THE NEUTRAL HYDROGEN BRIDGE BETWEEN M31 AND M33

The larger spiral galaxies of the Local Group, the Milky Way, M31, and M33, are growing through mergers with smaller systems. M31 and M33 contain numerous stellar streams (Ibata et al. 2001; Ferguson et al. 2002; McConnachie et al. 2009), and the Milky Way is currently accreting the Sagittarius dwarf galaxy, the Smith cloud, and probably the Magellanic Stream as well (Mathewson et al. 1974; Ibata et al. 1994; Putman et al. 2003; Lockman et al. 2008; McClure-Griffiths et al. 2008; Stanimirović et al. 2008; Nidever et al. 2010). M31, M33, and the Milky Way are also surrounded by clouds of neutral and ionized gas—the "high-velocity clouds" (HVCs)—that are not identifiable with any stellar system and may be dark matter subhalos, material stripped from satellites, accretion from a hot halo or the intergalactic medium, or something else entirely (Wakker & van Woerden 1997; Thilker et al. 2004; Westmeier et al. 2008; Grossi et al. 2008; Shull et al. 2009; Nichols & Bland-Hawthorn 2009).

Braun & Thilker (2004, hereafter B&T) mapped the H i over a large area around M31 and M33, and reported the detection of extremely faint 21 cm H i emission at the level log(N_H i) ≈ 17.0 cm⁻² that formed a partial bridge about 200 kpc in extent between the two galaxies. This gas lies well outside each galaxy's HVC system and has been interpreted as the neutral component of an intergalactic filament, or the remnant of a past encounter between M31 and M33 (B&T; Bekki 2008). Discussions of the likelihood and consequences of an interaction between M33 and M31 are given in recent papers by Putman et al. (2009), Davidge & Puzia (2011), and Peebles et al. (2011). For convenience we will refer to the structure reported by B&T as the M31 "bridge."

B&T made their measurements using the Westerbork Synthesis Radio Telescope configured as a group of single dishes to obtain very high sensitivity to low surface-brightness H i emission, but at the expense of angular resolution. They further smoothed the data in angle and velocity to detect this extremely weak emission more easily: their final maps had 49' angular resolution, equivalent to 11 kpc at the distance of M31, and a velocity resolution of 17 km s⁻¹. Even so, much of the signal they reported was just 2σ–3σ above the noise. Although they obtained a spectrum with the Green Bank Telescope (GBT) that was consistent with their Westerbork "single dish" measurements at one location, over much of the field the detections are only marginally significant.

Subsequently, Putman et al. (2009) questioned the existence of this H i bridge, noting its absence from the immediate vicinity of M33 at the level log(N_H i) ≳ 18.0 in data from the Arecibo radio telescope, and suggested that the B&T result might be blended Galactic H i mistakenly attributed to M31 or M33.

H i is typically the most massive component of tidal tails (e.g., Duc & Renaud 2011), so the existence of a neutral hydrogen bridge between M31 and M33 may be key in understanding the evolution of both galaxies. To determine the reality of the bridge and to study its properties at significantly higher angular resolution than B&T, we have observed a large area between M33 and M31 using the GBT at a factor ∼5 better resolution in angle and velocity. Importantly, the very clean optics of the GBT (Boothroyd et al. 2011) greatly reduce the chances that any faint 21 cm detection is a spurious signal that originates from the bright H i disks of M31 or M33 and has entered the receiver through a sidelobe.

Spectra in the map presented here have a 5σ sensitivity limit of log(N_H i) ≈ 18.0, so the gas in the B&T bridge would not be detected if it is smoothly distributed on scales ≈1° (14 kpc), but our hope was that the bridge would have much brighter small-scale structure, unresolved by the 49' observations of B&T, but well resolved by the 91 beam of the GBT. In addition to the map, three directions were selected for very deep observations capable of detecting H i emission at the level log(N_H i) ≈ 17.0. If the H i structure between M31 and M33 is actually diffuse, i.e., not highly spatially structured, and at the brightness implied by the B&T measurements, the emission would not be seen in the map, only in the deep pointings.

The goals of this project are thus: (1) to search a wide area between M31 and M33 for evidence of the H i bridge arising from clumps in the gas, and (2) to measure the H i at several locations at a sensitivity similar to that of B&T but with much greater angular resolution as a check on the basic existence of the bridge. Additional measurements were made toward two of M31's dwarf galaxies that lie within the mapped region. The observations are discussed in Section 2, results toward the dwarfs And II and And XV are presented in Section 3, H i emission that is widespread across the field but probably not related to the M31 bridge is discussed in Section 4, results that pertain to the H i bridge are presented in Section 5, the present measurements are discussed in relation to the HVC systems of M31 and M33 in Section 6, and the concluding discussion is in Section 7.

An area of approximately 50 deg² between M31 and M33 was observed with the 100 m diameter Robert C. Byrd Green Bank Telescope (Prestage et al. 2009), at an angular resolution of 91 using the L-band receiver, which has a typical system temperature at a zenith of 18 K in both linear polarizations. All observations were made over a bandwidth of 12.5 MHz at a channel spacing of 0.16 km s⁻¹. During the data reduction spectra were smoothed to coarser velocity resolution using a boxcar function and were resampled. Spectra were edited, then calibrated and corrected for stray radiation as described by Boothroyd et al. (2011), though at the velocities of interest here stray radiation is negligible. All intensities quoted here are brightness temperatures (T_b) averaged over the 91 main beam, corrected for atmospheric absorption. Three data sets were acquired: maps, follow-up observations, and deep pointings. The 5σ detection limits for a 25 km s⁻¹ (FWHM) line are given in Table 1 for both H i column density and H i mass within the GBT beam.

The GBT is an extremely sensitive instrument for 21 cm H i spectroscopy. Early suggestions of ≈10% "gain fluctuations" in 21 cm spectra (Robishaw & Heiles 2009) have not been confirmed in a series of rigorous checks (Boothroyd et al. 2011), while hundreds of thousands of H i spectra have been measured by different groups with accuracies limited only by noise, or quantifiable effects such as instrumental baseline variations (e.g., Lockman & Condon 2005; Hogg et al. 2007; Nidever et al. 2010; Chynoweth et al. 2011). Determinations of foreground Galactic N_H i toward active galactic nuclei (AGNs) using the GBT in the 21 cm line are in excellent agreement with values derived from Lyα absorption lines in the same direction (Wakker et al. 2011)

The observations were made by Doppler-correcting to a constant V_LSR at the center of each spectrum, and the final data cube was gridded in constant V_LSR channels. Conversion of velocities to heliocentric and Local Group Standard of Rest (LGSR) were made using the apex velocity and coordinates given by Karachentsev & Makarov (1996). For our data the uncertainties in the resulting V_LGSR are dominated by systematic uncertainties in the direction and velocity of the apex.

2.1. Maps of the Area between M31 and M33

The GBT was used to map 21 cm H i emission over a large area between M33 and M31 guided by the B&T results. For convenience, the survey region was divided into blocks of 2° × 2°. Spacing between samples on the sky was 35, somewhat finer than the Nyquist sampling interval of 36 for the GBT's 91 beam. Each position in a block was observed for 6 s, and each block was observed as many as six times. Data were acquired during more than two dozen independent observing sessions over a period of 20 months.

In-band frequency switching gave a useable velocity coverage between −600 and +470 km s⁻¹ (LSR). The spectra were smoothed from their intrinsic velocity resolution of 0.16 km s⁻¹ to a resolution of 2.90 km s⁻¹, and a third-order polynomial was fit to emission-free portions of each spectrum. The data were assembled into a cube as described by Mangum et al. (2007) and Boothroyd et al. (2011), and a third-order polynomial was removed from each pixel. The spectra suffered from occasional narrowband interference generated within the GBT receiver room. These signals were stable in frequency and appeared in only one spectral channel, so spectra were interpolated over the affected channels.

Occasionally, during some of the frequency-switched scans, one of the linearly polarized receiver channels had a poor spectral baseline likely caused by out-of-band interference. While it was possible in the "follow-up" and "deep" pointings to examine the data minute by minute for problems, the data over the mapped region consists of many tens of thousands of short measurements, and scrutiny of individual raw spectra was impractical. For this reason we retained only data from the good receiver channel for use in the maps. This problem did not appear in data taken by position-switching.

A channel map at a V_LSR that is free from 21 cm emission (except for M33) is shown in Figure 1. The rms noise over most of the map in a 2.9 km s⁻¹ channel is ≈20 mK except for the blocks at J2000 01^h33^m, +36° and 01^h10^m, +31°, where because of reduced integration time the noise is 44 mK. The area of this latter block, in the lower right of the figure, covers part of the Wright High-Velocity Cloud (Wright 1979) which confuses emission at negative velocities. It was not included in this analysis. For most of the map the 5σ detection limit for a 25 km s⁻¹ line (FWHM) is 1.5 × 10¹⁸ cm⁻². The characteristic angular size of the noise speckles in the map gives an indication of the angular resolution, 91.

Zoom In Zoom Out Reset image size

These mapping observations are 1.5–2.5 times more sensitive than the M31 HVC survey of Westmeier et al. (2008) which was made with a similar angular resolution over an area that overlaps the northwest portion of our map. The Arecibo GALFA-H i observations of the M33 region (Putman et al. 2009) have a much higher angular resolution than those presented here (34 against 91), but the GBT map spectra have a factor of ≈6 lower noise, giving our observations essentially equal sensitivity for structures with an angular size ≲ 3', and a factor of six more sensitivity to emission on scales ≳ 10'. The measurements of the galaxy M33 that resulted from our survey will be described elsewhere.

2.2. Follow-up Pointed Observations

2.3. Deep Pointed Observations

Two dwarf spheroidal galaxies associated with M31 lie within the boundaries of the survey, And II and And XV. As it is of interest to see if there is any connection between M31's dwarf galaxies and the H i bridge, we made pointed observations with the GBT toward both systems, and additionally, mapped a region around And II. No H i was detected that could be associated unambiguously with either galaxy. The limits are given in Table 2.

3.1. And XV

3.2. And II

And II, at a distance of 0.65 Mpc and V_HEL = −188 km s⁻¹ (V_LSR = −187 km s⁻¹) (McConnachie et al. 2005; Côté et al. 1999), is much closer to Galactic velocities than And XV. With an angular size ≈4', the And II stars, like those of And XV, are well covered by a GBT beam. Spectra taken directly toward this dwarf show weak H i emission (≈30 mK) at the velocity of the galaxy. However, similar emission is seen at surrounding positions up to several degrees away. Figure 2 shows the measured H i spectrum directly toward And II and the average of several H i spectra at positions offset in a ring 10'–15' from that galaxy. The spectra have similar intensities at the velocity of And II, regardless of whether the dwarf is in the beam or not. The difference between the H i spectrum toward And II and that toward the reference positions is shown in Figure 3. Any H i coincident in position and velocity with the stellar component of this dwarf must have a 5σ limit on N_H i ⩽ 7.7 × 10¹⁷ cm⁻² for a 25 km s⁻¹ line, with an associated 5σ H i mass limit M_H i ⩽ 1.4 × 10⁴M_☉.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Over the western portion of the mapped region there is H i at −150 ⩽ V_LSR ⩽ −100 km s⁻¹ in addition to the gas near And II that has −200 ⩽ V_LSR ⩽ −150 km s⁻¹ (both components are shown in the region of And II in Figure 2). The distribution of the −120 km s⁻¹ component over the mapped region is shown in Figure 4. Much lower-resolution, incompletely sampled data from the LAB survey (Kalberla et al. 2005) show that this material lies in a long stream approximately along the great circle that connects M31 and M33. Its velocity with respect to the LGSR is V_LGSR > +100 km s⁻¹, quite discrepant from that of M31, M33, and the H i bridge, which over the region of this emission has V_LGSR ≲ 0 km s⁻¹ (Section 5). Although in the existing data there are suggestions of a spatial correlation of this gas with the galaxies, its discrepant velocity indicates that it must have an origin quite different from that of the bridge.

Zoom In Zoom Out Reset image size

Figure 5 shows the survey area with various symbols indicating the location of emission peaks detected in the map (circles), detections in the deep pointings (triangles), the deep pointing without a detected line (inverted triangle) and the two dwarf galaxies (rectangles). Properties of the lines are listed in Table 3 along with limits from the third deep pointing and fiducial information on M31 and M33. Line parameters were derived from a Gaussian fit, either to the follow-up spectra for lines detected in the map (indicated by the word "Map" in Column 8), or to the deep spectra. The lines detected in the deep pointings are shown in Figures 6 and 7.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Velocities of the lines in the LGSR frame are plotted against angular distance from M31 in Figure 8. The errors are dominated by the uncertainties in the conversion of V_HEL to V_LGSR. Random errors in V_LGSR are typically the same order as those in V_LSR, a few km s⁻¹. With one exception, the detections lie between the velocities and positions of M31 and M33, and are thus likely related to these systems. The exception is the line at 01^h03^m219, + 40°33' which has a V_LSR and a V_LGSR more than 100 km s⁻¹ below that of the other emission. This gas is most likely related to an extension of the Magellanic Stream, which has a similar velocity in this part of the sky (Nidever et al. 2010; Stanimirović et al. 2008; B&T), and not to M31 or M33. It is labeled as such in Table 3 and will not be considered further here.

Zoom In Zoom Out Reset image size

The brightest H i lines detected in this survey have a column density about an order of magnitude larger than B&T found at the same positions, and appear to be unresolved, or only slightly extended to the GBT beam. Near 01^h20^m we have two detections, one from the map and one from a deep pointing taken fortuitously only 8' away. If the map observation is centered on an unresolved H i cloud the line brightness temperature at the offset position should be lower by a factor of 0.10. The observed ratio of line intensities, T_L, is 0.071 ± 0.013. Thus the observations are consistent, at the 3σ level, with the emission at 01^h20^m485, + 37°15' arising in an H i cloud that is <91 in angular size. At a distance of 0.8 Mpc, the GBT beam has a linear size of two kpc, implying H i masses within the GBT beam of 9.6 × 10⁴ M_☉ at 01^h20^m485, + 37°15' and 2.2 × 10⁵ M_☉ at 01^h08^m325, + 37°46'.

Recent studies of M31 and M33 have refined our knowledge of their populations of HVCs, and it is interesting to compare the current detections with those objects. The HVC system of M31 was discovered by Thilker et al. (2004) and investigated in depth by Westmeier et al. (2008). The HVCs are confined within a projected distance of 50 kpc despite sensitive searches of more distant areas; two-thirds of M31's HVCs are within a projected distance of 30 kpc. The situation is less clear for M33. A list of possible HVCs has been assembled by Grossi et al. (2008) with several additions from Putman et al. (2009) who note, however, that many from Grossi et al. (2008) seem connected to that galaxy's gaseous halo.

The combined HVC populations of M31 and M33 are shown in Figure 9 in angle from M31 versus V_LSGR, together with the bridge emission features detected here (circles). For clarity, HVCs on the opposite side of M31 from M33 are given a negative angular separation. Whereas the HVCs around each galaxy show a wide spread in velocity (those from M33 seem to trace a rotation curve) the bridge clouds have a velocity near the systemic velocity of each system. The velocity dispersion of the M31 HVCs about their mean V_LGSR is 130 km s⁻¹, and for M33 is either 76 or 50 km s⁻¹, depending on whether the entire sample or only those clouds from Putman et al. (2009) are used. For the bridge clouds the dispersion is only 13 km s⁻¹. It is thus probable that the bridge clouds arise from a very different source than the HVCs. Their kinematics are consistent with B&T's suggestion that they form a partial link between the two galaxies.

Zoom In Zoom Out Reset image size

Our observations have confirmed the existence of faint H i concentrations between the galaxies M31 and M33. They are found at projected distances of 50–115 kpc from M31, at least half the distance to M33, have kinematics consistent with the systemic velocity of M31 and M33, and appear to be distinct from the HVC populations of the individual galaxies. These results support the basic discovery of Braun & Thilker (2004). In two locations we detect H i an order of magnitude brighter than found by B&T, suggesting that the gas is highly clumped. At one position the emission is consistent with arising from a cloud with a size <2 kpc and an H i mass ∼10⁵ M_☉. This is similar to the size and H i content of the Milky Way dwarf galaxy Leo T (Irwin et al. 2007; Ryan-Weber et al. 2008), although no stellar system has been reported at the position of the H i feature.

We do not detect H i at one location in the southernmost extension of the B&T bridge to a 5σ limit of N_H i ⩽ 1.5 × 10¹⁷ cm⁻², inconsistent with the B&T map at the 3σ level. This implies that there is significant angular structure in the gas unresolved by the B&T measurements, or that the bridge is confined to δ > 36°, i.e., within 120 kpc of M31. As all of our detections are near localized peaks in the B&T map, we cannot confirm the existence of a very extended diffuse neutral H i bridge at levels ∼10¹⁷ cm⁻².

No H i was detected from the two known M31 dwarf galaxies covered by our survey, to limits of ≲ 10⁴ M_☉. This is not surprising, as both are dSph and located less than 200 kpc from M31, a proximity that is correlated with an absence of significant H i presumably because of ram-pressure stripping in a hot halo (Blitz & Robishaw 2000; Grcevich & Putman 2009; Nichols & Bland-Hawthorn 2011).

The H i emission in the M31–M33 bridge is extremely faint and beyond the reach of most radio telescopes because of limitations on sensitivity and the quality of instrumental baselines. The GBT spectrum in Figure 7 is among the highest-quality 21 cm H i emission spectra ever obtained at this low noise level. Further 21 cm observations with the GBT are planned to study the structure and kinematics of the M31–M33 bridge. It would also be extremely interesting to measure the bridge in UV absorption lines against distant AGNs to gain information on its metallicity and ionization stage, and the amount of ionized gas that is likely associated with the structure.

We thank the anonymous referee for useful suggestions. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under a cooperative agreement by Associated Universities, Inc. Part of this work was done while N.L.F. was a summer student at NRAO.

Facility: GBT - Green Bank Telescope