Thomas Edwards

I am a PhD student at GRAPPA (Gravitation AstroParticle Physics Amsterdam) centre of excellence. My work is focussed on detecting dark matter through astrophysical observations. Here is my personal webpage!


About Me

I work in theoretical astroparticle physics as part of the Institute of Physics at the University of Amsterdam, GRAPPA. Originally from south-east London I moved to Southampton in 2010 to complete my undergraduate degree in Physics with Astronomy after which I went to Cambridge to complete Part III of the mathematical tripos. I am now interested in probing the nature of dark matter (DM) through indirect detection methods as well theoretical studies.

The search for DM has led to many large experiments generating huge amounts of data in the hope of detecting a signal. I am interested in putting this information into a statistically rigorous framework allowing one to make accurate inferences from the data. Combining this with theoretical studies of the potential models will hopefully provide provide a window into this mysterious part of the Universe.

My Picture


Indirect Detection

DM is thought to only weakly interact with the Standard Model of particles but it may be able to self-annihilate and in the process produce electromagnetic waves which we can observe in the sky. In order to observe such a signal it is important to understand alternative processes that could lead something similar. In the case of indirect detection this means trying to accurately model all the astrophysical sources and disentangle the vast array of objects we find in our Universe.

Recently there has been a debate regarding the nature of an excess of gamma ray emission, consistent with a DM signal, from the centre of our galaxy observed by the Fermi LAT. It is also possible that a population of unresolved millisecond pulsars could produce a similar signal and much work is still needed to sway the debate one way or another

Dark Matter Distribution

Understanding DM distribution throughout the universe is a complex problem, leading researchers to perform large scale simulations of patches of the Universe. These simulations have become increasingly complex over the past several years, now attempting to correctly model galactic astrophysics such as supernova feedback to resolve DM distributions in the highly non-linear regime of gravitational interactions. Analysing the output of these simulations will help to provide an insight into potential observational targets in the hope of finding DM signals.

Statistics, Machine Learning, and Information Geometry

In between what we observe and our theoretical models sits statistics. Statistics allows us to quantify the amount of information we can infer about the parameters in our physical models. Since there has been no verified detection of dark matter to date, the name of the game is to set limits on the most theoretically motivated models. Obviously we hope to eventually find dark matter but in the mean time statistics can help to find the best methods to probe parts of the parameter space previously unknown.

Machine learning has in recent years become a central part of modern physics. Due to the complexity of our detectors and theoretical models, computationally expensive calculations are now the norm. The techniques developed in the field of machine learning allow physicists to perform the necessary complex calculations when a super-computer is not enough.

Information geometry is a field of mathematics that uses the techniques of differential geometry and applies them to probability distributions. It is an extremely promising avenue to apply similar techniques to physical problems to reveal optimal experimental design and data collection procedures.


Bayesian Model Comparison and Analysis of the Galactic Disk Population of Gamma-Ray Millisecond Pulsars [arxiv:1805.11097]

Dark Matter Model or Mass, but Not Both: Assessing Near-Future Direct Searches with Benchmark-free Forecasting [arxiv:1805.04117]

swordfish: Efficient Forecasting of New Physics Searches without Monte Carlo [arxiv:1712.05401]

A Fresh Approach to Forecasting in Astroparticle Physics and Dark Matter Searches [arxiv:1704.05458]

The Volumetric Rate of Superluminous Supernovae at z~1 [arxiv:1605.05250]


Accelerating the Search for Dark Matter with Mahine Learning, January, 2018

Title: Fast Forecasting for Counting Experiments

DANuCo, August, 2017

Title: Dark Information: Forecasting with the Fisher Matrix

TeVPA Particle Astrophysics, September, 2016

Title: Population synthesis of Fermi LAT sources: A Bayesian analysis using posterior predictive distributions

Amsterdam-Paris-Stockholm 6th Meeting, August, 2016

Title: Gamma-ray luminosity function of Millisecond Pulsars and 3FGL population study

21st Symposium on Astroparticle Physics in the Netherlands, April, 2016

Title: Bayesian Analysis of the Gamma-ray luminosity function of Millisecond Pulsars



In this coding package we implement some of the routines laid out in 1704.05458 for calculating exclusion limits and the discovery reach in a fast and efficient manner. In addition there are routines to visualise the Fisher matrix by plotting geodesics, streamlines or ellipses based off its major and minor eigenvectors. We are now working on a series of physics examples but the for now the functionality is explained in two jupyter examples.

My Picture My Picture


A useful wrapper for the YMW16 electron density model of the galaxy. Able to calculate the probability for a pulsar to exist at a some distance by performing a Monte-Carlo of the 30 uncertain YMW16 model parameters.


Feel free to send me an email!