution of Line Emission

Brγ, which traces ionized gas, and He i, which traces neutral gas, both appear in the N disk. These species both have moderately low velocity dispersion in the disk, and show no sign of being associated with a broad-line region. The coronal line [Si vi] is virtually non-existent in the N disk. However, 1.4 GHz images from Carilli & Taylor (2000) suggested that there may be a weak AGN present near N1. Similarly, the 6–7 keV extended emission detected near around the N nucleus may be due to enhanced 6.4 keV Fe k line emission, lending support to the presence of a heavily obscured AGN (Iwasawa et al. 2011a). We measure an increase in the velocity dispersion for H₂ along the minor axis and toward the SE, which gives direct evidence for the presence of biconical collimated molecular outflows from the N source.

There is a deficiency of Brγ and He i flux near N1 compared with the rest of the N disk. This could be due to ionization caused either by an AGN or by supernovae in the center of the disk. However, we find other putative supernovae sites that do not show a deficiency in Brγ, so factors such as extinction may contribute toward this phenomenon near N1. Our Kcb line data are unable to fully resolve the individual clumps in the N disk, but the [Fe ii] line and NIRC2 images resolve them at the 0 farcs 035 and 0 farcs 01 scales, respectively. In particular, N1 is bright in the K'- and H-band flux but relatively weak in [Fe ii], whereas the faint clump N2 and extended clump N3 have a higher [Fe ii]-to-continuum ratio. N3 has similar brightness as N1 in the H band, but is weaker than N1 in the K' band. The presence and clumpiness of [Fe ii] may indicate regions excited by supernova remnants (Alonso-Herrero et al. 1997). However, since N1 is weak in [Fe ii] and redder in color than N2 and N3, its ionizing source may be similar to the AGN in the SW component. Given that the supermassive black hole measured dynamically in Section 5.1.2 lies inside N1, we note the possibility that it may be an obscured AGN.

5.2. The Southwestern Nucleus

The SW component is at the location of the hard X-ray source from an updated Chandra map (Iwasawa et al. 2011a). It is red and unresolved in NICMOS imaging (Scoville et al. 2000), and shows 8.4 GHz flux consistent with a radio-quiet AGN (Condon et al. 1991). In our OSIRIS data, this SW region is compact and dominated by the continuum emission. It shows strong [Si vi] and moderate H₂ emission, but is weak in Brγ flux. At the 0 farcs 01 pixel⁻¹ scale of the NIRC2 data, the SW source is resolved and exhibits a N–S extended, asymmetric low surface brightness structure. The presence of hot dust as inferred from the rising continuum slope agrees with the redness seen at this source in the NICMOS JHK-band images (Scoville et al. 2000) as well as with the Spitzer Infrared Spectrograph (IRS) results (Armus et al. 2007), consistent with the presence of an AGN.

The SW nucleus features low detection from H₂, Brγ, and He i, all potentially partially destroyed by a strong ionizing source. [Si vi] emission is detected in this region, but is not resolved. It usually indicates the presence of the hard ionizing spectrum of an AGN, consistent with the Chandra results that an AGN resides in this source (Iwasawa et al. 2011a).

5.3. The Southeastern Component and its Bridge

Though the SE component has been observed in several bands, its nature has remained unclear and controversial. From HST Advanced Camera for Surveys (ACS) and NICMOS images, the SE component resembles a star cluster (Scoville et al. 2000) and is identified as a candidate blue star cluster (CSC1; Iwasawa et al. 2011a). Radio images show prominent emission and extended morphology consistent with AGN jets, but lack the compact core indicative of an AGN. The radio data show a steep spectrum (α = 1.4 ± 0.2), which could point to a starburst origin (Carilli & Taylor 2000; Bondi et al. 2005), though the extended nature (∼230 pc) suggests that it is larger than a single star cluster.

Our spectral decomposition of the SE component in the near-infrared adds new information. The SE component appears weak and diffuse in the NIRC2 H- and K'-band continuum images, but does appear clearly in several emission lines: H₂ (Figure 7), [Si vi] (Figure 10), and [Fe ii] (Figure 12) in our IFU data. In particular, we note the strong [Si vi] emission; as [Si vi] has the highest ionization potential of the observed lines (167 eV), its presence gives a strong clue to the origin of the SE component. The use of [Si vi] as a tracer of the coronal-line region (CLR) has been demonstrated by Müller-Sánchez et al. (2011) in seven nearby Seyfert galaxies. They found that the CLRs in these low-power Seyferts to be ∼80–150 pc in radius and are associated with outflows, providing circumstantial evidence for the existence of extended CLRs and highly ionized gas in outflows from Seyfert nuclei like that in Mrk 273. Here we consider two possible mechanisms to ionize [Si vi]: photoionization and shock-heating. We first consider the [Si vi]/Brγ flux ratio in our models; the map of this ratio along with the radial profiles drawn in the direction toward the N and the SW components, respectively, are shown in Figure 14. We compare the [Si vi] flux to Brγ flux since the ratio may give clues to the ionization mechanisms involved. We find that, along the northwest boundary between the SE component and its bridge, the [Si vi]/Brγ ratio reaches 7.9 near the N nucleus, and decreases over ∼0 farcs 3 to 1.2 near the SE component. We observe a regression slope of 6.6 × 10⁻³, which is statistically significant with p-value = 0.03 by permutation. For comparison, we compute the regression slope for a line drawn toward the SW nucleus to be −3.9 × 10⁻⁴, which is not significantly different from zero. This SE–N line gradient therefore suggests that the north end of the SE component being more highly ionized, and that the origin of the ionization mechanism may be external; we consider below potential sources for the ionization.

**Figure 14.** Left: [Si vi]/Brγ ratio map with contours showing K-band continuum emission. The ratio appears strongest near the northwest boundary between the SE component and its bridge and decreases toward the SE component. This gradient (solid+dotted yellow line) suggests that the origin of the ionization mechanism may be external rather than local. For comparison, a dashed+dotted yellow line is drawn in the direction pointing to the SW nucleus. Bar denotes 05; N points up. Right: the radial profile of the [Si vi]/Brγ ratio plotted as a function of projected distance from the SE component. The solid and dashed lines correspond to those in the left panel.
Download figure:
Standard image High-resolution image

5.3.1. Photoionized [Si vi]

If the [Si vi] is photoionized, the hard ionizing radiation may come from either an AGN or nearby O and B stars. We compared our measured [Si vi]/Brγ flux ratio to those produced by CLOUDY models representing each scenario. We used CLOUDY version 08.00, described by Ferland et al. (1998).

First we confirmed that an AGN can plausibly produce our measured [Si vi]/Brγ flux ratio. Our incident AGN spectrum was a power-law continuum with a spectral index of −0.2, and we assumed that the gas is approximately solar metallicity (consistent with estimates from Huo et al. 2004). We varied the ionization parameter of the gas U = Q(H⁰)/(4πr²n_Hc), where Q(H⁰) is the flux of ionizing photons, r is the distance between the ionizing source and the illuminated edge of the cloud, n_H is the volume density of neutral hydrogen to ionize, and c is the speed of light. We found that, in order to produce a [Si vi]/Brγ flux ratio of ∼6, an ionization paramater U of at least 10^−2.7 is required. When varying the metallicity log(Z/Z_☉) from −0.5 to 1.0, we found that the required ionization parameter varies by less than 0.3 dex. We used n_H ∼ 10³ cm⁻³ as a conservative estimate of the density in the central regions of ULIRGs (but see Davies et al. 2003, whose models suggest the densities can be as high as 10⁴–10⁵ cm⁻³). We found that the required Q(H⁰) ∼ 4 × 10⁵⁴(r/750 pc)² s⁻¹. A typical Q(H⁰) for a quasar can reach nearly 10⁵⁷ s⁻¹ (Tadhunter 1996); lower levels of AGN activity would naturally produce fewer ionizing photons. These numbers suggest that photoionization from an AGN in the N disk could plausibly produce our measured [Si vi]/Brγ flux ratios, even at a distance of 750 pc. We note that both the N and the SW components, at distances of 640 and 750 pc, respectively, fall within this range. However, the spatial distribution of the [Si vi]/Brγ flux ratio (see Figure 14, where the gradient seems to be higher toward the N nucleus) is more consistent with photoionization from a N AGN than from the SW unless there is variable dust extinction surrounding the SE component that alters the angle of the gradient.

We then determine if our measured [Si vi]/Brγ flux ratios could instead be produced by nearby O and B stars due to intense star formation. To test this scenario, we used an incident O star spectrum from the TLUSTY models described in Lanz & Hubeny (2003). As an extreme case, we chose two of the hottest O star spectra available, with effective temperatures of 55,000 K and 52,500 K, at solar metallicity and with log (g) of 4.5 and 4.25, respectively. Again we varied the ionization parameter U and the metallicity of the gas to find the required Q(H⁰) to produce our measured [Si vi]/Brγ flux ratios. In this scenario, we find that an ionization parameter U of at least ∼10¹² is required for our measured [Si vi]/Brγ flux ratio. This U value is much higher than is required for an AGN because the AGN spectrum is much harder than that of an O star. To produce this high of an ionization parameter from stars present within 5 pc requires Q(H⁰) ∼ 8 × 10⁶⁴(r/5 pc)² s⁻¹. A single O star's Q(H⁰) ∼ 10⁵⁰ s⁻¹ (see models by Sternberg et al. 2003); for this line ratio to be produced by photoionization from stars, we would need 10¹⁴ O stars within 5 pc—a completely unphysical scenario. We conclude that if our measured [Si vi]/Brγ flux ratios are caused by photoionization, they must be from an AGN.

5.3.2. Shock-heated [Si vi]

[Si vi] may be ionized in gas heated by shocks and is sometimes observed in novae and supernova remnants (Gerardy & Fesen 2001). Several pieces of evidence in our observations suggest the possibility of outflows that could be causing shocks near the SE component. The H₂ and [Fe ii] emission bridge, the increase in H₂ velocity dispersion along the minor axis, as well as the noted [Si vi]/Brγ gradient decreasing along the bridge from the N to the SE component, would be explained by outflows associated with the central powering source within the N nucleus in this scenario—AGN jets or stellar winds. Although the [Si vi] emission appears diffuse, the H₂ velocity map in this region shows a clear shift from the systemic velocity of the system with Δv ∼ −150 km s⁻¹ along the bridge and Δv ∼ −75 km s⁻¹ at the SE component. The spectrum of the H i 21 cm absorption shows a double-peaked profile suggesting infall at 200 km s⁻¹ on scales ⩽40 pc (Carilli & Taylor 2000). With this evidence of gas motion, it is likely that shocks may exist; these shocks could be responsible for some or all of the measured [Si vi]/Brγ flux ratio (⩽7.9 ± 3.4).

To determine if our observed [Si vi]/Brγ flux ratio may be plausibly caused by shocks, we compared it to the shock models of Allen et al. (2008) using the IDL SHOCKPLOT widget. In particular, we adopted the radiative shock plus photoionized precursor model with solar abundance and density n = 10³ cm⁻³ as a conservative estimate for a ULIRG, and calculated line ratios produced for models within a grid of parameter space. The parameter space in magnetic parameters B/n^1/2 and shock velocities v_s of the model grid is B/n^1/2 = 10⁻⁴–10 μG cm^3/2 and v_s = 200–1000 km s⁻¹. The model grid is shown in Figure 15, with a green symbol marking the value of the line ratios observed in the SE component; it does not appear to converge on any shock model generated within our parameter space. The more robust [Si vi]/Brγ ratio (which is less sensitive to dust or cross-filter photometric calibrations) is only produced by models with a high B/n^1/2 value and a shock velocity to v_s ⩾ 550 km s⁻¹, indicating that only fast shocks in the densest material, if at all, could produce the line ratios that are observed in the SE component.

**Figure 15.** Diagnostic plot of [Fe ii]/Brγ vs. [Si vi]/Brγ showing a grid of shock models with solar abundance and density n = 10³ cm⁻³ (Allen et al. 2008). The parameter space spans magnetic parameters B/n^1/2 = 10⁻⁴–10 μG cm^3/2 and shock velocities *v_s* = 200–1000 km s⁻¹. The green symbol marks the value of the line ratios associated with the SE component, and it does not appear to converge on any shock model generated within our parameter space. The more robust [Si vi]/Brγratio (which is less sensitive to dust or cross-filter photometric calibrations) is only produced by models with a high B/n^1/2 value and high shock velocity, suggesting that only fast shocks in the densest material, if at all, could be responsible for its excitation.
Download figure:
Standard image High-resolution image

5.3.3. Energy Flow from N1

The [Si vi]/Brγ ratio, seen as a gradient from the N and decreasing toward the SE, in combination with the bridge between SE and N seen in the other flux maps, strongly suggest an external origin for the source of the observed energy flow, possibly from the center of the N disk. We find that the SE component is likely to be gas from a tidal feature or stellar cluster that is being affected in one of two ways. The easiest way to ionize the [Si vi] is with a buried AGN in the N disk or the SW AGN with variable extinction. The AGN would not be isotropically obscured, and thus could photoionize gas in certain directions. However, we find that the observed [Si vi]/Brγ ratios may also be produced in extreme shock situations. Here we investigate if such shocks are consistent with our kinematic data, and whether star formation is able to provide sufficient energy to power them.

The spectral maps and kinematics in the region of the bridge suggest its origin as an outflow originating at the center of the N disk. It extends along the minor axis of the disk, following the region of increased velocity dispersion in molecular hydrogen, suggesting that a biconical turbulent outflow is present. The increased H₂ emission toward the SE has a peculiar velocity of −200 km s⁻¹ relative to the systemic velocity of the system; this gas appears to be outflowing. We used simple energy arguments to estimate the amount of energy input required, to determine whether it is likely driven by stellar winds or from AGN feedback. The energy contained in the outflow is 1/2mv², where m is the mass contained in the outflow and v is the peculiar velocity, −200 km s⁻¹. To estimate the mass contained in the outflow, we adopted solar metallicity and a density of 1000 cm⁻³ (a conservative estimate for (U)LIRGs; see Davies et al. 2003), and used the size of the extended emission (see Table 1) assuming an equal depth and width, resulting in a mass of 7700 M_☉. To determine an energy input rate, we estimated the amount of time the outflow has been traveling from the distance between the SE and the center of the N disk. To travel this distance of 640 pc at 200 km s⁻¹, the outflow requires ∼3.3 Myr. Assuming the energy contained in the outflow was spread out over this timescale, the energy input rate required to drive this outflow is 1.3 × 10⁴³ erg s⁻¹.

Is there enough star formation to provide this energy input in stellar winds? We estimate the available energy input from stellar winds as follows: the luminosity of Brγ in the N disk is 2.1 × 10⁴⁰ erg s⁻¹, implying a star formation rate of 17 M_☉ yr⁻¹ (Kennicutt 1998) and therefore a bolometric luminosity of 1.7 × 10¹¹ L_☉ (Krumholz & Tan 2007). Approximately 1% of the bolometric luminosity of a starburst is available to drive stellar winds (Heckman 1994), divided by 2 to account for the collimated biconical outflow, yielding an energy input rate available from the N disk's star formation of approximately 3 × 10⁴² erg s⁻¹. However, we note that the computed star formation rate may be underestimated since Brγ in the N disk may be subject to dust extinction. There may also exist an uncertainty of a factor of a few in the luminosity of starburst-driven winds. Thus, although the energy from star formation seems insufficient by a factor of four compared to the required energy input rate, starbursts cannot be definitively ruled out as a source of driving the winds.

It is also important to consider whether the energy available to drive outflows from star formation is sufficient to shock-heat the [Si vi] seen in the SE component. The incoming kinetic energy rate is $e_k = 1/2 \rho v_s^3A$ , where ρ is the density (as above, 1000 cm⁻³), v_s is the shock velocity (⩾550 km s⁻¹, calculated from the SHOCKPLOT models in Section 5.3.2), and A is the surface area of the gas being shocked. A lower limit on the shocked area is a hemisphere on the surface of the SE component: 2π (70 pc)². (In reality, the shock front would likely have structure and therefore larger surface area; this provides a conservative lower limit on the energy input required to shock this [Si vi].) These estimates require an energy input of 1.7 × 10⁴⁴ erg s⁻¹, two and a half orders of magnitude larger than 5 × 10⁴¹ erg s⁻¹, the available energy for outflows from star formation in the N disk. This suggests that the observed [Si vi] emission cannot be caused solely by shock-heating due to stellar winds. Stronger shocks may be plausible with AGN-driven outflows, but because shock models cannot produce the observed [Si vi]/Brγ ratio, at least some of the [Si vi] emission must be caused by AGN photoionization (as discussed in Section 5.3.1). This scenario is consistent with and supported by the nuclear bipolar superbubble outflow observed in neutral and ionized gas as potentially powered by an obscured AGN in the N nucleus (Rupke & Veilleux 2013).

Though photoionization from an AGN is the simplest way to produce the [Si vi] emission in the SE, it is likely produced by a combination of the two processes: e.g., a weaker AGN to photoionize some gas, and less extreme shocks to heat the rest. Our observed double-peaked [Si vi] velocity profile indicates that there are likely two kinematic components in this region—with broad and narrow velocity dispersions, respectively. The broad component contributes roughly half the [Si vi] flux, resulting in a [Si vi]/Brγ ratio of 3.9 ± 1.7, or 0.59 $^{+0.16}_{-0.25}$ dex in logarithmic units. This constrains the shock models to velocities v_s ∼ 350–450 km s⁻¹, within the range of the velocity dispersion of the broad component. The two narrow and broad components may have originated from two different heating mechanisms, lending support to the combined AGN and shocks scenario. A simple explanation might involve a combination of shocks from the outflow and photoionization by a buried AGN within the N disk.

The SW AGN is also a potential origin for the photoionizing flux, even though it is further in projection from the SE component. In this alternative scenario, the observed spatial gradient in the [Si vi] excitation toward the N nucleus could be caused by differential extinction in the dust between the SW and the SE components, by differential gas densities within the SE component, or by geometric projection effect. Current observations are unable to distinguish between these proposed scenarios, and thus further investigations will be needed to confirm the nature of the ionizing source within Mrk 273's nucleus.

6. SUMMARY

High spatial resolution imaging and integral-field spectroscopy with Keck LGS AO has allowed us to study the properties of the atomic and molecular emission lines of Mrk 273 in the H and K' bands. The nuclear region is resolved into three primary components (N, SW, and SE) with a gas bridge between the N and SE components. These near-infrared high-resolution maps are well suited for investigating the gas dynamics and the location of the hardest photons from AGN and fast shocks; they complement other multiwavelength data for determining the ultimate power associated with each component.

We have shown that the N component is a rotating gaseous disk with P.A. ∼330°. The characteristic size of the disk varies depending on the emission line studied, but the effective radius is roughly 0 farcs 30 (240 pc). Using the observed strong rotation, we measure from the kinematics of [Fe ii] a central mass of M = 1.04 ± 0.1 × 10⁹ M_☉, consistent with the mass measurement from OH maser kinematics (Klöckner & Baan 2004). This enclosed mass likely indicates the presence of a supermassive black hole at the center of the N disk, suggesting that the N component is the remnant nucleus of a progenitor galaxy. The gaseous disk may be a remnant from the progenitor galaxy or may be recently formed with new gas from the merger. The black hole is heavily obscured and may be an AGN and/or surrounded by additional star formation, consistent with conclusions from X-ray (Iwasawa et al. 2011a) and radio data (Carilli & Taylor 2000; Bondi et al. 2005). An increase in the velocity dispersion of H₂ along the minor axis of the N disk and toward the SE suggests the presence of bipolar molecular outflow originating from the N source. Extended emission of shocked H₂ toward the SE also reveals the presence of a collimated outflow. Energy arguments associated with the size and kinematics of this outflow suggest that it is too massive to be produced by star formation alone, and also point toward the scenario in which the N black hole is a buried AGN. H₂ line ratios offer insight into the rotational and temperature conditions as well as their excitation mechanisms. We derived T_vib ≃ 2591 ± 79 K and T_rot ≃ 1604 ± 113 K, which fall within the range of temperatures for Seyfert galaxies (Reunanen et al. 2002).

The SW component is bright and compact in the K'- and H-band continua, and shows emission of the coronal line [Si vi]. We suggest that a non-thermal photon source is ionizing the silicon and producing the [Si vi] emission that we observe; it would also heat the hot dust that we observe in the continuum. Our near-infrared data support the notion that the SW nucleus hosts the known hard X-ray AGN in this system (Iwasawa et al. 2011a).

The SE component presents a new piece in the puzzle of the merger's nuclear region. While it has previously been detected in the radio regime (Smith et al. 1998), it has not been detected in optical and near-infrared images until now due to the dominance of the broadband continuum. With our new AO-assisted NIRC2 images and OSIRIS integral-field spectroscopy, the SE component is detected in several emission lines, and is prominent in both [Fe ii] and [Si vi]. The double-Gaussian fit to this [Si vi] line indicates that there are two velocity components involved, perhaps due to outflows. We discussed several potential scenarios for the origin of the SE component: photoionization by AGN or O and B stars, shock-heating due to stellar winds or AGN-driven outflows, or a combination of both. We favor the case where the SE component is a clump of gas or tidal feature photoionized by an obscured AGN in the N disk and/or shock-heated by outflows from the same AGN and its surrounding star formation. Alternatively, the photoionizing flux affecting the SE component may originate from the SW AGN. Confirming the N black hole as a buried AGN and the nature of the source for the SE component's ionizing radiation will require further investigations. While AGN photoionization alone could produce the observed [Si vi]/Brγ ratio, only the most extreme shocks in the shock-heated scenario could account for these [Si vi]/Brγ values. We suggest that a combination of photoionization and shocks could best produce both the observed line ratios and the two kinematic components of [Si vi].

As a late-stage gas-rich merger, Mrk 273 offers a unique time in the merger sequence where the nuclei have not completely merged and yet one AGN has already been triggered. We have identified a second black hole, located in the center of the N disk, which may be an obscured AGN with supporting evidence of outflows. The gas disk is likely the remnant nuclear disk from one of the progenitor galaxies, or perhaps a young disk formed during the merger. An understanding of the unveiling of this second black hole as an unburied AGN may lead to a deeper understanding of the frequency of dual AGNs in merging gas-rich galaxies.

We thank the anonymous referee for a helpful discussion on alternative cases for photoionizing the SE component, and for various suggestions on strengthening the scenarios put forth in the paper. We acknowledge H. Inami and S. Stierwalt for providing their Spitzer-IRS spectra toward ease of target selection for our observing program, M. Ammons for help with data reduction and PSF estimation, R. da Silva for help with CLOUDY and shock-plot modeling, and D. Rupke for helpful conversations toward the results and their interpretations. We also thank the Keck staff for help with carrying out the observations and the UH/UC TACs for granting this observing time. The data presented herein were obtained at the W.M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W. M. Keck Foundation. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain. V.U. acknowledges funding support from the NASA Harriet G. Jenkins Predoctoral Fellowship Project and the Smithsonian Astrophysical Observatory Predoctoral Fellowship. A.M. acknowledges funding support from the NSF Graduate Fellowship. K.I. thanks support from Spanish Ministerio de Ciencia e Innovación (MICINN) through grant (AYA2010-21782-C03-01). This work is partially supported by the JPL Contract/IRAC GTO Grant No. 1256790. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc.

THE INNER KILOPARSEC OF Mrk 273 WITH KECK ADAPTIVE OPTICS

Article metrics

Permissions

Author e-mails

Author affiliations

Author notes

Dates

ABSTRACT

1. INTRODUCTION