Dense, Parsec-Scale Clumps near the Great Annihilator
E. J. Hodges-Kluck, M. W. Pound, A. I. Harris, J. W. Lamb, M. W. Hodges
aa r X i v : . [ a s t r o - ph . GA ] F e b Accepted by the Astrophysical Journal
Dense, Parsec-Scale Clumps near the Great Annihilator
Edmund Hodges-Kluck , Marc W. Pound , Andrew I. Harris , James W. Lamb & Mark Hodges Department of Astronomy, University of Maryland, College Park, MD 20742-2421 Owens Valley Radio Observatory, California Institute of Technology, Big Pine, CA 93513
ABSTRACT
We report on Combined Array for Research in Millimeter-Wave Astronomy(CARMA) and James Clerk Maxwell Telescope (JCMT) observations toward the
Ein-stein source 1E 1740.7-2942, a low-mass X-ray binary (LMXB) commonly known as the“Great Annihilator.” The Great Annihilator is known to be near a small, bright molec-ular cloud in a region largely devoid of emission in CO surveys of the Galactic center.This region is of interest because it is interior to the dust lanes which may be the shockzones where atomic gas from the HI nuclear disk is converted into molecular gas. Wefind that the region is populated with a large number of dense ( n ∼ cm − ) regionsof excited gas with small filling factors. The gas appears to have turbulent support andmay be the result of sprays of material from collisions in the shock zone. We estimatethat ∼ − × M ⊙ of shocked gas resides in our r ∼ ′ , ∆ v LSR = 100 km s − field.If this gas has recently shocked and is interior to the inner Lindblad resonance of thedominant bar, it is in transit to the x disk, suggesting that a significant amount ofmass may be transported to the disk by a low filling factor population of molecularclouds with low surface brightness in larger surveys. Subject headings:
Galaxy: center, ISM: clouds
1. Introduction
In spite of the difficulties of determining the structure and dynamics of the inner 400 pc of theGalaxy, remarkable progress has been made by comparing surveys in different tracers of moleculargas. These surveys have revealed the average properties of the gas (Bania 1977; Liszt & Burton1978; Bally et al. 1987; Stark et al. 1988; Bitran et al. 1997; Oka et al. 1998; Tsuboi et al. 1999;Martin et al. 2004). The gas is excited, dense, and exists in conditions unique in the Galaxy,including high external pressures and strong magnetic fields. The most prominent line emissioncomes from the various rotational energy levels of CO, and surveys exist from J = 1 − J = 7 −
6. It is evident from these degree-scale results that the molecular medium is not welldescribed by its average properties, but over the past few decades a convincing picture of the 2 –Galactic center has emerged. At least one stellar bar crosses the Galaxy, and the x and x orbitfamilies resulting from its potential contain atomic and molecular gas respectively (Binney et al.1991). The molecular gas on the x orbits is largely found in a “Galactic center ring” (GCR) andis both dense and turbulent (Fux 1999). Because gas on the GCR is forming stars, it is unlikelythat the molecular medium inside the x disk is primordial, and therefore some mechanism mustexist for converting atomic gas to molecular gas. The region in between the atomic disk and theGCR, the site of the bar’s inner Lindblad resonance (ILR), is where the transition occurs, and isthought to contain two dust lanes in small spiral arms where molecule-forming shocks and spraysoccur (Rodriguez-Fernandez et al. 2006). The GCR and surrounding molecular gas near these dustlanes are collectively known as the “central molecular zone” (CMZ). The most obvious failure ofthis model is that it has not yet explained the asymmetric distribution of gas around the dynamicalcenter of the Galaxy, Sgr A*, although Rodriguez-Fernandez & Combes (2008) argue that the GCRmay be accreting from only one side.This simple picture is remarkable considering the observed complexity of the region. How-ever, it necessarily overlooks details critical to a complete understanding. In particular, we areinterested in the specific nature of the process converting atomic gas to molecular gas. Thatthe dust lanes are the site of molecule-forming shocks is likely given the absence of star forma-tion, and studies in line ratios tracing shock chemistry indicate that the lanes contain shocked gas(Rodriguez-Fernandez et al. 2006), but the shock environment has yet to be explored in detail. Pre-and post-shock environments appear as relative voids in ( l, v ) diagrams; the emptiness of the regionvaries with the choice of spectral line, but even in CO (1–0), it is clear that there are empty regionswithin the CMZ, inside what Binney et al. (1991) saw as the “CO parallelogram” and earlier studiesidentified as an “expanding molecular ring.” This boundary is likely several structures includingthe molecule-forming shock zones (Rodriguez-Fernandez & Combes 2008), and since shocked gasmust fall farther into the Galactic potential, the relative paucity of emission interior to the par-allelogram is interesting. There are several possible explanations for the voids including (1) theshocked gas primarily joins the x disk at certain regions (e.g. the l = +1 . ◦ cloud), (2) the fillingfactor of shocked gas in these regions is too low to be seen by large-scale surveys (along the lines ofthe suggestion of Oka et al. (1998)), (3) the newly-formed molecular bundles are extremely diffuseor not emissive, or a combination of the above. We expect that studying the post-shock side of thedust lanes will provide information regarding the state in which the gas arrives on the x disk, andmay also be a useful contrast to molecular cloud formation in dust lane shocks in the disk. Motivated by published data (Bally & Leventhal 1991), we chose to observe the area near abright molecular cloud near the far side dust lane at l = − . ◦ , b = − . ◦ , and v LSR = − − which lies in such a void in CO (1–0) and HCN (1–0) surveys (Bally et al. 1987; Lee1996). Historically, this cloud has been of interest because of its coincidence on the sky with the 3 –low mass X-ray binary (LMXB) candidate known as the “Great Annihilator” (
Einstein source 1E1740.7-2942). The Great Annihilator is a bright X-ray and γ -ray source and was originally thoughtto be a black hole accreting directly from a host molecular cloud which, in turn, was brightened bythe association (Bally & Leventhal 1991; Mirabel et al. 1991). The original hypothesis of Bondi-Hoyle accretion from the ISM is almost certainly incorrect given the present understanding of thesystem as a LMXB (Main et al. 1999), but as Weidenspointner et al. (2008) proposed a correlationin spatial distribution of LMXB systems and diffuse γ -ray emission in the Galaxy, the strong andnarrow annihilation line still makes the possibility of physical association intriguing. Historically,the two strongest pieces of evidence for such an association were the stated small probability ( ∼ + (1–0) emission from the jets seen with the VLA (Phillips & Lazio1995). This offset was attributed to the high levels of ionization which would destroy the HCO + close to the black hole, and Lepp & Dalgarno (1996) suggested that HCN may be able to survivein more highly-ionized gas, proposing future observations.Our interest in the gas dynamics of the inner Galaxy led us to consider the Great Annihilatorregion as a target for a pathfinding observation for two reasons: (1) the bright cloud is in a “post-shock void” region and (2) it has a well documented brightness and spectrum in several molecularlines (Bally & Leventhal 1991; Mirabel et al. 1991; Phillips & Lazio 1995). This cloud is thus areliable pathfinder target for a larger map in the region near the shock zones. It is desirable to havea known bright target included in a wide field both for orientation and determining the suitability ofa given instrument to our science goals. Additionally, even supposing the LMXB is responsible forthe brightness of the v LSR = −
140 km s − cloud, it is not responsible for the presence of the densegas. The detection of one bright cloud, therefore, suggests the presence of additional molecular gas. However, it is imperative to consider whether the small region near the Great Annihilator istypical of the void regions we wish to study; determining whether a relative void is “typical” re-quires choosing a scale. We do this by considering both an observational strategy and our broaderscience goals. We have reason to expect the detection of emission in these voids based on thepublically-available Bally et al. (1987) CO (1–0) data. In the published l − v maps of this data,which have been integrated over several 1 ′ grid points in latitude, the relative voids appear empty,but an examination of the l − v maps for each slice in latitude shows faint molecular emissionin parts of these regions. An absolute void across all latitude slices is probably actually empty,but when integrating across several slices, a faint structure with little extent in latitude will be-come fainter relative to extended structures nearby. For example, in the Bally et al. (1987) databetween − . ◦ < b < − . ◦ , there appears to be a bridge of emission between the CO paral-lelogram shock lanes and the x orbits at l = − . ◦ , v LSR = 120 km s − . This emission occursin a region we would expect to be mostly empty according to bar parameters such as those in 4 –Rodriguez-Fernandez & Combes (2008), but may be related to the highly excited bundles of gasseen in earlier Jenkins & Binney (1994) work. Closer to the Galactic plane, this bridge does notappear, whereas the surrounding x disk and shock lane structures persist at the same longitudes;the region appears in the integrated map only as a void. Interior to the CO parallelogram, a “typi-cal” region for detailed study ought to be large enough to probe both the apparently empty regionsas well as some of the faint emission.For a detailed interferometric study of the region, the scale should clearly be much largerthan an individual molecular cloud’s size, and we postulate that the projected radius of the cloudassociated with the Great Annihilator is typical. Assuming interaction with the LMXB as wellas some intrinsic brightness, this measured radius is only influenced by the presence of the LMXBto the extent that the LMXB is responsible for the brightness of the cloud, i.e. the ratio betweenexcitation produced by the Great Annihilator’s particle emission and that produced by other means.Based on the evidence for a well-defined spectral peak at v LSR = −
140 km s − centered spatially atsome distance from the LMXB, as well as the physical arguments for how far the positrons couldtravel (Phillips & Lazio 1995), it seems likely that the measured radius is close to the physical one.As discussed in § u − v coverage and sensitivity can be achieved over the entire field ofview. A happy spatial medium exists for fields several arcminutes in radius; the known cloud has aradius on the order of 1 pc (24 . ′′ for a distance of 8.5 kpc) and a field several arcminutes (1 ′ ≈ . l − v maps of the Bally et al. (1987)data and the theoretical models we wish to test (Binney et al. 1991). An inspection of the CO(1–0) data suggests that the easiest way to view a sizable segment of the post-shock region at onceis to observe near one of the vertical segments of the shock lanes as projected in the l − v planethanks to our large range in ∆ v compared to ∆ l or ∆ b . We also wish to avoid contamination fromline-of-sight molecular emission and larger scale structures of the inner Galaxy. For this study, wemapped a region near the Great Annihilator described by r ∼ ′ and ∆ v LSR ∼
100 km s − whichmeets these criteria.In § § §
2. Observations & Data Reduction2.1. CARMA D & E Arrays
The Combined Array for Research in Millimeter-wave Astronomy (CARMA) is a heterogenousinterferometric array made up of six 10.4 and nine 6.1 m radio dishes located at 2195 m at Cedar Flatin the Inyo Mountains of California. The observatory operates in several different configurations.We report here on two tracks in the compact August 2007 E-Array and one track in the April2008 D-Array (Table 1) towards the Great Annihilator cloud in HCN J = 1 − + J = 1 − + (1-0) map at 17:40:43.0-29:43:25.0 (1950.0). The HCN and HCO + observations were conducted simultaneously with bothlines in the upper sideband and with a velocity range of -90 to -190 km s − . For one of the E-arraytracks (baselines 8–66 m) and the D-Array track (baselines 8–108 m), the weather was excellent,whereas the second E-Array track produced usable data requiring substantial flagging. Visibilitiesin the E-array which had one antenna shadowing another were discarded. While the shortestbaselines tracing the largest structures are lost in this procedure, the E-Array data considered byitself nonetheless reproduces the D-Array detections wheree the fields overlap. Importantly, theE-array includes numerous shorter, non-shadowed baselines significantly different from those inthe D-Array, producing significantly improved u − v coverage. To calibrate and analyze the data,we employed the MIRIAD software package (Sault et al. 1995). We customized a standard scriptprovided by CARMA to inspect the data, apply calibrations and flags, and extract clean maps. Weused an SDI CLEAN algorithm to generate the maps presented here; a check against a maximumentropy deconvolution scheme showed no significant differences.Figs. 1 & 2 show the integrated amplitude maps for the combined calibrated D & E-Arraydata, and it is immediately obvious that there is a large number of features, some of which arespatially resolved, and that many of these features are weak or absent in the HCO + data. Theclouds are dense and the spatial coverage of our r ∼ ′ field between -90 and -190 km s − is about25%. At a glance, it is clear that we see a large number of features which may have small fillingfactors in larger beams or survey grids. We clearly detect the Great Annihilator cloud (labeledcloud 1) to the left of map center. Note that because there were fewer E-Array mosaic pointingsthan in D-Array (Table 1), the SNR decreases towards the map edge. Because the map containsnegative amplitudes, and features at different velocities may overlap spatially, the integrated mapsdo not accurately reflect the strength of emission in any one cloud; we found clumps that have nolabels in Figs. 1 & 2. These clumps are evident in the channel maps (Figs. 5 & 6). 6 – We obtained CO J = 6 − T ∗ A for a forward coupling to a Jupiter-size ( ∼ ′′ ) source. Further information on observing and calibration methods may be found inHarris et al. (1995).The central field of Fig. 3 is pointed towards the peak of HCO + (1–0) emission in Phillips & Lazio(1995), and the remainder of the spectra are taken from fields offset by 20 ′′ in α and δ ; each pointinglasted for 240 s. Additional data were obtained in CO(6–5) toward the Phillips & Lazio peak,but no line was detected with an upper limit 20 times smaller than the CO (6-5) line brightnesstoward the same position.The detection of the cloud in the J = 6 − + (1–0) map.
3. Observed Properties
The presence of a number of bright regions in Figs. 1 & 2 and the suggestion that theseregions may be quite excited is an important result. To measure the properties of these regions,we used the clumpfind algorithm (Williams et al. 1994) using the average rms noise per spectralchannel within 3 ′ of map center and requiring 60 pixels per clump (about twice the area). Clumpsfound at r > ′ reside in the low-signal map edge, and were therefore rejected, although we foundadditional structure in the map edge we do not further discuss here. The clumpfind results aregiven in Table 2 and are consistent with a visual inspection in that the same regions are identified.Because clumpfind treats pixels along the v LSR -axis no differently than the spatial pixels, manyof the clumps reported are spatially unresolved at the 2 σ level, and in extracting spectra for thesecases we use the beamsize instead of the reported clumpfind radii. In Fig. 4 we show the spectraextracted for these clumps; for convenience, where clumpfind found multiple clumps which appearto be associated along all three axes we have extracted one average spectrum for the region usingthe MIRIAD imspect task. These spectra were extracted from roughly the 2 σ contour tracing theassociation boundaries. Using clumpfind introduced a systematic bias into the sizes of the clumpsalong the spatial and v LSR -axis, and we discuss the implications of this bias on our derived massesin § For our analysis, we divide our clumps into resolved associations and unresolved individualclumps, where the former are identified as groups of clumps in the same region of α , δ , and v LSR .The resolved clumps have characteristic scales of r ∼ . ′′ at a distance of 8.5 kpc) and,in some cases, several resolved regions may themselves be associated; we consider cloud 2 to bea molecular cloud containing multiple dense regions since we see nearby negative dips at similarpeak velocities. Hereafter we adopt “cloud” to refer to resolved associations of clumps found with clumpfind and “clump” to refer to any individual component or unresolved feature. The resolvedclouds are labeled 1, 2, 3, 4, 5, and 6, and these are the brightest features in the map; the fainterclouds are unresolved, but may have extended structure we would see with a deeper integration. Inthe case of clump 2b, the algorithm found two clumps in the HCN data which are spatially coincidentand differ by 5 km s − along the v LSR -axis, but a detailed inspection of the channel maps (Figs. 5& 6) convince us that clumpfind ought not to have found two clumps. Similarly, two clumps werefound in cloud 3 where a visual inspection suggests that they should be considered one structure.The clumps are listed as the algorithm found them in Table 2, and we use the results from thealgorithm through the rest of the analysis, but we note that there may be systematic inaccuracieswith our approach. All other clumps found by clumpfind passed visual inspection.The HCN and HCO + line strength is comparable in about half of the clumps; perhaps inter-estingly, only cloud 1 has a significantly stronger HCO + line. On average, where HCO + emissionis present, the FWHM line widths are comparable to the HCN despite differing line strengths.The lines are roughly symmetric, but we lack the signal to rigorously investigate the line profiles.Measured FWHM line widths for individual clumps range from 3–14 km s − whereas average linewidths for the clouds as presented in Fig. 4 range from 8–30 km s − , assuming the line is adequatelyrepresented by a single Gaussian component.As expected, many of the clumps and associations are seen within the faint emission describedin § x orbits and may notbe part of the same structure. Another group of clumps lies near -180 km s − and may lie on the x orbits, but the grouping in l and b suggests that these clumps are actually on non-circular orbits,since they are grouped between − . ◦ < b < − . ◦ and there is a relative lack of emission wherethe GCR intersects our region in the Bally et al. (1987) data. The apparent clustering in Fig. 7is largely a result of plotting all the clumps found with clumpfind and, when associations such ascloud 2 are considered as one object at some average ( l, v ), the clumps appear randomly distributed(in latitude, they are distributed preferentially towards the Galactic equator, and we do not havedata for b > − . ◦ ). If the clouds were on x orbits, we would expect them to be clustered towardslarger negative v LSR near -180 km s − at l ≈ . ◦ . The isolated clumps and associations of gas areisolated from one another along all three axes such that it seems unlikely that they are part of onelarger complex. This implies that the “bridges” seen in l − v slices of the Bally et al. (1987) dataare not coherent structures, but rather, the coherent structures exist on smaller scales. We remarkon the individual clouds below. 8 – − Cloud 1 is the molecular cloud associated with 1E 1740.7-2942 and is the onlycloud in the sample for which the HCO + line is much stronger than the HCN line. In the CARMAmaps, we do not see the extended structure to the south which Phillips & Lazio (1995) attribute todifferent ionization regimes near the jets of 1E 1740.7-2942, but we do see some low-signal emissionto the southeast (clump 15) at about -130 km s − which may be the “ridge” they describe. TheJCMT map does not extend far enough to assess the presence of clump 15 in CO J = 6 −
5, butclearly shows that the highly-excited CO traces the HCO + well in the region observed and thatthe cloud is compact. Clumps 11a and 11b at -180 km s − are spatially coincident with cloud 1,and may be physically associated with each other. In §
2: -100 km s − The clumps labeled 2a-2d are each resolved individually, but are close on the sky,have nearly identical peak velocities, and are near strong negatives in the deconvolved maps whichalso appear at the same velocity. These negatives indicate large-scale structure resolved out by theinterferometer and motivate a physical picture of several dense clumps within a diffuse envelope.The HCN and HCO + line strengths are roughly equal in each cloud, and 2b has the brightestpeak flux in the map (Fig. 4 shows the HCO + in cloud 1 as stronger because it is summed overan extraction box). However, cloud 2b coincides on the sky with a smaller clump at -180 km s − which makes the integrated amplitude in the region in Figs. 1 and 2 significantly brighter than thecontribution from 2b alone. Since the HCO + emission is weaker than the HCN, 2c is not identifiedas a clump in the HCO + data and is rather split into two tails extending from 2b and 2d. Weassume that because it is identified as a separate clump in HCN that it ought to be identifiedseparately from the two brighter emission cores.
3: -115 km s − This cloud is curious, since the average FWHM line width over the associationis significantly smaller in HCO + than HCN despite the strong emission in both lines. This effect isnoticeable in the individual clumps found in the region, and is pronounced even when the spectrumis extracted from a box within contours > σ , indicating that perhaps the HCO + in the cloud isconfined to the densest regions. − Clouds 4a and 4b have similar peak velocities and line widths in theiraverage spectra, and both are much weaker in the HCO + line than the HCN. The clouds arenotable for their broad lines which clumpfind breaks up into strings of smaller clumps at similarvelocities. The similar peak velocities mean that 4a and 4b are physically separated by < ′ andsuggest they are or were part of the same complex, although we do not see negative amplitudesthat would point to a diffuse envelope. Cloud 6 is similar in line width, size, and line strength toeither component of cloud 4, but it is several arcminutes away. 9 –
5: -180 km s − This cloud is close on the sky to cloud 6 and is similarly weak in HCO + , so theextracted spectrum includes a small portion of cloud 6 which appears as a weak feature at -125 kms − . Cloud 5 is the only certainly resolved cloud at -180 km s − , although clump 17 behind cloud2b may barely be resolved. As mentioned above, several clumps listed in Table 2 appear in the channel maps and not inthe integrated amplitude maps. This occurs either when the clump coincides with another cloudon the sky (as in the case of clump 17 behind 2b) or with a negative at a different v LSR (as in thecase of clump 9). The “unresolved” clumps do not appear to be a distinct population from theresolved clouds, and are instead characterized by weaker emission. We extracted average spectrafrom regions of about a beamsize; clumpfind does find smaller spatial sizes for these clumps thanfor the ones comprising resolved associations. The line widths of the unresolved clumps are inagreement with those of the resolved ones.
4. Derived Properties
We have detected ∼
20 clumps of gas in a 3 ′ -radius field. These clumps have r ≤ v = 100 km s − , the clumps cover about 25% ofour field. The gas tends to emit more strongly in HCN than HCO + , and at least one of these clouds(cloud 1) produces a CO (6–5) line which traces the HCO + and HCN gas quite well, meaning thecloud is highly excited. Surveys in CO (4–3) emission show that clouds near our field are excited,although we know from comparing the JCMT CO (6–5) and unpublished CO (2–1) data thatnot all gas in the region is so highly excited.We now use these measured properties to derive additional quantities for our clouds, and onthe basis of these results, propose that the sources are dense ( n H > cm − ), and that they haveshocked recently. Because of the weakness of the HCO + in many of the clumps, our argumentsregarding the dense gas rely on the HCN data. Mass estimates invariably reflect their underlying physical assumptions, so further testingof these assumptions is required to assess their accuracy. The virial mass in particular is veryuncertain for Galactic center clumps. Although the external pressures appear to be an order of 10 –magnitude too low to bind clumps on our scales (Miyazaki & Tsuboi 2000), the clumps may bebound by the strong magnetic fields known to exist in the Galactic center, or may be unbound.The virial mass is primarily useful as a way to compare our results to other work in the absence ofbetter estimators but adopting values from previous work is fraught with peril, since the structureobserved is necessarily dependent on the beamsize.The sensitivity of the virial mass to line width means that whether we take M vir = Σ i m vir ,i ,where a cloud is made up of clumps with mass m vir,i , or M vir = ¯ σ v R/G , where ¯ σ v is taken from anaverage description such as Fig. 4, will influence our results. The line width–size relation (Larson1981; Miyazaki & Tsuboi 2000) cannot be used to resolve such ambiguity because in our case it issystematic. Reported line widths also generally appear to reflect the beamsize used. We investigatedthe extent to which noise influences clumpfind in order to determine whether multiple clumps foundin resolved clouds accurately describe substructure. There is almost certainly substructure present:the question is whether clumpfind accurately detects it. A visual inspection of the channel mapsin addition to testing clumpfind ’s behavior with artificial resolved clouds in a featureless corner ofour map convinced us that the two clumps found in both 2b and 3 in the HCN are not real, and thata single clump is a better description. On the other hand, using the line width from the averagespectra in Fig. 4 likely overestimates the mass more severely than clumpfind underestimates it.We therefore report the clumpfind results and posit that the masses and densities contained in theresolved clouds are somewhat higher. Table 3 contains the results for each individual clumpfind detection, m vir ,i . In determining ¯ n H for each clump, we used µ = 1 . m H .Is using the virial mass reasonable? We know the clouds must be dense, since typical criticaldensities for HCN (1-0) are n ≥ cm − , so the assumption that they are gravitationally boundis not wholly unreasonable, although a shock compression might induce similar densities whichwould then dissipate. The line widths indicate macroscopic turbulent support and are roughlysymmetric, so if the clouds are bound, they are unlikely to be much more concentrated than atvirialized relaxation. We posit that for dense, non-collapsing gas the virial mass is an adequate orderof magnitude description, but note that other mass estimates from studies with larger beamsizesdisagree with the virial mass by up to an order of magnitude (Miyazaki & Tsuboi 2000).In the resolved clouds our derived densities range from 10 . − . cm − and, as we stated, webelieve our results understimate the virial mass. If the LTE mass is a better descriptor of the“real” mass (Miyazaki & Tsuboi 2000), then our densities are ∼
10 times too high and, therefore,typical for Galactic center clumps with radii 2–10 times larger than those of our clumps. If ourclumps are smaller structures inside larger diffuse clouds such as those resolved with larger beams,then they must either be much more dense than 10 cm − , or else comprise most of the massand volume of their hosts; for clumps ∼
10 times larger in radius the latter seems unlikely. Theconvergence of the critical density and size arguments lends some credence to the virial estimates asan order of magnitude estimate, so we believe our resolved clouds contain large amounts of gas atdensities exceeding 10 cm − . The brightnesses and line widths of the unresolved clumps suggestthat they contain gas at similar densities, so we estimate that the total mass contained in our 11 –cube is 1 − × M ⊙ , contained in parsec-scale bundles with a small filling factor. We note thatsince the HCO + line widths generally disagree with the HCN in a given cloud, the HCO + virialmasses would be different, but the HCO + lines have a lower SNR, making the measurements moreuncertain. Independently of the virial mass, we know the density to be quite high from the critical densityrequired to excite HCN (1–0). In an inspection of 2MASS J, H, and K-band images of our field, wedo not see enhanced star formation activity associated with our resolved clouds; the clouds have notyet collapsed, so they are either unbound and will decompress in a crossing time or bound and closeto virialized. In either case, the clumps must be quite young if they are starless, since they appearto have turbulent support. The crossing time, t cross ∼ R/ ∆ v FWHM , is about 10 yr (Table 3) forour resolved clouds. Assuming a circular orbit at 180 pc from the Galactic center with a velocityof 200 km s − (the highest LSR velocities seen in the CO parallelogram Binney et al. (1991)), theorbital period is ∼ × yr, over an order of magnitude greater than the dynamical timescale. Itis likely that the clumps have recently fallen onto their present orbits, so they probably experienceda recent shock. However, the constituents of cloud 2 suggest that condensation and collapse mayoccur.
5. Discussion5.1. The Post-Shock Voids
The Binney et al. (1991) hypothesis identifying the CO parallelogram with the ILR of theGalactic bar may not be correct in detail (Ferri`ere et al. 2007), but the identification of the COparallelogram with starless dust lane shock zones (Rodriguez-Fernandez et al. 2006) strengthensthe proposed mechanism for molecule formation in shocks near the ILR. The large surface densityof molecular gas in the Galactic center and orbit families in the barred potential make it all butcertain that continuous inflow from the HI disk occurs.The density, excitation, and ( l, v ) distribution of our clumps suggests that we are indeedseeing molecular gas which has recently formed in these shock zones. Owing to their density, theclumps represent a significant amount of mass ( > M ⊙ in our field); this mass is in a regionwhere the orbits guarantee eventual transport to the x disk. This mass is transported in smallbundles, possibly as a spray of material from streamer collisions. If these bundles are unbound anddecompress, diffuse molecular gas may rain down onto the GCR, otherwise they may experienceballistic impacts with the diffuse molecular envelopes in the GCR. The small filling factor of thedense regions may also explain the voids seen in larger surveys, although, as noted in § x disk.There are two challenges to the proposal outlined above. First, the clumps and clouds maybe associated by a larger, bridge-like structure. This would imply a directed mass flow at certainpoints on the CO parallelogram rather than a large number of small bundles of shocked gas dis-tributed throughout the region in between the shock lanes and the x orbits. Second, the multipleconstituents of cloud 2 in a larger envelope (but still small on the scales of large surveys) suggestthat condensation into dense clumps, as seen in Galactic molecular clouds, may take place underhigher pressures. The first case is difficult to assess because a close inspection of individual latitudeslices of the Bally et al. (1987) CO (1–0) data shows many small features in between the COparallelogram of Binney et al. (1991) and the GCR. We consider it unlikely that these representsteady-state channels for mass flow because of the nature of the self-intersecting x orbits interiorto the ILR. However, it is possible that material from molecular and atomic gas clouds shredded inthe shock lanes maintains some coherent structure as it dissipates angular momentum. The secondchallenge may fit neatly into the first if larger structures fall onto the self-intersecting x orbits, butaside from the constituents of cloud 2, we see little evidence for dense bridges between otherwisedistinct clumps. Furthermore, we see no evidence for star formation, so even internal turbulentshocks must have been recent. If a mechanism for recent shocks is required, the proximity to theshock lanes provides a natural explanation.To determine whether the Great Annihilator region appears typical, we applied the scaleof our map ( r ∼ ′ , ∆ v LSR = 100 km s − ) to other areas along the shock lanes labeled inRodriguez-Fernandez et al. (2006) using the Bally et al. (1987) data cube. For the diagonal por-tions of the CO parallelogram, we stretched our scale in longitude and shrunk it in ∆ v . Ignoringthe regions near v LSR = 0 km s − and obvious large scale structures such as the GCR, we findthat the 1E 1740.7-2942 region for our observation is typical of the post-shock regions in the CO(1–0) map. The emission associated with this region in the Bally et al. (1987) data appears tocome from a structure covering ∆ b ∼ . ◦ , ∆ l ∼ . ◦ , and ∆ v ∼
100 km s − . We think itunlikely that the structure is bridged to the x orbits based on a detailed inspection of the regionat different emission contour intervals. We therefore have no reason to believe that the presenceof the Great Annihilator in the region detracts from a gas-dynamical analysis. However, we notethat this analysis assumes that the observed similarities between regions in the Bally et al. (1987) CO (1–0) maps correspond to physical similarities in the environments of the shock lanes. Al-though the region is small, our results are largely consistent with the Binney et al. (1991) theoryof molecule-forming shocks at the ILR even though, in detail, the scenario is more complex. Ifwe are seeing clumps that have recently shocked, it is harder to fit them into alternative picturesexplaining the CO parallelogram, e.g. an “expanding molecular ring” or the Stark et al. (2004)proposal of a stalled ring of gas accumulating from the true ILR farther away from the Galactic 13 –center.We know that the GMCs on the x disk are forming stars from cores in clumps, but we donot know whether the clumps primarily form via condensation or external perturbation, nor thefilling factor of small clumps in gas accreting onto these orbits. Jenkins & Binney (1994) find intheir sticky-particle hydrodynamical models that the steady-state x disk contains a large numberof strongly shocked clumps of gas, although the decompression timescale suggests that if the cloudsare still dense, they must be at least gravitationally bound if they experience no additional shocksinterior to the dust lanes. Their simulations also find that although there are places where shockedgas preferentially joins the x disk (e.g. the l = 1 . ◦ near- side molecular complex), bundles ofgas fall in from the shock zone at the ILR at many angles. Jenkins & Binney (1994) admit thattheir results are only marginally successful at reproducing the features of the Galactic center, andsubsequent models (e.g. Rodriguez-Fernandez & Combes (2008)) have done much better. Yet theresults we present here suggest that at least the accretion of material onto the x disk in the formof dense bundles is still a possibility. Our data provide useful contrasts to two of the arguments made previously for a physicalassociation between 1E 1740.7-2942 and cloud 1, but we cannot rule out a possible association.The argument in favor of physical association due to a small chance of coincidental superpositionof a black hole with a bright molecular cloud depends in detail on choosing a cutoff brightness. The5-7% chance previously reported (Bally & Leventhal 1991; Mirabel et al. 1991) relied on studyingthe density of peaks in molecular line surveys near the Galactic center using some brightnesscriterion. Without establishing a similar criterion, we cannot directly compute the probabilityof a chance association. However, the presence of a large number of clumps in the region withsimilar line widths and a wide range of peak fluxes suggests that a highly excited clump is commonenough not to require association with a black hole. At approximately the 2 σ contour, the chanceof coincidental superposition with any clump in our field is 25%. That the HCO + is noticeablystronger in cloud 1 is unusual, but there are smaller clumps where HCO + is stronger than HCN.With a larger velocity range, the chance of coincidental superposition at arbitrary velocity maywell increase.The Phillips & Lazio (1995) argument in favor of physical proximity relies on the interpretationof a ridge of HCO + emission parallel to the VLA jets associated with 1E 1740.7-2942 as evidencefor an ionization rate gradient. Their ridge is significant ( > σ ) and separated from the VLA jetsby 15 ′′ with a peak velocity close to -140 km s − , suggesting that it is part of the same structureas cloud 1, shown in Fig. 2 in Phillips & Lazio (1995). The primary detection of cloud 1 in theirOVRO map agrees well with both our CARMA and JCMT data in spatial extent, but we do notdetect the southern ridge apparent in their map. Instead, we see clump 15 at -130 km s − tothe south east of cloud 1 – it is closer to the VLA jets than 15 ′′ . An inspection of clump 15 in 14 –( l, v ) space shows a clear separation between it and cloud 1. Moreover, this clump is detected inthe JCMT CO (2–1) data (not shown) whereas we do not see it in CO (6–5). Mirabel et al.(1991) also detect the clump in CO (2–1) and not in CS (2–1). There is undoubtedly emissionto the south of cloud 1, but it is not as excited or warm as cloud 1, nor do we see evidence for a“ridge” linking it to cloud 1. These results call into question whether an ionization rate gradientexists. Furthermore, Lepp & Dalgarno (1996) argue, based on their model, that if the HCO + ridgeis caused by an ionization rate gradient, we would expect to see HCN molecules surviving closerto the black hole. There is no evidence in our maps that HCN emission associated with cloud 1is closer to the Chandra
X-ray source than the HCO + . Our results make it more difficult to makethe case for physical proximity between the Great Annihilator and cloud 1.
6. Conclusions
We have used CARMA D & E-Array observations in tandem with JCMT data to study theregion near 1E 1740.7-2942, a relative “void” in ( l, v ) diagrams of CO survey data. The mostimportant result of our work is the discovery that even in the regions with low average emissivity,there is a substantial amount of mass ( > M ⊙ ) contained in small, dense, excited regions implyinga recent shock. These bundles have scales of r ∼ l, v )space, so shocks or sprays in the nearby dust lanes of the CO parallelogram naturally explain ourobservations. The small filling factor of the dense bundles accounts for why they are not seen oronly faintly present in large surveys. We investigate in detail the relationship between cloud 1 andthe LMXB 1E 1740.7-2942 and argue that the probability of coincidental superposition with excitedgas is higher than originally estimates, and that we see no evidence for a proposed ionization rategradient; we can explain why cloud 1 is excited in the JCMT map without invoking a black hole.This research was funded in part by the Astronomy Department of the University of Marylandat College Park. The authors thank M. Leventhal for useful discussion of the diffuse 511 keVannihilation radiation in the Galactic plane, and R. Genzel & J. Zmuidzinas for motivation andsupport of the JCMT observations. We used 7-m Bell Labs data made publically available by J.Bally.Support for CARMA construction was derived from the states of California, Illinois, and Mary-land, the Gordon and Betty Moore Foundation, the Kenneth T. and Eileen L. Norris Foundation,the Associates of the California Institute of Technology, and the National Science Foundation.Ongoing CARMA development and operations are supported by the National Science Foundationunder a cooperative agreement, and by the CARMA partner universities. This material is basedon work supported by the National Science Foundation under grant numbers AST-0540450 andAST-0540399. 15 – REFERENCES
Bally, J., Stark, A., Wilson, R., Henkel, C. 1987, ApJS, 65, 13BBally, J. & Leventhal, M. 1991, Nature, 353, 234BBania, T.M. 1977, ApJ, 216, 381Binney, J. Gerhard, O., Stark, A., Bally, J. & Uchida, K. 1991, MNRAS, 252, 210Bitran, M., Alvarez, H., Bronfman, L., May, J. & Thaddeus, P. 1997, A&AS, 125, 99BFerri`ere, K., Gillard, W. & Jean, P. 2007, A&A, 467, 611Fux, R. 1999, A&A, 345, 787FHarris, A.I, Schuster, K.F., Genzel, R., Plathner, B. & Gundlach K.H. 1994, IJIMW, 15, 1465HHarris, A.I., Avery, L.W., Schuster, K.F., Tacconi, L.J. & Genzel, R. 1995, ApJ, 446L, 85HJenkins, A. & Binney, J. 1994, MNRAS, 270, 703Larson, R.B. 1981, MNRAS, 194, 809LLee, C.W. 1996, ApJS, 105, 129Lepp, S. & Dalgarno, A. 1996, A&A, 306, L21Linke, R., Stark., A, Frerking, M. 1981, ApJ, 243, 147LLiszt, H.S. & Burton, W.B. 1978, ApJ, 225, 815Main, D.S., Smith, D.M., Heindl, W.A., Swank, J., Leventhal, M., Mirabel, I.F. & Rodr´ıguez, L.F.1999, ApJ, 525, 901Martin, C., Walsh, W., Xiao, K., Lane, A., Walker, C. & Stark, A. 2004, ApJS, 150, 239Mirabel, I., Morris, M., Wink, J., Paul, J. & Cordier, B. 1991, A&A, 251, L43Miyazaki, A. & Tsuboi, M. 2000, ApJ, 536, 357MOka, T., Hasegawa, T., Sato, F., Tsuboi, M., Miyazaki, A. 1998, ApJS, 118, 455OPhillips, J. & Lazio, T. 1995, ApJ, 442, L37Rodriguez-Fernandez, N.J., Combes, F., Martin-Pintado, J., Wilson, T.L. & Apponi, A. 2006,A&A, 455, 963Rodriguez-Fernandez, N.J. & Combes, F. 2008, A&A, 489, 115 16 –Qin, S., Zhao, J., Moran, J., Marrone, D., Patel, N., Wang, J., Liu, S., & Kuan, Y. 2008, ApJ, 677,353Sault, R.J., Teuben, P.J. & Wright, M.C.H. 1995, ASPC, 77, 433SStark, A., Bally, J., Knapp, G. & Wilson, R. 1988, LNP, 315, 303SStark, A., Martin, C., Walsh, W., Xiao, K., Lane, A. & Walker, C. 2004, ApJ, 614, L41Tsuboi, M., Handa, Toshihiro & Ukita, Nobuharu 1999, ApJS, 120, 1TWeidenspointner, G., Skinner, G., Jean, P., Kn¨odlseder, J., von Ballmoos, P., Bignami, G., Diehl,R., Strong, A., Cordier, B., Schanne, S. & Winkler, C. 2008, Nature, 451, 159Williams, J., de Geus, E. & Blitz, L. 1994, ApJ, 428, 693W
This preprint was prepared with the AAS L A TEX macros v5.2.
17 –Fig. 1.—: Contour map of integrated amplitude of HCN (1–0) for combined D and E-Array data.Dashed contours indicate negatives in the map. The mean RMS for the integrated map is 2.6Jy/beam · km s − ; a 1-contour feature is roughly a 2 σ detection. Some real detections appearweaker in the map as a result of spatial coincidence with negative features, others are behindbrighter clouds–see text. Clumps in Table 2 have been labeled. A crosshair near cloud 1 shows thelocation of 1E 1740.7-2942 (Mirabel et al. 1991), and the beamsize is at bottom left. The Galacticequator is not in our field, but is to the top right. 18 –Fig. 2.—: Contour map of integrated amplitude of HCO + (1–0) for combined D and E-Arraydata. Dashed contours indicate negatives in the map. The mean RMS for the integrated map is1.7 Jy/beam · km s − ; a 1-contour feature is roughly a 2 σ detection. Some real detections appearweaker in the map as a result of spatial coincidence with negative features, others are behindbrighter clouds–see text. Clumps in Table 2 have been labeled. A crosshair near cloud 1 shows thelocation of 1E 1740.7-2942 (Mirabel et al. 1991), and the beamsize is at bottom left. The Galacticequator is not in our field, but is to the top right. 19 –Fig. 3.—: JCMT CO J = 6 − σ contour. The spectra were extracted from the combinedD and E-Array data in each line. The label at the bottom of each frame marks the clump orcloud for which the spectrum was extracted; the same feature in a different frame has lower signal.In cases where the 2 σ contour was smaller than a beamsize, the spectrum was extracted from arectangle with the height and width of the major and minor axes of the synthesized beam. 21 –Fig. 5.—: Velocity channel maps for the combined D and E-Array HCN (1-0) data. The 63 originalchannels have been binned into 30 channels here. 22 –Fig. 6.—: Velocity channel maps for the combined D and E-Array HCO + (1-0) data. The 63original channels have been binned into 30 channels here. 23 –Fig. 7.—: Bally et al. (1987) CO (1–0) l − v maps for (a) − . ◦ < b < − . ◦ and (b) − . ◦
Array Config. Source Int. Time + Beam HCO + PA HCN Beam HCN PA(hr) (arcsec) (deg.) (arcsec) (deg.)D08A 2.36 37 1733-130 2148+069 MWC349 15.4 ± × ± ± ± × ± ± ± × ± ± ± × ± ± mospsf task within the central 20 ′′ × ′′ of the combined data sets.
25 –Table 2. CARMA Measured HCN/HCO + Clump Properties
Label Line α δ v
LSR T peak R clump ∆ v F WHM (J2000) (J2000) (km s − ) (K) (arcsec) (km s − )1 HCO +
17 43 56.1 -29 44 19.0 -141.7 2.47 26.9 4.81 HCN 17 43 55.8 -29 44 22.0 -145.0 1.61 24.1 4.12a HCO +
17 43 48.9 -29 43 25.0 -105.6 1.90 25.4 9.52a HCN 17 43 48.2 -29 43 55.0 -107.0 1.43 25.0 10.62b HCO +
17 43 44.8 -29 43 16.0 -103.9 2.51 26.9 4.82b HCN 17 43 44.8 -29 43 13.0 -102.0 2.58 24.2 7.22b HCN 17 43 44.8 -29 43 13.0 -107.0 2.08 20.5 5.22c HCO +
17 43 44.8 -29 43 16.0 -103.9 2.51 26.9 4.82c HCN 17 43 42.7 -29 42 58.0 -92.1 1.24 23.0 11.62d HCO +
17 43 43.2 -29 43 25.0 -105.6 1.90 25.4 9.52d HCN 17 43 42.9 -29 43 22.0 -102.0 2.31 23.5 11.83 HCO +
17 43 47.5 -29 41 37.9 -117.1 1.95 21.3 6.33 HCN 17 43 48.0 -29 41 34.0 -116.9 2.27 22.1 9.73 HCN 17 43 47.5 -29 41 25.0 -111.9 1.89 18.9 8.84a HCN 17 43 40.9 -29 43 40.0 -123.5 1.53 17.3 11.04a HCN 17 43 41.1 -29 43 40.0 -128.4 1.20 14.6 6.54b HCN 17 43 38.3 -29 43 55.0 -121.8 1.40 16.4 5.24b HCN 17 43 38.3 -29 43 52.0 -133.4 1.33 15.8 8.14b HCN 17 43 38.3 -29 43 55.0 -128.4 1.30 15.7 3.65 HCN 17 43 45.9 -29 45 10.0 -178.0 1.24 20.0 9.66 HCO +
17 43 45.7 -29 46 16.0 -125.3 0.92 9.7 5.96 HCN 17 43 45.9 -29 46 13.0 -126.8 1.28 18.6 5.86 HCN 17 43 45.9 -29 46 07.0 -121.8 1.03 12.5 2.77 HCN 17 43 50.5 -29 42 04.0 -186.3 1.50 15.8 3.08 HCN 17 43 52.1 -29 41 19.0 -103.7 1.48 17.9 13.89 HCO +
17 43 46.2 -29 42 28.0 -123.6 1.41 12.9 3.49 HCN 17 43 45.7 -29 42 28.0 -120.2 1.29 13.4 5.110 HCO +
17 43 50.5 -29 43 13.0 -130.2 1.01 11.3 2.110 HCN 17 43 51.0 -29 43 31.0 -115.2 1.11 13.0 3.511a HCN 17 43 55.6 -29 43 43.0 -182.9 1.06 13.1 3.711b HCN 17 43 55.1 -29 44 34.0 -184.6 0.85 16.3 7.612 HCO +
17 43 57.9 -29 43 07.0 -126.9 1.22 14.8 1.812 HCN 17 43 58.1 -29 43 13.0 -123.49 0.87 13.0 5.613 HCN 17 43 57.7 -29 43 10.0 -168.1 0.99 10.4 4.314 HCN 17 43 50.8 -29 44 34.0 -115.2 0.80 11.7 3.115 HCO +
17 43 56.3 -29 44 58.0 -130.2 1.05 17.1 3.216 HCN 17 43 40.2 -29 41 28.0 -105.3 2.17 11.1 4.116 HCN 17 43 41.1 -29 41 28.0 -107.0 1.82 12.3 4.916 HCN 17 43 41.6 -29 41 19.0 -111.9 1.38 9.9 7.116 HCN 17 43 41.3 -29 41 22.0 -97.1 1.30 9.1 6.016 HCN 17 43 41.3 -29 41 37.0 -115.2 1.20 16.3 10.517 HCN 17 43 45.2 -29 43 01.0 -181.3 1.33 16.1 6.817 HCN 17 43 45.2 -29 42 55.0 -188.0 1.26 9.9 4.417 HCN 17 43 45.2 -29 42 58.0 -191.2 1.23 13.7 2.718 HCN 17 43 50.1 -29 42 13.0 -97.1 0.93 12.5 3.3
26 –Table 2—Continued
Label Line α δ v
LSR T peak R clump ∆ v F WHM (J2000) (J2000) (km s − ) (K) (arcsec) (km s − )18 HCN 17 43 49.8 -29 42 25.0 -88.8 0.90 10.3 3.319 HCN 17 43 51.0 -29 43 31.0 -115.2 1.11 13.0 3.5 The HCO + clumpfind run connected 2b and 2c, whereas the HCN did not. Clump 16 was halfway outside our criterion for rejection in the map edge; the rms erroron the peak flux is greater than the other clumps, hence more clumps were found with thealgorithm.Note. — clumpfind results for our data cube. Clumps 1-6 we consider “resolved associations”or “clouds” whereas 7-19 are “unresolved clumps” based on the 2 σ contours in the integratedmap. All features are found within 3 ′ of the map center. Table 3. CARMA HCN Derived Properties
Label ∆ v F WHM
R M vir log [¯ n (H )] t cross (km s − ) (pc) (10 M ⊙ ) log[(cm − )] (105