A School of Emergent Priorities paper shows that the residual error in a simulated universe behaves exactly like a physical ingredient of that universe, and that one choice in the code decides which ingredient it becomes.
Which equation a simulation uses for the expansion rate decides whether its error looks like curved space or a cosmological constant
The School of Emergent Priorities has extended its interest in what accumulates unnoticed to the largest setting on offer: the expansion of a model universe.
The Error Has an Equation of State, by doctoral candidate Gunnhild Ekwueme and Associate Professor Kwame Lindqvist, Director, Master of Priority Studies, follows the small discrepancies a computer introduces when it calculates how a universe grows. The paper shows that such a discrepancy does not stay a discrepancy. It behaves, exactly, like an extra ingredient of the universe being simulated, and a single line in the code decides which one: a cosmological constant that never fades, or a curvature of space that thins out as the universe expands. In one model, an error of one part in a million at the start came to account for a third of the universe by the end of the run.
The universe is a very long document, and most of us only ever read the margins. I am grateful to have been handed such a wide margin so early in my candidature.
— Gunnhild Ekwueme, Doctoral Candidate, School of Emergent Priorities
Working across two families of integration method and a spectrum of evolution settings, the team built an end-to-end account of how residual error is carried forward through a simulated cosmic history, and identified the settings under which it settles and those under which it compounds.
The University sees in the paper a welcome widening of the School’s horizon, in the most literal sense available to it. “Every institution carries forward something from its first day that it never quite meant to,” said Associate Professor Lindqvist. “It has been a pleasure to watch a candidate find the same habit at work so much further from home.”
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.nzjg65.