FOCUS GROUP

For our local galaxy population we have used the same SDSS galaxy images predented in earlier chapters (see Section 3.4). We have used the HST WFC3/IR data from the CANDELS and 3D- HST multi-cycle treasury programme (Grogin et al. 2011; Koekemoer et al. 2011) centred on the COSMOS field (Scoville et al. 2007; Skelton et al. 2014). According to the position of the centroids for each galaxy provided by the 3D-HST catalogues, we take cut-outs from theJ(F125W) andH(F160W) filter images. Corresponding weight maps and segmentation maps are also cut using the same centroids and sizes. For the SDSS sample, bar galaxies (previously identified in Gadotti (2008)) were removed from the final sample. For the COSMOS sample we have visually removed galaxies with obvious bars but as the bar fraction decreases with redshift (Sheth et al., 2008) the problem of finding a bared galaxy is rare.

5.4.1

Redshift range and morphologicalk-correction

Galaxies have very different structures depending on what wavelength range they are observed at (Bohlin et al. 1991; Kuchinski et al. 2001; Windhorst et al. 2002; Papovich et al. 2003; Conselice 2004). This shift in the qualitative and quantitative appearance of a galaxy constitutes as a morpho- logical k-correction. A primary problem in understanding the evolution of galaxy structures over cosmic time is constraining the effects of the morphological k-correction.

Most deep high resolution imaging is done in the observed-frame optical wavelength range, probing up toλ_∼1µm, and allowing for a sampling of the rest-frame optical (>4000Å) wavelength range only up to a redshift ofz∼ 1.5 (Conselice 2004). Above this redshift, the rest-frame ultra- violet begins to be observed where predominantly young stars with ages<100 Myrs are sampled. With that in mind, the largest differences would be observed in galaxies composed of young and old stars which are not spatially mixed, such as early type spiral galaxies (see Windhorst et al. 2002).

Morphological k-corrections depend on both the galaxy type and redshift. In Huertas-Company et al. (2009) the effects of this morphological k-correction at 1 < z < 2 was quantified by com- paring morphologies in the K and I-bands in the COSMOS field. Their classification was based on

5.4. Sample selection

a machine learning approach quantifying the galaxy morphology on non-parametric values. They found that I-band classifications are less able to identify early-type galaxies than the K-band, thus underestimating the elliptical population.

With the above information in mind, it is clear that selecting the appropriate redshift range to probe the same rest-frame wavelength range within different filters at higher redshifts is fundamen- tal to avoid any misconceptions in interpreting our results due to morphology k-correction. Thus, to properly compare our local sample (0.04<z<0.06) in the SDSS i-band to the two HST WFC3/IR band filters we use images in the redshift ranges 0.5<z<0.85 and 0.85<z<1.27 for theJ and

H bands respectively. The redshift ranges are defined when the effective wavelength of the i-band is traced past the minimum and maximum wavelengths of both theJ andH band.

5.4.2

Selecting the progenitors of local galaxies

Evolutionary processes, make linking a galaxy in the local Universe with its progenitors difficult. When inferring the evolution of individual galaxies and linking them with progenitors without being strongly affected by progenitor bias is a key issue many previous studies have devoted time in solving (Carollo et al. 2013; Sonnenfeld et al. 2014; Shankar et al. 2015). It has been shown that the stellar mass function (SMF) has evolved significantly sincez∼3₋4 (Pérez-González et al. 2008; Muzzin et al. 2013; Ilbert et al. 2013) so a sample selection based on a fixed stellar mass can be affected by galaxies entering in at lower redshifts. Another method to link progenitors or descendants is by linking galaxy populations at a constant luminosity (Wake et al. 2006) or by isolating specific galaxy populations such as bright cluster galaxies (Lidman et al. 2012; Lin et al. 2013; Shankar et al. 2015). A popularised progenitor linking method in recent years is the selection using a fixed number density (van Dokkum et al. 2010; Bezanson et al. 2011; Conselice et al. 2013; Patel et al. 2013; Mundy et al. 2015; Torrey et al. 2016). The underlying assumption is for example, massive galaxies at higher redshifts will evolve into massive galaxies in the local Universe. As galaxy masses and other properties are thought to evolve significantly over time, the number density is thought to stay reasonably staticas long as the mass rank order among a galaxy population is preserved(Torrey et al. 2016).

We wanted to confirm the likely mass evolution of a Milky-Way like galaxy (as well as a M_∗ =

1010Mgalaxy) with a toy model consisting of a bulge and a disc component. Using a set of stellar population models we randomly traced a galaxy of mass logM_∗/M =10.66 back to a redshift of

Figure 5.2: The mass evolution of galaxy populations is shown tracked from a redshiftz=2 toz=0. The coloured bands are mass estimates from stellar population models combined according to a bulge-to-total ratio so that the result is matches a galaxy with logM_∗/M=10 (left panel) and Milky- Way sized (logM_∗/M=10.66; right panel) galaxy atz_∼0. The stellar populations are randomised for each component of the galaxy from: a single stellar burst model, an exponentially declining SFH and a constant SFH. The dashed line is the mass evolution function for a logM_∗/M =10.66 Milky-Way type galaxy from van Dokkum et al. (2013). The dashed-dotted lines are the mass evolution functions for galaxies with mass logM_∗/M = 10.27, 11.2 from Ilbert et al (2013) and Patel et al (2013) respectively. The vertical dotted lines indicate the redshift ranges obtained from the morphological k-correction analysis described in the above text.

5.4. Sample selection

& Charlot 2003) of a single burst, and exponentially declining and a constant SFH. We also altered the bulge-to-total mass ratio for each toy model. Figure 5.2 shows the stellar mass evolution of a logM_∗/M = 10 and logM_∗/M = 10.66 galaxy out toz =2 comparing it to the mass evolution function from a fixed number density approach from van Dokkum et al. (2010). We also varied the formation time of the systems between the redshiftsz∈ {2 : 8}. Albeit with a significant scatter at

z ¦1, the mass-redshift function of van Dokkum et al. (2010) matches our toy model. We want

to stress here that these toy models encompass our ignorance in how nearby galaxies can be traced back through time.

However, there are two problems with the comoving number density analysis (and subsequently our toy models): 1) galaxy mergers will change the total number density of galaxies (Ownsworth et al. 2014) and 2) stochastic growth rates, this makes a galaxy’s evolution more of a random process. Torrey et al. (2016) studied the number density evolution of galaxy populations was tracked through the cosmological hydrodynamical simulation Illustris (Vogelsberger et al. 2014; Genel et al. 2014; Nelson et al. 2015). This tracked number density evolution incorporated both the impact of scattered growth rates and galaxy mergers. In a following paper, Wellons & Torrey (2017) then capture how properties of galaxies within the simulations evolved over time and showed that the constant number density method performs poorly in recovering the evolution of galaxy properties. Wellons & Torrey (2017) suggest that their probabilistic number density method works best for tracing galaxy progenitors. Comparing Figure 5.2 to the mass evolution presented in Wellons & Torrey (2017), we find that we over estimate the evolution over time, as expected, due to the random selection of the SFHs for the galaxies.

5.4.3

Final sample

Figure 5.3 shows the final redshift and stellar mass bins for our COSMOS sample. We use the toy models as a rough indicator of how galaxies might evolve. Our local Universe sample in SDSS has a minimum mass limit of logM_∗/M=10, and a peak in the mass distribution logM_∗/M∼10.6, which is conveniently close to the mass of the Milky-way. This motivated us to split the local sample in two mass bins, so that we have two populations of galaxies, with either less than the mass of the Milky-Way or greater than. We then use the toy models to determine a minimum mass in both redshift ranges defined by evolving a galaxy with M_∗ = MMW and a galaxy with M∗ = 1010M

(minimum mass in SDSS sample) back to define their likely progenitor masses (see Fig. 5.2). The blue boxes in Fig. 5.3 represent the samples that will always evolve to masses less than the mass of the Milky-Way and vice verse for the red boxes. The green boxes in Fig. 5.3 are then the sample of

Figure 5.3: Stellar mass as a function of photometric redshift for the galaxies in the COSMOS field described in the catalogues of the 3D-HST project. The boxes indicate our cuts in mass and redshift according to the morphological k-correction of the SDSS i-band filter to the CANDELS WFC3/IR filters and the mass evolutions from the toy models. The blue boxes show galaxy masses that will evolve to become galaxies with masses logM_∗/M < 10.66 at z ∼ 0, and vice versa for the red boxes i.e. logM_∗/M>10.66. The green boxes show galaxies that have the potential to evolve to become logM_∗/M =10.66 or Milky-Way like masses. Mass ranges are all estimated from the toy models explained in the above text. For reference, the dashed line is the mass evolution function for a logM_∗/M=10.66 Milky-Way type galaxy from van Dokkum et al. (2013).