What this covers

Sean Carroll speaks with cosmologist Andrew Pontzen about how simulations work in modern cosmology and what they can and cannot tell us about the universe. The conversation spans the history of computational astronomy—from a 1940s light-bulb apparatus that exploited inverse-square-law physics to simulate colliding galaxies, through decades of digital refinement—and settles on a central claim: cosmological simulations are not transparent calculations of physics but creative constructions laden with approximations, especially in their treatment of sub-grid physics. Because the universe is too large and complex to fit inside any computer, simulators must invent hand-crafted rules for processes like star formation and supernova feedback, then tune those rules against real observations until the virtual universe resembles the actual one. Understanding those approximations is essential to interpreting what simulations reveal.

The episode explores where approximations matter most. Pontzen explains how the cosmic microwave background cannot be used directly as simulation starting conditions, why sub-grid treatment of feedback has repeatedly forced reinterpretation of observations (the Hubble Deep Field, early galaxies seen by JWST), and how dark matter has become the preferred framework not because it is proven correct but because it has been remarkably predictive—a position he distinguishes carefully from declaring modified gravity wrong. The discussion also ventures into philosophy: whether simulations count as theory, experiment, or something spanning both; how the line between simulation and real experiment erodes under scrutiny; and how independent corollary testing rather than recalibration against surprising data represents proper scientific method. Outstanding questions in black hole growth and galaxy formation receive treatment as illustrations of what simulations currently cannot resolve.

Sharpest takeaway

Pontzen argues that cosmological simulations are not straightforward computations of known physics but creative, approximation-laden constructions whose key uncertainties live in 'sub-grid' rules; understanding those approximations is essential to correctly interpreting what simulations tell us about the universe.

  • The universe cannot fit inside a computer, so simulators must invent sub-grid rules calibrated against real observations to force phenomena like star formation to occur.
  • Dark matter is favored over modified gravity primarily because it has been a remarkably predictive paradigm across vast scales.
  • Surprising results (Hubble Deep Field, JWST early galaxies) reflect mistreatment of feedback and sub-grid physics rather than failures of the underlying Big Bang model.

The argument · threads1 threads · 37 claims
0.85

Early massive galaxies observed by JWST are probably rare outliers in the right conditions rather than systematic formation gaps.

1 pointcentrality 5/5
  • The most natural way to interpret JWST's massive early galaxies may be that they exist but are rare—a 'fast track' for roughly 1% of galaxies that happen to be in the right place at the right time—rather than a missing major feature of galaxy formation; this connects to the open question of how supermassive black holes grew so quickly.

    “it's not necessarily something that these massive galaxies at early times exist but are rare... there's a fast track for the 1% that if you're really in the right place at the right time, you can make a more massive galaxy than we would have expected”