Molecular Gas Feeding the Circumnuclear Disk of the Galactic Center

The interaction between a supermassive black hole (SMBH) and the surrounding material is of primary importance in modern astrophysics. The detection of the molecular 2-pc circumnuclear disk (CND) immediately around the Milky Way SMBH, SgrA*, provides an unique opportunity to study SMBH accretion at sub-parsec scales. Our new wide-field CS(2-1) map toward the Galactic center (GC) reveals multiple dense molecular streamers originated from the ambient clouds 20-pc further out, and connecting to the central 2 parsecs of the CND. These dense gas streamers appear to carry gas directly toward the nuclear region and might be captured by the central potential. Our phase-plot analysis indicates that these streamers show a signature of rotation and inward radial motion with progressively higher velocities as the gas approaches the CND and finally ends up co-rotating with the CND. Our results might suggest a possible mechanism of gas feeding the CND from 20 pc around 2 pc in the GC. In this paper, we discuss the morphology and the kinematics of these streamers. As the nearest observable Galactic nucleus, this feeding process may have implications for understanding the processes in extragalactic nuclei.


INTRODUCTION
The origin of the 2-pc CND in the GC has remained unclear in spite of intensive study for the past decades (e.g., Guesten et al. 1987;Jackson et al. 1993;Amo-Baladrón et al. 2011;Harris et al. 1985;Mezger et al. 1989;Etxaluze et al. 2011;Lau et al. 2013;Wright et al. 2001;Montero-Castaño et al. 2009;Martín et al. 2012;Herrnstein & Ho 2002, 2005aChristopher et al. 2005;Requena-Torres et al. 2012;Mills et al. 2013). The CND is a ring-like molecular structure rotating with respect to the supermassive black hole SgrA*, within which are the arc-shape ionized gas streamers called SgrA West (Roberts & Goss 1993). The SgrA West arms converge at SgrA* and have been proposed to originate from the inner edge of the CND. The CND, being the closest molecular reservoir in the GC, is critical on the understanding of the feeding of the nucleus. The replenishment of the CND itself, therefore, is an important problem. Thus, many observations have been made to detect the kinematic connections between nearby molecular clouds and the CND with the main purpose of detecting the inflow of gas to the CND. Previous NH 3 (3,3) observations carried out with the VLA have detected several "streamers" (Okumura et al. 1989;Ho et al. 1991;Coil & Ho 2000;McGary et al. 2001) (Figure 1). The 20 km s −1 cloud (hereafter 20 MC) and the 50 km s −1 cloud (hereafter 50 MC) lying 20 pc south of SgrA * , appear to morphologically connect to the CND, with several NH 3 (3,3) streamers called the "southern streamer", "SE1 streamer", and the "northern ridge". The southern streamer extends northward from the 20 MC toward the southeastern edge of the CND (Okumura et al. 1989;Ho et al. 1991;Coil & Ho 2000). The northern ridge originated from the inner edge of the 50 MC is also connected to the CND with a velocity gradient of 0.5 km s −1 arcsec −1 spanning 110 (4 pc). SE1 extends northward and kinematically connects to the eastern lobe of the CND.
The northern ridge, southern streamer, and SE1 were proposed to feed the CND based on observed increasing line widths and heating as they approach the GC. However, the understanding of the accretion of the gas and that of the CND were limited by the previous lower-transitions and lower-resolution data (e.g. Tsuboi et al. 1999). For instance, the CND and the region interiors were not well sampled by the NH 3 (3,3) line, but are prominent in the NH 3 (6,6), HCN and CS lines (e.g. Montero-Castaño et al. 2009, also see Figure 1). This suggests that from the CND inward, the gas is dense and hot. The streamers also become faint toward the CND, and this suggests that the HCN and CS dense gas tracers are more reliable to detect the connecting material. Previous interferometric data also suffered from the negative bowel effect due to abundant missing information of the large-scale structures.
Thus, single-dish mapping is essential to recover the large-scale component. Moreover, wide-field mapping is also important to trace the origin of the streamers from the ambient clouds (20/50 MC). Therefore, we observed the central 30 pc with the Nobeyama Radio Observatory (NRO) 45-m and the Caltech Submillimeter Observatory (CSO) with the CS J u = 5, 4, 2 lines. The details of the observations are described in Hsieh et al. (2015Hsieh et al. ( , 2016, where we presented the studies of the molecular outflow in the central 30 pc of the GC with 40 resolution maps. In this paper, we focus on the investigation of the molecular inflow associated with the CND utilizing the higher resolution of the 20 in the CS(J = 2 − 1) map.

Comparing NRO 45-m/CSO and SMA data -Revisiting Structures
In Figure 1 and Figure 3 we show the integrated intensity and the channel maps of our NRO-45m CS(J = 2 − 1), CSO CS(J = 5 − 4) and the SMA HCN(J = 4 − 3) data (Montero-Castaño et al. 2009;Liu et al. 2012), respectively. While the channel maps in Hsieh et al. (2016) are shown with a low-velocity resolution of 40 km s −1 , we display maps here with 5 km s −1 resolution. Here, we note the well-studied components in the central 30 pc of the GC. In the HCN(J = 4 − 3) integrated intensity map (beam= 5.87 × 4.44 ), the CND appears as a ring-like structure. The western streamers outside the CND are resolved into protrusions called W-1 to W-3 in the SMA map. The southern lobe and the northern ridge are also present in the HCN(J = 4 − 3) map. The W-4 component is proposed to connect to the southern lobe in Liu et al. (2012). This is also called the negative-longitude extension (NLE) by Oka et al. (2011) and Takekawa et al. (2017). The HCN(J = 4 − 3) emission of the southern streamer is fainter than its NH 3 (3,3) emission. The 20/50 MC located south of the CND are clearly seen in the CS maps.
A comparison between the CS and HCN(J = 4 − 3) channel maps reveals that there is a streamer in CS, a counterpart to the western streamer, which is visually connected to the 20 MC through a linear feature around −18 km s −1 to 2.5 km s −1 (bounded by two cyan lines in Figure 3). We call this linear feature "extended western streamer (extW-streamer)" in this paper. The extW-streamer can also be seen in the previous dense gas maps (Tsuboi et al. 1999;Amo-Baladrón et al. 2011;Liu et al. 2012), but it is not mentioned before due to its lower brightness in several J = 1 − 0 lines. Our new CS maps are more sensitive to both diffuse and compact components.
In Figure 4, we present the CS(J = 2 − 1) map integrated from −24 km s −1 to −11 km s −1 (grey) overlaid on the integrated HCN(J = 4 − 3) map. The velocity ranges are selected to avoid the contamination of the 20 MC. The extW-streamer appears as a smooth ridge-like feature visually connecting the 20 MC and the western streamer. The resolved W-1 to W-3 protrusions appear as "compact cores" within the extW-streamer near the CND. The spectra of the extW-streamer are shown in Figure 4. The SMA spectra are smoothed to the same resolution (20 ) as the NRO 45-m data for comparison. The extW-streamer shows narrow line widths (full width half maximum (FWHM) of 30 km s −1 ) at the positions 7, 8, 9 and becomes broader as it becomes the western streamer near the CND (FWHM of 80 km s −1 ). Note that at the positions 7, 8 and 9, a feature called connecting ridge (CR) appears at the velocity around 40 km s −1 as discussed in Hsieh et al. (2015). The CR is proposed to be elevated disk emission connecting to the extraplanar feature called the polar arc (PA). This CR appears to have no physical association with the CND nor the streamers. The CS(J = 2 − 1) spectra show a wind component in the western streamer, which is absent in the HCN(J = 4 − 3) spectrum. This wind component might be filtered out in the interferometric data if the broad wind component is extended.
The NRO 45-m CS(2-1) data can be found on the following link. This streamer is the ext-W streamer (squares) and the western streamer (yellow and red squares in the upper and lower panels, respectively). The other streamer remains at a constant velocity around −16 km s −1 (yellow and red lines in the upper and lower panels, respectively). The line widths of the extW and the Western streamer are also increasing as they approach the CND by about a factor of two. The −16 km s −1 -streamer, on the other hand, maintains the same line width. The lv-diagrams indicate that the extW-streamer originates from the ambient cloud and moves towards the CND. Our wide-field map reveals this structural connection better than previous maps. Moreover, the extW-streamer also exists in CS(J = 5 − 4), while the −16 km s −1 streamer disappears. This suggests a high-excitation nature of the extW-streamer, consistent with being physically located in the GC.
Eastern streamer - Figure 6 displays CS(J = 2 − 1) and CS(J = 5 − 4) lv-diagrams sliced along the eastern part of the CND (the sense of "eastern" is in the conventional celestial coordinate relative to the western streamer). We find that there is a high velocity-component from 60 km s −1 to −100 km s −1 in Figure 6 to (359.9 • , −0.055 • ). The spectra in Figure 7 show two velocity components. The higher velocity component traces the southern lobe of the CND and connects to the protrusions east of the CND in the celestial plane. This component shows a kinematic behavior similar to the western streamer. We call this feature "eastern streamer". The eastern streamer is also seen in CO(J = 3 − 2) (Oka et al. 2011) but it is affected by foreground absorption.
Significantly more extended and connected features are seen in the CS(J = 2 − 1) map.
In Figure 7, the eastern streamer seems to connect to the northern ridge, which is a linear feature located at a velocity of ∼ −10 km s −1 . McGary et al. (2001) reported that the northern ridge shows a kinematic connection to the CND. However, in Figure 6, the origin of the eastern streamer might be from the 50 MC. A part of the 50 MC shows elongation in the lv-diagrams from b = −0.067 • and seems to be stretched towards the eastern streamer.
The physical associations of the 50 MC and the northern ridge are unclear because of the interaction with the supernova remnant SgrA East (Serabyn et al. 1992;Lee et al. 2003).
An additional component, called C1 by Oka et al. (2011), shows a kinematic behavior similar to the eastern streamer.
Note that the eastern streamer is different from the southern streamer. The southern streamer has velocities of ∼ 20 − 40 km s −1 . This suggests that the southern streamer is superimposed on the eastern streamer along the line of sight. The kinematics of the southern streamer show no significant velocity gradients and do not appear to be directly associated with neither the CND nor the nucleus (Herrnstein & Ho 2005b), unless the southern streamer is moving perpendicular to the line of sight.

Longitude-Velocity Diagrams
In Figure 8, we compare the extW-streamer, the eastern streamer, and the CND in lv-diagrams in HCN(J = 4 − 3) and in CS(J = 2 − 1). The extW-streamer is averaged Similar lv-diagrams of the ext-W/eastern streamers are presented by Oka et al. (2011) in CO data. In their interpretation, the extW/eastern streamers are the outer part of the CND, which is still infalling to the CND with an infall velocity of 50 km s −1 . However, the CO data can be contaminated by foreground absorption, and Oka et al. (2011) extracted the highly excited gas with CO(J = 3 − 2)/CO(J = 1 − 0) ratios ≥ 1.5. This extraction relying on intensity ratios can be uncertain and arbitrary. Since our dense gas map traced by CS(J = 2 − 1) is less affected by foreground emission, we can more cleanly depict the kinematic structures. Nevertheless, if the high-velocity "ring" is physically located in the GC, the lv-diagrams are not suitable for comparison because one fixed l corresponds to different radii of the ring. Therefore, in order to clarify the kinematics of the connection between the streamers and the ambient clouds, we show position-velocity diagrams drawn on trajectories along the "ring" (phase plot) (e.g. Jackson et al. 1993;Martín et al. 2012) in the next section.

Phase Plots: Kinematics of the CND, the Eastern and extW-Streamers
In Figure 9 and Figure   we found amount to 20-70 km/s. This range in infall motion defines a band in the phase plot that appears to capture the extended emission of the streamers.
In Figure 11 and Figure 12 we show the phase plots of the CND along a projected circle with radius of r = 3.5 pc for the upper panels and r = 2.5 pc for the lower panels.
The Keplerian curves are overlaid on the phase plots of the CND with r =3.5 pc ( We also find that the western streamers intersect with the CND in Figure 11 and 12, which provides some evidence that the western streamers are "converging and feeding" toward the CND. For the eastern streamer, however, we notice that it is fainter than the western streamers and it does not show a clear signature of transition into the CND. One of the possible reasons is that the feedback of the supernova remnant (SNR) SgrA East (Mezger et al. 1989;Ekers et al. 1983) may disturb the eastern streamer. The SgrA East is known to interact with the 50 MC (e.g., Genzel et al. 1990;Ho et al. 1991;Serabyn et al. 1992;Zylka et al. 1999). The age of the SgrA East (10 4 years)  is consistent with the dynamical time scale of the eastern streamer. Owing to the large velocity gradient and elevated temperature of the western streamer, McGary et al. (2001) also proposed that the western streamer appears to have been swept up by the expanding shell of the SgrA East. The western streamer shows expanding motion outwards with the SNR shell. In this regard, the "infall velocity" implemented in our simple model may also be interpreted as an expanding motion. However, we do not see significant morphological and kinematic disturbances of the western streamer in our SMA and NRO 45-m data.
In this paper, we alternatively propose that the radial motion of the eastern/western streamers may be attributed to the infall toward the CND. The infall picture based on our data is sketched in Figure 13. The molecular gas originated from the ambient clouds is tidally stretched into long streamers (protrusions). These streamers are gravitationally captured by the central potential and spirally rotate as they converge toward the CND.
This picture is consistent with the past theoretical modelings, which predict that the CND is formed by the tidal capture and disruption of nearby molecular clouds (Sanders 1998;Vollmer & Duschl 2002;Wardle & Yusef-Zadeh 2008;Mapelli & Trani 2016).

Lower limit of the infall rate
We estimate the mass inflow rate based on the NRO 45-m data because the SMA map suffers from missing flux. We estimate the infall rate of the eastern streamer assuming that it is currently stripped away from the 50 MC and approaching the CND. The molecular gas mass of the eastern streamer under local thermal equilibrium (LTE) assumption is ∼ 120 M (CS(J = 2 − 1)/(J = 1 − 0) intensity ratio of 1). The mass accretion rate over 7 × 10 4 years is thus ∼ 2 × 10 −3 M yr −1 . This accretion rate is consistent with the simulation in Vollmer & Duschl (2002). The total accreting gas mass is therefore . The CND may be transient and will dissolve if its density is lower than the tidal threshold (∼ 10 7 cm −3 ). In this paper, we find that the replenishment of the gas accreting onto the CND is on going, we then estimate whether this accretion rate is significant to The clump densities are on the order of 10 3−4 cm −3 , which is lower than the tidal threshold limit. These observational results suggest that the 50 MC is a possible "seeding factory" to continue feeding the CND over its crossing time (a few 10 5 years to cross 20 pc). We therefore conclude that in spite of the density arguments on the life time of the CND (Christopher et al. 2005;Montero-Castaño et al. 2009;Requena-Torres et al. 2012;Mills et al. 2013;Smith & Wardle 2014;Harada et al. 2015), a precise measurement of the CND mass will be also crucial to determine its life time.

Conclusions
In this paper, we present a dynamical picture of the infall scenario connecting from 20 pc to the central 2 pc, which might also have implications for the nuclei feeding in nearby galaxies. With our wide-field map of CS(J = 2 − 1), we find that the western streamer identified in the interferometric maps has a counterpart called extW-streamer. We also find a feature called eastern streamer in CS(J = 2 − 1) which shows a physical connection to the 50 MC via tidally stretched clouds. The extW-and eastern streamers are the outskirt of the CND and show a slower rotation than the CND. We also find that the extW-and the eastern streamers can be described by a simple Keplerian rotation and infall model. We thank the reviewer for a thoughtful review and constructive comments to improve CND consists of features called the southwest lobe, the southern extension, the northeast arm, and the northeast lobe (Christopher et al. 2005;Montero-Castaño et al. 2009;Liu et al. 2012). Several protrusions called W-1, W-2, W-3, and W-4 are also labeled (Liu et al. 2012).
W-1, W-2, and W-3 are the counterparts of the western streamer seen in the NH 3 (3,3) line.        ext-W streamer. The western streamer intersects with the CND as shown in the phase plots.