The Extraordinary Outburst in the Massive Protostellar System NGC 6334I-MM1: Flaring of the Water Masers in a North–South Bipolar Outflow Driven by MM1B

The VLA 22.235 GHz water maser spectral line data presented in this paper arise from three epochs: 2011.7, 2017.0, and 2017.8, as described in Table 1. The 2011.7 epoch data were also presented in Brogan et al. (2016). The K-band 1.3 cm continuum from the two newer epochs will be the subject of a forthcoming paper. The C-band 5 cm continuum data from epoch 2016.9 used in this paper were originally described in Hunter et al. (2018). The two new epochs of 1.3 cm data were calibrated using the VLA pipeline¹¹ in the Common Astronomy Software Applications (CASA) package. The VLA pipeline applies Hanning smoothing to the data which reduces ringing from strong spectral features and coarsens the spectral resolution to twice the observed channel width. A strong maser feature was used to iteratively self-calibrate the spectral line data for each dataset. Stokes I cubes of the maser emission were made with a channel spacing (and effective spectral resolution) of 0.25 km s⁻¹. In order to better match the 2017 H₂O data, we re-imaged the H₂O maser data from 2011 September in order to create cubes sampled with narrower channels (0.25 km s⁻¹ spacing, with 0.40 km s⁻¹ effective spectral resolution) than our previous analysis in Brogan et al. (2016). The resulting noise in a typical 2011.7 epoch signal-free channel is ∼20 mJy beam⁻¹. The angular resolution achieved for each epoch is given in Table 1. The full width half maximum (FWHM) of the VLA primary beam at 22.235 GHz is 2'. Primary beam correction was applied to correct the extracted maser flux densities.

Table 1.  VLA Observing Parameters

Parameter	Epoch 1	Epoch 2	Epoch 3
Project code	493_1	16B-402	17B-258
Observing date	2011 Sep 09	2017 Jan 08	2017 Oct 08
Mean epoch	2011.7	2017.0	2017.8
Antenna configuration	A	A	B
Time on source (minutes)	43	41	49
FWHM primary beam	2'	2'	2'
Polarization products	Dual circular	Dual circular	Dual circular
Gain calibrator	J1717-3342	J1717-3342	J1717-3342
Bandpass calibrator	J1924-2914	J1924-2914	J1924-2914
Flux calibrator	3C286	3C286	3C286
Bandwidth (km s−1)	222	215	215
Native channel width (km s−1)a	0.21	0.105	0.105
Spectral resolution (km s−1)	0.42	0.21	0.21
Cube channel width (km s−1)	0.25	0.25	0.25
Synthesized beam ('' × '' (P.A))	041 × 022 (−24)	030 × 023 (−18)	079 × 022 (−6)
rms noise: min, median, max (mJy beam−1)b	16, 24, 680	8, 12, 1300	4, 6, 990
Relative position uncertainty [R.A. (''), Decl. ('')]c	0.019, 0.034	0.020, 0.025	0.019, 0.066

Notes.

^aThe native spectral resolution is two times larger than the native channel width due to the application of Hanning smoothing before calibration. ^bThe rms noise varies significantly with channel number due to dynamic range limitations; the first number (the minimum) is the most representative value for the noise values measured in empty channels. ^cThe relative position uncertainty in R.A. and decl. for a 6σ maser detection threshold.

Download table as:  ASCIITypeset image

Due to the high peak signal strength of the water maser emission and the fact that it occupies much of the bandwidth of the high-resolution spectral window employed, a few channels of the maser data (from −71 to −59 km s⁻¹) are affected by spectral artifacts due to phase serialization in the correlator and usage of the default values of the local oscillator modulation frequencies (${f}_{\mathrm{shift}}$, Sault & Sowinski 2013). Fortunately, only the two strongest maser positions (near CM2) show contamination, which takes the form of −36 dB ghosts of the peak signals in the −17 to −5 km s⁻¹ range of channels, i.e., shifted by one-quarter of the spectral window bandwidth (54 km s⁻¹). We have removed these false components from our analysis.

The ALMA 1 mm observations presented in this paper are from project 2015.A.00022.T and have been described in earlier publications, which showed the 1 mm dust continuum image, as well as the integrated intensity image from a methanol transition (Hunter et al. 2018), and spectra that exhibited various emission lines of methoxymethanol (McGuire et al. 2017). Due to the copious hot core line emission emanating from MM1 and MM2, extracting a list of relatively line-free continuum channels to perform UV-continuum subtraction for this source is challenging.

To this end, we developed a technique of identifying relatively line-free channels that first uses the corrected sigma clipping method (c-SCM) approach of STATCONT (Sánchez-Monge et al. 2018) to define the continuum level in each spectral window, then choose a list of channels that best match that level over a representative ensemble of individual spatial pixels. These channels are then used to perform UV-continuum subtraction in CASA. In more detail, we first extract spectra from a sample of pixels (∼10–20) in the dirty line+continuum cube that represent the range of emission/absorption properties present in the field of view (spectral line peak positions, continuum peaks, and outflows, for example). For each spectrum, the c-SCM method is employed to find the sigma-clip continuum level (SCL) for that pixel using parameter α = 3. The SCL is then corrected for noise by $-0.5{\sigma }_{\mathrm{SCM}}$ (case D of STATCONT). Second, we compute a spectrum of the absolute deviation from the SCL for each pixel. We then form an aggregate deviation spectrum (ADS) from the maximum value per channel over all deviation spectra. Finally, we define the line-free channels as those whose ADS value is below a specified level, which is adjusted manually until the line contamination for all the representative spectra is low, while retaining sufficient line-free channels spanning the window to enable successful UV-continuum subtraction. We note that while UV-continuum subtraction is sensitive to sampling across the whole spectral window, it is relatively insensitive to the total fraction of bandwidth included. For NGC 6334I, the continuum bandwidth identified using this method is typically 10%–20% of the total observed spectral bandwidth. This method ensures that significant line emission is avoided across the range of spatio-spectral properties present, while still permitting UV-continuum subtraction, a method which provides distinct advantages when one needs to image lines in the presence of strong continuum emission (Rupen 1999).

For this paper, we also present for the first time images of the CS (6–5) spectral line transition at 293.9120865 GHz (Gottlieb et al. 2003) that were imaged with a spectral resolution of 2 km s⁻¹. These data are sensitive to smooth spatial structures ≲6'' in size. The full width at half power of the primary beam of the ALMA 12 m antennas at this frequency is only 20''; primary beam correction was applied to the images presented here. The synthesized beam of the 1 mm continuum image is 024 × 017 (−84°), while the synthesized beam of the CS (6–5) image is 028 × 021 at −84°. The sensitivity of the 1 mm continuum is 0.6 mJy beam⁻¹, while the sensitivity of the CS (6–5) channels free of significant emission is 3 mJy beam⁻¹.

We fit each channel of the water maser cube that had significant emission (≥6σ) with an appropriate number of Gaussian point sources, using the pixel position of each emission peak as the initial guess for the fitted parameters. A few weak sources that could be attributed to imaging artifacts in the dynamic-range limited channels were discarded despite otherwise meeting the detection threshold. A few of the most complicated channels exhibiting many closely spaced components required a manual iterative procedure to fix the decl. position of one or two components in order for the simultaneous fit for all components to converge. The fitted positions and flux densities of the water masers are given in Tables 2–4 for the 2011.7, 2017.0 and 2017.8 data, respectively. The fitted strength of the emission was corrected for primary beam attenuation.

The locations of the fitted maser positions are shown in Figure 1. To summarize the different maser locations in a compact format, we have identified maser "groups" toward the centimeter and millimeter continuum sources that had water masers detected toward them: MM1, CM2, MM3-UCHII, MM4, MM7, and a few outliers called NWflow and I-South. When the water masers within these groups are concentrated into spatially distinct "associations" they are further labeled with the suffix -W1, -W2, etc. for ease of referral. The naming of these associations parallels that used for the 6.7 GHz methanol and 6.035 GHz OH masers reported in Hunter et al. (2018) toward this source. The assignment of each fitted maser spot to an association is provided in columns in Tables 2–4. These data show that the masers have flared significantly toward the 5 cm synchrotron source CM2, and that the broad range of maser velocity components observed previously persists in the current epochs (see Figure 1, as well as Forster & Caswell 1999; Breen et al. 2010; Brogan et al. 2016; Titmarsh et al. 2016). The change in the total maser emission and the changes in the individual associations are summarized by the integrated spectra in Figure 2 and in the summary Tables 5 and 6. We next discuss the individual groups (and associations within them) in more detail.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 5.  Brightness Summary of 22 GHz Water Maser Associations

Association	Mean position (J2000)a		Integrated flux			Integrated Flux Ratios		Highest Flux Densityb
			(Jy km s−1)					(Jy)
	R.A.	Decl.	2011.7	2017.0	2017.8	[2017.0/2011.7]	[2017.8/2011.7]	2011.7	2017.0	2017.8
I-South	17:20:53.182	−35:47:29.10	⋯	2.57	0.881	New	⋯	⋯	1.66	0.856
CM2-W1	17:20:53.380	−35:46:55.73	1790	11100	12000	6.2	6.7	398	3920	4230
CM2-W2	17:20:53.431	−35:46:55.42	55.4	399	39.0	7.2	0.70	34.8	403	42.0
CM2 Total	⋯	⋯	1840	11500	12000	6.3	6.5	⋯	⋯	⋯
MM1-W1	17:20:53.416	−35:46:58.03	94.5	7.84	5.79	0.083	0.061	26.9	2.77	2.27
MM1-W2	17:20:53.410	−35:46:57.73	53.4	1.25	⋯	0.023	vanished	19.4	1.18	⋯
MM1-W3	17:20:53.422	−35:46:57.44	63.7	22.5	12.3	0.35	0.19	25.3	5.33	2.56
MM1-W4	17:20:53.447	−35:46:57.28	⋯	⋯	1.60	⋯	new	⋯	⋯	1.28
MM1 Total	⋯	⋯	212	31.6	19.7	0.15	0.093	⋯	⋯	⋯
MM4	17:20:53.104	−35:47:03.36	4.73	⋯	0.709	Vanished	0.15	2.07	⋯	0.295
MM7	17:20:53.645	−35:46:54.85	⋯	⋯	3.81	⋯	new	⋯	⋯	4.51
NWflow	17:20:53.107	−35:46:54.68	⋯	⋯	0.292	⋯	new	⋯	⋯	0.235
UCHII-W1	17:20:53.437	−35:47:01.04	701	755	366	1.1	0.52	66.5	237	83.3
UCHII-W2	17:20:53.451	−35:47:00.61	126	49.4	135	0.39	1.1	80.2	30.1	68.6
UCHII-W3	17:20:53.452	−35:46:59.81	⋯	109	60.5	New	⋯	⋯	46.5	60.5
UCHII-W4	17:20:53.405	−35:46:59.54	⋯	7.68	10.4	New	⋯	⋯	15.9	19.1
UCHII-W5	17:20:53.806	−35:47:01.21	⋯	⋯	0.465	⋯	new	⋯	⋯	0.461
UCHII Total	⋯	⋯	827	921	572	1.1	0.69	⋯	⋯	⋯
Grand Total	⋯	⋯	2880	12500	12600	4.3	4.4	⋯	⋯	⋯

Notes.

^aThis is the flux-weighted centroid of all spots in the epoch of first appearance. The differences in centroids between epochs are less than 008 with the following exceptions: UCHII-W1 and W2 which show bulk southward motion described in Section 4.4; MM1-W1 which shows a large drop in total emission described in Section 4.3; and MM4 which shows a new spatial component in 2017.8 (Figure 6). ^bThese are the highest values for the respective components from the Flux Density column in Tables 2–4.

Download table as:  ASCIITypeset image

3.1.1. CM2

In all three VLA epochs presented here, the masers toward and close to the 5 cm point source CM2 are the strongest in the region (Figures 1, 3; CM2-W1 and CM2-W2). Between mid-2011 and 2017, they exhibited a large increase in peak flux density, with the peak increasing from <156 Jy in 2011.4 (Titmarsh et al. 2016) and 398 Jy in 2011.7 (Brogan et al. 2016) to 3910 Jy in 2017.0 and 4208 Jy in 2017.8. The peak intensity in 2011.7 is similar to the value of 246 Jy observed at a comparable velocity in the first VLA observations of this region with 1.32 km s⁻¹ channels (epoch 1984.5, Forster & Caswell 1999), which further extends the baseline of the pre-outburst emission from CM2 to three decades. Figure 3 shows a close-up view of the CM2 water masers (with fixed symbol size) for the three epochs. This figure demonstrates that especially during the 2017.0 and 2017.8 epochs, the water masers coincident with CM2-W1 trace out a distinctive bow shock morphology pointing back toward MM1B, with the most blueshifted masers forming the apex of the bow shock. The nature of this bow shock is discussed further in Section 4.2.

Zoom In Zoom Out Reset image size

3.1.2. MM1

In MM1, we label the initial (epoch 2011.7) associations of masers as follows: the association toward and south of MM1B as MM1-W1, the association between MM1B and MM1D as MM1-W2, and the association north of MM1D as MM1-W3. Comparison of the 2011.7 and 2017.0 epochs reveals that all the 2011.7 maser sites have significantly weakened. Most notably, MM1-W2 disappears entirely by the 2017.8 epoch, see Figure 4. In contrast, new blueshifted features have appeared within MM1-W3 (to the northwest of MM1D) in the 2017.0 epoch, persisting into 2017.8. Also, a new association of redshifted features has appeared northeast and east of MM1D in epoch 2017.8, which we label MM1-W4. The peak feature of MM1-W4 is 1.28 Jy at 8.25 km s⁻¹, which would have been a 100σ detection had it been present in that channel in the 2017.0 data. Interestingly, the MM1-W4 masers coincide with the slight eastward extension of the 5 cm emission that lies just north of MM1D.

Zoom In Zoom Out Reset image size

3.1.3. MM3-UCHII Region

As shown in Figure 5, in the first epoch (2011.7) only two maser associations were detected toward MM3-UCHII: UCHII-W1 and UCHII-W2, each forming a roughly linear structure, in roughly perpendicular directions. These associations persisted to the 2017.0 and 2017.8 epochs, though in the 2017.8 epoch they had lost much of their ordered morphology. The two later epochs also show two new maser associations further north but still coincident with the UCHII region: UCHII-W3 and UCHII-W4. Of particular note is the fact that the UCHII-W1 and UCHII-W2 associations appear to have undergone bulk motion toward the south–southeast. This is demonstrated by the white boxes in Figure 5, which encompass these associations in the 2011.7 epoch; in contrast, the cyan boxes encompass these associations in the 2017.0 epoch data. The offset between the colored boxes is the proposed bulk motion. The more chaotic morphology of these associations by 2017.8 precludes a definitive judgment, but their locations in this epoch are also in general agreement with the bulk motion interpretation. This idea is explored further in Section 4.4.

Zoom In Zoom Out Reset image size

3.1.4. MM4

In both the 2011.7 and 2017.8 epochs, weak water masers are detected near the MM4 continuum peak (Figure 6). In the more recent epoch, the masers form two linear features that are roughly perpendicular to the bipolar outflow direction described in Section 3.3. The two linear features are offset from each other in R.A. by ∼100 au; the more easterly feature is slightly redshifted, while the western feature is blueshifted. Only the eastern feature was detected in the 2011.7 epoch, and its masers were significantly stronger then, by a factor of almost seven compared to the 2017.8 epoch (see Table 5). The nondetection of masers in the MM4 group in the 2017.0 epoch must be due to variability, as the strongest 2011.7 epoch maser (0.72 Jy beam⁻¹ at +2.25 km s⁻¹) would have been a 21.6σ detection in the 2017.0 epoch. In 2017.8, masers in the eastern feature are detected at 6–11 σ.

Zoom In Zoom Out Reset image size

3.1.5. MM7

In the 2017.8 epoch, water masers are detected for the first time toward the millimeter core MM7, with a peak intensity of 4.51 Jy beam⁻¹ at −10.25 km s⁻¹ (Figure 7). The 3σ upper limits in this channel for the earlier epochs are 0.54 and 0.87 Jy beam⁻¹, respectively. Considering all channels with detections in 2017.8, the highest intensity increase factors relative to epochs 2011.7 and 2017.0 are >9.3 (at −17.50 km s⁻¹) and >11.9 (at −17.75 km s⁻¹), respectively.

Zoom In Zoom Out Reset image size

3.1.6. NWflow

3.1.7. NGC 6334I-South

A few CH₃OH maser lines fall within our coarse spectral resolution windows, such as the Class II 2(1)–3(0) E transition at 19.9674 GHz (Menten & Batrla 1989; Ellingsen et al. 2004). We imaged the corresponding channel of our coarse resolution data and found that the positions of the emission, toward MM2 and the MM3-UCHII region, agree with the VLA positions reported by Brogan et al. (2016). No other new positions were detected.

The CS (6–5) transition is well suited to trace moderately dense, warm gas, with an upper state energy ${E}_{\mathrm{up}}=49.4$ K and a critical density of ${n}_{\mathrm{crit}}=9\times {10}^{6}$ cm⁻³ using

$$\begin{eqnarray}&&{n}_{\mathrm{cr}}=\displaystyle \frac{{A}_{\mathrm{ul}}}{{\gamma }_{\mathrm{ul}}},\end{eqnarray} \tag{ 1 }$$

where A_ul is the Einstein A coefficient of the transition (5.23 × 10⁻⁴ s⁻¹) and ${\gamma }_{\mathrm{ul}}$ is the collisional coefficient at 200 K ($6.02\,\times \,{10}^{-11}$ cm³ s⁻¹, Schöier et al. 2005; Lique et al. 2006). Figure 8(a) shows the integrated intensity of CS (6–5) emission toward NGC 6334I, including the velocity ranges of −88 to −78 km s⁻¹, −40 to −12 km s⁻¹, and −4 to +40 km s⁻¹ (these ranges exclude channels near the systemic velocity of ∼−7 km s⁻¹, and channels with significant confusing hot core line emission toward MM1). Figure 8(b) shows the same integrated intensity velocity ranges, but color-coded red and blue. From these images it is clear that several distinct collimated outflow structures are detected in the CS (6–5) transition.

Zoom In Zoom Out Reset image size

On the largest size scale, the inner portion of the large-scale northeast–southwest outflow, well known from previous single-dish studies (McCutcheon et al. 2000; Leurini et al. 2006; Qiu et al. 2011), is readily apparent in these data, though its full projected extent (∼50'') lies well beyond the primary beam of these single pointing ALMA data (FWHM 20''). The origin of this outflow has long been debated because at lower angular resolution it could reasonably emanate from either a protostar in MM1 or MM2 (e.g., Beuther et al. 2008). These new high-resolution and fidelity ALMA data reveal that indeed the majority of emission arises from one of the MM1 protostars, especially the redshifted northeast lobe, but there is also a one-sided blueshifted component from MM2 that extends to the southwest coincident with the larger blueshifted southwest lobe. Qiu et al. (2011) estimate a dynamical age for the large northeast–southwest outflow of 2.6 × 10³ years.

A number of other smaller, highly collimated outflows are evident from these data for the first time. The brightest emanates from MM1B in a roughly north–south direction (PA = −6°) with a projected linear extent of ∼62 (8060 au), and velocity extent from −40 to +18 km s⁻¹. This outflow will hereafter be called the MM1B north–south outflow. The majority of water masers in the NGC 6334I region are coincident with bright knots of CS (6–5) emission in this outflow (see Figure 8(a)). Since it also has the same north–south orientation as the 5 cm jet-like emission centered on MM1B (see Figures 4, 8(b), and Brogan et al. 2016), we suggest the 5 cm jet arises from the ionized base of the MM1B north–south outflow. Interestingly, toward the northern lobe, a bright CS (6–5) knot lies just south of the CM2 water maser bow shock shown in Figure 3. As demonstrated in Figure 8(b), the red- and blueshifted emission from the MM1B north–south outflow are superposed along the line of sight as expected for an outflow that lies nearly in the plane of the sky. For a maximum velocity extent of 33 km s⁻¹ (from systemic) and a symmetric single-lobe extent of 31 we estimate the dynamical time is ∼580 tan(θ_inc) years, where θ_inc is the inclination angle between the outflow axis and the plane of the sky. For θ_inc = 10°, the dynamical time for the MM1B north–south outflow is only 102 years. Interestingly, in Band 10 ALMA data we have also recently detected this outflow in CS (18-17)  (E_up = 402 K, n_crit = 2 × 10⁸ cm⁻³, and ν = 880.9 GHz) and the ground state transition of HDO (1_0,0 − 0_0,0) at ν = 893.6 GHz, i.e., thermal water emission (McGuire et al. 2018). The strong detection of CS (18-17) suggests the physical conditions in this outflow are warm and dense given the high energy and critical density of this transition, in agreement with past Herschel studies of high J molecular line transitions toward protostellar outflows (see for example Goicoechea et al. 2015). However, since the observed spatial scales of the Band 10 data are not well matched to the CS (6–5) data we have not attempted to model the CS line emission.

A third, predominantly blueshifted, outflow lobe is detected extending to the northwest of the MM1 region for ∼63 (8100 au), over a velocity extent of −94 to +2 km s⁻¹ at a position angle of +138° (Figure 8). Note that while the full velocity range of this outflow is not shown in Figure 8 (in order to avoid strong confusing hot core lines toward MM1), its overall morphology is well represented. The Atacama Pathfinder EXperiment (APEX) single-dish observations of Qiu et al. (2011) also detected high-velocity blueshifted emission in CO (9-8) toward this area, albeit with an angular resolution of 64, such that the emission was unresolved. The opposing side of this outflow is not obvious in the CS (6–5) data, possibly because it is confused with the redshifted emission from the large-scale northeast–southwest flow. This outflow also appears to point back toward MM1B (so it is henceforth called MM1B NW), and two water masers (NWflow) are detected toward it in the 2017.8 epoch data (Figure 8(a)). An isolated water maser ∼5'' southeast of MM1 (UCHII-W5), with a redshifted velocity of +4.25 km s⁻¹, falls along a line joining MM1B and the NWflow water masers, suggesting that perhaps this maser traces the opposing side of the outflow, rather than the edge of the MM3-UCHII region, or perhaps that this location marks an interaction between the two structures. From the radio recombination line velocity of the MM3-UCHII region (−5 to −7 km s⁻¹, Hunter et al. 2018), it is plausibly co-distant with MM1. We estimate a dynamical time for the MM1B northwest outflow of ∼450 tan(θ_inc) years.

Finally, in CS (6–5) we have discovered a highly collimated bipolar southeast–northwest outflow emanating from MM4 (MM4 southeast–northwest) at a position angle of +102° (see Figure 8). The velocity extent of this outflow is −28 to +18 km s⁻¹. The full (projected) linear extent is ∼91 (11830 au), however, the blueshifted emission extends 52, somewhat further than the red side. The smaller extent of the redshifted lobe may be due to interaction with the MM3-UCHII region if they are co-distant (which is currently unknown). Using the projected length of the blueshifted emission and maximum velocity extent from systemic of 25 km s⁻¹ we estimate a dynamical time of ∼1280 tan(θ_inc) years. This outflow would have been spatially confused with the large-scale MM1 northeast–southwest bipolar outflow in past lower resolution molecular line data.

Figure 9(a) shows the locations of the 2017.8 epoch water masers, as well as the 2016.9 epoch 6.7 GHz methanol and 6.035 GHz excited OH masers reported by Hunter et al. (2018) with respect to the 1 mm dust continuum, 5 cm emission, and the north–south outflow traced by CS (6–5). The north–south morphology of the ensemble of maser species is striking. Figures 9(b) and (c) show close-up views of the two regions where the three maser species are in close (projected) proximity to each other—the northern and southern extremes of the CS (6–5) north–south outflow. Additional information about these relationships is given in the following sections.

Zoom In Zoom Out Reset image size

4.2.1. The Water Maser Bow Shock

4.2.2. Excited OH and CH₃OH Masers Downstream of the Bow Shock

4.2.3. The Maser/Outflow Connection and Synchrotron Nature of CM2

In general, maser pumping schemes contain radiative and collisional components, though often one dominates. Additionally, it is important to recognize that while water maser transitions are in the centimeter to submillimeter regime, many of the water transitions that lead to the overpopulation of the masing states occur in the mid-infrared (e.g., Neufeld & Melnick 1991). In the context of an accretion outburst, immediately after the rise in protostellar luminosity, the dust efficiently absorbs continuum radiation and thus heats up more quickly than the gas, which absorbs only in relatively narrow lines. As a result, the temperature of the grains surrounding the flaring protostar (T_dust) can rise dramatically (Johnstone et al. 2013). At this stage, the dust emission considerably increases the density of the infrared radiation field. But since the gas has not heated up yet, the efficiency of the collisional source of pumping initially stays at the same level. The dust plays an important role in the radiative part of the pumping and this role is twofold: dust radiation provides the source of photons for the pumping while dust absorption provides the sink for photon energy in the pumping network (Strelnitskii 1981; Sobolev & Gray 2012). It has been shown that infrared emission alone cannot provide an efficient source for the pumping of the 22 GHz water masers due to energetic considerations (Shmeld 1976; Deguchi 1981; Strelnitskii 1984). Thus, it is often stated that collisional pumping dominates for water masers. However, if the dust is too warm to provide an efficient sink for the infrared transitions that take part in the overpopulation of the masing state, then the maser pumping becomes less efficient, regardless of whether the collisional pump remains favorable (Strelnitskii 1977; Deguchi 1981). For example, calculations by Yates et al. (1997) show that water maser gain is reduced by an order of magnitude at T_dust = 300 K compared to 100 K. In the context of water masers originating in J-shocks, an additional possible effect resulting from an elevated T_dust is that free hydrogen atoms in the post-shock gas that encounter dust grains will evaporate from the grains before being able to form molecules (Hollenbach & McKee 1979), which will also reduce the efficiency of the collisional pump. In summary, the rapid rise of the infrared radiation from the central protostar and the subsequent heating of the surrounding dust should result in a net decrease of water maser brightness, which can readily explain the drop we have seen in the central parts of MM1, close to MM1B and MM1D.

In contrast, the 6.7 GHz methanol maser requires infrared radiation to be pumped along with lower densities in the range of 10⁴–10⁸ cm⁻³ (Cragg et al. 2005). Thus, for a region with density satisfying the requirements for both masers (∼10⁸ cm⁻³) and T_dust around 100 K, an increase in T_dust would favor the 6.7 GHz line and depress the 22 GHz line, while a decrease in T_dust would have the opposite effect. This scenario thus readily and consistently explains the onset of strong 6.7 GHz masers surrounding MM1 coupled with the drop in water maser emission close to MM1 as a result of the accretion outburst. Interestingly, recent monitoring of the 22 and 6.7 GHz transitions toward the intermediate mass protostar G107.298+5.639 shows a remarkable pattern of anti-correlated alternating flares of these maser species that coincide in position and velocity within 360 au (Szymczak et al. 2016), which suggests periodic changes in the radiation field from the central object as a simple explanation (e.g., in contrast to superradiance operating at different timescales (Rajabi & Houde 2017)).

As demonstrated in Figure 5, the UCHII-W1 and UCHII-W2 maser associations appear to have undergone bulk motion to the south–southeast between 2011.7 and 2017.0. Between these two epochs, we find a separation of Δ R.A.=+0036 ± 001 and Δ decl. = −009 ± 001, and a position angle of −221. These maser associations are coincident with the southernmost CS (6–5) knot in the north–south outflow (see Figure 9). Strong evidence for the persistent motion of a complex association of water masers has been reported in other protostellar outflows (e.g., Gallimore et al. 2003; Goddi et al. 2006). If we interpret the shift in position of the maser associations between epochs as bulk motion over the intervening 5.33 years, then the implied vector sum proper motion on the sky plane is 0.0182 ± 00019 yr⁻¹, in turn yielding a sky plane velocity estimate of 112 ± 12 km s⁻¹ and a dynamical time of 170 ± 18 years, consistent with a high-velocity outflow. The most blueshifted masers in UCHII-W1 and UCHII-W2 have velocities of −39.0 and −40.25 km s⁻¹, respectively, yielding a mean maximum radial velocity of 32.6 km s⁻¹ with respect to the systemic velocity (−7 km s⁻¹) and MM1B. The inferred inclination angle of the maser motion, ${\theta }_{{{\rm{H}}}_{2}{\rm{O}}}$, derived from the two component velocities is then ${\mathtt{\arctan }}(32.6/112)=16\pm 2^\circ$, and the 3D vector velocity is v_jet = 117 ± 12 km s⁻¹. In Section 3.3, we estimated a dynamical time for the north–south outflow of ∼580 tan(θ_inc), which, by assuming ${\theta }_{\mathrm{inc}}={\theta }_{{{\rm{H}}}_{2}{\rm{O}}}$, we can now refine to a dynamical time estimate of ∼166 ± 22 years.

In light of this result, we also searched for bulk motion between epochs toward the CM2-W1 bow shock. The masers at the apex have moved northward by no more than 002 = 26 au, which is similar to the relative uncertainty in the fitted positions. This upper limit yields a proper motion upper limit of <3.8 mas yr⁻¹, which corresponds to a sky plane velocity of <23 km s⁻¹ and (correcting for the outflow inclination angle θ_inc) a bow shock advancement velocity (v_bs) of <24 km s⁻¹. The ratio of v_jet/v_bs is then >5, which implies a density ratio between the jet and the ambient medium of η < 0.066 (see Equation (1) of Blondin et al. 1990). Thus, the jet is a "light" (η ≪ 1), high Mach number jet, consistent with other observed protostellar jets (Devine et al. 1997; Gómez de Castro & Robles 1999; Rodríguez-Kamenetzky et al. 2017) and the simulations of Downes & Cabrit (2003).

Interestingly, although the excited OH 6.035 GHz masers have long been thought to be related to the MM3-UCHII region, the UCHII-OH5, UCHII-OH6, UCHII-OH7 maser associations are also coincident with the southern end of the north–south outflow as traced by CS(6–5) (see Figures 9(a), (c)). These three maser associations are also precisely the ones that show a reversal of the Zeeman B_los magnetic field direction compared to all the masers coincident with the more southerly parts of the UCHII region, as reported by Hunter et al. (2018) and Caswell et al. (2011). Moreover, the UCHII-OH7 maser association is in close proximity (<100 au in projection) to the water masers of the UCHII-W2 association. Unfortunately, too few excited OH 6.035 GHz masers are detected in the Caswell et al. (2011) data (epoch 2001.0) to discern if the OH masers show the same bulk motion as the water maser association UCHII-W2. Together, these clues suggest that, like the water masers, the UCHII-OH5, UCHII-OH6, and UCHII-OH7 maser associations are excited by shocks within the outflow, or possibly by interaction between the southern lobe of the outflow and the UCHII region.

The NGC 6334I-South masers were detected in both the 2017.0 and 2017.8 water maser epochs. We searched SIMBAD for counterparts to NGC 6334I-South (Table 5), and the only nearby object is the X-ray source CXOU 172053.21-354726.4, which is offset 27 north of the maser. The X-ray source has been identified as a possible counterpart to the infrared star TPR-9 (Tapia et al. 1996) as their respective positions are within 1''. It is possible that the maser is associated with an outflow from this star. We note that the separation from MM1B is 318, which corresponds to a projected distance of 239 light days. Since our VLA maser observations occurred more than this many days after the origin of the outburst (MacLeod et al. 2018), we cannot exclude the possibility that this maser was somehow pumped by narrowly beamed radiation from MM1B.