Pauli Semiwebs in ZX-Calculus

• Pauli semiwebs are ZX-calculus constructions that decorate wires with Pauli operators subject to local compatibility constraints, allowing controlled phase shifts that create defects.
• They generalize Pauli webs by relaxing strict eigenstate conditions, thereby enabling a ZX-native flow (ZX-flow) for deterministic computation on arbitrary ZX-diagrams.
• Their algebraic structure, decomposable into basic and edge semiwebs, ensures stability under ZX-calculus rewrites and supports fault-tolerant, non-Clifford operations.

Searching arXiv for papers on Pauli semiwebs and closely related ZX-calculus/Pauli web work. Pauli semiwebs are a ZX-calculus construction that decorates the wires of a ZX-diagram with Pauli operators subject to local compatibility constraints, while allowing controlled local phase shifts at spiders. They were introduced as a generalisation of Pauli webs in order to formulate a genuinely ZX-native flow criterion, ZX-flow, for deterministic computation on arbitrary ZX-diagrams (Kissinger et al., 10 Mar 2026). In this framework, a Pauli semiweb is not required to stabilise every local spider state exactly; instead, at certain spiders it may induce a nonzero phase shift, creating a “defect.” This relaxation is the central distinction from Pauli webs and is what makes semiwebs applicable to non-Clifford diagrams, where ordinary Pauli-web conditions are too restrictive (Kissinger et al., 10 Mar 2026). The notion also sits in continuity with earlier Pauli-web reasoning for fault-tolerant surface-code computations, where coloured Pauli decorations were used to track logical correlators through Clifford ZX-diagrams (Wan et al., 2 Feb 2025).

1. Definition and basic intuition

In the formalism of "ZX-Flow: A Flexible Criterion for Deterministic Computation with ZX-Diagrams" (Kissinger et al., 10 Mar 2026), a ZX-diagram $D$ has a set of nodes $N$ and wires $W$. For each spider $\nu$, the paper associates a local state $|\nu\rangle$, and also a twisted local state $|\nu_\alpha\rangle$, defined so that twisting changes the spider phase by $\alpha$, while having no effect on H-gates. A Pauli operator on all wires is written $\vec w \in \mathcal P_{|W|}$, with $\vec w_\nu$ denoting its restriction to the wires adjacent to $\nu$ (Kissinger et al., 10 Mar 2026).

The paper defines a Pauli semiweb by the condition that, for every node $N$0, there exists an angle $N$1 such that

$N$2

If $N$3, then $N$4 is called an $N$5-defect, or simply a defect, of the semiweb (Kissinger et al., 10 Mar 2026). Thus a semiweb is a global Pauli decoration that preserves the local spider form but may shift its phase.

This construction is combinatorial as well as algebraic. The same paper proves that a Pauli operator is a semiweb if and only if it satisfies the H and all-or-nothing conditions inherited from the Pauli-web formalism (Kissinger et al., 10 Mar 2026). The H condition constrains the colouring around H-gates to the allowed neighbourhoods $N$6, while the all-or-nothing condition requires that, for each spider, support of the opposite Pauli type is either absent on all adjacent wires or present on all adjacent wires (Kissinger et al., 10 Mar 2026).

A plausible implication is that semiwebs should be understood as the minimal extension of Pauli webs that preserves the local propagation rules of Pauli structure through ZX syntax, while dropping only those constraints that fail at non-Clifford phases.

2. Relation to Pauli webs and defects

The relevant antecedent is the Pauli web. In (Kissinger et al., 10 Mar 2026), a Pauli web is a Pauli operator $N$7 such that each local state $N$8 is an eigenstate of $N$9. The paper characterises Pauli webs by four conditions: H, all-or-nothing, parity, and Clifford (Kissinger et al., 10 Mar 2026). Semiwebs retain only the first two.

This difference is structural. Every Pauli web is a Pauli semiweb with no defects, but a semiweb may violate the parity and Clifford conditions, and those violations appear precisely as defects with specific twist angles (Kissinger et al., 10 Mar 2026). The paper states that, at a spider with phase $W$0, violation of the Clifford condition gives a $W$1-defect, violation of the parity condition gives a $W$2-defect, and violation of both gives a $W$3-defect (Kissinger et al., 10 Mar 2026).

This makes the web/semiweb distinction exact rather than heuristic. Webs encode genuine local Pauli eigenstructure; semiwebs encode Pauli action together with permissible angle shifts. That feature is essential for non-Clifford reasoning, because opposite-type Pauli support at a non-Clifford spider generally does not stabilise the spider state. Instead, it changes the phase, and semiwebs record that change as a defect (Kissinger et al., 10 Mar 2026).

The earlier paper "Pauli webs spun by transversal $W$4 state initialisation" (Wan et al., 2 Feb 2025) uses Pauli webs, not semiwebs, to analyse the CCLP fold-transversal $W$5 gate on the rotated surface code. There, a logical $W$6 correlator is shown graphically to become a logical $W$7 correlator under the circuit, with red and green web components overlapping at the output (Wan et al., 2 Feb 2025). That work explicitly does not define “Pauli semiweb.” It instead suggests, by inference, that a semiweb could be viewed as a partial or restricted Pauli web, for example one confined to a logical sector, a subset of time slices, or a single Pauli type (Wan et al., 2 Feb 2025). This suggests a terminological continuity: the later formal notion in (Kissinger et al., 10 Mar 2026) generalises the earlier web language by relaxing local eigenstate conditions, while the earlier paper motivates semiweb-like partial reasoning in fault-tolerant diagrams without naming it as such.

3. Generation, decomposition, and algebraic structure

Pauli semiwebs in (Kissinger et al., 10 Mar 2026) have a concrete algebraic structure. The paper proves that the product of Pauli semiwebs is again a Pauli semiweb (Kissinger et al., 10 Mar 2026). This closure property is central to later focusing and correction constructions.

Two generating families are defined. A basic semiweb $W$8 for a spider $W$9 is the smallest semiweb having support of the opposite colour on a wire adjacent to $\nu$0. An edge semiweb is a semiweb supported on a single plain edge with $\nu$1 or $\nu$2, or on a single H-edge with $\nu$3 (Kissinger et al., 10 Mar 2026). The paper then proves that any Pauli semiweb can be written as

$\nu$4

where $\nu$5 is a product of edge semiwebs and $\nu$6 is a generating set of spiders (Kissinger et al., 10 Mar 2026).

For graph-like diagrams, the generating set $\nu$7 is unique (Kissinger et al., 10 Mar 2026). This uniqueness is important because it permits direct comparison with Pauli-flow correction sets, where correction data are also specified by vertex subsets.

The paper’s examples indicate that a basic semiweb colours all edges in a connected fusable component and induces $\nu$8-defects on boundary spiders, while edge semiwebs act as minimal local decorations that may create defects at the two incident spiders (Kissinger et al., 10 Mar 2026). This suggests a layered interpretation: basic semiwebs provide the large-scale propagation skeleton, while edge semiwebs account for local adjustments.

Object	Definition in (Kissinger et al., 10 Mar 2026)	Role
Basic semiweb $\nu$9	Smallest semiweb with opposite-type support adjacent to $|\nu\rangle$0	Generates nonlocal propagation from a spider
Edge semiweb	Semiweb on a single plain edge or single H-edge	Local correction / local support
General semiweb	$|\nu\rangle$1	Arbitrary semiweb decomposition

4. ZX-flow and deterministic computation

The main application of Pauli semiwebs is ZX-flow, introduced in (Kissinger et al., 10 Mar 2026) as a flow criterion for arbitrary ZX-diagrams. A diagram has ZX-flow if there exists a partial order on the non-Clifford spiders together with two types of semiwebs: logical semiwebs $|\nu\rangle$2 for each input wire, and flow semiwebs $|\nu\rangle$3 for each non-Clifford spider $|\nu\rangle$4 (Kissinger et al., 10 Mar 2026).

The logical semiwebs restrict on the inputs to $|\nu\rangle$5 and $|\nu\rangle$6, and may have defects only at non-Clifford spiders (Kissinger et al., 10 Mar 2026). Each flow semiweb $|\nu\rangle$7 must have a $|\nu\rangle$8-defect at $|\nu\rangle$9, while all its other defects occur only at non-Clifford spiders $|\nu_\alpha\rangle$0 satisfying $|\nu_\alpha\rangle$1 (Kissinger et al., 10 Mar 2026). This implements a causal feed-forward condition directly at the level of semiweb defects.

The same paper also defines strong ZX-flow, in which every spider, not only every non-Clifford spider, is assigned a flow semiweb (Kissinger et al., 10 Mar 2026). It further defines focused ZX-flow, where logical semiwebs satisfy the parity condition everywhere and each flow semiweb satisfies parity everywhere except possibly at its own spider (Kissinger et al., 10 Mar 2026). The paper proves that ZX-flow exists if and only if focused ZX-flow exists, and that strong ZX-flow implies ZX-flow (Kissinger et al., 10 Mar 2026).

Operationally, a flow semiweb encodes how an undesired measurement branch can be corrected. Its $|\nu_\alpha\rangle$2-defect at $|\nu_\alpha\rangle$3 flips the phase at that spider, while its future defects indicate which later non-Clifford measurement angles must be adapted (Kissinger et al., 10 Mar 2026). This is the direct analogue of feed-forward in MBQC, but stated in a ZX-native language rather than on graph states.

A plausible implication is that semiweb defects play the role of an angle-domain correction calculus: rather than merely specifying where Pauli operators propagate, they specify where propagation induces compensating changes in future measurement parameters.

5. Equivalence with Pauli flow and preservation under ZX rewrites

One of the main results of (Kissinger et al., 10 Mar 2026) is that semiweb-based ZX-flow is not merely analogous to earlier flow notions but equivalent to them on graph-like diagrams. The paper proves that a graph-like ZX-diagram has strong ZX-flow if and only if its induced labelled open graph has Pauli flow (Kissinger et al., 10 Mar 2026). In this equivalence, the generating set of a flow semiweb corresponds exactly to the Pauli-flow correction set.

The same paper then proves a broader characterisation: a ZX-diagram has ZX-flow if and only if it is Clifford-equivalent to a graph-like ZX-diagram with Pauli flow (Kissinger et al., 10 Mar 2026). This theorem is the formal basis for the claim that ZX-flow is a ZX-native generalisation of graph-state flow criteria.

Preservation under rewrite is another central feature. The paper proves that the rules of the extended Clifford ZX-calculus preserve ZX-flow (Kissinger et al., 10 Mar 2026). It also proves that spider fusion between a pair of non-Clifford spiders preserves ZX-flow from left to right, and preserves it from right to left provided one does not unfuse a Clifford spider into a pair of non-Clifford spiders (Kissinger et al., 10 Mar 2026).

These preservation theorems are significant because older flow criteria such as causal flow, gflow, and Pauli flow were formulated for graph-state-like forms and are fragile under basic ZX rewrites. By contrast, ZX-flow is designed to survive general Clifford rewrites precisely because semiwebs are defined directly on arbitrary ZX-diagrams (Kissinger et al., 10 Mar 2026). This suggests that semiwebs function as rewrite-stable Pauli annotations.

The relation to later or parallel ZX work is also explicit. The paper frames Pauli semiwebs as a generalisation of Pauli webs, which had already been used in reasoning about fault-tolerant computations in the ZX-calculus (Kissinger et al., 10 Mar 2026). The fault-tolerant surface-code analysis in (Wan et al., 2 Feb 2025) is a concrete example of that earlier web-based methodology.

6. Computational interpretation, extraction, and broader context

A diagram with focused ZX-flow admits two computational interpretations in (Kissinger et al., 10 Mar 2026). First, it can be read as a deterministic MBQC pattern. In that view, non-Clifford spiders are measurements, and the flow semiweb of $|\nu_\alpha\rangle$4 specifies how to correct an undesired measurement outcome by acting with a global Pauli and updating the angles of future measurements according to the defect pattern (Kissinger et al., 10 Mar 2026).

Second, the paper proves that any diagram with focused ZX-flow can be rewritten as a Clifford isometry followed by a sequence of Pauli exponentials (Kissinger et al., 10 Mar 2026). The extraction theorem states that, after ordering the non-Clifford spiders compatibly with the partial order, the diagram decomposes into a Clifford isometry $|\nu_\alpha\rangle$5 and a product of Pauli exponentials determined by the output colourings of the corresponding flow semiwebs (Kissinger et al., 10 Mar 2026). The logical semiwebs determine the logical $|\nu_\alpha\rangle$6 and $|\nu_\alpha\rangle$7 operators of the Clifford isometry (Kissinger et al., 10 Mar 2026).

This extraction result positions semiwebs as an interface between ZX-diagram rewriting, MBQC determinism, and circuit synthesis. The same paper suggests that semiweb constraints are linear over $|\nu_\alpha\rangle$8, which indicates the possibility of algorithmic ZX-flow finding, though the paper presents this as a future direction rather than a completed algorithmic theory (Kissinger et al., 10 Mar 2026).

The broader literature supplied here uses “Pauli” language in several other ways, but those are conceptually distinct. "Transforming Collections of Pauli Operators into Equivalent Collections of Pauli Operators over Minimal Registers" (2206.13040) studies collections of Pauli operators through binary symplectic representation, commutation matrices, and the minimal number of qubits needed to realise a given commutation pattern. That paper does not define Pauli semiwebs, but it does describe a commutation-pattern object that can be interpreted informally as a “Pauli semiweb,” encoded by the commutation matrix $|\nu_\alpha\rangle$9, with intrinsic minimal size

$\alpha$0

for $\alpha$1 and $\alpha$2 (2206.13040). This suggests an algebraic analogy: both semiwebs in ZX-flow and minimal-register Pauli collections isolate the invariant structure of Pauli interaction independently of a particular embedding.

By contrast, "Pauli operators and the d-bar-Neumann problem" (Haslinger, 2017) concerns magnetic Schrödinger/Pauli operators $\alpha$3 and their unitary equivalence to weighted $\alpha$4-Neumann Laplacians (Haslinger, 2017), while "About the Dedekind psi function in Pauli graphs" (Planat, 2010) studies finite-geometric commutation graphs of Pauli observables and counts maximal commuting sets using $\alpha$5 and $\alpha$6 functions (Planat, 2010). Neither paper defines Pauli semiwebs in the ZX sense. Their inclusion is therefore useful mainly for disambiguation: “Pauli semiwebs” is a specific ZX-calculus notion introduced in (Kissinger et al., 10 Mar 2026), not a generic term for Pauli-operator networks across all subfields.

A common misconception is that Pauli semiwebs are merely incomplete Pauli webs. That interpretation appears as an inference in (Wan et al., 2 Feb 2025), where “semiweb” is not a formal term and could naturally denote a partial web over selected qubits, time slices, or Pauli types. In the formal language of (Kissinger et al., 10 Mar 2026), however, a semiweb is not defined by incompleteness but by the weakened local condition that allows twisted local states and hence defects. The difference is substantive: the concept was introduced precisely to accommodate non-Clifford behaviour while preserving enough structure to define flow and support extraction.

In this sense, Pauli semiwebs occupy a specific place in the modern ZX-calculus toolkit. They extend the Pauli-web method from Clifford stabilisation to defect-tolerant phase propagation, provide the native data structure for ZX-flow, and connect arbitrary ZX-diagrams to deterministic MBQC and Clifford-plus-Pauli-exponential circuit forms (Kissinger et al., 10 Mar 2026).