WATER MASERS IN THE ANDROMEDA GALAXY. I. A SURVEY FOR WATER MASERS, AMMONIA, AND HYDROGEN RECOMBINATION LINES

The dominant galaxies in the Local Group, the Milky Way and the Andromeda galaxy (M31), are likely to merge in ${5.86}_{-0.72}^{+1.61}$ Gyr (van der Marel et al. 2012b). The key quantity for this prediction is the transverse velocity of M31 with respect to the Galaxy. Sohn et al. (2012) and van der Marel et al. (2012a) used Hubble Space Telescope observations of thousands of M31 stars referenced to hundreds of compact background galaxies in three fields (near the major and minor axes and in the Giant Southern Stream) over five to seven years to constrain the tangential velocity of M31 to ≤34.3 km s⁻¹ (1σ). The consistency of the M31-Milky Way velocity with a radial orbit, infalling at −109.3 ± 4.4 km s⁻¹ (van der Marel et al. 2012a), provides a new estimate of the Local Group timing mass of (4.93 ± 1.63) × 10¹² M_⊙ (van der Marel et al. 2012a); although van der Marel et al. (2012a) derive a best estimate of (3.17 ± 0.57) × 10¹² M_⊙, citing cosmic variance as an uncertainty floor on the timing argument estimation of the Local Group mass. The transverse velocity of M31 is thus essential to our understanding of the Local Group's future (and past) dynamical evolution, as well as the density profiles and distribution of dark matter halos (Peebles et al. 2001; Loeb et al. 2005; Reid et al. 2009).

The Sohn et al. (2012) and van der Marel et al. (2012a, 2012b) results can be independently verified and possibly refined using maser proper motions in M31. Molecular masers, while small in number, are compact high brightness temperature light sources arising in the molecular ring of M31 that provide ∼10 microarcsecond (μas) astrometry using very long baseline interferometry (VLBI). Brunthaler et al. (2005, 2007) have measured the proper motions of the Galactic analog 22 GHz water (H₂O) masers discovered in M31's satellite galaxies M33 (first by Churchwell et al. 1977) and IC 10 (Henkel et al. 1986), obtaining their three-dimensional velocities with respect to the Milky Way. Brunthaler et al. (2005) also measured the proper rotation of M33, which provides a geometric determination of its distance via a "rotational parallax." The same can be done with masers in M31 (Darling 2011).

Detecting Galactic analog masers of any type in M31 is challenging due to its distance and large angular size. The M31 distance reduces Galactic maser flux densities by a factor of ∼10⁴, so only the brightest Galactic masers could be detected in M31, provided that surveys are sensitive to ∼10 mJy spectral lines (Darling 2011). The large angular size of M31—the molecular ring is roughly 2° × 05—implies that surveys for masers, particularly H₂O, need to be pointed in order to reach adequate sensitivity. New radio telescope facilities with larger collecting areas or larger fields of view than are currently available may improve this situation.

Sullivan (1973) may have made the first search for H₂O masers in M31, making no detections toward the nucleus and "emission nebula #132." Greenhill et al. (1995) and Imai et al. (2001) report pointed H₂O maser surveys toward H ii regions in M31, reaching 1σ rms noise levels of 29 and 70 mJy, respectively. Claussen & Beasley (2011, private communication) surveyed the nuclear region and most of the molecular ring using the Very Large Array, making no H₂O maser detections and reaching an rms noise of 30 mJy per beam. Sjouwerman et al. (2010) detected the first maser of any type in M31, a Class II methanol (CH₃OH) maser at 6.7 GHz, but no associated water maser was detected (see also Darling 2011, and this work). Finally, Darling (2011) conducted a survey of 206 compact 24 μm sources in the molecular ring with ∼3 mJy rms noise per 3.3 km s⁻¹ channel and detected five water masers. The water maser non-detections prior to 2011 were all consistent with the aggregate star formation rate of M31, and these surveys were not sufficiently sensitive to detect a typical Galactic maser at the distance of M31 (Brunthaler et al. 2006). The historical lack of Galactic analog water maser detections in M31 has simply been a problem of sensitivity and of locating likely sites of maser activity, both of which now have remedies.

In this paper, we describe an extension of the Darling (2011) Green Bank Telescope³ (GBT) H₂O maser survey to include an additional 300 unresolved 24 μm sources in M31. We also report on the simultaneous observation of two ammonia (NH₃) lines and one hydrogen radio recombination line (H66α) toward all 506 locations in the combined surveys. In a companion paper, we examine the physical conditions most likely to produce detectable H₂O maser emission in M31 (Amiri & Darling 2016, Paper II).

Throughout this manuscript, we use heliocentric velocities (optical definition) and assume the following about M31: a systemic velocity of −300 km s⁻¹ (de Vaucouleurs et al. 1991), a distance of 780 kpc (McConnachie et al. 2005), and central coordinates 00:42:44.3, +41:16:09 (J2000).

A Spitzer 24 μm map of M31 (Gordon et al. 2008) guided the GBT observations in two important ways: source selection and completeness. The association of water masers with (ultra)compact H ii regions is well known, but the molecular gas as traced by optically thick CO emission does not necessarily indicate these regions. Likewise, Hα-selected H ii regions may select against dust-enshrouded ultra-compact H ii regions (but see Amiri & Darling 2016). Compact 24 μm emission, however, is a good indicator of likely sites of H₂O maser emission, and there is a rough relationship between far-IR emission and water maser luminosity (Jaffe et al. 1981; Urquhart et al. 2011; although the latter relies on bolometric luminosities). We therefore selected compact 24 μm sources associated with the dusty molecular—and presumably star-forming—regions in M31 ("compact" means unresolved at the 23.5 pc resolution of the Spitzer map). However, this method can also select objects that are unlikely to produce water masers, such as planetary nebulae, giant stars, and background galaxies.

We constructed a catalog of 506 unresolved 24 μm sources working from the brightest down to the point where most of the 24 μm emission becomes extended, about 4 MJy steradian⁻¹ at peak specific intensity. Although this pointed survey does not include all star formation in M31, it does include the majority of likely locations for strong H₂O maser activity. By working down in luminosity through the catalog of compact 24 μm sources, we have surveyed a large fraction of the total ongoing star formation in the galaxy (e.g., Tabatabaei & Berkhuijsen 2010; Ford et al. 2013) and thus most of the likely H₂O-bearing regions (Amiri & Darling 2016). Figure 1 shows the 24 μm map of M31 with the 506 pointing centers and primary beam size for our GBT water maser survey, including the masers detected by Darling (2011).

Zoom In Zoom Out Reset image size

Darling (2011) observed 206 24 μm sources in M31 using the GBT in 2010 October through December. The ${6}_{16}\mbox{--}{5}_{23}$ 22.23508 GHz ortho-water maser line observations were reported in Darling (2011), but simultaneous observations of the para-ammonia (NH₃) rotational ground-state inversion transitions in the metastable states (J, K) = (1, 1) and (2, 2) at 23.6945 and 23.72263 GHz, respectively, and the hydrogen recombination line H66α at 22.36417 GHz were not. We subsequently observed all four of these lines toward an additional 300 24 μm sources in 2011 October through 2012 January. The telescope configuration was identical to that described by Darling (2011): a 632–674 km s⁻¹ (50 MHz) bandpass was centered, for each of the four observed transitions, on a heliocentric velocity of −300 km s⁻¹ for 239 sources in the central parts of the galaxy and along the minor axis, on −100 km s⁻¹ for 136 sources in the redshifted northeast wedge of the galaxy, and on −500 km s⁻¹ for 131 sources in the blueshifted southwest wedge of the galaxy (Figure 1 and Table 1).

Observations were conducted with the dual K-band receivers in 2010 and beams 1 and 2 of the K-band focal plane array in 2011–2012 in a nodding mode in two circular polarizations with 0.15–0.16 km s⁻¹ (12.2 kHz) channels and 9-level sampling. The time on-source was 5 minutes, except for sources that were reobserved to confirm or refute possible lines (typically 10 minutes). A winking calibration diode and hourly atmospheric opacity estimates were used for flux density calibration. Opacities ranged from 0.03 to 0.14 nepers, but were typically 0.06 nepers. The estimated uncertainty in the flux density calibration is ∼20%. Pointing and focus corrections were made hourly. Pointing was typically good to within a few arcseconds, and the largest pointing corrections during observing sessions were no more than 6''. The resolution of the 24 μm Spitzer image is 6'' (Gordon et al. 2008), so the unresolved IR sources remained within the 33'' GBT beam even during the largest pointing drifts. The 33'' beam (FWHM) at 22 GHz spans 125 pc in M31.

After averaging polarizations, spectra were Hanning smoothed and subsequently Gaussian smoothed to obtain a final spectral resolution of 3.1–3.3 km s⁻¹ (244 kHz) for the line search. H66α was also smoothed to 9.8 km s⁻¹ (732 kHz) to search for broad recombination lines (e.g., Dieter 1967; Mezger & Höglund 1967). Polynomial baselines, typically of fifth order, were fit and subtracted to obtain flat and generally uniform-noise spectra. Spectral rms noise measurements for each pointing center and observed line are listed in Table 1. Non-detection spectra generally did not show any features greater than 3σ, corresponding to roughly 10 mJy, which is at least an order of magnitude more sensitive than previous surveys (Greenhill et al. 1995; Imai et al. 2001; Claussen & Beasley, 2011, private communication), and is likely the weakest line that can currently be used for VLBI proper motion studies. All data reduction was performed in GBTIDL.⁴

In order to assess the average and median line emission of the sample, we aligned H₂O, NH₃, and H66α spectra according to the Nieten et al. (2006) CO velocities, and median- and mean-stacked them to determine whether the median or mean object selected for the survey produces line emission in a given species an order of magnitude below the single-spectrum sensitivity. Spectra were noise-weighted in the mean stacks, and 300 spectra were used for both the mean and median stacks (except for the H₂O stacks, which used 299 spectra). Omitted spectra either had no CO velocity information, had unreliable CO velocity information, or have been identified to be something other than an H ii region (indicated with footnotes in Table 1; see also Section 4). Every object in Table 1 was manually inspected in the Nieten et al. (2006) CO velocity map to ensure that no mask edges, incomplete data, or high (instrumental) velocity dispersion regions were included in the stack. All objects in Table 1 that show a CO velocity can be considered reliable.

We have detected water maser emission in 5 out of 506 24 μm selected regions in M31 (Table 1). The water maser isotropic line luminosity sensitivity is 4.4 × 10⁻⁴ L_⊙, which is a 3σ peak line flux density limit of 9 mJy and width 3.3 km s⁻¹. The detected water masers were presented in Darling (2011) and are not reproduced here. Omitting the 44 objects likely to be planetary nebulae (9 objects; Merrett et al. 2006; Amiri & Darling 2016) or giant stars (35 objects; Amiri & Darling 2016), the overall detection rate is 1.1(0.5)%. The actual water maser fraction likely depends on star formation stage and conditions, and our overall detection rate is a strong function of the survey sensitivity; Amiri & Darling (2016) demonstrate that we detect only the high-luminosity tail of the underlying maser population. Conversely, it is unlikely that all the remaining 24 μm sources are compact H ii regions (e.g., Verley et al. 2007; Mould et al. 2008), so the maser fraction among star-forming regions may be higher than 1.1(0.5)% given the survey sensitivity. The maser detection rate would also be higher were we to use a more refined selection method, such as the one presented in Amiri & Darling (2016).

No NH₃ or H66α lines were detected in the survey; see Table 1 for rms noise values. Figure 2 shows the noise-weighted mean spectral stacks for the four observed spectral lines: H₂O, H66α, NH₃ (1, 1), and NH₃ (2, 2). The stacks of 300 spectra (299 spectra for H₂O) reached rms noise values of 0.17, 0.14, 0.15, and 0.15 mJy per 3.1–3.3 km s⁻¹ channel, respectively. The median stacks appear similar to the mean stacks, but with 12%, 21%, 20%, and 13% higher noise. No lines are detected in these stacks, which show the expected $\sqrt{N}$ noise improvement over single-object spectra.

Zoom In Zoom Out Reset image size

Neither the mean nor the median compact 24 μm emitting region selected for the water maser survey show detectable emission in the four observed lines. In all of the stacked spectra, errors in the CO velocity used to center the spectra, CO velocity variance, and CO line peak offsets from the denser molecular gas velocity can all contribute to incorrect spectral shifts in the stacking process. These factors will dilute the detectability of lines in spectral stacks, particularly in the mean.

The water maser fraction in Galactic surveys can be of order 50% (e.g., Kurtz & Hofner 2005; Urquhart et al. 2011), so one might expect to detect line emission from the median object in an adequately deep survey. While the line luminosity sensitivity of the M31 survey does not reach the sensitivity needed in Galactic surveys to detect the median water maser in individual spectra, the median spectral stack reaches a 3σ luminosity sensitivity of 2.8 × 10⁻⁵ L_⊙ (assuming a 3.3 km s⁻¹ wide maser). Because the median Urquhart et al. (2011) H₂O maser isotropic line luminosity is 1 × 10⁻⁵ L_⊙ and the CO velocity alignment of the stacks lacks the velocity precision needed for alignment of few km s⁻¹ lines, the M31 spectral stack cannot quite detect the median Galactic H ii water maser equivalent.

For the H66α line, the non-detection in stacked spectra probably implies that the mean or median region simply does not emit in this line. Aside from selecting for 22 GHz radio continuum backlights, the 24 μm selection method is sub-optimal for hydrogen radio recombination lines (the H66α line was included in the GBT observations to allow serendipitous detection).

NH₃ emission, by contrast, is common in H ii regions; Urquhart et al. (2011) obtain a detection rate of ∼80%, fairly independent of H ii region luminosity, and Dunham et al. (2011) detected NH₃ (1, 1) toward 72% of 1.1 mm continuum sources in the inner Galaxy. One would expect to detect NH₃ lines in mean or median spectra of adequate depth, provided the NH₃ filling factor does not become too small at the distance of M31. We discuss these issues below.

5.1. Water Maser Detection Rate and Star Formation

5.2. Ammonia

We have expanded a survey for H₂O masers in M31 based on pointed observations of 24 μm selected regions. While the 24 μm selection method clearly succeeds in identifying water masers (albeit at the low rate of 1.1(0.5)%), 24 μm emission is a necessary but not a sufficient condition for water maser activity (see Amiri & Darling 2016 for a study of the properties of star-forming regions most likely to produce water masers). In Darling (2011), we suggested that the catalog of water masers in M31 is incomplete and that an exhaustive survey of IR-luminous regions would identify additional masers. The prediction was too optimistic, and it should be noted that the detected masers are consistent with the aggregate star formation rate of M31, given its distance and the sensitivity of the survey. New masers could still be detected in deeper surveys or in future surveys of similar depth, given the variable nature of water masers in star-forming regions. However, variability also implies that the currently detected masers could be flaring and therefore fade, limiting their astrometric utility.

For now, the five known water masers in M31 will have to serve as the basis for proper motion studies, which will probably minimally constrain the three expected proper motion signals (plus random peculiar maser motions) described by Darling (2011): (1) the systemic proper motion (already constrained by Sohn et al. 2012; van der Marel et al. 2012a), (2) proper rotation, and (3) proper expansion caused by the −300 km s⁻¹ (heliocentric) approach of M31 (see also Darling 2013). We predict that proper rotation and systemic proper motion will be measurable in a few years with current facilities, and that the "moving cluster" divergence of maser spots as M31 approaches the Galaxy may be detectable in a decade (Darling 2013). However a possible previous interaction between M31 and M32 may have produced peculiar radial motions in the molecular ring of M31 that would mask this signal (Block et al. 2006; Dierickx et al. 2014).

All authors acknowledge support from the NSF grant AST-1109078. The authors thank K. Gordon for the Spitzer map, M. Claussen and T. Beasley for sharing their results, and the anonymous referee for helpful comments. This research has made use of NASA's Astrophysics Data System Bibliographic Services and the NASA/IPAC Extragalactic Database (NED), and uses observations made with the Spitzer Space Telescope, both of which are operated by the Jet Propulsion Laboratory, California Institute of Technology, under a contract with NASA.

Facility: GBT - Green Bank Telescope.