COLD MOLECULAR GAS IN THE INNER TWO KILOPARSECS OF NGC 4151^*

The cleaned millimeter continuum map derived with uniform weighting is shown in Figure 2. The beam size is 28 × 22 and the rms noise is 0.3 mJy beam ⁻¹. The total integrated flux density is 14 mJy and the peak flux is 12 mJy beam⁻¹. The millimeter continuum peaks at R.A. $=\rm 12^h10^m32\mbox{$.\!\!^{\mathrm s}$}56$ and decl. = +39°24'2106 (J2000.0) corresponding to the AGN location (Mundell et al. 2003), with an offset of 025, consistent with our astrometric uncertainty of about 04. At this resolution it is barely resolved: a two-dimensional Gaussian fit gives a convolved source size of 32 × 23 with a P.A. of 93°. While the elongation of the continuum is similar to the shape of the clean beam, this suggests a small extent in east–west direction roughly consistent with the orientation of the radio jet that is elongated along a P.A. of 77° (Mundell et al. 1995).

Zoom In Zoom Out Reset image size

We used the task MOMENTS in GIPSY to construct the intensity, velocity, and velocity dispersion maps of the CO (1-0) line emission, using a flux clipping level of 3σ = 6 mJy beam⁻¹ per channel and considering emission as real only if it appears in at least two consecutive channels. The width of the two channels is 13 km s⁻¹, smaller than the typical line width (20 km s⁻¹) seen throughout the data cube. The three moment maps of the naturally weighted data cube are shown in Figure 3.

Zoom In Zoom Out Reset image size

3.2.1. Molecular Gas Morphology and Masses

The integrated intensity map of the CO line emission (Figure 3, top panel) shows a fairly regular geometry. About 15'' north and south of the nucleus, the molecular gas forms two gas lanes that run roughly in east–west direction over a length of 16'' (north) and 20'' (south). The southern lane is about twice as bright as the northern lane. In addition to these extended structures, several fainter and more compact clumps are detected. All major components are indicated in Figure 4. The most prominent clump is closest to the nucleus (labeled as central clump in Figure 4) at a distance of 9'' (≈600 pc) to the north. Four additional smaller clumps are present; three near the northern lane and the last one at about 25'' west (western clump on Figure 4) of the center of NGC 4151.

Zoom In Zoom Out Reset image size

We measured the CO fluxes within different components marked in Figure 4 using the tasks BLOT (to define the regions) and FLUX in GIPSY. All the fluxes presented here are corrected for primary beam attenuation. We then converted the CO (1-0) flux into molecular gas mass for each component, using a standard Galactic CO-to-H₂ conversion factor of X_CO = 2 × 10²⁰ cm⁻² (K km s⁻¹)⁻¹ (Solomon & Barrett 1991) and applying a mass correction for helium of 36%. The derived CO fluxes and molecular gas masses are listed in Table 3. In our entire map, the total flux recovered is 14.7 Jy km s⁻¹ and the corresponding cold molecular gas mass is about 4.3 × 10⁷ M_☉, with 60% of this mass being contained in the southern lane. Our PdBI data recover more flux than the JCMT single dish measurements (Rigopoulou et al. 1997). They measure a total CO (2-1) flux of I_CO = 1.56 K km s⁻¹, corresponding to S_CO(1−0) = 11 Jy km s⁻¹ in integrated flux, assuming optically thick molecular gas, in the inner 23'' (∼1.5 kpc). The total H₂ mass derived from the single dish data is then M(H₂)=2.4 × 10⁷ M_☉ which corresponds to a total molecular mass of M_mol = 3.2 × 10⁷ M_☉, after applying the mass correction for helium. Therefore, the JCMT data missed about 25% of the flux coming from the outer edge of the CO gas lanes.

Table 3. CO Fluxes and Molecular Gas Masses

Component	SCO dV	Mmol
	(Jy km s−1)	(107 M☉)
Total	14.7	4.3
Southern lane	8.7	2.5
Northern lane	5.0	1.5
Central clump	0.4	0.12
Western clump	0.63	0.18

Notes. Integrated CO (1-0) line fluxes and corresponding molecular gas masses for different components identified in the CO distribution. The location and extent of the individual components are indicated in Figure 4. The fluxes have been corrected for primary beam attenuation and the masses include 36% of Helium fraction.

Download table as:  ASCIITypeset image

Interestingly, we detect no CO emission in the central 3'', where warm H₂ emission is observed via its NIR emission lines (Fernandez et al. 1999; Storchi-Bergmann et al. 2009). We discuss the discrepancy between the CO and NIR H₂ observations in the central regions in Section 4.2.

3.2.2. CO Kinematics

The CO (1-0) velocity field and the velocity dispersion map are presented in the middle and bottom panels of Figure 3. The velocity field in the extended southern and northern lanes shows an ordered gradient along the lanes with values similar to those derived from H i observations (Asif et al. 1998; Mundell et al. 1999) and an average velocity dispersion of approximately 7 km s⁻¹. Figure 5 shows the integrated spectra of the three main components: the northern and southern gas lanes as well as the central clump. The integrated spectrum of the northern lane is centered at 80 km s⁻¹ and the central clump at 55 km s⁻¹. The spectrum of the southern lane shows the presence of two kinematic components originating from each half of the lane—one centered at 0 km s⁻¹, corresponding to the eastern part of the lane where it curves to the north, and the other centered at −40 km s⁻¹ arising from its most prominent western part—separated by a region at the center of the lane that shows an increased velocity dispersion, close to the kink point. Figure 6 shows a selection of spectra taken close to this high velocity dispersion region. Asymmetry in the line profiles is clear, suggesting multiple kinematic components, but this region is spatially only marginally resolved while a second line component can be easily identified in several spectra. This boundary region between the two halves of the gas lane therefore may represent a region in which gas clouds with distinct velocities overlap along the line of sight or a genuinely mixed, disturbed zone. Larger velocity dispersion in the eastern part of the northern lane may also indicate that the eastern part of this lane is a distinct cloud, and although the line profiles in this region deviate from a simple Gaussian shape they show no clear evidence of two components in this region.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We extracted pv diagrams along three different positions that are presented in Figure 7. The pv diagram along the major kinematic axis of P.A. = 22° shows strong emission from the gas lanes at $\rm +80$ km s⁻¹ and $\rm -50$ km s⁻¹. The CO (1-0) line emission shows a velocity gradient indicating that the gas is participating in circular motion. Interestingly, this gradient is not symmetric with respect to the systemic velocity of 995 km s⁻¹ from Pedlar et al. (1992), it presents an offset of +15 km s⁻¹. In order to investigate the gas motion within the gas lanes, two slits were placed along them (numbers 2 and 3, Figure 7). While the velocity gradient is very low in the northern lane with a mean velocity of about +80 km s⁻¹, two distinct features are present in the southern lane. The eastern part of the lane shows constant velocities consistent with the systemic velocity while the western part exhibits a clear velocity gradient from −20 to −60 km s⁻¹.

Zoom In Zoom Out Reset image size

The CO (1-0) rotation curve of NGC 4151 was derived from the velocity field using the task ROTCUR in GIPSY. This task fits the velocity, assumed to be circular, within tilted rings. We run ROTCUR for rings 2'' wide from 3'' to 19'' radial distances centered on the AGN as traced by the peak of the 3 mm continuum. The region defined by an angle of 5° around the minor axis was excluded for the fit. The inclination and position angle were fixed at i = 21° and P.A. = 22° as derived from the H i data by Mundell et al. (1999). The systemic velocity, v_sys, was allowed to vary in the first step and found to be at 1010 km s⁻¹, i.e., offset by about +15 km s⁻¹ from the value derived by Pedlar et al. (1992). This offset is consistent with the offset in the velocity gradient along the major axis, as discussed above. Indeed, with a systemic velocity of 1010 km s⁻¹, the pv diagram along the major kinematic axis (P.A. = 22°) presents a more symmetric velocity gradient, the southern and northern CO lanes lying then at −65 and + 65 km s⁻¹, respectively.

The resulting rotation curve, extracted with v_sys =  1010 km s⁻¹, is presented in Figure 8. The point at 9'' presents a velocity lower than expected for circular rotation. This point corresponds to the northeastern part of the southern lane, which has been shown to have kinematics distinct from the western part of the lane (Figure 5). The low velocity of the first point at 3'' is due to beam smearing. Comparison to the H i rotation curve (Mundell et al. 1999) shows a good agreement between the CO and H i rotation curves (Figure 8, right).

Zoom In Zoom Out Reset image size

Therefore, the overall CO kinematics are generally consistent with rotation in the plane of the galaxy. To demonstrate this further, Figure 9 shows velocity as a function of position angle measured directly along the CO arcs. The sinusoidal curves overplotted show the expected variation of velocity along a circular annulus in the plane of the galaxy (i = 21°, P.A. = 22°) at a radius of 18'' corresponding to the semimajor axis of the dust/CO ellipse. Although not exact, the correspondence between the plotted curves is good, ruling out the interpretation that the ellipticity traced by the CO arcs is a result of gas lying in a highly inclined circular disk tilted out of the main galactic plane as well as strong non-circular motions within the disk. The morphology is therefore intrinsically elliptical in the plane of the galaxy, as suggested by the analysis of the H i velocities along the dust arcs (Figure 3 of Asif et al. 1998) and consistent with predictions of gas flows in bars (Athanassoula 1992). The rotation curve in Figure 8 is thus only an approximation to the true kinematics, and the offset between the derived CO and H i systemic velocities is a likely indication of disturbed CO gaseous kinematics in these arcs.

Zoom In Zoom Out Reset image size

Finally, the velocities within the central clump present larger deviations from circular motions, up to 50 km s⁻¹ below the circular velocity at a radius of 5'' (Figure 9, right). The central clump shows redshifted kinematics (see Figure 3) significantly inconsistent with those expected for circular motion in this region, close to the minor kinematic axis of the galaxy. These observed deviations are suggestive of streaming motions, however, the uncertainties are too large to derive a more quantitative measure. Very deep CO imaging in the future is required to confirm whether this clump is an isolated cloud or part of a more continuous fluid flow toward the nucleus.

In order to analyze the importance of the large-scale stellar bar or oval for fueling the AGN, we need to establish the link between the bar and the molecular gas observed in the central 2.5 kpc. The detailed H i study of Mundell & Shone (1999) showed that the stellar bar/oval is unusually gas-rich and that its H i gas dynamics are consistent with the presence of x₂ orbits in an oval bar; sharp velocity changes across the bright regions close to the leading edges of the bar are found which can be explained by offset shocks predicted by simulations (Athanassoula 1992) and provide direct evidence for the presence of these x₁ and x₂ families of stellar orbits. The Hα image (Knapen et al. 2004) reveals H ii regions along the leading sides of the large-scale oval coincident with the peaks in the H i surface density. These H ii regions are associated with the ionized and neutral gas in the streaming shocks (Asif et al. 2005) that connect to the CO gas lanes to form an inner gaseous spiral (Figure 10, left). Mundell et al. (1999) derived the resonance curves from the H i velocity field and estimated the Inner Lindblad Resonance (ILR) to be at 2.8 ± 0.6 kpc. The ILR (red circle in Figure 10, left) coincides with the H ii regions, while the CO lanes are well inside the resonance. This picture is consistent with the molecular gas tracing a spiral density wave, driven by the large scale bar and extending from the ILR down to the center (Englmaier & Shlosman 2000). As we do not detect CO all the way down to the nucleus, this may argue in favor of the presence of a second ILR that limits the gaseous wave to exist between the two resonances (Englmaier & Shlosman 2000). However, the sparse distribution of the CO gas, the lack of H i in the central regions, and the nonlinear behavior of gas in a barred potential prevent useful constraints on the presence and location of an inner ILR from being derived directly from the gaseous rotation curves. Full dynamical modeling and numerical simulations constrained by millimeter, radio, and optical velocity fields and K-band derived mass models will be presented in a future paper (G. Dumas et al. 2010, in preparation) and may provide insight into the location of an inner ILR and its effect on the gas dynamics.

Zoom In Zoom Out Reset image size

The CO gas lanes also coincide with two dust lanes seen in a V−I color map (Asif et al. 1998), as shown in Figure 10 (right). The two red arcs delineate a 11'' × 18'' dust ellipse inside the large oval/bar and also coincide with H i emission (Asif et al. 1998). The peak column densities in the H i arcs are $(0.7\hbox{--}1.8)\times 10^{21}\,\rm cm^{-2}$ implying an atomic gas mass of $M_{\rm H\, \mathsc {i}}$ ∼(0.13–3.3) × 10⁶ M_☉. Thus, the ratio of atomic to molecular gas is 0.3 in the northern lane and 0.006 in the southern lane, which is in agreement with the values of the atomic-to-molecular gas ratio observed in the central regions of spiral galaxies (Schuster et al. 2007; Leroy et al. 2008), with the southern lane being at the lower end. Moreover, the kinematics of the CO gas lanes and the H i arcs are in good agreement as shown in the previous section. Therefore, the morphology and kinematics of the CO northern and southern lanes suggest that the molecular gas lanes are the continuation of the atomic and dusty ring, along elliptical orbits of the stellar bar.

It is interesting to note that while the structure of the gas lanes is very well matched, no obvious counter-structure is present in the V − I map for the nuclear clump. High-resolution Hubble Space Telescope/Advanced Camera for Surveys imaging shows the presence of dust in the central 1.5 kpc, but no clear structure is seen in the dust (Bentz et al. 2006). Connecting the gas flows in the large scale bar with the nuclear dust structures therefore remains a challenge.

In the central 25, excited H₂ line emission centered on the nucleus and extended along the minor axis of the galaxy (08 ≈ 50 pc) has been observed by Fernandez et al. (1999) and more recently by Storchi-Bergmann et al. (2009). The later showed that H₂ is thermally excited either by X-rays from the AGN or by shocks along the accretion flow toward the nucleus. This central H₂ emission has been interpreted as tracing the reservoir of (cold) gas from which the super-massive black hole in NGC 4151 is being fed. In particular, Storchi-Bergmann et al. (2009) found a mass of hot H₂ gas of 240 M_☉, and based on the conversion of warm-to-cold H₂ mass ranging between 10⁻⁷ and 10⁻⁵ (Dale et al. 2005), they estimated the total mass of molecular gas (hot plus cold) to be in the range of 2.4 × (10⁷–10⁹) M_☉. We should have easily detected such an amount of molecular gas. Indeed, the 1σ upper limit for the CO (1-0) emission within the central 3'' is 0.033 Jy km s⁻¹, which translates into a limit for the cold molecular gas mass of 10⁵ M_☉, well below the lower limit expected from the NIR H₂ line.

This discrepancy between warm and cold molecular gas mass can be interpreted in two opposite ways. On one hand, it indicates that no significant amount of cold molecular gas is present close to the nucleus. This implies that the warm molecular gas observed there is not a good tracer for the general amount of (cold) molecular gas present. Thus, the use of a generalized ratio of warm-to-cold H₂ could be misleading in the vicinity of AGN, which would explain the large overestimate of cold molecular gas mass by Storchi-Bergmann et al. (2009). On the other hand, in X-ray-dominated regions, equilibrium molecular abundances may be affected by the X-ray radiation from the AGN. This leads in particular to a depletion of CO abundance with respect to other molecules such as HCN (Boger & Sternberg 2005). Indeed, many Seyfert galaxies show enhanced central HCN(1-0) emission such as, e.g., NGC 1068 (Jackson et al. 1993; Tacconi et al. 1994; Helfer & Blitz 1995; Usero et al. 2004), NGC 1097 (Kohno et al. 2003), NGC 5194 (Kohno et al. 1996), NGC 6951 (Krips et al. 2007a). In particular, the latter study showed that the CO line emission largely underestimates the amount of cold molecular gas in the central region of the Seyfert 2 NGC 6951. Therefore, CO itself may be a poor tracer of cold molecular gas close to an active nucleus. Under this assumption, the change of chemistry in the central kiloparsec because of the AGN itself could explain the lack of CO emission inside the dust/gas lanes and the excitation of the NIR H₂ would be then driven by the AGN rather than the accretion flow. Further study of molecular abundances in the central 3'' of NGC 4151 is needed to quantify the cold molecular mass in the central region of NGC 4151 and to explain this discrepancy between (cold) molecular gas masses inferred from NIR H₂ and CO line emission.

The presence of non-circular motion in the central clump indicates that fueling of the AGN or at least a replenishment of the gas reservoir closer to the AGN might be ongoing. Analysis of near-IR images reveals no evidence for a second inner stellar bar (Onken et al. 2007). This suggests that the disturbances of the regular motions might be caused by a transient event, such as an interaction with a nearby galaxy. Such an interaction has been proposed by Mundell et al. (1999) to explain the distortion of the large two-arm stellar spiral. NGC 4156 and UGC07188 have small projected distances to NGC 4151 and therefore may be candidate companions for such a galaxy interaction, although at, respectively, v_sys = 6750 km s⁻¹ and 6850 km s⁻¹ the systemic velocities of NGC 4156 and UGC07188 are significantly larger than that of NGC 4151.

Our findings are in agreement with recent results from the NUGA survey, which found no direct indication for ongoing fueling within the central hundreds of parsecs for most of their targets (García-Burillo et al. 2005; Boone et al. 2007; Lindt-Krieg et al. 2008; Casasola et al. 2008; Haan et al. 2009). It is, however, interesting that most NUGA galaxies have a molecular gas reservoir within the inner kiloparsec with mass between 10⁷–10¹⁰ M_☉ (García-Burillo et al. 2003b, 2005, 2009; Combes et al. 2004, 2009; Krips et al. 2005, 2007b; Boone et al. 2007; Casasola et al. 2008). It is also interesting to note that most NUGA targets are either Seyfert type 2 or LINER galaxies. While gravitational mechanisms, driven, e.g., by large scale stellar bars or galaxy interaction, are efficient in transporting gas to the central kiloparsec of galaxies, other processes (e.g., viscous torques, García-Burillo et al. 2005; Haan et al. 2009) must take over to bring the gas closer to the nucleus and fuel the AGN. Moreover, recent nuclear star formation and AGN activity have been shown to be linked (González Delgado et al. 2001; Riffel et al. 2007; Davies et al. 2007), and stellar outflows, from starburst on tens of parsec scales, can also support the nuclear activity (Davies et al. 2007; Hicks et al. 2009; Schartmann et al. 2009). NGC 4151 presents no evidence for a recent nuclear starburst (Sturm et al. 1999). In fact, the star formation rate across the whole galaxy is unusually low given the presence of significant quantities of H i gas (Asif et al. 2005), therefore this scenario appears to be unlikely for this particular Seyfert galaxy. Finally, Dumas et al. (2007) showed that the ionized gaseous kinematics of Seyferts are more disturbed with respect to the stellar kinematics in the central kiloparsec than those of inactive galaxies, i.e., the kinematic disturbance scales with accretion rate in the inner kiloparsec where dynamical and activity timescales become comparable.