Simultaneous Measurement of Commuting Pauli Operators

23 concepts17 papers1 proposal

In variational quantum algorithms (VQAs) like the Variational Quantum Eigensolver (VQE), calculating the expectation value of a target Hamiltonian is a primary computational bottleneck. A typical molecular or physical Hamiltonian is expressed as a linear combination of thousands or millions of individual Pauli strings:

$$H = \sum_{i} c_i P_i$$

where each $P_i$ is a tensor product of Pauli operators ($X, Y, Z, I$) acting on individual qubits. Because quantum mechanics only yields probabilistic outcomes upon measurement, estimating the expectation value $\langle H \rangle$ to chemical accuracy requires running the underlying quantum circuit and sampling outcomes millions or even billions of times. As noted by Marco Majland et al. in their work on vibrational structure optimization, this massive sampling requirement creates a severe measurement problem that limits the practicality of near-term quantum applications.

To mitigate this overhead, researchers exploit a fundamental algebraic property of quantum mechanics: commutativity. If two Pauli operators commute ($[P_i, P_j] = 0$), they share a common eigenbasis, meaning they can be measured simultaneously from the exact same copy of a quantum state. By partitioning a large Hamiltonian into groups of mutually commuting operators, the total number of independent circuit evaluations ("shots") can be drastically reduced.

The Intuition Behind Commutativity Partitioning

To understand why grouping works, consider measuring two simple two-qubit operators: $Z \otimes I$ and $I \otimes Z$. Because they act on different qubits, they commute. If we prepare a quantum state and measure both qubits in the standard computational ($Z$) basis, a single shot yields a bitstring (e.g., 01). This single bitstring simultaneously provides an eigenvalue sample for $Z \otimes I$ (from the first bit) and $I \otimes Z$ (from the second bit).

Early foundational work by Jonathan Romero et al. in The theory of variational hybrid quantum-classical algorithms demonstrated that grouping commuting operators for correlated sampling significantly lowers the total quantum resource cost. Instead of executing separate runs for every single term in a Hamiltonian, terms are batched together, allowing a single circuit execution to serve multiple estimators at once.

In practice, commutativity grouping generally falls into two primary paradigms, each presenting a distinct trade-off between classical grouping efficiency and quantum circuit complexity.

Qubit-Wise Commutativity (QWC)

Two Pauli strings are qubit-wise commuting if, for every single qubit index, the corresponding Pauli factors commute. Since identical Paulis commute with themselves and any Pauli commutes with the identity ($I$), strings like $X \otimes Y$ and $X \otimes I$ are qubit-wise commuting.

The core advantage of QWC is that it requires no extra entangling gates. To measure a QWC group, you simply apply single-qubit rotations at the end of the circuit to align each qubit to the correct measurement basis (e.g., a Hadamard gate to measure in the $X$ basis). Marco Majland et al. demonstrated that in vibrational quantum chemistry, simpler QWC schemes often achieve performance parity with more complex methods without introducing high-depth basis-change circuits that could succumb to hardware noise.

Full Commutativity (FC)

Two Pauli strings can commute globally even if they do not commute qubit-wise. For example, $X \otimes X$ and $Z \otimes Z$ commute because the two sign flips introduced by their non-commuting local components cancel each other out ($(-1) \times (-1) = 1$).

Full Commutativity allows for much larger groups, which theoretically minimizes the number of distinct measurement setups. However, measuring an FC group requires mapping its joint eigenstates to the standard computational basis. This requires appending an additional Clifford circuit (containing entangling gates like CNOTs) to the end of the VQE ansatz. On noisy intermediate-scale quantum (NISQ) devices, these extra gates introduce physical errors that can degrade the accuracy of the final estimate.

Advanced Grouping Frameworks and Hybrid Strategies

To navigate the rigid trade-offs between QWC and FC, recent research focuses on adaptive and unified frameworks that optimize measurement allocation based on hardware constraints and statistical variance.

                  ┌──────────────────────────────┐
                  │    Hamiltonian Pauli Terms   │
                  └──────────────┬───────────────┘
                                 │
                  ┌──────────────┴───────────────┐
                  │   Commutativity Partition    │
                  └──────────────┬───────────────┘
                                 │
         ┌───────────────────────┼───────────────────────┐
         ▼                       ▼                       ▼
┌─────────────────┐    ┌──────────────────┐    ┌──────────────────┐
│ Qubit-Wise (QWC)│    │ Full Commuting   │    │  Hybrid (GALIC)  │
│  • Zero extra   │    │  • Max grouping  │    │  • Hardware-     │
│    gates        │    │  • Requires      │    │    aware         │
│  • More groups  │    │    Clifford gates│    │  • Balanced risk │
└─────────────────┘    └──────────────────┘    └──────────────────┘

Hardware-Aware and Overlapped Grouping

Rather than treating grouping as a purely mathematical graph-coloring problem, modern techniques incorporate hardware topology and noise. Chenxu Liu et al. introduced GALIC, a hybrid multi-qubitwise Pauli grouping strategy that balances the trade-offs between FC and QWC. By adapting to specific hardware connectivity and noise profiles, it achieves lower estimator variance than pure QWC while avoiding the deep, error-prone circuits demanded by full FC.

Simultaneously, Bujiao Wu et al. proposed Overlapped Grouping Measurement (OGM), a unified framework that optimizes observable compatibility. OGM allows Pauli strings to belong to multiple overlapping measurement groups, maximizing parallel efficiency and significantly reducing the total number of shots required to estimate complex Hamiltonians.

Active Learning and Importance Sampling

Beyond structural grouping, efficiency can be boosted by dynamically allocating shots to the groups that matter most. Ruidi Zhu et al. developed an active learning framework that adaptively assigns measurement repetitions to commuting groups based on real-time uncertainty reduction. Instead of sampling all groups uniformly, the algorithm prioritizes groups that contribute the highest variance to the total energy estimate.

In a similar vein, classical post-processing techniques like importance sampling can filter out insignificant terms. Youle Wang et al. demonstrated in their entropy estimation algorithms that evaluating only the most significant mathematical terms via importance sampling drastically cuts down the required quantum circuit evaluations.

Broad Applications Across Quantum Computing

The benefits of commuting observable partitioning extend far beyond standard molecular energy estimation in VQE. It has become a foundational optimization layer across various quantum computing domains:

  • Quantum Chemistry and Drug Discovery: In real-world benchmarks, such as the ground-state estimation of the hydroxyl cation ($OH^+$) analyzed by Zhiding Liang et al. in the Quantum Computing for Drug Discovery Challenge, grouping commuting terms lowered measurement shots by more than an order of magnitude. This reduction is vital for running accurate simulations within the coherence times of current noisy hardware, a benefit also leveraged by Wanqi Sun et al. to accelerate VQE convergence.
  • Circuit Wire Cutting: When breaking a large quantum circuit into smaller fragments to fit on limited hardware, the number of required subcircuits typically explodes exponentially. To combat this, Xinpeng Li et al. designed an efficient wire-cutting method based on commuting groups. By partitioning the cut Pauli observables into the minimum possible number of commuting groups, they extracted multiple expectation values from a single subcircuit execution, mitigating the exponential shot explosion.
  • Distributed Quantum Computing: For scaling algorithms across multiple quantum processors, Qian Yang et al. developed Shuffle-QUDIO. This framework integrates distributed optimization with commutation-based measurement reduction, allowing grouped observables to execute simultaneously across independent worker nodes, minimizing communication bottlenecks.
  • Partial Differential Equations (PDEs): In the domain of quantum machine learning, Chanyoung Kim et al. utilized direct Pauli measurements and commuting groups in their spectral operator learning framework (NVQLS). By replacing expensive controlled-unitary operations with grouped Pauli measurements, they drastically reduced the circuit evaluations needed for loss estimation.
  • Hamiltonian Learning and Tomography: For characterization tasks, Arkopal Dutt et al. introduced "Pauli hashing" in Testing and Learning Structured Quantum Hamiltonians, using commutation relations to enable memoryless sparsity testing. For state reconstruction, alternative measurement structures like product-structured positive operator-valued measures (POVMs) developed by Yong Wang et al. similarly reduce the storage and computational overhead inherent to quantum state tomography.

Ultimately, whether applied to optimizing parameter landscapes via meta-learning as explored by Junyong Lee et al., or learning system dynamics via Koopman operators as shown by Di Luo et al., minimizing the reliance on extensive quantum circuit evaluations remains essential. Commutativity-based grouping provides a mathematically rigorous, hardware-compatible mechanism that directly addresses the measurement bottleneck, making near-term quantum algorithms viable.

Key Foundational Work

The theory of variational hybrid quantum-classical algorithms (2016) — contributed 3 concept elements to this direction.
Generation of accessible sets for a class of quantum spin networks (2019) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_measurement_overhead_reduction_e30d9c08,
  title = {measurement overhead reduction},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/measurement-overhead-reduction/rdr_e30d9c08325d4caab9bd1ae7f70281f7},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_measurement_overhead_reduction_e30d9c08_v1,
  title = {measurement overhead reduction},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/measurement-overhead-reduction/rdr_e30d9c08325d4caab9bd1ae7f70281f7/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake