A CATALOG OF LOW-MASS STAR-FORMING CORES OBSERVED WITH SHARC-II AT 350 μm

Stars form in dense cores of dust and molecular gas (e.g., Di Francesco et al. 2007; Ward-Thompson et al. 2007). The ultraviolet, optical, and infrared radiation from both stars forming inside cores and the interstellar radiation field (ISRF) is absorbed by the dust within these cores, heating the dust to typical temperatures of 10–20 K (e.g., Di Francesco et al. 2007). The dust then re-radiates this emission at submillimeter and millimeter wavelengths. Thus, to study the very roots of star formation it is necessary to observe dense cores at these wavelengths.

Most nearby low-mass star-forming regions have been extensively surveyed at wavelengths between 850 μm and 1.3 mm due to the large number of bolometers available at these wavelengths and the high availability of weather suitable for ∼1 mm observations at most telescope sites (e.g., Motte et al. 1998, 2001; Testi & Sargent 1998; Johnstone et al. 2000, 2001; Shirley et al. 2000; Motte & André 2001; Young et al. 2003, 2006a, 2006b; Kirk et al. 2005; Enoch et al. 2006, 2007; Stanke et al. 2006; Kauffmann et al. 2008). Because they are both optically thin and in the Rayleigh–Jeans limit for 10–20 K dust, continuum observations at ∼1 mm are ideally suited for tracing the total dust mass and easily pick out the dense star-forming cores (e.g., Enoch et al. 2007). However, since the peaks of 10–20 K blackbodies occur at 250–500 μm, these millimeter wavelength surveys do not constrain the peaks of the spectral energy distributions (SEDs). Observations at shorter submillimeter wavelengths are needed to measure total source luminosities (Dunham et al. 2008, 2013, 2014; Enoch et al. 2009), to separate the contributions from internal (from a protostar) and external (from the ISRF) heating (Dunham et al. 2006), and to accurately classify protostars into an evolutionary sequence (Andre et al. 1993; Chen et al. 1995; Dunham et al. 2008, 2014; Frimann et al. 2015). Such observations are being provided by the Herschel Gould Belt Survey, which has obtained 70–500 μm images of all of the nearby, star-forming clouds in the Gould Belt (e.g., André et al. 2010). While this survey is providing unprecedented coverage and sensitivity of nearby star-forming regions at submillimeter wavelengths, it is doing so at relatively low angular resolution (ranging from approximately 9'' at 70 μm to 36'' at 500 μm).

In an effort to provide a submillimeter catalog of dense cores with high spatial resolution, we present in this paper 350 μm continuum observations of low-mass protostellar cores taken with the Submillimeter High Angular Resolution Camera II (SHARC-II) at the Caltech Submillimeter Observatory (CSO) on Mauna Kea, Hawaii. SHARC-II was a 350 and 450 μm "CCD-style" bolometer array of 12 × 32 pixels giving an instantaneous field of view (FOV) of 259 × 097, with the pixels filling over 90% of the focal plane and separated by their projected size on the sky of 485 (Dowell et al. 2003). With good focus and pointing, the 350 μm beam has a full-width at half-maximum (FWHM) of 85. In practice the effective resolution at 350 μm is 10'' (see Section 3.1), providing 2.5 times higher angular resolution than the Herschel Gould Belt survey at the same wavelength (albeit at lower sensitivity, especially to extended emission). Targets were selected to provide complementary data to several large surveys of nearby, low-mass star-forming regions, including the Spitzer Space Telescope c2d (Evans et al. 2003, 2009) and Gould Belt (Gutermuth et al. 2008; Dunham et al. 2015) Legacy Surveys and the Herschel Space Observatory DIGIT (e.g., Sturm et al. 2010; van Kempen et al. 2010; Green et al. 2013, 2016) open time Key Project. Some of the observations presented here were originally published by Wu et al. (2007), but we have re-analyzed and included them here using an updated version of the data reduction software and improved pointing model corrections. In general, we present observations of regions that are already well documented at other wavelengths so that this catalog will complement existing data in studies of the characteristics of protostellar regions. Though this is the primary goal of the paper, we also use our data to examine the sensitivity of SHARC-II to extended emission and to the observing mode used on the telescope, and to assess two different methods of classifying protostars.

We organize this paper as follows. First we describe the target selection and observation strategy in Sections 2.1 and 2.2, respectively. We then discuss the data reduction and calibration processes in Sections 3.1 and 3.2, respectively. We discuss the source extraction procedure in Section 3.3, including an analysis of the effects of optimizing the data reduction pipeline for the recovery of extended emission. We present our source catalog in Section 4, including source positions, flux densities, and radii. A comparison of the results from two different observing modes is presented in Section 5. In Section 6 we discuss the sensitivity of our observations to extended emission, and in Section 7 we investigate the effects of including SHARC-II 350 μm photometry when classifying protostars. A summary of our results is presented in Section 8.

2.1. Target Selection

2.2. Observations

3.1. Data Reduction

The data were reduced using the Comprehensive Reduction Utility for SHARC-II (CRUSH) version 2.12-1, a publicly available¹⁰ , Java-based software package that iteratively solves for both the source signal and the various correlated noise components (e.g., Kovács 2008a, 2008b). For increased redundancy, we use CRUSH to add together the bolometer time streams from individual scans before obtaining the solutions, taking into account various noise elements and changing atmospheric conditions for each scan. The only exception to this general method is NGC 1333, which had to be broken into five smaller pieces first due to computer memory limitations. These five pieces were then coadded together using the coadd function of the CRUSH package, after which point they were handled exactly the same way as the other sources. As described in more detail in Section 3.3 below, two versions of each map were produced: one with the extended flag given to CRUSH to optimize the software pipeline for the recovery of extended emission, and one without it given. The differences between and uses of the two versions of each map are discussed below.

The atmospheric opacity during each scan was determined from an online database¹¹ of measurements of the zenith optical depth at 225 GHz, ${\tau }_{225\mathrm{GHz}}$, obtained in ten minute intervals. This database includes polynomial fits to ${\tau }_{225\mathrm{GHz}}$ versus time for each night, where the orders of the polynomials were treated as free parameters. The orders range from 3 to 90 over the full database, but are typically less than 20, with a mean value of 13. We visually inspected plots of ${\tau }_{225\mathrm{GHz}}$ and the resulting polynomial fits for each night to verify that the fits accurately trace the variations in ${\tau }_{225\mathrm{GHz}}$ throughout the night. CRUSH uses these polynomial fits to calculate the optical depth at the time each observation scan was taken and uses this value to correct the signal. Pointing corrections were applied to each scan with CRUSH to correct for residual telescope pointing errors. These corrections were determined using a publicly available model¹² fit to all pointing data from that observing run. We applied this model using the option to also correct for a random drift with time by evaluating model residuals for pointing scans taken within a few hours before and after each science scan. The final maps were generated with 15 pixels. A Gaussian smoothing function with a FWHM of 4 pixels was applied to each map to reduce pixelation artifacts, resulting in a final effective beam of approximately 10''.

In order to determine the residual pointing uncertainty after applying the pointing model corrections, Figure 1 shows a histogram of the distances between measured peak positions for several sources that were observed twice on two different dates, after applying the pointing model corrections. The mean distance is 29. In order to interpret this in terms of a residual rms uncertainty in the pointing model, we construct a Monte Carlo model in which a source is placed on a polar coordinate grid twice, with each angle drawn randomly from a uniform distribution and each radius drawn randomly from a Gaussian distribtuion with σ = 1. Note that this σ has no units as we are only concerned with obtaining a dimensionless ratio that will characterize the pointing model uncertainty, as described below. The distribution of distances between the two "observations" of each source in the Monte Carlo model is shown in Figure 2; this distribution has a mean of 1.2. Since the input to the simulation was a pointing model with an assumed rms of 1, and the resulting mean of the distance distribution is 20% larger, we thus infer that our observed mean distance between peak positions of 29 implies an underlying pointing model residual rms of 24. Given the small number of sources observed twice and uncertainties in the assumptions of the Monte Carlo model, we thus conservatively estimate that the pointing model corrections are good to within ∼3''.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Once the maps were created using CRUSH, we used the imagetool function of the CRUSH package to eliminate map edges with increased noise by removing all pixels in the map with less than 25% of the maximum integration time. The average 1σ rms for each map was calculated by using the sky routine in the IDL Astronomy Library to measure the standard deviation of all off-source pixels and then calibrating with the peak calibration factor, as defined below.

3.2. Calibration

While CRUSH adopts an approximate calibration factor to produce maps in calibrated units of Jy beam⁻¹, it does not account for the fact that the instrument calibration changes both with observing conditions and randomly with time.¹³ Furthermore, the presence of beam sidelobes mean that the flux of a point source measured in apertures of increasing size will also increase, whereas ideally the flux of a point source should be independent of aperture size. We thus recalibrate all of our data as follows.

To derive calibration factors, we used observations of Mars, Neptune, and Uranus, which were observed every few hours to check and update the telescope pointing. These planet scans, which we hereafter refer to as calibration scans, were reduced with CRUSH and used to calculate the calibration factors. All calibration scans were observed using the Lissajous observing mode on the telescope, regardless of what observing mode was used for the science sources to which these calibrations were applied. As shown in Section 5, the calibration factors do not depend on observing mode, validating this strategy. For each calibration scan, we measured the peak intensity and flux densities in 20'' and 40'' diameter apertures using custom IDL routines. We choose these aperture sizes to match previous (sub)millimeter continuum surveys that measure flux densities in standard apertures of 20'', 40'', 80'', and 120'' in diameter (e.g., Enoch et al. 2006, 2007; Young et al. 2006a, 2006b; Wu et al. 2007; Kauffmann et al. 2008). Here we only adopt the two smallest apertures since our calibration images are too small to use larger apertures. By comparing the measured flux densities to the known fluxes of these planets, we calculate three calibration factors: C_peak, C_20'', and C_40''. C_peak is simply the factor required to obtain calibrated maps in units of Jy beam⁻¹. Since the maps are already calibrated with an approximate calibration factor, the values of C_peak are unitless and are generally close to unity, as seen below. C_20'' and C_40'' are "aperture calibration factors" and have units of Jy/(Jy beam⁻¹) = beam. Multiplying the flux densities measured through aperture photometry by the aperture calibration factors for the same size aperture will give the flux density in that aperture in Jy. As mentioned above, this method corrects for the beam sidelobes such that the measured flux density of a point source is independent of aperture size (e.g., Shirley et al. 2000; Enoch et al. 2006).

In practice, C_peak, C_20'', and C_40'' are calculated as follows. For C_peak, the expected peak intensity from the planet was calculated by convolving the SHARC-II beam (assumed to be Gaussian) with a disk of uniform brightness, using the known size and flux density of each planet on that observation date. The resulting values are then divided by the measured peak intensity in each calibration scan. For C_20'' and C_40'', the known total flux density of the planet on each observation date was divided by the measured flux density in 20'' and 40'' diameter apertures, respectively, using the uncalibrated maps. The flux densities were measured using standard aperture photometry with no sky subtraction since CRUSH removes the background sky emission. Table 2 lists our derived calibration factors for each observation night, and Table 3 lists the means and standard deviations of the calibration factors for each observing run. Each map is calibrated using the mean calibration factors for that run; maps consisting of scans obtained over multiple runs are calibrated using the mean values over those runs. For maps observed in runs where no planet scans were taken, the mean calibration factors over all runs were used.

Table 2.  Calibration Scans and Factors

Date	Calibrator	Cpeak	C20	C40
2003 May 17	Mars	1.3	0.021	0.015
2003 May 18	Uranus	1.2	0.027	0.022
2003 Sep 29	Uranus	1.7	0.038	0.031
2003 Sep 30	Uranus	2.3	0.045	0.034
2003 Oct 1	Uranus	1.8	0.038	0.031
2004 Jun 19	Uranus	2.5	0.049	0.035
2004 Jun 19	Uranus	2.3	0.044	0.032
2004 Jun 19	Neptune	1.9	0.044	0.036
2004 Jun 19	Neptune	1.9	0.040	0.030
2004 Sep 25	Uranus	1.6	0.034	0.019
2005 Jun 16	Uranus	1.6	0.037	0.026
2005 Jun 16	Uranus	1.3	0.032	0.022
2005 Jun 16	Neptune	1.6	0.038	0.026
2005 Jun 17	Uranus	1.3	0.032	0.023
2005 Jun 17	Neptune	1.2	0.032	0.024
2005 Nov 3	Uranus	1.2	0.028	0.020
2005 Nov 3	Uranus	1.4	0.026	0.020
2005 Nov 3	Uranus	1.2	0.032	0.025
2005 Nov 4	Uranus	1.4	0.027	0.020
2005 Nov 4	Uranus	1.6	0.032	0.025
2005 Nov 4	Uranus	1.3	0.034	0.025
2005 Nov 5	Uranus	1.7	0.032	0.025
2005 Nov 5	Uranus	1.2	0.032	0.025
2005 Nov 5	Uranus	1.4	0.038	0.028
2005 Nov 12	Uranus	1.4	0.031	0.021
2005 Nov 12	Uranus	1.5	0.032	0.023
2005 Nov 12	Uranus	1.6	0.035	0.024
2006 Dec 14	Uranus	1.2	0.034	0.026
2006 Dec 15	Uranus	1.3	0.027	0.022
2006 Dec 15	Uranus	1.3	0.029	0.023
2006 Dec 15	Uranus	1.3	0.028	0.022
2007 Oct 21	Mars	1.3	0.021	0.016
2007 Oct 21	Uranus	1.7	0.038	0.031
2007 Oct 21	Uranus	1.7	0.037	0.030
2007 Oct 21	Uranus	1.5	0.034	0.027
2007 Oct 21	Neptune	1.5	0.034	0.030
2007 Oct 21	Neptune	1.5	0.034	0.029
2007 Oct 21	Neptune	1.6	0.036	0.031
2007 Oct 21	Neptune	1.5	0.034	0.029
2007 Oct 21	Neptune	1.5	0.034	0.029
2007 Oct 21	Neptune	1.5	0.033	0.029
2007 Oct 22	Uranus	1.5	0.031	0.025
2007 Oct 22	Uranus	1.4	0.032	0.026
2007 Oct 22	Neptune	1.5	0.036	0.031
2007 Oct 23	Mars	1.9	0.029	0.022
2007 Oct 23	Mars	1.8	0.029	0.022
2007 Oct 23	Mars	1.8	0.028	0.021
2007 Oct 23	Uranus	2.1	0.044	0.036
2007 Oct 23	Neptune	1.7	0.086	0.032
2007 Oct 23	Neptune	1.6	0.037	0.030
2007 Oct 23	Neptune	1.2	0.028	0.023
2007 Oct 23	Neptune	1.4	0.031	0.025
2007 Oct 24	Uranus	1.4	0.031	0.026
2007 Oct 24	Uranus	1.6	0.034	0.027
2007 Oct 24	Neptune	1.7	0.040	0.032
2007 Oct 24	Neptune	1.7	0.038	0.032
2007 Oct 27	Uranus	1.3	0.029	0.024
2007 Oct 27	Uranus	1.4	0.031	0.026
2007 Oct 27	Uranus	1.3	0.029	0.024
2008 Jul 3	Uranus	1.4	0.030	0.023
2008 Jul 3	Uranus	1.3	0.029	0.023
2008 Jul 3	Neptune	1.2	0.030	0.026
2008 Jul 3	Neptune	1.1	0.028	0.022
2008 Jul 3	Neptune	1.4	0.034	0.028
2008 Sep 30	Uranus	1.5	0.032	0.025
2008 Oct 3	Uranus	1.3	0.030	0.024
2008 Oct 4	Uranus	1.6	0.035	0.027
2008 Oct 4	Neptune	1.5	0.035	0.028
2009 Sep 8	Uranus	1.5	0.032	0.026
2009 Sep 8	Uranus	1.4	0.032	0.026
2009 Sep 8	Neptune	1.5	0.035	0.029
2009 Sep 8	Neptune	1.2	0.030	0.025
2009 Sep 9	Uranus	1.3	0.030	0.025
2009 Sep 9	Uranus	2.0	0.044	0.034
2009 Sep 9	Neptune	1.5	0.037	0.028
2009 Sep 9	Neptune	1.5	0.034	0.028
2009 Sep 9	Neptune	1.4	0.034	0.029
2009 Sep 9	Neptune	1.4	0.033	0.027
2009 Sep 10	Uranus	1.1	0.024	0.019
2009 Dec 29	Uranus	1.0	0.024	0.019
2009 Dec 30	Uranus	1.3	0.030	0.023
2009 Dec 30	Uranus	1.2	0.029	0.023
2010 Jul 23	Uranus	1.5	0.034	0.028
2010 Jul 23	Uranus	1.4	0.033	0.027
2010 Jul 23	Uranus	1.5	0.033	0.027
2010 Jul 23	Uranus	1.1	0.026	0.021
2010 Jul 23	Uranus	1.1	0.024	0.020
2010 Jul 23	Uranus	1.2	0.028	0.023
2010 Jul 23	Neptune	1.1	0.026	0.022
2010 Jul 23	Neptune	1.0	0.026	0.022
2010 Jul 25	Uranus	1.2	0.023	0.017
2010 Jul 25	Uranus	1.6	0.028	0.019
2010 Jul 25	Neptune	2.0	0.039	0.032
2010 Jul 28	Uranus	1.2	0.026	0.022
2010 Jul 28	Neptune	1.1	0.027	0.023
2010 Jul 31	Uranus	1.3	0.028	0.022
2010 Jul 31	Neptune	1.6	0.037	0.031
2010 Jul 31	Neptune	1.5	0.036	0.029
2010 Jul 31	Neptune	1.5	0.034	0.028
2010 Jul 31	Neptune	1.4	0.033	0.027
2010 Jul 31	Neptune	1.5	0.034	0.027
2010 Jul 31	Neptune	1.5	0.034	0.027
2010 Jul 31	Neptune	1.5	0.035	0.028
2010 Aug 1	Uranus	1.3	0.029	0.023
2010 Aug 1	Neptune	1.3	0.032	0.025
2010 Aug 1	Neptune	1.3	0.031	0.025
2010 Aug 1	Neptune	1.4	0.032	0.025
2010 Aug 1	Neptune	1.3	0.032	0.025
2010 Dec 6	Uranus	1.4	0.031	0.022
2010 Dec 6	Uranus	1.4	0.031	0.023
2010 Dec 6	Uranus	1.4	0.031	0.023
2010 Dec 7	Uranus	1.3	0.027	0.020
2010 Dec 7	Uranus	1.3	0.028	0.021
2010 Dec 7	Uranus	1.4	0.028	0.021
2010 Dec 7	Uranus	1.3	0.030	0.022
2010 Dec 7	Uranus	1.4	0.031	0.022

Download table as:  ASCIITypeset images: 1 2

Table 3.  Average Calibrator Factors Per Observing Run

Date	Cpeak	C20	C40
2003 May	1.259 ± 0.035	0.024 ± 0.004	0.019 ± 0.005
2003 Sep/Oct	1.948 ± 0.352	0.041 ± 0.004	0.032 ± 0.002
2004 Jun	2.159 ± 0.294	0.044 ± 0.004	0.033 ± 0.003
2004 Sep	1.597 ± 0.000	0.034 ± 0.000	0.019 ± 0.000
2005 Jun	1.409 ± 0.159	0.034 ± 0.003	0.024 ± 0.002
2005 Nov	1.424 ± 0.168	0.032 ± 0.003	0.024 ± 0.003
2006 Dec	1.243 ± 0.056	0.029 ± 0.003	0.023 ± 0.002
2007 Oct	1.558 ± 0.191	0.035 ± 0.011	0.027 ± 0.004
2008 Jul	1.305 ± 0.132	0.030 ± 0.002	0.024 ± 0.002
2008 Sep/Oct	1.333 ± 0.306	0.023 ± 0.016	0.018 ± 0.012
2009 Sep	1.440 ± 0.220	0.033 ± 0.005	0.027 ± 0.004
2009 Dec	1.177 ± 0.135	0.028 ± 0.003	0.022 ± 0.002
2010 Jul/Aug	1.362 ± 0.213	0.031 ± 0.004	0.025 ± 0.004
2010 Dec	1.377 ± 0.053	0.030 ± 0.002	0.022 ± 0.001
All Runs	1.462 ± 0.258	0.033 ± 0.007	0.025 ± 0.004

Download table as:  ASCIITypeset image

Figure 3 shows histograms for each of the calibration factors. To investigate whether the derived calibration factors vary with time or depend on the properties of the calibration source, Figures 4–6 plot the calibration factors versus observation date, angular size of the calibration source, and total flux of the calibration source. No systematic variations are seen, and linear least squares fits to each panel in Figures 5 and 6 give better fits (lower reduced χ² values) for zeroth order polynomial than for first order polynomials, indicating that any dependences of the calibration factors on source properties are smaller than the overall calibration uncertainties. We calculate these calibration uncertainties by dividing the standard deviations by the means, resulting in values of 18% for C_peak, 21% for C_20'', and 17% for C_40''. We thus conservatively adopt an overall calibration uncertainty of 25%.

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

3.3. Source Extraction

As noted above, two versions of each map were produced, one with and one without the extended flag given to CRUSH. This flag optimizes the CRUSH pipeline for extended sources and results in maps that preserve extended emission at the cost of increased noise (see Kovács 2008a, for details). Since our targets are star-forming regions expected to feature extended emission, we investigated the possibility of using this flag to recover more extended emission. However, we found that the added noise is dominated by sky noise that is temporally correlated in the time stream and thus spatially correlated in the final maps, rather than random, and thus especially problematic for source extraction. Many false sources are detected and extracted regardless of the detailed implementation of source extraction. Figure 7 shows an example of a map reduced with and without the extended flag. Similar color figures are presented for all of the regions mapped below. All maps are displayed with a linear intensity scaling with the minimum (black) and maximum (white) intensities given in their respective figure captions. Figure 8 displays a normalized intensity color scale bar that, combined with the minimum and maximum intensities given in the figure captions, can be used to determine the absolute intensities of each map.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 7 illustrates that the extended flag results in large areas of correlated noise that are extracted as sources. Thus, in order to ensure a reliable catalog of sources, we perform source extraction on the maps produced without the extended flag (such as the one shown in the right panel of Figure 7). We extracted sources from each map using the Bolocat¹⁴ source extraction routine (Rosolowsky et al. 2010). Bolocat works by identifying regions of statistically significant emission based on their significance with respect to a local estimate of the noise in the maps. These regions of high significance are then subdivided into multiple sources based on the presence of local maxima within the originally defined regions, with each pixel assigned to one of the sources using a seeded watershed algorithm, similar to the Clumpfind or Source EXtractor algorithms (Williams et al. 1994; Bertin & Arnouts 1996). Bolocat was previously used to extract sources from SHARC-II images of massive star-forming clumps by Merello et al. (2015), demonstrating the feasibility of using this source extraction routine on data from the SHARC-II instrument.

Bolocat requires three input parameters, all of which are measured in units of the map rms: P_amp, the minimum required amplitude for a source to be extracted; P_base, the base level of emission out to which the initial detected source is expanded; and P_deb, a source deblending parameter. In practice, Bolocat first masks all regions of the map below P_amp and extracts one or more initial sources based on the number of regions of contiguous pixels remaining in the masked map. It then expands each of these initial sources to encompass all contiguous pixels down to the level P_base, with the rationale being that marginally significant emission levels spatially connected to those of higher significance are likely to be real. Finally, each initial source is deblended into one or more sources by finding pairs of local maxima and identifying them as separate sources if the lower of the two is at least P_deb above the highest contour level that contains both maxima (see Rosolowsky et al. 2010, for details). We adopted values of P_amp = 3, P_base = 3, and P_deb = 2 based on matching "by-eye" extractions in an initial exploration of the parameter space (note that, since P_amp = P_base, we did not expand the sources beyond their initial detections). While Merello et al. (2015) also adopted P_amp = 3, they adopted lower values of the other two parameters (P_base = 1 and P_deb = 1). We found that the adoption of such low values for P_base and P_deb resulted in extractions that did not match either our best "by-eye" results or known sources from catalogs at other wavelengths. In particular, the lower value of P_deb resulted in several objects that were clearly single being broken into multiple objects due to small noise variations. The advantage of studying known regions with copious multiwavelength data in the literature, as opposed to Merello et al. (2015), allowed us to refine the best values of the source extraction parameters.

After extracting sources and measuring their deconvolved radii using the maps reduced without the extended flag, aperture photometry was performed on the maps produced with the extended flag, at the peak position of each extracted source. We used custom IDL routines to measure the peak intensities and flux densities in 20'' and 40'' diameter apertures. In cases of overlapping apertures with nearby sources, only those up to the largest in which overlap did not occur were kept. We measured the uncertainties in the flux densities in each aperture by adding in quadrature the statistical uncertainty from the measurement itself and the overall calibration uncertainty of 25%.

This method of using both sets of maps ensures that a reliable catalog of sources is extracted while also preserving as much extended emission as possible in the measured flux densities. Figure 9 shows the distribution of ratios of flux densities measured in the maps with the extended flag to those measured in the maps without the extended flag, for the peak intensities as well as the flux densities measured in 20'' and 40'' diameter apertures. The mean (median) ratios are 1.7 (1.5), 2.2 (1.8), and 3.4 (2.6) for the peak, 20'', and 40'' measurements. Thus, as expected, we see that the extended flag improves the flux recovery, especially in the larger apertures that are more sensitive to the amount of extended emission.

Zoom In Zoom Out Reset image size

Figures 10–30 show contour maps overlaid on images for each of the 81 regions listed in Table 1. Since the source extraction routine uses a local noise measurement but the contours are plotted using a global noise measurement, weak sources in low-noise regions may lack associated 3σ contours. The images are of the maps reduced without the extended emission flag. Most maps are displayed in six-panel figures except for the largest and most crowded maps, which are instead displayed as one-panel figures. Maps with multiple sources have each source labeled. All of the reduced FITS files are available in calibrated units of Jy beam⁻¹ following the calibration procedures described above; versions with and without the extended flag can be accessed through the Data Behind the Figures (DBF) feature of the journal.

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

We detect a total of 164 sources in the 81 maps listed in Table 1 and presented in Figures 10–30. Table 4 lists, for each detected source, the name of the source, the map in which the source is covered, the peak position of the source, the deconvolved angular source radius as determined by Bolocat (see Rosolowsky et al. 2010, for details), the peak intensity, the flux density in a 20'' diameter aperture, the flux density in a 40'' diameter aperture, and a flag noting whether each source is starless or protostellar, determined mostly from a search for an infrared point source in Spitzer Space Telescope images from the c2d (e.g., Evans et al. 2009) and Gould Belt (e.g., Dunham et al. 2015) legacy projects. The peak intensities are given in units of Jy beam⁻¹ (for the SHARC-II beam, 1 Jy beam⁻¹ = 519.7 MJy sr⁻¹).

As noted in Section 2.2, all observations prior to 2006 were taken in the Lissajous scan mode. However, over the course of taking observations we were made aware of a different observing mode, the box scan mode, that might yield better results. In particular, maps observed in the box scan mode have much larger areas than those observed in the Lissajous mode. Since all emission on scales larger than the map size will be treated as sky emission and removed by CRUSH, the larger map areas provided by the box scans may provide better recovery of extended emission. The use of the box scan mode also allows us to map larger areas in reasonable amounts of time. Thus, all science data obtained during and after the 2008 July observing run were obtained exclusively in the box scan mode, with observations between 2006 December and 2008 July obtained in both modes for testing purposes.

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

Calibration scans were taken with the Lissajous mode, even after 2008 July, since one Lissajous mode scan can be completed in much less time than one box scan. To justify this choice, we obtained several calibration scans in the box-scan observing mode. The calibration factors obtained from these scans are listed in Table 5, using the same method as above for the Lissajous calibration scans. Averaged over all of the box calibration scans, we calculate C_beam = 1.63 ± 0.15, C₂₀ = 0.034 ± 0.004, and C₄₀ = 0.025 ± 0.003, where the uncertainties are the standard deviations. Comparison to Table 3 shows that the mean Lissajous and box-scan calibration factors agree to within 10%, well within the overall calibration uncertainty of 25%, indicating that any dependence on observing mode in the SHARC-II instrument calibration has a negligible impact on our results.

We also compared the SHARC-II beams resulting from the Lissajous and box scans. Figure 31 compares the size and shape of the beams derived from the Lissajous and box calibration scans. The beam profiles are derived by deconvolving the measured properties (sizes and elongations) of the calibration sources with their known, intrinsic properties. For the beam FWHM, the means and standard deviations of the means are 93 ± 02 for the Lissajous scans and 100 ± 02 for the box scans. For the beam aspect ratio, the means and standard deviations of the means are 1.06 ± 0.01 for the Lissajous scans and 1.12 ± 0.01 for the box scans. Thus, observations in the box scan mode have a beam profile that is, on average, 8% larger and 6% more elongated than observations in the Lissajous mode. While the degradation of the beam profile in the box scan mode is a statistically robust finding, resulting from both the faster scan rates used by the box scan modes and astrometric errors in the SHARC-II pixel plate scale that build up over larger scan areas, the overall effects are less than 10% and have no significant impact on our results.

Zoom In Zoom Out Reset image size

Finally, we also investigated whether the measured flux densities of our science targets depended on the observing mode. To do this, we observed several science sources in both observing modes. For the 23 maps observed in both modes, Tables 1 and 4 only present information for the box scans; the target and source information for the Lissajous scans are given in Tables 6 and 7. We extracted sources from these extra Lissajous maps and measured flux densities using the same methods as described above. Figures 32–35 show contour maps overlaid on images for these extra Lissajous maps, using the maps reduced without the extended emission flag. All of the reduced FITS files used to produce these figures are available through the Data Behind the Figures (DBF) feature of the journal. These maps are given in calibrated units of Jy beam⁻¹ following the calibration procedures described above, and both the versions with and without the extended flag are provided. For the sources detected in both modes, Figure 36 plots the ratios of the flux densities calculated in maps observed in the box mode to those observed in the Lissajous mode versus the radius of each source determined from the box scan observations. The means and standard deviations of these ratios are 0.93 ± 0.18, 0.97 ± 0.17, and 1.09 ± 0.29 for the peak intensities, 20'' flux densities, and 40'' flux densities, respectively. In all three cases the mean ratios are within 10% of unity. Given the overall calibration uncertainty of 25% and the fact that the ratios show no dependence on source radius, we conclude that similar amounts of flux are recovered between the two observing modes on scales up to at least 40''.

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 addition to the detected sources listed in Table 4, Table 8 lists an additional 48 cores covered by our maps but not detected in our SHARC-II observations. These are cores identified by other observations at submillimeter and millimeter wavelengths, based on SIMBAD¹⁵ searches of the total area covered by our maps. Combining the information from these tables, our maps cover a total of 137 protostellar cores, 130 of which are detected, and a total of 75 starless cores, 34 of which are detected. Our detection rate of 95% for protostellar cores is thus much higher than our detection rate of 45% for starless cores. These results suggest that SHARC-II is well suited for identifying and characterizing protostellar cores, but is not ideal for studying starless cores.

Of the maps observed in both the Lissajous and box scan modes, there are six starless cores. Two are detected in both modes (L1455-IRS2E and B18-4), two are detected in only the box scan mode (L1544 and L694-2), and two are undetected in both observing modes (Perseus Bolo 107 and L1521B). The lower detection rate for starless cores in the Lissajous mode (33%) versus the box scan mode (45%), coupled with the fact that two starless cores detected in the box scan mode are not detected in the Lissajous mode, suggest that the box scan mode is somewhat better suited to detecting starless cores, although we caution that the sample sizes are very small.

The low detection rate for starless cores is likely explained by the fact that, compared to protostellar cores, starless cores feature flatter density profiles and colder temperatures in their central regions (e.g., Ward-Thompson et al. 2007; Evans et al. 2001). Consequently starless cores exhibit 350 μm intensity profiles that are significantly shallower and less centrally condensed than those for protostellar sources, as confirmed by Wu et al. (2007) using simple, one-dimensional radiative transfer models. The more extended nature of their emission profiles make them harder to separate from sky emission, thus they are less reliably detected. To further quantify this effect, we examined all of the starless cores covered by our maps that were identified in 1.1 mm Bolocam surveys of Perseus (Enoch et al. 2006), Ophiuchus (Young et al. 2006b), and Serpens (Enoch et al. 2007). Figure 37 shows histograms of the core masses, sizes, and mean densities, as derived from the Bolocam observations, for both the detected and undetected populations of starless cores in our dataset. The detected starless sources span nearly the full range of masses, FWHM angular sizes, and mean densities, indicating there is no one unique property that determines the detectability of the core.

Zoom In Zoom Out Reset image size

Figure 38 plots the FWHM angular size versus core mass for these same starless cores, again taking the properties from the 1.1 mm Bolocam surveys cited above. Inspection of this figure shows that, for low-mass cores (M < ∼ 2 M_⊙), only the most compact cores are detected. These cores will have the steepest intensity profiles among all cores with such masses, allowing them to be separated from sky emission and reliably detected. Only for relatively high-mass (and thus relatively bright) cores (M > ∼3 M_⊙) are more extended cores able to be separated from sky emission and detected by SHARC-II observations. Thus, starless cores revealed by SHARC-II surveys of star-forming regions are biased toward the most compact or highest mass cores, and even in cases where starless cores are detected, the extended nature of their intensity profiles means that the measured flux densities are likely lower limits to the true flux densities. These cautions should be kept in mind when interpreting the results from Zhang et al. (2015), who derive a prestellar core mass function based on SHARC-II observations of Ophiuchus. This core mass function for detected prestellar cores may not be representative of the full population of such cores in this cloud, and furthermore the measured masses of the detected prestellar cores likely underestimate their true masses. Since the detectability of a starless core with SHARC-II is not a simple function of the total flux density of the core, but also its emission profile, measured upper limits for undetected cores do not necessarily represent true limits to the flux densities of these cores. While we do list the 1σ upper limits in Table 8 (taken directly from Table 1), this caution should be kept in mind when interpreting the non-detections.

Zoom In Zoom Out Reset image size

While we caution that the measured flux densities of starless cores are likely lower limits, the measured values for the protostellar cores are much more reliable. Wu et al. (2007) used simple, one-dimensional radiative transfer models to show that protostellar cores exhibit significantly steeper intensity profiles compared to starless cores, even for protostars with luminosities as low as 0.1 L_⊙. These steeper intensity profiles lead to more compact emission that is fully recovered by SHARC-II, at least on scales up to the 40'' considered here. This is confirmed by comparing the SHARC-II observations of IRAM 04191+1522 and L1521F, two protostars with luminosities less than 0.1 L_⊙, to published radiative transfer models; in both cases the observed flux densities in 40'' diameter apertures agree with the models to within 2σ (Bourke et al. 2006; Dunham et al. 2006).

Protostars are commonly classified into one of two classes, Class 0 and Class I, based on observational signatures that trace the underlying evolutionary state. Class 0 protostars were first defined by Andre et al. (1993), who defined such objects observationally as protostars emitting a relatively large fraction (greater than 0.5%) of their total luminosity at wavelengths λ ≥ 350 μm. Defining such luminosity as the submillimeter luminosity, L_smm, Class 0 objects are then protostars with L_smm/L_bol > 0.005. The corresponding physical Stage 0 objects are young, embedded protostars with greater than 50% of their total system mass still in the core (Andre et al. 1993). Another quantity used to classify protostars is the bolometric temperature T_bol, defined by Myers & Ladd (1993) as the temperature of a blackbody with the same flux-weighted mean frequency as the source. By calculating T_bol for a large sample of young stars, Chen et al. (1995) showed that Class 0 objects have T_bol < 70 K whereas Class I protostars have T_bol ≥ 70 K. Since L_smm has historically been difficult to calculate accurately due to the difficulty in obtaining high-quality submillimeter data from the ground, particularly at 350 μm, the T_bol criterion introduced by Chen et al. is often used instead for classifying protostars into their two Classes (e.g., Enoch et al. 2009; Dunham et al. 2013; Tobin et al. 2016). However, as several studies have shown that L_smm/L_bol is less sensitive to viewing geometry and a better tracer of underlying physical stage than T_bol (Andre et al. 2000; Young & Evans 2005; Dunham et al. 2010b; Frimann et al. 2015), protostellar classification must be revisited as additional data becomes available.

With the 350 μm photometry presented here, we can now accurately calculate (to within 20%–60%; see Dunham et al. 2008; Enoch et al. 2009, for details) both T_bol and L_smm/L_bol and compare classification via the two quantities. To ensure as uniform a dataset as possible, we consider only the protostellar cores in Perseus, and construct SEDs for each source consisting of Spitzer Space Telescope 3.6–70 μm photometry from Evans et al. (2009) and Dunham et al. (2015) and Bolocam 1.1 mm photometry from Enoch et al. (2006). We calculate T_bol twice, once without the SHARC-II 350 μm photometry included and once with it included, and we also calculate L_smm/L_bol.

Figure 39 compares the two values of T_bol, calculated with and without the 350 μm photometry included. Leaving out the 350 μm photometry increases the value of T_bol, with the effect growing in significance with decreasing evolutionary stage (colder values of T_bol). This result is explained by the fact that, for more deeply embedded protostars, the emission peaks at longer wavelengths, and more of this emission is lost when no submillimeter photometry is available. As demonstrated by Figure 39, Class 0 protostars with very low values of T_bol can masquerade as more evolved objects when 350 μm photometry is lacking. These results are in qualitative agreement with earlier investigations by Dunham et al. (2008) and Enoch et al. (2009).

Zoom In Zoom Out Reset image size

Figure 40 plots L_bol/L_smm versus T_bol for these same protostars as a way of comparing the two different classification methods. Such a comparison is only possible when 350 μm data is available, since accurate calculation of L_smm requires 350 μm data. With the T_bol Class boundaries as defined by Chen et al. (1995) and the L_smm/L_bol Class boundaries as defined by Andre et al. (1993), the two classification methods agree for only 65% of the protostars considered (35 out of 54). In particular, there are many protostars classified as Class I by T_bol (T_bol > 70 K) but Class 0 by L_smm/L_bol (L_smm/L_bol > 0.005). Revised L_smm/L_bol Class 0/I boundaries have been proposed in the literature, including 0.01 (Andre et al. 2000; Sadavoy et al. 2014) and 0.03 (Maury et al. 2011). Adopting these boundaries increases the agreement between classification methods to 67% and 74%, respectively.

Zoom In Zoom Out Reset image size

While the primary focus of this publication is to provide a catalog of SHARC-II 350 μm observations of nearby star-forming regions, these results demonstrate the critical role played by 350 μm observations in both accurate classification of protostars and assessing the reliability of different classification methods in tracing the underlying evolutionary stages of protostars. A more complete investigation of protostellar classification, using complete far-infrared and submillimeter SEDs provided by both ground-based surveys such as this effort and surveys with the Herschel Space Observatory, will be presented in a future paper (M. M. Dunham et al. 2016, in preparation).

In this paper we have presented a catalog of low-mass star-forming cores observed with the SHARC-II instrument at 350 μm. Our observations have an effective angular resolution of 10'', approximately three times higher than observations at the same wavelength obtained with the Herschel Space Observatory. A summary of our results is as follows:

• 1.  
  We present 81 maps covering a total of 164 detected sources. We tabulate basic source properties including position, peak intensity, flux density in fixed apertures, and radius.
• 2.  
  We examine the uncertainties in the pointing model applied to all SHARC-II data and conservatively find that the model corrections are good to within ∼3'', approximately 1/3 of the SHARC-II beam.
• 3.  
  We examined the differences between the Lissajous and box scan observing modes. We find that the calibration factors, beam size, and beam shape are similar between the two modes, and we also show that the same flux densities are measured when sources are observed in the two different modes. Thus we conclude that there are no systematic effects in our catalog introduced by switching observing modes during the course of taking observations.
• 4.  
  We find that less than half of the starless cores observed are detected by SHARC-II (45% to be precise), and show that the detections are biased toward the most compact or highest mass starless cores. We argue that, even for the detected starless cores, the measured flux densities are likely lower limits to the intrinsic flux densities.
• 5.  
  For protostellar cores, our SHARC-II observations fully recover the emission, at least up to the 40'' scales considered here.
• 6.  
  We demonstrate that the inclusion of 350 μm photometry significantly improves the accuracy of calculated values of T_bol, and enables comparison between two different measures of protostellar Class, T_bol and L_smm/L_bol. The latter can only be calculated when 350 μm photometry is available.

We thank the referee for helpful comments that have improved the quality of this publication. We gratefully acknowledge the assistance provided by the staff of the CSO in obtaining the observations presented here. We also acknowledge the numerous students and postdocs from the University of Texas at Austin who participated in observing runs over the years, and we thank Darren Dowell and Attila Kovács for technical assistance with SHARC-II and CRUSH. Finally, we express our profound gratitude to everybody who played a role in the construction, commissioning, and operation of the CSO over its three decades of operation. This work is based on data obtained with the Caltech Submillimeter Observatory (CSO), which was operated by the California Institute of Technology under cooperative agreement with the National Science Foundation (AST-0838261). This publication makes use of data products from the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. These data were provided by the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. This research has made use of NASA's Astrophysics Data System (ADS) Abstract Service, the IDL Astronomy Library hosted by the NASA Goddard Space Flight Center, and the SIMBAD database operated at CDS, Strasbourg, France. MMD acknowledges support from the Submillimeter Array as an SMA postdoctoral fellow, and from NASA ADAP grant NNX13AE54G. NJE acknowledges support from NSF Grant AST-1109116 to the University of Texas at Austin.