PaperlayerView exact revision on arXiv

Source-linked AI summary

The Data Analysis Pipeline for the SDSS-IV MaNGA IFU Galaxy Survey: Overview

Kyle B. Westfall, Michele Cappellari, Matthew A. Bershady, Kevin Bundy, Francesco Belfiore, Xihan Ji, David R. Law, Adam Schaefer, Shravan Shetty, Christy A. Tremonti, Renbin Yan, Brett H. Andrews, Joel R. Brownstein, Brian Cherinka, Lodovico Coccato, Niv Drory, Claudia Maraston, Taniya Parikh, José R. Sánchez-Gallego, Daniel Thomas, Anne-Marie Weijmans, Jorge Barrera-Ballesteros, Cheng Du, Daniel Goddard, Niu Li, Karen Masters, Héctor Javier Ibarra Medel, Sebastián F. Sánchez, Meng Yang, Zheng Zheng, Shuang Zhou

arXiv:1901.00856v2astro-ph.GA

TL;DR

MaNGA's spatially resolved datacubes require standardized analysis products that remain accessible and reusable across users and studies. This paper presents the DAP's modular workflow and publicly released products, emphasizing measurements close to the data. The DAP robustly measures key galaxy properties across the DR15 data, while users must account for quality flags and astrometric limitations.

  • Problem

    The complexity of spatially resolved MaNGA data creates demand for uniform analysis products that reduce duplicated common tasks while supporting broader scientific use.

  • Method

    The paper presents a modular DAP that uses pPXF-based spectral fitting, template libraries, masking, and configurable workflow components to derive data-close measurements.

  • Results

    The DAP provides accurate stellar kinematics to S/Ng∼10 and ∼50 km s−1 minimum velocity dispersion, with random spectral-index errors behaving well to S/Ng≳2.

  • Takeaways & Limitations

    The released products provide stellar kinematics, emission-line properties, and spectral indices for MaNGA analyses, with calibrated error guidance across signal-to-noise.

  • Takeaways & Limitations

    Users should apply MAPS quality bitmasks and account for astrometric registration errors that increase repeat-observation scatter in kinematic measurements.

Abstract

from arXiv · show

Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) is acquiring integral-field spectroscopy for the largest sample of galaxies to date. By 2020, the MaNGA Survey --- one of three core programs in the fourth-generation Sloan Digital Sky Survey (SDSS-IV) --- will have observed a statistically representative sample of 10$^4$ galaxies in the local Universe ($z\lesssim0.15$). In addition to a robust data-reduction pipeline (DRP), MaNGA has developed a data-analysis pipeline (DAP) that provides higher-level data products. To accompany the first public release of its code base and data products, we provide an overview of the MaNGA DAP, including its software design, workflow, measurement procedures and algorithms, performance, and output data model. In conjunction with our companion paper Belfiore et al., we also assess the DAP output provided for 4718 observations of 4648 unique galaxies in the recent SDSS Data Release 15 (DR15). These analysis products focus on measurements that are close to the data and require minimal model-based assumptions. Namely, we provide stellar kinematics (velocity and velocity dispersion), emission-line properties (kinematics, fluxes, and equivalent widths), and spectral indices (e.g., D4000 and the Lick indices). We find that the DAP provides robust measurements and errors for the vast majority ($>$99%) of analyzed spectra. We summarize assessments of the precision and accuracy of our measurements as a function of signal-to-noise, and provide specific guidance to users regarding the limitations of the data. The MaNGA DAP software is publicly available and we encourage community involvement in its development.

1. INTRODUCTION

The MaNGA DAP addresses the complexity of spatially resolved spectroscopy by providing uniform, publicly accessible analysis products while retaining flexible, reusable algorithms. This paper presents the DAP's design, workflow, measurements, assessments, and DR15 data products.

  • Publicly released data products broaden access to astronomical observations and support analyses combining observations across wavelengths and telescopes.
  • The DAP provides uniform analysis products for common tasks, helping users design samples rapidly and combine standard measurements with custom analyses.
  • The DAP's core algorithms are designed to be independent of MaNGA-specific use, enabling users to build new Python scripts for varied data sets.
  • The first public DR15 release includes DAP code version 2.2.1 and output products accessible through the SDSS website and Marvin.
  • Primary outputs include stellar kinematics, emission-line fluxes and kinematics, and continuum spectral indices, with measurements focused close to the spectra and minimal model-based assumptions.
  • The paper overviews the DAP's algorithmic structure, inputs, templates, analysis modules, performance assessments, and output data model.

2. DAP QUICK-START GUIDE

The quick-start guide uses representative DAP maps and model spectra to show the products, workflow, binning choices, masks, and cautions needed for interpretation. Users are directed to inspect the underlying spectra and model LOGCUBE when map features appear physically counter-intuitive.

  • Figure 1 samples DAP maps spanning signal-to-noise, stellar kinematics, emission-line properties, and spectral indices for one representative DR15 datacube.
  • The MAPS file stores DAP measurements as spatial maps matching the DRP datacube, with associated inverse variance, quality assessments, and bitmasks.
  • Map masks reflect low signal-to-noise, low flux, or quality-assurance criteria, while DAP masking is otherwise largely limited to numerical or computational issues.
  • The 0.′′5 × 0.′′5 spaxels are substantially smaller than MaNGA's 2.′′5 spatial-resolution FWHM, reflecting a compromise involving covariance and spatial sampling.
  • Hybrid binning measures stellar kinematics in spectra meeting S/Ng≳10 while measuring emission lines and spectral indices in individual spaxels.
  • The DAP analyzes each datacube under two DR15 analysis approaches, so users must choose which DAPTYPE products to use.
  • When map features appear physically counter-intuitive, users should inspect the MaNGA spectra and associated model LOGCUBE rather than rely on visual maps alone.
  • Completing both DR15 analysis approaches takes median times of 2.0, 3.1, 4.6, 6.6, and 8.7 hours for 19-, 37-, 61-, 91-, and 127-fiber datacubes, respectively.Some 127-fiber datacubes required up to 25 hours.

3. DAP INPUTS

MaNGA provides calibrated spectra and datacubes from dithered fiber observations, which serve as the primary inputs to the DAP. Photometric and redshift information, plus analysis plans, supply additional execution context.

  • MaNGA uses 17 fiber bundles containing 19, 37, 61, 91, or 127 fibers, with 2′′ on-sky fiber diameters.
  • Each plate is observed with a three-point dither pattern designed to fill IFU gaps and improve field-of-view sampling uniformity.
  • The DRP produces wavelength-, flux-, and astrometrically calibrated spectra, incorporating adjustments for MaNGA's observing configuration.
  • Spectra are combined into row-stacked files and uniformly sampled datacubes, with 0.′′5 spaxels and a velocity scale of ΔV = 69 km s−1.
  • The DAP uses photometric geometry to calculate elliptical polar coordinates R and θ, with radius expressed in arcseconds and effective-radius units.

4. WORKFLOW

The DAP processes eligible DRP datacubes through a modular, configurable workflow whose results feed MAPS and model LOGCUBE products. Its execution is governed by analysis plans and quality flags, with intermediate reference files supporting reuse and recovery.

  • At survey level, the DAP runs once per PLATEIFU datacube for galaxy observations with an initial redshift estimate.
  • The workflow writes products into MAPS and model LOGCUBE extensions, while Figure 3 encodes module objectives, dependencies, and execution order.
  • For DR15, the DAP analyzed 4731 observations, including datacubes marked as critical failures whose quality flags are propagated for cautious use.
  • A DAPTYPE is an analysis plan composed of keywords selecting preset configurations for the primary DAP modules.
  • The six primary modules assess data quality, spatially bin spectra, measure stellar kinematics, compute emission-line moments, fit emission lines, and measure spectral indices.
  • Modules execute serially and generally depend on preceding results; emission-line moments are computed both before and after emission-line modeling.
  • Reference files let the DAP reuse shared results, restart after failures, and expose more information than the two primary output files.

5. HIERARCHICAL CLUSTERING OF A SPECTRAL-TEMPLATE LIBRARY

The paper distills the 985-star MILES library into compact spectral templates using hierarchical clustering based on spectral similarity, preserving positive template weights while reducing computation. For DR15, this produces 42 MILES-HC templates with an approximately 25-fold execution-time gain and acceptable effects on kinematics and fit quality.

  • Motivation: Spectral-library distillation is critical for large surveys because pPXF execution time grows slightly faster than O(N_tpl) with the number of templates.The DAP uses pPXF for both stellar-kinematics measurement and emission-line modeling.
  • Method: The approach avoids parameter-grid limitations and preserves positivity constraints that PCA-based subsampling would lose, while remaining applicable to spectral-template libraries generally.The authors describe the approach as simple and not necessarily optimal, but reasonable for their purposes.
  • Method: The adopted method hierarchically clusters spectra as vectors in channel space, using a pPXF residual-based distance matrix to capture relevant spectral information.The distance uses a robust residual standard-deviation estimate from fitting one spectrum with another as the template.
  • Results: 49 clustered spectra were obtained from the 985-star MILES library using dmax = 0.05, with cluster sizes ranging from individual spectra to hundreds of stars.Clusters concentrate in stellar-parameter space, although the relevant criterion is whether they capture the library’s information for full-spectrum fitting.
  • Results: Removing low-S/N or emission-line templates yields 42 MILES-HC spectra, providing an approximately 25-fold speedup with acceptable effects on stellar kinematics despite moderately worse fit statistics.MILES-HC fits to MaNGA spectra are generally no worse than fits using simple-stellar-population templates.
  • Results: Cluster number is not generally related to mean stellar parameters, and the final template spectra are ordered by cluster number rather than physical-parameter similarity.Figure 5 omits clusters removed for low S/N or emission lines from flaring late-type stars.

6. SPATIAL BINNING

The DAP bins neighboring spaxels with adaptive Voronoi methods to achieve unbiased stellar-kinematic measurements while accounting for spatial covariance introduced during datacube reconstruction. Covariance is spatially localized but substantially reduces the effective number of independent measurements, so ignoring it produces bins whose formal S/N is overestimated.

  • 6. SPATIAL BINNING: Adaptive Voronoi binning averages neighboring spaxels to reach a minimum S/N required for unbiased stellar-kinematic measurements, especially velocity dispersion.The DAP uses the Cappellari & Copin (2003) Voronoi algorithm for this purpose.
  • 6.2. Spectral Signal-to-Noise and Spatial Covariance: The spatial correlation matrix is nearly empty away from the diagonal, with non-zero correlations extending fewer than 5 or 6 spaxels.The observed structure is consistent with sampling a 2.′′5 fiber beam on 0.′′5 × 0.′′5 spaxels.
  • 6.2. Spectral Signal-to-Noise and Spatial Covariance: A Gaussian fit describes the correlation decline with spaxel separation, with scale parameter σ = 1.92 spaxels (0.′′96).The fit is derived from spaxels within the convex hull of the fiber-observation centers.
  • 6.2. Spectral Signal-to-Noise and Spatial Covariance: The effective number of independent measurements is comparable to the 127 fibers in the relevant IFU and substantially smaller than the 1905 independent fiber observations used to construct the datacube.This comparison quantifies the impact of datacube covariance on statistical independence.

7. STELLAR KINEMATICS

The DAP stellar-kinematics analysis uses pPXF with tailored template, resolution, and continuum choices, and evaluates their effects on velocity-dispersion measurements. These tests support DR15 choices that balance accuracy, robustness, and computational cost while identifying limits at low dispersion and low signal-to-noise.

  • Stellar-continuum fitting: pPXF models galaxy spectra as template mixtures convolved with a line-of-sight velocity distribution, providing the DAP’s stellar-continuum fitting workhorse.The DAP assigns stellar-continuum templates to a single kinematic component in the described implementation.
  • Spectral-resolution corrections: First-order instrumental-resolution corrections can underestimate the correction by a few percent, producing a σ∗ bias of ≲5% near 70 km s^-1 and larger relative bias at lower dispersion.The approximation assumes all spectral regions contribute equally, although spectral-feature influence depends roughly on equivalent width.
  • Template library comparison: A 25-fold execution-time reduction with minimal kinematic change motivated adopting the smaller MILES-HC library instead of the full MILES library for DR15.The full library uses 985 templates versus 42 MILES-HC templates, with execution time scaling roughly as O(Ntpl).
  • Template down-selection: The ALL and NZT template-selection approaches produce reasonably consistent results, with most differences below 50% of the measurement error.For the few spaxels with larger dispersion differences, changes are 2–4 km s^-1, or 1–2%; NZT reduces execution time by a factor of 2.5–3 and was adopted for DR15.
  • Continuum polynomials: Polynomial-order effects are strong for p < 3 but small below a few percent for p > 3, supporting the continued use of p = 8.Trends at p ≥ 9 are less evident with robust statistics, while p < 9 is relatively resistant to low-S/N systematics.
  • Spectral-resolution matching: Keeping templates at native resolution improves the error budget and reduces measurements reaching the σobs ≈ 0 boundary, especially when σ∗ is below the MaNGA instrumental dispersion.Template-resolution uncertainties propagate into σobs uncertainty, and higher-resolution templates are advantageous under nominal assumptions.

7.5. Performance

The DAP’s stellar-kinematic uncertainties depend on both dispersion and signal-to-noise, with repeat observations and simulations revealing boundary biases, error-calibration limits, and practical guidance for DR15 users.

  • 7.5.1. Empirical Uncertainties from Repeat Observations: Repeat observations show generally consistent stellar-kinematic maps, with residuals varying mainly beam-by-beam because of spatial covariance.At the lowest surface brightness, σ∗ residuals vary more between spaxels because of systematic error at low signal-to-noise and low dispersion.
  • 7.5.1. Empirical Uncertainties from Repeat Observations: For σ∗≳100 km s−1, velocity errors follow δ(∆V)≈⟨σ∗⟩(S/Ng)−1, while low-dispersion measurements have substantially larger fractional errors.At S/N=10, the lowest dispersion bin reaches approximately 30% velocity error and 60% σ∗ error; σ∗ errors also become weakly dependent on S/N in the low-dispersion, low-S/N regime.
  • 7.5.1. Empirical Uncertainties from Repeat Observations: Formal errors generally underestimate empirical errors, although the two are usually consistent within a factor of two except for σ∗ at low dispersion and S/Ng.The repeat-observation comparison assesses relative repeatability and formal-error reliability but cannot determine absolute systematic error because the true kinematics are unknown.
  • 7.5.2. Parameter-Recovery Simulations: Idealized simulations constrain intrinsic method limits but are not necessarily representative of real applications because template, resolution, and LOSVD mismatches are difficult to model.These limitations complicate direct transfer of simulation results to the broad parameter space of practical pPXF applications.
  • 7.5.2. Parameter-Recovery Simulations: Low σ∗ and low S/N produce increasing systematic error because solutions accumulate at the physical σ∗=0 boundary and template-resolution mismatches strongly affect σ∗.The average σ∗ is positively biased when the true value approaches zero, with the bias proportional to the σ∗ errors.
  • 7.5.3. Algorithm Implications: Binning stellar-kinematics spectra to S/Ng∼10 limits boundary-driven low-dispersion bias and is expected to keep reported-error differences near 10–20% at low S/Ng.This DR15 choice was based on the combined empirical and simulated performance assessments.
  • 7.7. Usage Guidance: Measurements of σ∗<50 km s−1 at S/N∼10 require extreme caution because systematic errors may reach 10–20% and random errors 30–40%.Measurements near σ∗∼35 km s−1 can be reasonable at S/N>20, but their occurrence is expected to be small.
  • 7.7. Usage Guidance: Astrometric registration errors add position-association uncertainty beyond the spectral error, suggesting an additional factor of log S/Ng over 10≲S/Ng≲100.Users associating measurements with spatial locations should include this increase unless they model astrometric errors directly.

8. BANDPASS INTEGRALS

The DAP uses a discrete bandpass-integral formalism for emission-line and spectral-index measurements, accounting for partial pixel coverage, masking, wavelength sampling, and pseudo-continuum estimation.

  • BANDPASS INTEGRALS: Bandpass integrals are computed as discrete sums over wavelength-sampled vector values, using each pixel’s included fraction and wavelength step.The formalism applies to both emission-line moments and spectral indices.
  • BANDPASS INTEGRALS: DR15 assigns zero weight to masked pixels within a bandpass and flags spectra containing such masked pixels rather than interpolating or replacing them with model values.This preserves the direct data-based treatment of the affected passband measurements.
  • BANDPASS INTEGRALS: The same integration framework, with luminosity weighting, supplies the aggregated stellar velocity dispersion within 1Re in the DAPall file.This connects the bandpass formalism to an aggregate stellar-kinematic quantity.
  • BANDPASS INTEGRALS: The DAP analyzes spectra sampled linearly in log wavelength, so each pixel’s wavelength step is proportional to its wavelength.This sampling condition is part of the stated measurement formalism.
  • BANDPASS INTEGRALS: When needed, the DAP estimates a linear pseudo-continuum from blue and red sidebands surrounding the main spectral feature.The sideband calculations use flux density in the shorter- and longer-wavelength passbands.

9. EMISSION-LINE MEASUREMENTS

The DAP models emission lines with a second pPXF full-spectrum fit that simultaneously fits gas features and the stellar continuum while holding stellar kinematics fixed. Its templates support configurable line profiles, flux ratios, and kinematic ties, with Fourier-space construction improving accuracy for narrow or under-sampled lines.

  • Emission-line modeling: The DAP’s second pPXF fit simultaneously models all emission-line features and re-optimizes stellar-continuum template weights while fixing stellar kinematics from the first fit.
  • Emission-line modeling: pPXF represents gas emission as spectral templates, unlike GANDALF, which adds Gaussian emission lines to the stellar spectrum.The template-based approach is more general and flexible but requires initial template construction.
  • Emission-line templates: The DR15 MILES library restricts the modeled wavelength range, excluding the [S III] lines at 9071.1 Å and 9533.2 Å.The same library is used for stellar kinematics but is more limiting for emission-line modeling.
  • Spatial binning: DR15 spatial binning targets broad-band g-filter S/N for stellar-kinematics precision and may therefore be inappropriate for emission-line-focused science.
  • Emission-line templates: DR15 template construction specifies line-profile forms, fixed flux ratios, and kinematic ties that impose shared velocities or velocity dispersions.The adopted fixed ratios and ties are documented in Table 3.
  • Emission-line templates: Templates use narrow profiles based on the minimum instrumental dispersion, with a quadrature offset setting the minimum profile dispersion to 10 km s^-1.This helps pPXF return nonzero dispersion measurements while accommodating line-spread-function variation across the datacube.
  • Sub-pixel templates: Fourier-space template construction produces numerically accurate Gaussian profiles regardless of line width and permits Dirac delta profiles when σx = 0.For well-sampled lines with σx ≳1, the profile reduces to a normalized Gaussian summing to unity.
  • Sub-pixel templates: Ignoring integration over spectral pixels is equivalent to adopting an instrumental dispersion approximately 3% below the true value for a well-sampled line-spread function.

10. SPECTRAL INDICES

The DAP spectral-index measurements have predictable noise behavior and generally subdominant systematic errors, but reported random errors are underestimated and some model-driven biases remain. A tested combination method reproduces direct measurements on binned spectra for most indices.

  • 10.2. Possible Model-Driven Biases: Emission-line model errors can bias indices spanning emission-line regions, while velocity-dispersion corrections introduce additional model-dependent systematic errors.The latter can arise from biased dispersion measurements or local inaccuracies in reproducing the observed spectrum.
  • 10.3.2. Error Calibration: D4000 formal errors are accurate, unlike absorption-line errors, whose propagation is inadequate partly because of continuum-calculation uncertainties.For D4000, the assessment finds δϵ = 1.
  • 10.3.2. Error Calibration: At S/Ng∼10, typical errors are ∼0.5–1 Å or ≲0.03 dex, while DAP-reported errors are smaller than empirical estimates by ∼30–100%.The typical values apply to indices measured in Å or magnitudes, respectively.
  • 10.3.3. Binning Spectral Indices: Except for D4000, direct measurements on binned spectra agree with combined individual-spectrum measurements to better than 10% of the expected error for most calculations.The observed systematic shift is at most a few percent of the expected error.
  • 10.3.1. Idealized Simulations: Systematic errors are always smaller than random errors in idealized simulations and typically below 10% of the random error.This holds across index, template, velocity dispersion, and S/Ng, although systematic errors can matter relative to the index at low S/N and low equivalent width.
  • 10.3.2. Error Calibration: Random spectral-index errors follow an inverse proportionality with S/Ng, allowing users to select the S/Ng needed for a target error.The calibrated proportionality data are provided in Table 4.
  • 10.3.3. Binning Spectral Indices: The DAP provides a method for combining index measurements from multiple spectra without recomputing the index on the binned spectrum.The method is assessed against direct binned-spectrum measurements.
  • 10.4. Flagging: Spectral-index measurements beyond the fitted template range can receive NOCORRECTION flags when velocity-dispersion corrections cannot be calculated.For DR15, this is critical for passbands at λ ≳7400 Å.

11. PERFORMANCE

The DAP succeeds on nearly all attempted DR15 analyses and its fitting metrics broadly meet expectations for both stellar-continuum and emission-line modules. Remaining failures and poor fits are concentrated in identifiable edge cases, including broad AGN emission, cosmic rays, foreground stars, and overlapping non-target galaxies.

  • 11.1. Success Rate: Only 22 of 9462 DR15 DAP executions failed, a failure rate of 0.2%.Some observations succeeded with the VOR10-GAU-MILESHC method but failed with the HYB10-GAU-MILESHC approach.
  • 11.2. Statistical Performance: For well-fit spectra, fRMS should decrease inversely with S/Ng and reduced χ2 should remain near unity under robust flux-error estimates.These expectations provide the basis for assessing the full-spectrum fits.
  • 11.2. Statistical Performance: The expected fRMS and reduced χ2 behavior is well met by both stellar-continuum and emission-line fitting modules.The metrics are evaluated across fitted DR15 spectra as a function of S/Ng.
  • 11.2. Statistical Performance: The median reduced χ2 is approximately 0.9, consistent with slightly overestimated data errors by about 5% or possible mild over-fitting.At S/Ng≲5, large fRMS values occur in spectra with very low flux levels.
  • 11.2. Statistical Performance: Poor fits represent fewer than 1% of spectra, with coherent outlier structures often arising from a small number of datacubes.The paper identifies unmasked foreground stars and non-targeted galaxies among the relevant categories.
  • 11.2. Statistical Performance: AGN-focused studies should treat DAP products cautiously because broad, multi-component emission lines are poorly handled.Other analysis products may be required for these cases.
  • 11.2. Statistical Performance: Cosmic rays near emission lines can produce emission-line-module outliers even when the underlying model is reasonable over most of the spectrum.Their impact is localized, so direct assessment is needed for affected cases.

12. OUTPUT PRODUCTS

The DAP produces spatially organized MAPS and model LOGCUBE files, plus the DAPall summary catalog, with quality flags and usage limitations that users must consider.

  • Output organization: Each DAP module produces a reference file, while final outputs are reformatted from spectrum-oriented structures into spatial formats matching DRP datacubes.Reference files can reconstruct module state and reduce redundant analysis steps.
  • Output organization: Each successfully analyzed DR15 datacube has two MAPS files and two model LOGCUBE files, one pair for each DAPTYPE.The detailed output data models are provided in Appendix C and through the DR15 website.
  • DAPall catalog: The DAPall catalog aids sample selection, but its measurements lack uncertainties and are not recommended for science without detailed understanding of data quality.Its methods are simple; for example, the star-formation rate does not account for attenuation.
  • MAPS files: The MAPS file provides spaxel-by-spaxel measurements as maps matching the DRP datacube, including single-map quantities and channelized emission-line measurements.The file organizes derived quantities in extensions whose spatial dimensions match the input datacube.
  • MAPS files: DAP measurements require provided corrections, quality-bitmask checks, and awareness that binned-spectrum results repeat across every spaxel in the bin.The BINID extension helps identify repeated measurements, which matters for modeling or radial binning.
  • Usage limitations: DR15 MAPS data lack covariance matrices despite significant spatial covariance between spaxels, and nonsmooth adjacent-spaxel results can indicate systematic error.A first-order correlation matrix can be approximated from spaxel separation, but users should inspect the data and models when maps appear unreasonable.

13. CONCLUSIONS

The paper presents the MaNGA DAP, its public products, performance, and design philosophy. The DAP robustly measures key spectral properties, while current limitations and future extensions define its scope.

  • Conclusions: The paper describes the DAP, its output products, and performance, and directs users to the quick-start guide and output-product documentation.These resources provide summary-level usage information and routes to detailed discussions of each data product.
  • Performance: The DAP robustly fits MaNGA spectra and provides accurate stellar kinematics to S/Ng∼10 down to a minimum velocity dispersion of ∼50 km s−1.Repeat observations and simulations show that formal stellar-kinematic errors are consistent with direct repeat-observation estimates.
  • Design philosophy: The DAP focuses on measurements made directly from individual spectra, and its subroutines are designed for adaptation to other data sets with similar wavelengths and spectral resolution.The pipeline itself is tuned to MaNGA data and format, but the underlying routines emphasize generality.
  • Limitations and extensions: Future DAP extensions may add model-derived stellar-population and gas-phase properties and models that capture spatial information across datacubes.Examples include stellar age and metallicity, ionization and gas-phase metallicity, dynamical models, and spatially dependent forward models.

A. SPECTRAL-RESOLUTION MATCHING

The resolution-matching procedure convolves spectra with Gaussian kernels to match template and data resolution in pixel space. It handles mismatched resolutions through explicit masking or offsets.

  • Resolution matching: The DAP convolves template flux density with a wavelength-dependent Gaussian kernel to match the spectral resolution of DRP-produced MaNGA spectra.The convolution is normalized over the observed spectral range, especially near its edges.
  • Resolution matching: For Gaussian line-spread functions, the algorithm derives kernel parameters from the relative spectral resolutions of the galaxy and template spectra.It defines resolution using R = λ/∆λ and converts wavelength-space widths into pixel units.
  • Mismatched resolutions: When existing resolution is better than the target, the algorithm can trim the spectral range, apply a constant offset, or allow wavelength-dependent differences.The code supports the first two approaches; standard practice is to trim the spectral range, while the third is unsupported.
  • Mismatched resolutions: The algorithm masks pixels whose required convolution width falls below the tolerance when using the trimming approach.This prevents the convolution from altering the spectral resolution of those pixels.
  • Mismatched resolutions: Constant offsets are retained in km s−1, allowing later adjustment across templates with different resolutions or for kinematic purposes.The offset is set to zero when the calculated squared offset is nonpositive.

B. PROPAGATION OF SPECTRAL-RESOLUTION ERRORS IN THE σ∗ERROR BUDGET

The paper derives how instrumental-resolution uncertainties propagate into observed stellar-dispersion errors and evaluates how higher-resolution templates reduce that influence.

  • Error propagation: The error analysis models the observed stellar velocity dispersion using Gaussian instrumental line-spread functions and stellar line-of-sight velocity distributions.The goal is to quantify and minimize the influence of uncertainties in galaxy and template instrumental dispersions.
  • Error propagation: Φ measures the ratio of fractional error in σobs to fractional error in σinst as a function of template-to-galaxy resolution ratio at fixed σ∗.The analysis assumes Gaussian errors and nominal error propagation.
  • Template resolution: Higher-resolution templates always reduce the influence of instrumental-resolution errors on σobs, with larger relative gains when σ∗ approaches or falls below the galaxy instrumental resolution.The reduction in Φ increases as σ∗/σg decreases.
  • Template resolution: Improvements in Φ can exceed a factor of 2 when σ∗∼σg through use of higher-resolution templates.This result concerns the influence of instrumental-resolution errors on the observed dispersion error budget.
  • Template resolution: A modest 16% resolution difference between MaNGA and MILES gives Φξ=1/Φ = 1.43 when σ∗=σg, while gains diminish for ξ above roughly 2–3.Here ξ is the ratio σg/σt.

C. MANGA DAP DATA MODEL

The MaNGA DAP data model organizes outputs into MAPS, model LOGCUBE, and DAPall files, with documented extensions, measurements, configuration metadata, and quality masks. The model includes stellar, emission-line, and spectral-index products together with flags describing coverage, reliability, fitting, and data-quality conditions.

  • Files and organization: The appendix presents the DAP data model through a series of tables covering file extensions, quantities, masks, and catalog data.The primary output files are MAPS and model LOGCUBE, while DAPall summarizes DAP parameters and global quantities.
  • Quality masks: MAPS and model LOGCUBE files include bitmask extensions using MANGA_DAPPIXMASK and MANGA_DAPSPECMASK, while MANGA_DAPQUAL records global DAP quality.Bit values, names, and descriptions are provided in the corresponding mask tables.
  • Emission-line measurements: Emission-line products provide summed or Gaussian fluxes, equivalent widths, velocities, velocity dispersions, instrumental dispersion, and template dispersion.The products include inverse-variance and data-quality-mask extensions for the principal measurements.
  • Emission-line measurements: Gaussian emission-line velocities are tied across lines, and several line velocity dispersions share tied parameters rather than representing independent measurements.The listed velocity inverse variances should likewise not be combined as independent measurements.
  • Spectral indices: Spectral-index outputs include measurements, inverse variances, data-quality masks, and corrections for velocity dispersion and Doppler broadening.The spectral-index extensions are identified as SPECINDEX, SPECINDEX_IVAR, SPECINDEX_MASK, and SPECINDEX_CORR.
  • Quality masks: DAP quality flags identify conditions such as absent or low coverage, dead fibers, foreground stars, ignored or unreliable values, mathematical errors, failed fits, boundary proximity, and multiple components.DONOTUSE consolidates several problematic spaxel flags, including coverage, fiber, foreground-star, ignored-value, mathematical-error, fit-failure, and near-boundary conditions.
Loading 1901.00856v2…