Galaxy luminosity functions from far-UV to submillimetre at z=0 in the COLIBRE simulations

arXiv:2605.02022 · astro-ph.GA · Submitted 2026-05-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.

Vera: Today's paper: "Galaxy luminosity functions from far-UV to submillimetre at z=0 in the COLIBRE simulations".

Jocelyn: This research presents predictions from recent COLIBRE cosmological hydrodynamical simulations, post-processed with a calibration-free radiative transfer code called skirt,

Vera: First, who's behind it and why it matters.

Title and authors: Vera: So, we're talking about this paper titled "Galaxy luminosity functions from far-UV to submillimetre at z=zero in the COLIBRE simulations." It seems like they are looking at how galaxies look across a huge range of wavelengths, from the ultraviolet right down to the submillimetre.

Jocelyn: That title suggests a very broad view of galaxy populations, Vera; it's not just focusing on one specific light band but covering everything from young stars to warm dust emission. Who are the authors on this project? I want to know who is behind these cosmological simulations.

Subrahmanyan: The authors include folks like Shengdong Lu, Carlos S. Frenk, Cedric G. Lacey, Andrea Gebek, Joop Schaye, Shaun Cole, Sownak Bose, Nick Andreadis, Maarten Baes, Alejandro Ben´ıtez-Llambay4 and Evgenii Chaikin1 and Robert A. Crain5 among others. These are definitely heavy hitters in the simulation community.

Vera: Exactly; having names like Frenk and Schaye on the list tells us we’re dealing with a top-tier cosmological simulation effort, which is exciting because they're using COLIBRE simulations to generate these luminosity functions.

Jocelyn: And what does that mean for us observing the sky? It implies they are connecting the complex physics happening inside galaxies in those simulations directly to what we actually see across different telescopes.

Subrahmanyan: From a theoretical side, having such a large team suggests they have integrated multiple complex physical processes into their model, which is crucial when trying to understand how structure grows in the universe.

Vera: It really is; the implication here is that we're getting a way to test our models of galaxy formation using these simulated LFs as a benchmark for real observations.

Jocelyn: So, it sounds like this paper isn't just running simulations; they’re trying to create a direct link between the theoretical model and observable data across the spectrum.

The paper's summary: Vera: The core of this paper, "Galaxy luminosity functions from far-UV to submillimetre at z=zero in the COLIBRE simulations," boils down to them showing predictions for galaxy luminosity functions across a wide range of wavelengths at redshift zero.

Jocelyn: That's pretty impressive because those LFs cover everything from the far-ultraviolet all the way into the submillimetre, which is a massive spectral range to cover. What exactly are these LFs they’re calculating?

Subrahmanyan: They are using the output from their COLIBRE simulations, which they describe as state-of-the-art cosmological hydrodynamical simulations that explicitly model the multiphase interstellar medium and dust formation self-consistently.

Vera: That self-consistent modeling of dust is a big deal because it means they aren't just guessing how much light is absorbed or re-emitted; they are letting the simulation tell them about the dust distribution itself.

Jocelyn: And what's the main result they highlight? The summary mentions that this combined approach successfully predicts stellar population properties and the distribution of interstellar dust with "unprecedented agreement" across all wavelengths.

Subrahmanyan: That level of agreement suggests that their framework, which couples COLIBRE with a calibration-free radiative transfer code called skirt, is a very robust way to model galaxy formation in this cosmological context.

Vera: It’s the fact that they managed to get stellar population properties and dust distribution matching observations across such a wide spectral range that makes this study quite significant.

Jocelyn: So, in simpler terms, they've taken a complex simulation suite and paired it with a post-processing tool to predict how galaxies look light-wise from the ultraviolet to the submillimetre at today's cosmic epoch.

The paper's improvements: Vera: Now let’s talk about what the authors suggest as improvements for this research, because they aren't just presenting a final result; they are pointing out where things could get better.

Jocelyn: What kind of improvements are they suggesting? Are we talking about tweaking the input physics in COLIBRE, or is it more about refining how we process the data afterward?

Subrahmanyan: The paper points to several sources of uncertainty they systematically test, including the effect of different AGN feedback models, which showed little influence on LFs across all bands at redshift zero.

Vera: That’s interesting; so they found that changing how much feedback happens from active galactic nuclei didn't really shake the final luminosity function results across the spectrum.

Jocelyn: Then they also looked at star-forming region resampling, and that only affected the far-UV LF by up to zero point four dex, which is a pretty small adjustment compared to other uncertainties.

Subrahmanyan: They also tested how aperture size impacts the results, finding it primarily affects the bright end in the far-UV, r, and K bands. This suggests that how you define your measurement volume matters depending on what part of the spectrum you're looking at.

Vera: So, they’re showing us exactly where these modeling assumptions have their biggest impact—it’s not everywhere equally across all wavelengths or physical parameters.

Jocelyn: It sounds like the suggestion is really to focus on tuning those specific subgrid physics, like the dust grain size distribution or star formation efficiency, which are deeply tied into how the simulation runs.

Conclusion: Vera: So, to wrap things up on this paper, it seems they’ve established a very strong connection between their complex simulation setup and what we observe in terms of galaxy luminosity functions from far-UV to submillimetre at z=zero in the COLIBRE simulations.

Jocelyn: They’ve shown that when you combine COLIBRE with the skirt pipeline, you get predictions for LFs that match observations remarkably closely across the FUV to K band and even into those longer IR bands like seventy µm to eight hundred fifty µm.

Subrahmanyan: The implication is that this combination of simulation and radiative transfer modeling captures the bulk of the relevant physics governing stellar emission, dust attenuation, and dust re-emission in a broadly realistic manner for these galaxies.

Vera: They did find one specific limitation they wanted us to be aware of: a noteworthy result is that colibre-skirt predicts galaxies that are not bright enough at the bright end from eight to twenty-four µm, which leads to a truncated bright end relative to the observations.

Jocelyn: And they also mentioned that while the total infrared LF agrees well with observations at the faint end, it shows an underprediction of the bright end for those same IR bands.

Subrahmanyan: Overall, this paper on "Galaxy luminosity functions from far-UV to submillimetre at z=zero in the COLIBRE simulations" provides a very solid test for models of galaxy formation in a cosmological context by demonstrating how well their framework can reproduce observed properties.

Vera: It’s a really compelling piece of work that validates the use of this simulation pipeline for studying galaxy evolution across such a wide spectral range, and it makes us think about what's next on the horizon for these simulations.

Institute for Computational Cosmology, Department of Physics, University of Durham; Department of Physics and Astronomy, Universiteit Gent; Leiden Observatory, Leiden University; Dipartimento di Fisica G. Occhialini, Universita degli Studi di Milano Bicocca; Astrophysics Research Institute, Liverpool John Moores University; Department of Astrophysics, University of Vienna; Centre for Data Science, Artificial Intelligence and Modelling, University of Hull; E. A. Milne Centre for Astrophysics, University of Hull; Lorentz Institute for Theoretical Physics, Leiden University; Institute of Cosmology and Gravitation, University of Portsmouth

astro-ph.GA

Submitted: 2026-05-03

Updated: 2026-09-30

Comments: 21 pages, 11 figures in the main text. Accepted for publication in MNRAS

Project page: https://icc.dur.ac.uk/data

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 83/100

The gist: This research presents predictions from recent COLIBRE cosmological hydrodynamical simulations, post-processed with a calibration-free radiative transfer code called skirt, to study galaxy luminosity

Key concepts

COLIBRE Suite
A new cosmological simulation suite that explicitly models the multiphase interstellar medium, including cold gas components. It self-consistently tracks the growth and destruction of dust grains throughout cosmic time using energy-density smoothed particle hydrodynamics (SPH).
Calibration-Free Radiative Transfer Pipeline
A post-processing framework where dust parameters are directly derived from the COLIBRE simulation results rather than being tuned. It models how starlight interacts with dust based on the pre-calculated, self-consistent distribution of dust predicted by the simulation.
Luminosity Functions (LFs)
Statistical tools used to describe the distribution of galaxy brightness across different wavelengths. This study compares simulated LFs against observations from far-ultraviolet (FUV) through submillimetre bands to test how galaxies emit light.
Self-Consistent Dust Modeling
The simulation models dust formation, growth, and destruction simultaneously using a grain evolution model that accounts for multiple grain sizes and species. This ensures the dust properties used in subsequent calculations accurately reflect the physical processes occurring within the simulated galaxy.

Terminology

Summary

This research presents predictions from recent COLIBRE cosmological hydrodynamical simulations, post-processed with a calibration-free radiative transfer code called skirt, to study galaxy luminosity functions (LFs) across wavelengths from the far-ultraviolet to the submillimetre at redshift zero. This work is significant because it demonstrates that COLIBRE, coupled with this framework, successfully predicts stellar population properties and the distribution of interstellar dust with unprecedented agreement across all wavelengths, providing a robust test for models of galaxy formation in a cosmological context.

Simulation Framework and Post-Processing

The study utilizes the COLIBRE suite (Schaye et al. 2026), which is described as a "new state-of-the-art cosmological simulation suite, the first generation of large volume simulations that explicitly models the multiphase ISM, including a cold ISM component, and simultaneously models the formation, growth, and destruction of dust in a self-consistent manner. The simulations employ the energy-density smoothed particle hydrodynamics (SPH) method Sphenix. A key aspect of this model is that radiative cooling of gas is followed down to temperatures of about 10 K (Ploeckinger et al. 2025), allowing for the explicit modeling of the multiphase interstellar medium. Furthermore, dust formation, growth, and destruction are modeled self-consistently during the simulation using a dust grain evolution model that follows multiple grain sizes and species."

Calibration-Free Radiative Transfer Pipeline

The paper employs a calibration-free skirt post-processing framework, which is built upon the skirt code (version 9). This pipeline is crucial because it models the interaction between starlight and dust based on dust distributions predicted self-consistently by the colibre simulations prior to the radiative transfer calculations. The decomposition of luminosity sources into three components—evolved stars, star-forming regions, and dust—is central to this approach. For instance, evolved stars are modeled using simple stellar population (SSP) templates from BPASS, while star-forming regions are computed using the toddlers library. The researchers note that the colibre-skirt pipeline is calibration-free, meaning all dust-related parameters used in the pipeline are directly derived or calculated from the colibre simulations prior to the radiative transfer calculations.

Sample Selection and Data Products

To compute LFs, a representative subsample of galaxies is selected from three COLIBRE simulations at different resolutions: L025m5, L200m6, and L400m7. The selection process involves binning galaxies by their stellar mass and instantaneous SFR (with a bin size of 0.5 dex for both), and then randomly selecting a target number of galaxies from each bin to maintain computational efficiency while preserving population statistics. The calculation of luminosities involves following the practice of Camps et al. (2016) by adopting filter-weighted mean luminosities across all bands from the FUV to the submillimetre.

Key Findings on Luminosity Functions

The main results show that the LFs, especially at the bright end, from the FUV to the K band are influenced by the choice of aperture size, while those in IR bands are not affected. The combined LFs from different simulations exhibit excellent agreement over large luminosity ranges from FUV to K band, and good agreement at the bright end for FIR and submillimetre bands. Specifically, the paper finds that the colibre-skirt LFs match remarkably closely with the observed LFs from the FUV to the K band and in the FIR to submillimetre bands (70 µm to 850 µm). However, a noteworthy result is that colibre-skirt predicts galaxies that are not bright enough at the bright end from 8 to 24 µm, leading to a truncated bright end relative to the observations. The total infrared (TIR) LF agrees well with observations at the faint end but shows an underprediction of the bright end.

Model Uncertainties and Future Directions

The study systematically tests several sources of uncertainty, including:

  1. The effect of different AGN feedback models, which showed little influence on the LFs across all bands at z = 0.

  2. The effect of star-forming region resampling, which only affected the FUV LF by up to 0.4 dex.

  3. The effect of aperture size, which primarily impacted the bright end in FUV, r, and K bands.

The overall good agreement suggests that "the combination of the colibre galaxy population and the present dust-radiative-transfer modelling captures the bulk of the relevant physics governing stellar emission, dust attenuation, and dust re-emission in a broadly realistic manner. The remaining discrepancies in MIR bands are attributed to modelling uncertainties," including potential contributions from nebular continuum and AGN emission.

Improvements for AI systems

As a fastidious and diligent researcher, I have analyzed this scientific paper to extract key insights into galaxy formation physics, observational constraints, and simulation methodology. Here are specific improvements for AI systems that could be derived from this research:

)Specific Improvements for AI Systems:

  1. AI-driven Cosmological Parameter Inference & Model Calibration:

  2. AI-driven Multi-Wavelength Luminosity Function (LF) Prediction & Error Quantification:

  3. AI-driven Subgrid Physics Parameter Tuning and Sensitivity Analysis:

  4. Physics-Informed Neural Networks (PINNs) for Radiative Transfer Modeling:

)What the Improved AI System Can Do:

  1. AI-driven Cosmological Parameter Inference & Model Calibration:

  2. An improved system can use observed galaxy luminosity functions (FUV to submillimetre) as a primary constraint to infer cosmological parameters, particularly the Hubble constant and potentially dark energy evolution, by comparing simulated LFs across different redshift bins (as suggested by Appendix D). It could also be used to systematically calibrate the free parameters of complex hydrodynamical simulations like COLIBRE (e.g., AGN feedback coupling efficiencies) against observed scaling relations (GSMF, SSMR) using Gaussian Process Emulators or Bayesian methods, moving beyond simple manual tuning.

  3. An improved system can perform high-fidelity forward modeling of simulated galaxy properties to predict multi-wavelength LFs across the entire spectrum (FUV to submillimetre). It can output not just a predicted LF, but also rigorous uncertainty quantification (error bars) derived from the sensitivity tests performed in Appendix A and C. This allows AI systems to quantify exactly how much uncertainty stems from modeling assumptions like PAH abundance, dust grain size distribution, or smoothing lengths.

  4. A Physics-Informed Neural Network (PINN) system can be developed to replace or augment traditional radiative transfer codes (like SKIRT) for calculating galaxy luminosities from simulation outputs. By incorporating the known physics of radiation transport and dust interaction into the loss function, the PINN can learn to model complex, wavelength-dependent spectral energy distributions more accurately than purely empirical models. This system could then be used to rapidly generate LFs for new simulation runs or explore parameter spaces much faster than running full radiative transfer codes.

Sources

Related papers