Scaling Relations of Galactic Outflows Across Cosmic Time: New Insights from Cosmic Noon

This paper investigates the scaling relations of galactic outflows across cosmic time, with a focus on the Cosmic Noon epoch (z ∼ 2–3). The authors construct a joint sample of 387 galaxies from z ∼ 0–9, including 98 new Cosmic Noon galaxies from the Keck Baryonic Structure Survey (KBSS) and the Keck Lyman Continuum Spectroscopic Survey (KLCS). Using high signal-to-noise emission lines (SNR > 50 for Hα or [O iii] λ5007) from Keck/MOSFIRE spectra, they detect warm-ionized outflows by decomposing lines into narrow and broad components.

Key findings:

1. Outflow occurrence rate: The overall detection rate of outflows is 30% (98 of 323 galaxies with SNR > 50 detections of Hα or [O iii]) based on the existence of the broad component in nebular emission lines.

2. Outflow velocity: On average, outflows at Cosmic Noon have higher maximum velocities (Vmax ≈ 200–800 km s−1) than those at low redshift (Vmax ≈ 100–500 km s−1) for a fixed stellar mass (M⋆), SFR, or SFR surface density (ΣSFR). The maximum outflow velocity can be up to a factor of 3 higher at Cosmic Noon for fixed galaxy properties.

3. Mass outflow rate: Cosmic Noon galaxies have higher mass outflow rates (Ṁout) for a fixed M⋆, with offsets of 1.5–3 dex compared to local galaxies. The offset is larger at lower masses, implying that young, lower-mass galaxies at Cosmic Noon host stronger outflows than galaxies with the same stellar mass today.

4. Mass loading factor: Despite faster and stronger outflows, there is no evolution in the mass loading factor (ηm = Ṁout/SFR) for a fixed M⋆. The mass loading factor lies between 0.3 and 10, and shows an anti-correlation with stellar mass and a positive correlation with ΣSFR at Cosmic Noon.

5. Galactic fountains: The majority of outflowing gas is recycled at a radius of ∼0.03 Rvir. While 71.4% of the fastest outflows (Vmax) can travel beyond Rvir, only 19.4% of the 75th percentile outflows (V75) do so, suggesting most outflowing gas is likely recycled back into the central galaxy.

6. Outflow geometry and detectability: Taking a simple disk-like approximation, the mean inclination of the galaxies is ≈ 67.8°, meaning measured velocities are likely not the down-the-barrel velocities. If outflows are bi-conical, the outflow detection rate of 30% corresponds to an average outflow opening angle of θ > 42° and a total outflow occurrence rate of at least 30%. However, previous KBSS studies have not found significant evidence of such geometry, and if galaxies at Cosmic Noon host irregular or spherical outflows, the outflow occurrence rate would be foutflow = fbroad = 30%.

7. Redshift evolution: For the joint sample, there is a redshift evolution in Ṁout as a function of M⋆, where lower-mass galaxies at high redshifts expel more mass for a fixed M⋆, but become consistent with local galaxies at higher M⋆. This is consistent with the suppression of feedback in massive galaxies by gravitational and hot CGM confinement, regardless of redshift.

The paper concludes that large samples of high–SNR individual galaxies are critical for making robust measurements of outflow properties. The authors present the largest outflow sample of star-forming galaxies at Cosmic Noon to date, finding that warm-ionized outflows at Cosmic Noon play a significant role in mass ejection from the galaxy, driven by a heightened mass-outflow rate. The majority of the outflowing gas in a single feedback cycle is likely to remain bound, but over multiple feedback cycles, a significant fraction of baryons can be lost from their host halos.

Improvements to AI Systems Based on This Paper:

1. Redshift-Aware Galaxy Evolution Models

• Improvement: Train AI models to incorporate redshift-dependent scaling relations (e.g., outflow velocity, mass outflow rate, and mass loading factor as functions of stellar mass, SFR, and ΣSFR) rather than assuming local-universe relations.

• Capability: AI can predict outflow properties for galaxies at any redshift (z 0–9), enabling more accurate simulations of galaxy formation and feedback in early cosmic epochs.

2. Outflow Detection via Spectral Decomposition

• Improvement: Implement AI algorithms that automatically decompose emission lines (e.g., Hα, [O iii]) into narrow and broad components, using SNR thresholds (>50) and line-profile fitting, to identify warm-ionized outflows.

• Capability: AI can analyze large spectroscopic surveys (e.g., from JWST, Keck) to detect outflows in thousands of galaxies without manual inspection, flagging candidate outflows with quantified confidence.

3. Inclination and Geometry Correction

• Improvement: Build AI models that estimate galaxy inclination (mean ≈ 67.8°) and correct measured outflow velocities for projection effects, using disk-like or bi-conical outflow assumptions.

• Capability: AI can provide unbiased intrinsic outflow velocities and opening angles, improving comparisons across surveys and reducing systematic errors in kinematic studies.

4. Outflow Recycling and Escape Fraction Predictor

• Improvement: Train AI to predict the fraction of outflowing gas that escapes the virial radius versus that recycled (e.g., 71.4% of Vmax escapes, but only 19.4% of V75), based on galaxy mass, SFR, and outflow velocity distribution.

• Capability: AI can estimate baryon loss and recycling rates over multiple feedback cycles, informing models of galaxy quenching and chemical enrichment.

5. Mass Loading Factor Regression

• Improvement: Develop AI regressors that predict mass loading factor (ηm) as a function of stellar mass and ΣSFR, capturing the observed anti-correlation with mass and positive correlation with ΣSFR, while accounting for redshift independence.

• Capability: AI can generate realistic feedback prescriptions for cosmological simulations, improving predictions of galaxy stellar mass functions and star formation histories.

6. Outflow Occurrence Rate Estimator

• Improvement: Create AI classifiers that estimate the probability of outflow presence from galaxy properties (mass, SFR, redshift), incorporating the 30% detection rate and geometry-dependent corrections (e.g., opening angle > 42°).

• Capability: AI can prioritize targets for follow-up spectroscopy and predict outflow prevalence in untargeted galaxy populations, optimizing observational strategies.

7. Multi-Epoch Baryon Budget Tracker

• Improvement: Integrate the paper’s finding that lower-mass high-z galaxies expel more mass per fixed M⋆ into AI-based semi-analytic models, tracking cumulative baryon loss across cosmic time.

• Capability: AI can simulate the baryon cycle from z=9 to z=0, predicting how feedback shapes the mass–metallicity relation and the missing baryon problem.

8. Uncertainty-Aware Outflow Property Inference

• Improvement: Use Bayesian neural networks or Gaussian processes to infer outflow velocities, mass outflow rates, and mass loading factors with robust uncertainties, given the scatter in the observed scaling relations.

• Capability: AI can provide probabilistic predictions for galaxy outflows, enabling more reliable comparisons with theoretical models and future observations.

Galactic outflows play a significant role in regulating galaxy evolution. Scaling relations between stellar mass (M), star formation (SFR), and outflow properties have been extensively studied at low redshifts (z<1) but less so beyond z about 2. We construct a joint sample of 387 galaxies from z about 0-9, including 98 new Cosmic Noon galaxies from the Keck Baryonic Structure Survey and the Keck Lyman Continuum Spectroscopic Survey. Using high signal-to-noise emission lines (SNR > 50 for H alpha or [OIII] lambda5007) from Keck/MOSFIRE spectra, we detect warm-ionized outflows by decomposing lines into narrow and broad components. With the joint sample, we explore the redshift evolution of outflow scaling relations. On average, outflows at Cosmic Noon have higher maximum velocities than those at low-z by up to a factor of 3 for a fixed M, SFR, or SFR surface density. They also have higher mass outflow rates for a fixed M. Despite faster and stronger outflows, there is no evolution in the mass loading factor for a fixed M. We find evidence for galactic fountains, as the majority of outflowing gas is recycled at a radius of about 0.03 R vir. We also constrain the overall outflow occurrence rate in our galaxy sample to be at least 30 % when taking galaxy orientation and outflow geometry into account. By analyzing the largest sample of warm-ionized outflows at Cosmic Noon to date and compiling large galaxy samples across all redshifts, we present a comprehensive analysis of galactic outflows throughout cosmic time.