Extended Tully- Fisher relations using HI stacking Scott Meyer, Martin Meyer, Danail Obreschkow, Lister Staveley-Smith
Summary Intro TFR Where traditional methods fail HI Stacking Calibrations and simulated data Analytical galaxies Simulated galaxies (from the S 3 -SAX model) Real data test case (HIPASS and HICAT) HIPASS direct detections Extension work to 6df / SKA and precursors
Tully-Fisher relation (TFR) What is the TFR? The TFR is an empirical relation between absolute magnitude and intrinsic HI emission line width (as a tracer of rotation velocity) that spiral galaxies follow. What s happening here? This particular plot demonstrates a novel way of getting TFR parameters without having galaxy inclinations (See Obreschkow & Meyer 013 for more info) Figure 4. Absolute K-band (left) and B-band (right) magnitudes, plotted against v (red big dots, i > TFRs with Obreschkow 1-σ scatter. The & red Meyer, dashed ApJ, lines show 013, the TFRs Vol 777, obtained pg by140 applying the MLE using in to the red big dots. In turn, the blue solid lines are obtained without inclination data, by applying the inclinations, Extended i.e., ρ(i) Tully-Fisher = sin i.inother relations words, using these HI fits stacking arebased - Scott purely Meyer onthescattered 3 bluesmalld (A color version of this figure is available in the online journal.)
Current limitations Optical vs Radio rotation measures The TFR can be inves<gated in op<cal or radio. Op<cal probes deep, but does not necessarily retrieve the full velocity profile. Radio probes much more of the velocity profile but is limited to z < 0.1. Wavelength Observed Quanty Max Redshi8 Op<cal Hα Z ~ 1-1.5 low Radius Probed Radio HΙ z < 0.1 High NGC 6946 lez: Op<cal, right: Radio 4
HI stacking Fabello et. al. 011 1 What? This technique seeks to creates an HI emission spectrum with increased improved signal-to-noise by combining several HI spectra (usually in an attempt to get a detection from a collection of nondetections). How? Rest-frame HI spectra from galaxies are created using optical redshift (or radio redshifts if available). These are then co-added. Signal increases proportional to N (the number of galaxies included) while noise increases as N. Movie at https://vimeo.com/690671 5
Simulated galaxies 1 Analytically solvable galaxy profiles A single galaxy was modeled as a disk with constant linear velocity (not constant rotation). This galaxy was projected to various different inclinations and convolved with Gaussians of varying σ/v max values to simulate different random gas velocities (dispersion). Random projections of this galaxy were stacked to investigate the relation between the FWHM of the stack, and the FWHM of the deprojected galaxy. 6
Simulated galaxies Correction factor In order to reproduce the TFR, our measured stack widths must relate in some predictable way to the deprojected widths of the input galaxies. W stack 50 = width of stacked profile W50 ref = width of single analytical spectra, deprojected and dedispersed Blue = high dispersion (random gas velocities) Red = low dispersion 7
Simulated galaxies 3 Normalised flux density 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0. 0.1 Galaxies with differential rotation Galaxies were given an exponential rotation curve of the form: V (r) =V max 1 e r/r flat Since galaxies no longer have constant velocity profiles, we prescribe a HI surface density of the form: σ/v max = 1/10 Constant rotation, edge-on Constant rotation, stack Differential rotation, edge-on Differential rotation, stack HI (r) = M HI Fθ = W part 50 /W 50 ref 1.15 1.1 1.05 1 0.95 rhi e r/r HI σ/v max = 1/5 σ/v max = 1/10 σ/v max = 1/0 0-1 -0.5 0 0.5 1 Normalised velocity v = V/V max 0.9 0 10 0 30 40 50 60 70 80 90 Minimal inclination of stack i min [deg] 8
Simulated galaxies 4 Low number statistic corrections Because galaxies enter the stack without being corrected for inclination, having a sample too small introduces an offset and scatter into W stack. 50 9
Simulated galaxies 5 S 3 -SAX simulation HI and H semianalytical model made by post-processing the Millennium simulation. Obreschkow 009 Figure 9. Simulated sky field covering 3 1arcmin. The three panels correspond to three different redshift slices, with an identical comoving depth of 40 Mpc. Figure 8. Comparison of a simulated normalized emission line Ψ( )(solid = 10
Simulated galaxies 6 S 3 -SAX Tully-Fisher relation 1. Million galaxies Blue diamonds = White diamonds = W50 stack before correction Black circles = averages from individual galaxies W50 stack K band absolute magnitude -19-0 -1 - -3-4 -5 W stack,c 50 W stack 50 ref W 50 40 50 60 80 100 00 V rot [km s 1 ] 10 3.6 10 3 10.4 10 1.8 10 1. 10 0.6 10 0 11
Simulated galaxies 7 Simulated HIPASS (b) Vctotal Vchalo Vcdisk 500 h-1mpc (a) -4 - Obreschkow et al. stack 50 stack,c W50 Wind 50-0 The Astrophysical Journal, 766:137 (11pp), 013 April 1 K band absolute magnitude -6 5 independent regions of the S3-SAX box were selected and HIPASS was reproduced. TFR reproduced extremely well using HI stacking technique. W HI -18 Vcbulge 5 0 10 15 radius [kpc] 500 40 h-1 Mp c (c) -1Mpc 500 h 50 60 80 100 00 300 1 [km s ] Obreschkow 009 Figure 3. The five colors represent our five virtual HIPASS volumes fitted inside the cubic box of the Millennium simulation. The box obeys periodic boundary conditions, such that the truncated volumes (blue and purple) are in fact simply connected. The five HIPASS volumes do not overlap and can hence be used to estimate the magnitude of cosmic variance. of turbulent/thermal dispersion. The linewidth W50 at the 50%level of the peak flux density is then measured from the 1 1
Real data (HIPASS & HICAT) HIPASS Galaxy selection was kept to a minimum: Galaxies to be stacked were only cut based on angle from CMB equator. Red galaxies were cut based on angle from CMB equator as well as a 45 degree cut. Blue = Red = HIPASS W stack 50 galaxies White = without corrections W stack 50 B band absolute magnitude -16-18 -0 - W stack 50 W stack,c 50 HIPASS i > 45 deg K band absolute magnitude -18-0 - -4-6 W stack 50 W stack,c 50 HIPASS i > 45 deg 0 30 50 70 100 00 300 [km s 1 ] 0 30 50 70 100 00 300 [km s 1 ] 13
Extension work Non-detections Work is on-going in the HIPASS region by stacking 6df galaxies (non-detections in radio). SKA With significantly deeper data sets such as DINGO, WALLABY and LADUMA, stacking can be used to squeeze more science out of these instruments. Other uses for stacking So far stacking has only been used to recover total HI flux (and now average rotation velocities), how far can we take this technique? Optical and Radio survey overlap This work relies heavily on overlapping optical (or IR) and radio surveys. 14
Summary Calibrations and simulated data Analytical galaxies Simulated galaxies (from the S 3 -SAX model) Real data test case (HIPASS and HICAT) HIPASS direct detections Extension work Non-detections SKA Other stacking work 15
Extra plots HIPASS simula<on Millennium 16
Extra plots W50ref = deprojected but WITH dispersion Millennium data WITH ellip<cal galaxies K band absolute magnitude -19-0 -1 - -3-4 -5 W stack,c 50 W stack,c,e 50 40 50 60 80 100 00 V rot [km s 1 ] 10 3.6 10 3 10.4 10 1.8 10 1. 10 0.6 10 0 17