Optimal gentle measurements of finite-dimensional quantum states

Quantum state tomography typically involves destructive measurements that collapse the state of a system. This paper investigates the statistical limits of estimating an unknown d-dimensional quantum state (qudit) when the measurement process is constrained to be alpha-gentle. A measurement is alpha-gentle if the trace distance between the state before and after the measurement is at most alpha, effectively limiting the amount of information extracted to preserve the state's integrity.

The authors analyze the minimax estimation risk for these gentle measurements, drawing a formal connection between quantum gentleness and quantum differential privacy. They construct a gentle projected least squares estimator and prove its optimality by developing a new lower bound framework based on Assouad’s method. To demonstrate physical feasibility, they propose an implementation using an ancillary state and a CNOT gate to entangle it with the initial state, showing that the resulting measurement outcomes satisfy local differential privacy.

The study establishes that the optimal minimax rate for estimating a rank-r state is proportional to rd^2/(nalpha^2). For full-rank states (r=d), this results in a rate of d^3/(nalpha^2). A surprising finding is that the penalty for gentleness, d/alpha^2, scales with the ambient dimension of the Hilbert space. This is fundamentally different from classical differential privacy, where the penalty typically scales linearly with the number of parameters (rd). This suggests that the geometry of quantum state space permits significantly more efficient measurement strategies than those found in classical statistical models.