The Optical and Infrared Are Connected
Christian Kragh Jespersen, Peter Melchior, David N. Spergel, Andy D. Goulding, ChangHoon Hahn, Kartheik G. Iyer · 2025
Read The Optical and Infrared Are Connected by Christian Kragh Jespersen, Peter Melchior, David N. Spergel, Andy D. Goulding, ChangHoon Hahn, Kartheik G. Iyer (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-2503-03816.txt.
Introduction
Christian Jespersen and co-authors wrote a paper in 2025 showing that the optical and infrared light of galaxies are connected. Galaxies are often modeled as mixes of separate parts, each with its own spectral signature. That picture implies that two wavelength ranges are only weakly linked with each other.
The researchers present a data-driven model that uses subtle links between physical processes. From a compact neural summary of a galaxy's optical spectrum, the model predicts its infrared photometry well across all measured bands, with good colors as well. It can tightly constrain properties often derived from that light, such as the total luminosity of an active galactic nucleus and parameters of interstellar dust. Current standard fitting methods, tested against the same links, produced biased and overconfident predictions. The team then names which features of the optical spectrum drive the improved results, and lists several spectral lines.
Catalog record
| CREATOR | Christian Kragh Jespersen, Peter Melchior, David N. Spergel, Andy D. Goulding, ChangHoon Hahn, Kartheik G. Iyer |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 95 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-03816
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The Optical and Infrared Are Connected Christian Kragh Jespersen, Peter Melchior, David N. Spergel, Andy D. Goulding, ChangHoon Hahn, Kartheik G. Iyer Abstract Galaxies are often modelled as composites of separable components with distinct spectral signatures, implying that different wavelength ranges are only weakly correlated. They are not. We present a data-driven model which exploits subtle correlations between physical processes to accurately predict infrared (IR) WISE photometry from a neural summary of optical SDSS spectra. The model achieves accuracies of χ N 2 ≈ 1 \chi^{2}_{N}\approx 1 for all photometric bands in WISE, as well as good colors. We are able to tightly constrain typically IR-derived properties, e.g., the bolometric luminosities of AGN and dust parameters such as q PAH \mathrm{q_{PAH}} . We also test whether current SED-fitting methods reproduce such panchromatic relations, but find their predictions biased and overconfident, likely due to model misspecification, with correlated biases in star-formation rates and AGN luminosities being most evident. To help improve SED models, we determine which features of the optical spectrum are responsible for our improved predictions, and identify several lines (CaII, SrII, FeI, [OII] and H α \alpha ), which point to the complex chronology of star formation and chemical enrichment being incorrectly modelled. I Introduction Understanding the properties of galaxies and their evolution is one of the major focus point for the upcoming decades of astronomy [ 90 ] , as galaxies are both fascinating in their own right, and a fundamental tool for cosmology [ 51 , 12 ] . In the current paradigm of galaxy formation theory, galaxies develop through hierarchical growth, eventually forming many distinct components, such as the disk, bulge and central super-massive black hole (SMBH) [ 108 ] . These components are often modelled as evolving independently from each other, e.g. the stars in the outer disk and the circumgalactic medium are not directly affecting the SMBH. This assumption is often made implicitly, as modelling the interaction of different galaxy components is beyond our current understanding. Large-scale galaxy simulations are getting ever closer to being able to explore the interplay of these components, but for now they still rely on insufficient, resolution-limited subgrid models [ 127 ] . The implicit assumption of separability of many of the components of a given galaxy also implies that different regions in wavelength space are largely separable, since the different components emit across different wavelengths. The assumption that different components are approximately separable is propagated through to galaxy spectral energy distribution (SED) fitting codes, the most common method of deriving the physical properties of galaxies [ 94 , 30 , 31 , 62 , 95 , 57 ] . SED models comprise templates for separate components, which usually do not interact, except through a few constraints, e.g. to balance the amount of energy absorbed by dust at short wavelengths with its re-emission at longer wavelengths [ 35 ] . This constraint is based on the success of simple relations such as the star-forming \mathrm{H}\alpha –FIR connection [ 56 , 67 ] . Early works, such as [ 102 , 101 ] , also discussed the possible connections between optical properties and IR emission. As the astronomical community is now in possession of many hundreds of thousands of objects with coverage from the ultraviolet (UV) to the mid-infrared (MIR) [ 104 , 123 , e.g., GSWLC, COSMOS2020,] , as well as tools for large-scale data analysis, it is possible to find more general relations if these relations exist . In this paper we explore correlations across broad wavelength ranges. We do this by connecting the optical spectroscopy from the Sloan Digital Sky Survey [ 133 , 50 , SDSS;] to near- and mid-infrared (NIR/MIR) photometry from the Wide-field Infrared Survey Explorer [ 126 , WISE;] satellite. SDSS conducted spectroscopy of half a million galaxies between 3800 and 9200 Å at R \sim 2000 resolution, while WISE conducted all-sky four-band photometry from 3 to 27 \mathrm{\mu m} . SDSS was designed to be sensitive to the main stellar population in a typical galaxy, whereas WISE covers the transition from emission being dominated by older stellar populations to dust tori from AGN and asymptotic giant branch (AGB) stars, polycyclic aromatic hydrocarbons (PAHs) and hot dust. The combination of these two datasets is therefore ideal to investigate whether galaxy components are truly separable. We compare our results to results from two state-of-the-art SED fitting codes, CIGALE [ 7 , 132 ] and prospector [ 75 , 62 ] . Testing a physical model’s ability to predict quantities not directly included in the fit is a well-established validation approach. For example, [ 75 ] validated the prospector - \alpha model by comparing photometry-only predictions against spectroscopic quantities excluded from their fits, including H \alpha luminosities and Balmer decrements. Our approach is analogous: we ask whether models constrained by optical data produce physically reasonable predictions in the infrared, a natural self-consistency test for any model providing a panchromatic representation of galaxy emission. The paper is structured as follows. In § II we describe the SDSS and WISE data used in this study, as well as the preprocessing of the data. In § III we introduce the methods used to create the empirical mapping from optical spectra to IR photometry, as well as our setup for using CIGALE and prospector . § IV shows results from our empirical mapping and the SED fitting codes. We furthermore demonstrate that physical quantities typically derived from the IR can be robustly predicted using the optical spectrum alone, directly showing that the current picture of separable components is incorrect. In § V , we investigate what information is used for these accurate predictions and where SED models must improve. We furthermore demonstrate that the separability assumption in SED models leads to biases and imprecision in parameter inference. In § VI we conclude on our work and outline future directions. II Data The two data sources, SDSS and WISE, cover wavelengths spanning a factor of \approx 60 ( 0.4-24\mu m). The total dataset consists of 5.1\cdot 10^{5} crossmatched objects at z\leq 0.5 . II.1 SDSS spectra Figure 1: The redshift distribution of the objects used in this work. We utilize 5.1\cdot 10^{5} spectra from the Main Galaxy Sample of the Sloan Digital Sky Survey [ 114 , SDSS-II] . The spectroscopic part of the main galaxy sample from the SDSS survey was a magnitude-limited survey with high completeness. The targets were relatively bright galaxies with m_{AB,i}\leq 17.8 [ 114 ] . Figure 1 shows the distribution of redshifts. Table 1: Classifications of the spectra utilised in this work. “None” refers to any spectrum not falling into any other category, which mainly covers quiescent galaxies. None Star-forming Starburst AGN Narrow line 318,795 148,750 23,754 10,731 Broad line 7,922 673 37 1,350 Due to the magnitude-limited nature of the survey, the Main Galaxy Sample includes a high diversity of galaxies. The spectroscopic classifications (as determined by SDSS) are outlined in Table 1 . The high brightness limit allowed SDSS to generally obtain very accurate fluxes, since the main continuum foreground, the airglow emission of the night sky, varies between 20th and 22nd AB magnitude in the i-band [ 103 , 60 ]