A DEEP CHANDRA ACIS STUDY OF NGC 4151. II. THE INNERMOST EMISSION LINE REGION AND STRONG EVIDENCE FOR RADIO JET–NLR CLOUD COLLISION

The 2–7 keV emission is dominated by the unresolved nucleus (Yang et al. 2001; Wang et al. 2009b). Since we are interested in the extended X-ray emission, we used the energy range below 2 keV where this emission is prominent (see Paper I). To investigate the general spectral dependency of the morphology of this soft X-ray emission, images in three spectral bands below 2 keV were extracted from the merged data: 0.3–0.7 keV ("soft band"), 0.7–1.0 keV ("medium band"), and 1–2 keV ("hard band"). Following Wang et al. (2009a), exposure maps were created for the individual bands to obtain exposure-corrected flux images.

Figure 1 presents the resulting false color composite image of the central 7'' × 7'' (450 pc on a side) region of NGC 4151, where the soft, medium, and hard band images are shown in red, green, and blue, respectively. The images in the three band have been smoothed with an FWHM = 03 Gaussian kernel. It clearly shows bright structured soft X-ray emission along the northeast (NE)–southwest (SW) direction, which is also the direction of the radio jet. In particular, the medium band image shows a jet-like ∼2'' linear extension (contours in Figure 1). This medium band is dominated by the Ne ix emission (see below). In the following we concentrate on the analysis of this inner 2''radius circumnuclear region.

Zoom In Zoom Out Reset image size

High spectral resolution grating spectra of NGC 4151 (Ogle et al. 2000; Schurch et al. 2004; Armentrout et al. 2007) have shown that the emission in these soft spectral bands is almost entirely dominated by lines. Although the spectral resolution of ACIS CCD cannot provide unique identifications of the strongest soft X-ray emission lines (<2 keV) seen in the nuclear spectrum (Wang et al. 2010b), we can identify the dominant transitions guided by the HETG observations (Ogle et al. 2000). Most notably, the blended lines appear as three strong lines in the ACIS spectrum, approximately centered at 0.57 keV, 0.68 keV, 0.91 keV (cf. Figure 4 in Yang et al. 2001; Figure 3 in Wang et al. 2010b). We then extracted the X-ray emission in three narrow energy intervals (0.53–0.63 keV, blended O vii f,i,r; 0.63–0.73 keV, blended O viii Lyα and O vii RRC; 0.85–0.95 keV, blended Ne ix f,i,r) to create line strength images, highlighting regions of these prominent emission lines. This is a reasonable approach since the line emission dominates over the weak underlying continuum (Ogle et al. 2000; Schurch et al. 2004) in these narrow bands and reveal substructures that are not obvious in the broadband images (e.g., Bianchi et al. 2010; Gonzalez-Martin et al. 2010).

The resulting images are shown in Figures 2 and 3, showing the X-ray emission-line structure close to the nucleus (the central r = 2'', ∼130 pc) in the context of the radio outflow and the optical emission line clouds. The alignment between these images is done using the peak of X-ray emission, [O iii] emission, and the radio core. Since the astrometry of our Chandra image is accurate to 03, our interpretation of these features is not affected by significant alignment uncertainties. The position of the nucleus is indicated with a cross. The O vii, O viii, and Ne ix emission line images all show extended morphology and some structures, closely following the P.A. of the large scale extended NLR traced by the [O iii] emission (e.g., Winge et al. 1997; Kaiser et al. 2000).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In particular, the linear feature seen in the medium band is clearly present in the Ne ix image. There appears to be two X-ray enhancements bracketed by the optical clouds (indicated by arrows in Figure 3) that are close to the radio knot features C2 and C5 (Mundell et al. 1995). These "hot spots" are better visualized in the Ne ix/O vii ratio image, shown in Figure 4 together with the VLBA jet (Mundell et al. 2003) and the NIR [Fe ii] line emission (Storchi-Bergmann et al. 2009). The locations, where the jet appears to be intercepted by the optical clouds and where prominent [Fe ii] emission arise, are characterized by an Ne ix/O vii ratio of 2.8 ± 0.2, which is significantly (∼6σ) higher than the ratio in the surrounding regions (1.2 ± 0.2). We also examined the optical clouds with high velocity dispersion, possibly associated with jet–cloud impact (marked as crosses in Figure 4; see Figure 7 in Mundell et al. 2003), and find an elevated Ne ix/O vii ratio of 2.1 ± 0.4 (with marginal significance). The eastern Ne ix/O vii hot spot appears to be associated with clear interactions between the jet and the gas cloud, where shock heating becomes important. We note that the observed line flux of O vii is more suppressed than Ne ix in the presence of the Galactic absorption column (N_H = 2 × 10²⁰ cm⁻²), and XSPEC (Arnaud 1996) simulations show that the absorption corrected ratio may systematically decrease by 10%. Thus, the presence of Ne ix/O vii hot spots is not affected, unless there is significant differential absorption in the region.

Zoom In Zoom Out Reset image size

We extracted X-ray spectra from both the high Ne ix/O vii emission line ratio regions and the surrounding low ratio region (Figure 4(b)), and fit them jointly (with the same model components but normalizations set free) for comparison using XSPEC (version 12.5; Arnaud 1996). Spectra and instrument responses were generated using the CIAO tool specextract⁶. The background spectrum is taken from a nearby source-free region on the same CCD node. Spectra were grouped to have a minimum of 20 counts per energy bin to allow for χ² fitting. The contribution from the bright nuclear emission cannot be neglected here; to have a self-consistent model, we included a fraction of the scattered emission predicted from the PSF simulation of the point-like NGC 4151 nucleus.

We have made use of the Cloudy photoionization modeling code, last described by Ferland et al. (1998), to model the soft X-ray emission. Using Cloudy version C08.00⁷, which enables a Cloudy/XSPEC interface (Porter et al. 2006), we attempted to produce the soft part of the X-ray spectrum assuming an open plane-parallel geometry ("slab"). The dimensionless ionization parameter (Osterbrock & Ferland 2006) is defined as U = Q/(4πr²cn_H), where n_H is the hydrogen number density, r is the distance to the inner face of the model slab, c is the speed of light, and Q = ∫^∞_13.6eVL_ν/hν is the emitting rate of hydrogen ionizing photons (s⁻¹) by the ionizing source. It was previously noted that the spectral fit to the soft X-ray continuum of the nucleus is quite uncertain because of heavy absorption (Armentrout et al. 2007), therefore we adopted the broken power-law form in Kraemer et al. (2005) for the AGN continuum (Gamma = 2.3 for the energy range between 13.6 eV and 0.5 keV, and Gamma = 1.5 for E > 0.5 keV). Normalizing to the observed X-ray luminosity of the nucleus (Wang et al. 2010b), we obtain Q ∼ 7 × 10⁵³ (photons s⁻¹). We varied U and the column density N_H of the model slab to create spectral model grids, which were fed to XSPEC.

We quickly find that using a single photoionized component alone cannot reproduce the observed soft X-ray emission in NGC 4151. With a high photoionization component (log U ∼ 1.9), we were able to produce the hydrogen-like neon and oxygen line emission (e.g., O viii, Ne x), which are unambiguously present in all the grating spectra (e.g., Ogle et al. 2000; Schurch et al. 2004; Armentrout et al. 2007). However, the helium-like transitions, in particular the O vii and Ne ix emission line features, are not acceptably fitted. Adding a second lower ionization photoionized component (log U ∼ 0), as detected in Armentrout et al. (2007), we were able to obtain a significantly improved fit (Δχ = 200), except for residuals in the 0.7–1.1 keV range that are particularly strong for the hot spot spectrum. Adding additional photoionized components to the model does not further improve the fit when an F-test is performed. Allowing different absorption columns for the models also does not give a better fit (reduced χ²_ν = 1.7), indicating that the high Ne ix/O vii ratio is not due to higher obscuration. The absorption column required by the fits is consistent with the low Galactic extinction toward the NLR region found in Crenshaw & Kraemer (2005). Instead, these residuals disappear when a thermal emission component (APEC; Smith et al. 2001) with a temperature of kT = 0.58 ± 0.05 keV is added to the model, suggesting the presence of collisionally ionized Fe L emission. The spectral fitting results are summarized in Table 1 and shown in Figure 5. The absorption column is consistent with the low Galactic extinction toward the NLR region found in Crenshaw & Kraemer (2005).

Zoom In Zoom Out Reset image size

Lastly, we note that fitting the data with only combinations of absorbed optically thin thermal emission (APEC) gave strong residuals of emission line features (reduced χ²_ν ≫ 1), even when the abundance Z is allowed to vary.

The subpixel resolution ACIS image also allows a similar comparison of the X-ray emission with the optical NLR clouds resolved in the Hubble Space Telescope (HST) image. Taking advantage of the spectral capability of ACIS, we measured the 0.3–2 keV counts for the same [O iii] clouds identified in Wang et al. (2010b) and derived their 0.5–2 keV X-ray fluxes using the spectral model above. To subtract the nuclear contribution, we used a local region at the same radii but with different azimuthal angles from the clouds. The results are listed in Table 2. Figure 6 shows these [O iii] to soft X-ray ratios, comparing the measurements with HRC results in Wang et al. (2010b) and photoionization models in Bianchi et al. (2006).

Zoom In Zoom Out Reset image size

The [O iii]/X-ray ratios at the two X-ray hot spot locations (corresponding to radio knots C2 and C5 in Mundell et al. 1995) were also measured and found to be uniformly low, ∼3. This was previously noted in Wang et al. (2009b) and implies enhanced X-ray emission compared to other clouds under nuclear photoionization, which is likely associated with the jet–cloud interaction and consistent with their association with high Ne ix/O vii ratio regions.

The 6.4 keV fluorescent iron Kα emission of neutral or mildly ionized cold material is readily visible in the ACIS spectra of the nucleus (Wang et al. 2010b) and the extended region (Figure 7(a)). Another fluorescent feature, SiI Kα emission at 1.74 keV, is also clearly present in the extended emission (Figure 7(b)). Both spectra of extended emission were extracted from a 2'' < r < 30'' annular region centered on the nucleus. Is this "extended" fluorescent line emission truly originating from outside the nuclear region or is it the result of the PSF wing spreading out point-like nuclear features?

Zoom In Zoom Out Reset image size

To determine the spatial distribution of the Fe Kα line emitting material, we followed the procedure in Smith & Wilson (2001): a continuum image, taken to be the average of the counts in the 5.9–6.2 keV and 6.5–6.8 keV bands, was subtracted from the image extracted in the 6.2–6.5 keV band, resulting in a "pure" Fe Kα line image, shown in Figure 8(a). The Fe Kα emission appears mostly circular in r ⩽ 2'', but shows faint extended emission toward the north and west.

Zoom In Zoom Out Reset image size

The Si Kα line is blended with the Mg xii Lyβ and Si xiiif, r emission (Ogle et al. 2000), thus the above continuum subtraction is not applied. We extracted emission in the 1.7–1.8 keV range to obtain the Si Kα emission map (Figure 8(b)). It is mostly concentrated within an r ∼ 2'' circle, with a slight elongation along the optical bi-cone direction (northeast–southwest).

However, at such energies the wings of the energy-dependent telescope PSF spread a small but non-negligible amount of the nuclear photons to large radii. Thus, the presence of "extended" Fe Kα emission or Si Kα emission in the r > 1'' region may in fact come from the PSF-scattered unresolved nuclear emission.

To investigate this point, we further compared the observed radial profile of the Fe Kα emission band (6.2–6.5 keV) and the radial profile in the same band for a point-like nucleus simulated with MARX⁸ (see Wang et al. 2010b), shown in Figure 9(a). A similar comparison for the Si Kα (1.7–1.8 keV) is shown in Figure 9(b). The extracted observed profiles shown in Figure 9 agree well with the PSF profiles, except for the inner r < 2'' where the simulation overpredicted the observed emission due to heavy pile-up in the PSF core.

Zoom In Zoom Out Reset image size

The simulated PSFs indicate that the r ⩽ 2'' region contains 89% of the point-source emission in the 6.2–6.5 keV band and 94% of the point-source emission in the 1.7–1.8 keV band. Approximately 6% of the nuclear Fe Kα emission is expected to be spread within the 2'' ≲ r ≲ 6'' extended regions, the extent of Fe Kα seen in Figure 7(a). Taking into account the contribution from the PSF wings, the remaining extended Fe Kα emission is 5% ± 1% of the total Fe Kα emission. This should be considered an upper limit for the truly extended emission, because the nuclear Fe Kα emission is likely to be underestimated due to pile-up. Similarly, we estimated that the extended Si Kα emission is ≲ 4% ± 1% of the total Si Kα emission. As pointed out by the referee, the measured extended emission could easily disappear with a slightly broader simulated PSF given that the DitherBlur parameter in the MARX simulation is not known a priori.

Therefore, we conclude that although there is faint extended emission seen in the Fe and Si fluorescent line images, at most 5% of the observed Fe Kα emission and 4% of the Si Kα emission can be truly spatially extended; most of these emissions are consistent with the PSF scattering of a strong nuclear component.

Both our previous work using HRC image of NGC 4151 nuclear region (Wang et al. 2009b) and this work using ACIS find that most of the NLR clouds in the central 150 pc radius region of NGC 4151 have a relatively constant [O iii] to soft X-ray ratio (∼10; Figure 6 and Table 2), no matter the distance of the clouds from the nucleus. This ratio indicates a uniform ionization parameter and a density decreasing as r⁻². This may be consistent with a photoionized wind scenario (e.g., Bianchi et al. 2006) as the NLR clouds are outflowing (e.g., Kaiser et al. 2000). The [O iii]/X-ray ratios at the locations of jet–cloud collision are lower (∼3), leading to the suggestion of enhanced X-ray emission in these regions due to shock heating in addition to photoionization.

This conclusion is supported by the results of multiwavelength imaging studies. The high spatial resolution emission line images of O vii, O viii, and Ne ix emission show extended structures, in particular two "hot spots" in the Ne ix image, which are close to the radio knot features (Figures 2 and 3). At these locations (1'' from the nucleus), Mundell et al. (2003) noted that the morphology of the highly collimated radio outflows appears more disturbed, suggesting interaction with some dense clouds here. There are also clearly optical clouds with high velocity dispersion associated with the jet–cloud impact (Mundell et al. 2003). The radio knots are near the location of strong [Fe ii] emission (Storchi-Bergmann et al. 2009), falling between the peaks of the [Fe ii] emission.

Altogether, the morphology suggests a scenario where the radio ejecta runs into a denser cloud in the inhomogeneous interstellar medium (ISM) and results in locally enhanced shock heating (e.g., Capetti et al. 1999). This jet–cloud interaction has been suggested to explain the enhanced [Fe ii] emission, which could be due to the iron unlocked from grains by the shocks (Storchi-Bergmann et al. 2009). The Ne ix emission, which traces the higher ionization gas, becomes more prominent with the extra collisional ionization from the jet–cloud interaction. Armentrout et al. (2007) also noted that the poor fit to the O vii line profile in the grating spectra could be due to the presence of a non-photoionization component. This possibility is further supported by our spectral fits, which suggest that the Ne ix/O vii enhancements have a thermal origin, requiring the presence of a collisionally ionized plasma (the kT = 0.58 ± 0.05 keV component) in addition to a photoionized nuclear outflow (Yang et al. 2001; Wang et al. 2009b).

The emission measure of the APEC component allows us to estimate the electron density n_e (≈n_H) and the thermal pressure p_th (∼2n_ekT). The emitting volume (V) of the hot gas, assuming that the depth along the line of sight is comparable to the other dimensions, is 5.6 × 10⁶⁰ cm³ for the high Ne ix/O vii ratio hot spots. The filling factor η is assumed to be 100% and both n_e and p_th have a weak dependence on η (p∝η^−1/2). There is ∼10% uncertainty in kT, which is propagated to the estimated pressure and energy. For the kT = 0.58 keV emission, we derived an average density n_e = 0.06 cm⁻³, a hot gas pressure p_th = 6.8 × 10⁻¹⁰ dyne cm⁻², a thermal energy content of E_th = 4 × 10⁵¹ erg, and a cooling time of τ_c = E_th/L_APEC ∼ 10⁵ yr. The estimated internal pressure of the radio jet based on the synchrotron minimum energy assumption ranges between 10⁻⁷ and 10⁻⁹ dyne cm⁻² (Pedlar et al. 1993), to be confined by the hot gas pressure of 10⁻⁹ dyne cm⁻², which further supports our collisional ionization scenario.

Assuming T = 10⁴ K for the photoionized clouds and n_e = 220 cm⁻³ (Penston et al. 1990), the thermal pressure from the photoionized gas is p_ph = 3 × 10⁻¹⁰ dyne cm⁻², which is comparable to the thermal pressure of the hot ISM, implying a possible pressure equilibrium between the collisionally ionized hot gas and the photoionized line-emitting cool clouds.

We further estimate the age of the interaction from the approximate crossing time of the 05 region (dimension of the radio knot; Pedlar et al. 1993) with a characteristic velocity of c_s = 200 km s⁻¹, the local sound speed. This is approximately the thermal velocity for the 0.58 keV gas (v_th = 240 km s⁻¹) and also at the order of velocity dispersion of emission line gas seen in Storchi-Bergmann et al. (2010). Assuming strong shock jump conditions, a crude estimate of the shock velocity v_sh can be obtained using T ≈ 1.5 × 10⁵ K (v_sh/100 km s⁻¹)² (Raga et al. 2002). For the T ∼ 6.7 × 10⁶ K X-ray emitting gas, a v_sh of ∼700 km s⁻¹ relative to the downstream material is required. This appears consistent with a supersonic but relatively slow radio jet, which is constrained to be sub-relativistic from radio proper motion estimates (e.g., v_jet < 12,000 km s⁻¹; Ulvestad et al. 2005). Thus, we obtain a characteristic timescale of t_cross ∼ 10⁵ yr.

If the jet–cloud interaction converts kinematic energy into heating of the hot gas, a lower limit can be placed on the kinematic luminosity of the jet, L_K.E. ≳ E_th/t_cross = 1 × 10³⁹ erg s⁻¹. It is interesting to compare this observed L_K.E. to the jet power directly inferred from the synchrotron emission. Using the P_jet–P_radio scaling relation (Equation (1)) in Cavagnolo et al. (2010) and a total radio luminosity at 1.4 GHz (νL_ν ≈ 4 × 10³⁷ erg s⁻¹; Mundell et al. 1999), we find P_jet ∼ 10⁴² erg s⁻¹, suggesting that ≳ 0.1% of the jet power is deposited in the ISM. However, recalling that the P_jet–P_radio relation was derived from a sample of radio loud galaxies with X-ray cavities (see Cavagnolo et al. 2010 for review on the P_jet–P_radio scaling relation) while here NGC 4151 is of low radio power, we also estimate directly the jet energy flux using the pressure in the knots (Pedlar et al. 1993) based on the minimal energy assumption, giving a similar P_jet = 5 × 10⁴² erg s⁻¹.

We further note that the crossing time t_cross is comparable to the cooling time τ_c, implying that the hot gas is prominent only locally (close to the radio knots). Indeed, along the path of the radio jet, the clouds that are spatially close to the jet impact spots (e.g., clouds 3, 4 bracketing knot C2; clouds 9, 10 for knot C5) show [O iii]/X-ray ratios indistinguishable from others (Figure 6). If these clouds are physically related to the ISM interacting with the jet, instead of being projected close to the knots in our line of sight, their X-ray emission shows little evidence for excess shock heating in addition to nuclear photoionization. This localized heating is perhaps related to the highly collimated radio ejecta in NGC 4151 (Mundell et al. 2003), in contrast to the more expanded lobe-like radio outflow in Mrk 3 (Capetti et al. 1999), where the NLR gas appears broadly impacted. We plan to investigate in future work whether the [O iii]/X-ray ratio, such as the ≲4 measured here and seen in Mrk 3 and NGC 1386 (Bianchi et al. 2006), together with other line ratios (e.g., [Fe ii]/[PII]; Oliva et al. 2001), could be useful diagnostics for strong jet–ISM interaction in Seyfert galaxies.

In terms of energetics, we summarize our findings in NGC 4151: (1) The jet power estimated from the radio power is ∼10⁴² erg s⁻¹, a few percent of the current AGN energy output (L_bol = 7 × 10⁴³ erg s⁻¹; Kaspi et al. 2005), which is close to the canonical ∼5% of the radiant energy adopted in theoretical models for efficient AGN feedback (e.g., Di Matteo et al. 2005; Hopkins et al. 2006); (2) ≳ 0.1% of the jet power is observed to have been deposited into the host ISM in the nuclear region through the interaction between the radio jet and the dense medium; and (3) the shock heating due to the jet–cloud interaction appears localized to the impact spots, and most clouds are consistent with being photoionized by the nucleus.

Ogle et al. (2000) claimed that the narrow iron line emission was spatially resolved in the Chandra HETG observation and 65% ± 9% of the Fe i Kα emission comes from the ENLR at distance up to 6'' extent (∼400 pc across). This is intriguing, since the putative pc-scale torus or more compact broad-line region (Liu et al. 2010 and references therein) is expected to be the primary location of narrow Fe i Kα emission. Schurch et al. (2003) cautioned that due to the sensitivity of the short grating observation in Ogle et al. (2000), the 1–3 arcsec region off the peak of the cross-dispersion profile contains few photons. Therefore, the detection of spatially extended iron line emission cannot be significant. A high signal-to-noise XMM-Newton spectrum of the neutral Fe Kα suggests that all line flux originates in a nearly Compton thick torus that is not resolvable at current resolution (Schurch & Warwick 2002; Schurch et al. 2003).

Our results show that the extended Fe Kα emission, if present, is only at the ∼5% level of the observed Fe Kα emission when the contribution from the PSF wings is taken into account. This is consistent with our constraint on the extent of the Si fluorescent emission and supports the conclusions of Schurch et al. (2003), while it is in strong disagreement with the 65% reported in Ogle et al. (2000). We note that in Ogle et al. (2000) the PSF is represented as a narrow Gaussian with FWHM = 09, when the Chandra mission was new. The MARX model is now much more advanced so that a more reliable estimate becomes possible. At 6.4 keV, the encircled energy fraction of a point source is only 63.7% (see Chandra Proposers' Observatory Guide,⁹ Chapter 4, Table 4.2) in such a PSF core. Hence a significant fraction of the nuclear emission could have been measured in Ogle et al. as spatially extended Fe Kα emission.