The role of symmetries in adiabatic quantum algorithms

Exploiting the similarity between adiabatic quantum algorithms and quantum phase transitions, we argue that second-order transitions -- typically associated with broken or restored symmetries -- should be advantageous in comparison to first-order transitions. Guided by simple examples we construct an alternative adiabatic algorithm for the NP-complete problem {\em Exact Cover 3}. We show numerically that its average performance (for the considered cases up to ${\cal O}\{20\}$ qubits) is better than that of the conventional scheme. The run-time of adiabatic algorithms is not just determined by the minimum value of the fundamental energy gap (between the ground state and the exited states), but also by its curvature at the critical point. The proposed symmetry-restoring adiabatic quantum algorithm only contains contributions linear and quadratic in the Pauli matrices and can be generalized to other problem Hamiltonians which are decomposed of terms involving one and two qubits. We show how the factoring problem can be cast into such a quadratic form. These findings suggest that adiabatic quantum algorithms can solve a large class of NP problems much faster than the Grover search routine (which corresponds to a first-order transition and yields a quadratic enhancement only).