Conditional on the Riemann hypothesis, Selberg showed in 1943 that the average size of the squares of differences between consecutive primes less than x is O(log(x)^4). Unconditional results still fall far short of this conjectured bound: Peck gave a bound of O(x^{0.25+epsilon}) in 1996 and to date this is the best known bound obtained using only methods from classical analytic number theory.


In this talk we discuss how sieve theory (in the form of Harman's sieve) can be combined with classical methods to improve bounds on the number of short intervals which contain no primes, thus improving the unconditional bound on the mean square gap between primes to O(x^{0.23+epsilon}).