PaperlayerView exact revision on arXiv

Source-linked AI summary

Resolving galaxies in time and space: I: Applying STARLIGHT to CALIFA data cubes

R. Cid Fernandes, E. Perez, R. Garcia Benito, R. M. Gonzalez Delgado, A. L. de Amorim, S. F. Sanchez, B. Husemann, J. Falcon Barroso, P. Sanchez-Blazquez, C. J. Walcher, D. Mast

arXiv:1304.5788v1astro-ph.COastro-ph.IM

TL;DR

Integrated galaxy spectra provide global information but cannot resolve internal structures and processes. This paper builds and documents a CALIFA–STARLIGHT pipeline that handles correlated-error binning, spectral fitting, data organization, and multidimensional visualization. Using NGC 2916, it demonstrates the richness of spatially resolved stellar-population products while noting limits from single-galaxy scope and temporal resolution.

  • Problem

    Integrated spectra lack spatial resolution, limiting studies of galaxy evolution to global views rather than internal structures and processes.

  • Method

    The paper develops an end-to-end pipeline that preprocesses CALIFA datacubes, corrects correlated errors during Voronoi binning, fits zones with STARLIGHT, and organizes multidimensional outputs with PyCASSO.

  • Results

    The NGC 2916 case study yields spatially resolved maps, temporal and radial profiles, and space–time visualizations of stellar populations and mass growth.

  • Takeaways & Limitations

    Combining IFS with spectral synthesis provides a rich framework for analyzing galaxy stellar populations across spatial and temporal dimensions.

  • Takeaways & Limitations

    The demonstrated comparison between global and spatially resolved results is based on one example galaxy, so its generality remains uncertain.

Abstract

from arXiv · show

Fossil record methods based on spectral synthesis techniques have matured over the past decade, and their application to integrated galaxy spectra fostered substantial advances on the understanding of galaxies and their evolution. Yet, because of the lack of spatial resolution, these studies are limited to a global view, providing no information about the internal physics of galaxies. Motivated by the CALIFA survey, which is gathering Integral Field Spectroscopy over the full optical extent of 600 galaxies, we have developed an end-to-end pipeline which: (i) partitions the observed data cube into Voronoi zones in order to, when necessary and taking due account of correlated errors, increase the S/N, (ii) extracts spectra, including propagated errors and bad-pixel flags, (iii) feeds the spectra into the STARLIGHT spectral synthesis code, (iv) packs the results for all galaxy zones into a single file, (v) performs a series of post-processing operations, including zone-to-pixel image reconstruction and unpacking the spectral and stellar population properties into multi-dimensional time, metallicity, and spatial coordinates. This paper provides an illustrated description of this whole pipeline and its products. Using data for the nearby spiral NGC 2916 as a show case, we go through each of the steps involved, presenting ways of visualizing and analyzing this manifold. These include 2D maps of properties such as the v-field, stellar extinction, mean ages and metallicities, mass surface densities, star formation rates on different time scales and normalized in different ways, 1D averages in the temporal and spatial dimensions, projections of the stellar light and mass growth (x,y,t) cubes onto radius-age diagrams, etc. The results illustrate the richness of the combination of IFS data with spectral synthesis, providing a glimpse of what is to come from CALIFA and future surveys. (Abridged)

1. Introduction

Integrated spectra reveal global galaxy properties but lack spatial resolution needed to study internal structures and processes. This paper develops a CALIFA–STARLIGHT methodology for turning spatially resolved spectra into multidimensional stellar-population products.

  • Motivation: Integrated galaxy spectra lack spatial resolution and aperture coverage, preventing isolation of components and mapping of internal evolutionary processes.Examples include bulges, discs, bars, mergers, and stellar migration.
  • Motivation: CALIFA uses Integral Field Spectroscopy to map stellar populations, ionized gas, and kinematics across the optical extent of 600 nearby galaxies.The survey is designed to provide spatially resolved information rather than only a global view.
  • Scientific opportunity: Applying spectral synthesis to CALIFA data can recover spatially resolved star-formation histories and reveal inside-out growth in massive galaxies.Earlier CALIFA work found faster growth in inner regions and a relative growth efficiency maximum near 7 × 10^11 M⊙ for intermediate masses.
  • Contribution: The paper presents an end-to-end methodology from CALIFA datacubes through STARLIGHT analysis and multidimensional products, emphasizing processing and visualization rather than a full scientific exploration.The authors position the work as a methodological foundation for later studies and for IFS spectral-synthesis analyses more generally.

2. Data

The study uses CALIFA, a nearby-galaxy IFS survey whose reduced datacubes provide calibrated spectra, errors, and bad-pixel flags suitable for spatially resolved synthesis.

  • Survey: CALIFA’s mother sample contains 939 galaxies at 0.005–0.03 redshift, with observations planned for approximately 600 galaxies.The sample was selected from SDSS imaging to span the local color–magnitude diagram down to Mr < −18 mag.
  • Observations: Each galaxy is observed with PMAS/PPAK using V500 and V1200 setups, with the V500 providing broad optical coverage at approximately 6 Å resolution.The V500 range is approximately 3700–7500 Å, while V1200 covers approximately 3650–4600 Å at approximately 2.3 Å resolution.
  • Data products: Reduced cubes sample flux densities on a 1′′ × 1′′ × 2 Å grid and include propagated errors plus bad-pixel flags for unreliable entries.Fluxes are absolutely calibrated and corrected for Galactic extinction.

3. Pre-processing steps: From datacubes to STARLIGHT input

The preprocessing pipeline converts reduced CALIFA datacubes into reliable STARLIGHT inputs by masking poor regions, repairing flags and errors, rest-framing, resampling, and binning spectra with correlated-noise corrections.

  • Overview: Preprocessing comprises spatial masking, flag and error refinement, rest-framing, spatial binning, and wavelength resampling before spectral fitting.The operations are organized as a general preparation sequence for STARLIGHT or similar 3D fitting codes.
  • Spatial masks: The pipeline masks foreground sources, artefacts, and low-S/N regions, combining automated source detection with visual corrections that cannot be fully automated.A final outer S/N threshold is applied automatically; manual masking remains feasible but laborious for CALIFA’s scale.
  • Flags and errors: Spectra are rest-framed, flagged around strong sky and atmospheric features, and screened for anomalously large or small errors before fitting.The rest-framing applies a (1 + z)^3 correction, while the S/N estimate uses the 5590–5680 Å window and excludes flagged pixels.
  • Segmentation and binning: The datacube’s 5635 Å image defines HLR and supplies signal and noise maps for Voronoi zones targeting S/N = 20, with little binning inside 1 HLR.For CALIFA 277, 3649 non-masked spaxels were grouped into 1638 zones.
  • Correlated errors: Because dithering correlates neighboring spaxel errors, an empirical βz(Nz) correction produces larger zones and achieves the target S/N more reliably than assuming independent spaxels.The βz relation is measured from zone spectra and shows an initial rise followed by flattening as less-correlated spaxels are aggregated.

4. STARLIGHT runs and PyCASSO

STARLIGHT fits each zone with a stellar-population basis while PyCASSO organizes the many zone-level outputs into galaxy-wide files and analysis-ready spatial products.

  • STARLIGHT: STARLIGHT combines a population base to match each observed spectrum while fitting stellar extinction, velocity offset, and velocity dispersion.The analysis uses a base with four metallicities and 39 ages spanning 10^6 to 1.4 × 10^10 yr.
  • STARLIGHT: The fits cover 3800–6850 Å, use the Cardelli extinction curve with RV = 3.1, and mask major emission lines and Na I D.The selected base combines Granada and Vazdekis models and assumes a Salpeter initial mass function.
  • Outputs: STARLIGHT outputs include fit-quality, kinematic, extinction, mass, and population-vector quantities that can be post-processed into star-formation histories and mass-growth measures.The outputs are stored as separate ASCII files before PyCASSO reorganizes them for IFS analysis.
  • PyCASSO: PyCASSO addresses the organizational burden of IFS by packing all zone fits and preprocessing information into one file, reading it into accessible structures, and performing common post-processing.Its post-processing supports mappings and multidimensional stellar-population analyses.
  • Quality assessment: The pipeline includes maps of spectral-fit quality indicators such as mean relative deviation, χ2 per fitted flux, usable flux counts, and clipped-flux counts.The paper notes that mean relative deviation is especially useful because it does not explicitly require the error spectrum and has an intuitive interpretation.

5. Results

The NGC 2916 showcase demonstrates how spectral-fit quality can be assessed spatially, while highlighting masking, flagging, clipping, and resolution-related caveats.

  • 5.1. Fit quality assessment: The median fit residual is 3.0% for NGC 2916, with Δ spanning 1.0–5.6% and increasing toward lower-S/N outer regions.The Δ metric is preferred over χ² because it is easier to interpret and does not explicitly require the often unavailable error spectrum.
  • 5.1. Fit quality assessment: Across roughly 10^5 CALIFA zone spectra, the median Δ is 4%, fewer than 2% exceed 10%, and this success depends on carefully derived errors and flags.The authors identify non-masked emission lines and artifacts as common explanations for the highest-Δ spectra, although such bugs are rare in CALIFA.
  • 5.1. Fit quality assessment: A Δ threshold of 8–10% is suggested for spatial quality control, while results beyond 2–3 HLR require greater care because residuals tend to rise outward.No Δ-based cut is needed for NGC 2916.
  • 5.1.1. Subtleties and caveats on quality control: Large residuals can reflect poor data or inadequate masks and flags, so blind quality-control failures do not necessarily indicate bad spectra.Unmasked emission lines, velocity shifts, sky residuals, and missed artifacts can all inflate Δ.
  • 5.1.1. Subtleties and caveats on quality control: The pipeline uses conservative 4σ clipping, while custom residual-based emission-line masks remain unavailable for CALIFA.Clipping is intended to identify pixels that are difficult to fit and may be nonstellar or spurious.

5.2. 2D maps: Stellar light, mass and extinction

The 2D products reconstruct luminosity, extinction, and stellar mass across NGC 2916, revealing similar light and mass structure but weaker spiral-arm contrast in mass.

  • 5.2. 2D maps: Stellar light, mass and extinction: The reconstructed products map surface brightness at 5635 Å, stellar extinction, dereddened luminosity, and stellar mass surface density.These products illustrate the conversion from fitted stellar light to mass in spatially resolved data.
  • 5.2. 2D maps: Stellar light, mass and extinction: Voronoi effects are strongest in the extinction map, whereas luminosity and mass images are dezonified using each spaxel’s fractional zone flux.Intensive quantities such as extinction, mean ages, and velocity dispersion cannot be dezonified.
  • 5.2. 2D maps: Stellar light, mass and extinction: Stellar extinction is low at approximately 0.1–0.2 mag, with slight enhancements in the nucleus and spiral arms and limited overall variation.The map shows relatively little extinction variation across the galaxy face.
  • 5.2. 2D maps: Stellar light, mass and extinction: Light and mass both show a bulge within roughly 0.5 HLR and a disc beyond it, but spiral arms are less prominent in mass than in light.The lower mass contrast of the arms reflects their higher luminosity-to-mass ratios.
  • 5.2. 2D maps: Stellar light, mass and extinction: The total present-day stellar mass is 6.5 × 10^10 M⊙, excluding the masked region around a foreground star.Profile-based hole filling would increase the quoted mass by 4% for CALIFA 277.

5.4. 2D maps: Mean ages and metallicities

The synthesis compresses stellar-population distributions into light- and mass-weighted mean ages and metallicities, revealing central-to-outer gradients and age–metallicity structure in CALIFA 277.

  • Mean-age maps increase toward the center, while metallicity shows negative gradients that flatten within the bulge region.
  • Mass-weighted ages span a smaller dynamical range than light-weighted ages because old populations carry substantial weight.
  • The spiral arms appear as lower-age regions outside 1 HLR, especially in the western half of the galaxy.
  • Central populations show an age–metallicity anticorrelation, although the broader positive relation suggests limited impact from age–metallicity degeneracy.
  • Young, intermediate, and old light-fraction maps add detail but largely reproduce the information already visible in the light-weighted mean-age map.

5.6. 2D maps: Star formation rates and IFS-based variations over Scalo’s b parameter

The paper defines spatially resolved star-formation rates over multiple lookback times and normalizes them against local, galactic, or contemporaneous regional references. These complementary maps reveal spatially and temporally varying activity, including bursty star formation and distinct arm structures.

  • SFR surface density is estimated by averaging stellar mass formed since a chosen lookback time, but the estimator becomes less useful as that time approaches the full 14 Gyr base.
  • The local birthrate parameter compares recent SFR with the lifetime average at the same spatial location, whose interpretation differs from a galaxy-wide measure because old stars may have migrated.
  • The definitions compare present activity with different temporal and spatial reference values, including local, whole-galaxy, and regional baselines.
  • The b_gal maps suggest bursty rather than continuous star formation because the 142 Myr map is nearly featureless while shorter and longer windows retain arm structure.The logarithmic age resolution limits recognition of short bursts to the youngest sampled ages.
  • The novel c index compares present star formation at each location with present star formation across the galaxy, producing time-windowed spatial snapshots.At 14 Myr it highlights ongoing spiral-arm activity; longer windows blur the structures and approach the corresponding b_gal maps.
  • The three SFR normalizations provide different and complementary views despite sharing the same numerator.

5.7. 1D spatial maps: radial profiles

Pycasso compresses spatially resolved maps into radial profiles using circular or elliptical coordinate transformations and bin-wise averaging. The resulting profiles retain individual-spaxel variation while showing radial means and dispersions.

  • Pycasso provides circular and elliptical xy-to-R conversion tensors to support azimuthal averaging of 2D data.
  • Radial averaging can either aggregate mass or light before division or directly average spaxel surface-density values; this paper uses the latter for uniformity.The two methods almost always give nearly identical results, but area averaging does not apply to quantities such as extinction or velocity dispersion.
  • Radial profiles display individual spaxel values together with bin-wise means and dispersions, allowing 2D map features to be inspected as functions of radius.

5.8. 1D temporal maps: evolutionary curves

The evolutionary curves compare how light, mass, and star formation change with time across radial regions, revealing faster inner growth and important limits from logarithmic age resolution and dust variation.

  • 5.8. 1D temporal maps: evolutionary curves: The 39-age basis is smoothed to 0.5 dex in log t because fossil-record methods have logarithmic age resolution and require SFH compression.Age resolutions between 0.5 and 1 dex are described as reasonable.
  • 5.8. 1D temporal maps: evolutionary curves: The mass half-radius is 3.4 kpc, smaller than the 4.8 kpc light half-radius because stellar-population gradients make light and mass differently distributed.Regions inside and outside 1 HLR contain different masses despite containing equal light by definition.
  • 5.8. 1D temporal maps: evolutionary curves: The nucleus reached 80% of its mass by 8.5 Gyr, whereas regions beyond 1 HLR reached this level at 1.9 Gyr, indicating inside-out mass assembly.The paper notes that this ordering applies to essentially all massive galaxies.
  • 5.8. 1D temporal maps: evolutionary curves: The SFR per unit area generally decreases outward, but reverses this trend during the last approximately 500 Myr.Smoothing makes SFR(t) continuous and interpretable as an instantaneous SFR.
  • 5.9. The whole versus the sum of the parts: Integrated and spatially summed stellar properties agree closely: total stellar masses differ by 1%, while weighted ages and metallicities agree within 0.05 dex.In this example, integrated extinction, mean ages, and metallicities also approximate the spatially resolved values near 1 HLR.
  • 5.9. The whole versus the sum of the parts: The integrated spectrum reproduces the summed spatial SFH in this galaxy, but this agreement may not generalize when dust varies strongly across the system.Spatially varying extinction can drive age-extinction compensation, while population gradients and kinematics introduce additional uncertainty.

5.10. Space × time diagrams: SFHs in 2D

The space–time diagrams compress spatial and metallicity dimensions to visualize radially resolved galaxy evolution, showing distinct light, mass, and star-formation histories across radius and age.

  • 5.10. Space × time diagrams: SFHs in 2D: The diagrams are made by compressing x and y into radius, collapsing metallicity, smoothing in log t, and producing radially averaged SFH(R,t) maps.The resulting panels represent luminosity density, formed stellar-mass surface density, and time-dependent SFR per unit area.
  • 5.10. Space × time diagrams: SFHs in 2D: The R-t map shows old stellar light concentrated in the bulge, while young stars shine approximately equally across radius.Disc populations older than about 1 Gyr are more smoothly distributed, whereas the bulge contains concentrations near 2.5 Gyr and above 10 Gyr.
  • 5.10. Space × time diagrams: SFHs in 2D: SFR(R,t) and formed-mass M′R,t diagrams contain the same information, but the t^-1 factor makes their visual patterns differ.Because age sampling is logarithmic, SFR is proportional to stellar mass formed at t divided by t.
  • 5.10. Space × time diagrams: SFHs in 2D: The logarithmic age resolution prevents fossil-record methods from distinguishing bursts much shorter than their current age.Apparently similar recent and ancient peaks can span time intervals differing by roughly a factor of a thousand.
  • 5.10. Space × time diagrams: SFHs in 2D: The diagrams provide evidence for inside-out growth: outer regions assemble their mass more slowly and over a more extended period.They are presented as a novel visualization of galaxy evolution across both space and time.
  • 5.10. Space × time diagrams: SFHs in 2D: The figures include pseudo-snapshots and radius–age projections of stellar light, formed mass, and SFR, preserving spatial information through radial averaging.The whole-galaxy evolution is reconstructed by summing the spatial components.

5.11. Space vs. time ”snapshots”

The pipeline visualizes spatially resolved stellar mass assembly across age, but these age maps describe where stars of different ages are located today rather than historical snapshots. Residual spectra additionally expose spatial trends in ionized-gas emission.

  • Age slices through the cumulative mass-growth cube reveal the full two-dimensional structure of CALIFA 277’s mass assembly history.
  • These age sequences can represent several quantities, including mass, luminosity, star formation rate, and metallicity, but not extinction or stellar kinematics as functions of age.
  • The age panels map the present-day locations of stars with a given age rather than showing the galaxy as it appeared at different look-back times.
  • Residual spectra from the stellar fits isolate emission lines for diagnosing ionized-gas kinematics, extinction, and chemical abundance.
  • The residual cube shows brighter emission lines toward spiral arms, while relatively faint nuclear lines indicate a low-luminosity AGN despite the active nucleus.

6. Summary

The paper presents an end-to-end pipeline that converts reduced IFS cubes into spatially resolved spectral-synthesis products and multidimensional stellar-population views. Applied to NGC 2916, it yields maps, profiles, and age–radius visualizations that reveal structured galaxy evolution while retaining important interpretive limits.

  • The pipeline processes reduced IFS data through STARLIGHT, transforming (x, y, λ) cubes into multidimensional maps of synthesis products.
  • Pre-processing combines spatial masking, flag and error adjustments, rest-framing, resampling, and Voronoi binning targeting spectra with S/N ≥20.
  • Spatially resolved products include extinction and stellar-mass maps, mean ages and metallicities, star-formation measures, and IFS-specific diagnostics comparing present and past activity across regions.
  • Temporal and radial profiles reveal negative age and metallicity gradients and spatially varying mass-growth speeds compatible with inside-out formation in the example galaxy.
  • The tools are intended for future studies of the astrophysical implications of spatially resolved spectral-synthesis results across the CALIFA survey.
Loading 1304.5788v1…