Institute for Astronomy
Institute for Astronomy
Contact us

MScR Projects

MScR Projects for 2026

Testing Model Dependence of BAO constraints in non-ΛCDM cosmologies

Prof Florian Beutler

This 12-month project will quantify how much standard BAO (Baryon Acoustic Oscillation) analyses depend on the choice of fiducial cosmology by modifying the current BAO pipeline to natively support non-ΛCDM models. We will assess potential biases in inferred distance measures when using non-ΛCDM fiducials, determine when such biases are negligible relative to statistical uncertainties, and produce practical guidance for robust, model-flexible BAO inference.

Motivation and Background:

BAO provides a robust standard ruler, routinely used to infer distances and expansion rates through measurements of:

  • Isotropic: \(D_V(z)/r_d\)

  • Anisotropic: \(D_M(z)/r_d\) and \(H(z),r_d\)

Most pipelines rely on a ΛCDM fiducial to:

  • Convert redshifts and angles to comoving coordinates

  • Perform density-field reconstruction [1]

  • Build anisotropic BAO templates and apply Alcock–Paczynski (AP) scalings

  • Define the sound horizon at drag, \(r_d\)

In ideal circumstances, results are reported as scaling parameters relative to the fiducial:

  • \[\alpha_\perp = \frac{(D_M/r_d)}{(D_M/r_d)_{\rm fid}}\]
  • \[\alpha_\parallel = \frac{(H,r_d)_{\rm fid}}{H,r_d}\]
  • \[\alpha_{\rm iso} \approx (\alpha_\perp^2 \alpha_\parallel)^{1/3}\]

If all steps are perfectly consistent and sufficiently flexible, downstream cosmological fits should be independent of the fiducial. In practice, approximations (e.g., fixed damping scales, polynomial marginalisations, fixed covariances, simplified reconstruction assumptions, and often fixed \(r_d\)) can introduce small but non-negligible model dependence, especially for non-ΛCDM models with different expansion histories or early-time physics (curvature, \(w_0w_a \) dark energy, early dark energy, varying \(N_{\rm eff}\), massive neutrinos, or modified gravity).

Research questions:

  1. How robust are standard BAO measurements to the choice of non-ΛCDM fiducial?

  2. Which components of the pipeline (data coordinates, reconstruction, BAO template, priors on \(r_d\), covariance) drive any residual model dependence?

  3. For what parameter ranges do biases become comparable to current and near-future (DESI/Euclid) statistical errors?

  4. What best practices ensure BAO results are portable and minimally model-dependent?

Objectives:

  • Generalise the BAO pipeline to accept arbitrary late- and early-time cosmologies as the fiducial, including \(\Omega_k \neq 0\), \(w_0w_a\), EDE, varying \(N_{\rm eff}\) and massive neutrinos.

  • Compute \(r_d\) consistently from the chosen cosmology and propagate into templates and AP scalings.

  • Quantify shifts in \(\alpha_\perp\), \(\alpha_\parallel\), \(D_M/r_d\), and \(H,r_d\) across a grid of non-ΛCDM fiducials using mocks and real data.

  • Diagnose which modelling choices (e.g., reconstruction smoothing scale, BAO damping model \(\Sigma_{\rm nl}\), polynomial marginalisation, covariance re-use) most affect robustness.

Given the recent hints for evolving dark energy in the DESI DR1 [2] and DR2 [3] data, the robustness of BAO data is crucial. This work will be conducted within the DESI collaboration, meaning the student will become a member of DESI and will use the DESI BAO fitting pipeline. This will provide insights into how a large Astronomy collaboration works. One (ambitious) goal of this project could be to support the official DESI DR4 BAO analysis, which will be well aligned with the duration of this project. Further reading on this subject could be [4] and [5].

References:

[1] Monthly Notices of the Royal Astronomical Society, 427, 2132-2145 (2012)

[2] Journal of Cosmology and Astroparticle Physics, 2025, 021 (2025)

[3] Arxiv:2503.14738

[4] Monthly Notices of the Royal Astronomical Society, 534, 1 (2024)

[5] Arxiv:2603.03443

Mapping the CGM in the Simba Simulations

Prof Romeel Davé

Gas in the circum-galactic medium (CGM) traces the baryon cycle of galactic inflows and outflows that govern the growth and evolution of galaxies over cosmic time.  Historically it has been difficult to directly image this diffuse gas, but with modern instruments such as VLT/MUSE this is now becoming feasible.  But what are we learning from such observations, and how do we connect them to baryon cycling processes?  Using the Simba galaxy formation simulations, this project will involve generating mock emission-line data cubes of the CGM around galaxies at different masses and redshifts in various tracers such as H\(\alpha\), CIV, and OVI, and then use past & future particle tracking to understand which tracers correspond to which components of the baryon cycle.  This will inform current observations which can barely detect the CGM, but will be particularly important for making predictions for the upcoming 39m E-ELT with which such detections will become routine.  The project involves significant python code development, learning to work with simulation data, exploring CGM science, and assembling a publication for submission to a journal.  

Stellar abundances with machine learning

Prof Sergey Koposov

This research project sits at the intersection of astrophysics and machine learning, focusing on the chemical analysis of the Dark Energy Spectroscopic Instrument (DESI) survey. The primary objective is to test and deploy the Julia-based Korg spectral synthesis code to fit DESI stellar spectra.  Because running traditional spectral synthesis on DESI’s massive catalog of 20 million stars is computationally prohibitive, a key component of this work involves constructing and training neural network emulators based on Korg’s output to accelerate the fitting process. Using this pipeline, you will investigate and establish exactly which elemental abundances can be robustly extracted from the DESI data. The ultimate goal is to develop and run a scalable, AI-driven pipeline capable of processing the entire multi-million star dataset to map stellar populations and chemical evolution.

Minimal Seeds and Subcritical Transitions in Stellar Dynamos

Dr Calum Skene & Prof Steve Tobias

Magnetic fields in planets and stars are sustained by dynamo action driven by rotating convection, yet in some regimes dynamo onset may be subcritical. In this case, the purely hydrodynamic state is linearly stable, and a dynamo is only triggered when the system is seeded with a sufficiently large and suitably structured magnetic perturbation. Recent work on the geodynamo has shown how nonlinear optimisation can be used to numerically identify such triggering perturbations (“minimal seeds”) and to reveal the dynamical pathways that connect non-magnetic and magnetic attracting states [1]. This project builds on those ideas in a stellar context, where stratification is important. It will exploit the recent development of automatic differentiation capabilities in the open-source Dedalus framework to implement the optimisation procedure efficiently, enabling systematic searches for minimal seeds [2]. The overall aim is to map the dynamical landscape of subcritical stellar dynamos and improve our understanding of how self-sustained magnetism can arise and persist in linearly stable regimes.

References:

[1] Skene, C.S., Marcotte, F., Tobias, S.M., ‘On nonlinear transitions, minimal seeds and exact solutions for the geodynamo’, Journal of Fluid Mechanics, 2025

[2] Skene, C.S., Burns, K.J., ‘Fast automated adjoints for spectral PDE solvers’, Proceedings of the National Academy of Sciences, 2026

 

This article was published on