Abstract
Superconducting circuits have emerged as a promising platform to build quantum processors. The challenge of designing a circuit is to compromise between realizing a set of performance metrics and reducing circuit complexity and noise sensitivity. At the same time, one needs to explore a large design space, and computational approaches often yield long simulation times. Here we automate the circuit design task using SCILLA, a software for automated discovery of superconducting circuits. SCILLA performs a parallelized, closed-loop optimization to design circuit diagrams that match pre-defined properties such as spectral features and noise sensitivities. We employ it to discover 4-local couplers for superconducting flux qubits and identify a circuit that outperforms an existing proposal with similar circuit structure in terms of coupling strength and noise resilience for experimentally accessible parameters. This work demonstrates how automated discovery can facilitate the design of complex circuit architectures for quantum information processing.
Introduction
As quantum computing moves toward larger-scale architectures, the complexity of designing the underlying superconducting circuits increases exponentially. Traditionally, this process relies on human intuition—a "forward design" approach where engineers propose a topology and simulate its properties. However, the vast search space of circuit topologies and component values (inductances, capacitances, and Josephson junctions) makes it difficult to find truly optimal solutions. This paper introduces an "inverse design" methodology, where desired physical properties are specified first, and an automated system discovers the corresponding circuit diagram.
The SCILLA Framework
To automate this discovery, the researchers developed SCILLA (Superconducting Circuit Innovation by Log-Loop Automation). SCILLA utilizes a modular, closed-loop workflow to navigate the non-convex landscape of circuit design.
The workflow consists of three primary modules:
- Design Module: Proposes circuit layouts and parameters using a mix of random search, gradient-based fine-tuning (L-BFGS-B), and Particle Swarm Optimization (PSO) to avoid local minima.
- Property Computation Module: Automatically constructs the Hamiltonian of the proposed circuit and performs diagonalization to extract energy spectra and noise sensitivities.
- Merit Evaluation Module: Scores the circuit based on its adherence to target performance metrics.
To manage the high computational cost of simulating thousands of circuits, SCILLA employs an asynchronous database system (SQLite), allowing optimization and simulation tasks to run in parallel across high-performance computing clusters.
Application: 4-Local Coupler Design
A critical challenge in quantum annealing and simulation is implementing 4-local interactions, where four qubits interact simultaneously. The researchers framed this as a "spectral engineering" problem: a coupler circuit that exhibits a specific "double-well" ground state energy profile as a function of external flux will mathematically mediate the required 4-qubit interaction.
By targeting this specific energy spectrum, SCILLA was tasked with searching for three-node circuits that could provide the necessary coupling strength while remaining resilient to flux noise.
Key Results and Discovery
The automated search successfully identified a novel two-loop circuit, designated "Circuit C," which significantly outperforms previous human-designed proposals. Key findings include:
- Enhanced Coupling: The discovered circuit achieved a coupling strength of 573 MHz, far exceeding prior theoretical benchmarks.
- Noise Resilience: The design demonstrated a 3.3x reduction in spectral asymmetry caused by flux noise, a critical factor for maintaining coherence in real-world hardware.
- Galvanic Advantage: SCILLA revealed that galvanic (direct) connections between inductive loops provide stronger coupling than traditional mutual induction, all while using component values that are easier to fabricate.
- Efficiency: The system evaluated over 10,000 circuits in less than 7 hours, demonstrating the power of parallelized automated discovery over manual iteration.
Conclusion
This work demonstrates that automated discovery can effectively bridge the gap between theoretical quantum requirements and practical circuit implementation. While computational scaling remains a challenge for larger circuits, the use of SCILLA to find optimized "building blocks" like the 4-local coupler marks a significant step toward the automated synthesis of large-scale quantum processors. Future developments will focus on integrating geometric fabrication constraints directly into the optimization loop.
Summary updated