GraphQ: High-Coherence Superconducting Circuit Optimization using Graph Machine Learning*

High coherence times in the superconducting qubits are required to achieve scalable quantum computing. Researchers attempt to achieve this through material engineering [1] or modifying the Hamiltonian mechanics, like in the zero-pi qubit [2]. Current state-of-the-art simulators [3] can compute the circuit eigensystem but are too slow for a brute-force search of the best-performing qubit design. In this work, we propose GraphQ, a machine-learning-based approach to explore high-coherence qubits. We formulate the qubit design as a graph optimization problem where we maximize the coherence time and gate speed. The proposed framework actively queries a simulator to explore multi-node architectures efficiently [4]. We present our recent progress in obtaining high-performance candidate qubit designs.

 

 

 

[1] Place, Alexander PM, et al. "New material platform for superconducting transmon qubits with coherence times exceeding 0.3 milliseconds." Nature Communications 12.1 (2021): 1779.

 

[2] Gyenis, András, et al. "Experimental realization of a protected superconducting circuit derived from the 0–π qubit." PRX Quantum 2.1 (2021): 010339.

 

[3] Chitta, Sai Pavan, et al. "Computer-aided quantization and numerical analysis of superconducting circuits." New Journal of Physics 24.10 (2022): 103020.

 

[4] Tuli, Shikhar, and Jha, Niraj K. "BREATHE: Second-Order Gradients and Heteroscedastic Emulation based Design Space Exploration." arXiv preprint arXiv:2308.08666 (2023).