Resilience Beyond the Light Cone: Error-Detected Primitives for Practical Dynamic Circuits

Dynamic quantum circuits, which integrate mid-circuit measurements and classical feedforward, offer a powerful way to generate long-range entanglement in constant depth. However, these circuits are often limited by the accumulation of errors from mid-circuit measurements and feedforward latency. This paper addresses this challenge by introducing a general framework for error-detected dynamic circuits, enabling the robust implementation of various quantum primitives without requiring additional ancilla qubits.

The core of the authors' approach is the distribute-collapse framework. The distribute primitive encodes a logical control into a distributed GHZ-like state across the device, while the collapse primitive reduces this state back to a single qubit. By augmenting the collapse primitive with a set of local check gadgets, the authors enable the detection of both gate and measurement errors. They distinguish between two types of checks: explicit checks, which directly verify stabilizer constraints, and implicit checks, which leverage redundant measurement outcomes to detect readout errors during the collapse process.

The authors demonstrate the utility of their framework through several key applications, including long-range CNOT gates, multi-qubit Pauli rotations, and the Hadamard test. Experimental results on an IBM superconducting processor show that this error-detection scheme allows for the preparation of entangled Bell pairs across 100 qubits with a fidelity of 0.59, surpassing the entanglement-certification threshold of 0.5. Furthermore, the authors demonstrate the constant-depth preparation of W states of up to 20 qubits, achieving absolute fidelity improvements of approximately 0.2 compared to baseline implementations.