Gravitational lensing
Gravity is a peculiar phenomena. Gravity is weak, but on large scale it is a wonder-working winner. It distorts and twists anything with energy. Light carries energy, so light is also distorted by gravity’s relentless rule. Much like your glasses (or your friend’s, if you don’t have any yourself!) distort the path of light due to the different refractive indexes of glass and air, so the gravitational field of a galaxy or a cluster of galaxies — or even just a star — distorts the path of light. This is called gravitational lensing.
In a series of papers, Dr. Chris Clarkson has devised a formalism for looking at both weak and strong gravitational lensing under one scope; the Roulette formalism:
[1603.04698] Roulettes: A weak lensing formalism for strong lensing - I. Overview (arxiv.org)
[1904.04471] Recursion relations for gravitational lensing (arxiv.org)
Slowly, but steadily, we are working on implementing this with Machine Learning to see if the whispers of the night sky makes ever so slightly more sense when perceived through gravitational glasses.
Take a look:
[Bachelor thesis]: A study of gravitational lensing through simulation and machine learning
[Bachelor thesis]: Intergalactic machine Vision (to appear)
The group consists of two far-fetched physicists (Kenny Solevåg Hoti & Ben David Normann) and one guru of computing (Hans Georg Schaatun), alongside undergraduate and graduate students.
Looking for a thesis project? If you intend to write your Master’s or bachelor’s thesis on gravitational lensing, you may find the following list interesting.
A comparison between the Roulette formalism and standard approaches: The Roulette formalism is a weak-lensing approach to (weak and) strong lensing. Why is the gravitational potential used in Clarkson’s derivation different from the standard definitions? Can the Roulette formalism solve the mass-sheet degeneracy?
Improving the Roulette formalism: The current Roulette formalism is computed within the thin-lens and weak-field approximations. Do we see interesting effects if we discard of these?
Real data: To make our models useful, we would have to implement with real data. To this end there is an array of more technical issues that calls for attention (databases, noice reduction).
Visualisation tool: There is a lot to do to improve the visualisation tool. The goal here is to create a desktop application that can be run, with all the necessary qualities to effectively show-case gravitational lensing in general and the Roulette formalism in particular.
Model building: We are particularly interested in cluster-lens profiles.
Lens-Mass reconstruction: How do we actually obtain the lens mass from the roulette amplitudes provided by the machine-learning algorithms? Some sort of interpolation method or the like seems like a good guess.
Lensing of gravitational waves.
Why this study? As mentioned in the previous project, we want to understand dark matter. Ad-mass retrieval of information from gravitationally lensed light is an excellent probe on the distribution of dark matter.