Abstract
Quantum computation architecture based on $d$-level systems, or qudits, has attracted considerable attention recently due to their enlarged Hilbert space. Extensive theoretical and experimental studies have addressed aspects of algorithms and benchmarking techniques for qudit-based quantum computation and quantum information processing. Here, we report a physical realization of qudit with upto 4 embedded levels in a superconducting transmon, demonstrating high-fidelity initialization, manipulation, and simultaneous multi-level readout. In addition to constructing SU($d$) operations and benchmarking protocols for quantum state tomography, quantum process tomography, and randomized benchmarking etc, we experimentally carry out these operations for $d=3$ and $d=4$. Moreover, we perform prototypical quantum algorithms and observe outcomes consistent with expectations. Our work will hopefully stimulate further research interest in developing manipulation protocols and efficient applications for quantum processors with qudits.
Introduction
The prevailing paradigm in quantum computing relies on the qubit, a two-level system. However, the information capacity of a quantum processor scales exponentially with the number of levels per physical site. While $N$ qubits span a Hilbert space of dimension $2^N$, a system of $N$ qudits ($d$-level systems) spans $d^N$. This expanded space offers significant advantages, including more efficient algorithm encoding, simplified circuit topologies for complex gates, and enhanced pathways for quantum error correction.
Despite these benefits, implementing high-fidelity qudits in superconducting circuits presents substantial challenges. Higher energy levels in transmons are typically more susceptible to decoherence and charge noise. Furthermore, as the number of levels increases, the frequency spectrum becomes crowded, making selective control difficult. This work addresses these hurdles by developing a comprehensive toolbox for the control, readout, and benchmarking of a four-level ($d=4$) transmon qudit.
Experimental Implementation
The researchers utilized a superconducting transmon circuit designed for long coherence times and minimal charge-parity sensitivity. The system's energy levels are housed within a weakly anharmonic potential, where the first four levels ($|0\rangle, |1\rangle, |2\rangle, |3\rangle$) are employed for computation.
To achieve universal $SU(d)$ control, the team implemented a gate decomposition strategy based on Gaussian elimination. Specifically, they utilized a "bubbling" approach that breaks down arbitrary multi-level operations into a sequence of $SU(2)$ rotations between adjacent energy levels. This method is particularly advantageous for transmons because single-photon transitions (e.g., $|1\rangle \leftrightarrow |2\rangle$) are faster and less prone to error than multi-photon transitions. Additionally, the "Virtual Z" gate technique was extended to the qudit regime, allowing for instantaneous, perfect-fidelity phase shifts by adjusting the phase of subsequent microwave pulses.
Readout was performed using a single readout resonator coupled to the transmon. By analyzing the transmitted signal in the quadrature (IQ) plane, the system can simultaneously distinguish between all four states with high precision, avoiding the time-consuming sequential "shelving" methods often used in higher-level systems.
Benchmarking and Gate Fidelity
The performance of the qudit toolbox was rigorously evaluated using a suite of benchmarking protocols, including Quantum State Tomography (QST), Quantum Process Tomography (QPT), and Randomized Benchmarking (RB).
Randomized benchmarking revealed exceptional gate performance. For the $d=3$ (qutrit) system, the average Clifford gate fidelity was measured at 0.9946(1), with $\pi/2$ pulse errors as low as $7.6 \times 10^{-4}$ for certain transitions. In the $d=4$ (ququart) case, the Clifford fidelity remained high at 0.942(2). Simultaneous readout fidelities for all four states were maintained above 91.1%, and a complex four-level superposition state was prepared with a fidelity of 99.64%.
Algorithmic Demonstrations
To demonstrate the practical utility of the qudit platform, the researchers executed several prototypical quantum algorithms:
- Discrete Fourier Transform (DFT): A fundamental building block for many quantum algorithms, the DFT was implemented for both $d=3$ and $d=4$, achieving a process fidelity of 87.64% for the latter.
- Grover’s Search Algorithm: The algorithm successfully identified "labeled" states across all four levels, demonstrating the ability to perform search operations within a single physical qudit.
- Variational Quantum Eigensolver (VQE): The team simulated the ground state energies of $H_2$ and $HeH^+$ molecules. By mapping the molecular orbitals to the four levels of the qudit, they were able to reproduce theoretical energy curves within "chemical accuracy" across various bond distances, showcasing the resource efficiency of qudit-based simulation.
Conclusion
This work demonstrates a high-fidelity, four-level superconducting transmon qudit capable of executing universal $SU(d)$ operations and complex algorithms. While challenges such as shortened coherence times at higher levels and frequency crowding remain, the results show that qudits are a viable and efficient alternative to multi-qubit systems for specific tasks. Future improvements in pulse shaping and scaling to multi-qudit architectures will be essential to fully realize the potential of high-dimensional quantum information processing.
Summary updated