Improved stack-slide searches for gravitational-wave pulsars

We formulate and optimize a computational search strategy for detecting gravitational waves from isolated, previously unknown neutron stars (that is, neutron stars with unknown sky positions, spin frequencies, and spin-down parameters). It is well known that fully coherent searches over the relevant parameter-space volumes are not computationally feasible, and so more computationally efficient methods are called for. The first step in this direction was taken by Brady and Creighton (2000), who proposed and optimized a two-stage, stack-slide search algorithm. We generalize and otherwise improve upon the Brady-Creighton scheme in several ways. Like Brady and Creighton, we consider a stack-slide scheme, but here with an arbitrary number of semicoherent stages and with a coherent follow-up stage at the end. We find that searches with three semicoherent stages are significantly more efficient than two-stage searches (requiring about 2-5 times less computational power for the same sensitivity) and are only slightly less efficient than searches with four or more stages. We calculate the signal-to-noise ratio required for detection, as a function of computing power and neutron star spin-down-age, using our optimized searches.