Infalling Streamer-Disk-Outow System in a Nearby High-mass Star Formation Region

Theoretical models and numerical simulations suggest that high mass star (with mass > 8 M ⊙ ) can be formed either via monolithic collapse of a massive core 1 , 2 or competitive accretion 3 , 4 , but the dominant mechanism is currently unclear 5 , 6 . Although recent high resolution observations with the Atacama Large Mil-limeter/submillimeter Array (ALMA) have detected physical and kinematic features, such as disks 7 − 9 , outﬂows 10 , 11 and ﬁlamentary structures 12 , 13 surrounding the high mass young stellar objects (H-MYSO), direct detection of the infalling gas towards the HMYSO is still the key to distinguish the different scenarios. Chemically fresh gas inﬂows have been detected towards low-mass stars 14 be-ing formed, which are consistent with the accretion-disk-outﬂow process 15 . In this work we report the detection of a chemically fresh inﬂow which is feeding HMYSO growth in the nearby high-mass star-forming region G352.63-1.07. High quality images of the dust and molecular lines from both ALMA and the Submil-limeter Array (SMA) have consistently revealed a gravitationally-controlled gas inﬂow towards a rotating structure (disk or torus) around the HMYSO. The HMYSO is also observed to have an outﬂow, but it can be clearly separated from the inﬂow. These kinematic features provide observational evidence to support the con-jecture that high-mass stars can be formed in a similar process to that observed in the low-mass counterparts. The chemically fresh infalling streamers could also be related with the disk conﬁguration, fragmentation and accretion

shows the ALMA 3 mm and SMA 1.1 mm dust continuum emissions (see Method), which have a spatial resolution of ∼0.01 pc and should closely trace the dense-core gas. Their contours are overlaid on the 2MASS 2 µm Ks and three-color GLIMPSE IRAC images. Five dust emission components can be clearly identified, labelled as M1 to M5. They should represent individual high-mass dense cores. M1 is the most compact and strongest one therein. It is also associated with strong 6.7 GHz methanol maser emission 16 , which is supposed to be exclusively associated with HMYSOs 17 . The 2MASS 2µm image around M1 (see Figure 1) exhibits a bright bipolar emission feature with a roughly symmetrical distribution over M1. The intense 2MASS 2µm emission usually traces the shock-excited molecular H2 and the emission feature also has a comparable morphology with the shocked outflow in IRAS 20126+4104 (ref. 18). As another well known shock tracer, the IRAC 4.5 µm emission also shows a largely overlapped spatial extent with the 2µm emission. The two emission features thus both strongly indicate a bipolar outflow from M1 (see Figure 1). In contrast to the shock emissions, the dense and warm gas tracers (e.g. CH3CN and CH3OH transitions) are only detected toward M1, suggesting that warm compact gas is only associated with M1. The CH3CN transitions provide a gas temperature estimate of ∼170 K (see Method). M1 thus should be a compact hot molecular core driving a bipolar outflow.
The bipolar outflow is also clearly detected in the molecular transitions (see Figure 2). The ALMA HCO + and SMA CO lines all exhibit high-velocity emission wings on their spectra (Figure 2 right panels) with a speed up to ∼ 15 km s −1 with respect to light of sight (LOS). The emission regions of these high-velocity components are closely overlapped with the infrared shock emissions, and also have a symmetric distribution with respect to M1. As denoted with dashed lines, this bipolar-outflow has a large opening angle (∼ 60 • ). Considering its spatial separation of ∼20000 au (or 0.1 pc) from the driving source, and assuming an inclination angle of 60 • with respect to LOS (same as the angle of inclination of the disk/torus rotational axis; see below), it then implies a dynamic ages of ∼3600 years. With this time scale, the mass-loss rate for the M1 outflow is estimated to be 1.4 × 10 −5 M⊙ year −1 (see Method).
On the other hand, although the molecular outflow lobes and infrared shock emissions have similar spatial extents, the lobes actually have clumpy structures and are less dispersed than the infrared emission. As an explanation, the outflow components ejected from M1 may not be entirely interacting with the surrounding gas. The infrared emission at 2 µm and 4.5 µm would trace considerably more energetic fraction of the outflow that is sufficient to generate shocks. The less intense molecular outflow could propagate to a larger distance, particularly as indicated by the southwestern HCO + lobes (Figure 2 upper panel), which have a projected distance of 0.1 pc from M1.
The SMA images of the molecules DCN, DCO + and H2CO show that these three molecular lines are emitting from a linear structure aligned with M1, M2 and M3 at a position angle of ∼165 • (see main panel of Figure 3). The DCN and H2CO emission is mostly located in the region between M2 and M3 and peaks at M1, but DCO + is only observed in the M2 and M3 regions. Their position-velocity (P-V) profiles along the direction of M1-M2-M3 are shown in the lower panels of Figure 3. The velocity field of the blue-and red-shifted sides both exhibit an symmetric increasing feature towards the center, which is similar to that seen in other Keplerian-rotating disks 19 . It suggests the presence of a rotating structure associated with M1. The enclosed mass calculated by fitting a Keplerian rotation model to the P-V profiles is 9.0/sin 2 θ M⊙ (see Method), where θ is the angle of inclination of the disk/torus rotational axis with respect to the LOS. Combining the structures toward the disk/torus edge at M3 traced by DCO + and dust emission, an inclination angle θ of ∼ 60 • is inferred, indicating an enclosed mass of ∼12 M⊙ in system. This mass is consistent with that estimated from the 1.1 mm dust emission detected with the SMA (see Method).
The molecular emissions show the rotational system to have a radius of ∼ 10 ′′ (7000 au), which is larger than the typical size of a HMYSO revealed by ALMA observations (∼ 1000 au) 7−9,20,21 . Instead, the rotating gas has a larger spatial size, a less regular shape, but is still flattened to some degree onto the mid-plane of the rotational axis. It could be rotating torus surrounding the inner disk, which has a radius of < 1 ′′ (or < 700 au). There is also a smooth temperature increasing trend towards the center as inferred from the H2CO transitions (see Method for details). This also suggests the rotating torus to be the surrounding structure closely associated with the inner HMYSO. A similar circumstellar torus with comparable radius (8000-12000 AU) was detected around the clump SMA1 in S255N (ref. 22).
The hypothesis of rotating torus with decreasing temperature is also supported by the DCO + emission, which only detected towards the outer edge of the rotational system, but absent in the inner region. This molecule is found to trace coldest disk locations 23,24 and the outer disk ring in a low-mass protostar 25 . Taken together, we suggest that the condensations of DCO + close to the M2 and M3 dust emission reflects higher material accumulation at those locations due to an edgebrightening effect along the LOS, or perhaps a thickened ring at the edge (similar to a 'doughnut'). The orientation of the rotational system perpendicular to the bipolar-outflow driven by M1 (see Figure 2), also supports the hypothesis that it is a rotational disk or torus.
It should be noted that the centre of the rotating system is not coincident with the peak of M1 emission, but with an offset of of ∼ 1000 au (or ∼ −1.5 ′′ ; see lower panel of Figure 3) from the position of the M1 peak along the disk/torus axis. The position of the center of gravitation is likely to coincide with the position of the YSO and displacement with respect to the dust emission peak may be caused by inhomogeneous material distribution in the disk/torus system. The differences in the distributions of DCN (detected primarily toward the northwest) and H2CO (detected primarily toward the southeast) also suggests different gas density and temperature environments, as well as chemical diversity over the disk/torus structure.
As shown in Figure 4, the ALMA HC3N and CCH emissions exhibit an overall north-south oriented filament with a scale of ∼ 30 ′′ (0.1 pc, see upper panel of Figure 4). The two species both have a noticeable velocity variation around the M1 position, which may indicate a more extended gas motion, such as an accretion streamer towards M1. The HC3N and CCH are both chemically early-type molecules 29 , thus are expected to trace colder gas regions. The gas temperature derived from our observations of two CCH hyperfine transitions is typically ∼ 10 K (see Method), which is consistent with that measured in other sources 23,30 . These species are likely to be depleted in more evolved dense cores, resulting in less close correspondence between the molecular emission from HC3N and CCH and the dust continuum emission. It is noted that another streamer with distinct velocities are also seen along M4 and M5. It could be driven by the outflow from M4 and eventually enters the main filament at the M5 position.
We inspected the different molecular emissions closer to M1 and found streamers along two more directions traced by the chemically fresh molecues HC3N, CCH and HCO + , as shown in Figure 4 (middle and lower panels). As shown in middle panels, the ALMA HCO + emission exhibits a noticeable spatial extension from southwest to northeast over M1, denoted as PV2 direction. The PV diagram along this direction exhibits velocity gradient with increasing blueshift towards M1. The SMA HCO + emission is less extended than its ALMA counterpart, which is possibly due to the less complete spatial coverage of the SMA baselines. But it shows an overall consistent velocity gradient on the PV diagram. Such velocity increasing feature toward M1 is different from either outflow or core rotation. Moreover, the HCO + extension is almost perpendicular to the mid-plane of the rotating torus. The HCO + feature is therefore more likely to indicate an infall motion towards the mid-plane.
As shown in lower panels, the ALMA HC3N and CCH emissions both exhibit another extension towards northeast as indicated by PV-3 direction. And their PV diagrams again exhibit a steep but smooth gradient towards the mid-plane with a velocity change of V −Vsys=0 to −4 km s −1 over the range r = 20 ′′ to 0 ′′ , The corresponding velocity gradient is approximately 60 km s −1 pc −1 , which is comparable with that seen in Orion 31 .
Assuming that the streamers are free falling due to the gravity of a central mass, we can fit the P-V diagrams to a very consistent degree by adjusting the inclination angle of the streamer, and the projection height of the landing point of the streamer onto the disk/torus. The best-fit central mass (∼ 12 M⊙) is also very consistent with the value derived from the rotational disk/torus system (see Method for details).
The P-V distributions for the chemically fresh molecular species show gravitationally-controlled infalling streamers feeding material onto the high-mass protostar embedded in M1 from at least two directions. The two streamers are also likely to have different landing points. But the landing points are both located on the mid-plane of the rotating structure (torus). The landing points appear to be offset by ∼ 3000 au from the central protostar. We estimate the typical free-fall timescale for these streamers. They are estimated to have t f f ∼ 4.6 × 10 4 years. The total infall rate of material from these streamers accordingly is ∼ 1 × 10 −4 M⊙ year −1 (see Method for details). This implies that the infalling streamers can contribute up to 40% of accreting material in the accretion process see Method).
Numerical simulations predict that asymmetric accretion flows could have a more significant impact on the accretion rate of a protostar compared axisymmetric infalling-envelope 26−28 . Direct imaging of such structures is thus crucial but still rare so far. The presence of chemically fresh accretion streamers has been detected on scales of 10000 au towards a low-mass star forming region in recent AL-MA observations of carbon-chain molecular species 14 . There are no confirmed previous observations of this kind of an accretion flow in a high-mass star formation region, although filament or hub structures have been detected in a number of regions 12,13 .
Our results show clear signatures of infalling streamers towards a rotating disk/torus system around the HMYSO. And the streamers should mainly consist cold dense gas due to their detection in the chemically fresh molecules. The HMYSO is in a sparsely clustered en-vironment and also associated with a set of bipolar-outflows. But the other gas components can all be clearly distinguished from the infalling flows. These observations support the hypothesis that high-mass stars can efficiently form via a monolithic accretion provided that infalling material can be transported into the center in an efficient process. The chemically fresh infalling streamers detected here are also similar to those observed in low-mass star formation, suggesting that they play an important role in both low-and high-mass star formation.
It is also noted that the streamers are falling onto the rotating structure instead of directly onto the central star. The infall in this manner may support the disk accretion with a continuous mass supply, and eventually more efficiently transfer the mass onto the star. But on the other hand, the impact from the infalling gas could also cause a disk fragmentation, generate instabilities, including accretion bursts or episode accretion phenomena, as has been detected in a number of HMYSOs (ref. 32 and references therein). New numerical calculations are required to investigate such kinematic processes and similar high angular resolution observations in additional regions are required to clarify the frequency and importance of these streamers during the high-mass star formation process.   The visibility data were calibrated using the IDL superset MIR. The imaging and analysis were undertaken in MIRIAD. The synthesized beam size, i.e., the angular resolution of the image data are approximately 5.8 ′′ × 2.4 ′′ at 230 GHz and 4.7 ′′ × 1.9 ′′ at 285 GHz, with a position angle of −12.3 • northwest. The typical rms noise is ∼ 50 mJy beam −1 in a spectral channel for the molecular line data. We used the line-free channels in the frequency range 281-289 GHz to produce the 1.1 mm dust emission image. The rms noise for the dust emission image is ∼ 4 mJy beam −1 .
Gas temperature in the hot molecular core in M1. Hot molecular core tracers (such as CH3CN and CH3OH) were only detected towards a compact region (not resolved by either the ALMA nor SMA observations) towards the dust component M1. We have assumed that the molecular gas is in local thermodynamic equilibrium (LTE) to estimate physical parameters. The column density can be calculated using a general formula 34,35 where k is the Boltzmann constant in erg K −1 , W is the observed line integrated intensity of the species in units of K km s −1 , ν is the frequency of the transition in Hz, and Sµ 2 is the product of the total torsion-rotational line strength and the square of the electric dipole mo-ment, Eu/k is the upper level energy in K and Q(Tex) is the partition function at temperature Tex.
When the molecular emission is optically thin and in LTE, we can determine the rotation temperature (Trot) from rotation diagram analysis. The standard formula for such analysis can be derived from Equation (1) as follows: A straight line fitted to a plot of ln(3κW/8π 3 νSµ 2 ) versus Eu/κ will have a slope of 1/Trot. We used eight transitions of CH3CN (from 120 −110 to 127 −117 with frequencies in the range 220.5-220.8 GHz) detected in the SMA observations to estimate the rotation temperature of the hot gas in M1 (see Extended Figure 1). The derived gas rotation temperature is 165 ± 45 K, confirming that M1 corresponds to a hot molecular core. Kepler rotation model for the M1-M2-M3 disk/torus system. In the case of system undergoing Keplerian rotation, the speed of rotation along the line-of-sight for a given enclosed mass (M ) and radius (R) along the disk/torus satisfies the following relation: where G is the gravitational constant, v⋆ is the systemic velocity of the disk/torus system, θ is the inclination angle of the disk/torus rotational axis, and R0 is the offset from the zero position in the P-V diagram (i.e. the M1 position). By adjusting the parameters M, v⋆ and R0 to fit the P-V diagrams of the disk tracers H2CO and DCN, we find M=9.0/sin 2 θ M⊙, v⋆ = −0.3 km s −1 and R0 = −1000 au (or −1.5 ′′ ) provide the best fits to the data. Furthermore, if we assume an inclination angle of θ =∼ 60 • (based on the distorted structures traced by DCO + and dust emission at M3), we then estimate an enclosed mass of ∼12 M⊙ in this rotational system. It should be noted that in our fit, R0 = −1000 au, so the gravitational center of the rotational system is not at the M1 peak, but at position of 1000 au northward away from the M1 peak along the P-V slice. This may be caused by inhomogeneous material distribution in the disk/torus system. Mass estimated from the 1.1 mm dust emission. We estimated the rotational system mass using the 1.1 mm dust emission under the assumption that it is optically thin. Under these conditions, the mass can be found using the relation where Sint is the integrated flux density of the 1.1mm dust emission, D is the distance to the source, κ d is the mass absorption coefficient per unit mass of dust, Bν (T dust ) is the Planck function at temperature T dust , and R d is the dust-to-gas mass ratio. Here we have used κ d =1.14 cm 2 g −1 for 1.1 mm (ref. 36) and a dust-to-gas ratio (R d ) of 1:100 in our calculations. For M1 (Sint=0.1 Jy), which is a hot molecular core with a gas temperature of ∼165 K (see above), we have assumed the dust temperature (T dust ) is equal to the gas temperature under LTE. Based on this, we estimate that the mass of M1 is only 0.05 M⊙. Even considering a dust temperature of 20 K, the mass of M1 is ∼ 0.6 M⊙. These are inconsistent with a high-mass protostar and more likely to trace the mass of the internal hot disk around protostar embedded in M1.
For M3 (Sint=0.15 Jy), located at the disk/torus edge, a cold environment we have assumed a dust temperature of 10 K, which is equal to the cold gas temperature derived from the CCH observations (see below). Under these assumptions we estimate the mass in the region is about 2.5 M⊙. Considering that the gravitation center of the disk/torus rotational system has an offset R0 = −1.5 ′′ from M1 towards M3, we can estimate a total mass of the rotational system to be ∼ 15 M⊙ according the motion of the center of mass. This is consistent with that derived from the Keplerian rotation model (see above). Excluding the mass contributions from the outer torus traced by M2 and M3, the mass of the protostar embedded in M1 is likely in the range 10-12 M⊙.
Gas temperature distribution along the disk/torus. The gas kinetic temperatures along the disk/torus are estimated from the line ratios of para-H2CO 303 −202/322 −221 (ref. 37) by performing the nonlocal thermodynamic equilibrium (non-LTE) radiation transfer code RADEX 38 coupled with collision rates from the Leiden Atomic and Molecular Database calculated by the ref. 39. In the calculations, the number density of molecular hydrogen is 10 5 cm −3 , the column density of para-H2CO is 5×10 13 cm −2 , the line width is 2 km s −1 and the background temperature is 2.73 K. The line ratios lower than 1.67 corresponding to the temperatures higher than 360 K are totally ignored because the line ratios of para-H2CO 303 − 202/322 − 221 are not sensitive for gas kinetic temperature measurements. Extended figure 2 shows the derived gas kinetic temperature distributions and gradient along the disk/torus plane. Notably, we can not determine the temperatures towards northwester region of the H2CO emission, because the above method is invalid due to the lower ratios between the two lines.
Gravitationally-controlled infalling streamers. In the case of gravitationally-controlled infall, it is possible to estimate the speed of free-fall for an enclosed mass (M ) and radius (R) using the relation: where θ is the inclination angle of the streamer along the line-of-sight, v0 is the velocity at the starting position of the streamer, and R0 is the projection height of the landing point of the streamer onto the disk/torus. We have assumed M ∼ 12M⊙, the mass derived from the the rotational disk/torus system and adjusted the parameters θ, v0 and R0 to fit the P-V distributions of the infalling streamers shown in Figure 4. These P-V diagrams can be well fit using this method, suggesting that the streamers are moving towards the disk/torus under gravitational attraction. The values estimated for θ, v0 and R0 are different for each streamer, indicating that the infalling directions of these streamers and the landing point on the disk/torus are different. The free-fall timescale of the infalling streamers can be derived as For these streamers, the typical observed length R in the sky-plane is equal to ∼ 1.0 × 10 4 au, and for M = 12 M⊙, the free-fall timescale is ∼ 4.6 × 10 4 years. Cold gas properties of the streamers. Under LTE conditions, we assume an excitation temperature (Tex) equal to the gas kinematic temperature. We can derive the excitation temperatures from the CCH observations (one of the chemically fresh molecules in the streamers) using the ratio between its two hyperfine lines (N = 1 − 0, J = 3/2 − 1/2, F = 2 − 1; N = 1 − 0, J = 3/2 − 1/2, F = 1 − 0). Firstly, the optical depth of the main line (F = 2 − 1) can be estimated using the following relation: where TB,m and TB,s are the observed brightness temperature in K for the main and satellite lines, respectively. We have used that the expected opacity ratio (r) between the main and satellite lines is τm/τs = 2, under optically thin conditions. We can further calculate the excitation temperature using the relation: where T bg (=2.73 K) is the background brightness temperature and T0 = hν/k. We obtain the ratio between the peak intensities of the CCH maps of the two hyperfine transitions (from ALMA observations) to calculate the optical depth and excitation temperature in each pixel. The derived excitation temperature distributions are illustrated in Extended Figure 2. The excitation temperatures are the range 8-20 K, with a typical value of ∼10 K. The streamers can almost be traced by the CCH (in particular for the northern one), indicating that the streamers are chemically fresh and in a cold environment.
Infall material rates of the streamers. Based on the optical depth (τ ) and excitation temperature (Tex), we can determine the column density in the streamer regions. Using Eq. (1). we can derive the column density in each pixel of the CCH streamer. We summed the CCH gas mass in the northwestern streamer region and estimate the total gas mass to be ∼ 2.5 M⊙ (assuming a typical CCH abundance of [CCH/H2] = 8 × 10 −9 (ref. 23)) and a mean mass per particle of µ = 2.33 mH (ref. 40).
The CCH emission associated with the southeastern streamer is much weaker. Using Eq. (1), we have derived the gas mass along this streamer using the ALMA HCO + (3-2) observations, assuming the emission is optically thin and at a gas temperature of 10 K. This gives a total gas mass of ∼ 1.5 M⊙ in the south-eastern streamer region, assuming a typical HCO + abundance of [HCO + /H2] = 1.2 × 10 −9 (ref. 30) Using the calculated free-fall timescale of 4.6 × 10 4 years, we can estimate the infall material rates along these streamers to be about 5.5× 10 −5 M⊙ year −1 for the northwestern stream and 3.1 × 10 −5 M⊙ year −1 for the southeastern stream. This implies a total of infall rate of 8.6 × 10 −5 M⊙ year −1 from the two streamers.
Assuming the that accreting timescale is equal to the free-fall timescale of the infalling streamer (4.6 × 10 4 years; see above), and that the mass of the protostar in M1 is about ∼10 M⊙, this suggests that the overall accretion rate is ∼ 2.1 × 10 −4 M⊙ year −1 . Therefore, it also suggests that the detected infalling streamers can contribute up to 40% of the material in the accretion process.
Mass-loss rate from the outflow driven by M1. We estimate the mass-loss rate from the outflow driven by M1 using the CO outflow. Assuming that the CO emission in the outflow lobes is optically thin at a gas temperature of 20 K (which is a typical temperature in a cold CO gas region, and a bit higher than the CO evaporation temperature of 15 K; ref. 41), we can use Eq. (1) to derive the column density for each pixel. A total of mass of ∼ 0.05 M⊙ is obtained summing over the CO emission in the northeastern outflow lobe region under the assumption of a CO abundance of [CO/H2] = 1.0 × 10 −4 . If we consider the masses obtained for the blue-and red-shifted components are nearly same and a dynamic age of ∼ 3600 years for the outflow (see main text), it suggests a mass-loss rate of ∼ 1.4 × 10 −5 M⊙ year −1 from the outflow driven by M1. Notably, considering that part of CO emission in the outflow regions may be optically thick and mass of the energetic part of the outflow traced by the infrared emission, the above estimation of mass-loss rate can be treated as a lower limit.

Data availability
The data from both ALMA and SMA that support the plots within this paper and other findings of this study are available from the corresponding author upon reasonable request.    Molecular out ow images driven by M1 detected with ALMA and SMA. The molecular out ow images (blue contours for blueshifted components and red contours for red-shifted components) are given in left panels, which are obtained from the integrated emission of the corresponding molecules in the blue-and red-shifted velocity ranges given in right panels. The background in the left panels is the three-color GLIMPSE IRAC image same as lower panel of Figure 1, and black contours show the 3 mm dust emission detected with ALMA, with contours same as Figure 1. Upper panels present the HCO+ (J = 1 -0) out ow detected with ALMA, with contour levels of (3, 5, 7, 9, 11, 13)x0:1 Jy beam−1 km s−1. Lower panels show the CO (J = 2 -1) out ow detected with SMA, with contour levels of (1,2,4,6,8,10,12,14)x6 Jy beam−1 km s−1. The dust component M1 is marked with plus. The beam shapes of the two observations are given by the ellipses. The overall out ow structure of is described by the dashed straight lines.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. extended g1.pdf extended g2.pdf extended g3.pdf