Working with measurement-based computations on qudits

Abstract: Measurement-based quantum computing is a universal model of quantum computation in which successive product measurements of an entangled resource state drive the computation. The non-deterministic nature of measurements necessitates adaptivity to ensure an overall deterministic computation. Flow structures characterise cases in which such an adaptive correction procedure is possible. Recently, flow has been defined in a setting where the resource states are prime-dimensional qudit graph states rather than the usual qubit graph states. Yet, this qudit flow definition is more burdensome to work with than analogous definitions for qubits. Here, we give a simpler definition of qudit flow and consider various useful properties of this flow, drawing on results for the qubit case. In particular, we show how to focus qudit flow and argue that focused flow is canonical. We improve the previous algebraic formulation to capture focused flow and use it to obtain an $O(n^3)$ flow-finding algorithm (where $n$ is the number of qudits), matching the best known complexity for qubit flows and improving on the previous $O(n^4)$ result for qudits. Furthermore, we explore multiple flow-preserving transformations, thus opening a pathway to using flow for optimisation. These transformations include pivoting, removal and insertion of certain types of vertices, and reversibility of flow. Lastly, we propose an algorithmic approach to generating large qudit computations with flow, for testing or machine learning.

• The paper presents a novel definition of qudit flow, simplifying MBQC by establishing a canonical focused flow structure for error correction.
• It introduces an efficient O(n^3) algorithm that leverages matrix algebra for deterministic MBQC on prime-dimensional qudits.
• It generalizes flow-preserving rewrites, facilitating scalable quantum circuit optimization and benchmarking on multilevel systems.

Measurement-Based Quantum Computation on Qudits: Focused Flow Structures and Algorithms

Introduction

Measurement-based quantum computation (MBQC) utilizes sequential adaptive measurements on entangled resource states (typically graph states) as an alternative and universal model to the traditional circuit-based paradigm. While most MBQC research has focused on qubits ($d=2$), the extension to qudits (systems with $d>2$ where $d$ is prime) is both of theoretical and practical importance, given emerging physical implementations where multilevel systems are more natural. One of the critical requirements for MBQC is deterministic operation despite the inherent randomness of quantum measurement. This is captured through the flow (or generalized flow/gflow) structure, which determines if, and how, a correction procedure can be designed to render the computation deterministic.

The paper "Working with measurement-based computations on qudits" (2606.30525) makes substantial progress in systematizing, simplifying, and algorithmically advancing the theory of flow for prime-dimensional qudit MBQC. It provides a new, more tractable definition of qudit flow, introduces the notion of focused flow as canonical, establishes an efficient $O(n^3)$ flow-finding algorithm, and generalizes essential flow-preserving rewrite rules, thus creating a pathway for scalable MBQC optimization and benchmarking on qudits.

New Definition of Qudit Flow and Its Properties

The standard definition of flow (and its qudit generalization as $F_d$-flow) encodes a strategy for measurement adaptivity and error correction using a correction matrix $C$ and a partial order $\prec$ on the measurement sequence. Existing definitions for qudit flow, as introduced by Booth et al. [boothOutcomeDeterminismMeasurementbased2023], were more cumbersome than their qubit counterparts, largely due to additional conditions involving arbitrary totalizations of the underlying partial order.

This paper refines the definition by focusing on an algebraic perspective: a qudit MBQC pattern has $F_d$-flow if and only if there exists a correction matrix $C$ and strict partial order $\prec$ such that (i) for every measured vertex the correction is a valid Pauli operator dependent on measurement labels and graph topology, (ii) the correction matrix is zero for rows indexed by inputs and columns indexed by outputs, and (iii) $d>2$0 and $d>2$1 are lower triangular with respect to any totalization of $d>2$2. This characterization significantly expedites practical verification and structural analysis, eliminating the overhead of working with all possible totalizations directly.

Focused Flow: Canonical Structure

Borrowing from prior results in the qubit context (specifically focused gflow [backensThereBackAgain2021]), the notion of focused flow in the qudit setting restricts the correction by-products so that $d>2$3-type measurements are associated only with $d>2$4 corrections and $d>2$5 or $d>2$6-type with $d>2$7 corrections. The authors rigorously prove that whenever $d>2$8-flow exists, a focused flow also exists, and, when $d>2$9, the focused flow is unique up to refinements of $d$0. This renders focused flow a canonical representative and simplifies downstream algorithmic and structural reasoning.

Algebraic Formulation and Algorithmic Advances

The main technical innovation lies in expressing the focused flow conditions purely in terms of matrix algebra. The paper introduces the flow-demand ($d$1) and order-demand ($d$2) matrices—constructed from the resource graph's adjacency matrix and measurement labels. The existence of focused flow is equivalent to the existence of a correction matrix $d$3 such that $d$4 is the identity and $d$5 describes a directed acyclic graph (DAG).

This algebraic formulation makes the flow-finding problem amenable to polynomial-time algorithms relying on well-developed finite field linear algebra. The authors present an explicit $d$6 algorithm for flow-finding, matching the best-known qubit result and improving substantially on the prior $d$7 for qudits, while detailing its implementation over arbitrary finite fields. The paper also shows that the kernel of the flow-demand matrix determines the space of "do-nothing" stabilizers via focused vectors, generalizing known results for qubits.

Flow-Preserving Rewriting and Transformations

The flexibility and modularity of MBQC patterns are critical for optimization and resource management. The paper pursues the generalization of flow-preserving graph transformations from the qubit regime—local scaling, local complementation, and pivoting—to qudit systems, leveraging their connections to local Clifford operations. The transformations are accompanied by precise update rules for the measurement labels and corrections, ensuring that determinism and flow properties are maintained.

The reversibility of flow is rigorously established for the $d$8 case, facilitating transformations between input/output assignments and improving the versatility of MBQC pattern manipulation. Furthermore, the work develops procedures for efficient $d$9-like vertex deletions and insertions, allowing for dynamic circuit resizing and resource-efficient circuit synthesis while preserving focused flow whenever possible.

Implications for Benchmarking, Optimization, and Machine Learning

The algebraic and algorithmic results have immediate ramifications for both theory and practical deployment. The efficient flow-finding algorithm enables MBQC compilers (such as Graphix and t|ket⟩) to handle large qudit graphs. Flow-preserving rewrites are extrapolated to support optimization strategies analogous to those in the qubit case (e.g., depth reduction, T-count minimization, and ancilla management), with potential applications in hardware-aware transpilation and quantum resource estimation.

The canonical nature of focused flow in the $O(n^3)$0 regime ensures that pattern generation and benchmarking (including for variational algorithms and machine learning approaches employing MBQC architectures [ewenApplication2025]) can utilize randomly constructed instances with guaranteed existence of flow, broadening the diversity and generality of MBQC testbeds.

The algebraic framework also sets the stage for further theoretical exploration, such as possible extensions to continuous-variable systems or systems with non-prime $O(n^3)$1, and pursuing advances akin to Pauli flow and causal flow for qudits. Moreover, the possibilities for formal flow-preserving rewriting system completeness (as recently established in the qubit case [backensCompleteness2026]) now become tractable for qudits as well.

Conclusion

This paper (2606.30525) delivers a comprehensive, technically robust treatment of flow structures in measurement-based quantum computation on prime-dimensional qudit systems. By providing a simplified, canonical algebraic foundation for flow, along with an efficient algorithm and an array of flow-preserving transformation rules, it removes significant practical and theoretical barriers to the adoption of MBQC beyond qubits. The implications are broad—encompassing scalable quantum circuit optimization, advances in quantum software tooling, and the development of a unified language for qudit MBQC for future benchmarking, resource estimation, and machine learning paradigms. Further research directions include the full extension of causal/Pauli flow characterizations to general $O(n^3)$2, the comprehensive structural analysis of MBQC as diagrammatic quantum computation, and generalizing these insights to hybrid and continuous-variable architectures.