We review the basic theoretical underpinnings of the Grover algorithm, providing a rigorous and well motivated derivation. We then present a generalization of Grover's algorithm that searches an arbitrary subspace of the multi-dimensional Hilbert space using a diffusion operation and an amplitude amplification procedure that has been biased by unitary {\em steering operators}. We also outline a generalized Grover's algorithm that takes into account higher level correlations that could exist between database elements. In the traditional Grover algorithm, the Hadamard gate selects a uniform sample of computational basis elements when performing the phase selection and diffusion. In contrast, steered operators bias the selection process, thereby providing more flexibility in selecting the target state. Our method is a generalization of the recently proposal pattern matching algorithm of Hiroyuki et al..
Grover's quantum search algorithm evolves a quantum system from a known source state
∣s⟩ to an unknown target state
∣t⟩ using the selective phase inversions,
Is and
It, of these two states. In one of the generalizations of Grover's algorithm,
Is is replaced by a general diffusion operator
Ds having
∣s⟩ as an eigenstate and
It is replaced by a general selective phase rotation
Itϕ. A fast quantum search is possible as long as the operator
Ds and the angle
ϕ satisfies certain conditions. These conditions are very restrictive in nature. Specifically, suppose
∣ℓ⟩ denote the eigenstates of
Ds corresponding to the eigenphases
θℓ. Then the sum of the terms
∣⟨ℓ∣t⟩∣2cot(θℓ/2) over all
ℓ=s has to be almost equal to
cot(ϕ/2) for a fast quantum search. In this paper, we show that this condition can be significantly relaxed by introducing appropriate modifications of the algorithm. This allows access to a more general class of diffusion operators for fast quantum search.