A Hard X-Ray Test of HCN Enhancements As a Tracer of Embedded Black Hole Growth

The selection of these four sources is based on their global HCN/HCO+ line ratio. It has been noted that resolution effects can in some circumstances result in "contamination" of any HCN/HCO⁺ AGN signature by surrounding starburst activity (Izumi et al. 2016). This is a result of the general behavior that starbursts have HCN/HCO⁺ ratios ∼0.9–1.2 and many AGN have ratios ≳2. Measurement of the line ratio in increasingly larger apertures around AGN with intrinsically high ratios will be diluted due to the inclusion of regions of star formation with lower ratios. This would act to reduce the measured line ratio in larger apertures, which include increasing contribution from the starburst, and may explain some instances where coarse resolution measurements of AGN show "starburst" line ratios.

In contrast, globally enhanced HCN/HCO⁺ line ratios in starburst galaxies are not straightforward to explain via contamination from extended emission associated with star formation. As any contamination to these global apertures from a starburst decreases the line ratio (as described above), attempting to correct for the contamination would increase the inferred HCN/HCO⁺ line ratio. Global line ratios that are enhanced relative to the values typical for star formation are likely reliable indicators of nuclear enhancements (e.g., Privon et al. 2017). Thus, the use of single-dish spectra to select these objects as having elevated HCN/HCO⁺ line ratios is unlikely to result in false positives due to contamination from extended star formation.

The four targets were observed by NuSTAR for ∼20 ks each (see Table 1), in Cycle 3 (Project 3190, PI: G. C. Privon). The data obtained by the two NuSTAR focal plane modules, FPMA and FPMB, were processed using the NuSTAR Data Analysis Software nustardas v1.8 within Heasoft v6.24, using the calibration files released on UT 2018 August 18. The cleaned and calibrated event files were produced using nupipeline following standard guidelines. For all sources we extracted a spectrum at the position of the counterpart, selecting circular regions of 45'' radius centered on the position of the source. For the background spectra we used an annular region centered on the source with inner and outer radii of 90'' and 150'', respectively.

Two of the four sources (IRAS F06076–2139 and NGC 7591 ) have archival Chandra (Weisskopf et al. 2000) ACIS (Garmire et al. 2003) observations. The Chandra ACIS data were reduced following the standard procedures, using CIAO v.4.7.9. We reprocessed the data using the chandra_repro task and extracted source spectra using circular regions of 5'' radius. For the background we used circular regions of radius 10'', selected in regions where no other source was present. Both objects show point-like sources coincident with the optical counterparts.

The other two sources, NGC 5104 and NGC 6907 were observed several times by the XRT detector (Burrows et al. 2005) on board the Neil Gehrels Swift Observatory (Gehrels et al. 2004). We used all the data available (for a total of 10.3 ks and 19.9 ks, respectively), processing it using the xrtpipeline within heasoft v6.24 following the standard guidelines.

Out of the four sources, an indication of an obscured AGN is only present for IRAS F06076–2139. However, the AGN contributes a minor fraction (3%) to the total luminosity of the system and is likely not spatially associated with the molecular gas.

The NuSTAR upper limits enable us to place joint constraints on ${L}_{\mathrm{AGN}}$ and ${N}_{{\rm{H}}}$ for the other three sources. We take the ${L}_{3\mbox{--}24\mathrm{keV}}$ upper limits and apply corrections for a range of obscuring columns determined from the Mytorus models of Murphy & Yaqoob (2009). We then convert this to an upper limit on ${L}_{\mathrm{AGN}}$ (for that assumed ${N}_{{\rm{H}}}$) using a ${L}_{3\mbox{--}24\mathrm{keV}}$ bolometric correction of 13.6 derived from Vasudevan et al. (2009). In the bottom row of Figure 1 we show the joint constraints on ${N}_{{\rm{H}}}$ and ${L}_{\mathrm{AGN}}$ based on this approach. These NuSTAR observations indicate that AGN cannot contribute more than ∼10% unless the obscuration is >10²⁵ ${\mathrm{cm}}^{-2}$.

In the bottom row of Figure 1 we also show the average bolometric AGN fractional contributions, f_AGN, computed by performing a survival analysis on the individual AGN fractions derived from both atomic and dust MIR tracers of AGN activity (Diaz-Santos et al. 2017). In the case of these four objects, none had [Ne v] detections. Two sources (NGC 5104 and NGC 6907) had [O iv] detections, though the inferred AGN fractions are consistent with no contribution of the AGN to the mid-infrared luminosity. The PAH EQW (see Table 1 and Stierwalt et al. 2013) was also included in the AGN fractions computed by Diaz-Santos et al. (2017). The AGN fraction derived from the PAH EQW requires the adoption of a baseline value for a source free of supermassive black hole (SMBH) accretion. In our original identification of these sources as pure starburst or starburst-dominated (Privon et al. 2015), we adopted the pure starburst value from Brandl et al. (2006). In contrast, when computing the AGN fractions, Diaz-Santos et al. (2017) assumed a higher PAH EQW reference value typical of normal star-forming galaxies, following Stierwalt et al. (2013) and Veilleux et al. (2009). The baseline EQW value for a pure starburst is lower than that of a normal star-forming galaxy because in the former, a larger fraction of the starlight is reprocessed by dust and subsequently emitted as infrared continuum, resulting in brighter continuum and lower EQW.

As the Diaz-Santos et al. (2017) mean AGN fractions were computed using a higher PAH EQW than is seen in pure starbursts, we consider the f_AGN values in Table 1 and Figure 1 as upper limits to the true AGN fraction for these systems. We defer a detailed examination of the agreement between hard X-ray and other multiwavelength AGN indicators (including those in the MIR) to C. Ricci et al. (2020, in preparation), but later briefly comment on scenarios varying the distribution of gas and dust, and the impact on these AGN diagnostics (Section 3.3.2). If these MIR indicators are accurate, they would imply the NuSTAR observation would be consistent with ${N}_{{\rm{H}}}$ > 10²⁵ ${\mathrm{cm}}^{-2}$. However, such column densities imply that the MIR would be optically thick, and therefore diagnostics that use spectral features in this range may not be accurate. In addition, modeling of dust obscuration of starburst and AGN energy sources suggests that modest unobscured lines of sight to a starburst can dilute mid-infrared signatures of AGN (Marshall et al. 2018).

One possible explanation for the lack of detected hard X-ray emission in the three sources is extremely high column densities, in excess of ∼10²⁵ ${\mathrm{cm}}^{-2}$. Such high levels have been reported for other systems (Sakamoto et al. 2013; Scoville et al. 2015; Aalto et al. 2019) and it would be extremely difficult to see X-rays from a sufficiently buried AGN, particularly if the covering factor is ∼1. However, a sufficiently luminous, buried energy source may radiatively pump HCN molecules, leading to HCN-VIB emission (e.g., Ziurys & Turner 1986; Aalto et al. 2015). Thus, if there are energetically important AGN with sufficiently high obscuration that we do not see it with NuSTAR, we may expect HCN-VIB emission in the (sub)millimeter.

Limits on the HCN-VIB (4–3) emission will be presented by D. Jeff et al. (2020, in preparation), based on ALMA Band 7 observations of these four sources. We use those HCN-VIB upper limits in conjunction with Figure 13 of González-Alfonso & Sakamoto (2019) to determine upper limits for ${L}_{\mathrm{AGN}}$/${L}_{\mathrm{IR}}$ as a function of column density.²⁷ These constraints are shown as the dotted lines in the bottom row of Figure 1 and suggest that any AGN with ${N}_{{\rm{H}}}\gt {10}^{25}$ ${\mathrm{cm}}^{-2}$ would likely contribute less than 10% of the bolometric luminosity to each system.

If the column density is high enough to effectively prevent any X-rays from escaping, they may also be unable to penetrate sufficiently deeply into the molecular gas to drive changes in the chemistry. Based on the NuSTAR curves of Figure 1, we tentatively rule out the existence of AGN contributing more than ∼10% of ${L}_{{\rm{bol}}}$ in NGC 5104, NGC 6907, and NGC 7591, even if the AGN is heavily obscured. The Meijerink et al. (2007) XDR models indicate that HCN/HCO${}^{+}\gt 1$ due to X-ray driven chemistry is expected only when F_X > 10 erg s⁻¹ cm⁻² (incident flux on the cloud surface, over the energy range of 1–100 keV, where they assume a power-law energy spectrum of F(E) ∝ E^−0.9). We use the observed upper limits on the hard X-ray flux from the AGN to place limits on the X-ray flux at the location of the circumnuclear material where the HCN and HCO⁺ line emission is expected to arise. Taking our 3–24 keV limits and correcting to the 1–100 keV energy range (assuming the same α = 0.9) we estimate the flux experienced by molecular clouds in the nuclear regions of these systems is at least a factor of 10 below what is predicted to lead to HCN enhancements. This suggests that the X-ray flux currently incident on circumnuclear clouds is unlikely to be sufficient to drive the chemistry necessary to explain the HCN enhancements. In Section 4.1 we discuss the timescales of significant AGN variability and the chemical evolution of the gas.

For a scenario in which there is an AGN, the MIR and hard X-ray discrepancy may be resolved if the X-rays are obscured by gas that is largely within the dust sublimation radius. In this case, the small-scale dust-free gas would obscure the X-rays, but the dust on several parsec scales and larger would feel the influence of the reprocessed optical/UV radiation. This picture is potentially supported by observations which find that optical extinctions underestimate the X-ray obscuring columns toward AGN (e.g., Alonso-Herrero et al. 1997; Merloni et al. 2014; Burtscher et al. 2016), though as those authors discuss, other scenarios can also explain the data.

The NuSTAR upper limits for the three nondetections indicate that values of ${L}_{{\rm{bol}}}$ ≳ 10⁴³ erg s⁻¹ would require ${N}_{{\rm{H}}}$ ≳ 10²⁵ ${\mathrm{cm}}^{-2}$ (Figure 1). This bolometric luminosity corresponds to a dust sublimation radius of 0.04 pc (see Nenkova et al. 2008). Assuming a spherically symmetric gas configuration with ${N}_{{\rm{H}}}$ = 10²⁵ ${\mathrm{cm}}^{-2}$ contained within the dust sublimation radius suggests a mean density of ≈8 × 10⁷ cm⁻³.

A consequence of this scenario is that the heavy attenuation of the X-rays on small size scales would ensure that no significant X-ray flux would be seen by the molecular clouds in the nuclear regions (e.g., Section 3.3.1). Thus the X-rays from an AGN with such a high dust-free obscuring column could play no role in driving the chemistry of molecular clouds or influencing the HCN/HCO⁺ line ratio.

We ran a radiative transfer experiment with CLOUDY (Ferland et al. 2017) of a ${L}_{{\rm{bol}}}$ = 10⁴³ erg s⁻¹ luminosity AGN embedded in a uniform, dust-free, solar metallicity gas cloud with a total column density of ${N}_{{\rm{H}}}$ = 10²⁵ ${\mathrm{cm}}^{-2}$ and an outer size of 0.04 pc. Effectively, no ionizing photons (E > 13.6 eV) escape this obscuration. This lack of ionizing photons would likely prevent MIR indicators such as [Ne v] and [O iv] from being present. We thus conclude that the scenario of dust-free obscuring columns is not a viable mechanism for explaining any apparent tension between the MIR and X-ray results for these systems.

We argue in this work that there is no evidence that HCN enhancements are convincingly linked to presently observed AGN activity (diagnosed by the hard X-rays). However, it is possible that the AGN can have an impact on the molecular gas properties, without molecular tracers being useful diagnostics of whether SMBH accretion is present. If the timescale for any of the above physical processes is significantly different from the AGN turn on or shutdown timescales, the HCN enhancements would not be a good tracer of AGN. Thus, assessing the timescales for the above physical processes is relevant to understanding the physical origin of HCN enhancements (despite the fact that these enhancements do not appear to be robust in identifying presently accreting SMBHs). Looking toward X-ray and CR-driven chemistry, Viti (2017) studied the time-dependent chemical evolution of molecular gas after an enhancement of cosmic-ray flux, meant to mimic the impact of cosmic-rays and/or X-rays from an AGN. Their chemical model consisted of two phases, where Phase I simulated the formation of molecular dense gas under galactic conditions, while in Phase II the cosmic-ray flux was varied from the galactic value to several factors higher. Viti (2017) showed that, if the cosmic-ray ionization rate is not enhanced compared to the galactic one, even with constant cosmic-ray ionization rate, the simulated molecular clouds had not reached a chemical steady state after 1 Myr.

To explore the chemical evolution after the cessation of accretion, we extend the modeling of Viti (2017) by adding a Phase III, where the enhanced cosmic-ray flux is reduced back to the average Galactic flux and the chemical evolution is followed for an additional 1 Myr (S. Viti et al. 2020, in preparation). For the details of the included physical processes we refer the reader to Viti (2017). In Figure 4 we show the time evolution of the HCN/HCO⁺ ratio³¹ for several J_upper levels as a function of time. Phase II (enhanced cosmic-ray flux) occurs between 0–1 Myr and Phase III (Galactic cosmic-ray flux) is between 1 and 2 Myr. The line ratios show significant evolution in both phases, but we note that during and after Phase III the line ratios do not return to their state prior to the onset of the enhanced cosmic-ray flux (i.e., Prior to Phase II). This lends further support to the conclusion that there is no one-to-one mapping of line ratios with AGN activity.

Zoom In Zoom Out Reset image size

The last relevant physical mechanism is emission from HCN-VIB. This requires a warm (T > 100 K), high column density source; also, the strength of HCN-VIB is likely promoted if HCN has an enhanced abundance. The radiative timescale for HCN-VIB is short, so the lifetime of HCN-VIB emission is set by the cooling time for the central molecular gas, as the primary consideration is the existence of the T_b > 100 K radiation field. We use DESPOTIC (Krumholz 2014) to estimate the cooling time for a typical HCN-VIB source. We include CO, C, O, and H₂O as cooling species, with properties drawn from Schöier et al. (2005). All species were set to standard abundance values, except H₂O, which was set to a relative abundance of 10⁻⁶ based on modeling of Herschel observations in similar sources (González-Alfonso et al. 2014). The CO, ortho, and para-H₂O species together dominate the cooling luminosity and all contribute approximately in equal measure (${\rm{\Lambda }}\approx 8\times {10}^{-24}$ erg s⁻¹ nucleus⁻¹ each). Adopting fiducial values for the highly obscured CONs (${T}_{\mathrm{gas}}={T}_{\mathrm{dust}}\,=200$ K; n_H = 10⁵ ${\mathrm{cm}}^{-2}$; ${N}_{{\rm{H}}}$ = 10²⁵${\mathrm{cm}}^{-2}$; Aalto et al. 2019), we find a cooling time of ∼100 yr. This is considerably shorter than AGN flickering timescale.

For highly obscured systems (${N}_{{\rm{H}}}$ > 10²⁵ ${\mathrm{cm}}^{-2}$) the radiative diffusion timescale is ∼10⁴ yr (González-Alfonso & Sakamoto 2019) and the H₂O lines are typically seen in absorption (González-Alfonso et al. 2012, 2013) and thus will not contribute to cooling. If line cooling is inefficient, then the radiative diffusion time will set the cooling time and it will be ∼10⁴ yr. Together the cooling time and photon diffusion time suggest HCN-VIB emission (and associated blending with HCN v = 0 lines) should disappear approximately in step with SMBH accretion (if accretion is the dominant energy source).

Though the HCN-VIB emission would likely disappear on timescales similar to that of >1 dex AGN flickering, the chemical evolution may be much slower. Thus, HCN enhancements can persist long past the time in which an AGN has turned off, consistent with our finding that HCN enhancements are poor indicators of current SMBH growth.