Nascent Embedded-protostar Survey in Taurus (NEST) II: Measuring Dust Mass, Disk Size, and Gas Mass

This paper presents the first dust+gas study of all consistently classified Class 0/I systems in a single star-forming region: the Nascent Embedded-protostar Survey in Taurus (NEST), which includes 26 embedded systems in Taurus (23 Class I and 3 Class 0) and excludes Flat Spectrum sources. The authors combine ALMA Band 7 (345 GHz; 870 µm) continuum and 13CO and C18O J=3–2 line observations with VLA Ka-band (33 GHz; 9 mm) continuum, with angular resolutions of 0.3″ (42 AU) at 345 GHz and 0.2″ (30 AU) at 33 GHz.

Key results:

• At 345 GHz, the median flux density, dust mass, and radius are 71 mJy, 5.5 M⊕, and 28 AU, with 68% ranges of 54–107 mJy, 3.9–9.4 M⊕, and 25–39 AU. At 33 GHz, the corresponding medians are 0.43 mJy, 39 M⊕, and 32 AU, with ranges of 0.41-0.80 mJy, 34–52 M⊕, and 29–33 AU.

• Taurus Class I disks are fainter and less massive than those in Orion, comparable to Perseus Class I disks but fainter than Perseus Class 0 disks, and brighter and more massive than those in Ophiuchus.

• Within Taurus, Class 0/I disks are brighter than Class II disks at both frequencies, although their inferred dust masses are comparable at 345 GHz and slightly higher at 33 GHz.

• Radiative-transfer modeling of CO isotopologue emission yields a median gas mass of Mgas,med = 6.7 × 10−4 M⊙.

• The resulting CO-inferred gas-to-dust ratios span a broad range, with a mean of 147±75, a median of 26, and a 16th–84th percentile range of 8–147. This distribution overlaps the Taurus Class II population at the low end and ISM-like or higher values, including the AGE-PRO Ophiuchus Class 0/I population, at the high end.

Improvements to AI Systems:

1. Multiwavelength Disk Property Estimator

• Train a model on the paired 345 GHz and 33 GHz continuum data to predict disk dust mass, radius, and optical depth from single-wavelength observations, with uncertainty quantification (e.g., Bayesian neural networks).

• Capability: Automatically classify embedded protostellar disks (Class 0/I vs. II) and estimate physical properties from future surveys (e.g., ngVLA, SKA) without requiring full radiative-transfer modeling.

2. Environment-Aware Star Formation Simulator

• Use the NEST sample (Taurus vs. Orion, Perseus, Ophiuchus) as training data to build a generative model that predicts disk mass distributions and gas-to-dust ratios as functions of star-forming region density, metallicity, and age.

• Capability: Simulate realistic protostellar populations for galaxy-scale models, improving predictions of planet formation potential in different galactic environments.

3. Gas-to-Dust Ratio Anomaly Detector

• Develop an unsupervised anomaly detection system using the broad range of CO-inferred gas-to-dust ratios (8–147) to flag disks where CO is depleted (e.g., freeze-out, photodissociation) or dust growth is enhanced.

• Capability: Automatically identify chemically or physically unusual disks in large surveys, prioritizing follow-up observations with JWST or ALMA for carbon chemistry and grain evolution studies.

4. Resolution-Aware Flux Extrapolator

• Train a convolutional neural network on the matched-resolution 345 GHz and 33 GHz images to learn the spatial scaling between dust emission at different frequencies, accounting for the 0.3″ vs. 0.2″ beam difference.

• Capability: Reconstruct high-resolution 9 mm maps from lower-resolution 870 µm data (or vice versa), enabling physical analysis of disks observed with heterogeneous arrays.

5. Evolutionary Stage Classifier with Mass Priors

• Build a classifier that combines continuum flux, spectral index, and gas mass (from CO) to assign evolutionary stages (Class 0/I/II) with probabilistic outputs, trained on the NEST labels and excluding Flat Spectrum sources.

• Capability: Automatically stage embedded protostars in real-time from ALMA/VLA observations, reducing human bias and enabling large-scale statistical studies of disk evolution.

6. Comparative Region Transfer Learner

• Use the inter-region comparisons (Taurus vs. Orion/Perseus/Ophiuchus) to develop a domain-adaptation model that corrects for environmental biases (e.g., distance, extinction, background contamination) when estimating disk properties from unresolved observations.

• Capability: Apply a model trained on one star-forming region to another with minimal re-calibration, improving cross-survey consistency for legacy datasets.

7. CO Depletion Correction Module

• Train a regression model on the gas-to-dust ratio distribution (mean 147, median 26) to predict the true gas mass from CO line luminosity, using dust mass and disk radius as auxiliary inputs to correct for CO optical depth and freeze-out effects.

• Capability: Provide more accurate gas mass estimates for planet formation models, especially for disks with low CO emission where standard assumptions fail.

Envelope-embedded protostellar disks represent the earliest stage of protoplanetary disk evolution, but their masses and sizes are difficult to measure because disk emission is entangled with the envelope. We analyze 26 protostellar disk systems in Taurus using ALMA Band 7 (345 GHz; 0.3'') and VLA Ka-band (33 GHz; 0.2'') continuum observations, together with molecular-line data to constrain disk gas masses. At 345 GHz, the median flux density, dust mass, and radius are 71 mJy, 5.5 M Mearth, and 28 AU, with 68% ranges of 54-107 mJy, 3.9-9.4 M Mearth, and 25-39 AU. At 33 GHz, the corresponding medians are 0.43 mJy, 39 M Mearth, and 32 AU, with ranges of 0.41-0.80 mJy, 34-52 M Mearth, and 29-33 AU. Taurus Class I disks are fainter and less massive than those in Orion, comparable to Perseus Class I disks but fainter than Perseus Class 0 disks, and brighter and more massive than those in Ophiuchus. Within Taurus, Class 0/I disks are brighter than Class II disks at both frequencies, although their inferred dust masses are comparable at 345 GHz and slightly higher at 33 GHz. Radiative-transfer modeling of CO isotopologue emission yields a median gas mass of 6.7 x 10-4 M Msun. The resulting CO-inferred gas-to-dust ratios span a broad range, with a mean of 147 +/- 75, a median of 26, and a 16th-84th percentile range of 8-147. This distribution overlaps the Taurus Class II population at the low end and ISM-like or higher values, including the AGE-PRO Ophiuchus Class 0/I population, at the high end.