Trace the Self-Gravitating Gas Using CO Isotopologues

The paper presents a new method for estimating the mass of gravitationally bound gas in molecular clouds using multi-line CO isotopologue observations, specifically combining 12CO, 13CO, and C18O J=1-0 data. The study builds on previous work showing that the star formation rate (SFR) correlates tightly and linearly with the mass of gravitationally bound gas, which can be identified from the power-law tail of the column-density probability distribution function (N-PDF) derived from dust emission observations.

The sample includes 16 molecular clouds with robust detections in all three CO isotopologue transitions, spanning both massive inner Galaxy clouds and nearby star-forming regions. The authors developed an optical-depth-correction framework that combines multiple CO isotopologues to trace gas continuously across a wider range of column densities than a single CO line, allowing better recovery of the full N-PDF shape, including its power-law tail.

The method works as follows: In regions where C18O is not detected, 13CO is assumed to be optically thin, and its column density is modeled using radiative transfer theory. In regions where C18O is detected, the 13CO/C18O line ratio is used to estimate the optical depth of 13CO, allowing an optical-depth-corrected column density to be derived. The excitation temperature is estimated from the peak brightness temperature of 12CO, assuming it is optically thick. The 13CO/C18O abundance ratio is determined by stacking spectra in optically thin regions. Finally, the 13CO column density is converted to molecular hydrogen column density using a Galactic gradient model for 12C/13C and a metallicity-dependent CO-to-H2 abundance ratio.

The authors find that the N-PDFs derived from combined CO isotopologue data recover the characteristic log-normal plus power-law profiles seen in dust-based studies. For the nearby clouds (Orion A, Orion B, and Aquila), the CO-based N-PDFs agree well with their dust-based counterparts over both the log-normal component and the first power-law tail. The normalized transition column densities inferred from the two tracers are broadly consistent, indicating that the combined CO isotopologue method recovers the onset of the self-gravitating component traced by dust.

For the distant clouds, most CO-based N-PDFs still exhibit the characteristic combination of a log-normal component and a power-law tail, though their agreement with dust-based N-PDFs is generally poorer due to overlapping cloud components along the line of sight. The dust-based maps integrate emission from all structures along the line of sight, whereas the CO-based maps use velocity information to isolate the principal molecular cloud component.

The spatial consistency of the bound structures was quantified using intersection-over-union (IoU) measurements. For the nearby clouds, all three sources yield IoU values above 0.5, with Orion A reaching 0.63. Among the distant clouds, 69.2% have IoU values above 0.5, and this fraction increases to 75% when the nearby clouds are included. If a threshold of 0.4 is taken as the criterion for good agreement, the matching fractions increase to 100% for the full sample.

The mass comparison shows that the two Mbound estimates agree well, with a fitted relation having a slope close to unity (0.97 ± 0.11) and most sources lying within approximately a factor-of-two scatter around this relation. The total cloud masses derived from the two methods also show good agreement across the sample.

The authors also compared their CO-based method with conventional dense-gas tracers. They found that the total gas mass inferred from 13CO alone is systematically higher than the bound-gas mass estimates, with poor spatial overlap (mean IoU of 0.18). The total gas mass derived from C18O alone is typically lower than the bound gas mass, with larger cloud-to-cloud dispersion and heterogeneous spatial agreement (mean IoU of 0.36). The CO-combination method, which uses the N-PDF rather than a single-line integrated intensity, provides a more stable way to identify and quantify the bound gas component.

The CO-based bound gas masses reproduce a linear correlation with SFRs, with a best-fit slope of 1.08+0.11/−0.10, in excellent agreement with the dust-based results from S. Jiao et al. (2025). This further supports multi-line CO modeling as a robust tracer of gravitationally bound gas.

The paper also presents the derived 13CO/C18O abundance ratios as a function of Galactocentric radius, finding a linear gradient for distant clouds: 13CO/C18O = (0.75 ± 0.14) × (Rgc/kpc) + (3.76 ± 0.72). The solar neighborhood values vary strongly relative to the large-scale Galactic trend.

The authors tested the extension of the method to the 12CO+13CO pair using Ophiuchus and S287, finding that the derived N-PDFs show log-normal components at low column densities and power-law tails at high column densities, with IoU values of 0.5 compared to dust-based bound structures.

In conclusion, the paper demonstrates that the CO isotopologue combination can robustly trace the self-gravitating component via the N-PDF method and provides a reliable, scalable, and velocity-resolved alternative to dust emission for identifying the star-forming gas in molecular clouds. A key advantage of the CO-based method lies in its intrinsic velocity resolution, which is particularly important for large-sample analyses in the Galactic plane where line-of-sight confusion is widespread.

Improvements to AI Systems:

1. Velocity-Resolved Column Density Reconstruction: Implement a multi-line radiative transfer model (12CO, 13CO, C18O) that uses velocity information to deproject overlapping molecular clouds along the line of sight. The AI can separate blended emission from multiple clouds, reconstructing individual N-PDFs even in confused Galactic plane regions where dust-based methods fail.

2. Optical-Depth-Aware Tracer Selection: Train a classifier that dynamically selects the optimal CO isotopologue pair (e.g., 13CO+C18O vs. 12CO+13CO) based on local signal-to-noise and detection thresholds. This improves mass estimation accuracy by automatically applying the correct optical-depth correction framework (radiative transfer vs. line-ratio-based) without manual intervention.

3. Self-Gravitating Gas Identification via N-PDF Fitting: Develop an AI module that fits the log-normal + power-law functional form to CO-derived N-PDFs, automatically extracting the transition column density (onset of self-gravity) and the power-law slope. This enables real-time identification of bound gas mass (Mbound) from CO data alone, bypassing the need for dust emission maps.

4. Galactocentric Abundance Gradient Calibration: Integrate the derived 13CO/C18O abundance gradient (linear with Rgc) into AI pipelines for CO-to-H2 conversion. The AI can self-calibrate abundance ratios per cloud based on its Galactocentric radius, reducing systematic biases in mass estimates for distant clouds.

5. Spatial Consistency Scoring (IoU-based Validation): Add an automated intersection-over-union (IoU) metric to compare AI-predicted bound structures against dust-based references. This enables the AI to flag low-confidence detections (IoU < 0.4) and trigger re-analysis with alternative tracer combinations or velocity windows.

6. Scalable Star Formation Rate Prediction: Use the CO-based Mbound estimates to train a regression model that predicts SFR with a slope of 1.08, matching dust-based results. The AI can then forecast SFR for large samples of clouds in the Milky Way without requiring expensive dust observations, enabling statistical studies of star formation across the Galactic disk.

What the Improved AI System Can Do:

• Map the full column-density distribution of molecular clouds in 3D (position-position-velocity) using CO isotopologues, recovering both low-density envelopes and dense, self-gravitating cores.

• Automatically isolate a single cloud's emission from line-of-sight confusion, producing clean N-PDFs for thousands of clouds in the Galactic plane.

• Estimate bound gas mass and star formation potential with accuracy comparable to dust-based methods, but at higher spatial resolution and with velocity discrimination.

• Self-correct for abundance variations across the Galaxy, improving mass accuracy for clouds at different Galactocentric radii.

• Provide real-time quality metrics (IoU, slope agreement) to flag unreliable measurements and suggest alternative tracers or fitting strategies.

• Enable large-scale, dust-independent surveys of star-forming gas, facilitating studies of cloud evolution, star formation efficiency, and Galactic-scale feedback without reliance on Herschel or ALMA dust continuum data.

Recent studies have shown that the star formation rate (SFR) correlates tightly and linearly with the mass of gravitationally bound gas, which can be delineated from the power-law tail of the column-density probability distribution function (N-PDF) derived from dust emission observations. This relationship holds across four orders of magnitude within the Milky Way--spanning low-mass to high-mass star-forming regions and encompassing the extreme environment of the Central Molecular Zone. Building on this framework, we present a new approach for estimating the mass of gravitationally bound gas in molecular clouds using multi-line CO isotopologue observations. Our sample includes 16 molecular clouds with robust detections in 12 CO, 13 CO, and C 18 O J = 1-0, spanning both massive inner Galaxy clouds and nearby star-forming regions. We find that the N-PDFs derived from combined CO isotopologue data recover the characteristic log-normal plus power-law profiles seen in dust-based studies. The mass and spatial distribution of the self-gravitating structures estimated from both dust-based and CO-based methods agree well throughout the sample. This indicates that the CO isotopologue combination can robustly trace the self-gravitating component via the N-PDF method and provides a reliable, scalable, and velocity-resolved alternative to dust emission for identifying the star-forming gas in molecular clouds.