Furst, Merrick, James B. Saxe, and Michael Sipser. “Parity, circuits, and the polynomial-time hierarchy.” Theory of Computing Systems 17.1 (1984): 13-27.

Synopsis

Parity cannot be decided by a circuit in \(AC^0\) (see https://en.wikipedia.org/wiki/AC0).

I skipped section 2, which has some more results about the polynomial-time hierarchy.

The proof proceeds by contradiction: a polynomial sized depth \(d\) parity circuit can be used to construct a polynomial sized depth \(d - 1\) parity circuit, which means there can’t be a “smallest” \(d\).

Tricks I Learned

You can prove the existence of something by demonstrating that the probability of its existence is non-zero (sounds trivial when you say it :) ). This can be easier than, say, counting arguments, in some cases.

You can sometimes do “artificial” case analysis on a function by splitting its domain into convenient subsets.