What’s New in DR17

We maintain detailed change logs for all of our software products. Locations of detailed change logs are found at the following locations.

Data Reduction Pipeline (DRP)	RELEASE_NOTES
Data Analysis Pipeline (DAP)	CHANGES.md	ReadTheDocs
Marvin	CHANGELOG	ReadTheDocs

The MaNGA DRPall catalog file has been updated in DR17 to split information about MaNGA data cubes from information about MaStar stellar spectra. Previously these were grouped together in a single extension; in DR17 MaNGA data cubes are listed in Extension 1, and MaStar stellar spectra in Extension 2.

With a spectral resolution R ~ 2000 corresponding to an instrumental width of about 70 km/s (1-sigma), accurate measurement of the MaNGA spectral line-spread function (LSF) is critical for the computation of reliable velocity dispersions. However, estimating the LSF is made significantly more challenging by the near-critical sampling of the detector pixels and the time-variable nature of gravitational flexure within the BOSS spectrographs. As such, substantial effort has been dedicated over the life of the MaNGA survey to progressively refining our understanding of the LSF.

As discussed in detail by Law et al. 2021, in DR17 the pipeline-estimated LSF has been improved to sub-percent accuracy based on comparison against data from spectrographs with substantially higher spectral resolution. At high astrophysical velocity dispersions (i.e., greater than the ~ 70 km/s instrumental line width) the impact of these updated LSF estimates will be minimal. At low astrophysical velocity dispersions however the values recovered by the DAP (which subtracts the instrumental LSF from the measured signal in quadrature) have changed significantly and are now robust down to 20 km/s or below for sufficiently high SNR emission lines. Any analyses of the cold gas velocity dispersion for instance should thus use DR17 data instead of DR15 or earlier results.

The various data models for intermediate and final-stage DRP data products have been changed accordingly to introduce new extensions and/or modify the names of existing LSF-related extensions accordingly (e.g., changing the DISP and PREDISP extension names to the more descriptive LSFPRE and LSFPOST for the pre-pixellized and post-pixellized LSF estimates). See the MaNGA Data Model for further details.

Multiple pipeline changes since DR15 have produced a few-percent change in the photometric calibration of the MaNGA data products. These changes include using the Fitzpatrick-99 extinction curve for standard star calibrations instead of the O'Donnell-94 curve, changes to the pixel reference flatfields, and adjustment of the IFU fiber bundle metrology (i.e., the locations of the individual fibers within the hexagonal IFU ferrules).

As the MaNGA sample size has built up over the course of the survey, it has become possible to self-calibrate the metrology using on-sky observations compared to previous SDSS broadband imaging. This exercise revealed a ~ 2.5% systematic error in the overall scale factor of the original laboratory-based metrology. After correcting for this factor in the MaNGA reference metadata the overall flux normalization of the survey data increased by about 3% due to the changed aperture loss correction factor for the calibration minibundles. See Law et al. 2021 (their Appendix A) for details.

DRP data cubes with significant reduction problems identified by the pipeline are flagged in DR17 and all previous DRs by the CRITICAL data quality bit. For DR17 we have manually inspected each of these cases and (where possible) rejected problematic exposures in order to recover the highest-quality science data. As a result, there are some data cubes that were flagged in previous DRs as poor quality but are good quality in DR17. We have additionally introduced an UNUSUAL data quality bit to indicate data cubes that are different from ordinary MaNGA data quality but still useful for science analysis. This new UNUSUAL category includes cases for which large regions of the data cube are blank (e.g., 127-fiber IFUs missing 1/4 of the field of view due to a problem with one of the V-groove blocks), and cases with severe 'blowtorch' contamination which has been corrected by the pipeline but may have lower SNR and some artifacts in the 6000-6200 Angstrom region.

These 'blowtorch' cases result from a severe electronic artifact on one of the BOSS cameras during the final year of survey operations which required a significant revision to the DRP to satisfactorily process the data. See Law et al. 2021 (their Appendix B) for details.

In total, of the 11,273 data cubes in MPL-11 there are 69 flagged with the UNUSUAL data quality bit and 86 flagged with the CRITICAL data quality bit.

The total number of galaxies in the final MaNGA data release is difficult to define, but is broadly in the vicinity of 10,000. DR17 includes 11,273 calibrated data cubes; many of these data cubes however are part of ancillary programs targeting either regions within large extended galaxies (e.g., IC342 and M31) or non-galaxy targets (e.g., globular clusters, intracluster light in the Coma cluster, etc). If we use the MaNGA drpall summary file to identify only galaxy-type targets (i.e., those with non-zero MANGA_TARGET1 or MANGA_TARGET3 bitmasks) that are not in the Coma, IC342, M31, or globular cluster programs (ancillary target bits 19, 20, 21, 27 respectively):

drpall=mrdfits('drpall-v3_1_1.fits',1)
galaxies=drpall[where(((drpall.mngtarg1 ne 0)or(drpall.mngtarg3 ne 0)) $
    and((drpall.mngtarg3 and 2L^19L+2L^20L+2L^21L+2L^27L) eq 0))]

we are left with 10,296 data cubes. Not all of these have the highest data quality though for a variety of observational reasons. If we restrict ourselves only to those data cubes that do not have either the UNUSUAL or CRITICAL MANGA_DRP3QUAL quality flags set (bits 14 and 30 respectively):

highqual=galaxies[where((galaxies.drp3qual and 2L^14L+2L^30L) eq 0)] 

we are left with 10,145 data cubes. Not all of these are unique galaxies though as some galaxies were observed multiple times. If we use the MANGAID galaxy identifier to weed out only unique galaxies:

mangaid=highqual.mangaid
uniquegals=highqual[uniq(mangaid,sort(mangaid))]

we are left with a sample of 10,010 unique, high-quality galaxy data cubes. Note, however, that there is still some potential confusion that can arise from treatment of merging galaxy pairs; see Caveats for details.

DR17 introduces new sets of template spectra used to fit the continuum during the combined continuum+emission-line fit. Recall that the DAP fits the MaNGA spectra twice, once primarily to fit the stellar kinematics and once to fit the emission lines. In DR15/DR16, the same templates were used for both (MILESHC; see Section 5 from Westfall et al. (2019, AJ, 158, 231)). In DR17, we continue to use the MILESHC templates for the stellar kinematics, but switch to templates based on the MaStar stellar library in the emission-line module. The majority of the DR17 DAP products use a sub-selection of the Simple Stellar Population models from Maraston et al. (2020, MNRAS, 496, 2962). In the DAP, we refer to these MaStar SSP templates as the MASTARSSP library, which are limited to the following set of stellar-population parameters:

• Age [Gyr]: 0.00316228, 0.01, 0.03162278, 0.100, 0.300, 1.000, 3.000, 9.000, 14.000
• log(Z/Z_sun): -1.35, -1., -0.7, -0.33, 0, 0.35
• IMF Slope: 2.35 (Salpeter)

We also performed a clustering analysis of MaStar spectra to construct a set of composite stellar templates, which we refer to as the MASTARHC2 library. These templates are only provided for a single binning type in DR17. See additional discussion here.

Comparisons between the emission-line results based on the MASTARSSP and MASTARHC2 libraries can be used to assess systematic errors due to the chosen template library (cf. Law et al. (2021, in press).

DR17 introduces numerous changes to the DAP data model, both with respect to the names of the core output files and their content. The data model pages for the MAPS and model LOGCUBE files each provide short sections listing the important changes made with respect to DR15/DR16. The changes to the relevant data models accommodate:

• an expanded set of emission lines with measured properties,
• a new approach to our spectral index measurements,
• a better organized set of model spectra,

and many other improvements. See the updated MaNGA DAP Data Model and ReadTheDocs for details.