EXTENDED X-RAY EMISSION IN THE H i CAVITY OF NGC 4151: GALAXY-SCALE ACTIVE GALACTIC NUCLEUS FEEDBACK?

Figure 1(a) shows the resulting soft X-ray image (3'' × 3'', ∼12 kpc on a side). Note that the non-detection of X-ray emission toward the northernmost and southernmost part of the image is artificial, because this area is out of the ACIS-S field of view for our observation (Figure 1(a)). Figure 1(b) presents a composite image of the same region in NGC 4151, consisting of soft X-ray (0.3–1 keV), exposure-corrected Chandra ACIS image (blue), with the H i λ21 cm map (red) and a continuum-subtracted Hα image (green; Knapen et al. 2004).

Zoom In Zoom Out Reset image size

The kiloparsec-scale H i distribution (Mundell & Shone 1999) appears as a ring with brightest emission in the NW and SE (white contours in Figure 1(a)). The Very Large Array fluxes (Mundell et al. 1999) agree well with lower resolution single dish measurements (Pedlar et al. 1992), supporting that the apparent central cavity is truly devoid of H i. Toward the nuclear region, only localized H i absorption toward the radio jet is found (Mundell et al. 1995, 1999). Inside this oval H i distribution, CO line emission (Dumas et al. 2010) is prominent in two lanes ∼1 kpc north and south of the nucleus (red contours in Figure 1(a)) and coincident with the circum-nuclear dust ring (Asif et al. 1998).

The ionized gas traced by Hα is mainly located in the ∼20'' long biconical ENLR (cyan = Hα+X-ray in Figure 1(b)) along the NE–SW direction centered on the nucleus, which closely follows the high excitation emission line gas (e.g., [O iii]5007; Kaiser et al. 2000; magenta contours in Figure 1(a)) photoionized by the AGN radiation. There is also Hα emission associated with a string of H ii regions located along the NW and SE edges of the large-scale stellar bar (yellow = Hα+H i; Pérez-Fournon & Wilson 1990).

The soft X-ray emission is brightest in the nuclear region and central 15'', extending along the same direction as the ENLR bicone (P.A. ∼ 65°/230°; appearing as the cyan rectangle in Figure 1(b)). This emission and the associated morphological features will be discussed in detail in a companion paper (J. Wang et al. 2010, in preparation).

Two X-ray clumps are present at the terminals of the bicone (marked as "interaction" in Figure 1(b)), where the outflows appear to encounter dense materials (CO gas lane in the NE and H i clump in the SW). In addition, there are two regions of X-ray enhancements in the H i ring (marked as "H ii" in Figure 1(b)), spatially coincident with known H ii regions (Pérez-Fournon & Wilson 1990). There are also some anti-correlations in the spatial distribution of X-ray emission relative to the H i material due to obscuration of the soft X-rays. Beyond the H i material (∼60'' from the nucleus), the X-ray emission becomes weak to the east and absent to the west.

Here, we focus on the presence of the previously unknown, faint soft X-ray emission, which is seen extending beyond a radius of 30'' (∼2 kpc), filling the cavity in the H i distribution.

We extracted X-ray data from an elliptical region centered at the nucleus (yellow polygon in Figure 1(a)), which covers the H i cavity and excludes the inner brighter emission (r ≲ 15'') associated with the nucleus and the optical outflow. We measured the background level to be 0.050 ± 0.006 counts per ACIS pixel, selecting alternative background regions in the image 2''–3'' away from NGC 4151. We find that the extended emission (0.3–1 keV) in the cavity region is significant (1562 counts versus 728 counts expected from background) at 11σ (F_{0.5–2 keV} = 3.2 × 10⁻¹⁴ erg s⁻¹ cm⁻², L_{0.5–2 keV} ∼ 10³⁹ erg s⁻¹). Moreover, we performed ChaRT⁸ and MARX⁹ simulations of the strong nuclear source and the resolved extended emission using Chandra ACIS spectra (Yang et al. 2001; Wang et al. 2010), which demonstrate that the point-spread function (PSF) wings of the nuclear emission and the extended emission can only contribute 157 ± 12 counts and 42 ± 6 counts in the 0.3–1 keV band in the extraction region, respectively.

We extracted the ACIS spectrum of the soft X-ray emission in the same region. The background-subtracted X-ray spectrum is poorly fitted with an absorbed power-law model that is consistent with the nuclear spectrum (Γ = 1.68; Wang et al. 2010), showing significant line emission residuals in the 0.3–1 keV range. Moreover, the spectrum can neither be fitted by a single photoionized emission model (photemis; see XSTAR¹⁰) nor by a thermal plasma (APEC; Smith et al. 2001) model (χ²/dof = 182/83 and 178/83, respectively).

The fit is much improved when a power-law component with a photoionized emission component is used (log ξ = 1.7 ± 0.2), where ionization parameter ξ ≡ L/nR² (Kallman & McCray 1982). The fit is slightly improved when a thermal plasma (APEC; Smith et al. 2001) component (kT = 0.25^+0.04_−0.03 keV) is adopted. In both two-component fits, the power-law component contributes ∼50% of the 0.3–2 keV flux. The results are shown in Figure 2 and summarized in Table 1.

Zoom In Zoom Out Reset image size

The surface brightness profile of the extended emission is shown in Figure 3, using Δr = 5'' concentric annuli starting at r = 15'' from the nucleus. The small excess at r ∼ 160 pixel (80'') is due to a faint extended (d ∼ 6'') X-ray enhancement (L_x ∼ 4 × 10³⁷ erg s⁻¹ assuming the same kT ∼ 0.25 keV) centered at α=12^h10^m399, δ=+39°24'10'' (P.A. ∼ 120°). It appears to be spatially coincident with the faint diffuse H i gas seen around the outer ends of the bar connecting to the spiral arms (Pedlar et al. 1992).

Zoom In Zoom Out Reset image size

We consider four hypotheses for the extended soft X-rays on spatial scales of ∼10⁴ light year.

• 1.  
  Unresolved point sources. Active star formation in the region of the H i cavity is at a low level, as indicated by the lack of Hα emission and weak polycyclic aromatic hydrocarbon emission (Asif et al. 2005). Adopting the empirical measured L_x/M_* ∼ 8.2 × 10²⁷ erg s⁻¹ M⁻¹_☉ (Revnivtsev et al. 2008) and a bulge mass of M_* ∼ 10⁹ M_☉ for NGC 4151 (Wandel 2002), the expected combined contribution of the unresolved old stellar population (low-mass X-ray binaries and cataclysmic variables) is L_{0.5–2 keV} ∼ 10³⁷ erg s⁻¹, two orders of magnitude lower than the detected soft emission (L_{0.5–2 keV} ∼ 10³⁹ erg s⁻¹). Moreover, the K-band starlight profile (Knapen et al. 2003) decreases faster than the X-ray emission profile (Figure 3). Thus, we conclude that the contribution of stellar sources is negligible.
• 2.  
  Electron scattered nuclear emission. The hot plasma around the nucleus could electron scatter a fraction of the nuclear flux at a larger radii (e.g., NGC 1068; Elvis et al. 1990). Assuming a solid angle Ω = 4π for a scattering medium, the observed L_x/L_x,nuc implies an electron-scattering optical depth of τ_es = 0.01 and a mean column density of N_H ∼ 10²² cm⁻². Adopting 200 pc for the depth, the typical galactic disk scale height (e.g., Padoan et al. 2001), this corresponds to a volume density of n_e ∼ 15 cm⁻³. This is ∼100 times higher than n_e inferred for the thermal plasma and so thermal emission (which scales as n²_e) will dominate the emission unless it is much cooler than 10⁶ K. The scattering medium must also be highly ionized, requiring T ≳ 3 × 10⁶ K if this is achieved thermally. For the electron-scattering model to be self-consistent, a photoionized medium (eliminating T discrepancy) or a historic outburst (∼10⁴ yr ago) during which the nucleus was 100 times brighter than currently observed (eliminating n_e discrepancy) is needed.However, the poor fit with the nuclear power-law model (Section 3.2) does not support this scenario. Moreover, the required power-law component in either two-component fits (L_{0.5–2 keV} ∼ 4 × 10³⁸ erg s⁻¹) is comparable to the expected total contribution (L_{0.5–2 keV} ∼ 2.5 × 10³⁸ erg s⁻¹) from PSF scattering and unresolved point sources, which allows little contribution from the electron-scattered component. We conclude that an electron-scattered component is not important for the extended emission.
• 3.  
  Photoionized gas. The faint extended emission in NGC 4151 could be due to gas photoionized by a more luminous AGN in the past. Such a "quasar-relic" scenario was proposed for NGC 5252, which has a spectacular ionization cone extending ∼10 kpc from the nucleus (Dadina et al. 2010). Observable signatures of such "afterglow" from an outburst in a radio quiet quasar (L ∼ 10⁴⁶ erg s⁻¹) have been studied by Wang et al. (2005). In this context, NGC 4151 belongs to the regime of the lower luminosity sources, for which Wang et al. (2005) suggest an X-ray spectrum dominated by Lyα and Lyβ lines of Ne x and O viii. Under this assumption, the 0.3–1 keV emission is dominated by line emission blended at ACIS' spectral resolution, which in turn implies an ionization parameter of log ξ ∼ 1.6–2.1 (Kallman & McCray 1982; Yang et al. 2001) consistent with log ξ = 1.7 measured in XSTAR. Adopting an electron density n = 2 cm⁻³ (equal to the H i density) and R = 3 kpc, the required ionizing luminosity of the nucleus is L_ion ∼ 6 × 10⁴⁵ erg s⁻¹ ∼ L_Edd (M_BH = 4 × 10⁷ M_☉; Bentz et al. 2006). Based on detailed photoionization modeling of the extended optical emission of NGC 4151 (Schulz & Komossa 1993), the photoionizing continuum could be anisotropic and the ionizing flux toward the ENLR may be ∼10 times higher than that in the direction of the Earth. This is still insufficient to power the more extended X-ray emission. Therefore, an Eddington limit outburst of NGC 4151 ∼10⁴ yr in the past is required.The same photoionized medium would produce [O iii] emission. If we adopt an observed [O iii]/soft X-ray ratio of ∼10 for log ξ ∼ 1.7 photoionized gas (Bianchi et al. 2006), the expected [O iii] surface brightness is ∼10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻², fainter than the sensitivity limit of existing [O iii] images (e.g., F_lim ∼ 10⁻¹⁵ to 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻²; Pérez-Fournon & Wilson 1990; Evans et al. 1993; Kaiser et al. 2000). Thus, a photoionized origin is still open to direct testing.
• 4.  
  Confined hot gas. We model the background-subtracted surface brightness profile with a one-dimensional β model provided in SHERPA¹¹ (Σ_x ∝ [1 + (r/r₀)²]^−3β+1/2, where r₀ is the core radius). This model is often adopted in studies of the hot gas in early-type galaxies (e.g., Trinchieri et al. 1986). The best-fit index is β = 0.39^+0.03_−0.02, implying a brightness profile that decreases slower than an adiabatically expanding wind (Σ_x ∝ r⁻³) and confinement of the hot gas. We attempted fitting spectra extracted from two concentric annuli at radii Δr = 15''–30'' and 30''–45''. Within the uncertainties, the thermal fits give the same temperature (kT = 0.25 ± 0.07 keV and kT = 0.22^+0.05_−0.03 keV), implying that any hot gas is approximately isothermal.

Under this assumption, we derived the pressure radial profile of the hot gas based on the surface brightness profile, considering two cases for the volume in which the hot gas is contained: a cylinder or a sphere. For the cylinder, we assumed a depth of 200 pc. To calculate the volume emission density of spherically symmetric gas, we account for the projection effect by successively subtracting outer shells to smaller radii (Kriss et al. 1983). The electron density distribution is then derived from the emission per unit volume, adopting the simple approximation ≈ 6.2 × 10⁻¹⁹T^−0.6n²_p erg s⁻¹ cm⁻³ (McKee & Cowie 1977) for the emissivity of a kT = 0.25 keV thermal gas (Section 3.2).

The pressure of the X-ray gas gradually decreases from 10⁻¹¹ to approximately ∼5 × 10⁻¹² dyne cm⁻² when reaching the H i gas lanes. Without confinement, the gas will inevitably expand, cooling too much to be observed in the X-rays. We estimate that the thermal pressure of CO is ∼10⁻¹⁵ dyne cm⁻², using the mass measured in Dumas et al. (2010), thus the molecular gas present (T ∼ 10 K) cannot provide the pressure balance. The estimated thermal pressure of the H i material is ∼2 × 10⁻¹² dyne cm⁻², assuming T ∼ 8000 K¹² and a column density N_H = 1.5 × 10²¹ cm⁻² (Mundell et al. 1999). This is well matched to the pressure required to provide the confinement needed to prevent the hot gas from expanding in the disk plane where atomic gas is most dense.

Mundell et al. (1999) showed that the isovelocity contours of the H i emission deviate from a circularly rotating disk and identified kinematic evidence for the presence of shocks in inflowing gas along the stellar bar. Taking an average inflow velocity of v ∼ 40 km s⁻¹ (Mundell et al. 1999) and a particle density of n = 2 cm⁻³ in the H i gas lanes, the dynamic pressure as a result of the inflowing motion is ρv² ∼ 10⁻¹¹ dyne cm⁻², where ρ = nm_p is the density. This provides additional pressure for the neutral material to be in equilibrium with the hot gas.

The presence of soft X-ray emission (either thermal or photoionized) on a ≳2 kpc scale is interesting. If the X-ray emission originates from hot gas heated by the AGN outflow, it would be strong evidence for AGN feedback to the ISM on galaxy scales, resembling the larger scale AGN feedback from radio bubbles interacting with the intracluster medium seen in the Perseus Cluster (Fabian et al. 2003).

In the photoionized scenario, considering the light-travel time from the nucleus to the H i gas lanes (10⁴ yr) and the photoionization recombination timescale t_rec ≈ 200(n/10⁹)⁻¹(T/10⁵)^0.7 s (∼1.5 × 10⁴ yr under the above assumptions; Reynolds 1997), we can constrain the timescale when NGC 4151 was experiencing such an Eddington limit outburst phase to be t_Edd < 2.5 × 10⁴ yr ago. Otherwise the X-ray gas would no longer emit recombination lines.

If the soft diffuse emission is due to shock heating associated with a nuclear outflow instead, the current L_bol = 7.3 × 10⁴³ (Kaspi et al. 2005) indicates L_bol/L_Edd ∼ 0.01. Assuming that ∼5% of the power has gone to create the H i cavity (e.g., Silk & Rees 1998; Hopkins et al. 2005), ∼4 × 10⁴ yr of such AGN heating is needed to produce the thermal energy content of the hot gas (E_th ∼ 3 × 10⁵⁴ erg). This is much shorter than the current cooling time of the hot gas (τ_c ∼ 10⁸ yr). The efficiency at which the AGN outflow deposits its kinematic energy in the ISM can be as low as 10⁻⁴, provided that the hot gas is confined and the duration of the strong nuclear outflow is τ_outflow ≪ τ_c.

However, perpendicular to the plane, the gas could expand and cool efficiently via adiabatic expansion unless some other confinement exists. Assuming an outflow velocity of v_outflow ∼ 10³ km s⁻¹ (typical of the NLR clouds; Kaiser et al. 2000) for the free expansion perpendicular to the disk (cf. thermal velocity dispersion $v_{{\rm th}}=\sqrt{k_BT/m_p}\sim 100$ km s⁻¹), the relevant timescale is the time span in which the hot gas expands to the scale height above disk plane ΔR/v_outflow ∼ 10⁵ yr. This timescale again implies that the timescale to the last AGN outflow heating must be <10⁵ yr, in order to replenish the X-ray gas expanding out of the plane that is cooling rapidly.

Compared to the estimated AGN lifetime (10⁷–10⁸ yr), the fact that we observe the AGN–host interaction is intriguing given the short timescale to the last episode of activity. Such episodic outbursts must occur frequently (Ciotti et al. 2010), and our finding implies that they may occupy ≳ 1% of the AGN lifetime.