At the end of last century measurements of distant exploding stars, or supernovae, showed us that the expansion of the Universe is speeding up. This led to the discovery of dark energy. During my first summer of Laidlaw, I worked on a machine learning algorithm that could be used for Supernovae classification in astrophysical surveys. This is of great importance to determine the properties of dark energy. For greater precision measurements of the parameters that define it, we need to find supernovae at further distances from us. However, we encounter the issue that explosions at greater distances are fainter and therefore harder to detect. To solve this problem, we can make use of gravitational lensing.
Gravitational lensing occurs when light from a distant object passes through a large galaxy. The galaxy’s gravitational pull forces the light to bend, which results in the light source appearing distorted and magnified to the observer. In many cases, this process causes multiple images of the object to appear in the foreground lensing galaxy. The study of gravitationally-lensed supernovae (gLSNe) could answer a number of questions in astrophysics, providing high- precision constraints on cosmological parameters that define the nature of dark energy.
My research during my second summer focused on one main problem. Our calculations say that the Zwicky Transient Facility (ZTF) should be detecting about 3 gLSNe per year. However, in the last 3 years of survey none have been detected. There exist two possible explanations for this: either our calculations are wrong or we are not using the right method to filter and detect them. My work during my LiA consisted of simulating a large set of gLSNe with the ZTF “simsurvey” simulation code and studying their characteristics. We then checked how many of these would be disregarded if we applied the current filters used and looked for possible correlations between their properties.
My final results indicated that, as we expected, the filters on brightness and colour currently set to make our job of finding these Supernovae easier are actually making us ignore a large set of candidates. In the fainter colour bands over 50% of the possible candidates were lost. These results are promising, as they mean that with some changes on our current detection algorithms we will probably be able to find the number of gLSNe our calculations suggest. This would contribute to our search for possible answers to the greatest cosmological questions!
I would like to thank Dr. Ariel Goobar and PhD student Ana Sagués Carracedo, from the University of Stockholm for their supervision during the project, as well as the rest of the gLSNe research group. I would also like to thank my Summer 1 supervisor Prof. Kate Maguire for facilitating this opportunity, as well as the Laidlaw programme for supporting my work.