A graphical approach to measurement-based quantum computing

Quantum computations are easily represented in the graphical notation known as the ZX-calculus, a.k.a. the red-green calculus. We demonstrate its use in reasoning about measurement-based quantum computing, where the graphical syntax directly captures the structure of the entangled states used to represent computations, and show that the notion of information flow within the entangled states gives rise to rewriting strategies for proving the correctness of quantum programs.

💡 Research Summary

The paper presents a novel, graph‑theoretic methodology for reasoning about measurement‑based quantum computing (MBQC) by exploiting the ZX‑calculus, also known as the red‑green calculus. The ZX‑calculus is a categorical diagrammatic language whose basic generators are Z‑spiders (green nodes) and X‑spiders (red nodes), together with Hadamard edges that change colour. These generators correspond to multi‑input, multi‑output linear maps with an arbitrary phase parameter, and a small set of rewrite rules—spider fusion, colour change, and bialgebraic interaction—captures the full algebra of quantum circuits. Because the rewrite rules are sound and complete for stabiliser quantum mechanics, any diagrammatic transformation preserves the underlying quantum operation.

The authors first recast the standard MBQC model in this language. In MBQC a large entangled resource state—typically a 2‑dimensional cluster state—is prepared, and computation proceeds by measuring individual qubits in bases that may depend on previous measurement outcomes. The resource state can be expressed as a ZX‑diagram: each lattice vertex is a Z‑spider, each edge a Hadamard (colour‑changing) wire, and the whole graph encodes the stabiliser structure of the cluster. Measurements are introduced by attaching an X‑spider (or a Z‑spider, depending on the basis) with a phase equal to the measurement angle. Classical feed‑forward, which in the circuit model appears as conditional Pauli corrections, becomes a simple phase‑adjustment on downstream spiders.

A central contribution of the paper is the identification of the MBQC “flow” (and its generalisation, gflow) with directed paths in the ZX‑diagram. The flow determines a causal order for eliminating measured spiders while propagating their phases forward. The authors define two systematic rewrite strategies:

1. Flow‑based reduction – Starting from the measured spider, fuse it with its neighbour according to the flow, apply spider fusion, and use colour‑change rules to push Hadamards past the remaining structure. The phase of the measured spider is transferred to the target spider, reproducing the classical feed‑forward. Repeating this along the flow eliminates all measured nodes, leaving only the input and output spiders. The resulting diagram is exactly the unitary circuit that the MBQC protocol implements.

2. gflow‑based reduction – When a strict flow does not exist, the authors introduce auxiliary spiders and additional Hadamard edges to create a temporary flow. By applying the same fusion and colour‑change steps, the auxiliary structure can be eliminated, yielding the same final circuit. This demonstrates that the gflow condition, previously defined algebraically, has a natural graphical realisation.

The paper illustrates these strategies with three canonical examples. First, a one‑dimensional linear cluster is used to teleport an arbitrary qubit state; the diagrammatic reduction reproduces the standard teleportation circuit. Second, a two‑dimensional lattice implements a CNOT gate; the flow‑based rewrite shows how the entanglement pattern and measurement angles give rise to the familiar control‑target interaction. Third, arbitrary single‑qubit rotations are achieved by choosing appropriate measurement angles; the gflow‑based reduction confirms that any unitary in SU(2) can be realised.

Beyond manual derivations, the authors discuss the compatibility of their approach with automated diagram‑rewriting tools such as Quantomatic. Because the rewrite system is finite and confluent for stabiliser fragments, large‑scale MBQC protocols can be verified automatically, a task that is cumbersome with matrix‑based proofs. The graphical perspective also clarifies optimisation opportunities: redundant spiders can be fused early, and unnecessary Hadamards can be eliminated, leading to more resource‑efficient implementations.

In the discussion, the authors acknowledge current limitations. While the ZX‑calculus handles stabiliser‑type MBQC elegantly, extending the framework to non‑Clifford resources (e.g., magic‑state injection) requires additional generators or enriched rewrite rules. Moreover, the complexity of gflow‑based reductions grows with lattice size, suggesting the need for heuristic algorithms to choose optimal auxiliary structures.

The conclusion emphasises that representing MBQC entirely within the ZX‑calculus provides a unified, visual language that captures both the entanglement structure of the resource state and the classical control flow of measurements. By translating flow and gflow into concrete rewrite strategies, the paper offers a systematic, provably correct method for verifying quantum programs, paving the way for more transparent algorithm design, automated verification, and potentially new optimisation techniques in measurement‑based quantum computing.