PaperlayerView exact revision on arXiv

Source-linked AI summary

CIGALE: a python Code Investigating GALaxy Emission

M. Boquien, D. Burgarella, Y. Roehlly, V. Buat, L. Ciesla, D. Corre, A. K. Inoue, H. Salas

arXiv:1811.03094v1astro-ph.GAastro-ph.IM

TL;DR

Interpreting galaxy evolution requires extracting physical information from broad-band SEDs despite degeneracies and incomplete wavelength coverage. CIGALE addresses this through flexible multi-wavelength spectral modelling, energy balance, likelihood-based inference, and an extensible Python architecture. The resulting tool supports fast analysis of large model grids and galaxy samples, as well as model generation and library-based applications.

  • Problem

    Galaxy SED modelling is difficult because different physical properties can produce similar SEDs, while full wavelength coverage is seldom available.

  • Method

    CIGALE constructs flexible multi-wavelength SED models and estimates galaxy properties by fitting large model grids and analysing likelihood distributions.

  • Results

    CIGALE is a versatile, easy-to-use, high-speed tool that builds millions of models, analyses thousands of galaxies, and estimates physical properties.

  • Takeaways & Limitations

    CIGALE can support SED fitting, physical-property estimation, theoretical model generation, and development of new applications as a library.

  • Takeaways & Limitations

    The described software status is version 2018.1, and further developments are reserved for separate publications.

Abstract

from arXiv · show

Context. Measuring how the physical properties of galaxies change across cosmic times is essential to understand galaxy formation and evolution. With the advent of numerous ground-based and space-borne instruments launched over the past few decades we now have exquisite multi-wavelength observations of galaxies from the FUV to the radio domain. To tap into this mine of data and obtain new insight into the formation and evolution of galaxies, it is essential that we are able to extract information from their SED. Aims. We present a completely new implementation of CIGALE. Written in python, its main aims are to easily and efficiently model the FUV to radio spectrum of galaxies and estimate their physical properties such as star formation rate, attenuation, dust luminosity, stellar mass, and many other physical quantities. Methods. To compute the spectral models, CIGALE builds composite stellar populations from simple stellar populations combined with highly flexible star formation histories, calculates the emission from gas ionised by massive stars, and attenuates both the stars and the ionised gas with a highly flexible attenuation curve. Based on an energy balance principle, the absorbed energy is then re-emitted by the dust in the mid- and far-infrared domains while thermal and non-thermal components are also included, extending the spectrum far into the radio range. A large grid of models is then fitted to the data and the physical properties are estimated through the analysis of the likelihood distribution. Results. CIGALE is a versatile and easy-to-use tool that makes full use of the architecture of multi-core computers, building grids of millions of models and analysing samples of thousands of galaxies, both at high speed. Beyond fitting the SEDs of galaxies and parameter estimations, it can also be used as a model-generation tool or serve as a library to build new applications.

Context.

The paper is situated within data analysis, numerical methods, statistical methods, and general galaxy studies.

  • The work concerns data analysis methods.
  • It addresses numerical methods for galaxy research.
  • It also involves statistical methods and general galaxy studies.

1. Introduction

Galaxy SEDs encode information about interacting baryonic components, but extracting physical properties is difficult because different galaxies can have similar, incomplete-wavelength SEDs. The paper presents a new Python implementation of CIGALE that uses flexible, energy-balance-based modelling and Bayesian-like inference for efficient multi-wavelength analysis.

  • Galaxy SEDs capture the interplay of stars, gas, dust, and supermassive black holes across cosmic time.
  • Estimating physical properties is challenging because different galaxies can have similar SEDs, especially with restricted wavelength coverage.
  • Radiative-transfer models provide detailed modelling but can make large model grids prohibitively expensive computationally.
  • Energy-balance SED codes link dust infrared emission to stellar energy absorbed in the UV–optical range.
  • CIGALE weights models by goodness of fit to estimate physical properties while accounting for observational uncertainties and parameter degeneracies.
  • The new Python CIGALE implementation is designed to provide efficient, reliable, and versatile galaxy modelling for expanding multi-wavelength surveys.

2. Architecture

CIGALE is designed as a clear, modular, and efficient Python system for generating galaxy models, fitting observations, and estimating physical properties. Its architecture separates configuration, model computation, analysis, and output while supporting multi-core computation and extension as a library.

  • CIGALE prioritizes modularity, clarity, and efficiency for developers and users.
  • Its four main stages separate input handling, model computation, analysis, and output handling.
  • Modules can be replaced independently, such as changing the attenuation law without affecting the rest of the code.
  • Python provides accessible syntax and extensive scientific libraries for astronomy, numerical computation, storage, plotting, and data handling.
  • CIGALE is intended to generate theoretical models, fit observations, estimate galaxy properties, and support new applications as a library.
  • The pcigale executable performs model computation and property estimation, while companion tools support plotting and filter-database management.

3. Model creation modules

CIGALE builds galaxy SEDs through successive modules representing star formation, stellar populations, nebular emission, attenuation, dust, and AGN components. Its modular framework supports flexible and computationally tractable modelling across diverse galaxy histories and physical conditions.

  • Model construction: CIGALE progressively applies independent modules to compute star formation histories, stellar spectra, nebular emission, attenuation, dust emission, and other physical components.Individual spectral components and the combined spectrum are stored separately for subsequent computations.
  • Star formation history: Analytic and arbitrary star formation histories accommodate complex variations, but each run can use only one analytic SFH type.Available analytic cases include single or double exponentials, delayed histories with bursts, and periodic histories.
  • Star formation history: 1 Myr SFH sampling is refined into ten instantaneous episodes separated by 0.1 Myr to capture brief stellar evolutionary phases.The SFR is evaluated at the beginning of each age bin, while stellar contributions are computed at its end.
  • Attenuation: Flexible attenuation modules model diverse stellar and nebular attenuation behavior, including distinct treatment of young stars in birth clouds and older stars in the ISM.CIGALE combines flexible laws inspired by starburst attenuation with the Charlot & Fall model.
  • Dust and AGN emission: Three dust-emission model families provide varied infrared SED representations, while greater physical flexibility in the Draine & Li models increases parameter-space and computational costs.CIGALE includes Dale et al. (2014), Draine & Li (2007) with updates, and Casey (2012) models.
  • Dust and AGN emission: AGN emission can be represented with coarse rapid methods or more detailed slower models, including parameters for torus structure, viewing angle, and AGN fraction.The fritz2006 module provides detailed AGN accretion-disk, scattered, and thermal components.

4. Analysis modules

CIGALE’s analysis modules generate model grids, compute observables across spectral resolutions, and estimate physical properties by likelihood analysis. A fixed model grid supports efficient fitting and mock-catalogue reliability tests.

  • Analysis modules: CIGALE’s analysis modules generate model grids and outputs, then estimate physical properties from likelihood distributions rather than only the best-fitting model.The savefluxes module generates model grids, while pdf_analysis uses likelihood-weighted parameters and probability distributions.
  • Bandpasses: Band fluxes are obtained by integrating model spectra through transmission filters, with interpolation preserving spectral features and normalized transmissions avoiding redundant denominator calculations.Fluxes are converted to frequency units using a source-independent pivot wavelength and rescaled to flux densities in mJy.
  • Emission lines: Emission-line measurements use continuum-subtracting filters for low-resolution data, while high-resolution data use theoretical nebular emission templates corrected for dust extinction.The low-resolution filters include a positive line component and a negative continuum component, and are stretched by 1 + z to preserve rest-frame resolution.
  • Model generation: The savefluxes module computes spectra and physical properties for arbitrary theoretical galaxies by enumerating parameter combinations on an n-dimensional model grid.Parameters may instead be explicitly supplied in a file, with one line per SED and its associated model configuration.
  • Fixed-grid fitting: CIGALE’s fixed-grid strategy computes models once for all objects, enabling optimization and efficient fitting while requiring memory that grows with model, band, and property counts.The grid must be sufficiently sampled and remain in memory for likelihood-weighted estimation.
  • Reliability assessment: Mock-catalogue analysis tests estimation reliability by comparing known physical properties with likelihood-distribution estimates after perturbing best-fit quantities with Gaussian noise.This provides a direct assessment of whether the physical properties can be reliably recovered.

5. Examples of CIGALE use cases

CIGALE supports theoretical SED generation, physical-property estimation, and library-based applications. Examples show how model grids expose assumptions, enable interactive exploration, and achieve reliable estimates while retaining important scope limitations.

  • 5.1. Example of CIGALE as an SED generation tool: 22500 models combine 150 values for galaxy age with 150 values for the timescale of a delayed star-formation history.The example adopts Bruzual & Charlot (2003) models, a Salpeter IMF, and metallicity Z = 0.02.
  • 5.1. Example of CIGALE as an SED generation tool: Mid-to-near-IR flux ratios vary by less than 20% for nearby wavelengths but by up to a factor of two at WISE 22 µm or MIPS 24 µm.The longer-wavelength ratios depend strongly on star-formation history, especially for older early-type galaxies.
  • 5.1. Example of CIGALE as an SED generation tool: CIGALE generates theoretical SED grids for studying how star-formation histories and other assumptions affect observables and derived relations.It can also connect model generation to numerical and semi-analytic simulations.
  • 5.2. Example of CIGALE as a library: CIGALE can serve as a modular library for interactive exploration of dust attenuation across the FUV-to-far-IR spectrum.The demonstrated application uses sliders and is approximately 150 lines long.
  • 5.3. Example of CIGALE to estimate the physical properties of star-forming galaxies: In a KINGFISH comparison, CIGALE showed excellent consistency with other codes for average 100 Myr SFR, stellar mass, FUV attenuation, and dust luminosity.The comparison included results from MAGPHYS and the radiative-transfer code GRASIL.
  • 5.3. Example of CIGALE to estimate the physical properties of star-forming galaxies: CIGALE estimates physical quantities including FUV attenuation, dust luminosity, SFR, stellar mass, UV slope, and IRX.The example focuses on six commonly used properties for star-forming galaxies.
  • 5.3. Example of CIGALE to estimate the physical properties of star-forming galaxies: SFR, dust luminosity, and IRX show scatter below 0.1 dex, while β and AFUV have larger scatters of 0.217 and 0.157 dex, respectively.The exercise does not include uncertainties from model reliability, so these are lower limits on actual uncertainties.
  • 5.3. Example of CIGALE to estimate the physical properties of star-forming galaxies: CIGALE’s flexibility allows star-formation-history effects to be studied across galaxies from intense starbursts to elliptical systems.One application examines why more quiescent galaxies deviate from starburst relations.

6. Summary

The new generation of CIGALE is designed around modularity, clarity, and efficiency. It constructs panchromatic models using energy balance, estimates properties with likelihood-weighted statistics, and supports library-based extensions, while representing only a snapshot of an evolving code.

  • 6. Summary: CIGALE’s development is guided by modularity, clarity, and efficiency, including parallel execution on modern multicore processors.The modular design supports adding or swapping modules, while the code aims to remain understandable and easy to use.
  • 6. Summary: Energy balance makes absorbed UV-to-near-IR energy re-emerge self-consistently in the mid- and far-IR.Models extend from the FUV to radio and incorporate flexible star-formation histories, stellar populations, and ionised gas.
  • 6. Summary: Physical-property evaluation uses likelihood-weighted means and standard deviations while accounting for upper limits.Generated models can also be saved for theoretical studies or applied to observed objects.
  • 6. Summary: CIGALE can function as a library for building new applications beyond stand-alone SED fitting.The paper presents this as a consequence of the code’s versatility.
  • 6. Summary: The article describes CIGALE’s status as a snapshot of version 2018.1 rather than a final, static implementation.The code is described as constantly evolving, with future major developments anticipated.

Appendix B: Important implementation choices

CIGALE uses parallelisation, caching, and block-wise computation to improve performance and handle model grids that exceed available memory. These choices accelerate model generation and analysis while introducing scaling and memory trade-offs.

  • Parallelisation strategy: Parallelisation accelerates model generation and analysis by distributing work across processes and using shared arrays for fluxes, properties, and results.Shared arrays reduce communication and avoid copying data for individual subprocesses.
  • Parallelisation strategy: Computation speed initially increases with the number of physical cores, but memory–processor communication eventually causes sub-linear gains.Launching subprocesses also introduces a slight start-up delay.
  • Caching: Caching reuses model-creation modules and partially computed SEDs instead of repeatedly recomputing identical components or earlier model steps.Intermediate SEDs can support calculations at different redshifts without recomputing preceding steps.
  • Caching: Cache contents are ordered and pruned to retain only intermediate SEDs needed for remaining models, improving lookup speed while limiting memory usage.Unused intermediate SEDs are discarded as computation proceeds.
  • Block-wise computation: Block-wise computation lets CIGALE process model grids larger than computer memory and estimate target properties from each block's models.The overall likelihood-weighted standard deviation requires a dedicated combination of block-level standard deviations.

Appendix C.2.1: bc03

Appendix C.2.1 documents the bc03 module through separate tables of its input and output parameters.

  • Inputs: Table C.13 lists the input parameters of the bc03 module.
  • Module interface: Together, Tables C.13 and C.14 document the bc03 module's parameter interface.
  • Outputs: Table C.14 lists the output parameters of the bc03 module.

Appendix C.2.2: m2005

Appendix C.2.2 documents the m2005 module through separate tables of its input and output parameters.

  • Inputs: Table C.15 lists the input parameters of the m2005 module.
  • Module interface: Together, Tables C.15 and C.16 document the m2005 module's parameter interface.
  • Outputs: Table C.16 lists the output parameters of the m2005 module.

Appendix C.3.1: nebular

Appendix C.3.1 documents the nebular module through separate tables of its input and output parameters.

  • Inputs: Table C.17 lists the input parameters of the nebular module.
  • Module interface: Together, Tables C.17 and C.18 document the nebular module's parameter interface.
  • Outputs: Table C.18 lists the output parameters of the nebular module.

Appendix C.4.1: dustatt_modified_CF00

Appendix C.4.1 lists two tables documenting the input parameters of the dustatt_modified_CF00 module.

  • Tables C.19 and C.20 present the input parameters for dustatt_modified_CF00.

Appendix C.4.2: dustatt_modified_starburst

Appendix C.4.2 documents the dustatt_modified_starburst module through separate input- and output-parameter tables.

  • Table C.21 lists input parameters for dustatt_modified_starburst.
  • Table C.22 lists output parameters for dustatt_modified_starburst.

Appendix C.5.1: dale2014

Appendix C.5.1 documents the dale2014 module with separate tables for its inputs and outputs.

  • Table C.23 lists input parameters for the dale2014 module.
  • Table C.24 lists output parameters for the dale2014 module.

Appendix C.5.2: dl2007

The supplied material identifies dl2007 input and output parameter tables and gives the module's four named input parameters.

  • dl2007 uses four listed inputs: qpah, umin, umax, and gamma.qpah is the PAH mass fraction; umin and umax are the minimum and maximum radiation fields; gamma is the illuminated fraction.
  • Tables C.25 and C.27 are identified as input-parameter tables for dl2007.
  • Tables C.26 and C.28 are identified as output-parameter tables for dl2007.
  • CIGALE progressively applies modules, stores individual spectral components and the combined spectrum, and supports physical-property analysis and model generation.
Loading 1811.03094v1…