Reconstructing the Cosmic Web from the Lyman-alpha forest with machine learning

Supervisor: Oliver Hahn

Contact information: oliver.hahn@univie.ac.at 

Co-supervisor: Dr. Florian List

Expected duration: 9 months

Project description & Goals:

Neutral hydrogen clouds leave characteristic absorption features in the spectra of distant quasars and galaxies, caused by the Lyman-alpha transition of hydrogen. This Lyman-alpha "forest" allows probing the large-scale structure of the Universe and, if sufficiently many quasars or galaxies can be measured in a patch of the sky, even enables the reconstruction of 3D maps by combining multiple line-of-sights (see Ref. [1] for an interactive visualisation). Traditionally, a Wiener Filter is used for this purpose, which can be shown to be optimal for Gaussian fields. However, in view of the non-linearity of structure formation, more elaborate methods can extract additional information that is missed by the Wiener Filter (e.g. Ref. [2]). Recently, machine learning methods have successfully been employed for various tasks related to the Lyman-alpha forest, e.g. for estimating the Lyman-alpha flux based on dark matter-only N-body simulations (see Ref. [3]). Regarding the reconstruction of 3D maps from line-of-sight "skewers", machine learning methods can potentially exploit non-Gaussian information and provide uncertainty estimates for the reconstructed maps. The goal of this master thesis is to develop and implement a machine learning-based method for the Lyman-alpha flux reconstruction and to compare the results with those of traditional reconstruction methods using simulated data.

 

Working plan & Milestones (including final thesis):

  1. Literature study and introduction to the Lyman-alpha forest and the methodology
  2. Test 3D tomography reconstruction using Wiener Filter (self-implemented or using code package, e.g. https://github.com/caseywstark/dachshund)
  3. Develop and implement one (or possibly even several) machine learning method for 3D tomography reconstruction (could use Gaussian processes, neural networks, …)
  4. Validate the method using simulated Lyman-alpha data and compare the results with the traditional Wiener Filter-based method.
  5. Write thesis (and, depending on the progress, a short paper on the results

Requirements / special skills: 1) Should have followed lectures on cosmology and/or cosmic large-scale structure, 2) Advanced (Python) programming skills are a bonus 3) Some prior experience with machine learning techniques is a plus.

References:

 

[1] sketchfab.com/3d-models/eboss-lya-line-of-sights-in-the-stripe-82-8850a9776c3944e7b70cf13509542936

[2] arxiv.org/pdf/2102.12306.pdf

[3] arxiv.org/pdf/2106.12662.pdf

[4] arxiv.org/pdf/1903.09049.pdf

[5] arxiv.org/pdf/2009.10673.pdf

[6] arxiv.org/pdf/astro-ph/0105196.pdf

[7] arxiv.org/pdf/2002.10676.pdf

[8] arxiv.org/pdf/2004.01448.pdf