ALMA Observations of Massive Clouds in the Central Molecular Zone: Ubiquitous Protostellar Outflows

The ALMA observations were carried out in the C40-3 and C40-5 configurations in 2017 April and July (project code: 2016.1.00243.S). Details of the sample selection, observation setup, and data calibration can be found in Paper I. The four clouds in the sample are listed in Table 1. The covered fields are chosen based on Submillimeter Array (SMA) and VLA observations that revealed potential sites of high-mass star formation including H₂O masers and massive dense cores of 0.2 pc scale (Lu et al. 2019b). We imaged the Band 6 (1.3 mm) spectral lines using CASA 5.4.0. The covered frequencies range from 217–221 GHz to 231–235 GHz with a uniform channel width of 0.977 MHz (1.3 km s⁻¹). The effective spectral resolution is 1.129 MHz (1.5 km s⁻¹) after a Hanning smoothing done by the observatory.

Table 1. Properties of the Clouds

Cloud	Vlsr	No. of Cores	No. of Outflows	Fraction with Outflows	$\bar{X}$(SiO)	$\bar{X}$(SO)	$\bar{X}$(CH3OH)	$\bar{X}$(H2CO)	$\bar{X}$(HC3N)	$\bar{X}$(HNCO)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
The 20 km s−1 cloud	12.5	471	20	0.042	1.69 × 10−9	7.10 × 10−9	0.99 × 10−7	0.80 × 10−8	0.65 × 10−9	4.40 × 10−9
Sgr B1-off	31.1	89	5	0.056	1.39 × 10−9	6.19 × 10−9	0.86 × 10−7	1.09 × 10−8	1.24 × 10−9	4.32 × 10−9
Sgr C	−52.6	274	18	0.066	2.39 × 10−9	9.87 × 10−9	1.33 × 10−7	1.84 × 10−8	1.48 × 10−9	5.44 × 10−9
All three clouds	⋯	834	43	0.052	2.05 × 10−9	8.49 × 10−9	1.15 × 10−7	1.40 × 10−8	1.16 × 10−9	4.95 × 10−9
The 50 km s−1 cloud	48.6	0	0	⋯	⋯	⋯	⋯	⋯	⋯	⋯

Note. Column (1): cloud name. Column (2): V_lsr of the cloud (Kauffmann et al. 2017). Column (3): number of identified cores (Paper I). Column (4): number of identified outflows in this work. Column (5): the fraction of the cores that have identified outflows. Columns (6)–(11): mean molecular abundances of all the outflows in the cloud. Only the abundances with independent measurements are considered (entries without notes in the last column in Table 3; see Section 4.1.2).

Download table as:  ASCIITypeset image

We first manually identified line-free channels in the visibility data and fed them to the uvcontsub task to subtract the continuum baseline. Then, we used the tclean task to image the spectral lines, with Briggs weighting and a robust parameter of 0.5, and multiscale algorithm with scales of [0, 5, 15, 50, 150] pixels and a pixel size of 004. The image reconstruction was carried out in a two-step manner: first, the auto-masking algorithm with the recommended parameters ¹⁷ was employed in tclean to automatically identify and clean signals; then, the tclean task was restarted using clean models and residuals from the previous step as input, and all pixels within the 20% primary beam response included in the clean mask, to a threshold of ∼5σ (8 mJy beam⁻¹ per channel), in order to clean any residual significant signals. In a few cases where strong spatially diffuse emission is detected (e.g., H₂CO in the 20 km s⁻¹ cloud), a threshold of 8 mJy beam⁻¹ may be too low and causes the clean algorithm to diverge. We elevated the threshold to 10 mJy beam⁻¹ for these lines. We have compared images produced by our automatic approach with those produced by a fine-tuned manual tclean of several lines and found that the images are almost identical.

The resulting synthesized beam is on average 028×019 (equivalent to 2200 au×1500 au) but slightly varies between lines. The largest recoverable angular scale is 10'' (∼0.4 pc) with the shortest baseline length of 17 m. The spectral line rms achieved is between 1.6–2.0 mJy beam⁻¹ (0.8–1.0 K in brightness temperatures) per 1.3 km s⁻¹ channel depending on frequencies and regions.

By their nature, interferometric observations do not recover structures on size scales larger than their largest recoverable angular scale (10'' or 0.4 pc for these observations). If structures larger than this exist in the field, the flux captured by interferometers will be smaller than the true flux, which is referred to as the missing flux problem. Spatially extended (≳0.4 pc) structures including outflows and filaments are seen in the spectral line images. In principle, we can combine our data with a shorter baseline as well as single-dish data to recover any spatially extended emission. Such data are available from the SMA and the Caltech Submillimeter Observatory (CSO) observations published in Lu et al. (2017, 2019b) and Battersby et al. (2020).

However, we note several issues that prevent us from efficiently combining the data: (i) the sensitivity of the SMA data is not optimal for the combination. For several regions, e.g., Sgr C, the SMA observation recovers an even smaller flux than the ALMA data, suggesting that some weaker emission is missed by the SMA data due to its lower sensitivity. (ii) Several regions of particular interest (e.g., the western part of Sgr C) are not well sampled by existing SMA observations.

We attempted to combine the ALMA and SMA data by concatenating the visibility data and imaging them together. The resulting image is not improved compared to the ALMA image, e.g., the rms becomes higher, and the integrated fluxes of several spatially extended structures do not increase significantly (in the extreme case of Sgr C, the fluxes even decrease).

Therefore, we conclude that the imaging of diffuse structures does not benefit from the addition of the SMA data. This does not rule out the possibility that better shorter baseline data may help. A longer SMA observation, or an ACA observation that provides better sensitivity than the SMA, would be necessary to be combined with the ALMA data to recover spatially extended emission.

Meanwhile, we compared the integrated fluxes recovered by the CSO and ALMA, focusing on the SiO emission in Sgr C which is the most spatially extended in our data and thus the most affected by the missing flux. In a circle of 50'' diameter centered in between the two clusters in Sgr C, the ALMA/CSO SiO integrated fluxes are measured to be 120/200 Jy km s⁻¹. The ALMA data recover 60% of the flux observed by the CSO. We thus estimate an upper limit of 40% for the missed flux in our ALMA data. In Section 4.1.4, we will see that this does not affect our estimate of outflow properties, as the dominant uncertainty is the molecular abundance that is unconstrained by several orders of magnitude.

Typical spectra toward a chemically active core and a relatively quiescent region spatially offset from any cores but with line emission are shown in Figure 1. The former shows characteristic hot molecular core chemistry, while the latter represents regions that are likely influenced by parsec-scale shocks prevailing in the CMZ. Note that there exist even more chemically active cores (e.g., the two ultracompact (UC) H ii regions in Sgr C), with many more spectral lines detected, mostly from complex organic molecules. Here, we focus on the spatially extended spectral line emission detected outside of the hot cores or UC H ii regions and leave the discussion of the line-rich chemically active cores to a future paper.

Zoom In Zoom Out Reset image size

We identified spatially extended line species and plotted their integrated intensities in Figures 2-5. The line species include the CO isotopolog C¹⁸O, a group of potential shock tracers (SiO, SO, HNCO, CH₃OH), and several dense gas tracers that are sometimes found in outflows (H₂CO, H${}_{2}^{13}$CO, CH₃CN, c-C₃H₂, HC₃N). ¹³CO emission is more spatially extended than C¹⁸O and is not plotted. Three features are clearly seen: (i) linear structures spatially associated with dust emission are prominent, which may be outflow lobes (black boxes in the figures). (ii) Multiple lines tracing similar filamentary structures that are spatially offset from any dust emission, including some that are more typically confined to hot cores in environments outside of the CMZ (e.g., CH₃CN), whose nature is unclear. An example is marked by the blue arrow in Figure 2. (iii) Point-like SiO emission with large line widths (>20 km s⁻¹) found toward two H₂O masers that have known AGB star counterparts (magenta crosses in Figures 2 and 4; see Lu et al. 2019b), with no associated dust emission within a radius of 0.1 pc, which probably originated from the atmosphere of AGB stars (González Delgado et al. 2003). Here we focus on the first feature, potential outflows, while leaving the discussion of the other features to future papers.

The two CO isotopolog lines, ¹³CO and C¹⁸O, present strong absorption at velocities of −55, −30, and −5 km s⁻¹ against strong continuum emission (see the inset in Figure 1). These are consistent with the absorption features seen in other line observations toward the Galactic Center (e.g., Jones et al. 2012) and are attributed to foreground gas in spiral arms along the line of sight. In particular, the absorption at −55 km s⁻¹ is close to the cloud velocity of Sgr C, which complicates the interpretation of the two lines in Sgr C. Many of the lines, including H₂CO, CH₃OH, SiO, and the two CO isotopologs themselves, also present absorption features at a few km s⁻¹ blueshifted with respect to the cloud velocity toward continuum emission peaks (e.g., the absorption at ∼12 km s⁻¹ in the inset of Figure 1). These features are likely owing to a combination of missing flux as a result of interferometer observations and self-absorption when the lines become optically thick.

Signatures of protostellar outflows have been detected toward Sgr B2(M) and (N), the two high-mass protoclusters in the Sgr B2 complex (Qin et al. 2008; Higuchi et al. 2015; Bonfand et al. 2017). Outside of Sgr B2, detecting protostellar outflows has been challenging in the CMZ because of the lack of resolution to spatially resolve outflows and the prevalence of broad-line-width gas produced by phenomena other than outflows (Henshaw et al. 2019; Sormani et al. 2019). Previous SMA observations have detected widespread emission of potential shock tracers (e.g., SiO, SO, CH₃OH) at 0.2 pc scales in CMZ clouds (Kauffmann et al. 2013; Lu et al. 2017; Battersby et al. 2020). However, limited by the angular resolution and the imaging sensitivity, it was unclear whether the emission seen by the SMA is owing to protostellar outflows or parsec-scale shocks prevailing in the CMZ. ALMA observations, with high resolution and high sensitivity, have only recently started to detect outflows in the CMZ, e.g., in the G0.253+0.025 cloud (Walker et al. 2021).

Our high-angular-resolution ALMA observations resolved substructures of 2000 au scale in the molecular line emission, enabling us to search for protostellar outflows. The integrated intensity maps in Figures 2–5 already reveal collimated emission spatially associated with cores, indicating the existence of outflows.

We then examined potential shock tracers detected by ALMA, including SiO, SO, H₂CO, HNCO, HC₃N, and CH₃OH. All these tracers have been previously found to be enhanced by at least one order of magnitude in shocked regions in protostellar outflows (e.g., Bachiller & Pérez Gutiérrez 1997). We applied the following criteria to identify outflows:

• (i)  
  We used the H₂O masers from Lu et al. (2019b) as a guide to search for associated shock tracer emission. First, we made integrated intensity maps of SiO across the full velocity range and searched for linear structures spatially associated with the masers. If linear emission is found, then, we made integrated intensity maps of blue- and redshifted components based on the V_lsr of the cloud (see Table 1) and checked whether the linear structures show symmetric blue- and redshifted emission with respect to the masers. Finally, we determined the systematic velocity of each individual outflow-driving source by using dense gas tracers in the ALMA data toward the maser position (CH₃CN, HC₃N, CH₃OH, or C¹⁸O, in decreasing order of preference; C¹⁸O was used only once, toward the 20 km s⁻¹ cloud-F #1) and remade integrated intensity maps of blue- and redshifted components based on the individual V_lsr. If the shock tracer emission exhibits blue- and redshifted components with respect to the systematic velocity at opposite positions to the maser position, it is considered an outflow.
• (ii)  
  In cases where H₂O masers are not present, we checked the emission of the six tracers around the 2000 au scale cores from Paper I following the same procedures to search for blue- and/or redshifted line emission spatially offset from the cores. We require the blue- and/or redshifted features to be seen in at least two shock tracers, including the canonical shock tracer SiO, plus any of the five supplemental tracers.

We identified 43 outflows and marked regions where they are detected with boxes in Figures 2–5. The zoomed-in views are in Figures 6–22, in which we plot the red-/blueshifted shock tracer emission with respect to the systematic velocity and highlight individual outflows with arrows. The position–velocity diagrams of the 43 outflows made from the SiO line are displayed in Figure 23. The numbers of outflows identified in the individual clouds are listed in Table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In several cases, lobes from different outflows spatially overlap with each other (e.g., outflows #5–#9 in region C of the 20 km s⁻¹ cloud; see Figure 8), but we were able to separate them apart unambiguously based on velocities (see Figure 23). Most of the other outflows are easily distinguished spatially from nearby outflows and the diffuse emission. All of these outflows are considered as "highly likely." However, there exist cases where the outflows cannot be robustly separated from other outflows or the diffuse emission, either spatially or kinematically. One example is the two blueshifted lobes in region B of the 20 km s⁻¹ cloud, where the lobes overlap in both projected locations and velocities (see Figures 7 and 23). Following the definitions in Li et al. (2020), we classified these ambiguous identifications as "candidates." The classifications are noted with asterisks in Table 3 and Figures 6–23. Among the 43 outflows, 37 are highly likely and 6 are candidates.

Zoom In Zoom Out Reset image size

We stress that this visual identification is likely to be incomplete. Potential outflows could have been missed if they cannot be distinguished from the background emission or other outflows, or if their emission is too weak. The actual (in)completeness, however, is difficult to quantify, as the identification is based on visual inspection and is subjective in nature. Recent ALMA surveys toward infrared-dark clouds in the Galactic disk using CO lines as the primary outflow tracer yield detection rates of 14%–22% (e.g., 62 out of 280 cores in Kong et al. 2019 and 41 out of 301 cores in Li et al. 2020 are identified with outflows). The detection rate using SiO as the primary tracer in this paper is 4%–7% (Table 1), which is much lower. It is infeasible to directly compare, e.g., the outflow mass sensitivities of previous surveys and ours, considering that the observations use different lines as outflow tracers and assume different abundances. Assuming that the observations are sufficiently sensitive to detect all existing outflows, the lower detection rate in our sample may suggest that we have missed a substantial number of outflows that are not traced by SiO, or may reflect the variation of outflow occurrence rates along the evolutionary stages.

Meanwhile, we also note that due to the complicated environment in the CMZ (e.g., the widespread shock tracer emission; Martín-Pintado et al. 1997) and possible contamination from the foreground, it is likely that false-positive identifications exist in our sample if such large-scale shock tracer emission accidentally lies upon cores. But such accidental spatial coincidence should be rare, as the velocities of the identified outflows and cores have a continuous overlap (Figure 23). Note that the cores associated with the outflows are all likely in the CMZ, given that the velocities of their line emission (e.g., C¹⁸O; see Figure 1 inset) are consistent with the overall velocity field in the CMZ (e.g., Henshaw et al. 2016), while spiral arm clouds along the line of sight are mostly seen as absorption in ¹³CO and C¹⁸O (Figure 1 inset), indicating that the overlapping spiral arm clouds consist of low-density gas and hence are unlikely to have dense cores.

SiO and SO seem to be the best outflow tracers among the six molecules. Their emission is usually well separated from the background and is usually collimated as expected for outflows. CH₃OH, H₂CO, and HNCO often suffer from contamination from the background or foreground emission, and thus are not tracing the outflows as well as SiO and SO. HC₃N traces both cores and outflows. Its emission is weaker than the other molecules, and therefore it is not an optimized outflow tracer either. As pointed out by several previous studies, CH₃OH, H₂CO, HC₃N, and HNCO may be released into the gas phase by slow (≲20 km s⁻¹) shocks that evaporate ice mantles of dust or be produced by gas-phase reactions in postshock regions, and therefore probe the widespread low-velocity shocks in the CMZ (Lu et al. 2017; Tanaka et al. 2018; Taniguchi et al. 2018). SiO and SO, on the other hand, may be released from the dust by sputtering of the grain core, and thus probe fast (≳20 km s⁻¹) shocks induced by the outflows better.

In this section, we calculate the column densities of the six molecules detected toward the identified outflows (Section 4.1.1), introduce the method to estimate the molecular abundances in the outflows (Section 4.1.2), and estimate the outflow masses, and where possible, the outflow energetics (Section 4.1.3). The uncertainties involved in the estimate of column densities, abundances, and masses are discussed in Section 4.1.4. We need an order-of-magnitude estimate on these parameters to discuss implications for astrochemistry and star formation in the following sections, so the unavoidable significant uncertainties in these results (one to two orders of magnitude, as detailed in Section 4.1.4) are acceptable.

4.1.1. Column Densities of the Shock Tracers

4.1.2. Molecular Abundances

4.1.3. Masses and Energetics of the Outflows

After the molecular column densities and abundances are derived, the outflow masses are estimated by assuming uniform abundances within each outflow, using the calcu toolkit. The results are listed in Table 3 and plotted in Figure 24. For the same outflow lobe, multiple shock tracers could be detected, in which case we derive more than one outflow mass. All of these masses are deemed to be worth reporting, as the different molecules may trace different components of the same outflow with different chemical environments or excitation conditions. Nevertheless, the masses of the same outflow are consistent within an order of magnitude as demonstrated in Figure 24.

Zoom In Zoom Out Reset image size

In a few cases where the outflow emission can be unambiguously separated from contaminations and the lobes are well collimated, we are able to measure the projected length and estimate a dynamical timescale. Then, the energetics of the outflows, e.g., the outflow rate and the outflow mechanical force, can be estimated following the procedures in Li et al. (2020). One example of such a well-defined outflow is the blueshifted lobe of the Sgr C region F #2 outflow. With a projected angular scale of 20'' (∼0.8 pc), and an outflow terminal velocity (the maximum velocity difference between the outflow and the core, as noted by the velocity range listed in Table 3) of 33.7 km s⁻¹, the dynamical timescale is $3.2\times {10}^{4}/\cos (\theta$) yr where θ is the inclination angle of the outflow lobe with respect to the plane of the sky. The outflow mass rate is then ∼$6\times {10}^{-4}\cos (\theta$) M_⊙ yr⁻¹. If the inclination angle is not too large (θ < 80°), the outflow mass rate is ≳10⁻⁴ M_⊙ yr⁻¹, which is usually found toward outflows around high-mass young stellar objects (Maud et al. 2015). It is also possible to estimate the accretion rate, with the same assumptions in Li et al. (2020), e.g., a wind speed of 500 km s⁻¹ from the disk and a ratio between the accretion rate and the mass ejection rate of 3, which leads to a value of $2.5\times {10}^{-5}\cos (\theta$) M_⊙ yr⁻¹. However, there are significant uncertainties in the outflow masses (see next section) and the true physical scale of the outflow lobes (contamination, potentially missed weak emission, inclination angle) on top of the unconstrained assumptions made in the calculation of the accretion rate (Li et al. 2020). We expect the uncertainty in the estimated outflow mass rate and accretion rate to be potentially three orders of magnitude or even greater, which is comparable to the dynamical range of our data (Figure 24). Therefore, we will not discuss these results further.

4.1.4. Uncertainties in the Estimated Parameters

We compare the relative abundances of the six shock tracers and investigate the shock chemistry in the outflows. This is the first spatially resolved astrochemical study toward outflows in the CMZ, and one of the very few such studies even including works toward Galactic disk targets.

The abundances of the six shock tracers are presented in Figure 25. For each molecule, we plot the median of its abundances in all the outflows as a horizontal orange line.

Zoom In Zoom Out Reset image size

To understand the relative abundances of the six molecules, we compare our result with similar studies toward nearby star-forming clouds. However, we note that spatially resolved observations toward outflows that include all six molecules, even for popular targets such as the Orion molecular cloud, are rare. For example, we are not able to find a paper that reports SiO, H₂CO, or HC₃N column densities or abundances in the explosive outflow around Orion KL, even though results of SO, CH₃OH, and HNCO are available (Feng et al. 2015). In the end, we are able to find only two representative outflows in nearby clouds: L1157, a well-studied, prototypical low-mass outflow around a low-mass protostar (e.g., Bachiller & Pérez Gutiérrez 1997; Rodríguez-Fernández et al. 2010; Podio et al. 2017; Holdship et al. 2019), and NGC 7538S, a prototypical massive outflow around a high-mass protostar (e.g., Naranjo-Romero et al. 2012; Feng et al. 2016).

We take the abundances of the six molecules toward L1157-B1 and B2 (two reference positions in the blueshifted lobe) from Bachiller & Pérez Gutiérrez (1997) and Rodríguez-Fernández et al. (2010). For NGC 7538S, we take the results from JetS, a reference position on the redshifted side of the protostar, from Feng et al. (2016), except for SiO, which is not observed by the authors. We instead obtain the SiO abundance at the reference position using data from Naranjo-Romero et al. (2012; L. Zapata 2021, private communication). The JetN position in Feng et al. (2016) does not show HC₃N emission, and therefore only an upper limit of its abundance can be detected, which prevents an appropriate comparison to our results as we use HC₃N to infer the abundances of the other molecules. All results from the above publications have assumed LTE conditions and optically thin line emission. The abundances are plotted in Figure 25(a).

In addition, we note that the abundances toward L1157-B1/B2 are estimated based on CO lines, which may become optically thick and therefore result in overestimated abundances for other molecules (Bachiller & Pérez Gutiérrez 1997). This may explain the systematically higher abundances for all six molecules in L1157-B1/B2 than the other targets in Figure 25(a). To eliminate such biases, we normalize the abundances with respect to that of HC₃N in the outflows and plot the relative abundances of the six molecules in Figure 25(b).

By comparing the two samples in Figure 25(b), the CMZ clouds versus the two nearby clouds, we do not find clear evidence of a difference between the relative abundances of the six shock tracers in the outflows. The relative abundances of the two nearby clouds usually fall within an order of magnitude apart from the medians of our CMZ outflow sample. However, given the significant uncertainty of the abundances and a limited sample from nearby clouds, it is premature to conclude any consistency between the two samples.

Protostellar outflows are ubiquitously detected in star-forming regions, suggesting active gas accretion around protostars (e.g., Shang et al. 2007; Bally 2016). Here we investigate star formation and chemistry in the four massive clouds in the CMZ based on our observations of the outflows.

The first implication, obviously, is that protostellar accretion disks ubiquitously exist in these clouds, as protostellar outflows are supposedly driven by disks (Shang et al. 2007). Direct observational evidence of protostellar accretion disks in the CMZ has been limited to the hot cores in the Sgr B2 cloud (e.g., Hollis et al. 2003; Higuchi et al. 2015), and even for these cases, the evidence is ambiguous given the complicated kinematic environments in Sgr B2. More recently, Walker et al. (2021) reported detection of protostellar outflows in G0.253+0.025 in the CMZ based on ALMA observations. Our finding of a large population of outflows (except in the 50 km s⁻¹ cloud, which is likely in a more evolved phase of star formation; Mills et al. 2011; Lu et al. 2019b), suggests that active accretion is ongoing around protostars in these CMZ clouds.

The second implication concerns the evolutionary phases of star formation in these clouds based on the statistics of the cores with or without star formation signatures. In Paper I, we identify 834 cores at 2000 au scales in the three CMZ clouds. Among them, only 43 are found to be associated with outflows. The remaining 791 cores are not associated with other signatures of star formation (masers, H ii regions) either, and therefore are candidates of starless cores (gravitationally bound and prestellar, or simply unbound). However, as mentioned in Section 3.2, the outflow sample is very likely to be incomplete because of the subjectivity of the identification, thus potential outflows, even those with sufficiently strong emission, may have been missed. In addition, deeper observations may reveal more signatures of star formation such as weaker outflows or new masers. Therefore, the fraction of protostellar and starless cores are highly uncertain. For individual clouds, the fractions of cores associated with outflows range from 0.04 to 0.07 (Table 1), although this is unlikely to suggest any evolutionary trend among the three clouds given the small numbers of the outflow detections and the potential incompleteness of the outflow sample.

If we base our analysis on the current observations, i.e., 4%–7% of the cores identified in Paper I are protostellar (Table 1), then we may put a constraint on the evolutionary phase of the clouds. The timescale needed to enter the protostellar phase is of the order 1–2 Myr for both low-mass and high-mass star-forming cores (Enoch et al. 2008; Könyves et al. 2015; Battersby et al. 2017). This timescale is similar to the proposed lifetime of molecular clouds in the CMZ (Jeffreson et al. 2018; Barnes et al. 2020). Assuming that all the cores we detected will eventually evolve into the protostellar phase in a timescale of 1–2 Myr, the small fraction of the currently identified protostellar cores may suggest an age of star formation in these clouds as short as ∼0.05–0.1 Myr. In such case, star formation may have started only recently, if the clouds just condensed out of a more diffuse state, possibly driven by tidal compression during their arrival in the CMZ or on their orbit around the Galactic Center (Longmore et al. 2013b; Kruijssen et al. 2015, 2019), or by the impact of adjacent expanding H ii regions (Kendrew et al. 2013; Barnes et al. 2020). Again, we stress that this estimate depends on the (in)completeness of the outflow sample, and the age of star formation in these clouds is likely longer as the fraction of protostellar cores is potentially higher.

The third implication is related to the result presented in Figure 24, where we find outflows from both high-mass (>100 M_⊙) and low-mass (<5 M_⊙) cores. Here the core masses have a strong dependence on the unconstrained dust temperature and may be overestimated by a factor of 3, as demonstrated in Paper I. Several of the high-mass cores are known to be forming high-mass protostars, with UC H ii regions and Class ii CH₃OH masers (Lu et al. 2019a, 2019b). The low-mass cores, on the other hand, are only capable of forming low-mass stars, with the current mass even assuming a high star formation efficiency of 50%. Meanwhile, the majority of the outflow masses lie in the range of 1–10 M_⊙, albeit with a large uncertainty of a factor of 70. This mass range is characteristically found around high-mass protostars (Zhang et al. 2005; Lu et al. 2018). Some of the outflows have lower masses of <1 M_⊙, which are typical for low-mass star-forming regions (Arce et al. 2010). Therefore, considering the mass ranges of the cores and the outflows, the detected outflows likely trace a mix of high-mass and low-mass star formation. Simultaneous low-mass and high-mass star formation has been observed ubiquitously in massive clouds in the Galactic disk (e.g., Cyganowski et al. 2017; Pillai et al. 2019; Sanhueza et al. 2019), which we now confirm to take place in the CMZ as well.

The last implication, as discussed in Section 4.2, is about the shock chemistry in the outflows. Given the large uncertainties involved in the abundances (at least one order of magnitude), we are not able to conclude consistency between the shock chemistry in the CMZ clouds and in nearby analogs, but we do not find evidence of difference either. This is in contrast to the situation on the cloud scale of a few parsec, where the chemistry in the CMZ is distinctly different from that in nearby clouds, e.g., an anomalous enhancement of complex organic molecules and shock tracers as compared to those in nearby clouds, suggesting the presence of widespread low-velocity shocks (Martín-Pintado et al. 1997; Requena-Torres et al. 2006; Menten et al. 2009). Several previous studies have pointed out that at the sub 0.1 pc scale, physical processes such as gas fragmentation and turbulent line widths in the CMZ and in nearby regions may start to converge (e.g., Kauffmann et al. 2017; Lu et al. 2019b; Paper I; Walker et al. 2021) despite distinct properties on larger scales. Our results set the first step toward a similar comparison of the shock chemistry in protostellar outflows in the CMZ and in nearby clouds. Multitransition spectral line observations toward the CMZ outflows that enable a more robust estimate of column densities and abundances, and a larger sample of resolved astrochemical studies toward outflows in nearby clouds, will help clarify whether the shock chemistry in the two environments are consistent or not.