Automated Auxiliary Qubit Allocation in High-Level Quantum Programming

TL;DR

This work tackles the problem of efficiently compiling high-level quantum programs by automating auxiliary qubit allocation for multi-qubit gate decompositions. The authors implement and evaluate this approach in the Ket platform via Libket, enabling the compiler to dynamically select decomposition algorithms based on circuit state and available hardware qubits, thereby reducing the final circuit's $CNOT$ count. Across Grover's algorithm and a state-preparation routine, Ket achieves substantial reductions in $CNOT$ gates compared with Qiskit, particularly when ample auxiliary qubits are available, and demonstrates robustness to noise in hardware-limited scenarios. The findings highlight the practical significance of automatic resource management in quantum compilation for near-term devices and point to future enhancements such as explicit programmer control of auxiliaries and broader algorithm coverage.

Abstract

We present a method for optimizing quantum circuit compilation by automating the allocation of auxiliary qubits for multi-qubit gate decompositions. This approach is implemented and evaluated within the high-level quantum programming platform Ket. Our results indicate that the decomposition of multi-qubit gates is more effectively handled by the compiler, which has access to all circuit parameters, rather than through a quantum programming API. To evaluate the approach, we compared our implementation against Qiskit, a widely used quantum programming platform, by analyzing two quantum algorithms. Using a 16-qubit QPU, we observed a reduction of 87% in the number of CNOT gates in Grover's algorithm for 9 qubits. For a state preparation algorithm with 7 qubits, the number of CNOT gates was reduced from $2.8\times10^7$ to $5.7\times10^3$, leveraging additional Ket optimizations for high-level quantum program constructions. Overall, a quadratic reduction in the number of CNOT gates in the final circuit was observed, with greater improvements achieved when more auxiliary qubits were available. These findings underscore the importance of automatic resource management, such as auxiliary qubit allocation, in optimizing quantum applications and improving their suitability for near-term quantum hardware.

Figures (15)

• Figure 1: Example of code implementation using the high-level quantum programming platform Ket.
• Figure 2: $\sqrt{X}$ gate implementation in Ket.
• Figure 3: $CU(2)$ decomposition, where $U = e^{i\theta}AXBXC$nielsenQuantumComputationQuantum2010.
• Figure 4: Network decomposition. The gate in the middle of the network can be further decomposed as shown in Figure \ref{['fig:cu2']}.
• Figure 5: V-Chain decomposition. Gates inside the dashed box are not required when using clean auxiliary qubits.