9

When I read about the Church-Turing thesis it seems to be a common claim that "physical reality is Turing-computable." What is the basis for this claim? Are there any theoretical results along these lines?

For context, I am a researcher who works on physical simulations, so of course I'm aware that many partial differential equations (PDEs) that would arise in nature (e.g. the heat equation, wave equation, etc) can be approximated by numerical methods like finite elements, and that for many PDEs a solution can be estimated to arbitrary accuracy given enough computation (by decreasing the space and time step sizes).

However, I also know that proving convergence of finite element methods is notoriously difficult for PDEs of any appreciable complexity, even "easy" PDEs like the mean curvature flow that describes the shape of a soap film. I also know that many "Zeno-type" situations arise in practice in physical systems, such as Euler's disk or inelastic collapse. Is there reason to believe that the solutions to all PDEs, or at least all PDEs that would arise in nature, are Turing-computable?

Raphael
  • 73,212
  • 30
  • 182
  • 400
user168715
  • 221
  • 1
  • 5

1 Answers1

4

The branch of mathematics and computer science that studies these questions is computable mathematics. The general answer is that things tend to be computable. I would add to that the observation that it often takes some work to establish computability. For instance, you mention finite element methods and the problems with their convergence. This proves absolutely nothing about computability of PDEs because there are, or might be, other methods for solving PDEs.

Some references that migth interest you, in order of relevance:

Anton Trunov
  • 3,499
  • 1
  • 19
  • 26
Andrej Bauer
  • 31,657
  • 1
  • 75
  • 121