The terms in Vaughan's identity are not random sums: they are convolutions of simpler and/or shorter range arithmetic functions. Convolution allows the separation of variables in the sum, and hence it allows a more efficient estimation of the sum. A classical and prime example is Dirichlet's hyperbola method.
GH from MO
- 105.4k
- 8
- 293
- 398