You try to explain why the universe behaves the way it does, using mathematics and computation rather than telescopes. The work spans a spectrum. At one end is pencil-and-paper theory — deriving how matter behaves near a black hole, or how structure grew in the early universe. At the other is large-scale computational astrophysics — building and running simulations of galaxy formation, stellar explosions, or planetary system dynamics on some of the world's most powerful computers, then mining the output for understanding. In practice most modern astrophysicists live somewhere in the middle, combining analytic insight with serious software.
The distinctive intellectual move is that your "experiment" is usually a simulation. Because you cannot rerun the formation of a galaxy, you build a model universe inside a computer, evolve it under the laws of physics, and compare what comes out against what telescopes actually see. When model and observation disagree, you have learned something — either the physics in your model is incomplete or the observation is telling you something new. That loop is the engine of the field, and it is why Discovery, not Creation, is the primary pull: the simulation is a tool for finding out, not an end in itself.
Explanation sits close underneath, because a theory that cannot be communicated — to observers, to students, to the wider field — does not travel. The best theoretical astrophysicists are often unusually good at making the abstract graspable.
Kitsune can talk through anything on this page — whether it might suit you, what to do next, questions this page doesn't answer. Everything here is yours to read either way.
The code is the science, and the software engineering is real. Some of the field's most important simulation codes have been developed continuously for decades, and a large part of the job is reading, understanding, extending, and debugging large scientific codebases. People who imagine theory as pure mathematics are often surprised by how much of the day is spent programming.
A negative or null result is common and rarely publishable, which can be quietly demoralising. You can spend months on a model that simply does not work, with little to show for it externally. Tolerance for that kind of ambiguity is an occupational requirement.
Physics, astrophysics, or maths degree, then a PhD in theoretical or computational astrophysics — STFC-funded in the UK [official_funder, STFC/UKRI 2025-26] — followed by postdoctoral positions on the same pay scale as the rest of the field [survey_aggregator, Prospects/PayScale 2025-26]. Heavy programming skill (Python, often C/C++ or Fortran for performance, and experience with high-performance computing) is essential and is also exactly what makes this the easiest astrophysics specialism to convert into a high-paying data or software career later.
ML emulators/SBI democratise theory, but the quality gap lands hardest here — miscalibration means confident wrong science.
Strengthening for ML-fluent theorists; 'code is the science' intensifies.
People drawn to Theoretical / Computational Astrophysicistare often drawn to these — in the order they're closest. The ones marked sit in a different field entirely.