Quantum computation mediated by an ancilla of arbitrary dimension

Timothy Proctor, Shane Dooley, Viv Kendon

Abstract: Models of universal quantum computation where the required interactions between computational qubits are mediated by some ancillary system are highly relevant to experimental realisations of a quantum computer. We introduce two such models of quantum computation, the first of which employs a d-level (qudit) ancilla and the second of which is based on the coherent states of n spins. The computation is implemented with geometric phase gates, created via controlled displacement operators defined on a discrete lattice phase space, in the case of a qudit ancilla, and a Bloch sphere, in the case of a spin coherent state ancilla. Both of these models are shown to be different possible generalisations to finite dimensions of the field-mode quantum bus (qubus) computer. We then consider the computational complexity of the qubus architecture and show that although in certain circumstances the model can be used to reduce the number of operations required to implement quantum circuits, contrary to previous claims [New J. Phys. 13, 095007, 2011] this is not the case for the quantum Fourier transform (QFT). Furthermore, using the generalisation of the qubus model to a qudit bus, we show that this reduction in the circuit complexity depends on the dimensionality of the bus.

Introduction

Building a universal quantum computer requires reliable interactions between qubits. In many physical implementations, direct coupling between distant qubits is technically challenging and prone to noise. Ancilla-Mediated Quantum Computation (AMQC) offers an alternative: a single ancillary system, or "quantum bus," interacts sequentially with register qubits to mediate gates. While previous "qubus" models utilized infinite-dimensional field modes (like photons in a cavity), this work generalizes the architecture to finite-dimensional ancillas, specifically $d$-level qudits and spin ensembles.

The Qudit Bus

The first model employs a $d$-dimensional qudit as the mediator. The authors define a discrete phase space for the qudit that possesses the topology of a torus. By applying controlled displacement operators, the ancilla is moved along closed loops in this phase space.

The toroidal phase space of a d-dimensional qudit ancilla, where controlled displacements generate geometric phases.

Due to the non-commutative nature of these displacements on a torus, the register qubits acquire a geometric phase. This mechanism allows for the implementation of universal gates, such as the Controlled-Phase gate. A significant advantage is that a qudit with $d > n$ can implement a controlled-U gate on $n$ qubits with only $2(n+1)$ interactions, demonstrating how dimensionality serves as a computational resource.

The Spin-Ensemble Bus

The second model utilizes an ensemble of $N$ spins (e.g., atomic ensembles or NV centers) as the ancilla. Here, the phase space is represented by a Bloch sphere.

Geometric phase accumulation on the Bloch sphere. The area enclosed by the path (green) determines the phase acquired by the register qubits.

Similar to the qudit model, rotations on the sphere act as displacements that generate geometric phases. The authors demonstrate that as $N \to \infty$, the spherical geometry locally approximates a flat plane, recovering the behavior of the continuous-variable field-mode qubus. Even for finite $N$, numerical simulations show that the "curvature" errors are minimal; for $10^7$ spins, gate fidelities exceed 0.9999.

Complexity and the Quantum Fourier Transform

A key contribution of this paper is the re-evaluation of the qubus model's computational complexity. Previous literature suggested that the qubus could significantly reduce the number of operations for the Quantum Fourier Transform (QFT). However, the authors prove that while the qubus architecture reduces the total number of interactions for certain multi-qubit gates, it does not fundamentally alter the $O(n^2)$ scaling of the QFT. This clarifies the specific types of algorithms where ancilla-mediated models provide a true efficiency boost.

Conclusion

By extending the qubus framework to finite-dimensional systems, this research provides a flexible blueprint for universal quantum computation. Whether using the discrete lattice of a qudit or the collective states of a spin ensemble, geometric phase gates offer a robust path toward scalable quantum architectures. Future work will focus on mitigating correlated errors that arise from the shared ancilla and exploring higher-symmetry ancillary systems.

Summary updated

Abstract

Models of universal quantum computation where the required interactions between computational qubits are mediated by some ancillary system are highly relevant to experimental realisations of a quantum computer. We introduce two such models of quantum computation, the first of which employs a d-level (qudit) ancilla and the second of which is based on the coherent states of n spins. The computation is implemented with geometric phase gates, created via controlled displacement operators defined on a discrete lattice phase space, in the case of a qudit ancilla, and a Bloch sphere, in the case of a spin coherent state ancilla. Both of these models are shown to be different possible generalisations to finite dimensions of the field-mode quantum bus (qubus) computer. We then consider the computational complexity of the qubus architecture and show that although in certain circumstances the model can be used to reduce the number of operations required to implement quantum circuits, contrary to previous claims [New J. Phys. 13, 095007, 2011] this is not the case for the quantum Fourier transform (QFT). Furthermore, using the generalisation of the qubus model to a qudit bus, we show that this reduction in the circuit complexity depends on the dimensionality of the bus.
BibTeX
@inproceedings{proctor2015quantum,
  author = {Timothy Proctor and Shane Dooley and Viv Kendon},
  title = {{Quantum computation mediated by an ancilla of arbitrary dimension}},
  booktitle = {Physical},
  year = {2015},
  doi = {10.1103/physreva.91.012308},
}
Figure