Reversibility

Here’s an Ars Technica article about a recent measurement at SLAC. It describes a kind of reaction that happens at the subatomic level which isn’t reversible in time. I like this, because it underlines a key point that discrete modelers sometimes forget: not everything in the universe is trivially symmetric. In this case, in order to maintain the larger picture of CPT symmetry, we expect particle interactions that aren’t symmetric in terms of charge or reflection to also not be symmetrical in time.

There are some digital models that start from the position of baking reversibility into the system in the hope that this will yield more consistently physics-like results. While these models have a lot to offer, and can yield some amazing effects, I remain unconvinced that they offer a deep parallel with nature. This is because such models don’t leave room for the kind of result that SLAC has revealed.

So isn’t reversibility important? Should we not be trying to build it into our simulations? Absolutely it’s important, but it’s also relatively easy to get reversibility to appear as an emergent property of a non-reversible algorithm. For an example, take the Jellyfish algorithm that I’ve covered in previous posts. By reproducing rotational and translational symmetries from bulk properties of a network, we also get time-reversal symmetry as a bulk property, even though the algorithm running the pseudo-particle only runs one-way. This enables us to build models that are temporally symmetric most of the time.

As a rule of thumb, I hold to a principle that Tommaso Bolognesi once stated to me. Where possible, aim   for emergence. It’s nice to have physical effects appear in a model, but the fewer of them we insert by fiat, the more likely our models are to surprise us with their results.

Advertisement