The Extended Mapping Obscuration to Reionization with ALMA (Ex-MORA) Survey: A Molecular Gas Line Search

This study provides a census of molecular gas content in galaxies spanning a broad cosmic range (0.5 < z < 5) through a spectral search within the Extended Mapping of Obscuration to Reionization Survey (Ex-MORA) ALMA Band 4 observations. The spectral axis of ALMA data in the Ex-MORA field, covering 577 arcmin2, enabled a dual-strategy molecular line search: an unbiased search for high-significance line emitters and a targeted search utilizing spectroscopic redshift priors. In total, 52 galaxies were detected in CO or [C I] line transitions, with two exhibiting two distinct lines. Physical properties were derived using CIGALE with non-parametric star formation history modeling, and molecular gas masses were estimated using metallicity-dependent conversion factors. Our sample is dominated by MS systems, characterized by high stellar masses (median log M∗ /M⊙ = 10.83) and substantial molecular gas reservoirs (> 1010 M⊙). The gas fraction (µgas) increases with redshift among active populations; however, this trend may be influenced by environmental factors, particularly a significant bias at high redshift due to a gas-rich protocluster at z ≈ 2.4. In contrast, the overdensity at z ≈ 0.73 is indistinguishable from the field population and does not bias the results. Five transitional Green Valley sources are identified, exhibiting a peak in µgas at cosmic noon, potentially due to stellar feedback during this epoch that suppresses star formation by heating the gas rather than immediate reservoir depletion. Trying to conclude that this process happens to all Green Valley galaxies is beyond the scope and capabilities of our survey, but our results show that at least some GV galaxies could be affected by feedback. Kinematic analysis of resolved sources indicates a predominance of rotating disk structures. Large-area line searches are highly effective at recovering gas-massive galaxies across diverse evolutionary states, including transitional phases often missed by traditional surveys. The prevalence of MS systems and high gas masses underscores this approach’s ability to identify and characterize the most prominent gas reservoirs.

Improvements to AI Systems Based on This Paper:

1. Multiwavelength Spectral Line Search Optimization

• Improve AI algorithms for simultaneous unbiased and redshift-prior-guided line searches in large interferometric datasets (e.g., ALMA).

• The improved system can automatically detect multiple CO and [C I] transitions with high significance, reducing false positives and recovering faint emitters in crowded fields.

2. Redshift- and Metallicity-Dependent Gas Mass Estimation

• Enhance AI regression models to apply metallicity-dependent CO-to-H2 conversion factors dynamically, rather than using fixed values.

• The improved system can predict molecular gas masses with lower systematic bias across 0.5 < z < 5, especially for low-metallicity or high-redshift galaxies.

3. Environmental Bias Detection and Correction

• Develop AI clustering algorithms to identify overdensities (e.g., protoclusters at z ≈ 2.4) and flag their influence on derived scaling relations.

• The improved system can automatically separate field galaxies from cluster members, providing unbiased gas fraction–redshift trends and quantifying environmental contamination.

4. Transitional Phase (Green Valley) Classification

• Train AI classifiers on multi-band photometry and spectral indices to identify Green Valley galaxies with suppressed star formation but high gas content.

• The improved system can pinpoint galaxies where feedback heats gas without depletion, enabling targeted follow-up and testing of feedback mechanisms across cosmic time.

5. Kinematic Morphology Classification from Resolved Line Maps

• Upgrade AI vision models to classify resolved velocity fields (e.g., rotating disks vs. mergers) from CO/[C I] cubes.

• The improved system can automatically measure rotation curves, velocity dispersions, and dynamical masses, distinguishing disk-dominated from disturbed systems without manual inspection.

6. Non-Parametric Star Formation History (SFH) Inference

• Improve AI-based SED fitting (e.g., CIGALE) to handle non-parametric SFH priors, enabling more accurate stellar mass and star formation rate estimates for gas-rich, high-z galaxies.

• The improved system can recover bursty or delayed SFHs, reducing degeneracies and improving physical property estimates for transitional and massive systems.

7. Large-Scale Survey Completeness and Selection Function Modeling

• Enhance AI to model selection biases in line searches (e.g., luminosity limits, line width effects) and correct for incompleteness.

• The improved system can produce volume-averaged gas mass functions and unbiased redshift evolution of gas fractions, even with heterogeneous survey footprints.

8. Multi-Line Correlation and Excitation Analysis

• Implement AI to jointly analyze multiple detected lines (e.g., CO(4-3) and [C I]) to constrain gas excitation conditions and radiation field.

• The improved system can automatically infer gas temperature, density, and radiation field intensity, providing deeper insight into the physical state of the molecular gas.

What the Improved AI System Can Do:

• Automatically process large ALMA/NOEMA surveys to deliver a complete census of gas-rich galaxies across cosmic time, including rare transitional phases.

• Provide unbiased, environment-corrected measurements of gas fractions, stellar masses, and SFRs for thousands of galaxies.

• Identify and characterize rotating disks vs. mergers in unresolved or partially resolved data.

• Flag potential protoclusters and overdensities, and correct for their bias in cosmological evolution studies.

• Predict which galaxies are likely to be affected by feedback-driven gas heating, guiding future JWST or ALMA follow-up observations.