Comment on “Grover search with pairs of trapped ions”
Abstract: In this Comment on Feng's paper [Phys. Rev. A 63, 052308 (2001)], we show that Grover's algorithm may be performed exactly using the gate set given, provided that small changes are made to the gate sequence. An analytic expression for the probability of success of Grover's algorithm for any unitary operator $U$ instead of Hadamard gate is presented.
Introduction
In 2001, M. Feng proposed a method to perform Grover’s quantum search algorithm using pairs of trapped ions, an approach intended to minimize phase errors caused by "superpositional wave function oscillations." However, Hill and Goan identified a critical flaw: Feng’s proposed gate sequence for the "diffusion" operator—the part of the algorithm that amplifies the correct answer—was mathematically incorrect.
Specifically, Feng used the same rotation gate twice ($W P W$) instead of a rotation followed by its inverse ($W P W^\dagger$). This error was masked in the original paper because the results were plotted as "amplitudes" rather than "probabilities," leading to an overestimation of the algorithm's effectiveness for systems with three or more qubits.
Correcting the Gate Sequence
The authors re-evaluate the Grover iteration through the lens of geometric rotations. The primary correction involves redefining the diffusion operator $D$ to include the adjoint (inverse) of the rotation gate: $D = W P W^\dagger$. In quantum computing, the direction of a rotation is vital; Feng’s original proposal failed to "undo" the initial basis change correctly, which is why the algorithm didn't scale properly for larger numbers of qubits.
As shown in the circuit above, the implementation requires precise alternating rotations. With this small change to the gate sequence, Grover’s algorithm can be performed exactly using the trapped ion operations originally introduced by Feng.
Geometric Analysis and the Bloch Sphere
Instead of relying strictly on the standard Hadamard gate, the authors analyze Grover’s algorithm using an arbitrary unitary operator $U$. They map the Grover iteration onto a two-dimensional subspace (a pseudo-spin basis) and show that the entire process is equivalent to a rotation on a Bloch sphere.
This visualization reinforces the idea that Grover’s algorithm doesn't require Hadamard gates. Any unitary transformation $U$ that creates a superposition can be used, provided the reflection step is built using $U^\dagger$. By treating the search as a rotation in SU(2), the authors show that each "Grover iteration" simply nudges the quantum state closer to the target state by a fixed angle on the Bloch sphere.
Key Results and Success Probability
This geometric approach allows for the derivation of a clean, analytic expression for the probability of success after a certain number of steps. The definitive success probability formula for a generalized Grover search is given by:
$$P_s = \sin^2[(2s + 1)\theta]$$
where $s$ is the number of iterations and $\sin \theta = 1/\sqrt{N}$ (where $N$ is the number of possible states).
A key takeaway for researchers is the importance of clear labeling and rigorous verification. Feng’s original graphs showed amplitudes (e.g., $\approx 0.62$), which looked promising, but the actual probability of success (the square of the amplitude) was only about $38\%$. The corrected sequence is essential for the algorithm to work for 3 or more qubits, ensuring the algorithm achieves its full theoretical potential in trapped ion systems.
Summary updated