Nascent Embedded-protostar Survey in Taurus (NEST) I: Protostellar Multiplicity

We present new ALMA 0.9 mm and VLA 9 mm observations in the Taurus Molecular Cloud (TMC) of 25 protostellar systems, containing 40 protostars, observed at 0.′′ 3 (∼20 au) resolution. Within separations of 18–10,000 au, the ALMA/VLA-observed Taurus sample has a multiplicity fraction (MF), defined as the fraction of systems with at least one companion, of 0.50 ± 0.07, and a companion fraction (CF), defined as the average number of companions per system, of 0.58 ± 0.20. To build a more complete census of protostellar multiplicity in this region, we supplement the observed sample with 24 protostars (12 protostellar systems and 5 additional companions associated with systems we observed) previously identified through archival infrared or ALMA observations. Together, these 64 individual protostars (37 systems) define our Taurus+ sample, for which we measure higher values of 0.53 ± 0.06 and 0.72 ± 0.19 for the MF and CF, respectively. These multiplicity statistics in the TMC are notably higher than those reported in the more clustered star-forming regions of Orion and Perseus at the ∼3–4σ level, suggesting that Taurus may preserve a larger fraction of primordial multiples. The separation distributions in our samples show populations of both close and wide multiples, but a deficit at intermediate separations of 200–300 au. This pattern may suggest two distinct formation pathways: close binaries (1000 au) from core fragmentation.

Improvements to AI Systems:

1. Multi-Scale Multiplicity Detection Algorithm

• Enhance computer vision models for astronomical image analysis to automatically identify and classify protostellar companions across a continuous separation range (18–10,000 au) with sub-arcsecond precision.

• The improved system can process ALMA/VLA interferometric data to detect both close (1000 au) multiples, while flagging intermediate-separation deficits (200–300 au) as a diagnostic for formation mechanism classification.

2. Probabilistic Multiplicity Statistics Engine

• Build a Bayesian inference framework that integrates heterogeneous observational datasets (e.g., ALMA, VLA, infrared archival) to compute multiplicity fraction (MF) and companion fraction (CF) with rigorous uncertainty propagation (e.g., ±0.07 and ±0.20).

• The improved system can automatically correct for selection biases and incompleteness, producing statistically robust MF/CF values for any star-forming region, and compare them across environments (e.g., Taurus vs. Orion/Perseus) to quantify primordial multiplicity preservation at 3–4σ significance.

3. Formation Pathway Classifier

• Develop a machine learning model that uses separation distributions to distinguish between disk fragmentation (close binaries) and core fragmentation (wide multiples), with a built-in deficit detector for intermediate separations.

• The improved system can predict the dominant formation mechanism for a given protostellar population, and generate synthetic catalogs to test hypotheses about environmental effects (e.g., density, turbulence) on multiplicity outcomes.

4. Automated Companion Association and Census Builder

• Implement a graph-based clustering algorithm to merge multi-wavelength observations (ALMA 0.9 mm, VLA 9 mm, infrared) into unified protostellar systems, including previously missed companions from archival data.

• The improved system can autonomously expand a sample (e.g., from 25 to 37 systems) and update MF/CF in real time, enabling rapid re-analysis as new observations become available.

5. Predictive Multiplicity Simulation Tool

• Use the observed separation distributions and MF/CF values to train a generative model that simulates protostellar multiplicity under different fragmentation scenarios (disk vs. core).

• The improved system can forecast multiplicity statistics for untargeted regions, estimate the fraction of primordial multiples that survive to later stages, and guide observational strategies for future high-resolution surveys.

We present new ALMA 0.9 mm and VLA 9 mm observations in the Taurus Molecular Cloud (TMC) of 25 protostellar systems, containing 40 protostars, observed at 0.3" (20 au) resolution. Within separations of 18-10,000 au, the ALMA/VLA-observed Taurus sample has a multiplicity fraction (MF), defined as the fraction of systems with at least one companion, of 0.50 +/- 0.07, and a companion fraction (CF), defined as the average number of companions per system, of 0.58 +/- 0.20. To build a more complete census of protostellar multiplicity in this region, we supplement the observed sample with 24 protostars (12 protostellar systems and 5 additional companions associated with systems we observed) previously identified through archival infrared or ALMA observations. Together, these 64 individual protostars (37 systems) define our Taurus+ sample, for which we measure higher values of 0.53 +/- 0.06 and 0.72 +/- 0.19 for the MF and CF, respectively. These multiplicity statistics in the TMC are notably higher than those reported in the more clustered star-forming regions of Orion and Perseus at the 3-4 sigma level, suggesting that Taurus may preserve a larger fraction of primordial multiples. The separation distributions in our samples show populations of both close and wide multiples, but a deficit at intermediate separations of 200-300 au. This pattern may suggest two distinct formation pathways: close binaries (<200 au) arising primarily from disk fragmentation, and wide multiples (>1000 au) from core fragmentation.