Down but Not Out: Properties of the Molecular Gas in the Stripped Virgo Cluster Early-type Galaxy NGC 4526

One of the outstanding questions in modern astrophysics is understanding the origins of the Hubble sequence and the surprising diversity of nearby galaxies. Fortunately, nearby galaxies preserve some clues to their histories in their current properties. Cold molecular gas in early-type (elliptical and lenticular) galaxies is particularly interesting in this context; it falls outside our simple paradigm that early-type galaxies are quiescent and free of cold gas. In fact, almost 25% of nearby early-type galaxies have retained some molecular gas and at least 40% of them host atomic and/or molecular gas (Young et al. 2014; Davis et al. 2019) at a level of ≳10⁻³ M_⋆.

Detailed studies of the gas in early-type galaxies have provided important insights into their evolutionary histories. Comparisons between gas and stellar kinematics reveal frequent kinematic misalignments between the gas and stars, such that it is not unusual for the cold gas in early-type galaxies to be counterrotating with respect to the stars. That counterrotating gas cannot have had a long symbiotic relationship with those stars. Up to 30%–50% of the cold gas in early-type galaxies is sufficiently misaligned that it must have come in from outside after the bulk of the stars were formed (Davis et al. 2011). More recently, Davis & Young (2019) moved toward using metallicities as another, complementary set of clues to the evolution of early-type galaxies. Signatures of gas accretion can be seen in the fact that the metallicity of the ionized gas in early-type galaxies is occasionally lower than the metallicity of the stars. Ultimately it would be useful also to study the isotopic abundance patterns in early-type galaxies as further clues to their gas accretion and nucleosynthetic enrichment histories, employing the types of chemical evolution models that are frequently used for the Milky Way (Côté et al. 2019; Romano et al. 2019).

Beyond their accretion histories, early-type galaxies also allow us to probe the processing of galaxies in clusters. And because they have generally been resident in clusters for a much longer time than the spirals that are more commonly the subject of ram pressure stripping studies, they give additional insights beyond studies of spirals. They sample longer timelines for cluster-driven processing.

In this paper we present an exploration of molecular line ratios in the Virgo Cluster galaxy NGC 4526. The target is an unusually molecule-rich early-type galaxy with a striking but small silhouette dust disk, 1 kpc in radius. We focus on new Atacama Large Millimeter/submillimeter Array (ALMA) data for continuum emission, CO isotopologues, and high-density molecular tracers in the 3 mm band, supplemented by archival ¹²CO(2−1) and ¹³CO(2−1) data. We carry out radiative transfer modeling of the available isotopologue data, coupled with Bayesian inference techniques. The current data constrain the excitation-corrected [¹²CO/¹³CO] and [¹³CO/C¹⁸O] abundance ratios; we find the former to be roughly constant in the disk but find the latter to have a strong radial gradient, and those patterns provide clues to the processes driving the abundance gradients. We also estimate the CO/H₂ conversion factors α_CO using the dust continuum and CO images, providing the first estimate of this type for an early-type galaxy. This analysis lays the groundwork for future explorations of isotopic abundances in early-type galaxies, especially those in cluster members, which should give insights into the evolution of early-type galaxies and the processing of galaxies in clusters.

NGC 4526 is a Virgo Cluster member whose prominent central dust disk (Figure 1) has long prompted interest in its cold gas content. CO emission was detected in the galaxy in single-dish surveys by Sage & Wrobel (1989) and Combes et al. (2007); the first resolved images of its CO emission were published by Young et al. (2008). It is also a member of the ATLAS^3D survey (Cappellari et al. 2011), which provides additional information on its stellar and ionized gas kinematics (Krajnović et al. 2011), stellar populations (McDermid et al. 2015), and star formation history (Krajnović et al. 2020). It is a fast rotator and one of the more massive galaxies in the ATLAS^3D sample, with a stellar mass $\mathrm{log}({M}_{\star }/{M}_{\odot })\sim 11.1$ and global colors that place it firmly in the red sequence (Young et al. 2014). As it is a member of the Virgo Cluster, we assume a distance of 16.4 Mpc (Cappellari et al. 2011). A more recent measurement based on the tip of the red giant branch gives 15.7 ± 0.2 ± 0.4 Mpc (the uncertainties are statistical and systematic, respectively; Hatt et al. 2018), which is only 6% smaller than the distance we assume.

Zoom In Zoom Out Reset image size

Davis et al. (2011) studied the kinematics of gas and stars in the ATLAS^3D sample and noted that field early-type galaxies often display kinematic misalignments between their gas and stars, suggesting the gas was recently accreted from some external source. In contrast, Virgo Cluster early-type galaxies and those in other dense groups tend to have relaxed and well-aligned prograde gas. In this respect NGC 4526 is typical of cluster members.

It would have been difficult for NGC 4526 to acquire gas after falling into the Virgo Cluster, because the relative velocities of typical interactions (characterized by the velocity dispersion of the cluster) are much higher than the internal escape velocities of the galaxies themselves. Thus, unlike the cold gas in many field early-type galaxies, this gas has most likely been in NGC 4526 for several gigayears and has been processed through a long, continuous symbiotic relationship with the stars in multiple generations of mass loss and star formation.

The molecular gas in NGC 4526 is also undergoing star formation with an efficiency not too different from that of typical spirals. Davis et al. (2014) estimated a star formation rate of 0.2 M_⊙ yr⁻¹ based on 22 μm and far-UV emission; this gives a specific star formation rate (SFR/M_⋆) of 10^−11.8 yr⁻¹, a gas depletion time of 2.6 Gyr, and a star formation rate surface density Σ_SFR ∼ 0.09 M_⊙ yr⁻¹ kpc⁻². These values are all averaged over the whole molecular disk.

Previous observations of the cold gas in the galaxy have revealed some unusual and extreme properties. For example, Crocker et al. (2012) used the IRAM 30 m telescope to study molecular tracers in a subset of the ATLAS^3D early-type galaxies. They found that, relative to ¹²CO, NGC 4526 has unusually bright ¹³CO and HCN and a high HCN/HCO⁺ line ratio. It is also undetected in Hi emission (Lucero & Young 2013), which produces a remarkably high H₂/Hi mass ratio >60. This extreme deficit of atomic gas suggests that the low-density interstellar medium in NGC 4526 has been stripped by the intracluster medium (ICM). Its molecular properties should therefore give a valuable perspective on the effects of long residence in a cluster.

NGC 4526 was observed in ALMA's compact configurations in band 3, projects 2017.1.01108.S and 2018.1.01599.S, in 2018 April and 2019 April. We made use of the standard pipeline calibrated raw data and carried out continuum subtraction, imaging, and cleaning ourselves. An additional round of phase self-calibration was not found to offer any improvement in the images because of the relatively modest signal-to-noise ratios. For continuum subtraction we used zero-order or first-order fits to the line-free channels in the visibility domain; for imaging, we took a wide variety of images at varying channel widths and resolutions as were necessary to optimize the resolution or to match resolutions for resolved line ratios. The times on source, along with the fiducial beam sizes and rms noise levels, are indicated in Table 1 for the detected spectral lines. Emission was cleaned down to about the 1σ level and a primary beam correction was applied.

We also used ¹²CO(2–1) observations obtained with the CARMA array (Davis et al. 2013b). Those observations are relatively high resolution data, and were initially imaged at 025 resolution; for the analysis here, we tapered the visibility data to make a lower-resolution image and smoothed in the image plane to match the resolution of the ALMA ¹²CO(1–0) data. Appendix A describes the short-spacing verification and continuum subtraction on the ¹²CO(2–1) data.

For NGC 4526 we have near-simultaneous observations in the same telescope configuration, covering frequencies from 87 GHz to 110 GHz (with gaps) with similar UV coverage and time on source, so these data are suitable for imaging the continuum intensity and estimating the spectral index of any emission. We used all of the available line-free frequencies in CASA's multiterm, multifrequency synthesis deconvolver (Rau & Cornwell 2011), and we solved for two terms, i.e., the intensity and the spectral slope. When taken at a relatively high resolution of 09 × 08, the resulting continuum image at 99.33 GHz shows only a nuclear point source of flux density 4.79 ± 0.03 mJy and spectral index –1.07 ± 0.02. This nuclear source is clearly synchrotron emission. Kim et al. (2019) commented that the galaxy was not previously known to host an active galactic nucleus (AGN); however, they detected a nuclear X-ray point source, and combined with a 5 GHz point source (Nyland et al. 2016) and the 99 GHz synchrotron emission, these data are all consistent with the presence of an AGN.

Imaging at a lower resolution of 20 also reveals continuum emission from a low surface brightness disk, as shown in the contours in Figure 2. The total flux density associated with NGC 4526 in the 20 image is 6.68 mJy ± 0.08 mJy (statistical) ± 0.33 mJy (absolute calibration), of which 4.8 mJy is the point source and the remaining 1.9 mJy is the disk. The disk emission is rather faint and uncertainties on its spectral index are large, but individual pixels have a mean spectral index of 1.8 and an rms ≈ 0.7. This strongly rising spectral index is confirmed by individual images at the extreme frequencies. Using a high-resolution image (as described above) to isolate the flux density of the point source from that of the extended dust disk, we find the dust disk to have a flux density of 1.60 ± 0.15 mJy at 90.30 GHz, 1.98 ± 0.12 mJy at 101.69 GHz, and 2.46 ± 0.20 mJy at 108.33 GHz. The disk and nuclear continuum flux densities are also shown in the broadband spectrum in Figure 3, where they can be compared to the far-IR (FIR) flux densities and modified blackbody fit from Dustpedia (Nersesian et al. 2019); additional details are provided in Section 8. The disk's flux density measurements are remarkably close to the long-wavelength extrapolation of the modified blackbody fit.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The accuracy of the extrapolation in Figure 3 suggests that the extended 3 mm disk in NGC 4526 might be, like the FIR disk, modified blackbody emission from the dust. On the other hand, the good match in Figure 3 could also be a coincidence, if the dust emissivity is overestimated at 3 mm and a free–free contribution masks the overestimate. Star-forming galaxies usually show a combination of free–free and dust emission at 3 mm (Peel et al. 2011). If we interpret the disk flux densities above as a combination of flat-spectrum free–free emission and dust with an effective spectral index of 4.0, then we can reproduce the observed spectral index between 90 and 108 GHz as arising from a combination that is 50% free–free and 50% dust at 90 GHz. Thus, we infer that somewhere between half and all of the disk's 3 mm flux density is attributable to thermal dust emission. This fraction is somewhat larger than would be usual for starburst and spiral galaxies (e.g., Bendo et al. 2015, 2016), but as this is an early-type galaxy, the dust emission might be more strongly driven by heating from an old stellar population than by heating from young stars. More precise interpretations of the 3 mm continuum emission in NGC 4526 will require additional data at ≈200 to 300 GHz and/or 30 GHz, to help constrain the spectral energy distribution.

Figure 4 presents an integrated ¹²CO(1−0) intensity image, produced by masking regions containing real emission and then integrating within the mask. The mask was made using CASA's auto-multithresh algorithm, which clips each channel at 3σ and then extends the clipping region spatially by about a beamwidth. The molecular gas distribution observed here is consistent with the results shown by Davis et al. (2013b) and Utomo et al. (2015), who studied ¹²CO(2−1) at higher resolution (03) and lower sensitivity. We find a poorly resolved central peak, which they showed to be a fast-rotating disk or ring with a radius of about 05 (40 pc). Outside the nuclear peak are a dip in the surface brightness, a bright molecular ring at r = 2'' to 7'' (160 to 550 pc), and lower column density emission extending to about 15'' (1.2 kpc). The bright molecular ring is just interior to some recent star formation activity that is visible as blue stars in Figure 4 or as white regions in the unsharp-masked image in Figure 2. The peak ¹²CO(1−0) integrated intensities are 2.71 Jy beam⁻¹ km s⁻¹ in the central peak and 2.74 Jy beam⁻¹ km s⁻¹ in the molecular ring, when measured at 1.1 × 08 resolution. Assuming a standard conversion factor of 2.0 × 10²⁰ cm⁻² (K km s⁻¹)⁻¹ or 4.3 M_⊙ pc⁻² (K km s⁻¹)⁻¹, including He, and deprojecting to face-on assuming an inclination of 78° (discussed below), these values correspond to 240 M_⊙ pc⁻².

Zoom In Zoom Out Reset image size

The peak brightness temperature in ¹²CO(1−0), at 1.1 × 08 (75 pc) resolution, is 6.1 K; it is found in the molecular ring at a radius of 59. Somewhat lower brightness temperatures are found in the outer disk, with values of 1.0 to 3.5 K at radii of ≈ 12''. Peak brightness temperatures are also quite low (1.0 to 1.5 K) in the nucleus of the galaxy, though the column densities there are high because of the large line widths.

We created velocity fields for the molecular gas in NGC 4526 using two different methods, one using an intensity-weighted mean velocity and another fitting fourth-order Gauss–Hermite functions (i.e., including skew and kurtosis) to the spectrum at each location. Significant beam-smearing effects mean that the skew and kurtosis terms are necessary to accurately reproduce the line profile shapes and identify the peak in each spectrum. Figure 5 shows a CO velocity field and Figure 6 shows the velocity dispersion.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In this paper we do not focus on detailed kinematic and dynamic analysis of the gas disk, but we still need estimates of the kinematic position angle and inclination for deprojection and analysis of radial trends in the molecular properties. For this purpose we use a kinemetric analysis of the ¹²CO(1−0) velocity field, based on the code of Krajnović et al. (2006). ⁶ The galaxy shows a well-defined kinematic center that is coincident with the 3 mm continuum peak to better than 005. The kinematic position angle exhibits a slight twist from 2922 ± 02 at r = 2'' to 2938 ± 01 at r = 126. The fitted inclination varies only from 790 ± 02 at r = 3'' to 763 ± 03 at r = 126. Thus, tilted-ring models should be highly accurate for both deprojection and dynamical analysis.

Figure 7 shows the integrated spectra of the galaxy in all the detected lines, and Table 2 gives the corresponding integrated line fluxes and line intensity ratios. The line fluxes presented here are consistent with, though associated with significantly higher resolution and signal-to-noise than, the corresponding values for ¹²CO, ¹³CO, C¹⁸O, CS, HCN, HCO⁺, and CH₃OH measured with the BIMA and IRAM 30 m telescopes by Young et al. (2008), Davis et al. (2013a), and Crocker et al. (2012). Table 3 also gives the peak column density estimates for the various molecular species, assuming local thermal equilibrium (LTE) at excitation temperatures of 20 K and 10 K (Section 7). Column densities were calculated using the methodology in Mangum & Shirley (2015, Equation (80)), with molecular data from the LAMDA database.

Zoom In Zoom Out Reset image size

Integrated line intensity images (Figure 8) show that the distributions of all the detected molecules in NGC 4526 are broadly similar, and the molecules differ mainly in their signal-to-noise ratio and degree of central concentration. For example, the position–velocity slice in Figure 9 shows that HCN is more centrally concentrated than ¹²CO, as the ¹²CO/HCN ratio increases with radius.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We quantified the radial variations in the line ratios using three main techniques, which all gave similar answers. In method 1 we created matched-resolution integrated intensity images, computed a line ratio directly for every pixel, and used the orientation and inclination of the disk to associate each pixel with its deprojected radial distance. Method 2 was to construct matched-resolution major-axis position–velocity slices (e.g., Figure 9), integrate over radial regions, and compute the ratios of the corresponding sums in radial bins. Uncertainties for this method can be estimated from standard error propagation on the number of independent data points in the bin and also by varying the clip threshhold used to select pixels with the signal. (We always used the same data volumes for both lines, to eliminate biases related to different line strengths.) Method 3 was to construct integrated spectra by summing the data cube over elliptical annuli of the appropriate position angle and inclination. We then carried out a least-squares optimization modeling a fainter line's annular spectrum as a multiple of that of the brighter line. This technique works because the distributions of all the species are broadly similar, so that the corresponding annular spectra all have the same shape. The uncertainty in the ratio can be characterized with standard χ² goodness-of-fit criteria. Figures 10 and 11 show comparisons of these methods for the CO isotopologue ratios and for a few other selected line ratios, indicating good agreement for all of them. Figures 12 and 13 show the radial trends in other line ratios.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In this section we present a large-scale overview of the relative intensities of the molecular lines in NGC 4526 with some qualitative comments on unusual aspects. In Section 7 we undertake more quantitative analysis of the physical conditions in the molecular gas. To summarize the qualitative aspects, we note that NGC 4526 is relatively bright in ¹³CO and the high-density tracers (HCN, HCO⁺, HNC, CN, and CS), when compared to many spirals. It has relatively faint emission in the shock tracers CH₃OH and HNCO, though it should be noted that existing observations of those species are biased toward active and starbursting galaxies. All of the line ratios between ¹²CO and other molecules, with the singular exception of ¹³CO, show strong radial gradients of factors of two to four over a 1 kpc radius. In contrast, there is limited or unmeasurable radial variation in ¹²CO(1−0)/¹³CO(1−0) and the line ratios of any two high-density tracers, such as HCN/HCO⁺, HCN/HNC, HCN/CS, HCO⁺/CS, HCN/HNCO, and CH₃OH/HNCO.

6.1. CO Isotopologues

The ¹²CO(1−0)/¹³CO(1−0) ratio observed in NGC 4526 is 3.4 ± 0.3, which is one of the lowest values observed outside of the Local Group (Figure 14). Beyond the Milky Way, such low values are rare and noteworthy; some examples include those for NGC 4429 and NGC 4459 (two other Virgo Cluster early-type galaxies) and parts of resolved galaxies like the Small and Large Magellanic Clouds, Centaurus A, and IRAS 04296 + 2923 (citations in Figure 14; Meier et al. 2014). Spatial variations in ¹²CO/¹³CO in NGC 4526 are also remarkably small, being less than 5% and smaller than the corresponding statistical uncertainties. Resolved studies of ¹²CO/¹³CO within spiral galaxies usually show greater variation than those, sometimes increasing with radius and sometimes decreasing (Cao et al. 2017; Cormier et al. 2018; Topal 2020), though usually over larger radial ranges.

Zoom In Zoom Out Reset image size

The ¹³CO/C¹⁸O line ratios in NGC 4526 are not particularly unusual, as they range from 4.7 ± 0.3 to 8.7 ± 1.8, and they fall near the median of the values observed in nearby galaxies (Figure 15). But unlike ¹²CO/¹³CO, the ¹²CO/C¹⁸O and ¹³CO/C¹⁸O line ratios have significant radial structures. Figure 10 shows that the ratios involving C¹⁸O rise by a factor of 1.8 over this compact disk. The figure also identifies the location of the bright molecular ring, and the gradients in the C¹⁸O ratios are markedly larger exterior to the ring than they are interior to the ring. Thus the rise from ¹³CO/C¹⁸O = 5 to 9 occurs over the radial range of 0.4 to 1.0 kpc. For comparison, the same rise in the Milky Way ¹³CO/C¹⁸O(1−0) line ratios occurs over a radial range of about 6 to 12 kpc (Wouterloot et al. 2008). Additional comparisons to other spiral galaxies are discussed in more detail in Section 9.2.

Zoom In Zoom Out Reset image size

Resolved measurements of the ¹²CO(2−1)/(1−0) line ratio mostly range between 0.4 and 0.7 (Figure 11). These values are very typical for spiral galaxies (Leroy et al. 2021; Brown et al. 2021). Indeed, the star formation efficiency of the molecular gas in NGC 4526 is consistent with the values in many spiral galaxies even though its gas content is lower than most of theirs. Evidence for any spatial variations in the (2−1)/(1−0) line ratio is not compelling; the ratio might be slightly elevated toward the nucleus, but because of the uncertainty associated with continuum subtraction in the (2−1) data (Appendix A), the issue should be reevaluated with new data. Like NGC 4526, many spirals show small radial variations in the (2−1)/(1−0) line ratio, and sometimes the spirals show arm/interarm variations (den Brok et al. 2021, 2022).

6.2. High-density and Photodissociation Region Tracers

The range of ¹²CO/HCN values we measure in NGC 4526 is 8.0 ± 0.5 to 30 ± 8. The values in the outer disk of NGC 4526 are comparable to those in the spirals in the EMPIRE survey (Jiménez-Donaire et al. 2019), but only a handful of points in the spirals have bright enough HCN to match the center of NGC 4526. Of course, it is worth noting that the EMPIRE data have lower resolution than we are using here (1 to 2 kpc compared to 200 pc). Similar results also hold for CN and CS. ¹²CO/CS in NGC 4526 ranges from 27 ± 2 to 51 ± 6; the spirals in Gallagher et al. (2018) have typical values of ¹²CO/CS ∼ 90 in their central kiloparsecs. For the ¹²CO/CN ratios (isolating just the J = 3/2−1/2 blend), we measure values ranging from 11.4 ± 0.2 to 49 ± 10, or 17 ± 0.7 for the global line ratio. Other global ¹²CO/CN ratios measured by Wilson (2018) range from 20 to 150. In short, the high-density tracers in NGC 4526 are not exceptionally bright in comparison to ¹²CO but, especially for CS, they are factors of roughly 2 brighter than the median values for nearby spirals.

The CN radical is expected to be formed through photodissociation of HCN and/or through reactions involving C and C⁺, so it should be enhanced in regions with strong radiation fields. It may also be particularly enhanced in X-ray-dominated regions (Boger & Sternberg 2005; Meijerink et al. 2007). Ledger et al. (2021) found some supporting evidence for factor of two enhancements in CN/HCN ratios toward the nuclei of some starbursting galaxies and ULIRGs. In contrast, we find no such evidence for variations in CN/HCN in NGC 4526, or in CN/CS or HCN/HCO⁺ for that matter, even though there is an AGN in the center of the galaxy and a faint X-ray point source as well (Kim et al. 2019). The AGN in NGC 4526 may simply be faint enough that its effects on molecular line ratios are restricted to regions smaller than our current resolution.

In light of evidence that CS emission is enhanced in photodissociation regions (PDRs) (e.g., Lintott et al. 2005; Meier & Turner 2005), it has been suggested that CS might trace star formation activity even better than a standard dense gas tracer such as HCN. For example, Davis et al. (2013a) found that CS is enhanced relative to HCN in galaxies whose ionization is more strongly dominated by star formation activity than by an AGN. That study used unresolved single-dish measurements of the molecular lines and traced star formation activity through integrated [O iii]/Hβ line ratios. With our new ALMA data we can also test whether this trend applies in an internal, spatially resolved manner. NGC 4526 displays a wide range of [O iii]/Hβ values (Figure 16), from log([O iii]/Hβ) = −0.85 in its northwest quadrant to −0.2 in its nucleus and even higher values outside the molecular disk. However, we find no evidence of variable CS/HCN ratios anywhere in the galaxy, and no significant difference between the northwest quadrant and other parts of the galaxy. While the integrated measurements in NGC 4526 are consistent with a trend between low [O iii]/Hβ and high CS/HCN, the internally resolved measurements do not follow such a trend.

Zoom In Zoom Out Reset image size

The above discussion also highlights the fact that it is not yet clear whether internal variations of [O iii]/Hβ in NGC 4526 trace stochasticity in the star formation rate in the last ∼10 Myr, shocks, metallicity variations, or something else. The lowest [O iii]/Hβ ratios are on the receding side of the galaxy, which is the side with brighter CO emission (Figure 7). However, the asymmetry in [O iii]/Hβ is not reflected in the 3 mm continuum emission or in the 24 μm dust emission (Young et al. 2009). Additional work on other star formation tracers will be required before NGC 4526 can give a clear conclusion about CS/HCN and star formation activity or about internal variations in star formation efficiency.

6.3. CH₃OH and HNCO—Shock Tracers

7.1. CO Isotopologues

7.2. HNCO

7.3. HCN/HNC

7.4. CN Hyperfine Blends and Implications for the Other High-density Tracers

With an assumed dust emissivity and a gas-to-dust mass ratio, the total dust and H₂ masses can be obtained from dust continuum emission (e.g., Hildebrand 1983; Leroy et al. 2011, and many others). In the case of NGC 4526, we have argued (Section 4) that at least half of the 3 mm continuum is dust emission, so the image in Figure 2 can give loose constraints on the resolved H₂ column density estimates and CO/H₂ conversion factors. Here we assume all of the extended 3 mm continuum is dust emission. We use the modified blackbody emissivity model for NGC 4526 from the Dustpedia results (Nersesian et al. 2019): they show a dust temperature of 25.1 ± 0.8 K and a mass of (4.38 ± 0.58) × 10⁶ M_⊙, assuming a distance of 15.35 Mpc. These values derive from an emissivity of 0.640 m² kg⁻¹ (250 μm/λ)^1.79, which also predicts 7.4 × 10⁻³ m² kg⁻¹ at 99 GHz and gives a remarkably accurate prediction of the observed 3 mm flux density. We also assume an H₂/dust mass ratio of 100, based on the results of Cortese et al. (2016) for the Virgo Cluster and Hi deficient galaxies of stellar mass above 10^10.5 M_⊙.

Figure 17 shows the resolved CO-to-H₂ conversion factors α_X inferred for the three isotopologues under the assumptions detailed above. Table 4 also gives representative statistics for all of the pixels that have 3 mm continuum emission detected at a signal-to-noise ratio > 3, and that are free of synchrotron contamination (projected radius r > 2.6'', which is 1.3 × FWHM). Our estimated α_CO value for ¹²CO(1−0) in NGC 4526 is 5.6 ± 1.2 M_⊙ pc⁻² (K km s⁻¹)⁻¹, which is just 30% larger than the "standard" Milky Way value of 4.3 in the same units (e.g., Leroy et al. 2011). The result suggests that ¹²CO emission in NGC 4526 is just a bit fainter than usual, for a given H₂ surface density. Estimated values for α_CO under different assumptions may be scaled by noting that it is proportional to the gas/dust ratio and the fraction of the continuum intensity that is assumed to be dust, and that it is inversely proportional to the dust emissivity. For example, if only half of the continuum comes from dust, our inferred value of α for ¹²CO(1−0) drops to 2.8 ± 0.6 M_⊙ pc⁻² (K km s⁻¹)⁻¹. As a consistency check, Utomo et al. (2015) found that an α_CO value of 4.4 M_⊙ pc⁻² (K km s⁻¹)⁻¹ suggests that most of the molecular clouds in NGC 4526 are in virial equilibrium.

Zoom In Zoom Out Reset image size

Furthermore, as the RADEX/MCMC analysis gives confidence intervals on the species' column densities, the H₂ column density can be used to infer the species' abundances. In a continuum image at 99.33 GHz and 20 resolution, the surface brightness at r = 5'' is 0.12 ± 0.01 mJy beam⁻¹. Using the dust emissivity and gas/dust ratio described above, this specific intensity corresponds to an H₂ column density of 1.4 × 10²² cm⁻², or 300 M_⊙ pc⁻² (including helium). Both of those values have been deprojected to face-on column densities assuming an inclination of 78° (Section 5). The dust-derived H₂ column densities then give the CO isotopologue abundances listed in the last column of Table 4. For ¹²CO, the inferred abundance ranges from [¹²CO/H₂] ≈ 8 × 10⁻⁶ to 3 × 10⁻⁵, roughly factors of 3 to 10 lower than the "standard" Milky Way abundance of 8.5 × 10⁻⁵ (Frerking et al. 1982). The inferred ¹²CO abundance is proportional to the dust emissivity but is inversely proportional to the gas/dust ratio and to the fraction of the continuum that is assumed to be dust. Thus, for example, if only half of the 3 mm continuum emission came from dust, the ¹²CO abundance would be ≈2 × 10⁻⁵ to 6 × 10⁻⁵ and would be closer to the Milky Way value.

Our quoted ¹²CO abundance in NGC 4526 is farther from the "standard" Milky Way value than one might expect, given that α_CO is fairly close to the Milky Way value. In this context it is worthwhile to note that the ¹²CO abundance calculation takes optical depths into account, because it is based on the RADEX/MCMC results. In contrast, the α_CO calculation uses the raw line intensities and does not account for variations in optical depth. Lower-than-usual ¹²CO optical depths in NGC 4526 might explain relatively normal α_CO values despite low ¹²CO abundances.

We do not place much significance on the differences between the α_CO and [¹²CO/H₂] values for NGC 4526 and the Milky Way, because the uncertainties associated with this analysis are very large. In Section 4 we suggest that a free–free contribution at 100 GHz might mean that the dust surface brightnesses are half of the observed intensities. Dust emissivities are also highly uncertain at these wavelengths (Cigan et al. 2019); the slightly different modified blackbody parameters of Auld et al. (2013) produced a difference of a factor of 3 in the inferred α_X values and the H₂ column density, and Clark et al. (2019) also found evidence for variations in the dust emissivity within an individual galaxy. The RADEX/MCMC column densities are also, strictly speaking, not as well constrained as the ratios N/Δv, and we have used line widths Δv estimated from the outer part of the disk, where there is little kinematic beam smearing affecting the line widths. Finally, estimates of gas/dust mass ratios find wide variations of at least a factor of 30, and sometimes 100, between different galaxies (Cortese et al. 2016; Lianou et al. 2019), and it is difficult to know how much of that variation is real and how much is noise. Recent theoretical work suggests that much of it may be real (Whitaker et al. 2021).

In the context of these uncertainties, we simply comment that the inferred α_X values and CO isotopologue abundances found here are roughly consistent with the values assumed in Section 5, and they are similar to those of the spirals of the Local Group (Leroy et al. 2011), and at our 150 pc resolution there is little to no evidence for strong internal variations within the disk of NGC 4526.

9.1. Whole-galaxy Integrated Properties

9.2. Structures within NGC 4526

We present new ALMA observations of continuum emission and several molecular species in the dusty disk of the Virgo Cluster early-type galaxy NGC 4526; the species include CO and its isotopologues, CN, CS, HCN, HCO⁺, HNC, HNCO, and CH₃OH.

The 3 mm continuum emission shows a nuclear point source with a spectral index of −1.07 ± 0.02, which is clearly synchrotron emission from an AGN that is also detected in X-rays and in lower-frequency radio emission. We also find low surface brightness continuum emission from an extended disk; this disk emission has a strongly rising spectral index, and its flux density is a remarkably good match to the long-wavelength extrapolation of a modified blackbody that fits the FIR flux densities. We therefore estimate that 50% to 100% of this extended 3 mm continuum is dust emission, and the remainder is probably free–free emission.

The resolved molecular disk structure shows a nuclear peak and a bright ring at r ∼ 5'' (0.4 kpc); assuming a CO/H₂ conversion factor of 4.3 M_⊙ pc⁻² (K km s⁻¹)⁻¹, the deprojected molecular surface densities in the nucleus and in the ring are 240 M_⊙ pc⁻² at 75 pc resolution. The molecular disk also shows modest asymmetries, kinematic disturbances, and evidence for a bar, and those kinematic features will be discussed in a future paper.

The ¹²CO(1−0)/¹³CO(1−0) line ratio in NGC 4526 is one of the lowest ever observed outside of the Local Group, at 3.5 ± 0.35, and it is remarkably constant throughout the disk. In contrast, ¹³CO(1−0)/C¹⁸O(1−0) shows typical values but an unusually steep gradient; it is nearly constant within the molecular ring (r < 5'') but exterior to that it rises by almost a factor of two over only a 0.6 kpc radius.

We carry out RADEX modeling of the CO isotopologues and employ Bayesian analysis with an MCMC sampler to estimate posterior probability distributions for the physical parameters in the CO-emitting gas. The current set of transitions strongly constrain the [¹³CO/C¹⁸O] abundance ratio to be ${5.7}_{-0.6}^{+0.7}$ in the inner disk (r < 2'') and ${9.4}_{-0.9}^{+1.1}$ at a radius of 1 kpc; those uncertainties are 68% confidence intervals. The [¹²CO/¹³CO] abundance ratio is poorly constrained in the innermost regions of the disk but is found to be ${7.8}_{-1.5}^{+2.7}$ at r ≈ 5'' and ${6.5}_{-1.3}^{+3.0}$ at r ≈ 12'', and these values are a factor of three lower than those typically found in nearby spirals.

We find volume densities ${n}_{{{\rm{H}}}_{2}}$ = 250 cm⁻³ ± 0.4 dex and T_kin > 10 K in the CO-emitting gas. The ¹³CO(1−0) line is moderately optically thick, with τ₁₃ ≈ 1 and possibly as high as 2. Our temperature estimates from the CO lines are consistent with those from the FIR [C i] transitions and from CN (see below). We also infer temperatures in the range of 5.4 K to 13.4 K from HNCO(5−4)/(4−3). These estimates roughly follow the expected pattern that high-density tracers should show lower temperatures than CO or [C i], reflecting the origins of high-density tracers in the darker and colder interiors of molecular clouds. The uniformly low temperatures are also consistent with general expectations based on the galaxy's low star formation rate.

NGC 4526 is relatively bright in the high-density tracers HCN, CN, and especially CS; it is relatively faint in the shock tracers CH₃OH and HNCO. We find no measurable enhancements in high-density ratios like CN/HCN or HCN/HCO⁺ (or indeed anything else) toward the AGN, when they are measured at 150 pc resolution. The one possible exception to that statement is a tentative rise in ¹²CO(2−1)/(1−0) toward the nucleus, but that analysis is complicated by uncertain continuum subtraction in the (2−1) data. It has also been suggested that CS/HCN might be enhanced in regions of star formation activity, but we find no evidence for any spatial variations in that ratio either. Line ratios between any two high-density tracers are remarkably constant within NGC 4526.

NGC 4526 also has an unusually high optical depth in the 3 mm CN(N = 1−0) lines. We find an extremely low (J = 3/2−1/2)/(J = 1/2−1/2) ratio of 1.12 ± 0.16 in the nucleus, rising to values consistent with 2.0 outside the molecular ring (r > 5''). An assumption of LTE requires temperatures ≤ 20 K in order to reproduce the low line ratio in the nucleus, and it suggests the optical depth in the 3/2−1/2 transition is τ_3/2 ≥ 4.5. For typical abundance ratios, this value then suggests that HCN(1−0), HCO⁺(1−0), and HNC(1−0) should also be very thick, with ${\tau }_{\mathrm{HCN}\ 1-0}$ in the range of 5 to 30. Extragalactic observations rarely find such high values, and it is not clear why NGC 4526 would be unusual in this respect.

Motivated by the good match between the modified blackbody fit and the extended 3 mm continuum flux density, we assume a dust emissivity and a gas/dust ratio in order to infer dust-based H₂ surface densities and CO/H₂ conversion factors α_CO. Column density estimates for the CO isotopologues, from the RADEX/MCMC analysis, then allow us to estimate the abundances of the CO species. The values have large uncertainties of at least a factor of a few, but at that level they are consistent with other values inferred for nearby spirals. We find no particular evidence for internal variations in α_CO or the gas/dust ratio, though we cannot estimate those quantities toward the nucleus because of the dominant synchrotron continuum component there.

In many respects, the molecular properties of NGC 4526 are consistent with its being a heavily stripped descendant of something that used to look like a large spiral galaxy. Its unusually low [¹²CO/¹³CO] abundance ratios and typical [¹³CO/C¹⁸O] ratios provoke questions about the processing of its molecular gas in the Virgo Cluster. Its deep potential well (high circular velocity) should protect its molecular gas from stripping more effectively than would be the case in lower-mass spirals, though, so it is not obvious whether the galaxies that are currently being stripped in the Virgo Cluster will end up looking like NGC 4526 in the future. Future work on isotopologues in other cluster galaxies, in combination with chemical enrichment models, will be particularly valuable for understanding the history and the processing of galaxies in clusters.

We thank Martin Bureau for many fascinating and productive conversations about early-type galaxies. This paper makes use of the following ALMA data: ADS/JAO.ALMA#2017.1.01108.S and ADS/JAO.ALMA#2018.1.01599.S. ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada), MOST and ASIAA (Taiwan), and KASI (Republic of Korea), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under a cooperative agreement by Associated Universities, Inc.

Facilities: ALMA, CARMA, IRAM.

The ¹²CO(2–1) data used here were obtained with the CARMA telescope at a resolution of 025 by Davis et al. (2013b). A smoothed version of the data was also used in the dynamical analysis of Davis & McDermid (2017). As the uv coverage of the data is significantly different from that of the other lines studied here, we check whether the CARMA images might have missed significant flux by comparing them to a single-dish spectrum from the IRAM 30 m telescope (Combes et al. 2007). We convolve the CARMA data cube to the 12'' resolution of the 30 m at 230 GHz, ensuring proper flux retention by appropriate normalization of the convolution kernel. The spectra of the individual pixels in the convolved cube then simulate what the 30 m would have observed if it were pointed at that location. The technique can account for both pointing offsets and the fact that the 30 m beam is smaller than the size of the molecular disk. The results are shown in Figure 18 for a location 16 northwest of the galaxy nucleus, a location chosen by eye to reproduce the degree of asymmetry in the 30 m spectrum. (Larger offsets produce stronger asymmetry and smaller offsets produce weaker asymmetry.) An offset of 1.6'' is well within the expected pointing performance of the 30 m telescope. ⁷ The good match achievable between the actual 30 m spectrum and the simulated one suggests the CARMA ¹²CO(2–1) data do not miss significant flux due to missing short spacings. Indeed, in a source like this with such strong rotation, the emission in each individual channel is effectively unresolved in the direction of the local velocity gradient.

Zoom In Zoom Out Reset image size

A second issue for the ¹²CO(2–1) images is that continuum subtraction was not performed on the original data because it was primarily used for kinematic work, and the limited available bandwidth did not provide enough sensitivity in line-free channels to detect any continuum emission. But we can now estimate the strength of the 230 GHz continuum, based on the 100 GHz continuum and FIR flux density measurements. As described in Section 4, the 99 GHz continuum image at 20 resolution (Figure 2) contains roughly 1.9 ± 0.1 mJy in an extended disk with a spectral index of +1.8 and 4.7 ± 0.1 mJy in a nuclear point source with a spectral index of −1. Uncertainties in the way the flux density is partitioned are difficult to estimate because they involve interpolating the disk under the point source, so the quoted uncertainties may well be underestimates. Extrapolating those flux densities using their estimated spectral indices suggests 11 mJy of total continuum emission at 230 GHz, of which roughly 2 to 4 mJy would be the nucleus and the rest would be the disk. Alternatively, the modified blackbody fits, which use a steeper effective spectral index for the disk emission, predict roughly 40 mJy at 230 GHz (Figure 3).

We construct a model 230 GHz continuum image including 30 mJy in an elliptical Gaussian mimicking the size and shape of the dust disk, plus 4 mJy in a nuclear point source. This level of continuum emission is below the detection limits of the CARMA data. We then subtract this model continuum image from the CARMA data cube and compare analyses with and without the estimated continuum. The continuum subtraction makes a 10% difference in the integrated ¹²CO(2–1) flux (221 ± 6 Jy km s⁻¹ compared to 245 ± 6 Jy km s⁻¹). It also makes a 35% difference in the nuclear peak integrated intensity (8.1 ± 0.7 Jy beam⁻¹ km s⁻¹ at 1.12'' resolution versus 11.0 ± 0.7 Jy beam⁻¹ km s⁻¹). Resolved line ratio measurements outside of the nucleus are indistinguishable in the two cases, as the disk has low surface brightness and the line emission outside of the nucleus has relatively narrow velocity support. The values presented in the paper are computed after the estimated continuum has been subtracted.

We investigate the physical conditions in the CO-emitting gas through the RADEX non-LTE code, which carries out iterative solutions of the level populations and the line radiation using an escape probability formalism (van der Tak et al. 2007). We compute the line intensities and optical depths for the J = 1−0 and 2−1 transitions in ¹²CO,¹³CO, and C¹⁸O for a grid of physical conditions, giving the relevant line ratios as functions of those physical conditions. The grid spans volume densities ${n}_{{{\rm{H}}}_{2}}$ = 50 to 10⁶ cm⁻³ in steps of 0.215 dex, kinetic temperatures T_kin = 3.0 to 80 K in steps of 0.071 dex, and column densities N(¹²CO) = 8 × 10¹⁵ to 8 × 10¹⁹ cm⁻² in steps of 0.067 dex. For a fiducial [¹²CO/H₂] abundance of 8 × 10⁻⁵ (e.g., Frerking et al. 1982), the corresponding total column densities are N(H₂) = 1 × 10²⁰ to 1 × 10²⁴ cm⁻². We also assume a fixed line width of 30 km s⁻¹, which is similar to the line width in the outer parts of the galaxy, where there is little beam smearing (Figure 6). In this context we note that the RADEX calculations in fact depend only on the ratio of the column density to the line width, so we show fiducial column densities N(¹²CO) in the figures but in fact the fundamental parameter of physical relevance is N(¹²CO)/Δv. The corresponding column densities can be easily calculated for other line widths with a simple scaling. The grid also spans [¹²CO/¹³CO] abundance ratios of 3.4 to 400 and [¹³CO/C¹⁸O] of 2.0 to 27, both in steps of 0.067 dex. This grid strategy is very similar to that employed by Teng et al. (2021), though independently coded.

We then compare the predicted line ratios to observed quantities and use an MCMC sampler (emcee; Foreman-Mackey et al. 2013) to estimate posterior probability distributions on the relevant physical parameters in the molecular gas. Sampling using the adaptive Metropolis algorithm of Cappellari et al. (2013) produces results that are indistinguishable from those of emcee. In both cases, the sampler interpolates in the RADEX grids to evaluate predicted intensities, line ratios, and the likelihood at each set of physical conditions it explores. We assume flat priors on the logs of the parameters, over the range of conditions spanned by the RADEX grids.

The observations we use as constraints are the ¹²CO(1−0) intensity and the line ratios ¹²CO(1−0)/¹³CO(1−0), ¹²CO(1−0)/C¹⁸O(1−0), ¹³CO(1−0)/C¹⁸O(1−0), ¹²CO(2−1)/¹²CO(1−0), and ¹²CO(2−1)/¹³CO(2−1). As the beam is larger than the typical size of a molecular cloud, beam dilution may make the observed ¹²CO(1−0) brightness temperature lower than the intrinsic radiation temperature for the line. We account for the beam filling factor by using an asymmetric distribution for the T_B contribution to the model likelihood. Specifically, the other measurements' contributions to the model likelihood are symmetric Gaussians, with their measurement uncertainties characterizing the width of the distributions, but for T_B we use a Gaussian below the measured value and a flat distribution above the measured value. All predicted line intensities that are higher than the measured value are thus considered equally likely, allowing for beam filling factors in the range 0 ≤ f ≤ 1. In attempting to reproduce the other line ratios we are effectively assuming this beam filling factor is the same for all the isotopologues.

The ¹²CO(2−1)/¹³CO(2−1) line ratio also deserves some comment; it is not measured using resolved ALMA data like the other line ratios, but with the IRAM 30 m telescope (Crocker et al. 2012) so it is an average over a 12'' beam. The fact that the resolved measurements of ¹²CO(1−0)/¹³CO(1−0) and ¹²CO(2−1)/¹²CO(1−0) show no significant radial variation lends confidence to the assumption that ¹²CO(2−1)/¹³CO(2−1) may also be roughly constant over the disk of NGC 4526.

We carry out this analysis for three representative sets of conditions in NGC 4526. One set of values corresponds to the nuclear regions (r ≲ 2''), where the C¹⁸O intensity is relatively high, so the ¹³CO/C¹⁸O intensity ratio is low, and the ¹²CO(1−0) brightness temperature is lower than elsewhere in the galaxy (Section 5). Another set corresponds to values measured in the molecular ring at r ≈ 5'', where the ¹³CO/C¹⁸O intensity ratio is still low as in the inner disk but the ¹²CO(1−0) brightness temperature is at its highest. The third set of values corresponds to the outer disk, where the ¹³CO/C¹⁸O intensity ratio is relatively high and ¹²CO(1−0) brightness temperatures are intermediate between those of the nucleus and the ring.

Figures 19, 20, and 21 show corner plots for the posterior probability distributions for these conditions. Figure 19 is for the nucleus; it illustrates the classic degeneracy between density and kinetic temperature for driving excitation, due to the limited set of currently available line ratios and the fact that the ¹²CO brightness temperature does not provide strong constraints in the nucleus. The inferred posterior probability distribution for the [¹²CO/¹³CO] abundance ratio peaks at 6.3 in the nucleus, though there is a long asymmetric tail such that higher values are also permitted. The inferred [¹³CO/C¹⁸O] abundance ratio is, however, tightly constrained and almost independent of the other parameters; it is ${5.7}_{-0.6}^{+0.7}$ (the quoted uncertainties correspond to a 68% confidence interval). That value is 15% higher than the observed line ratio, due to the optical depth of the ¹³CO(1−0) line.

Zoom In Zoom Out Reset image size

The conditions in the ring (Figure 20) include higher brightness temperatures, which give tighter constraints on the excitation. Here the line ratios suggest relatively low densities for the molecular gas, with ${n}_{{{\rm{H}}}_{2}}$ ∼ 250 cm⁻³ ± 0.4 dex and T_kin > 10 K. The improved constraints on ${n}_{{{\rm{H}}}_{2}}$ and T_kin also help to constrain the [¹²CO/¹³CO] abundance ratio, which we find to be ${7.8}_{-1.5}^{+2.7}$. We also find a [¹³CO/C¹⁸O] abundance of ${6.2}_{-0.6}^{+0.8}$, consistent with the value in the nucleus. These conditions require ¹³CO(1−0) to be moderately optically thick, with ${\tau }_{13}={1.1}_{-0.5}^{+0.9}$. As a check on the plausibility of those conditions, we note that the density ${n}_{{{\rm{H}}}_{2}}$, the column density N(¹²CO), and a typical CO abundance of 10⁻⁴ give a line-of-sight thickness of 12 pc. That value is reasonable for a giant molecular cloud; additional discussion can be found in Teng et al. (2021). Finally, in the outer disk (Figure 21) we infer ${n}_{{{\rm{H}}}_{2}}$, T_kin, and [¹²CO/¹³CO] values similar to those in the ring, with [¹²CO/¹³CO] = ${6.5}_{-1.3}^{+3.0},$ but substantially higher [¹³CO/C¹⁸O] abundances of ${9.4}_{-0.9}^{+1.1}.$

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In short, we find strong evidence for a radial gradient in the [¹³CO/C¹⁸O] abundance ratio, increasing to larger values at larger radii; the change is a factor of 1.6 over 1 kpc. The gradient is shallower in the inner disk and steeper in the outer disk, with the majority or perhaps all of the gradient occurring exterior to the ring. In contrast, we find no compelling evidence for a radial gradient in the [¹²CO/¹³CO] ratio, as our two inferred values of ${7.8}_{-1.5}^{+2.7}$ and ${6.5}_{-1.3}^{+3.0}$ are consistent with each other given their uncertainties. However, those [¹²CO/¹³CO] abundance ratios are unusually low. If they reflect an intrinsic [¹²C/¹³C] ratio ≈ 7, then this isotopologue ratio in NGC 4526 is at least a factor of three lower than those in the centers of most nearby spirals (e.g., Martín et al. 2019). Additional discussion of the implications of these isotopologue measurements is made in Section 9.

Figure 22 also presents (${n}_{{{\rm{H}}}_{2}}$ and T_kin) slices of the five-dimensional RADEX model volume, with contours indicating the measurements we are attempting to explain. We construct each panel by identifying the most likely values for the ¹²CO column density and [¹²CO/¹³CO] and [¹³CO/C¹⁸O] abundance ratios for each set of modeled conditions. (Specifically, we use the median values of the marginalized posterior probability distributions, as indicated in the histogram titles in Figures 19 and 20.) The slice locations are fixed at the grid values closest to the medians in those three parameters. As usual (e.g., Teng et al. 2021) the maxima in the individual one-dimensional marginalized distributions do not necessarily correspond to the maximum likelihood solution in the whole model volume, but they are close enough to be used as reasonable plausibility checks to illustrate the properties of our solutions. For improved constraints on the temperature of the gas, and thus on its density, we suspect that it would be most helpful to have high-resolution measurements of the ¹³CO(2−1) and C¹⁸O(2−1) lines.

Zoom In Zoom Out Reset image size