Mass Assembly of Stellar Systems and Their Evolution with the SMA (MASSES)—1.3 mm Subcompact Data Release

We wanted the targeted cloud to be nearby and have a large protostellar population so that one can statistically constrain protostellar evolution, but not be so large of a sample that a survey is impractical for the SMA (e.g., Orion). The Perseus molecular cloud has over 70 protostellar objects, ranging from candidate first hydrostatic cores that have just formed central, hydrostatic objects, all the way to evolved Class I systems near the end of the protostellar stage. For star-forming clouds within ∼350 pc, Perseus (and possibly Aquila; the distance to the cloud is uncertain) is the only star-forming cloud with more than 40 protostellar objects (Dunham et al. 2015). At decl. = +31°, Perseus is ideally located in the sky for maximum SMA visibility and can be targeted by most telescopes in the world. Aquila, on the other hand, has a declination near 0°, which makes it difficult to attain sufficient SMA uv coverage to produce high fidelity maps with the SMA. As one of the best-studied sites of nearby star formation, copious complementary data are available for Perseus to aid with analysis, including single-dish imaging at mid-IR (Spitzer), far-IR (Herschel), and (sub)mm (James Clerk Maxwell Telescope, Caltech Submillimeter Observatory; JCMT, CSO) wavelengths (e.g., Hatchell et al. 2005; Enoch et al. 2006; Jørgensen et al. 2006; Kirk et al. 2006; Evans et al. 2009; Sadavoy et al. 2014; Dunham et al. 2015; Chen et al. 2016; Zari et al. 2016). Finally, the VANDAM Perseus survey had already observed the same targets and revealed multiplicity down to a projected separation of 15 au (Tobin et al. 2016). The synergy between connecting the physical and kinematic properties of the dense gas and dust revealed by the SMA and the multiplicity revealed by the VLA is one of the key strengths of this survey.

From 2014 to 2017, we used the SMA to observe all known protostars in the Perseus molecular cloud. We targeted 74 protostellar systems (some "systems" are multiples not resolved by Spitzer). Spitzer was used to identify 66 of these targets, and they were identified as Per-emb-1 through Per-emb-66 (Enoch et al. 2009). Eight additional systems that were not identified in the Enoch et al. (2009) Spitzer survey were observed as well. These systems are B1bN and B1bS (e.g., Pezzuto et al. 2012), L1448-IRS2e (e.g., Chen et al. 2010), L1451-mm (e.g., Pineda et al. 2011), Per-Bolo-45 (e.g., Schnee et al. 2012), Per-Bolo-58 (e.g., Dunham et al. 2011), and SVS 13B and 13C (e.g., Chen et al. 2009). Except for SVS 13B and 13C, these systems are candidate first hydrostatic cores (see Dunham et al. 2014 for a brief discussion on first hydrostatic cores), though some of the aforementioned studies suggest they could be Class 0 protostars. The candidate first cores were not identified by Enoch et al. (2009) because they were deeply embedded and/or had low luminosities. The SVS 13B/13C sources were not identified because they lie near the SVS 13A diffraction spike and thus failed the 24 μm signal-to-noise criteria set out in Enoch et al. (2009). The vast majority of protostars are expected to be identified by Enoch et al. (2009), unless a large population of protostars with luminosities substantially below 0.1 L_⊙ exists (Dunham et al. 2008). A future Herschel catalog of protostellar sources would better constrain the completeness of the MASSES protostellar sample.

The angular separation between some of these 74 protostellar systems was small enough that a single pointing could observe both systems simultaneously. We needed a total of 68 pointings to survey every system. The phase centers of each target are given in Table 1. Accurate positions of the protostars themselves (which are typically within the SMA envelopes), along with their multiplicity (resolved to a projected separation of 15 au), are given in Tobin et al. (2016).

Observations for the MASSES survey (project code 2014A-S093; Co-PIs M. Dunham and I. Stephens; searchable in the SMA archive via Dunham) were conducted using the SMA (Ho et al. 2004), which is an eight-element array of 6.1 m antennas located on Maunakea. While the SMA has eight antennas, only seven antennas were typically available for these observations. For the MASSES survey, we made observations in both the subcompact (SUB) and extended (EXT) SMA array configurations. The baselines covered by the SUB and EXT configurations at 230 GHz were approximately 4–55 kλ and 20–165 kλ, respectively, although these ranges varied if certain antennas were missing from the array. The focus of this data release paper is on the SUB data, and the combined SUB plus EXT data will be presented in a forthcoming paper.

While the MASSES project was being observed, the SMA upgraded its correlator from the Application Specific Integrated Circuit (ASIC) correlator to the SMA Wideband Astronomical ROACH2 Machine (SWARM) correlator (Primiani et al. 2016). The SUB observations were predominantly done with the ASIC correlator. Twenty-eight ASIC SUB tracks and six SWARM SUB tracks had usable data. More information on each correlator will be discussed below.

The SMA can observe simultaneously with two receivers that can be tuned to different frequencies. The spectral setup and line rest frequencies are indicated in Table 2. In the SUB configuration, we used the dual receiver mode to tune the SMA's two receivers to different frequencies. For the ASIC data, the local oscillators of the receivers were tuned to 231.29 and 356.720 GHz. For the SWARM data, the receivers were tuned to 231.29 and 356.410 GHz. For some tracks, the higher-frequency 356 GHz tuning is missing due to technical difficulties with the SMA. In this paper, we focus solely on the 231.29 GHz subcompact data; the ∼356 GHz data will be presented in a future data release paper.

The ASIC correlator in dual receiver mode has a total bandwidth of 2 GHz for each sideband, and the center of each sideband is separated by 10 GHz. Each 2 GHz sideband is divided into 24 chunks, each of which has a bandwidth of 104 MHz. These 104 MHz chunks slightly overlap in frequency, making the "effective" bandwidth of each chunk 82 MHz.

In dual receiver mode, the ASIC correlator is divided into six blocks, each with four chunks. Each block is allowed up to 1024 channels, which can be distributed to each chunk by a power of 2 between 64 and 1024. Chunks are also allowed 0 channels. If the user specifies 1024 channels in one block of one receiver (e.g., the low-frequency receiver), the other receiver (e.g., the high-frequency receiver) can have no channels assigned to its block. To maximize spectral resolution for the chosen spectral lines, blocks 1 and 5 (chunks s01 to s04 and s16 to s19, respectively) had 1024 channels for the high-frequency receiver and blocks 4 and 6 (chunks s13 to s16 and s20 to s24, respectively) had 1024 channels for the low-frequency receiver. Blocks 2 and 3 (chunks s05 to s12) were used for the continuum. Chunk s14 in the lower sideband (lsb) was also used for the continuum because it did not contain any lines. Table 2 shows the ASIC chunk number(s) assigned to the continuum and for each spectral line, the amount of channels per chunk, and the velocity resolution of the uv data. The continuum has eight chunks (total bandwidth of 656 MHz) in the upper sideband (usb), and nine chunks in the lsb (738 MHz). Combining the sidebands together, the total continuum bandwidth for the ASIC correlator is 1.394 GHz.

The SWARM correlator allows for the SMA to observe 8 GHz for each sideband simultaneously at a uniform spectral resolution of 140 kHz (0.18 km s⁻¹ at 233 GHz) across the entire bandwidth. The center of each sideband is separated by 16 GHz. Each SWARM sideband is divided into four different chunks that slightly overlap in frequency, with each chunk containing 16384 channels. Combining the two sidebands together, the SWARM correlator provides a total bandwidth of 16 GHz, which allows for tracks using SWARM to reach much better sensitivities in the 1.3 mm continuum than those for ASIC. SWARM's high spectral resolution across its entire bandwidth, along with its additional frequency coverage increases the likelihood that additional spectral lines are detected. These spectral lines were identified by looking at the uv-averaged spectrum, but are not mapped in this paper. These identified lines detected toward some targets are listed in Table 3. For MASSES targets using the SWARM correlator, these lines were the strongest for the fields targeting Per-emb-15, Per-emb-44/SVS 13B, and SVS 13C, and very weak or undetected toward other fields. It is certainly possible that additional lines that are not listed in this table were also detected toward some targets.

Table 3.  Other SWARM Lines Detected toward Some Fields

Tracer	Transition	Frequency	Eu
		(GHz)	(K)
SO	JN = 55 − 44	215.22065	44.1
DCO+	J = 3 − 2	216.11258	20.7
DCN	J = 3 − 2	217.23854	20.9
c-C3H2	60,6-51,5	217.82215	38.6
c-C3H2	61,6-50,5	217.82215	38.6
H2CO	${J}_{{K}_{a},{K}_{b}}={3}_{\mathrm{0,3}}-{2}_{\mathrm{0,2}}$	218.22219	21.0
H2CO	${J}_{{K}_{a},{K}_{b}}={3}_{\mathrm{2,2}}-{2}_{\mathrm{2,1}}$	218.47563	68.1
H2CO	${J}_{{K}_{a},{K}_{b}}={3}_{\mathrm{2,1}}-{2}_{\mathrm{2,0}}$	218.76007	68.1
SO	JN = 65 − 54	219.94944	35.0

Note. Frequencies and upper energy levels are from Splatalogue (http://www.splatalogue.net/). While all these lines are certainly detected toward some sources, other lines may exist in the data. Images for these lines are not provided in this data release. Only the full SWARM visibilities are delivered.

Download table as:  ASCIITypeset image

The full SWARM uv data includes these additional lines. The SWARM frequency coverage for the lsb is approximately 214.5–222.5 GHz and for the usb is approximately 230.5–238.5 GHz.

For some observations, both correlators were used simultaneously and were tuned to similar frequencies. Given that using multiple correlators does not increase signal-to-noise (i.e., they use the same receivers), we discarded the ASIC correlator observations in these instances. These tracks are considered SWARM tracks in Table 1.

The names of the tracks, as defined in the SMA Archive, are given in the "Track(s)" column of Table 1. The format of the names is YYMMDD_STARTTIME, where YY is the year, MM is the month, DD is the day, and STARTTIME denotes the start time of the track. Tracks that were taken on the same day (i.e., have the same YYMMDD prefix) were combined together during the data reduction process. We also indicate in this table which antenna number(s) are missing from the track, where the eight SMA antennas are assigned numbers 1 through 8.

All MASSES data were calibrated using the MIR software package, partially following the general reduction process outlined in the MIR cookbook.¹⁴ We discuss our reduction process in detail here.

With MIR, we first applied a baseline correction to tracks requiring such a correction, as noted on the Radio Telescope Data Center website.¹⁵ We then flagged irrelevant data, such as slew and pointing scans. Then we applied a system temperature correction to the data. For bandpass calibration, we typically used 3C454.3 and/or 3C84. Sometimes the SMA operator observed additional bandpass calibrators in conjunction with these two, and if they were sufficiently bright and high-quality, we often included them in the bandpass calibration. We first applied a phase-only bandpass calibration, followed by an amplitude-only bandpass calibration.

Next, we used flux calibrators to measure the fluxes of our gain calibrator, 3C84. We typically used Uranus as a flux calibrator, but when it was not available, we used Neptune, Callisto, or Ganymede. To maximize the accuracy of the flux calibration, we first tried to find where 3C84 elevation and integration time best matched that of the flux calibrator. We then used only these parts of the 3C84 data, and flux-calibrated only on the shorter baselines (≲10 kλ). In doing so, we are only using the highest signal-to-noise data that is the most representative of the observing conditions of the flux calibrator. Using the Submillimeter Calibrator List,¹⁶ we checked the flux measured by the SMA for a nearby day, and found that our measured flux for 3C84 was similar (within ∼20%), even though the source is quite variable.

After the flux was measured, we used the 3C84 flux to scale the fluxes for the MASSES targets during gain calibration. This was done by specifying the measured flux of 3C84 during MIR's gain_cal task. We first performed phase-only gain calibration, followed by amplitude-only gain calibration.

At this point, the calibration was complete, and the data were converted to MIRIAD format (Sault et al. 1995). For ASIC data, the high spectral resolution chunks containing CO(2–1) (including the usb chunk s13), ¹³CO(2–1), C¹⁸O(2–1), and N₂D⁺(3–2) were exported with their full resolution, while the rest of the chunks were averaged together to generate the 1.3 mm continuum. For SWARM, which has a uniform spectral resolution across the entire spectrum, the data were exported in their entirety in MIRIAD. Lines were then split out of the bandwidth manually using MIRIAD, and the rest of the channels were averaged as the continuum. In some cases, other high signal-to-noise spectral lines (besides the four mentioned above) were serendipitously detected with SWARM (Table 3), and thus were also removed from the continuum.

All targets of the MASSES survey were imaged with MIRIAD after exporting the calibrated data from MIR. For imaging of spectral lines, we first used the MIRIAD task uvlin to subtract a zeroth order polynomial from the continuum. The delivered spectral line uv data all have had their continua subtracted.

Dirty maps were then created via an inverse Fourier transform using the MIRIAD task invert. If a particular target was observed over multiple days (see Table 1), the tracks were combined during this task. All targets were imaged using Briggs weighting with the robust parameter equal to 1. We also specified the pixel size to be 08 with 100 pixels on each side of the map, resulting in 80'' × 80'' maps. These maps image outside the full width at half maximum (FWHM) of the primary beam, which has a size of 48'' at 231 GHz. For each line, we also specified the imaged channel width, which was slightly wider than the channel spectral resolution. The imaged spectral resolutions are specified in Table 2 under the Δv_img column. The number of channels imaged is also listed in this table. For CO(2–1) data, which often have high-velocity components, an additional ASIC chunk is available in the usb (s13). This chunk provides for redshifted velocities of ∼26–160 km s⁻¹, but the user should note that for many targets, the noise in the chunk is extremely high at the edges. The s13 chunk has a slight overlap with the main CO(2–1) chunk, s14, which is imaged for velocities up to ∼48.5 km s⁻¹. In a case in which the CO(2–1) data were imaged with SWARM, we imaged channels for a larger velocity range (as indicated in Table 2) than could be used with a single ASIC chunk.

After creating dirty maps, we cleaned the maps using the MIRIAD task clean. For CO(2–1) and ¹³CO(2–1), which typically trace protostellar outflows, we used a three-step iterative cleaning algorithm. We first cleaned only the pixels and channels with emission to a level of 1.5 times the dirty map noise. This was completed by binning channels together with similar emission and creating multiple regions with the task cgcurs. When selecting such pixels and channels, we typically excluded emission near the systemic velocity, since the emission is often completely confused with the large-scale emission of the Perseus molecular cloud. For the second step, we selected the area with cgcurs where emission is present over the entire cube, and cleaned this entire region for all channels to 2 times the dirty map noise. For the third and final step, we cleaned over all pixels and channels to 2.5 times the dirty map noise. We find that this three-step cleaning process recovers the extended emission well while minimizing the creation of interferometric imaging artifacts.

For the continuum, C¹⁸O(2–1), and N₂D⁺(3–2), all of which typically trace protostellar envelopes, we used a two-step iterative cleaning algorithm. This algorithm is identical to that described above, except we skip the second step since that step had a negligible effect for imaging the compact emission that is usually found by these three tracers. For both the three-step and two-step cleaning processes, if no emission was obviously associated with a source, we only cleaned the channels to 3 times the dirty map noise. After deriving the clean components, we used the MIRIAD task restor to create clean maps.

The sensitivity (standard deviation) and synthesized beam parameters for the continuum and spectral lines for each map are listed in Table 4. The sensitivity for the continuum was measured using the IDL task sky.¹⁷ This task measures the standard deviation over the entire map, clips all values deviating from 3 times this standard deviation, and iteratively repeats this for a total of 5 iterations. Sensitivities for the continuum were often limited by dynamic range rather than integration time. For some sources, self-calibration may help the continuum sensitivity, but we choose to not self-calibrate because we want to deliver consistent data products to the user. We note that sensitivity improvements using self-calibration with the SMA are typically minimal (i.e., less than ∼10%), in part because the SMA is only an eight-element array. Note that the uv data are provided in case the user wants to apply the calibration. For the sensitivity of spectral line observations, we measured the standard deviation of the pixels over many emission-free channels and converted the standard deviation to a brightness temperature. The distribution of sensitivities is shown in Figure 1.

Zoom In Zoom Out Reset image size

Figure 2 shows the peak 1.3 mm pixel flux normalized by the sensitivity (a proxy for dynamic range) versus the sensitivity of the image. Targets that have higher dynamic ranges have worse sensitivities, even for good observing conditions, indicating that continuum sensitivities are often limited by dynamic range.

Zoom In Zoom Out Reset image size

Table 4 also gives the FWHM of the synthesized beam's major and minor axes, θ_maj, and θ_min, and the position angle of the beam's major axis, PA, which is measured counterclockwise (east) from north.

Table 5 shows whether the line is detected toward each MASSES field, as judged by analyzing the continuum images and the spectral cubes by eye. With the exception of some marginal C¹⁸O(2–1) detections, the three spectral lines CO(2–1), ¹³CO(2–1), and C¹⁸O(2–1) are detected toward every field. Nevertheless, these detections do not imply that the emission is always associated with the source, since the large-scale emission of the Perseus molecular cloud is detected with these observations. Indeed, spectral lines do not seem to be associated with many of the targets, which will be analyzed in more detail in Section 5.5.

We estimate the minimum mass of a compact source that we expect to detect for a given continuum observations based on the measured sensitivity σ_{1.3 mm}. Following Hildebrand (1983), the mass of a source for optically thin dust continuum flux is

$$\begin{eqnarray}&&M={R}_{\mathrm{gd}}\displaystyle \frac{{F}_{\nu }{d}^{2}}{{\kappa }_{\nu }{B}_{\nu }({T}_{\mathrm{dust}})},\end{eqnarray} \tag{ 1 }$$

where R_gd is the gas-to-dust-mass ratio, F_ν is the source's flux, d is the distance to the source, κ_ν is the dust opacity, and B_ν(T_dust) is the Planck function at dust temperature T_dust. We assume typical values of R_gd = 100, d = 235 pc (Hirota et al. 2008), and κ_{1.3 mm}= 0.899 cm² g⁻¹ (Ossenkopf & Henning 1994, assuming thin ice mantles and a gas density of 10⁶ cm⁻³). To be conservative with our minimum detected compact mass estimates, we assume T_dust = 10 K and require a 3σ detection with all the flux in a single beam, i.e., F_ν = 3(σ_{1.3 mm} × bm). Note that if the source is larger than the beam, this assumption for F_ν is not valid, i.e., we are only concerned with the detection limit of a source smaller than the beam. Given these assumptions, the mass detection limit for each field is

$$\begin{eqnarray}&&{M}_{\mathrm{limit}}=\left(\displaystyle \frac{{\sigma }_{1.3\mathrm{mm}}}{\mathrm{mJy}\,{\mathrm{bm}}^{-1}}\right)\times 0.010\,{M}_{\odot },\end{eqnarray} \tag{ 2 }$$

where σ_{1.3 mm} is given in Table 4. The mass sensitivity for the continuum varies dramatically, often due to dynamic range, but also due to observing conditions and the correlator that is used (ASIC versus SWARM). To demonstrate a pessimistic mass detection limit for the majority of the observations, we select the observation for Per-emb-4 as an illustrative example. In this field, no continuum source was detected (i.e., we were not limited by dynamic range), and our estimated thermal noise is a bit worse than most MASSES observations due to unfavorable observing conditions. For this field, σ_{1.3 mm} = 2.0 mJy bm⁻¹, so M_limit = 0.02 M_☉. Therefore, for most of our fields, we expect to detect a compact source greater than 0.020 M_☉ with >3σ significance, if it exists.

Many sources in Table 4 have σ_{1.3 mm} values that are much higher than 2.0 mJy bm⁻¹ because these observations were limited by dynamic range. In other words, while our detection limit is generally ∼0.02 M_☉, if there is a much brighter source in the field, we would not be able to detect a 0.02 M_☉ source in the same field.

We provide the delivered uv data for both the continuum observations and spectral line observations. If tracks were taken on the same day (i.e., had the same YYMMDD prefix; see Section 2.2), they are combined into a single data set during the MIR calibration. As explained in Section 3.2, lines were subtracted when generating the continuum, and the continuum was subtracted for each spectral line observation. The line uv data is delivered per spectral line for each track, while the continuum uv data is delivered separately for the lsb and usb.

Finally, for the SWARM data, we also provide the full-resolution uv data, which are not continuum-subtracted. Many additional spectral lines (Table 3) may be found in these cubes, as discussed in Section 2.2. These lines were typically not detected in the ASIC cubes in part due to the smaller bandwidth coverage, although the coarse spectral resolution also makes such detections with ASIC less obvious. These SWARM uv data are delivered separately for each of its 8 (4 per sideband) spectral chunks.

We also note that toward the edge of both the SWARM and ASIC chunks, the noise in the spectra is extremely high, so the user should use these channels with caution. These channels were not used for the delivered imaged cubes.

The uv data are delivered as uv-fits files. Examples of the delivered uv-fits file names are given in Section 4.3.

For the continuum and spectral line observations, we deliver both the images not corrected for the primary beam as well as the primary-beam-corrected images. The images that are not corrected for the primary beam are typically used to better show structure throughout the entire map. The primary-beam corrected images allow the user to make accurate flux measurements.

Multiple tracks were combined during the MIRIAD invert task (Section 3.2), allowing for two delivered products per continuum/spectral line (primary-beam-uncorrected and primary-beam-corrected). These data are delivered as fits files.

The units for the continuum images are Jy bm⁻¹, and the units for the spectral line cubes are Jy bm⁻¹ channel⁻¹.

An example of the delivered fits files for the Per-emb-24 data set is shown below.

The delivered uv data for Per-emb-24 are:

• 1.  
  Per24.sub.cont1.3 mm.lsb.151122.uvfits
• 2.  
  Per24.sub.cont1.3 mm.usb.151122.uvfits
• 3.  
  Per24.sub.cont1.3 mm.lsb.151127.uvfits
• 4.  
  Per24.sub.cont1.3 mm.usb.151127.uvfits
• 5.  
  Per24.sub.12CO21.151122.uvfits
• 6.  
  Per24.sub.12CO21.151127.uvfits
• 7.  
  Per24.sub.highvelCO.151122.uvfits
• 8.  
  Per24.sub.highvelCO.151127.uvfits
• 9.  
  Per24.sub.13CO21.151122.uvfits
• 10.  
  Per24.sub.13CO21.151127.uvfits
• 11.  
  Per24.sub.C18O21.151122.uvfits
• 12.  
  Per24.sub.C18O21.151127.uvfits
• 13.  
  Per24.sub.N2DP.151122.uvfits
• 14.  
  Per24.sub.N2DP.151127.uvfits

The delivered images/data cubes for Per-emb-24 are:

• 1.  
  Per24.sub.cont1.3 mm.fits
• 2.  
  Per24.sub.cont1.3 mm.pbcor.fits
• 3.  
  Per24.sub.12CO21.cube.fits
• 4.  
  Per24.sub.12CO21.cube.pbcor.fits
• 5.  
  Per24.sub.highvelCO.cube.fits
• 6.  
  Per24.sub.highvelCO.cube.pbcor.fits
• 7.  
  Per24.sub.13CO21.cube.fits
• 8.  
  Per24.sub.13CO21.cube.pbcor.fits
• 9.  
  Per24.sub.C18O21.cube.fits
• 10.  
  Per24.sub.C18O21.cube.pbcor.fits
• 11.  
  Per24.sub.N2DP.cube.fits
• 12.  
  Per24.sub.N2DP.cube.pbcor.fits

Tracks with SWARM data will include include additional full spectral resolution uv data, separated into 8 chunks (4 for each sideband). An example of the additional SWARM uv data for Per-emb-7 is shown below.

• 1.  
  Per7.sub.SWARM.lsb.s1.160925.uvfits
• 2.  
  Per7.sub.SWARM.lsb.s2.160925.uvfits
• 3.  
  Per7.sub.SWARM.lsb.s3.160925.uvfits
• 4.  
  Per7.sub.SWARM.lsb.s4.160925.uvfits
• 5.  
  Per7.sub.SWARM.usb.s1.160925.uvfits
• 6.  
  Per7.sub.SWARM.usb.s2.160925.uvfits
• 7.  
  Per7.sub.SWARM.usb.s3.160925.uvfits
• 8.  
  Per7.sub.SWARM.usb.s4.160925.uvfits

Note, the full SWARM uv data have not been continuum-subtracted because the subtraction is best done by fitting the continuum near each spectral line rather than across the entire bandwidth. If a user analyzes the SWARM visibility data, the first and last ∼1200 channels for each chunk also should probably be discarded, since fluxes toward the chunk edges of SWARM cannot be trusted. For sources with multiple protostellar systems in the same field, the names were combined in the file name, e.g., Per8Per55.

Figure 3 shows CO(2–1) moment 0 maps of all protostars in the MASSES sample. For some maps, outflows are quite clear, while for other maps, outflows can be confused with other emission (e.g., Per-emb-32), or they are completely absent (e.g., Per-emb-4). Note that we do not supply moment 0 maps because selecting integrated channel ranges for such maps depends on the features the user wants to extract from the map. Nevertheless, these maps can be reproduced by using the channel integration ranges shown in Table 5, and these ranges were selected by eye in attempt to best show the outflow emission. Spectra for many of these outflows are presented in Stephens et al. (2017).

Zoom In Zoom Out Reset image size

Measurements of most of the envelope masses of the protostars were originally estimated in Pokhrel et al. (2018) by converting the integrated 1.3 mm continuum flux to a gas mass. We briefly summarize the methodology below, and we refer the reader to Pokhrel et al. (2018) for additional details.

To measure the integrated fluxes of envelopes, we first analyzed the envelope's visibility plot of the amplitude versus uv-distance to determine which of three following models best describes the source. Some envelopes appeared more as point sources (amplitude versus uv-distance relation was flat), some were more Gaussian (relation is Gaussian), and others were a combination of the two (relation is a Gaussian with a flat tail). The total integrated flux of the source was then fit via the miriad task uvfit by specifying one of these models.

The targets have a variety of luminosities and evolutionary stages, and therefore many targets are likely to have different temperatures. Temperatures at radius r of each envelope are estimated via

$$\begin{eqnarray}&&T(r)=60{\left(\displaystyle \frac{r}{2\times {10}^{15}{\rm{m}}}\right)}^{-q}{\left(\displaystyle \frac{{L}_{\mathrm{bol}}}{{10}^{5}{L}_{\odot }}\right)}^{q/2}\,{\rm{K}},\end{eqnarray} \tag{ 3 }$$

where L_bol is the bolometric luminosity of the source (taken from Tobin et al. 2016) and q is related to the dust emissivity index via q = 2/(4 + β). For our envelopes, we calculate T_{d,1000 au}, which is the dust temperature at 1000 au (approximately the resolution of the MASSES observations) assuming q = 0.33. Each envelope's estimated Td,1000 au value is listed in Table 6. Targets with multiple envelopes in the same field (e.g., Per-emb-10 and Per-emb-10-SMM) are assigned with the same temperature T_{d,1000 au}. The integrated intensity and T_{d,1000 au} are then used to estimate the gas mass via Equation (2) (using the same values for R_gd, d, and κ_ν discussed in Section 3.3). These estimated envelope masses are given in Table 6.

Three envelope masses (for Per-emb-7, Per-emb-34, and Per-emb-38) were not reported in Pokhrel et al. (2018) because those data were not yet calibrated at the time of publication. In the same manner as Pokhrel et al. (2018), Per-emb-7 was fit with a Gaussian, while Per-emb-34 and Per-emb-38 were each fit with a point source plus a Gaussian.

The envelope estimates for many sources were updated after newer, calibrated baselines were added to the visibility data (Per-emb-15, Per-emb-30, Per-emb-44, Per-emb-47, Per-bolo-45, SVS 13B, and SVS 13C). The calculated masses were similar to those calculated in Pokhrel et al. (2018), except for Per-emb-44, whose mass was essentially halved. This change was due to a different dust temperature being used for Per-emb-44, 38 K, rather than the temperature of 21 K (i.e., the temperature of the adjacent source SVS 13B) used in Pokhrel et al. (2018). Moreover, we also update the Pokhrel et al. (2018) table with two other corrections; the measured Pokhrel et al. (2018) integrated flux for Per-emb-41 should have been for B1bS, and we also use update Per-emb-19-SMM's T_{d,1000 au} to be the same as Per-emb-19. Finally, we update all the masses to use 235 pc as the distance rather than 230 pc. These new masses do not change any of the conclusions found in Pokhrel et al. (2018).

We show the distribution of envelope masses in Figure 4. The range of masses is from 0.010 to 3.2 M_⊙. The median value for the measured envelope masses is 0.13 M_⊙. We do not provide the median value for all envelopes (i.e., including those undetected in the continuum), since we question the protostellar nature of some of these targets (see Section 5.5).

Zoom In Zoom Out Reset image size

In Figure 5, we show the measured envelope masses versus the bolometric temperatures, T_bol. Envelope masses tend to slightly decrease with increasing bolometric temperature. Since protostars with higher bolometric temperatures tend to be older sources, this trend may suggest that envelopes typically become smaller as the protostar evolves. The decreasing envelope masses between the Class 0 and I stages agree with the results based on C¹⁸O(1–0) observations in Arce & Sargent (2006) and submillimeter continuum observations in Jørgensen et al. (2009). This trend, along with the large observed scatter, was also found via simulations in Frimann et al. (2016).

Zoom In Zoom Out Reset image size

For the MASSES survey, C¹⁸O(2–1) is more frequently detected toward the peak of the continuum emission than N₂D⁺(3–2) and typically has a much higher signal to noise. Therefore, we fit C¹⁸O(2–1) spectra to estimate line widths toward individual protostellar sources.

We show the C¹⁸O(2–1) spectra toward each protostar in which we believe C¹⁸O(2–1) emission is associated with the protostar. We deemed emission as associated with the protostar if an integrated intensity map of C¹⁸O(2–1) overlapped significantly with the location of the protostar. In Figure 6, we show the C¹⁸O(2–1) spectra, which is taken at the 1.3 mm continuum peak of the protostar. If the 1.3 mm continuum is not detected (Per-emb-31, Per-emb-32, Per-emb-42, Per-emb-52, and Per-emb-54) we instead take the spectra from the protostar's position given in Tobin et al. (2016). Each spectrum is fit with a Gaussian, and the fit parameters (amplitude, systemic velocity v_systemic, and line width Δv) are presented in Table 6. Sources with no fits in the table have a 1.3 mm continuum detected, but C¹⁸O(2–1) is not obviously associated with the source.

Zoom In Zoom Out Reset image size

The fit parameters in Table 6 indicate that the amplitudes and line widths do not always have great fits. However, the systemic velocities are fairly well constrained. The median line width of the C¹⁸O(2–1) spectra toward the protostars' positions is 1.45 km s⁻¹, considering only fitted line widths with 3σ significance. These C¹⁸O(2–1) line widths are broader than the 0.6–1.0 km s⁻¹ line widths measured for the Perseus filaments via C¹⁸O(1–0) with the Five College Radio Astronomy Observatory 14 m telescope at 1' (0.07 pc) resolution (Hatchell et al. 2005). Moreover, Kirk et al. (2007) observed C¹⁸O(2–1) at Perseus cores using the IRAM 30 m telescope at 11'' (∼2600 au) resolution. They found that C¹⁸O(2–1) line widths toward Perseus cores are also typically lower than 1 km s⁻¹, with protostellar cores having slightly larger line widths than starless cores. Other cores in nearby molecular clouds also typically have lower than 1 km s⁻¹ C¹⁸O(1–0) line widths (Myers & Benson 1983).

We investigate whether multiplicity or the envelope mass (as measured by the 1.3 mm continuum; Table 6) affects the observed C¹⁸O(2–1) line widths. Tobin et al. (2016) investigated the multiplicity of all Perseus protostars with a resolution of ∼0065 (15 au), and reported the projected separations of all multiples out to 43'' (∼10,000 au). In Figure 7, we show the projected separation of the nearest known multiple of each protostar as a function of the C¹⁸O(2–1) line width. The size of each point in Figure 7 is directly proportional to the envelope mass. The protostellar systems with the largest line widths tend to be multiple systems within the SMA beam (∼1000 au), but there are still many close multiples with smaller line widths. Envelope mass does not seem to have a major effect on the observed line widths. Other factors such as rotation, infall, and/or outflows could also broaden the C¹⁸O(2–1) line. At this time, it is unclear why the envelope line widths are larger than the observed large-scale line widths.

Zoom In Zoom Out Reset image size

We do not show the N₂D⁺(3–2) spectra as we did with C¹⁸O(2–1) because the emission is typically anti-correlated with the continuum emission. Instead, we investigate the detection rates of N₂D⁺(3–2) near the protostar as a function of evolution.

Figure 8 shows a histogram of whether or not N₂D⁺(3–2) is detected toward each protostar, as indicated in Table 5. The approximate ages of these protostars are considered to be proportional to the bolometric temperature, T_bol (e.g., Myers & Ladd 1993), which is listed for each source in Table 1. Figure 8 bins the protostars by T_bol in four ∼equal-sized bins. The detection rates (with marginal detections considered as a non-detection) for the four bins, from left to right, are 84.2% ±8.4%, 88.9% ± 7.4%, 44.4% ± 11.7%, and 31.6% ± 10.7%. This indicates that protostars with higher T_bol, which are likely to be older protostars, are much more likely to have N₂D⁺(3–2) undetected. Since N₂D⁺(3–2) is expected to disappear at T ≳ 20 K (e.g., Jørgensen et al. 2011) or at least in hotter environments (Emprechtinger et al. 2009), such a trend is expected.

Zoom In Zoom Out Reset image size

As mentioned in Section 3.2, several sources have no spectral lines associated with the target. Some of these non-detected targets also do not have 1.3 mm continuum associated with them, nor are they detected with the VLA in Tobin et al. (2016). There are six sources in question: Per-emb-4, Per-emb-39, Per-emb-43, Per-emb-45, Per-emb-59, and Per-emb-60.

With the exception of Per-emb-4, the derived bolometric luminosities (listed in Enoch et al. 2009 and Tobin et al. 2016) are similar to or smaller than the error. Only Per-emb-39 and Per-emb-60 are marginally detected in the 1.3 mm continuum (Figure 3). Based on CSO SHARC-II (Suresh et al. 2016) and JCMT SCUBA-2 (Chen et al. 2016) single-dish observations, Per-emb-39 and Per-emb-60 also have evidence of compact structure, while Per-emb-4 questionably does as well. However, based on these single-dish observations and Herschel archival observations, no strong compact emission is evident for Per-emb-43, Per-emb-45, and Per-emb-59.

Envelope masses for these sources are likely to be less than 0.2 M_⊙, and in some cases, much smaller (see Section 3.3 and Table 6). These six sources may have spectral energy distributions that cause them to be undetected at (sub)millimeter wavelengths (e.g., they are not protostellar in nature), are misidentifications in the Enoch et al. (2009) paper, or are simply very low-mass sources that are undetected with the MASSES and VANDAM observations. If these targets are misidentifications, this may indicate that Enoch et al. (2009) sources with poor luminosity fits are less reliably protostars.

Many examples of MASSES observations have already been presented in a number of papers (Lee et al. 2015, 2016; Frimann et al. 2017; Stephens et al. 2017; Pokhrel et al. 2018). All of the previous papers showed results from the ASIC correlator. Here, we show an example of one of the brightest regions in the MASSES sample, SVS 13, in which we combine an ASIC-only track with a SWARM-only track. Looney et al. (2000) classified the three envelopes in the SVS 13 region, from northeast to southwest, as SVS 13A, SVS 13B, and SVS 13C. SVS 13A is referred to as Per-emb-44 in the MASSES sample, but we will refer to it as SVS 13A in this section.

In Figures 9 and 10 we show the MASSES observations of the SVS 13 region. The 1.3 mm continuum, C¹⁸O(2–1), and N₂D⁺(3–2) SWARM observations are combined with ASIC observations, while CO(2–1) and ¹³CO(2–1) are SWARM-only (chunks containing CO(2–1) and ¹³CO(2–1) were missing from the observations during the ASIC track). The four panels in Figure 9 are centered on SVS 13A, while the four panels in Figure 10 are centered on SVS 13C.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The top panels show integrated intensity (moment 0) maps of CO(2–1) and ¹³CO(2–1). These maps trace the protostellar outflows of the sources. The bipolar outflow for SVS 13A is quite extended, with an angular extent much larger than the mapped area. Maps of the entire outflows are shown via a CO(1–0) mosaic in Plunkett et al. (2013). We show CO(2–1) velocity maps in Figure 11. For the SVS 13A outflow, we find high-velocity components associated with both the blue and red lobes. We indicate the locations of these high-velocity components in Figure 9 and show the CO(2–1) spectra in Figure 12. Using the MIRIAD task maxfit, we find that the peak emission in the −150 km s⁻¹ channel is located at R.A. = 03:29:04.6 decl. = +31:15:43, and the peak in the 160 km s⁻¹ channel is located at R.A. = 03:29:03.4 +31:16:29. These high-velocity components were originally detected with the Caltech Submillimeter Observatory in Masson et al. (1990). Given that the systemic velocity of SVS 13 system is about 8 km s⁻¹ (see Figure 13), both lobes have line-of-sight velocity components of over 150 km s⁻¹ relative to the protostar's rest-frame.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The bottom left panels of Figures 9 and 10 show moment 0 maps of C¹⁸O(2–1) and N₂D⁺(3–2). Both spectral line moment maps show filamentary structures, with C¹⁸O(2–1) connecting six (SVS 13A is a very close binary) protostellar sources that were identified in the VANDAM survey (Tobin et al. 2016). As seen in Figure 10, the C¹⁸O(2–1) moment 0 map has four primary peaks. Three of these peaks are mostly coincident with continuum peaks, while the fourth traces a filamentary structure protruding southeast from SVS 13C. N₂D⁺(3–2) emission is not detected in this southeast structure. N₂D⁺(3–2) is absent near SVS 13A, which is the hottest of the three sources (T_bol = 188 K, c.f., ∼30 K for the other two). N₂D⁺(3–2) peaks toward SVS 13B, and is reduced toward SVS 13C. The lack of a deuterated species is indeed expected toward hotter sources.

The bottom right panels display the continuum, which show a chain of envelopes. The projected separation for the SVS 13A and SVS 13B 1.3 mm continuum peaks is 149 (∼3500 au), while for the SVS 13B and SVS 13C peaks, the projected separation is 198 (∼4700 au).

Figure 13 shows example spectra at the peak of SVS 13B. CO(2–1) and ¹³CO(2–1) emission spectra have outflow wings. The CO(2–1) spectrum has an obvious dip between the two outflow wings, with the emission becoming negative; this emission is not likely absorption, but rather reflects confusion (due to missing zero-spacing in the uv-plane) with the large-scale emission near the systemic velocity of the local molecular cloud. The C¹⁸O(2–1) and N₂D⁺(3–2) spectra look very similar to each other. A Gaussian fit to both of these spectra gives line widths of 1 km s⁻¹, with line centers for C¹⁸O(2–1) and N₂D⁺(3–2) of 8.3 and 8.2 km s⁻¹, respectively.

In general, MASSES observations of the C¹⁸O(2–1) and N₂D⁺(3–2) lines are more compact than those seen for the SVS 13 system. More images of MASSES C¹⁸O(2–1) observations can be found in Frimann et al. (2017).