A SAMI and MaNGA view on the stellar kinematics of galaxies on the star-forming main sequence
A. Fraser-McKelvie, L. Cortese, J. van de Sande, J. J. Bryant, B. Catinella, M. Colless, S. M. Croom, B. Groves, A. M. Medling, N. Scott, S. M. Sweet, J. Bland-Hawthorn, M. Goodwin, J. Lawrence, N. Lorente, M. S. Owers, S. N. Richards
MMNRAS , 000–000 (0000) Preprint 1 March 2021 Compiled using MNRAS L A TEX style file v3.0
A SAMI and MaNGA view on the stellar kinematics ofgalaxies on the star-forming main sequence
A. Fraser-McKelvie , (cid:63) , L. Cortese , , J. van de Sande , , J. J. Bryant , , , B. Catinella , , M. Colless , ,S. M. Croom , , B. Groves , , A. M. Medling , , N. Scott , , S. M. Sweet , , J. Bland-Hawthorn ,M. Goodwin , J. Lawrence , N. Lorente , M. S. Owers , , S. N. Richards . International Centre for Radio Astronomy Research, The University of Western Australia, 35 Stirling Hwy, 6009 Crawley, WA, Australia ARC Centre of Excellence for All Sky Astrophysics in 3 Dimensions (ASTRO 3D) Sydney Institute for Astronomy (SIfA), School of Physics, The University of Sydney, NSW 2006, Australia Australian Astronomical Optics, AAO-USydney, School of Physics, University of Sydney, NSW 2006, Australia Research School of Astronomy and Astrophysics, Australian National University, Canberra, ACT 2611 Ritter Astrophysical Research Center, University of Toledo, Toledo, OH 43606, USA School of Mathematics and Physics, University of Queensland, Brisbane, QLD 4072, Australia Australian Astronomical Optics – Macquarie, 105 Delhi Rd, North Ryde, NSW 2113, Australia AAO-MQ, Faculty of Science & Engineering, Macquarie University. 105 Delhi Rd, North Ryde, NSW 2113, Australia Department of Physics and Astronomy, Macquarie University, NSW 2109, Australia Astronomy, Astrophysics and Astrophotonics Research Centre, Macquarie University, Sydney, NSW 2109, Australia SOFIA Science Center, USRA, NASA Ames Research Center, Building N232, M/S 232-12, P.O. Box 1, Moffett Field, CA 94035-0001, USA
ABSTRACT
Galaxy internal structure growth has long been accused of inhibiting star formation in disc galaxies. We investigatethe potential physical connection between the growth of dispersion-supported stellar structures (e.g. classical bulges)and the position of galaxies on the star-forming main sequence at z ∼
0. Combining the might of the SAMI andMaNGA galaxy surveys, we measure the λ Re spin parameter for 3289 galaxies over 9 . < log M (cid:63) [M (cid:12) ] <
12. Atall stellar masses, galaxies at the locus of the main sequence possess λ Re values indicative of intrinsically flatteneddiscs. However, above log M (cid:63) [M (cid:12) ] ∼ . λ Re remains roughly constant and only at very high stellar masses(log M (cid:63) [M (cid:12) ] > λ Re once galaxies decline in star formation activity. If this trendis confirmed, it would be indicative of different quenching mechanisms acting on high- and low-mass galaxies. Theresults suggest that while a population of galaxies possessing some dispersion-supported structure is already presenton the star-forming main sequence, further growth would be required after the galaxy has quenched to match thekinematic properties observed in passive galaxies at z ∼ Key words: galaxies: evolution – galaxies: general – galaxies: bulges – galaxies: kinematics and dynamics
Galaxy physical appearance (or morphology) and star for-mation rate (SFR) are two of the most common propertiesused to classify galaxies. There is some linkage between thetwo such that frequently, we see that passive galaxies possesslarge galactic bulges, whereas star-forming galaxies are morediscy in appearance (e.g. Strateva et al. 2001; Driver et al. (cid:63) [email protected] M (cid:63) . This correlation means thatstar-forming galaxies are confined to a narrow sequence (withscatter of order ∼ . M (cid:63) ) plane, dubbedthe star-forming main sequence (SFMS; Noeske et al. 2007).This fundamental scaling relation covers several dex in stel- © a r X i v : . [ a s t r o - ph . GA ] F e b A. Fraser-McKelvie et al. lar mass and describes a (mostly) linear increase in log(SFR)with log( M (cid:63) ). This relation was in place early (Schreiber et al.2015; Leslie et al. 2020), and while the sequence is tight, thephysics of what drives the scatter in the SFMS (especially athigh stellar masses) is of great interest.Many recent works find that the SFMS relation is not lin-ear across the entire range of stellar masses mapped by extra-galactic surveys. Instead, it bends such that high-mass galax-ies (log M (cid:63) (M (cid:12) ) (cid:38) . z = 0) possess lower SFRs thanprojected for their mass based on an extrapolation of the re-lation for lower-mass galaxies (e.g. Noeske et al. 2007; Baueret al. 2013; Whitaker et al. 2014, 2015; Schreiber et al. 2015;Tomczak et al. 2016; Leslie et al. 2020). The reason for thisdecrease in SFR at high stellar masses is unknown, but atlow redshifts is thought to be due to a combination of theeffects of stellar mass, morphology, and environment (Erfani-anfar et al. 2016). Indeed, various works have studied the linkbetween main sequence bending and secular processes suchas gas depletion due to environmental effects (e.g. Gavazziet al. 2015), AGN feedback (e.g. Mancuso et al. 2016; Bren-nan et al. 2017), halo quenching (e.g. Popesso et al. 2019), ordisc rejuvenation (e.g. Mancini et al. 2019).The growth of a component that increases the stellar massof a galaxy but not its SFR could also cause the observed de-crease in galaxy specific SFR (sSFR) at high stellar masses.For this reason, bulges have also been proposed as a morpho-logical driver of SFMS bending (Wuyts et al. 2011; Abramsonet al. 2014; Lang et al. 2014; Whitaker et al. 2015; Erfanianfaret al. 2016). The growth of a dispersion-dominated bulge hasalso been linked to the cessation of star formation in a galaxyvia a morphological quenching pathway (Martig et al. 2009).In this manner, a disc may be stabilized against further frag-mentation through the growth of a central mass concentra-tion. However, this paradigm does not explain observations ofbulge-dominated galaxies residing in the highly star-formingregion of the SFR vs. M (cid:63) diagram (e.g. Wuyts et al. 2011;Morselli et al. 2017; Popesso et al. 2019).Bulges can form and grow via multiple pathways, includingmergers (e.g. Hopkins et al. 2010), or in a secular manner (e.g.Pfenniger & Norman 1990). Stellar bars are known to playan important role in bulge formation by driving gas into thecentral regions of galaxies (e.g. Quillen et al. 1995), resultingin starbursts (e.g. Spinoso et al. 2017), and contributing tocentral mass concentration growth (e.g. Wang et al. 2012).Given bars are disc phenomena, we expect the bulges formedby their influence to be rotation-supported by nature (e.g.Bittner et al. 2020).But just how can a bulge grow in an actively star-forminggalaxy without also quenching the galaxy? ‘Compaction’ de-scribes the growth of a bulge through the movement of galax-ies around the main sequence plane through both internal andexternal processes (e.g. Zolotov et al. 2015; Tacchella et al.2016). Star-forming galaxies may propagate upwards to beabove the SFMS line when an episode of gas infall is trig-gered (be that by mergers, counter-rotating streams, or vio-lent disc instabilities). During this episode, gas is funnelledto the central regions of a galaxy, where it is used up in aburst of star formation (e.g. Ellison et al. 2018), fuelling thegrowth of central regions to a saturation point. After thisstarburst ceases, a galaxy will drop down onto the SFMS (orbelow) as the gas-depleted galaxy waits to become replen-ished again. The complex interplay between depletion and replenishment times determines the position of a galaxy onthe SFR vs. M (cid:63) diagram today. In this manner, a galaxy willbuild up its bulge (and become more compact) through suc-cessive compaction events, whilst remaining on the SFMS.Importantly, the process of compaction sets no constraintson bulge kinematics.One of the results of compaction should be a population ofbulge-dominated galaxies that lie above the SFMS. (Morselliet al. 2017; Popesso et al. 2019). Some studies however, donot find this, (e.g. Cook et al. 2020), and rather attribute thebulge-dominated starbursting galaxies to poor bulge-disc de-compositions, often complicated by mergers and interactions.These same works that suggest bulge growth as the cause ofSFMS bending also report that this process is not sufficientto produce the amount of bending seen at high stellar masses(Popesso et al. 2019). Indeed, main sequence bending has alsobeen seen in populations of visually classified pure disc galax-ies (Guo et al. 2015). These studies suggest that a decreasein the SF activity of the disc is also required, and various en-vironmental mechanisms including virial shock heating (e.g.Birnboim & Dekel 2003; Kereˇs et al. 2005) or gravitationalinfall heating (e.g. Dekel & Birnboim 2008; Khochfar & Os-triker 2008) have been proposed to provide this additionalstar formation quenching.Whatever the cause of the SFMS bending, we do knowthat the scatter in the SFMS likely reflects a real diversityin star formation histories (Abramson et al. 2014; Matthee& Schaye 2019). Extending on this idea, we might also ex-pect a variety of stellar kinematics, indicative of a variety ofgalaxy formation pathways. Previous work has shown a linkbetween Hubble type and the spin parameter λ Re (Corteseet al. 2016; Falc´on-Barroso et al. 2019; Wang et al. 2020), V /σ (van de Sande et al. 2018), and specific angular momen-tum (Cortese et al. 2016), such that later-type spiral galaxiesare more rotationally-supported than earlier-type spirals andS0s. Wang et al. (2020) extended on this idea by examiningthe link between galaxy visual morphology and position onthe SFMS. A picture emerged in which galaxies lying on theSFMS were predominantly spirals with small bulges, whilebelow the SFMS, galaxy kinematics depended on stellar mass(though it should be noted that our own Milky Way violatesthis picture with a small bulge, but low SFR for its stel-lar mass e.g. Licquia & Newman 2015). Wang et al. (2020)reported a strong mass dependence below the SFMS suchthat low-mass galaxies were ‘fast rotator’ early-type galax-ies, while high-mass ( M (cid:63) > × M (cid:12) ) galaxies were ‘slowrotator’ spheroids.A dichotomy at z = 0 between star-forming, disc-dominated galaxies and passive, bulge-dominated galaxies isapparent. What is unclear however, is the order of these pro-cesses. Can a bulge form in a star-forming galaxy (and doesit have a role in the quenching of star formation), or is bulgebuild-up the realm of passive galaxies?In this paper, we investigate kinematic trends across theSFMS with IFS data, comparing galaxy spin parametersboth on and off the SFMS. For this sort of analysis, wewill benefit from the number statistics that the two largestIFS surveys to date can provide, and so we combine datafrom both the Sydney-AAO Multi-object Integral field spec-trograph (SAMI; Croom et al. 2012) galaxy survey and theMapping Nearby Objects at APO (MaNGA; Bundy et al.2015) galaxy survey. Given that the target selection of these MNRAS , 000–000 (0000) tellar kinematics of main sequence galaxies two surveys differ, we are able to probe more of the galaxy pa-rameter space, and compare whether or not trends seen in onedata set persist between the two. To enable the best compar-ison possible, we measure kinematic properties between thetwo surveys using a homogeneous set of structural parame-ters, SFR measurements and stellar mass indicators.This paper is organised as follows: in Section 2 we describethe SAMI and MaNGA IFS surveys, along with the homo-geneous structural parameters used to calculate kinematicmeasurements. We also describe the IFS sample, kinematicmeasurement and corrections, along with the definition of themain sequence line used. In Section 3 we present the results,and in Section 4 we discuss the implications of our findings.Throughout this paper we employ a ΛCDM cosmology, withΩ m = 0 .
3, Ω λ = 0 . H = 70 km s − Mpc − and a Chabrier(2003) IMF. The SAMI galaxy survey is an IFS survey on the Anglo-Australian Telescope (AAO) that observed 3068 galaxiesfrom 2013–2018 (Croom et al. 2012). SAMI uses 13 fusedfibre hexabundles (Bland-Hawthorn et al. 2011; Bryant et al.2014) with a high (75%) fill factor. Each bundle contains 61fibres of 1 . (cid:48)(cid:48) diameter resulting in each integral field unit(IFU) having a diameter of 15 (cid:48)(cid:48) . The IFUs, as well as 26 skyfibres, are plugged into pre-drilled plates using magnetic con-nectors. SAMI fibres are fed to the double-beam AAOmegaspectrograph (Sharp et al. 2015), which allows a range of dif-ferent resolutions and wavelength ranges. The SAMI Galaxysurvey employs the 570V grating at 3750–5750 ˚A giving aresolution of R=1810 ( σ = 70 . − ) at 4800 ˚A, andthe 1000R grating from 6300–7400 ˚A giving a resolution ofR=4260 ( σ = 29 . − ) at 6850 ˚A (Scott et al. 2018).83% of galaxies in the SAMI target catalogue have coverageout to 1 R e (Bryant et al. 2015).The SAMI survey is comprised of a sample drawn from theGAMA equatorial regions (Bryant et al. 2015), and an addi-tional sample of eight clusters (Owers et al. 2017). SAMI DataRelease 3 (DR3; Croom et al. 2021) contains observations of3068 galaxies and is the final data release of the SAMI survey.SAMI DR3 includes observations spanning 0 . < z < . . < log M (cid:63) [M (cid:12) ] < .
89 (corresponding to an r -bandmagnitude range of 18 . < m r < . . (cid:48)(cid:48) spaxel − , and the average seeing FWHM is ∼ (cid:48)(cid:48) .Here, we employ the two-moment Gaussian line of sight ve-locity distribution (LOSVD) stellar kinematic maps (van deSande et al. 2017b), including rotational velocity, and veloc-ity dispersion ( σ ) maps. We use the adaptively binned maps,in which spaxels are binned to a signal-to-noise (S/N) of 10using the Voronoi binning code of Cappellari & Copin (2003).The S/N is calculated from the flux and variance spectra ofeach spaxel as the median across the entire blue wavelengthrange (Scott et al. 2018), and spaxels with S/N >
10 are notbinned.
The MaNGA Galaxy Survey is an IFS survey that ob-served > ∼ σ ∼
70 km s − ). MaNGA’s tar-get galaxies were chosen to include a wide range of galaxymasses and colours, over the redshift range 0 . < z < . ∼ . R e for ∼
66% of the total sample, and the remainder (dubbed theSecondary sample) are observed out to ∼ . R e , generally athigher redshifts than the Primary+ sample. SDSS-IV datarelease 15 (DR15; Aguado et al. 2019) contains 4621 uniquegalaxies, selected in the range 7 . < log M (cid:63) [M (cid:12) ] < .
1, (cor-responding to 18 . < m r < . . (cid:48)(cid:48) spaxel − , and the average seeing conditions throughoutthe survey were such that the r -band PSF FWHM is ∼ . (cid:48)(cid:48) .We employ the two-moment LOSVD stellar velocity anddispersion maps using the Voronoi binning scheme to ensureeach bin reaches a target S/N of 10. We also apply the ve-locity dispersion correction provided to account for MaNGAinstrumental dispersion (see Westfall et al. 2019). In this analysis we wish to compare trends in the spin pa-rameter λ Re with current star formation activity in galax-ies. Given there may be observational biases that are un-accounted for between the two surveys, we report trends inSAMI and MaNGA data separately. However, to determinea robust star-forming main sequence line, we wish to be ableto place the two surveys on a homogeneous SFR- M (cid:63) plane.For this reason, we match both SAMI DR3 and MaNGADR15 to the GALEX -Sloan-
WISE legacy catalogue 2(GSWLC-2; Salim et al. 2016, 2018) using a sky match withmaximum separation of 2 (cid:48)(cid:48) . GSWLC-2 provides UV–optical–mid-infrared (IR) SED-derived stellar masses and SFRs for659,229 galaxies within the SDSS footprint and z < .
3, withphotometry provided by
GALEX , SDSS, and the Wide-FieldSurvey Explorer (
WISE ). We utilise the GSWLC-X2 cat-alogue, which uses the deepest
GALEX photometry avail-able (selected from the shallow ‘all-sky’, medium-deep, anddeep catalogues) for a source in the SED fit. SED fittingwas performed using the Code Investigating GALaxy Emis-sion (CIGALE; Noll et al. 2009; Boquien et al. 2019), whichconstrains SED fits with IR luminosity, which they termSED+LIR fitting.3901 MaNGA galaxies have matches to the GSWLC-2, and1832 SAMI galaxies. Unfortunately many of the galaxies lostbelong to the SAMI cluster sample, though we note that fourclusters have GSWLC-2 coverage.
MNRAS , 000–000 (0000)
A. Fraser-McKelvie et al.
To enable a comparison between SAMI and MaNGA kine-matic quantities, we require the structural parameters usedto define the apertures to be identical. Indeed, a small changein aperture size can result in an appreciable difference in λ Re values for a given galaxy. For this reason, we match both sur-veys to the NASA-Sloan Atlas (NSA; Blanton et al. 2011),and use the elliptical Petrosian values for effective radius( R e ), axis ratio ( b/a , which we use to define the ellipticity, (cid:15) ,as (cid:15) = 1 − b/a ), photometric galaxy position angle ( φ ), andthe S´ersic index (n) from a single S´ersic fit. As MaNGA’s tar-geting catalogue was the NSA, all galaxies have these valuesavailable. 1831 SAMI galaxies with GSWLC-2 data also havecounterparts in the NSA. We define the SFMS line for the SAMI and MaNGA galaxiesused in this work by fitting a curve to the points at which thenumber density is highest in the SFR vs. M (cid:63) diagram in binsof stellar mass over the mass range 9 < log M (cid:63) [M (cid:12) ] < . ∼ ∼ , (cid:104) SF R [M (cid:12) yr − ] (cid:105) ) = S − a t − log (cid:32) M (cid:48) t M (cid:33) , (1) M (cid:48) t = M − a t, (2)where M is (cid:104) log( M (cid:63) / M (cid:12) ) (cid:105) , M (cid:48) t is the turnover mass, and t is the age of the Universe in Gyr. For star-forming galaxies,Leslie et al. (2020) find S = 2 . +0 . − . , M = 11 . +0 . − . , a =0 . +0 . − . , and a = 0 . +0 . − . . We use these values, along with t = 13 . SF R ) > . × log( M (cid:63) ) − .
21) in the GSWLC-2 using scipy’s optimize.curvefit package. The best fit main sequence line for the SAMI andMaNGA galaxies is:log(
SF R [M (cid:12) yr − ]) = 0 . − log (cid:18) . M (cid:19) , (3)where M is as defined above.We note that Equation 3 deviates at high stellar massestowards slightly higher SFRs compared to the Leslie et al.(2020) curve, as shown in Figure 1; similar SFMS behaviouris also reported in Thorne et al. (2020). The Leslie et al.(2020) SFMS relation is derived from 3GHz radio continuumimaging of the COSMOS field, and is extrapolated below z ∼ .
3. The high-mass objects observed at low-redhift in thiswork are not present in the Leslie et al. (2020) sample, and log ( M [M ]) l o g ( S F R [ M y r ]) Max frequency SFR in mass binCurved SFMS fitLinear SFMS fitLeslie et al. (2020, z = 0) 0.000.250.500.751.001.251.501.752.00 l o g n g a l Figure 1.
A 2D histogram of the number of galaxies (n gal ) in binsof stellar mass and star formation rate for the combined SAMI andMaNGA sample. Contours of the the overall distribution of thesample are overlaid in white. The SFMS line is fit to GSWLC-2galaxies that match the overall redshift distribution of SAMI andMaNGA, and magenta markers denote the peak of the SFR distri-bution for that mass bin. The SFMS line of Equation 3 is shown ingreen and the linear SFMS line fit of Equation 4 is shown in blue.For comparison, the SFMS fit of Leslie et al. (2020) extrapolatedto z = 0 is shown in red. we speculate this is the reason behind the discrepancies atthe high-mass end.As a comparison, we also fit a linear main sequence line toinvestigate any biases introduced by the assumption that themain sequence bends at high stellar masses. We use only low-mass galaxies in the linear main sequence fit, (where there isno obvious deviation from a straight line, see Figure 1) withinthe mass range 9 . < log M (cid:63) [M (cid:12) ] < .
0. The best fit straightline to the GSWLC-2 galaxies scaled to match the redshiftdistribution of the combined SAMI and MaNGA sample is:log(
SF R [M (cid:12) yr − ]) = (0 . × M) − . , (4)where M is as defined above. The results of this paper usinga linear SFMS line are presented in Appendix A.We define the quantity ∆ MS as the difference in SFR fromthe prediction of the SFMS curve of Equation 3 for a galaxyof the same mass. λ Re measurement Following Emsellem et al. (2007) and Emsellem et al. (2011),we define the spin parameter approximation, λ Re , as theflux-weighted ratio of ordered to disordered motion withina galaxy: λ R = (cid:104) R | V |(cid:105)(cid:104) R √ V + σ (cid:105) = (cid:80) N spx i =0 F i R i | V i | (cid:80) N spx i =0 F i R i (cid:112) V i + σ i , (5)where F is the flux, V the stellar rotational velocity, and σ thestellar velocity dispersion of the ith spaxel. In the same man-ner as Cortese et al. (2016) and van de Sande et al. (2017b),we define R as the semi-major axis of an ellipse on whichspaxel i lies. We chose to use the intrinsic radius rather than MNRAS000
SF R [M (cid:12) yr − ]) = (0 . × M) − . , (4)where M is as defined above. The results of this paper usinga linear SFMS line are presented in Appendix A.We define the quantity ∆ MS as the difference in SFR fromthe prediction of the SFMS curve of Equation 3 for a galaxyof the same mass. λ Re measurement Following Emsellem et al. (2007) and Emsellem et al. (2011),we define the spin parameter approximation, λ Re , as theflux-weighted ratio of ordered to disordered motion withina galaxy: λ R = (cid:104) R | V |(cid:105)(cid:104) R √ V + σ (cid:105) = (cid:80) N spx i =0 F i R i | V i | (cid:80) N spx i =0 F i R i (cid:112) V i + σ i , (5)where F is the flux, V the stellar rotational velocity, and σ thestellar velocity dispersion of the ith spaxel. In the same man-ner as Cortese et al. (2016) and van de Sande et al. (2017b),we define R as the semi-major axis of an ellipse on whichspaxel i lies. We chose to use the intrinsic radius rather than MNRAS000 , 000–000 (0000) tellar kinematics of main sequence galaxies N o . o f g a l a x i e s All SAMI & MaNGA w/ SFRs9.5 < log M < 10.0 1.0 0.5 0.0 0.5 1.00255075100125150175200 10.0 < log M < 10.51.0 0.5 0.0 0.5 1.0020406080100120140160 N o . o f g a l a x i e s MS [dex] N o . o f g a l a x i e s log M > 11.5 Figure 2.
The effects of the kinematic sample selection cuts onthe parent sample of SAMI and MaNGA galaxies with GSWLC-2SFRs. Each panel is a histogram of the distribution of the distanceof a galaxy to the main sequence line (∆ MS) in five mass bins,with the parent sample in grey, and the final sample of galaxiesused for kinematic analysis after quality cuts were made in colour.The majority of galaxies rejected from the kinematic analysis arelow-mass, and this is mostly due to low continuum S/N. the circular projected radius as it follows the galaxy lightprofile more accurately. We note that while this is the sametechnique used for SAMI galaxies by van de Sande et al.(2017b), the values of (cid:15) , R e , and φ used to define the ellipsewithin which λ Re is calculated are different. The reason forthis difference is that we wish to compare SAMI and MaNGAmeasurements in as close a manner as possible, and henceused the same catalogue (the NSA) for structural measure-ments of galaxies for both surveys. This discrepancy results ina small scatter of order ∼ .
05 in λ Re measurements (thoughimportantly, no offset) between the λ Re values from van deSande et al. (2017b) and those reported in this work.At this point, some cuts were also applied to the SAMIand MaNGA data to ensure only galaxies with reliable kine-matics were included in the λ Re catalogue. In both samples,we removed galaxies with R e less than the HWHM PSF ofthe observation. For MaNGA, we also removed galaxies forwhich more than 20% of spaxels within an ellipse of semi-major axis 1 R e were masked. The masking could be the re-sult of flags introduced in the data reduction process (as theDAP velocity and σ masks were applied to the maps prior toanalysis), or we also masked all spaxels where the corrected σ <
50 km s − , as Westfall et al. (2019) suggests that thisis the lower limit for which dispersion measurements can betrusted when S/N > σ cutsuch that bad spaxels are defined as those with σ error >σ × . − . We keep the same quality cuts as van deSande et al. (2017b), and reject any galaxy with >
25% badspaxels from the following analysis. In addition, we removedany galaxies for which R e is greater than the aperture size( ∼
16% of the sample), to avoid the need for aperture cor-rections, and those that were flagged as having unreliablekinematics in the SAMI DR3 kinematics catalogue.In Figure 2, we show histograms of the combined SAMIand MaNGA parent sample with GSWLC-2 SFRs (grey his-tograms), and the final kinematic sample used in this analysisafter all cuts are made (coloured histograms). Each panel ofFigure 2 represents a mass bin used in this work. Unsurpris-ingly, the greatest number of galaxies are lost from the lowstellar mass bins, mostly due to poor continuum S/N withinthe galaxy. Our final samples are representative and highlycomplete (84%) for log M (cid:63) [M (cid:12) ] >
10, though the complete-ness drops significantly (to 48%) for log M (cid:63) [M (cid:12) ] <
10. Whilewe still cover the entire range of SFRs of interest for ouranalysis, we recommend caution in extrapolating our findingsto the entire low mass population. In summary, 897 SAMIgalaxies and 2392 MaNGA galaxies have reliable λ Re mea-surements. Measurement of the λ Re parameter is influenced by both theFWHM of the PSF of the observation and the galaxy in-clination angle (e.g. Cappellari 2016; Graham et al. 2018),hence we attempt to account for both of these effects. Giventhe difference in average seeing conditions between the SAMI(FWHM ∼ (cid:48)(cid:48) ) and MaNGA (FWHM ∼ . (cid:48)(cid:48) ) surveys, it isessential to apply a seeing correction so that we may facili-tate as close a comparison in kinematic properties as possi-ble. There are several recent examples of seeing correctionsfor IFS data in the literature (e.g. Graham et al. 2018; Chunget al. 2020; Harborne et al. 2020). We decide to apply the see-ing correction of Harborne et al. (2020) , due in part to itsease of application to different IFS survey datasets. Briefly,the corrections of Harborne et al. (2020) take the S´ersic in-dex of the galaxy and FWHM of the IFS observation andprovide a value for λ Re that is corrected for seeing. Giventhe MaNGA PSF is on average ∼ . (cid:48)(cid:48) greater than that ofSAMI, the PSF corrections affect the MaNGA data more.Figure 3 shows the increase in median λ Re after both PSFcorrection (an increase of ∼ .
1, in line with Graham et al.2018) and deprojection corrections are applied as a functionof distance from the main sequence (∆ MS) for SAMI (navyblue lines) and MaNGA (green lines) galaxies. Although thecorrections change the absolute value of the median λ Re , theoverall shape of the curves are preserved, meaning that therelative ordering of the spin parameter values will not changegreatly when kinematic corrections are applied.Figure 4 shows that for the combined SAMI and MaNGA http://github.com/kateharborne/kinematic correctionsMNRAS , 000–000 (0000) A. Fraser-McKelvie et al.
MS [dex] M e d i a n R e SAMI PSF corr + deprojMaNGA PSF corr + deprojSAMI PSF corrMaNGA PSF corrSAMI rawMaNGA raw
Figure 3.
The effect of the PSF and inclination corrections on thespin parameter λ Re . Navy lines indicate SAMI median values, andgreen MaNGA. Dotted lines denote the raw median λ Re values,dashed lines are after the PSF-correction of Harborne et al. (2020),and solid lines are median λ Re values in bins of ∆ MS after PSF-correction and deprojection. MS [dex] P e t r o s i a n b / a SAMI/MaNGA median
Figure 4.
Petrosian axial ratio, b/a , of the combinedSAMI/MaNGA kinematic sample for bins of ∆MS. The shadedregion denotes the 25 th and 75 th percentiles of the distribution.Galaxies above the main sequence are on average more round thanthose on or below. sample, there is a dependence on galaxy axis ratio ( b/a ) with∆ MS such that galaxies above the main sequence are rounderthan those on the main sequence (assuming that b/a indicatesinclination and not intrinsic shape). As pointed out by Wanget al. (2020), without an inclination correction the raw ro-tational stellar velocity and σ values propagate to artificiallylow λ Re values above the main sequence, making these galax-ies appear more dispersion-dominated than they actually are.Many deprojection corrections exist in the literature rang-ing from a simple 1 / √ (cid:15) (e.g. Cortese et al. 2016), to morecomplicated functions (e.g. Falc´on-Barroso et al. 2019). Wechose the correction of Emsellem et al. (2011), as imple- mented by del Moral-Castro et al. (2020): λ deprojR = λ R (cid:112) C − λ R ( C −
1) ; (6) C = sin i (cid:112) − β cos i ; (7)cos i = (cid:115) ( b/a ) − q − q (8)where b/a is the axis ratio of the galaxy, and q is the intrinsicaxial ratio of an edge-on galaxy. As we are interested primar-ily in galaxies on or near the star-forming main sequence, wechoose q = 0 .
2, as used in Cortese et al. (2016) for galaxieswith a clear disc component. The anisotropy parameter, β ,varies slightly with Hubble type, but we use β = 0 .
3, whichis appropriate for disc galaxies (derived from Table B.1 ofKalinova et al. 2017).We note that previous studies have found variation in boththe q and β parameters with galaxy morphology (e.g. Cap-pellari et al. 2007; Chemin 2018). We tested the differencebetween SAMI λ Re, deproj values using fixed q and β andthose where q and β varied with galaxy morphology obtainedfrom the catalogue of Cortese et al. (2016). We found verylittle difference between the two methods, with the maximum∆ λ Re, deproj of ∼ .
03. Importantly, there are no trends inmedian λ Re, deproj with ∆ MS. In addition, van de Sande etal. (
MNRAS, in prep. ), show that there is very little offset(∆ λ Re < . − .
1) in deprojected λ Re measures betweenusing the simplistic assumptions presented above and a moredetailed approach as described by Cappellari et al. (2007).Given the similarity between the fixed and morphology-based deprojection values coupled with the fact that we donot have a homogeneous morphology catalogue for both theSAMI and MaNGA samples, we stick with the assumption of q = 0 . β = 0 . λ Re distribution as a function of ∆ MS remains simi-lar, on average, MaNGA λ Re values are slightly higher(∆ λ Re, PSF corr + deproj ∼ λ out to 1 R e , if MaNGA is sampling slightlymore spaxels per galaxy on average than SAMI, this may re-sult in a slightly higher median λ Re measurement. Finally,another possible reason for the λ Re discrepancy may be theway in which stellar velocity and σ were derived between sur-veys. While SAMI broadens their spectra to that of the tem-plates used for a continuum fit, MaNGA fits at the nativeresolution, then applies a dispersion correction after fittingto account for instrumental dispersion effects. Both of thesemethods produce velocity and σ measurements that conveythe astrophysical Doppler broadening, though it is possiblethat the differing techniques result in slight differences be-tween the resultant derived velocity and σ measurements.There is currently no galaxy that is observed in both SAMIDR3 and MaNGA DR15 releases, but we note that if thischanges in the future (indeed, Law et al. (2020) found 74 MNRAS000
1) in deprojected λ Re measures betweenusing the simplistic assumptions presented above and a moredetailed approach as described by Cappellari et al. (2007).Given the similarity between the fixed and morphology-based deprojection values coupled with the fact that we donot have a homogeneous morphology catalogue for both theSAMI and MaNGA samples, we stick with the assumption of q = 0 . β = 0 . λ Re distribution as a function of ∆ MS remains simi-lar, on average, MaNGA λ Re values are slightly higher(∆ λ Re, PSF corr + deproj ∼ λ out to 1 R e , if MaNGA is sampling slightlymore spaxels per galaxy on average than SAMI, this may re-sult in a slightly higher median λ Re measurement. Finally,another possible reason for the λ Re discrepancy may be theway in which stellar velocity and σ were derived between sur-veys. While SAMI broadens their spectra to that of the tem-plates used for a continuum fit, MaNGA fits at the nativeresolution, then applies a dispersion correction after fittingto account for instrumental dispersion effects. Both of thesemethods produce velocity and σ measurements that conveythe astrophysical Doppler broadening, though it is possiblethat the differing techniques result in slight differences be-tween the resultant derived velocity and σ measurements.There is currently no galaxy that is observed in both SAMIDR3 and MaNGA DR15 releases, but we note that if thischanges in the future (indeed, Law et al. (2020) found 74 MNRAS000 , 000–000 (0000) tellar kinematics of main sequence galaxies log ( M [M ]) l o g ( S F R [ M y r ]) Curved SFMS fitLinear SFMS fit 0.10.20.30.40.50.60.70.80.9 R e , P S F c o rr + d e p r o j Figure 5.
A 2D histogram of median λ Re for bins of stellar massand star formation rate for the combined SAMI and MaNGA sam-ple. The SFMS line of Equation 3 is shown in green, and for com-parison, the linear main sequence line fit to low-mass galaxies ofEquation 4 is shown in blue. galaxies in common between the internal MaNGA ProductLaunch 10 (MPL-10) and SAMI DR2), a detailed analysisinto any discrepancies between velocity and σ measures willbe extremely informative. Additionally, performing the anal-ysis of this work on simulated SAMI and MaNGA kinematicdata will give insight into the origin of any small differencesseen in the kinematics between the two surveys. λ Re on the SFMS After performing the various sample cuts described in Sec-tion 2, the SAMI sample spans 0 . < z < .
11, 9 . < log M (cid:63) [M (cid:12) ] < .
8, 17 . < m r < .
1, and the MaNGAsample 0 . < z < .
15, 9 . < log M (cid:63) [M (cid:12) ] < .
1, 17 . 10 to ∼ . < log M (cid:63) [M (cid:12) ] < λ Re such that the slope of the median λ Re as a function of ∆ MS becomes steeper at higher stellarmasses.Above the main sequence, the SAMI sample does not showany significant change in the median value of the stellarspin parameter. However, this sample does not probe be-yond ∆ MS ∼ ∼ . < log M (cid:63) [M (cid:12) ] < 11, we find marginal evidence for adecrease in λ Re in very strongly star-forming galaxies. Whileintriguing, this decrease is only marginally significant, andgiven the tendency of tidal interactions triggering starbursts,potentially more indicative of disturbances in the stellar ve-locity field than gradual thickening of the disk or build-up ofa dispersion-dominated stellar component.Indeed, if we remove the 27 galaxies that clearly show signsof gravitational interaction in their SDSS optical images fromthe highest two bins of ∆ MS for the MaNGA sample, thedecrease in stellar spin at high ∆ MS reduces somewhat. InFigure 7, solid lines depict the full MaNGA kinematic sam-ple, and dotted lines are the MaNGA sample with obviousinteractions removed from the highest two ∆ MS bins in theright panel. All disturbed SAMI galaxies were already re-moved from the sample when the quality control cuts wereapplied. Figure 7 confirms that especially for stellar masses10 < log M (cid:63) [M (cid:12) ] < 11, the median λ Re value flattens outslightly above the main sequence. We note that we removedonly the most obviously interacting systems whose SDSS im-ages showed extreme warping from tidal interaction. Thereare likely many interacting systems of varying degrees of tidaldisruption still remaining within the MaNGA kinematic sam-ple.Below the main sequence, the picture emerging isslightly different. For galaxies with stellar masses 9 . < log M (cid:63) [M (cid:12) ] < . 5, stellar spin seems to remain roughly con-stant up to ∼ λ Re relation so that withincreasing mass, low stellar spin galaxies become more fre-quent closer to the locus of the main sequence. We firstly discuss trends seen for star-forming galaxies lo-cated on the main sequence. The most striking observation MNRAS , 000–000 (0000) A. Fraser-McKelvie et al. R e , P S F c o rr + d e p r o j SAMI M [M ] < 10 10 < log M [M ] < 10.5 MS [dex] R e , P S F c o rr + d e p r o j M [M ] < 11 MS [dex] 11 < log M [M ] < 11.5 Figure 6. Median λ Re for SAMI galaxies as a function of distance from the SFMS line of Equation 3 (∆ MS) in bins of stellar mass.Each panel highlights a different mass bin, with all other mass bins shown in grey for comparison. Shaded regions denote the 25 th and75 th percentiles for each mass bin. The vertical dashed line denotes the locus of the main sequence. In general, the SFMS is populated bydiscy galaxies, with little change in λ Re within ± from Figures 6 and 7 is that apart from the highest stellarmass bin of log M (cid:63) [M (cid:12) ] > . 5, all galaxies on the main se-quence (∆ MS = ± . 25 dex) possess λ Re values indicative ofdiscy galaxies (0 . < λ Re, PSF corr + deproj < . λ Re spin parameter and the in-trinsic shape of a galaxy (e.g. Foster et al. 2017), we cantherefore infer that galaxies on the main sequence are intrin-sically flattened and axisymmetric discs. Apart from perhapsthe highest mass bins, galaxies on the main sequence are asrotationally-supported and intrinsically flattened as they willget; the main sequence is populated by the disciest galaxies.For the MaNGA galaxies in Figure 7, we see a small massdependence at ∆ MS = 0 such that apart from the highestmass bin (within which dispersion-dominated structures maybe beginning to dominate) there is the trend that on aver-age, higher-mass galaxies have greater values of λ Re than lowmass. Catinella et al. (2006) show that the rotation curves ofhigh-mass galaxies reach their peaks at shorter disc scale-lengths than low-mass dwarfs, making it more likely that theflat region of their rotation curves are contained within 1 R e .Given the λ Re metric effectively normalises galaxy spin bystellar mass (thanks to the σ in the denominator of Equa-tion 5), on average, the peak velocity of a galaxy’s rotation curve should be contained within the 1 R e aperture of high-mass galaxies more frequently than for their low-mass coun-terparts. Hence, the resultant λ Re value will be greater. Thisobservation may explain the mass dependence seen at thelocus of the main sequence. Recent photometric studies of galaxies above the main se-quence report that the bulge-to-total ratio ( B/T ) increasessuch that starbursting galaxies are more bulge-dominatedthan their neighbours on the main sequence (Morselli et al.2017; Popesso et al. 2019). These authors find that starburst-ing galaxies possess highly star-forming central regions, whichfrom concentration measures they infer are resultant from thegrowth of classical bulges. We are able to test this theory froma kinematic standpoint.While we do observe a slight reduction in median λ Re above the main sequence, this reduces when we remove the27 galaxies from the MaNGA sample that are obviously in-teracting according to their SDSS colour images. These in-teracting galaxies will likely be highly dispersion-supported MNRAS000 B/T ) increasessuch that starbursting galaxies are more bulge-dominatedthan their neighbours on the main sequence (Morselli et al.2017; Popesso et al. 2019). These authors find that starburst-ing galaxies possess highly star-forming central regions, whichfrom concentration measures they infer are resultant from thegrowth of classical bulges. We are able to test this theory froma kinematic standpoint.While we do observe a slight reduction in median λ Re above the main sequence, this reduces when we remove the27 galaxies from the MaNGA sample that are obviously in-teracting according to their SDSS colour images. These in-teracting galaxies will likely be highly dispersion-supported MNRAS000 , 000–000 (0000) tellar kinematics of main sequence galaxies R e , P S F c o rr + d e p r o j MaNGA M [M ] < 10.0 10 < log M [M ] < 10.5 R e , P S F c o rr + d e p r o j M [M ] < 11 11 < log M [M ] < 11.5 MS [dex] R e , P S F c o rr + d e p r o j log M [M ] > 11.5 Figure 7. Median λ Re for MaNGA galaxies as a function of distance from the SFMS line of Equation 3 (∆ MS) in bins of stellar mass.Each panel highlights a different mass bin, with all other mass bins shown in grey for comparison. Shaded regions denote the 25 th and 75 th percentiles for each mass bin. Solid lines depict the full MaNGA kinematic sample, and dotted lines with white circles are the MaNGAsample with obvious interactions removed from the highest two ∆ MS bins. The extra high-mass bin for MaNGA galaxies is due to thegreater number of high-mass galaxies in this sample. due to the random motions of stars induced by merger ac-tivity. Of course interactions act to decrease the rotationalsupport of a galaxy whilst often inducing starburst activ-ity, however these motions are not necessarily indicative ofclassical bulge growth. Galaxy interactions may therefore be artificially lowering the median spin parameter value abovethe main sequence.We re-examine trends in the λ Re − ∆ MS relation of Fig-ure 7 above the main sequence once interacting galaxies areremoved. For all but the highest and lowest stellar mass bins,the median λ Re curves flatten somewhat such that they are MNRAS , 000–000 (0000) A. Fraser-McKelvie et al. MS [dex] F r e q u e n c y MS [dex] 10 < log M [M ] < 10.5 MS [dex] MS [dex] F r e q u e n c y 11 < log M [M ] < 11.5 MS [dex] log M [M ] > 11.5 Curved SFMSFull sample Re > 0.6 Re < 0.6 Figure 8. Histograms and their associated kernel density estimate (KDE) plots of the combined SAMI and MaNGA kinematic sample(grey), galaxies with λ Re > . λ Re < . − < ∆ MS < 1. ∆ MS = 0 is shown in black. Atlow stellar masses, there is no deviation away from the main sequence. At higher stellar masses, low λ Re galaxies deviate towards lowersSFRs, while high λ Re galaxies move above the main sequence line. We interpret these trends as evidence that more dispersion-dominatedgalaxies populate the ‘bending’ region of the SFMS. similar both on and above the main sequence, and these val-ues are for that of dynamically cold, discy systems. Our re-sults show that there is very little decrease in λ Re for themajority of non-interacting systems above the main sequence,and from this we imply that we do not see evidence of clas-sical bulge growth in this regime.The finding that dispersion-dominated bulges are not grow-ing above the main sequence for the majority of the galaxypopulation does not preclude a compaction scenario. Rather,it is constrained such that the episode of gas infall must occurin a manner so as not to disrupt the rotationally-supportednature of the inner regions of a galaxy. We speculate that thegas infall event that triggers a central burst of star formationmust be ordered. An investigation into the ordered and ran-dom motions of gas in the central regions of galaxies abovethe main sequence should reveal just how turbulent the gasinfall episode is.We note that the SAMI and MaNGA samples do not probethe extreme starbursting galaxy population. Our results arestatistically significant up to +0 . B/T trends up to +1 dex above. Figure 5 of Morselli et al. (2017)shows that the steepest increase in B/T above the main se- quence occurs between +0 . B/T measurements across a largemass range. From a careful structural decomposition of a rel-atively small sample of galaxies, Cook et al. (2020) founda monotonic decrease in B/T above the main sequence forall but the lowest-mass galaxies in their sample. They at-tributed the observed differences to spurious structural mea-surements stemming from the limited model validation avail-able for large (SDSS-sized) catalogues of bulge-disc decom-positions. Indeed, mergers and interacting galaxies are tra-ditionally very difficult to fit with simple bulge+disc models(e.g. Mezcua et al. 2014). Our work agrees qualitatively withthat of Cook et al. (2020): we do not find evidence of a pop- MNRAS000 MNRAS000 , 000–000 (0000) tellar kinematics of main sequence galaxies ulation of starburst galaxies with systematically higher B/T in the local Universe.A caveat to this work is the spatial resolution of the IFSobservations; it is possible that these galaxies on and abovethe main sequence do not have bulges large enough to be seenin the stellar kinematics. The PSF of SAMI and MaNGAare ∼ (cid:48)(cid:48) and ∼ . (cid:48)(cid:48) respectively, which both correspondto 2.0 kpc at the median redshifts of the kinematic samplesused in this work of z = 0 . 05 and z = 0 . 04. Dispersion-dominated bulges significantly smaller than 2 kpc may bewashed out through beam-smearing effects. While this shouldnot be a problem for higher-mass galaxies, classical bulgeslocated in lower-mass galaxies can indeed possess sub-kpcbulge effective radii (Gadotti 2009).These results can be linked to the structural growth andmorphological transformation within galaxies in the contextof star formation. Given that we see no growth of dispersion-supported structure on the SFMS, and yet passive galaxieshost such structures (especially at high stellar masses), wemay say something about the link between galaxy quenchingand morphological transformation via dispersion-dominatedbulge growth. Our results are consistent with two scenar-ios: the first where initial quenching must take place before morphological transformation, and the second where if thesetwo processes are concurrent, then the timescales differ suchthat morphological transformation occurs more slowly thanquenching (or at least the galaxy moving off the main se-quence; e.g. Cortese et al. 2019). We are not in a positionto say which scenario is occurring, but Croom et al. ( MN-RAS, submitted ) takes a different approach in attempting toexplain the formation of S0 galaxies via a combination ofphotometric concentration measures and kinematic disper-sion parameterisation. In this manner, they find that S0 for-mation can be explained via a simple disk fading model takinginto account progenitor bias. These results may provide cluesabout bulge growth in the wider galaxy population. Figures 6 and 7 show a steepening of the λ Re –∆ MS relationwith increasing stellar mass below the SFMS. The reason be-hind this steepening is unclear: while it seems to be revealingan increase in dispersion-supported structure dominance, itcould also be the result of an upwards scatter in SFRs dueto the inherent unreliability of SFR indicators at low sSFRs.Separating star-forming and passive galaxies becomes in-creasingly difficult at higher stellar masses. As can be seenfrom Figure 1, the clear bi-modality of star-forming and pas-sive populations seen between 10 . < log M (cid:63) [M (cid:12) ] < . λ Re –∆ MS relation may be explained by a por-tion of passive galaxies (with dominant dispersion-supportedstructure) contaminating the λ Re measures below the SFMS.Indeed, the steepening of the λ Re –∆ MS relation with masspractically disappears if we use a linear fit to the main se-quence.If the observed steepening of the λ Re –∆ MS slope is real,then this would suggest that the mechanisms acting on high-and low-mass galaxies as they become more passive are differ-ent: one produces passive galaxies with similar disc structureas when they were on the main sequence, while the othermust dramatically alter the kinematics of a galaxy. The obvi-ous mechanism that will destroy or thicken a disk is mergers.Interestingly, the vast majority of slow rotator galaxies pos-sess high stellar masses (e.g. Emsellem et al. 2007; van deSande et al. 2017a; Graham et al. 2018; van de Sande et al.2020; Wang et al. 2020). It is tempting to speculate that thereason for the λ Re steepening in high-mass galaxies only maybe that either the mergers required to create them only oc-cur in high-mass galaxies, or perhaps the processes of massbuild up as the result of mergers differ with stellar mass (e.g.Robotham et al. 2014). Both of these processes must beginwhile the galaxy is still on the SFMS.One subject that this work does not touch on is the effectof environment on the degree of dispersion support withingalaxies as a function of their sSFR. Hence, an exciting av-enue for follow-up work on this topic is through exploringtrends with centrals vs. satellite galaxies. Many works propose the growth of bulges as the driver ofmain-sequence bending (e.g. Abramson et al. 2014; Popessoet al. 2019). Already we see a hint in Figure 7 that thehighest-mass galaxies (the regime in which we expect thegreatest deviation from a linear main sequence) are moredispersion-dominated. We are in a unique position to test thistheory from a kinematic standpoint by examining whether wesee any differences in the ∆ MS values of high- and low- λ Re galaxies.We split the combined SAMI and MaNGA sample be-tween − . < ∆ MS < . λ Re < . 6) and high( λ Re > . λ Re sub-samples. We note here that the low λ Re sample does not consist solely of dispersion-dominated sys-tems, rather they are simply more dispersion-supported thanthe high λ Re systems. There are also trends present withstellar mass such that higher-mass galaxies are more likelyto possess greater dispersion support. This means that therewill be a greater number of high-mass galaxies in the low- λ Re sample, and lower-mass galaxies in the high- λ Re sample. InFigure 8, we plot the distribution of ∆ MS for low λ Re (redline) and high λ Re (blue line) galaxies as a function of dis-tance from the curved SFMS line defined in Equation 3. Asa comparison, we plot the distribution of the overall samplein grey. The locus of the SFMS is shown by a black dashedline.At low stellar masses we see that the ∆ MS distributionis very similar for all values of λ Re , though the low- λ Re sys-tems begin to deviate above log M (cid:63) [M (cid:12) ] = 10, and at high MNRAS , 000–000 (0000) A. Fraser-McKelvie et al. masses are preferentially located below the SFMS line. Sim-ilarly, above log M (cid:63) [M (cid:12) ] = 11, the high- λ Re systems beginto deviate above the overall ∆ MS distribution. At high-mass, systems with greater dispersion dominance preferen-tially populate regions below the SFMS line (however it isdefined), whilst rotation-dominated systems sit above. Weinterpret these trends as evidence that the ‘bending’ regionof the SFMS is populated by galaxies of greater dispersionsupport – high-mass galaxies with greater dispersion sup-port are more likely to possess lower SFRs than their morerotationally-dominated counterparts.Our findings suggest that dispersion-dominated bulges arealready present in massive galaxies on the main sequence.This is not surprising, given that the existence of visuallyclassified early-type (i.e. possessing a prominent bulge com-ponent) star-forming spirals has been known since the estab-lishment of the Hubble morphological sequence. That said,the growth of a dispersion dominated bulge is not the onlypossible cause of a decrease in λ Re : disc thickening will alsodecrease λ Re . When our results are coupled with photometricwork highlighting the redistribution of stars towards centralregions below the main sequence however (e.g. Morselli et al.2017; Popesso et al. 2019), they are sufficient to expect thatat least some of the λ Re decrease is due to bulge growth.It is very tempting to push the interpretation of our resultsfurther and wonder if they provide direct evidence of a phys-ical link between lower SFRs and the growth of dispersion-dominated structure in high-mass galaxies. The morpholog-ical quenching argument of Martig et al. (2009) suffices inexplaining the lower SFRs seen in high-mass galaxies withgreater dispersion support. These galaxies possess lower SFRsbecause their bulges are large enough that they have begunto stabilise galaxy discs against further star formation. Asimilar explanation was put forward by both Whitaker et al.(2015) and Erfanianfar et al. (2016) to explain the morphol-ogy dependence on the scatter in the main sequence, and aflatter main sequence for galaxies with high S´ersic index re-spectively. It is also possible that the lower sSFR is due tothe growth of a non-star-forming component that adds to thestellar mass of a galaxy without increasing its SFR. In thiscase, the growth of a bulge and the cessation of star forma-tion do not need to be linked. Whatever the cause, we are leftwith an intriguing hint of the role of morphology in regulat-ing a galaxy’s star formation. We can certainly conclude thatthe bending of the SFMS at high stellar masses is coincidentwith a population of galaxies that possess classical bulges. We search for evidence of kinematic transformation in galax-ies on the SFMS by examining the link between galaxy SFRand stellar kinematics from IFS observations. Combining themight of the SAMI and MaNGA IFS galaxy surveys, we cal-culate the spin parameter, λ Re , in a homogenised manner for3289 galaxies. Our main results are:(i) Galaxies on the SFMS possess λ Re values indica-tive of intrinsically flattened discs. There is a small masstrend such that higher-mass galaxies appear to have higher λ Re values than lower-mass galaxies, which we expect is dueto the peak of low-mass galaxy velocity fields being morelikely to occur outside the 1 R e aperture used in this work. For the highest stellar mass bin (log M (cid:63) [M (cid:12) ] > . No decrease in λ Re above the SFMS. Onceinteracting galaxies are removed, λ Re measurements upto +0 . . < log M (cid:63) [M (cid:12) ] < A decrease in λ Re below the SFMS for high-mass ( log M (cid:63) [M (cid:12) ] > ) galaxies. One possibility for thedecrease in median λ Re below the SFMS may be that theSFR indicator is unreliable at low sSFRs, scattering somegreen valley galaxies to higher SFRs than they should be.If the trend is real however, then quenching mechanismsmust differ between high- and low-mass galaxies: low-massgalaxies are quenching without structure growth, while somemechanism is acting to both quench a galaxy and dramati-cally adjust the stellar kinematics at log M (cid:63) [M (cid:12) ] > 11. Thelikely culprit is gravitational interactions.(iv) Evidence for a tantalising phenomenologicalconnection between the bending of the SFMS andan increase in galaxy dispersion support. Lower λ Re galaxies are preferentially located on or below the SFMS linefor log M (cid:63) [M (cid:12) ] > . 5. More rotationally-supported systems( λ Re > . 6) better follow a linear SFMS line. The bendingof the SFMS is primarily due to the fact that lower λ Re galaxies start dominating the galaxy budget of the SFMS athigh stellar masses, which we speculate is evidence for thegrowth of classical bulges.Our results indicate that bulge growth is occurring in high-mass galaxies on and just below the SFMS to some degree.In addition, we see evidence that the growth of a dispersion-dominated bulge is linked to the bending of the SFMS athigh stellar masses. While extremely promising, we note thatfurther investigation is still required to precisely identify thelink between the SFMS bending and an increase in dynamicalpressure support. Despite our observations, bulge growth isminor for the majority of galaxies on the SFMS. Given thatmost extremely massive passive galaxies are slow rotators, wefind that extra bulge growth is still required once a galaxyhas quenched to produce the red and dead S0s observed inthe local Universe today. The SAMI Galaxy Survey is based on observations madeat the Anglo-Australian Telescope. The Sydney-AAO Multi-object Integral field spectrograph (SAMI) was developedjointly by the University of Sydney and the Australian Astro-nomical Observatory. The SAMI input catalogue is based ondata taken from the Sloan Digital Sky Survey, the GAMA MNRAS , 000–000 (0000) tellar kinematics of main sequence galaxies Survey and the VST ATLAS Survey. The SAMI GalaxySurvey is supported by the Australian Research CouncilCentre of Excellence for All Sky Astrophysics in 3 Dimen-sions (ASTRO 3D), through project number CE170100013,the Australian Research Council Centre of Excellence forAll-sky Astrophysics (CAASTRO), through project num-ber CE110001020, and other participating institutions. TheSAMI Galaxy Survey website is http://sami-survey.org/. LCis the recipient of an Australian Research Council FutureFellowship (FT180100066) funded by the Australian Govern-ment. JvdS acknowledges support of an Australian ResearchCouncil Discovery Early Career Research Award (projectnumber DE200100461) funded by the Australian Govern-ment. NS acknowledges support of an Australian ResearchCouncil Discovery Early Career Research Award (projectnumber DE190100375) funded by the Australian Govern-ment and a University of Sydney Postdoctoral Research Fel-lowship. Parts of this research were conducted by the Aus-tralian Research Council Centre of Excellence for All SkyAstrophysics in 3 Dimensions (ASTRO 3D), through projectnumber CE170100013. JJB acknowledges support of an Aus-tralian Research Council Future Fellowship (FT180100231).JBH is supported by an ARC Laureate Fellowship and anARC Federation Fellowship that funded the SAMI proto-type. AMM acknowledges support from the National ScienceFoundation under Grant No. 2009416. M.S.O. acknowledgesthe funding support from the Australian Research Councilthrough a Future Fellowship (FT140100255). DATA AVAILABILITY APPENDIX A: LINEAR MAIN SEQUENCE We present median λ Re in bins of stellar mass as a functionof ∆ MS using the linear definition of the SFMS line fromEquation 4. Figure A1 shows the SAMI results, and A2 arethe MaNGA results.It is worth noting that the increase of λ Re with stellarmass at the locus of the main sequence described in Section 3remains even if ∆ MS is measured from the linear fit to theSFMS. The only difference is the change in behaviour at thehighest stellar mass bins, simply because we no longer havegalaxies at these stellar masses on the SFMS.Interestingly, the trend of a steepening of the ∆ MS– λ Re relation below the main sequence almost entirely disappears(or is at least pushed towards higher distances from the mainsequence) when a linear fit to the main sequence is used. REFERENCES Abramson L. E., Kelson D. D., Dressler A., Poggianti B., GladdersM. D., Oemler Jr. A., Vulcani B., 2014, The AstrophysicalJournal Letters, 785, L36 Aguado D. S., et al., 2019, The Astrophysical Journal SupplementSeries, 240, 23Bamford S. P., et al., 2009, Monthly Notices of the Royal Astro-nomical Society, 393, 1324Bauer A. E., et al., 2013, Monthly Notices of the Royal Astronom-ical Society, 434, 209Birnboim Y., Dekel A., 2003, Monthly Notices of the Royal Astro-nomical Society, 345, 349Bittner A., et al., 2020, arXiv e-prints, 2009, arXiv:2009.01856Bland-Hawthorn J., et al., 2011, Optics Express, 19, 2649Blanton M. R., Kazin E., Muna D., Weaver B. A., Price-WhelanA., 2011, The Astronomical Journal, 142, 31Blanton M. R., et al., 2017, The Astronomical Journal, 154, 28Bluck A. F. L., Mendel J. T., Ellison S. L., Moreno J., Simard L.,Patton D. R., Starkenburg E., 2014, Monthly Notices of theRoyal Astronomical Society, 441, 599Boquien M., Burgarella D., Roehlly Y., Buat V., Ciesla L., CorreD., Inoue A. K., Salas H., 2019, Astronomy and Astrophysics,622, A103Brennan R., et al., 2017, Monthly Notices of the Royal Astronom-ical Society, 465, 619Bryant J. J., Bland-Hawthorn J., Fogarty L. M. R., Lawrence J. S.,Croom S. M., 2014, Monthly Notices of the Royal AstronomicalSociety, 438, 869Bryant J. J., et al., 2015, Monthly Notices of the Royal Astronom-ical Society, 447, 2857Bundy K., et al., 2015, The Astrophysical Journal, 798, 7Cappellari M., 2016, Annual Review of Astronomy and Astro-physics, 54, 597Cappellari M., Copin Y., 2003, Monthly Notices of the Royal As-tronomical Society, 342, 345Cappellari M., et al., 2007, Monthly Notices of the Royal Astro-nomical Society, 379, 418Catinella B., Giovanelli R., Haynes M. P., 2006, The AstrophysicalJournal, 640, 751Chabrier G., 2003, Publications of the Astronomical Society of thePacific, 115, 763Chemin L., 2018, Astronomy and Astrophysics, 618, A121Chung H., Park C., Park Y.-S., 2020, arXiv:2008.04313 [astro-ph]Cook R. H. W., Cortese L., Catinella B., Robotham A., 2020,Monthly Notices of the Royal Astronomical Society, 493, 5596Cortese L., et al., 2016, Monthly Notices of the Royal AstronomicalSociety, 463, 170Cortese L., et al., 2019, Monthly Notices of the Royal AstronomicalSociety, 485, 2656Croom S. M., et al., 2012, Monthly Notices of the Royal Astro-nomical Society, 421, 872Croom S. M., et al., 2021, arXiv e-prints, 2101, arXiv:2101.12224Dekel A., Birnboim Y., 2008, Monthly Notices of the Royal Astro-nomical Society, 383, 119Driver S. P., et al., 2006, Monthly Notices of the Royal Astronom-ical Society, 368, 414Drory N., et al., 2015, The Astronomical Journal, 149, 77Ellison S. L., Sanchez S. F., Ibarra-Medel H., Antonio B., MendelJ. T., Barrera-Ballesteros J., 2018, Monthly Notices of theRoyal Astronomical Society, 474, 2039Emsellem E., et al., 2007, Monthly Notices of the Royal Astronom-ical Society, 379, 401Emsellem E., et al., 2011, Monthly Notices of the Royal Astronom-ical Society, 414, 888Erfanianfar G., et al., 2016, Monthly Notices of the Royal Astro-nomical Society, 455, 2839Falc´on-Barroso J., et al., 2019, Astronomy and Astrophysics, 632,A59Foster C., et al., 2017, Monthly Notices of the Royal AstronomicalSociety, 472, 966Gadotti D. A., 2009, p. 22Gavazzi G., et al., 2015, Astronomy & Astrophysics, VolumeMNRAS , 000–000 (0000) A. Fraser-McKelvie et al. R e , P S F c o rr + d e p r o j SAMI M [M ] < 10 10 < log M [M ] < 10.5 MS [dex] R e , P S F c o rr + d e p r o j M [M ] < 11 MS [dex] 11 < log M [M ] < 11.5 Figure A1. Same as Figure 6, but using the linear SFMS line Equation 4 to calculate ∆ MS. Given there are fewer galaxies above thelinear SFMS line, we are not able to probe as far above the main sequence as with the curved line.580, id.A116, < NUMPAGES > < /NUMPAGES > pp., 580,A116Graham M. T., et al., 2018, Monthly Notices of the Royal Astro-nomical Society, 477, 4711Gunn J. E., et al., 2006, The Astronomical Journal, 131, 2332Guo K., Zheng X. Z., Wang T., Fu H., 2015, The AstrophysicalJournal Letters, 808, L49Harborne K. E., van de Sande J., Cortese L., Power C., RobothamA. S. G., Lagos C. D. P., Croom S., 2020, Monthly Notices ofthe Royal Astronomical SocietyHopkins P. F., et al., 2010, The Astrophysical Journal, 715, 202Kalinova V., et al., 2017, Monthly Notices of the Royal Astronom-ical Society, 469, 2539Kereˇs D., Katz N., Weinberg D. H., Dav´e R., 2005, Monthly No-tices of the Royal Astronomical Society, 363, 2Khochfar S., Ostriker J. P., 2008, The Astrophysical Journal, 680,54Lang P., et al., 2014, The Astrophysical Journal, 788, 11Law D. R., et al., 2015, The Astronomical Journal, 150, 19Law D. R., et al., 2020, arXiv e-prints, 2011, arXiv:2011.04675Lee N., et al., 2015, The Astrophysical Journal, 801, 80Leslie S. K., et al., 2020, The Astrophysical Journal, 899, 58Licquia T. C., Newman J. A., 2015, The Astrophysical Journal,806, 96Mancini C., et al., 2019, Monthly Notices of the Royal Astronom-ical Society, 489, 1265Mancuso C., Lapi A., Shi J., Cai Z.-Y., Gonzalez-Nuevo J., B´ethermin M., Danese L., 2016, The Astrophysical Journal,833, 152Martig M., Bournaud F., Teyssier R., Dekel A., 2009, The Astro-physical Journal, 707, 250Matthee J., Schaye J., 2019, arXiv:1805.05956 [astro-ph10.1093/mnras/stz030Mezcua M., Lobanov A. P., Mediavilla E., Karouzos M., 2014, TheAstrophysical Journal, 784, 16Morselli L., Popesso P., Erfanianfar G., Concas A., 2017, Astron-omy and Astrophysics, 597, A97Noeske K. G., et al., 2007, The Astrophysical Journal Letters, 660,L43Noll S., Burgarella D., Giovannoli E., Buat V., Marcillac D.,Mu˜noz-Mateos J. C., 2009, Astronomy and Astrophysics, 507,1793Oh S., et al., 2020, Monthly Notices of the Royal AstronomicalSociety, 495, 4638Owers M. S., et al., 2017, Monthly Notices of the Royal Astronom-ical Society, 468, 1824Pfenniger D., Norman C., 1990, The Astrophysical Journal, 363,391Popesso P., et al., 2019, Monthly Notices of the Royal AstronomicalSociety, 483, 3213Quillen A. C., Frogel J. A., Kenney J. D. P., Pogge R. W., DepoyD. L., 1995, The Astrophysical Journal, 441, 549Robotham A. S. G., et al., 2014, Monthly Notices of the RoyalAstronomical Society, 444, 3986MNRAS000 Same as Figure 6, but using the linear SFMS line Equation 4 to calculate ∆ MS. Given there are fewer galaxies above thelinear SFMS line, we are not able to probe as far above the main sequence as with the curved line.580, id.A116, < NUMPAGES > < /NUMPAGES > pp., 580,A116Graham M. T., et al., 2018, Monthly Notices of the Royal Astro-nomical Society, 477, 4711Gunn J. E., et al., 2006, The Astronomical Journal, 131, 2332Guo K., Zheng X. Z., Wang T., Fu H., 2015, The AstrophysicalJournal Letters, 808, L49Harborne K. E., van de Sande J., Cortese L., Power C., RobothamA. S. G., Lagos C. D. P., Croom S., 2020, Monthly Notices ofthe Royal Astronomical SocietyHopkins P. F., et al., 2010, The Astrophysical Journal, 715, 202Kalinova V., et al., 2017, Monthly Notices of the Royal Astronom-ical Society, 469, 2539Kereˇs D., Katz N., Weinberg D. H., Dav´e R., 2005, Monthly No-tices of the Royal Astronomical Society, 363, 2Khochfar S., Ostriker J. P., 2008, The Astrophysical Journal, 680,54Lang P., et al., 2014, The Astrophysical Journal, 788, 11Law D. R., et al., 2015, The Astronomical Journal, 150, 19Law D. R., et al., 2020, arXiv e-prints, 2011, arXiv:2011.04675Lee N., et al., 2015, The Astrophysical Journal, 801, 80Leslie S. K., et al., 2020, The Astrophysical Journal, 899, 58Licquia T. C., Newman J. A., 2015, The Astrophysical Journal,806, 96Mancini C., et al., 2019, Monthly Notices of the Royal Astronom-ical Society, 489, 1265Mancuso C., Lapi A., Shi J., Cai Z.-Y., Gonzalez-Nuevo J., B´ethermin M., Danese L., 2016, The Astrophysical Journal,833, 152Martig M., Bournaud F., Teyssier R., Dekel A., 2009, The Astro-physical Journal, 707, 250Matthee J., Schaye J., 2019, arXiv:1805.05956 [astro-ph10.1093/mnras/stz030Mezcua M., Lobanov A. P., Mediavilla E., Karouzos M., 2014, TheAstrophysical Journal, 784, 16Morselli L., Popesso P., Erfanianfar G., Concas A., 2017, Astron-omy and Astrophysics, 597, A97Noeske K. G., et al., 2007, The Astrophysical Journal Letters, 660,L43Noll S., Burgarella D., Giovannoli E., Buat V., Marcillac D.,Mu˜noz-Mateos J. C., 2009, Astronomy and Astrophysics, 507,1793Oh S., et al., 2020, Monthly Notices of the Royal AstronomicalSociety, 495, 4638Owers M. S., et al., 2017, Monthly Notices of the Royal Astronom-ical Society, 468, 1824Pfenniger D., Norman C., 1990, The Astrophysical Journal, 363,391Popesso P., et al., 2019, Monthly Notices of the Royal AstronomicalSociety, 483, 3213Quillen A. C., Frogel J. A., Kenney J. D. P., Pogge R. W., DepoyD. L., 1995, The Astrophysical Journal, 441, 549Robotham A. S. G., et al., 2014, Monthly Notices of the RoyalAstronomical Society, 444, 3986MNRAS000 , 000–000 (0000) tellar kinematics of main sequence galaxies R e , P S F c o rr + d e p r o j MaNGA M [M ] < 10.0 10 < log M [M ] < 10.5 R e , P S F c o rr + d e p r o j M [M ] < 11 11 < log M [M ] < 11.5 MS [dex] R e , P S F c o rr + d e p r o j log M [M ] > 11.5 Figure A2. Same as Figure 7, but using the linear main sequence line of Equation 4 to calculate ∆ MS. Given there are fewer galaxiesabove the linear SFMS line, we are not able to probe as far above the main sequence as with the curved line.Salim S., et al., 2016, The Astrophysical Journal Supplement Se-ries, 227, 2Salim S., Boquien M., Lee J. C., 2018, The Astrophysical Journal,859, 11Schreiber C., et al., 2015, Astronomy and Astrophysics, 575, A74Scott N., et al., 2018, Monthly Notices of the Royal AstronomicalSociety, 481, 2299Sharp R., et al., 2015, Monthly Notices of the Royal AstronomicalSociety, 446, 1551 Smee S. A., et al., 2013, The Astronomical Journal, 146, 32Speagle J. S., Steinhardt C. L., Capak P. L., Silverman J. D., 2014,The Astrophysical Journal Supplement Series, 214, 15Spinoso D., Bonoli S., Dotti M., Mayer L., Madau P., BellovaryJ., 2017, Monthly Notices of the Royal Astronomical Society,465, 3729Strateva I., et al., 2001, The Astronomical Journal, 122, 1861Tacchella S., Dekel A., Carollo C. M., Ceverino D., DeGraf C., Lap-iner S., Mandelker N., Primack Joel R., 2016, Monthly NoticesMNRAS , 000–000 (0000) A. Fraser-McKelvie et al. of the Royal Astronomical Society, 457, 2790Thorne J. E., et al., 2020, arXiv e-prints, 2011, arXiv:2011.13605Tomczak A. R., et al., 2016, The Astrophysical Journal, 817, 118Wake D. A., et al., 2017, The Astronomical Journal, 154, 86Wang J., et al., 2012, Monthly Notices of the Royal AstronomicalSociety, 423, 3486Wang B., Cappellari M., Peng Y., Graham M., 2020,arXiv:2002.09011 [astro-ph 10.1093/mnras/staa1325Westfall K. B., et al., 2019, The Astronomical Journal, 158, 231Whitaker K. E., et al., 2014, The Astrophysical Journal, 795, 104Whitaker K. E., et al., 2015, The Astrophysical Journal Letters,811, L12Wuyts S., et al., 2011, The Astrophysical Journal, 742, 96Yan R., et al., 2016a, The Astronomical Journal, 151, 8Yan R., et al., 2016b, The Astronomical Journal, 152, 197Zolotov A., et al., 2015, Monthly Notices of the Royal AstronomicalSociety, 450, 2327del Moral-Castro I., et al., 2020, Astronomy and Astrophysics, 639,L9van de Sande J., et al., 2017a, Monthly Notices of the Royal As-tronomical Society, 472, 1272van de Sande J., et al., 2017b, The Astrophysical Journal, 835, 104van de Sande J., et al., 2018, Nature Astronomy, 2, 483van de Sande J., et al., 2020, arXiv e-prints, 2011, arXiv:2011.08199MNRAS000