- Hardware
- Also: mid-circuit measurement and reset
- Also: MCMR
- Also: dynamic circuit measurement
- Also: feed-forward measurement
Mid-Circuit Measurement
A mid-circuit measurement is a quantum measurement applied to one or more qubits during a quantum circuit without terminating the computation, enabling dynamic circuits and real-time classical feedback.
Traditional quantum circuit models execute all gates, then measure all qubits at the end. Mid-circuit measurement breaks this model: a qubit is measured partway through the circuit, the measurement outcome is recorded classically, and the circuit continues running on the remaining qubits. The measured qubit may be reset and reused, or simply abandoned. This capability enables a class of computations called dynamic circuits, where later gates depend on earlier measurement outcomes.
What happens physically
Measuring a qubit mid-circuit collapses it to a definite or state. This is a projective measurement, identical to a final measurement, and it causes the same wavefunction collapse. Qubits that were entangled with the measured qubit instantaneously (in the quantum mechanical sense) have their shared entanglement resolved. The remaining unmeasured qubits continue to evolve coherently.
The classical outcome of the measurement is available to the control hardware in real time. Classical logic can then condition subsequent gate operations on this outcome, implementing classically controlled quantum gates. This real-time feedback loop is the defining feature of dynamic circuits and distinguishes mid-circuit measurement from simple intermediate decoherence.
Hardware requirements
Mid-circuit measurement is technically demanding. It requires:
- Fast readout. The measurement must complete quickly enough that other qubits do not decohere while waiting. For superconducting qubits, dispersive readout via a coupled resonator achieves measurement in 100 to 500 nanoseconds.
- Low measurement crosstalk. Activating the readout resonator for one qubit must not disturb neighboring qubits. Crosstalk-induced dephasing on unmeasured qubits is a primary hardware challenge.
- Fast classical control. The classical feedback decision must be made and the resulting gate issued before the coherence of waiting qubits degrades. Round-trip latencies below 1 microsecond are typically required.
- Active qubit reset. After measurement, if the qubit is to be reused, it must be reset to before the next gate sequence.
IBM, Google, Quantinuum (trapped-ion), and IonQ have all demonstrated mid-circuit measurement with varying degrees of fidelity and feedback latency.
Applications
Quantum error correction. Error correction codes like the surface code require syndrome measurements every few microseconds. These are mid-circuit measurements on ancilla qubits, whose outcomes feed into a classical decoder that determines which corrections to apply to data qubits. Without mid-circuit measurement, real-time error correction is impossible.
Quantum teleportation protocols. Teleportation requires measuring the Bell state of two qubits and applying corrections conditioned on the two-bit outcome. In a circuit that is part of a larger computation, this measurement is a mid-circuit operation.
Adaptive quantum algorithms. Some variational algorithms benefit from intermediate measurements to implement adaptive state preparation, where early measurements guide the choice of later gates. This can reduce the total number of parameters that need to be optimized.
Qubit reuse. In circuits that require more logical operations than available qubits, mid-circuit measurement followed by reset allows a qubit to serve multiple roles within a single circuit execution, effectively trading depth for width.
Dynamic circuits and circuit models
The ability to condition gates on mid-circuit measurement outcomes extends the expressiveness of quantum circuits beyond the standard unitary circuit model. The resulting model is sometimes called a classically controlled quantum circuit or dynamic circuit. It is strictly more powerful than the unitary-only model for certain tasks (like teleportation-based gate synthesis) and enables the measurement-based quantum computing model in which entangled resource states are consumed by a sequence of adaptive single-qubit measurements.