SE3D: Building a radiative transfer emulator to fit panchromatic resolved galaxy observations with 3D models of dust and stars 2025 SE3D: Building a radiative transfer emulator to fit panchromatic resolved galaxy observations with 3D models of dust and stars – C

Steven Ramnichal, Junkai Zhang, Stijn Wuyts, Cheng Li · 2025

Read SE3D: Building a radiative transfer emulator to fit panchromatic resolved galaxy observations with 3D models of dust and stars 2025 SE3D: Building a radiative transfer emulator to fit panchromatic resolved galaxy observations with 3D models of dust and stars – C by Steven Ramnichal, Junkai Zhang, Stijn Wuyts, Cheng Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-19623.txt.

Introduction

Steven Ramnichal, Junkai Zhang, Stijn Wuyts, and Cheng Li present SE3D, a framework for fitting resolved galaxy observations across all wavelengths at once. It models a galaxy's full energy output together with how its size, light profile, and shape change with wavelength. That joint view ties stars and dust into one consistent picture.

A machine learning emulator, a Bayesian neural network, stands in for costly three-dimensional dust radiation calculations. Trained on toy galaxies seen from many angles, it reproduces outputs to within about five hundredths of a dex. Sensitivity checks show the emulator has learned the real links between physical traits and what telescopes observe.

Catalog record

CREATORSteven Ramnichal, Junkai Zhang, Stijn Wuyts, Cheng Li
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE82 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-19623

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

SE3D: Building a radiative transfer emulator to fit panchromatic resolved galaxy observations with 3D models of dust and stars 2025 SE3D: Building a radiative transfer emulator to fit panchromatic resolved galaxy observations with 3D models of dust and stars – C Steven Ramnichal, Junkai Zhang, Stijn Wuyts, Cheng Li Abstract We present a framework for analysing panchromatic and spatially resolved galaxy observations, dubbed SE3D . SE3D simultaneously and self-consistently models a galaxy’s spectral energy distribution and its spectral distributions of global structural parameters: the wavelength-dependent galaxy size, light profile and projected axis ratio. To this end, it employs a machine learning emulator trained on a large library of toy model galaxies processed with 3D dust radiative transfer and mock-observed under a range of viewing angles. The toy models vary in their stellar and dust geometries, and include radial stellar population gradients. The computationally efficient machine learning emulator uses a Bayesian neural network architecture, and reproduces the spectral distributions at an accuracy of ∼ 0.05 \sim 0.05 dex or less across the dynamic range of input parameters, and across the rest-frame UVJ colour space spanned by observed galaxies. We carry out a sensitivity analysis demonstrating that the emulator has successfully learned the intricate mappings between galaxy physical properties and direct observables (fluxes, colours, sizes, size ratios between different wavebands, …). We further discuss the physical conditions giving rise to a range of total-to-selective attenuation ratios, R V R_{V} , with among them most prominently the projected dust surface mass density. 1 Introduction The past few decades have seen a flourishing of efforts to model the spectral energy distributions (SEDs) of galaxies with the aim of translating direct observables to intrinsic physical properties ( 20 , e.g.,) . In essence, the technique of modelling a galaxy’s integrated emission as a composite of stellar spectra traces back to the pioneering work by 72 . Commonly used libraries with spectra of the basic building blocks, the so-called single (or mono-age) stellar populations, have been composed by, e.g., 10 , 45 and 19 . These can be integrated according to a particular star formation history, and subject to dust attenuation. When paired with an energy balance approach (i.e., energy absorbed equals energy re-emitted) the predicted SEDs can further be extended into the far-infrared(FIR)/sub-mm regime ( 21 ; 8 ) . For the most part, applications of the above technique have focussed on galaxy-integrated SEDs, and necessarily imposed major simplifications compared to the true and complex nature of galaxy systems. These notably include, but are not limited to: (1) a simplified star formation history (SFH; see, e.g., 16 ; 37 for a discussion on parametrization), (2) the assumption of a universal attenuation law ( 11 , e.g.,) , and (3) a uniform foreground screen as dust configuration. Here, the attenuation law encodes the wavelength dependence of the combined effects of scattering and absorption, and incorporates the net effect of the typical star-dust geometry for the sample of galaxies on which it was calibrated. Its application via an assumed foreground screen of dust implies that dust reddening and attenuation are scaling linearly in the modelling. Where energy balance arguments are applied, these inherently assume isotropic emission across all wavelengths, or at least their basic ansatz is understood to only hold in an angle-averaged sense. That the above assumptions break down when considering galaxies in more detail should come as no surprise. Indeed, evidence for non-universality of the attenuation law shape has been presented by, e.g., 64 and 61 , although its inference is not free of modelling degeneracies ( 59 ) . Among sources of variation in attenuation law shape are intrinsic galaxy properties, such as their specific star formation rate ( 34 ; 60 ; 61 ) , but also observer-specific conditions such as the viewing angle. More inclined galaxies have been shown to feature greyer attenuation laws (i.e., with shallower wavelength dependence) compared to face-on counterparts, both on the basis of statistical samples of observed galaxies ( 80 ; 63 ; 6 ) and as demonstrated via radiative transfer calculations ( 74 ; 86 ) . As for the energy balance approach, isotropic emission is a reasonable expectation for all but the most extreme systems in the FIR regime ( 41 ) , but does not generally hold at shorter wavelengths ( 86 ) . Integrating over 4\pi steradian, energy conservation dictates that the amount of dust-obscured starlight must match the emerging dust reprocessed radiation. However, this argument does not need to hold for any individual viewing angle. Spatially resolved studies of galaxy structure across cosmic time have also progressed dramatically over the past two decades, thanks to space telescopes ( HST , JWST ) and ground-based interferometers (ALMA, NOEMA). Often, such studies address galaxy structural properties monochromatically and in 2D. Translation to 3D intrinsic shapes is feasible in a statistical sense for ensembles of galaxies that can be treated as similar aside from their orientation ( 76 ; 84 ; 86 ; 57 ) . With the advance from HST /ACS to HST /WFC3 and ultimately JWST /NIRCam, the rest-wavelengths probed for galaxies at cosmic noon ( 1\lesssim z\lesssim 3 ) progressed from rest-UV to rest-optical and now rest-NIR emission. Increasingly, such multi-wavelength structural measurements are complemented by longer wavelength resolved probes, in the mid-infrared (MIR) regime probing Polycyclic Aromatic Hydrocarbons (PAHs), a tracer of dust-obscured star formation ( 44 , e.g.,) , or on the Rayleigh-Jeans tail of the dust continuum ( 69 ; 71 , e.g.,) . One approach to leverage this multi-wavelength resolved information is to perform a pixel-by-pixel (or bin-by-bin) SED modelling ( 87 ; 81 ; 67 , e.g.,) . However, the ability to do so panchromatically (i.e., across UV-to-submm wavelengths) remains largely restricted to well-resolved, nearby galaxies ( 1 , see) . Even for nearby galaxies, such modelling is normally carried out in 2D, with few exceptions. 1 1 1 For examples of modelling in 3D, see the high-resolution dust radiative transfer applications by 22 ; 78 ; 55 ; 56 ; 79 ; 58 . The reason for the paucity of resolved panchromatic SED modelling studies at high-z owes to the heterogeneous resolution of the aforementioned instruments, the often marginally resolved nature of MIR and submm observations, and the frequently modest wavelength sampling of resolved observations for high-z galaxies. Consequently, rather than having robust UV-to-submm SEDs for every galaxy pixel, the more common situation high-z observers face is one in which a well-sampled galaxy-integrated SED is available, and sizes (as well as Sérsic indices under good signal-to-noise ratio conditions) are measured across a subset of the wavebands. Projected axis ratios can be extracted from high-resolution imaging too. Given an assumed intrinsic 3D shape, or one inferred via the aforementioned ensemble modelling approach, they can provide constraints on the observer’s viewing angle with respect to the galaxy under consideration. In this paper, we lay out a framework to exploit the full wealth of the accumulating multi-wavelength (semi-)resolved observations of distant galaxies. What is it that one may wish to learn from a joint analysis of the galaxy-integrated SED ( F_{\lambda}(\lambda) ), wavelength-dependent size ( R_{e}(\lambda) ), Sérsic index ( n(\lambda) ) and axis ratio ( q(\lambda) )? First,

Read the complete text