Speaker
Description
Michael Hall, Marcin Zwierz, Howard Wiseman and I set out to establish how accurately a completely unknown phase can be estimated. One difficulty in proving a universal bound is that a measurement can be exceptionally accurate for particular values of the phase. By averaging the squared error uniformly over the possible phases, we obtained bounds that apply to arbitrary probe states and measurements. For an optical phase shift, we showed that the minimum root-mean-square error scales inversely with the mean number of photon passes through the phase shifter, and determined the optimal constant in this scaling.
These bounds also apply to quantum algorithms that estimate an energy encoded in the eigenphase of a unitary operator. Each controlled application of the unitary to an eigenstate plays the role of a pass through a phase shifter. I will describe how controlling the direction of a quantum walk doubles the relative phase accumulated per step, halving the number of steps needed for a given precision.
The choice of control state depends on how the estimate will be used. Two control qubits suffice to implement the measurement that minimises the Holevo phase variance. When estimating a molecular ground-state energy from the lowest of several samples, however, a single large downward error can determine the result. In this setting, the aim is to minimise the probability that the error exceeds a specified tolerance. I will explain why the optimal control amplitudes form a discrete prolate spheroidal sequence and how three control qubits suffice to approximate the corresponding measurement closely.
| I am the presenting author | Yes |
|---|