SPATIALLY RESOLVED CHEMISTRY IN NEARBY GALAXIES. III. DENSE MOLECULAR GAS IN THE INNER DISK OF THE LIRG IRAS 04296+2923

The high ratios toward the CNZ are not extreme compared to other LIRGs and imply low-opacity CO gas. A range of LTE solutions for gas opacity and isotopic abundance are possible. The optical depth of CO(1–0), ¹²τ, can range from ≃ 4.0, (corresponding to [CO/¹³CO] ≃ 60, and [CO/C¹⁸O] ≃ 200) to ≃ 8.0, (for [CO/¹³CO] ≃ 120 and [CO/C¹⁸O] ≃ 375). Pushing ¹²τ larger than 8.0 requires very large [CO/¹³CO] abundance ratios, greater than 120. These values would be consistent with the low nuclear processing levels observed for gas in the outermost parts of the Galaxy, based on the Wilson & Rood (1994) extrapolated gradient (there are indications that this gradient may be too shallow in the far-outer Galaxy; Wouterloot & Brand 1996), and with values of [¹²C/¹³C] ∼100 recently suggested for the centers of local starburst galaxies (Martín et al. 2010; Henkel et al. 2014). On the other hand, pushing ¹²τ lower than 4.0 requires [CO/C¹⁸O] abundance ratios lower than seen anywhere in the Galaxy and lower than the fully massive star enriched values predicted from chemical evolution models (Henkel & Mauersberger 1993). Toward the starburst site, these constraints are even more dramatic if LTE applies (see below). Overall, low values of ¹²τ are favored for the CNZ based on a comparison with the northern arm (Section 4.1.2).

Low CO opacities in CO-bright nuclear starburst may seem surprising. Even ¹³CO has moderate, not low opacity in the more normal spirals, IC 342 and Maffei 2 (Wall & Jaffe 1990; Meier et al. 2000; Meier & Turner 2001; Meier et al. 2008). The presence of the starburst and its location at the center of a barred potential alters the situation somewhat: the CO line width of the nuclear emission is about twice the line widths of the bar ends. The broader line accounts for some of the increased CO intensity toward the nucleus, but not all. The C¹⁸O peak brightness temperatures toward the CNZ and the northern arm differ by less than a factor of two.

Non-LTE effects can also be responsible for the high $\mathbb {R}_{13}$ and $\mathbb {R}_{18}$ values, especially toward the starburst. The CO(1–0)/HCN(1–0) and HCN(1–0)/HCO⁺(1–0) line ratios over the CNZ argue that the gas is dense enough that sub-thermal CO emission can be neglected across the nucleus (Section 4.2.2). However, the chemical data indicate that the nuclear molecular emission may be partially influenced by photon-dominated regions (PDRs; Section 4.3). The radiation field from the starburst can heat the surfaces of the nearby clouds, preferentially exciting the optically thick (¹²CO) transitions relative to the optically thin ¹³CO and C¹⁸O. The high radiation fields can also preferentially photo-dissociate the optically thinner species. Both mechanisms raise $\mathbb {R}_{13}$ and $\mathbb {R}_{18}$ relative to their values in quiescent conditions (e.g., Meier et al. 2000). Since ¹³CO is not highly opaque, there should be a much weaker influence on $\mathbb {R}_{1318}$. Non-LTE effects are able to explain the elevated ratios toward the starburst relative to the CNZ, but must be extreme to change the conclusion that the CNZ has modest ¹²τ.

So even accounting for non-LTE effects associated with the starburst, it appears that the observed nuclear CO isotopic line ratios imply quite low ¹³CO opacities ¹³τ ≪ 1, [CO/¹³CO] ⩾ 60 and [CO/C¹⁸O] ≳ 200. The isotopic abundances ratios are consistent with an ISM enriched in ¹⁸O from recent massive star ejecta. Moreover, a relatively lower abundance of ¹³C suggests less long term nuclear processing, since C is a primary and ¹³C is a (mostly) secondary nucleus (Henkel & Mauersberger 1993).

Adopting [CO/C¹⁸O] ≃ 200 and an excitation temperature of 30 K along with a CO/H₂ abundance ratio of 8.5 × 10⁻⁵ (Frerking et al. 1982), we derive an H₂ column density of N(H₂) ≃ 1.0 × 10²³ cm⁻² toward the CNZ. This is about a factor of two lower than predicted based on a Galactic conversion factor of 2.0 × 10²⁰ cm⁻² (K km s⁻¹)⁻¹ (Hunter et al. 1997; Strong et al. 1988). The Galactic conversion factor can be accommodated either by adopting a nuclear excitation temperature of ∼50 K over ∼200 pc scales, or raising [CO/C¹⁸O] to ∼340.

Toward the bar arms, the isotopic line ratios are very low compared to what is typically observed for disks in other extragalactic systems (e.g., Paglione et al. 2001). Under LTE, low $\mathbb {R}_{13}$ and $\mathbb {R}_{18}$ imply large ¹²τ (∼65–110). It is counter-intuitive that the northern arm would have gas opacity at least an order of magnitude larger than the CNZ, while its CO intensity is more than an order of magnitude fainter. For these high opacities, the observed $\mathbb {R}_{1318}$ imply [¹³CO/C¹⁸O] ∼2.1, which is lower than favored for the CNZ and more than three times lower than found in either the local Galactic ISM or the Galactic center region.

Assuming LTE it is difficult to obtain a consistent solution for these line ratios. To simultaneously match the three ratios at the end of the northern arm, we require [CO/¹³CO] ≃ 90–140 (unusually high) and [CO/C¹⁸O] ≃ 185–300, as ¹²τ ranges from 65 to 110. Given the strong bar here it is possible that the radial inflow of relatively unprocessed outer disk gas could explain the high [CO/¹³CO] ratio observed for the bar arms, however this explanation would imply that the [CO/C¹⁸O] abundance ratio should also be raised, which is not observed. These [CO/¹³CO] ratios suggest that stellar processing on the longer timescales typical of intermediate-mass star lifetimes is rather low.

If we accept the high end of the opacity range for the CNZ and the low end of the range for the end of the northern arm (to minimize dramatic opacity differences) then the implied abundance ratios [CO/¹³CO] and [CO/C¹⁸O] would both decline with galacto-centric radius. This is at odds with the measured gradients in the Galaxy and chemical evolution models. To have [CO/¹³CO] and [CO/C¹⁸O] gradients in the right sense we are forced to favor low opacity for the CNZ and high opacity for the outer arms. However, even this is not particularly satisfying because with high ¹²τ in the arms, [CO/C¹⁸O] ≃ 300. Moreover, if we adopt T_ex ≃ 10 K and this abundance the implied N(H₂) column is almost five times that obtained using the Galactic conversion factor. Such high conversion factors may be seen in low metallicity systems, but not in solar metallicity gas (e.g., Bolatto et al. 2013)

The anomalies in the ratios demand that we consider non-LTE effects. The PDRs that were discussed in the nuclear region are not relevant in the bar arms because strong star formation is absent. Two other non-LTE effects worth considering are (1) sub-thermal excitation of the isotopologues and (2) ¹³CO chemical fractionation in the arm medium versus the CNZ. The second we dismiss because we observe [¹³CO/C¹⁸O] to be smaller than toward the CNZ. Since ¹⁸O is not expected to fractionate efficiently in the cold ISM, fractionation would act to raise the [¹³CO/C¹⁸O] abundance ratio relative to the nucleus, not lower it. Moreover, detailed studies find little evidence for ¹³CO fractionation actually being observed in the ISM (e.g., Milam et al. 2005).

Sub-thermal excitation is a viable explanation, especially since the clouds in the arms likely have lower densities than the nuclear disk. The large opacity of CO lowers its effective critical density relative to the CO isotopologues by the escape probability factor (β ∼ 1/¹²τ). Therefore when gas densities drop below ${\sim} n_{H_{2}} \le 10^{3.5}$ cm⁻³, the brightness temperatures of the CO isotopologues can drop relative to CO, inflating $\mathbb {R}_{13}$ and $\mathbb {R}_{18}$ (Meier & Turner 2001). This agrees with the observed behavior only if the true $\mathbb {R}_{13}$ and $\mathbb {R}_{18}$ are unity (¹²τ → ∞).

If we adopt the ¹²τ ≃ 80 solution for the end of the northern arm, the lowest that keeps the correct sense of the abundance gradient, then [CO/¹³CO] ≃ 110 and [CO/C¹⁸O] ≃ 225. For an excitation temperature of 10 K, the observed ¹³CO and C¹⁸O intensities imply areal filling factors of f_a ∼ 0.03. This is reasonable for the bright, extended molecular gas disk characteristic of IRAS 04296. Adopting the above abundances and temperature for the arm, N(H₂) ≃ 3.8 × 10²² cm⁻² at the outer end of the northern arm. This column is larger than that from a Galactic conversion factor but is not extreme; we will use it to estimate abundances toward the northern arm. Until J = 2–1 transitions are observed so the gas excitation can be directly measured, implied conversion factors must be regarded as suspect, but it is clear that the conversion factor for the CNZ is lower than in the arms (Meier et al. 2010), as it is in many galactic centers.

We have seen that CO isotopic abundance ratios in IRAS 04296 are anomalous compared to the Galaxy and other local galaxies. The observations further show that the physical conditions of the molecular gas in the arms are very different from that observed in the CNZ. Taking into account gas columns, absolute abundance ratios, and their variation with galacto-centric radius, we conclude that ¹²τ ∼ 4–6, [CO/¹³CO] ≃ 60 and [CO/C¹⁸O] ≃ 200 toward the CNZ. At the outer end of the bar arm these values have changed to ∼80, 110, and 225, respectively. The [CO/C¹⁸O] ratios are fairly low, consistent with the inner Galaxy value and with enrichment from massive star ejecta over much of the inner r ∼ 3.5 kpc. However, the inner disk [CO/¹³CO] values are like those ≳10 kpc out in the Galactic disk. Recent observations suggest high ¹²C to ¹³C isotopic abundance ratios of [¹²C/¹³C] ∼ 100 may be common in other starburst nuclei (Martín et al. 2010; Henkel et al. 2014). Hence, if [CO/C¹³O] can be a proxy of longer term nucleosynthetic processing, the underlying disk of IRAS 04296 and possibly other starbursts are less processed than most of the Galaxy.

Meier et al. (2010) concluded from the molecular gas and dynamical masses that IRAS 04296 must be experiencing one of its first major starburst episodes. The isotopic abundances agree with this initial burst scenario, but also show that the current burst is mature enough to have enriched the nuclear disk and potentially the inner bar with massive star ejecta. In this picture IRAS 04296's evolutionary state parallels the Large Magellanic Cloud (LMC): a relatively unprocessed galaxy with a current burst of massive star formation. However, the LMC appears to have a completely different abundance pattern, being somewhat enriched in ¹³CO, extremely depressed in C¹⁸O (e.g., Wang et al. 2009), and with no sign of local isotopologue variation across the galaxy (A. Hughes et al. 2014, in preparation). Finding the metallicity of IRAS 04296 could help determine the stage of enrichment.

Since CO(1–0) traces low density molecular gas and HCN(1–0) high density, the CO(1–0)/HCN(1–0) line ratio is a measure of the fraction of gas that is dense. The tight correlation observed between HCN(1–0) and the SFR (Section 4.2.3), together with the compactness of nuclear star formation in IRAS 04296 (Meier et al. 2010), suggests that the dense gas fraction ought to increase toward the starburst.

This is indeed seen across the CNZ, with CO(1–0)/HCN(1–0) decreasing from ⩾12 at the outer edge of the nuclear disk to ∼7 at the starburst. Just southwest of the starburst the ratio even drops below six, at the low end of that found by (single-dish) surveys of nearby [U]/LIRGs (e.g., Solomon et al. 1992; Aalto et al. 1995; Gao & Solomon 2004; Baan et al. 2008). Even averaged over the inner 20'' diameter, (close to single-dish sampling scales), the CO(1–0)/HCN(1–0) line ratio is <8. We see that the fraction of dense molecular gas, $n_{H_2}\gtrsim 10^{4.5}$ cm⁻³, in the CNZ is high near the nuclear starburst, and falls away from the starburst.

Along the arms HCN(1–0) is only tentatively detected in a few locations (see Figure 5), so there are just a few isolated regions with elevated dense gas fractions. Given the much fainter CO(1–0) at these arm locations, they have CO(1–0)/HCN(1–0) limits that are comparable to or slightly higher than the starburst values. However, the CO isotopologues are likely better tracers of gas column (Section 4.1). If ¹³CO/HCN or C¹⁸O/HCN is used to constrain the dense gas fraction, then there is a pronounced decrease in the dense gas fraction between the CNZ and the bar arms, as expected given the lower star formation rate there.

Zoom In Zoom Out Reset image size

Line ratios between HCN(1–0), HCO⁺(1–0), HNC(1–0), and CN(1–0; 3/2–1/2) depend on gas physical conditions as well as chemistry. The dominant physical and chemical processes controlling the line intensities of these transitions have been extensively discussed in Paper II (Meier & Turner 2012) and the literature (e.g., Aalto et al. 2002; Graciá-Carpio et al. 2006; Meijerink et al. 2007; Papadopoulos 2007; Baan et al. 2008; Krips et al. 2008; Loenen et al. 2008; Kazandjian et al. 2012), so here we only briefly summarize.

The first-order physical parameter controlling these ratios is gas density. The critical density of HCO⁺(1–0) is nearly an order of magnitude lower than HCN(1–0), while HNC(1–0) has a critical density slightly lower than HCN(1–0). Therefore in the density range from 10^4–6 cm⁻³, the HCN/HCO⁺ depend on gas density, with HCO⁺ favored at lower densities relative to HCN. Furthermore, HCO⁺, being a molecular ion, has its abundance decreased in high density gas due to faster recombination with electrons (e.g., Papadopoulos 2007). Together these two effects suggest that HCN should be significantly brighter than HCO⁺ in high density gas. In normal gas phase chemistry the HCN/HNC intensity ratio is driven to unity through the mutual formation reaction HCNH⁺ + e⁻ → HNC/HCN + H (e.g., Talbi & Herbst 1998; Shiba et al. 1998). In environments that are hot or have experienced strong shocks or PDR irradiation, HNC can be rapidly converted to HCN so this ratio can deviate significantly from unity (e.g., Schilke et al. 1992; Turner et al. 1997). Thus we expect HCN/HNC intensity ratios to be near unity over a wide range of high density gas conditions. Where deviating from unity we expect HCN to be strongly favored in hot, disturbed gas. This is generally consistent with what we observe toward the more moderate starbursts, IC 342 and Maffei 2 in Papers I and II (Meier & Turner 2005, 2012). CN is expected to trace PDR gas and its chemistry is discussed in more detail in Section 4.2.4. Finally it should be noted that high optical depth in these lines will act to hide any differences in the physical and chemical behavior, driving all the ratios to unity.

Ratios of these three transitions are displayed in Figure 5. Both the HCN(1–0)/HCO⁺(1–0) and HCN(1–0)/HNC(1–0) line ratios exhibit the same east—west gradient across the CNZ, with high values at the western side. HCN(1–0)/HCO⁺(1–0) ranges from 0.7 to 1.5 and HCN(1–0)/HNC(1–0) from 1.3 to 2.5. The nuclear starburst resides approximately in the middle of the observed ratio gradient in both cases, so no clear evidence is seen for a distinct component directly associated with the compact starburst. The fact that the HCN/HCO⁺ intensity ratio is 1.0 toward the starburst suggests that densities are high enough to thermalize both transitions and that the opacities of both lines could be high. Large velocity gradient modeling of HCN (see Meier et al. 2008, for model) and HCO⁺ toward the starburst (not shown) imply that for a kinetic temperature of 40 K (the dust temperature; Table 1) densities are ∼2 × 10⁵ cm⁻³. The inferred HCN abundance (Table 3) is near that typically observed on large scales toward star forming clouds (e.g., Blake et al. 1987), so HCN opacities, while probably >1, are unlikely to be extreme.

Both HCN/HCO⁺ and HCN/HNC have the same sense of trend, rising toward the west across the CNZ. One possible model for these ratios is that the gas density is lower in the east away from the starburst, resulting in lower HCN/HCO⁺ there, and warm PDR or shocked gas is more prevalent in the west near the starburst, favoring HCN emission and producing the high HCN/HNC line ratio there. The starburst and CO column density peak, is located at the inner terminus of the northwestern bar arm, suggesting the western side of the nucleus to be hotter/more energetic (Meier et al. 2010). In this model, the elevated HCN/HCO⁺ is a better indicator of high gas density and HCN/HNC more reflective of high radiation fields. However opacity must also be considered. A second possible model is that line ratios near the starburst could instead be dominated by high HCN line opacity; if HCN is opaque but the two lower abundance species, HCO⁺ and HNC, transition from optically thick to thin from west to east, following the decline in CO column, then the decline in HCO⁺ and HNC brightness relative to HCN toward the east could be explained by their decline in opacity in this region. However, this second model is less favored since HCN/HCO⁺ (and HCN/HNC) on the western side are significantly above one, ratios naively inconsistent with high opacities.

Here we consider IRAS 04296, a spatially resolved LIRG, in the context of the global L_IR versus HCN(1–0) relations (e.g., Gao & Solomon 2004; Garcia-Burillo et al. 2012). The total HCN(1–0) luminosity over the CNZ is L_HCN = 7.4× 10⁷ K km s⁻¹ pc². This equates to L_IR/L_HCN = 1300 L_☉ (K km s⁻¹ pc²). This value is 1.5 times the global values of [U]LIRGs (Gao & Solomon 2004; Garcia-Burillo et al. 2012, with L_IR ≃ 10¹¹L_☉). However, the IR fluxes from IRAS cover a much larger area than the HCN emission. If we adopt L_IR(CNZ) ≃ 0.5 L_IR(tot) (Meier et al. 2010), then the observed nuclear value for IRAS 04296 is L_IR/L_HCN = 660, which is 40% smaller than the global values. That the CNZ in IRAS 04296 is slightly IR under-luminous normalized by the HCN compared to global averages is consistent with the determination focused tightly on the dense gas of the CNZ region.

Since HCN(1–0) is imaged at high resolution, a rough estimate is attempted to see if the correlation also persists within the galaxy. Meier et al. (2010) provide a detailed discussion of the rate, efficiency and distribution of star formation over the nucleus of IRAS 04296. However, the high frequency radio data does not sample the SFR on spatial scales comparable to HCN(1–0) and so we use a cruder SFR proxy, the 20 cm radio continuum map. Figure 3 displays the 20 cm radio continuum (Meier et al. 2010) compared with HCN(1–0) at matched spatial resolution. There is a clear linear correlation between the two. Adopting a q parameter value of 2.3 (q = log$[\rm L_{FIR}/3.75 \times 10^{12} W/m^{2}] - \rm log[S_{20cm}/10^{26} Jy]$), typical of what is commonly observed for star forming galaxies (Condon 1992), the observed L_IR/L_FIR ratio (with L_IR being the total luminosity from 8 to 1000 μm and L_FIR being the luminosity from 40 to 400 μm; Sanders et al. 2003), and Gao & Solomon (2004)'s conversion between L_IR and SFR allows the 20 cm map to be roughly converted to a resolved SFR map. The observed correlation between HCN(1–0) and 20 cm radio continuum intensity is

$$\begin{equation} \rm I_{20}(mJy \, bm^{-1}) = [0.38\pm 0.08] I_{{\rm HCN}}(K \,km\, s^{-1}) - [0.42\pm 0.35]. \end{equation} \tag{ 2 }$$

For the above normalization, after converting intensities to fluxes and luminosities, we find

$$\begin{eqnarray} \mathrm{S}FR(M_{\odot } \,{\rm yr}^{-1}) &=& [1.4\pm 0.3 \times 10^{-7}] L_{{\rm HCN}}(\rm K \,km\, s^{-1} \, pc^{2}) \nonumber\\ && - [0.02\pm 0.02]. \end{eqnarray} \tag{ 3 }$$

The quoted errors are statistical only and do not include (potentially large) systematic errors associated with uncertainties in adopted q and the conversion between L_IR and SFR. Discounting the very small zero-point offset, we find a slightly lower normalization than the SFR(M_☉ yr⁻¹) = (1.8 × 10⁻⁷) L_HCN(K km s⁻¹ pc²) value of Gao & Solomon (2004). Given the assumptions this is considered good agreement between the local and global values.

HCO⁺, which has a critical density lower than HCN's, shows a statistically identical relationship. Krumholz & Thompson (2007) suggest that the slope and normalization of the SFR vs. molecular gas tracer relationship should differ between species that are fully thermalized and those that are not. The fact that we do not see a difference is further evidence that the molecular gas localized to the starburst in IRAS 04296 has a density high enough to excite both HCN(1–0) and HCO⁺(1–0) equally well over much of the CNZ and that both may have moderate opacity. This finding is consistent with other observations of dense gas tracers in nearby, resolved star-forming galaxies (Zhang et al. 2014).

The interaction of UV photons from the starburst's massive stars with the molecular gas is expected to result in PDRs with distinctive chemical properties (e.g., Tielens & Hollenbach 1985; Sternberg & Dalgarno 1995). Given the strong nuclear star formation in IRAS 04296, does it show signs of radiative feedback from the starburst in the gas chemistry?

CN is expected to be abundant in PDRs. It forms from reactions between CH and N, with CH forming directly from H₂ and PDR-abundant C⁺(Sternberg & Dalgarno 1995). CN(1–0; 3/2–1/2) and HCN(1–0) have similar critical densities, so the HCN(1–0)/CN(1–0; 3/2–1/2) (hereafter HCN(1–0)/CN(1–0)) line ratio is considered to be an excellent tracer of dense PDRs (e.g., Sternberg & Dalgarno 1995; Boger & Sternberg 2005).

Toward the starburst in IRAS 04296 the HCN(1–0)/CN(1–0) line ratio reaches 2.4, which for LTE corresponds to an [HCN/CN] abundance ratio of ⩾0.6 and a CN abundance of 9.0× 10⁻⁹. (The abundance ratio is a lower limit because of the potentially large HCN opacity, whereas it is shown in Section 3 that CN is optically thin.) According to the models of Boger & Sternberg (2005), the above conditions are consistent with molecular gas at A_V ≃ 4–5 mag and a range of densities and radiation fields, χ. In the range A_V ∼ 4–6 the predicted column ratio changes dramatically (Boger & Sternberg 2005), so the implied A_V is not strongly sensitive to the optical depth of HCN as long as it is <10. Moderate densities (⩽10⁴ cm⁻³) and radiation fields (χ < 10³), where χ is the normalized Draine solar neighborhood radiation field at 1000 Å are slightly favored since as density and χ increase, peak CN abundances drop below 10⁻⁸ as ionization is pushed deeper into the cloud. Molecular gas at A_V ≃ 4–5 corresponds roughly to the transition between PDR and dark cloud conditions. So the starburst ISM traced by CN and HCN in IRAS 04296 appears to be influenced by PDRs, but not strongly. Furthermore, the fact that the HCN(1–0)/HNC(1–0) ratio > 1 over the CNZ favors PDR conditions. Localized chemical effects associated with X-ray dominated regions such as from a buried active galactic nucleus (AGN) are not pronounced (e.g., Meijerink et al. 2007).

To explain the total observed CN column density when $n_{H_{2}}\le 10^{4}$ cm⁻³ and χ ≲ 10³ with A_V ∼ 4–5 gas requires ∼45 PDR clouds along the line of sight. Using the same observed ratio, Boger & Sternberg (2005) and Fuente et al. (2005) estimate somewhat higher χs for the giant PDR in M 82, but fewer (10–20) numbers of PDR clumps. The number of clumps estimated here is 2–4 times larger than for M 82, the same factor by which the SFR is larger.

However, it is hard to explain why HCN(1–0)/CN(1–0) decreases toward the outer CNZ in the context of PDR gas. This implies that the abundance of the PDR tracer CN relative to HCN increases with distance from the starburst, contrary to expectation. One possible explanation is that the ionization remains high over the entire CNZ, via the weaker extended star formation component or elevated cosmic ray ionization rate and the gas away from the starburst being on average somewhat more diffuse. This model of more diffuse gas in the outer CNZ, away from the starburst, is also consistent with the lower HCN(1–0)/CO(1–0) ratio there, since diffuse clouds have lower total dense gas fractions.