The result of international collaboration, the Effort.jl programme is ushering in a new era for digital cosmology. It makes it possible to reproduce the evolution of large structures in the universe at unprecedented speed, giving researchers a more detailed view of the mechanisms that govern matter and space-time.
Observing the universe has always required colossal instruments, from giant telescopes to supercomputers. The more data is collected, the more complex and energy-intensive cosmological models become. This logic seemed inseparable from modern astronomy, until an international team proposed an unexpected alternative. Thanks to a unique cosmological simulator, designed to run on a simple laptop, exploring the birth of galaxies or probing dark energy suddenly becomes a matter of code, within easy reach.
Astronomy faces a data explosion
For two decades, observatories around the world have been scanning the sky with unprecedented precision. Giant programmes such as Euclid and DESI are accumulating terabytes of data on dark matter, galaxy formation and the expansion of the cosmos. However, this wealth of information comes at a cost. Each simulation of the universe becomes a colossal logistical and energy challenge.
Researchers often have to mobilise supercomputers for several days to test a single hypothesis on the dynamics of large-scale structures. This pace slows down analysis considerably in a field where the slightest variation in a parameter can alter an entire cosmological model.
For Marco Bonici, an astrophysicist at the Waterloo Centre for Astrophysics, this digital inertia threatened to slow down science itself. This observation led to the creation of Effort.jl, a tool capable of transforming a simple computer into a veritable cosmology laboratory. The team, bringing together the universities of Waterloo, SISSA and Paris-Saclay, set itself the task of making the simulation of the cosmos as fluid as navigating a file.
A cosmological simulator that fits in a laptop
Effort.jl, which stands for Effective Field Theory surrogaTe, is changing the way researchers explore the fabric of the cosmos. This open source tool, described in detail in the Journal of Cosmology and Astroparticle Physics, is based on a simple but revolutionary idea: emulating the physics of the universe rather than recalculating it each time.
They learn to reproduce the behaviour of complex models without going through all the equations again. In the case of Effort.jl, this approach makes it possible to analyse massive data sets in just a few hours, where previously it would have taken several days. Designed in the Julia language, Effort.jl relies on neural networks capable of grasping the subtleties of cosmological structures.
Its modular design allows it to take into account complex observational effects, such as distortions due to the speed of galaxies or the geometry of the observed sky. According to SciTechDaily, this optimisation now makes it possible to run the entire model on a simple laptop without losing accuracy.
This change of scale is revolutionising the practice of numerical cosmology. Researchers can adjust their hypotheses in real time, restart a calculation, compare several possible universes, or even correct biases in observations. Effort.jl thus becomes an extension of the scientific gaze, capable of interacting with data in real time.
What this miniaturisation means for fundamental research
Behind this technical feat lies a more profound change. Effort.jl marks a breakthrough in access to cosmological calculations. Whereas large collaborations once relied on heavy infrastructure, small teams can now test hypotheses with limited resources. Thanks to this new approach, young researchers are gaining autonomy and exploring models that were previously reserved for large laboratories.
This accessibility opens up new perspectives in the study of dark matter, neutrinos and fluctuations in the cosmic microwave background. Thanks to its open architecture, Effort.jl can already interface with other tools, such as Capse.jl for fossil radiation or Blast.jl for photometric survey analysis.
Beyond the field of cosmology, Marco Bonici and his collaborators envisage unexpected applications in climate modelling and weather forecasting. Their approach could make it possible to simulate chaotic systems with unprecedented efficiency by combining physical equations and machine learning.
In this way, the miniaturisation of scientific computing is sketching out another way of understanding the universe. A more agile, more accessible, but equally rigorous science. What was once the privilege of supercomputers is now simply a matter of imagination and code.
