Astrophysics PhD

Research

Some galaxies harbour a supermassive black hole at their centre that is actively pulling in surrounding matter. As matter falls in, enormous amounts of energy are released — and in the most extreme cases, a narrow beam of plasma is ejected at close to the speed of light. This beam, called a jet, can reach hundreds of thousands of light-years into intergalactic space.

These objects are called Active Galactic Nuclei (AGN). M87, shown in the video below, is one of the most studied: its jet was first observed over a century ago, and in 2019 it became the host of the first black hole ever directly imaged.

My research asks: what happens inside these jets to produce such violent, short-lived bursts of radiation? The source can brighten by orders of magnitude in a matter of hours — simultaneously across radio, optical, X-ray, and gamma-ray bands. The goal is to build a physical model that reproduces this behaviour and can be tested against real telescope measurements.

This video zooms into galaxy M87 using real visible light, X-ray and radio images of the galaxy, its jet of high-speed particles, and the shadow of its central black hole — the type of object this research is about. Credit: NASA's Goddard Space Flight Center.

The research problem

Problem

AGN jets produce sudden, intense flares visible across all wavelengths at once. What physical mechanism inside the jet triggers these events, and can it be predicted?

Approach

Simulate the jet numerically, compute the radiation it would produce at each frequency band, and compare the synthetic output against real telescope measurements.

Outputs

Predicted time-series light curves and resolved spatial maps, validated against independent multi-instrument observational data.

How the model works

A fast-moving disturbance travels down the jet and collides with a stationary compression region. The collision accelerates particles, which then radiate across the full electromagnetic spectrum — producing the observed multi-band flare.

Jet dynamics simulated numerically using relativistic fluid equations.
Radiation at each frequency computed from the simulated particle distribution.
Propagation effects — Doppler shift, light-travel delays — accounted for.
Synthetic outputs compared directly against telescope data.

Software

Two tools built or co-released during the thesis:

RIPTIDE: a Python pipeline that converts raw simulation data into predicted observable signals at each frequency — parallelised and designed for large output volumes.
AM³: contributed to the open-source public release of this spectral modelling code for the community.

AM³ code (GitLab)AM³ documentation

Collaboration & operations

Active member of H.E.S.S. — a network of ground-based gamma-ray telescopes operated by over 200 scientists across 40 institutions worldwide. Contributed to regular analysis operations and data monitoring processes.

Coordination between simulation, theory, and observational teams.
Shared data products with standardised, reproducible workflows.
Routine operations within a large distributed scientific infrastructure.

RES highlight — Barcelona Supercomputing Center

When a relativistic jet propagates through its host galaxy, it does not travel through empty space — it collides with stars, stellar winds, and interstellar gas. These interactions load the jet with mass, slow it down locally, and leave measurable imprints on the radiation it emits. Simulating this required running large-scale numerical experiments on MareNostrum5, one of Europe's most powerful supercomputers, operated by the Barcelona Supercomputing Center. The Spanish Supercomputing Network (RES) selected this work as a featured highlight.

The complex dance between relativistic outflows and their host galaxies ↗

Skills this work required

Physical simulation: ran large-scale numerical simulations of a high-speed magnetized flow on national supercomputing facilities, managing job scheduling, resource allocation, and output validation.
Signal synthesis and analysis: computed predicted emission across the full electromagnetic spectrum — from radio to gamma rays — as a function of time and spatial position, then compared against multi-instrument observational datasets.
Pipeline engineering: built RIPTIDE, a Python library handling the full chain from simulation output to observable prediction, with parallel processing and high-throughput I/O for large data volumes.
Parameter estimation: fitted model parameters to independent observational data, with systematic uncertainty quantification and clear documentation of model limits.
Open-source contribution: co-released AM³, a spectral modelling code now used by the international high-energy astrophysics community.

Thesis and publications

Full thesis and publication list for technical details.

Read the thesis (PDF)SciXplorer papers list Google Scholar

In brief

The environment is unforgiving: noisy data, extreme physical conditions, no possibility of a controlled experiment. The models either survive contact with real measurements or they don't. That standard — build it, test it rigorously, document where it fails — is what the work is built on, and what carries over to any domain where predictions need to be accountable.