Measurement Circuits

Definition

Quantum measurement is the process by which quantum information is extracted as classical data. Unlike classical measurement, quantum measurement is fundamentally probabilistic and destructive: measuring a qubit in the computational basis collapses its state to either $|0\rangle$ or $|1\rangle$, with probabilities determined by the state amplitudes. Measurement circuits are the physical and logical constructs that perform this operation while minimising readout error and preserving the ability to perform conditional (feedforward) operations.

Projective Measurement in the Computational Basis

The standard measurement projects an $n$-qubit state $|\psi\rangle$ onto the $Z$-basis (computational basis):

$\displaystyle |\psi\rangle = \sum_{i=0}^{2^n-1} \alpha_i |i\rangle \quad\longrightarrow\quad \text{outcome } i \text{ with probability } |\alpha_i|^2$

A measurement circuit consists of:

1. Qubit readout resonator (superconducting) or fluorescence detection (trapped ions) or single-photon detector (photonic).
2. Signal chain: Amplification (Josephson parametric amplifier, HEMT), filtering, digitisation (ADC).
3. Discrimination: Classifying the analog signal as $|0\rangle$ or $|1\rangle$ — typically using a matched filter or machine-learning classifier.

Measurement Fidelity

Measurement fidelity $F_m$ is the probability that the reported outcome matches the true state:

$\displaystyle F_m = 1 - \varepsilon_{\text{ro}}$

where $\varepsilon_{\text{ro}}$ is the readout error rate. State-of-the-art readout fidelities:

Readout error is asymmetric: $P(0|1)$ (measuring 0 when the qubit is 1) and $P(1|0)$ differ due to qubit relaxation during measurement. Calibration provides a $2\times2$ assignment matrix that corrects for this bias.

Measurement Bases

Beyond the $Z$ basis, measurements can be performed in any single-qubit basis:

• $X$ basis: Measure $\sigma_x$ eigenvalues $\{\ket{+}, \ket{-}\}$. Achieved by appending a Hadamard gate before $Z$ measurement.
• $Y$ basis: Measure $\sigma_y$ eigenvalues. Appending $S^\dagger H$ before $Z$ measurement rotates $Y$ eigenstates to the $Z$ basis.

For multi-qubit measurements, one can measure Pauli strings (tensor products of Pauli operators) by applying the appropriate basis rotation to each qubit and measuring in $Z$.

Tomographically Complete Measurement Sets

Full state tomography requires measuring in a set of bases that span the space of density matrices. For an $n$-qubit system, this requires $3^n$ measurement settings (one per qubit per Pauli basis). Adaptive tomography reduces this to $O(2^n)$ by using measurement outcomes to select subsequent settings.

Mid-Circuit Measurement and Conditional Operations

Mid-circuit measurement reads out a qubit during algorithm execution without collapsing the remaining qubits. This enables:

• Quantum teleportation: Measure two qubits of a three-qubit entangled state; send classical outcomes to perform corrective gates.
• Error syndrome extraction: In quantum error correction, ancilla qubits are measured mid-circuit to detect errors without disturbing the data qubits.
• Feedforward: Classical results from early measurements determine later gate operations. This is essential for fault-tolerant protocols and adaptive circuits.

Hardware Requirements

Mid-circuit measurement demands:

• Fast, high-fidelity readout — measurement must complete within the qubit coherence time.
• Reset capability: The measured qubit must be returned to $|0\rangle$ (or reused) without disturbing neighbours. Techniques include active reset (π-pulse conditioned on outcome) or passive reset (letting $T_1$ relaxation return the qubit to ground).
• Crosstalk isolation: Readout pulses on one qubit can leak into neighbouring qubit channels, causing dephasing. Mitigated by frequency multiplexing and readout pulse shaping.

Active Reset Circuit

   ┌───┐┌─┐
q: ┤ H ├┤M├───
   └───┘└╥┘
         ║  ┌───┐
c: 1 ════╩══╡ X ╞ (conditionally flip back to |0⟩)
              └───┘

If the measurement outcome is 1, an $X$ gate is applied to return the qubit to $|0\rangle$.

Readout Error Mitigation

Measurement Error Calibration

Execute known preparation circuits (all-$|0\rangle$, all-$|1\rangle$, separable $|+\rangle$) and build an assignment matrix $A$ where $A_{ij} = P(\text{report } i | \text{prepare } j)$. Invert $A$ to correct raw counts:

$\displaystyle p_{\text{corrected}} = A^{-1} p_{\text{raw}}$

Zero-Noise Extrapolation (ZNE)

Run the same circuit at multiple readout pulse amplitudes or integration times. Extrapolate the measurement outcome to the zero-readout-error limit using a polynomial fit.

Correlated Readout Error

Readout errors are not independent — crosstalk in the multiplexed readout chain creates correlations between qubits. Mitigated by:

• Readout-Error Mitigation (REM) in Qiskit: Builds a full $2^n \times 2^n$ assignment matrix from calibration circuits and inverts it via least-squares or iterative Bayesian unfolding.
• Tensor-product error models: Assume readout errors factorise per qubit (approximate but scalable to many qubits).

Resources and References

• DiVincenzo, "The physical implementation of quantum computation" (2000), Fortschritte der Physik 48, 771–783. [DiVincenzo criteria, including measurement]
• Bultink et al., "Active reset for superconducting qubits" (2020), Phys. Rev. Applied 14, 044050.
• Nation et al., "Scalable readout-error mitigation for near-term quantum computers" (2021), PRX Quantum 2, 040326.
• Quantum Tomography resources: qiskit.org/documentation/apidoc/ignis.verification.html