Graduate Student Highlights

By Carter Rhea (Chair, CASCA Graduate Student Committee)
(Cassiopeia – Winter / hivers 2020)

Each month, the GSC highlights the work of an outstanding Canadian graduate student by sharing their work with our members. Since the launch in February of 2020, we have highlighted several students from around the country. In this issue, we share the highlights of new students since the last issue of Cassiopeia.

Follow us on Twitter, Instagram, and Facebook under the handle casca_gsc.

Mainak Singha — University of Manitoba

Mainak’s research investigates how weakly accreting ‘Active Galactic Nuclei’ (AGN) can drive galaxy evolution processes. Most successful galaxy evolution models require the AGN to launch galactic scale outflows to drive the galaxy evolution processes. In order to trace the signs of outflows, he uses spectroscopic data (spectra) from SDSS (Sloan Digital Sky Survey). The emission lines from these spectra pin-point the evidences of ionization caused by the photons from the AGN accretion disks or the shocks from the AGN. Any asymmetry in the emission line profiles indicates the gas moving towards / moving away from us which are the signatures of outflows.

Figure 1


Figure 1 is a standard BPT diagram from SDSS DR7. The radio galaxy J142041+025930 lies in the LINER (Low Ionization Nuclear Emission Line Region) region suggesting it to be a Low Excitation radio galaxy (LERG).

Vivian Tan — York University

Vivian’s research is on the galaxies that reside within massive clusters at redshifts 0.25 < z < 0.6, in the Hubble Frontier Fields. Clusters are dynamic environments where galaxies interact and quench, which means transitioning from star-forming to quiescent. Quenching processes alter a galaxy’s morphology, which we want to measure not just with their light profiles but through their stellar mass distribution. Mapping where the stellar mass is in a galaxy is usually difficult at z > 0, but the Frontier Fields have deep multiband Hubble photometry. This means resolved stellar mass maps are possible even for galaxies as small as 108 solar masses. Galaxies with such low stellar masses have not been studied in a resolved way at z > 0. Because we can analyze morphology with resolved stellar mass maps, we found that quiescent galaxies which are less massive than 109.5 solar masses are more likely to be disk-dominated (Sersic index ~ 1 to 2), but quiescent galaxies are bulge-dominated above that mass limit (Sersic index of 4 or more). This was only found in clusters but not in the less dense “field” environments. This means different quenching processes must have occurred to transform these galaxies, and these quenching processes depend both on the galaxy’s mass and their environment.

Figure 2


Figure 2 shows the process of creating the resolved stellar mass maps through a process called SED-fitting. The galaxy is broken up into spatial bins, and a SED is fitted to photometric flux from multiple bands in each of the bins. The fitted SED can reveal what the stellar mass of that region of the galaxy is and putting it all together results in a resolved stellar mass map. Sersic index measurements for the stellar mass are obtained via parametrically fitting a 2-D Sersic profile directly to the map of stellar mass using GALFIT.

Jessica Campbell — University of Toronto

Jessica’s research focuses on the multiphase nature of our Galaxy’s magnetic field and how it connects between different phases of the interstellar medium (ISM). Whether it is the turbulent warm ionized medium (WIM) that fills much of the Galaxy or the cold neutral medium (CNM) often found in sheets and filaments, this complex ISM is permeated with high energy cosmic rays and magnetic fields. When accelerated by the magnetic field, these cosmic rays emit radio synchrotron radiation that is strongly linearly polarized. As this polarized emission passes through the foreground ISM, thermal electrons and magnetic fields in the WIM rotate the plane of polarization, an effect called Faraday rotation. These cosmic rays can also penetrate and ionize the densest regions of the ISM, causing even the predominantly neutral medium to be coupled to the magnetic field via linear 21 cm HI structures called ‘HI fibers.’ Despite the wealth of magnetic field information about the WIM and CNM, very little is known about how they relate to one another. Do the diffuse ionized and cold clumpy media share a common magnetic field? If so, how often does this occur, and under what circumstances? These are the questions driving Jessica’s research.

Figure 3


Figure 3 shows Planck dust emission at 353 GHz, where the coloured image is the total (unpolarized) intensity and the textured lines indicate the magnetic field orientation. The dust emission clearly contains the same knee and fork morphologies, and the overall field orientation is roughly parallel to the polarized filaments F1 and F3.

Robert Bickley — University of Victoria

Robert’s research focuses on the intersection between observational astronomy and machine learning, specifically, using machine vision techniques to identify galaxies that have recently undergone a merger with another galaxy. Mergers often leave behind a distinct visual signature, giving rise to unusual morphologies and leaving behind displaced streams of stars. To identify mergers using machine vision, he trains Convolutional Neural Networks (CNNs) on examples of mergers and non-mergers taken from a simulation (IllustrisTNG) and modified to look like real observations. He can then use the simulation data to identify where the CNNs are successful, and where they struggle.

Figure 4


Figure 4 shows how well a CNN identifies mergers and non-mergers as a function of the environment. If a galaxy has a neighbor very close by, it will have a small r_1 value. If there are no nearby neighbors, r_1 will be very large. The top panel shows the total number of post-mergers and controls (blue and orange histograms, respectively), further broken down as correctly and incorrectly classified (fp, brown: controls classified as post- mergers; tn, purple: correctly-classified controls; fn, red: post-mergers classified as controls; tp, green: correctly-classified post-mergers). The bottom panel shows the fraction of post-merger and control galaxy images correctly identified by the model.

The figure demonstrates that the model retains much of its ability to distinguish between mergers and non-mergers with a close neighbor down to 10 kiloparsecs, below which the visual degeneracy becomes prohibitive. However, such close neighbors are rare in both the simulation and the real Universe, and therefore do not present a significant source of contamination.

Bookmark the permalink.

Comments are closed.