In the realm of superconducting quantum computing, three-dimensional superconducting radio-frequency (SRF) cavities offer exceptionally long-lived electromagnetic modes, and when coupled with nonlinear elements such as transmon qubits, they become promising architectures for bosonic quantum information processing. The inverse design of these systems, i.e., recovering device geometries that produce specified electromagnetic and coupling targets, is generally a one-to-many problem. The qubit-cavity coupling strength is highly sensitive to both the transmon geometry and its position within the cavity's electromagnetic field.
As these systems scale up, the growing design parameter spaces make conventional iterative simulations prohibitively costly. We present two deep neural network (DNN) approaches that address this inverse design problem at complementary levels of the design stack. The first proposes SRF cavity geometries that yield target cavity observables. The second proposes transmon qubit designs that achieve target qubit-cavity parameters, including coupling rate, qubit frequency, and anharmonicity $(g, \nu_q, \alpha)$.
The recovered candidate designs match the targets within approximately 5% (cavity) and 2% (transmon), validated by end-to-end re-simulation. Both approaches map desired device behavior directly to candidate designs, providing a fast alternative to the iterative simulation studies usually required.
Blogger's Review: This article highlights the application of deep learning in the design of quantum computing devices, especially in efficiently navigating high-dimensional design spaces. By mapping desired behaviors directly to design parameters, it significantly reduces design time and costs, pushing forward the advancement of quantum computing technology. It will be interesting to see how the network structures can be further optimized for improved design accuracy and efficiency.