sub-grid
The set of hand-crafted, calibrated rules added to a simulation to represent physical processes that occur on scales smaller than the simulation's grid resolution and therefore cannot emerge naturally from the simulated equations.
As defined in 240 | Andrew Pontzen on Simulations and the Universe
Where it comes up · 1
Most important claims · 7 of 7Search all claims for “sub-grid” →
- The biggest source of uncertainty in simulations is the 'sub-grid': because computers can only divide the universe into a finite number of grid cells, crucial processes happening below that resolution (like star formation or convection-driven clouds in weather) cannot emerge naturally, so simulators must add hand-crafted rules and tune them against real observations until the virtual universe resembles the real one.240 | Andrew Pontzen on Simulations and the Universe
- Black hole formation and growth are difficult to simulate because they are sub-grid problems: stellar-mass black holes form at the end of star lives (already uncertain), and black holes grow by eating surroundings at rates that can't be directly resolved, while the energy released by feeding can blow away other gas—an extraordinarily complex feedback process that cannot simply be observed by zooming into a simulation.240 | Andrew Pontzen on Simulations and the Universe
- The Hubble Deep Field in the mid-1990s—pointing the telescope for a week at an apparently empty patch of sky to reveal it full of distant galaxies—was initially a catastrophe for simulators because predictions had said it would contain bright, large galaxies; the discrepancy was eventually traced not to a failure of the dark matter framework but to incorrect treatment of how and when stars form (sub-grid feedback details).240 | Andrew Pontzen on Simulations and the Universe
- Because no supercomputer can track individual stars among hundreds of billions per galaxy, simulators create 'star particles' representing whole collections of stars, formed when gas is sufficiently cold and dense; these particles move under gravity and feed energy and gas back into surroundings via stellar winds, magnetic fields, and supernova explosions.240 | Andrew Pontzen on Simulations and the Universe
- Ignoring sub-grid processes is doubly dangerous because they have feedback effects on the rest of the system: in weather, unmodeled clouds block sunlight, change ground temperatures, and alter winds, so missing the small-scale process corrupts the entire forecast—illustrating why small-scale physics cannot simply be neglected.240 | Andrew Pontzen on Simulations and the Universe
- One promising use of deep learning in cosmology is to run a very detailed simulation of a specific process (like a black hole eating gas) under many conditions, train a model on those results, and use it as a learned set of approximations feeding into the sub-grid—essentially automating and improving the approximations humans currently craft by hand.240 | Andrew Pontzen on Simulations and the Universe
- JWST's discovery of surprisingly massive early galaxies throws up interesting challenges but, as things stand, does not put the overall Big Bang model in jeopardy; it likely reflects the same pattern as the Hubble Deep Field—improving sub-grid simulations and refining data treatment will resolve it—though if the challenges prove insurmountable in a couple of years the discussion may change.240 | Andrew Pontzen on Simulations and the Universe