Deciphering The Launching of Multi-phase AGN-driven Outflows and Their (Spatially Resolved) Multi-scale Impact

This paper, a Science Case Development Document (SCDD) from the Habitable Worlds Observatory (HWO) AGN Working Group, proposes future HWO observations to solve key questions about the role of active galactic nucleus (AGN) feedback in galaxy evolution. The overarching science goal is to understand the role of feeding and feedback processes from growing black holes (i.e., active galactic nucleus; AGN) in shaping the cosmic ecosystem, which involves understanding the dynamics of gas flows in the interstellar (ISM), circumgalactic (CGM), intracluster (ICM), and intergalactic media (IGM).

The specific science objectives are to answer: "Which mechanism is dominant in triggering inflows/outflows through feedback? How is AGN activity triggered, and is it associated with circumnuclear star formation and what is the overall effect of AGN feedback on star formation (SF)? In AGN feedback, which mode is more influential and does AGN feedback operate similarly or differently in the local universe and at high redshift? The paper elaborates on these as: Which is more dominant in triggering inflows and outflows through feedback: AGN activity or star formation, and how does it work? How is AGN activity triggered, and is it associated with circumnuclear star formation? In AGN feedback, which mode is more influential: the momentum-driven (kinetic) mode or the energy-driven (radiative) mode? What is the overall effect of AGN feedback on star formation – does it predominantly inhibit or promote it (negative or positive feedback)? Does AGN feedback operate similarly in the local universe and at high redshift, or does it vary across cosmic time?"

To achieve these objectives, the paper states the need to "Resolve and characterize the spatial distribution of ionized and cold/warm molecular gas, especially those in inflows/outflows; Explore the spatial coupling and potential stratification of multi-phase inflows/outflows on different physical scales and their resolved and global correlations with AGN and/or SF activities; Investigate whether corresponding outflows/jets induce shocks and/or fluctuations that trigger or suppress the formation of molecular clouds and, hence, new stars. The required physical parameters to be measured are: Spatially resolved ionized and molecular gas content (M, from the flux of gas emission)", Spatially resolved ionized and molecular gas kinematics (v, σ from emission line profile), Spatially resolved ionized and molecular gas outflow rate/energy distribution (Ṁ, Ė), and Observation and modeling of AGN strength and SF distribution (LAGN, λAGN, SFR).

Regarding the observational bands, the paper argues that the capability of UV band spectroscopy is indispensable because "UV emission is pivotal for constraining the nature and intensity of AGN activity and for breaking the degeneracy between AGN emission and emission associated with the host galaxy given the UV peaked emission of AGN accretion disk. Furthermore, emission and absorption lines pertaining to UV spectra... can provide important diagnostics of the physical properties of AGN-driven outflows and estimates of how much mass and energy they actually carry. The paper also states that the relatively dust-extinction-immune NIR spectroscopy is critical for robustly constraining models of the torus and then ascertaining the launching site of the AGN-driven gas outflows, and that CO and H2 features in the NIR spectrum provide constraints on the mass and kinematics of cold/warm molecular gas outflows."

For the required sample size, the paper notes that both the observed momentum and energy loading factors of molecular outflows in AGN have a broad distribution, spanning over two to three orders of magnitude, and that an outflow may persist for more than an order of magnitude longer than the AGN episode that drives it, leading to a substantial population of ‘fossil’ outflows. Therefore, "a factor of ∼ 10 − 20 for the sampling of different loading factors and a factor of ∼ 50 − 100 for the sampling of different evolution stages of outflows under different AGN episodes are desired for exhaustive statistics. Namely, it is ideal to have a sample of ∼ (10 − 20) × (50 − 100), i.e., ∼ 500 − 2000, galaxies for HWO IFU observations."

The paper concludes that HWO’s capabilities will enable the spatially resolved analysis of multi-phase outflows and their connection to AGN activity across diverse galaxy types, while Existing facilities lack the required combination of high-throughput UV and NIR IFU capabilities with sufficient spatial resolution and sensitivity. Specifically, the paper states that there is currently no UV IFU available for the spatially-resolved spectroscopy survey, and that existing NIR IFUs overall have limited FOVs (∼ 3′′ × 3′′) and hence are not effective for mapping nearby galaxies. The required HWO capabilities are "UV and NIR IFU spectroscopies with ∼ 6′′ × 6′′ or better ∼ 10′′ × 10′′ FOV, ∼ 0.′′ 01 or better PSF size (with the resolved physical scale < 10 pc or better out to z = 0.1, i.e., 400 Mpc, and < 1 kpc at all redshifts), R ≈ 3000 – 30000 spectral resolution (δv ≈ 10 − 100 km s−1), and ∼ 10−19 erg s−1 cm−2 or higher sensitivity."

Based on the scientific paper, here are the specific improvements I can make to AI systems and what the improved systems can do:

Improvement: Train a deep learning model on spatially resolved IFU data to automatically classify gas components as inflows, outflows, or quiescent gas based on velocity dispersion and spatial coherence patterns.

Capability: The AI can process HWO IFU cubes (UV + NIR) to produce real-time kinematic maps that distinguish:

• Outflowing gas (velocity offsets > 10 km/s with coherent spatial structure)

• Inflowing gas (opposite velocity signatures)

• Turbulent/ambient gas (high dispersion, no coherent flow)

This directly addresses the paper's requirement to resolve and characterize the spatial distribution of ionized and cold/warm molecular gas, especially those in inflows/outflows.

Improvement: Develop a convolutional neural network that takes spatially resolved maps of gas kinematics, AGN luminosity distribution, and star formation rate maps to determine whether outflows originate from the AGN (central engine, <10 pc) or from star-forming regions (10-100 pc scales).

This directly answers the paper's question: Which is more dominant in triggering inflows and outflows through feedback: AGN activity or star formation?

Improvement: Train a classifier on synthetic outflow models (from simulations like those referenced in Zubovas & Nardini 2020) to distinguish between momentum-driven (kinetic) and energy-driven (radiative) outflows based on observable signatures: outflow velocity profiles, mass loading factors, and spatial stratification patterns.

Improvement: Implement a graph neural network that models the spatial relationships between gas phases (ionized, warm molecular, cold molecular) across scales from 1 pc to 10 kpc, learning how outflows couple between phases.

Improvement: Build a transformer-based time-series model trained on the scaling relations in Figure 6 (Fiore et al. 2017) and the loading factor distributions in Figure 7, incorporating AGN luminosity variability timescales (10 4–10 5 yr) and outflow lifetimes (10 6 yr).

Improvement: Train a domain-adaptation model that takes local universe HWO observations (well-resolved, multi-phase) and learns to predict what the same physical processes would look like at high redshift (z > 2) with degraded spatial resolution and different rest-frame wavelengths.

Improvement: Create a reinforcement learning agent that uses the paper's requirement of 500–2000 galaxies (sampling loading factors ×10–20 and evolutionary stages ×50–100) to optimize target selection from survey catalogs (e.g., UVEX).

Improvement: Develop a physics-informed neural network that accounts for the telescope PSF (diffraction-limited, stable) to recover true emission line profiles and velocity dispersions, correcting for any residual smearing effects.

Improvement: Train a variational autoencoder on AGN SED templates (Figure 4) combined with stellar population models to decompose observed UV-NIR spectra into AGN, stellar, and outflow components simultaneously.

Improvement: Implement a 3D convolutional neural network trained on hydrodynamical simulations of jets interacting with ISM gas to predict where shocks will occur, whether PAHs will be destroyed, and whether molecular cloud formation will be triggered or suppressed.

These improvements directly address the paper's science objectives and observational requirements, enabling the AI to extract maximum scientific return from HWO's UV and NIR IFU capabilities.

Beyond deepening our understanding of the formation, growth, and evolution of supermassive black holes, it is crucial to uncover the role of feeding and feedback processes from growing black holes (i.e., active galactic nucleus; AGN) in shaping the cosmic ecosystem. Such studies include understanding the dynamics of gas flows in the interstellar (ISM), circumgalactic (CGM), intracluster (ICM), and intergalactic media (IGM). As the output of a sub-group in Habitable Worlds Observatory (HWO) AGN Working Group, this Science Case Development Document (SCDD) proposes to use future HWO observations to solve the following questions. Which mechanism is dominant in triggering inflows/outflows through feedback? How is AGN activity triggered, and is it associated with circumnuclear star formation and what is the overall effect of AGN feedback on star formation (SF)? In AGN feedback, which mode is more influential and does AGN feedback operate similarly or differently in the local universe and at high redshift? To answer these questions, this SCDD proposes to use potential HWO observations as follows. Resolve and characterize the spatial distribution of ionized and cold/warm molecular gas, especially those in inflows/outflows; Explore the spatial coupling and potential stratification of multi-phase inflows/outflows on different physical scales and their resolved and global correlations with AGN and/or SF activities; Investigate whether corresponding outflows/jets induce shocks and/or fluctuations that trigger or suppress the formation of molecular clouds and hence new stars. Specifically, HWO's capabilities will enable us to achieve the above scientific goals while existing facilities lack the required combination of high-throughput ultraviolet (UV) and near-infrared (NIR) integral field unit (IFU) capabilities with simultaneously sufficient spatial resolution and sensitivity.