Merging galaxies and Deep Learning


Galaxy mergers are very important for the evolution of galaxies, and the overall appearance of our universe today. Together with Dr. Brian Nord and the Deep Skies Lab, I am working on a deep learning algorithm e.g. convolutional neural network (CNN), that will be able to distinguish between merging and non-merging galaxies. CNNs represent a deep learning technique, vastly used in image or video recognition, computer vision, or even recommendation systems, and natural language processing.

The CNN can be trained on simulated data, and through transfer learning this algorithm can later be applied to observational data. This will be very important for searching through large surveys, like DES or LSST. Finding many interacting galaxies in different merger stages, will be crucial for understanding these interacting objects, their physics and the role of mergers in galaxy evolution.

Read more in:

DeepMerge: Classifying High-redshift Merging Galaxies with Deep Neural Networks

Production of lithium in small irregular galaxies

One interesting product of cosmic-ray interactions with interstellar medium is lithium, which is one of the few elements which was produced during the Big Bang. Until recently measured lithium abundances were only available inside Milky Way, but first measurements outside our galaxy are available for small neighbouring galaxy – Small Magellanic Cloud (SMC), which was also detected in gamma rays. The observed gamma-ray flux of the SMC describes its present day cosmic-ray activity, while its lithium abundances are the result of all cosmic-ray activity in this galaxy during its lifetime.

It turns out that the observed lithium abundances in the SMC are much higher than in the MW halo stars, and are in fact just slightly below the expected primordial abundances. Trying to explain the observed lithium abundances in the SMC is an important issue, since it can shed light on our understanding of lithium abundances in the Universe, which proved to be very problematic until now. In order to link these two observations, we have linked the production of lithium in SMC and the pionic gamma-ray intensity produced by all SMC-like galaxies over the cosmic history (normalised to the present day SMC gamma-ray flux). We have shown that galactic cosmic rays produced in supernova remnants (considered to be the dominant cosmic-ray population in the SMC) can only explain a very small part of the observed abundance of lithium (less than 1%), if we assume that the entire present gamma-ray emissivity of the SMC that we observe also originates from the interaction of galactic cosmic rays with gas within the galaxy.

This leaves room for the possible existence of other interesting sources of lithium in the SMC, like other cosmic-ray populations which were present at some point in SMCs history, for example during close fly–bys of SMC and MW. This CR population would not be present for a long time compared to the life of the galaxy, so they could produce more lithium without violating the present day gamma-ray flux of the SMC. Comparing SMC’s lithium abundances and its gamma-ray flux suggests a possible important role of galaxy–galaxy interactions for production of cosmic rays inside galaxies, both in tidal shocks, as well as through triggering new star formation, and subsequent supernovae explosions.

Read more in:

Galactic Cosmic-Ray Induced Production of Lithium in the Small Magellanic Cloud


Structure-Formation Cosmic Rays

Accretion of new gas onto already virialised structures, for example in galaxy clusters, is a process that could potentially accelerate cosmic rays, since these are environments where both shocks and magnetic fields are present. Since accretion shocks appear and evolve during formation of large scale structures over cosmic time, production of structure-formation cosmic rays should also evolve, and consequently the production of gamma rays by the source cosmic rays. We have implemented, for the first time, the evolution of gamma-ray production, which is a consequence of the evolution of accretion shocks which appear during large scale structure formation. The developed models are more realistic, compared to previous models which use single redshift approximation for the gamma-ray origin. Models are used to derive the gamma-ray flux of all unresolved galaxy clusters, which is than compared to the isotropic diffuse gamma-ray background, measured by telescope Fermi-LAT.

Normalisation of our models was done by using gamma-ray flux upper limit for Coma cluster, which we assume is an average cluster. We find that these cosmic rays can have non-negligible contribution to the isotropic diffuse gamma-ray background. Gamma-ray models are also normalised using high-energy neutrinos detected by IceCube detector, since cosmic rays produce both gamma rays and neutrinos during their propagation through interstellar medium. We found that if the accretion shocks are predominantly strong, neutrino background is more limiting to the possible gamma-ray emissivity of these objects, compared to the gamma-ray background. After using neutrinos to constrain our models, derived uppermost limit on the structure-formation cosmic rays contribution (in case of strong shocks) to the isotropic diffuse gamma-ray background observed by Fermi is then 12−18%, which is similar to values derived by other authors.  If accretion shocks are not strong, than it would be possible to produce more gamma-rays without violating the expected neutrino production.

Cosmic rays accelerated in accretion shocks can have non-negligible contribution to the isotropic diffuse gamma-ray background and that this population of cosmic rays has to be taken into consideration in addition to other components that are thought to be major contributors, like for example, unresolved star–forming galaxies or blazars.

Read more in:

Diffuse Pionic Gamma-Ray Emission from Large-scale Structures in the Fermi Era

Neutrino Constraints to the Diffuse Gamma-Ray Emission from Accretion Shocks

Optical Observations of Nearby Galaxies

The observations are done through narrow band Hα and [SII] filters, and until now were mostly done on the 2m telescope at the National Astronomical Observatory (NAO) Rozhen, Bulgaria. We search for HII regions and supernova remnants. Supernova remnants are identified by their elevated [SII] /Hα > 0.4 ratios compared to HII  regions.

Star formation rate in galaxies is usually derived based on their Hα fluxes, which should correlate with number of HII regions within the galaxy. By using [SII] /Hα criterion we found that supernova remnants can contaminate Hα flux of the observed galaxies to a few percents. By only observing in Hα line we cannot exclude these supernova remnants. Therefore, star formation rate based on just Hα observations are not very precise and it is important to correct them for supernova remnant contamination.

Read more in:

Optical Observations of the Nearby Galaxy IC342 with Narrow Band [SII] and Hα Filters. II – The Detection of 16 Optically–Identified Supernova Remnant Candidates

Optical Observations of the Nearby Galaxy IC342 With Narrow Band [SII] and Hα Filters.I