Steven G. Murray

Steven G. Murray

Marie Sklodowska-Curie Fellow

Scuola Normale Superiore

Biography

I am a Marie Sklodowska-Curie (MSCA) Fellow at the Scuola Normale Superiore (SNS) in Pisa, Italy. My research interests are in 21cm cosmology, cosmological structure formation, Bayesian statistical analysis, and creating useful open-source software for scientific applications.

Interests
  • 21cm Cosmology
  • Bayesian methods
  • Open-source scientific software and practices
Education
  • PhD in Astrophysics, 2017

    University of Western Australia

  • BSc (Honours) in Physics, 2011

    University of Western Australia

  • BSc in Mathematics, 2009

    University of Queensland

Experience

 
 
 
 
 
Scuola Normale Superiore
Marie Sklodowska-Curie Fellow
Sep 2023 – Present Pisa, Italy
My MSCA fellowship project is called ‘FORWARD’, and will focus on developing Bayesian methods of understanding the full, correlated error budget of 21cm experiments like HERA, EDGES and the SKA. This will focus primarily on developing models of our instruments, and propagating our uncertainties on these models through to parameter inference.
 
 
 
 
 
Arizona State University
Research Scientist
Oct 2018 – Aug 2023 Arizona
At ASU, I worked with the HERA interferometer and the EDGES global signal experiment, as well as leading development of the 21cmFAST simulation package. For HERA, I helped lead the Validation team, designing and running some of the most sophisticated instrumental simulations in the world. For EDGES, I have developed a new robust analysis pipeline to help confirm the potential first-ever detection of Cosmic Dawn.
 
 
 
 
 
Curtin University
Postdoctoral Researcher
Nov 2015 – Sep 2018 Western Australia
At ICRAR-Curtin, I worked within the EoR group, developing theoretical models in the context of the Murchison Widefield Array (MWA). I contributed to models of the ionosphere, as well improving statistical models of bright foregrounds.

Projects

*
FORWARD
FORWARD is the name of my Marie Sklodowska-Curie Fellowship. It’s not an acronym: it is short for “Forward-Models of Cosmic Dawn: connecting 21cm simulations to the real world”. I am performing the fellowship at Scuola Normale Superiore in Pisa, Italy.
FORWARD
EDGES

The Experiment to Detect the Global Eor Signal is a “global 21cm signal” experiment. It reported the first evidence of the very first stars in the Universe in 2018.

I develop EDGES’ analysis pipeline for everything from data formats to calibration to data reduction and interpretation. I am particularly interested in validating our pipeline with Bayesian statistical models.

EDGES
HERA

The Hydrogen Epoch of Reionization Array is one of the largest radio interferometers in the world, and is looking to map the effects of the first stars, galaxies and black holes over the first billion years of the Universe.

I co-lead the Validation Team, developing increasingly realistic simulations of our observations at scale. I also lead the development of an interface to compute theoretical likelihoods on HERA data.

HERA
SKA

The SKA will be the largest radio telescope ever built, with a collecting area of over a square kilometre. It will be able to detect the faintest signals from the early Universe, and will be able to image the sky with unprecedented resolution.

I am a member of the EoR and Cosmic Dawn Science Working Group, contributing to the development of our tools and techniques, to be ready for the insane amount of data we’re going to get when SKA comes online.

SKA
21cmFAST

21cmFAST is the premiere semi-numerical 21cm cosmological simulator. It is currently used by every 21cm cosmology experiment to derive predictions of the spatial distribution of 21cm brightness temperature, to compare to observations.

I am a core developer in this project. My role has been primarily to wrap the fast C-code of the original simulator into Python, so that it can be more widely and easily used. This also has the benefit of adding a number of tests so that future development is safer and so that the code can be trusted by its many users.

21cmFAST

Codes

Browse the different codes I have created or am heavily involved in developing.

.js-id-theory
21cmSense
21cmSense computes sensitivity estimates for radio interferometers like HERA and the SKA. Originally written by Jonathan Pober, I have been the primary maintainer and developer of this code for the past few years.
21cmSense
hera-sim

hera_sim is one of the primary codes I contribute to in the HERA software suite. It is geared at simulating HERA-specific instrumental effects (thermal noise, cross-coupling, cable reflections) and injecting them into simulated visibilities.

This code is central to validating HERA’s analysis pipeline (as discussed in the linked paper).

hera-sim
21cmMC

21cmMC uses 21cmFAST to compare observations with theory, and predict astrophysical and cosmological parameters. This is a beast of a process, since it must run thousands of cosmological simulations to generate the predictions. Thus, 21cmMC is highly parallelized with MPI and is geared to run on supercomputers.

21cmMC was originally written by Brad Greig. My primary role has been to integrate it with 21cmFAST v3+, and to improve the interface and extensibility.

21cmMC
cosmotile
A simple library that is able to transform a periodic 3D coeval volume into a slice of an angular lightcone. The algorithm is presented in Boom Kittiwisit’s paper. I merely packaged up the code.
cosmotile
EDGES Analysis Pipeline

I am the primary developer of a new software pipeline for EDGES data. This involves a stack of codes for data reading/writing, calibration, analysis and parameter estimation.

The software is developed by a small team, and I am overseeing the development, in order to transform the EDGES analysis pipeline into something that is highly transparent, reproducible and accurate.

EDGES Analysis Pipeline
halomod

The halo model is a super-convenient analytic tool to predict the spatial distribution of “things” in the Universe. It uses dark matter halos as the basic scaffolding, and then associates other things with the halos — eg. galaxies or diffuse gas. The predictions can be compared to observations of the spatial layout of galaxies, for example, to determine either cosmological parameters or properties of the galaxies themselves.

halomod provides an extremely simple – but quite comprehensive and easy to extend and manipulate – framework for computing quantities from the halo model. It builds on hmf (another of my codes) to do this.

halomod
hankel

A small Python code that implements a way to do a very specific integral transform — the Hankel transform.

This transform is particularly common in astrophysics and cosmology (for instance, it is required to transform a correlation function into a power spectrum and vice versa). It is also a bit of a tricky integral to do, because it is highly oscillatory. This code implements a neat idea (not mine!) that is able to accurately perform this integral with good computational performance (in most cases).

hankel
21cmFAST

21cmFAST is the premiere semi-numerical 21cm cosmological simulator. It is currently used by every 21cm cosmology experiment to derive predictions of the spatial distribution of 21cm brightness temperature, to compare to observations.

I am a core developer in this project. My role has been primarily to wrap the fast C-code of the original simulator into Python, so that it can be more widely and easily used. This also has the benefit of adding a number of tests so that future development is safer and so that the code can be trusted by its many users.

21cmFAST
matvis
matvis is a visibility simulator: it computes what an interferometer (like HERA) would measure given a model of the sky. This process is incredibly computationally demanding for telescopes like HERA that have several hundred antennas (and, more importantly, several tens of thousands of pairs of antennas). matvis adopts a novel matrix-based approach to the computation that gets significant acceleration from modern GPUs. The original author of this code was Aaron Parsons. I have been the primary maintainer and performance-enhancer in recent times.
matvis
powerbox
A small Python code that is all about power spectra in “boxes” of arbitrary dimension. It can both create boxes with a given power spectrum, or determine the power spectrum of a given box. It does all of this with a lot of generality — we’re not just talking about 2- or 3-dimensional boxes, but boxes with an arbitrary number of dimensions (as long as your computer has enough memory!). Bring on the 5-dimensional cubes!
powerbox
pydftools
A package that implements a statistical method for robust estimation of density functions for observables, eg. luminosity functions and mass functions of galaxies or haloes. While I contributed to the statistical method itself, I was not the primary force behind the algorithm (that was Danail Obreschkow. This package is my Python implementation of the method.
pydftools
yabf
Yet Another Bayesian Framework. This package is a Python library that implements a simple way to define complicated hierarchical Bayesian models via static YAML files. It is used by edges-estimate to perform Bayesian inference for EDGES.
yabf
TheHaloMod

TheHaloMod is a web-application that uses my Python codes hmf and halomod to calculate halo model quantities, and serve up plots directly without ever having to install the codes, from the comfort of your web browser.

The purpose of TheHaloMod is to be useful both for researchers and also students approaching research into large-scale structure. TheHaloMod was formerly known as HMFcalc, and has been well-used since its creation in 2013. TheHaloMod adds extra goodies to the original HMFcalc — namely, the ability to calculate correlation functions and power spectra of galaxies.

From June 2021, TheHaloMod has become a shiny new single-page app, developed by a very talented group of Computer Science undergrads at ASU.

TheHaloMod
hmf

A python application that provides a flexible and simple way to calculate the Halo Mass Function for a range of varying parameters.

The Halo Mass Function is a critical quantity in determining properties of the large-scale structure of the Universe. It encodes the number of halos of a given mass in a given volume of the Universe, and can be predicted from some neat arguments given a cosmology that describes how the Universe is expanding. The predictions can be used to compare to observations of galaxy clusters to determine parameters of the large scale structure. The HMF is also a key ingredient in more complex predictions of the Universe, via the halo model.

hmf

Recent Publications

Quickly discover relevant content by filtering publications.

Contact