How Codex Is Redefining Black‑Hole Simulations and Expanding Scientific Frontiers

Background

Plasma surrounding a black‑hole event horizon is magnetized and highly relativistic. In regions of extreme temperature and low density, electrons and ions no longer collide frequently enough to be treated as a fluid; instead they follow helical trajectories along magnetic field lines. Accurate modeling therefore requires kinetic simulations that track the microscopic motion of a vast number of particles.

Computational challenge

State‑of‑the‑art simulations must follow up to trillions of electrons and ions with time steps small enough to resolve their gyromotion. On current supercomputers the majority of runtime is spent on these fine‑grained particle updates, making many scientifically interesting regimes effectively inaccessible.

Mathematical reformulation

Chi‑kwan Chan (University of Arizona & Steward Observatory) proposed to alter the mathematical description of particle motion so that the simulation no longer needs to resolve each individual spiral. The goal is to replace the explicit tracking of microscopic trajectories with a transformed set of equations that capture the same physics at a coarser level.

Use of Codex for algorithm generation

Chan employed OpenAI Codex to generate candidate coordinate transformations and numerical schemes. Codex produced a large set of possible formulations; each candidate was evaluated by comparing its output against analytically known solutions. Although many proposals were incorrect, the iterative testing process identified several viable alternatives.

Results

Among the successful outputs, a new coordinate change combined with a revised numerical integration method demonstrated a speed increase of up to 1000× relative to the original kinetic code. This acceleration makes simulations that were previously out of reach feasible. The team emphasizes that every Codex‑generated approximation must still be validated against benchmark problems to ensure scientific correctness.

Implications for black‑hole modeling

The speedup enables more realistic exploration of plasma dynamics near the event horizon, potentially improving the fidelity of synthetic images and theoretical predictions that complement observational efforts such as the Event Horizon Telescope.

References

https://openai.com/index/creating-new-simulations-black-holes/

https://openai.com/zh-Hans-CN/index/accelerating-science-gpt-5/

https://openai.com/index/using-codex-to-simulate-black-holes/

Code example

来源：ScienceAI
本文
约1500字
，建议阅读
5
分钟
对黑洞研究来说，这意味着模拟有机会变得更真实。