Skip to main content
  • UC Berkeley
  • College of Letters & Science
Berkeley University of California

URAP

Project Descriptions
Spring 2026

URAP Home Project Listings Application Contact

Cosmology with Roman space telescope spectra of Type Ia Supernovae

Saul Perlmutter, Professor  
Physics  

Applications for Spring 2026 are closed for this project.

It has been thought that the accelerated expansion of our universe is caused by an unknown "dark energy" which has a constant energy density. However, there are currently hints that dark energy is actually evolving over time. If this were true, there would be implications on fundamental physics and potentially even the ultimate fate of the universe - will it expand forever, or eventually collapse?

Precise measurements of supernova distances are required to distinguish between constant and evolving dark energy models. The Nancy Grace Roman space telescope will take tens of thousands of light curves and spectra of Type Ia supernovae (SN Ia). We have developed machine learning (ML) based methods to extract distances from spectra of SNe Ia with half the errorbars compared to traditional light-curve analyses. In order to prepare for the analysis of Roman SN spectra, we simulate SN spectra based on the instrument specifications (noise, resolution, wavelength ranges, ...) and want to find out which method delivers the best cosmological results under these circumstances.

Role: The student will:
1. Learn about the basics and current status of cosmology
2. Install and run one of the ML codes we developed
3. Understand (optional: re-derive) how we simulate Roman SN spectra, and adapt that to the code they are running
4. Perform distance measurements with the simulated SN spectra
5. Do whatever you are interested in! This is an open-ended project, it can evolve into something else the student is interested in after these points are completed.

Skills learned:
- Homogeneous cosmology, especially distances
- Understanding how experimental real-world constraints impact actual results
- Ability to work with large python codes
- Running ML code on an HPC cluster

Qualifications: Python programming
(Optional: Git/Linux/slurm. Familiarity with astronomy, cosmology, machine learning is nice-to-have, but not necessary - you will learn it!)

We are interested in students who might want to work with us for more than one semester - perhaps you will analyze real data with the methods you developed yourself in a few years!

Day-to-day supervisor for this project: Jannik Truong, Graduate Student

Hours: to be negotiated

Off-Campus Research Site: remote

 Engineering, Design & Technologies   Mathematical and Physical Sciences

Return to Project List

Office of Undergraduate Interdisciplinary Studies, Undergraduate Division
College of Letters & Science, University of California, Berkeley
Accessibility   Nondiscrimination   Privacy Policy