Spired Graph: Obfuscation & Quantum Walks

• Spired graph is an obfuscated graph family obtained by lifting a d-regular base graph and attaching balanced spires with randomized relabeling.
• Its construction hides local connectivity while preserving key spectral properties in a polynomial-dimensional Krylov subspace, enabling analysis via continuous-time quantum walks.
• The design underpins quantum hidden-graph identification, suggesting an exponential quantum speedup over classical methods in oracle query complexity.

Searching arXiv for recent and directly relevant papers on spired graphs and hidden-graph identification. arXiv search query: "spired graph hidden graph identification welded trees continuous-time quantum walk" A spired graph is an obfuscated graph family built from a simple, connected, $d$-regular base graph $G$ on $n$ vertices by lifting each base vertex into an exponentially large cluster, attaching a balanced spire to each cluster, and then randomly relabelling all vertices (Wocjan, 11 May 2026). In the formulation introduced for hidden-graph identification, the algorithmic task is not to traverse $G$ directly but to recover the hidden base graph from oracle access to the resulting graph $G_{\rm spire}$. The construction is designed so that local connectivity is obscured at the graph level while a continuous-time quantum walk from a designated apex remains confined to a polynomial-dimensional invariant subspace with an explicitly analyzable effective Hamiltonian (Wocjan, 11 May 2026).

1. Construction and basic definition

The construction takes as input a simple, connected, $d$-regular base graph $G=(V,E)$ on $n$ vertices with adjacency matrix $A\in\mathbb{R}^{n\times n}$, parameters $L\ge 1$ and $G$0, and a distinguished vertex $G$1 called the “entry” apex. Writing $G$2, the graph $G$3 is formed in three stages (Wocjan, 11 May 2026).

Stage	Operation	Resulting property
Lifted graph $G$4	Replace each $G$5 by a cluster $G$6, $G$7, and for each edge $G$8 sample a uniformly random $G$9-regular bipartite graph between $n$0 and $n$1	$n$2 and $n$3 is $n$4-regular
Add balanced $n$5-ary spires	Over each cluster attach an inverted perfect $n$6-ary tree of depth $n$7 with apex $n$8 and level sets $n$9	Level $G$0 matches the cluster $G$1
Random relabelling	Apply a random injective labelling $G$2 with $G$3	The graph is accessed only through an obfuscating oracle

The full graph is

$G$4

where each spire $G$5 has edges $G$6–$G$7 for $G$8 (Wocjan, 11 May 2026).

Two size relations are central. First, each cluster has size $G$9. Second, the number of spire vertices attached to a single base vertex is

$G_{\rm spire}$0

This gives the construction its characteristic geometry: each base vertex is replaced by an exponential tree-plus-cluster gadget (Wocjan, 11 May 2026).

A notable special case is that specializing to $G_{\rm spire}$1 recovers the welded-trees graph (Wocjan, 11 May 2026). This places spired graphs in direct continuity with the quantum-walk literature on oracle separations based on highly obfuscated graph structure.

2. Oracle model and obfuscation mechanism

The graph is presented through an oracle $G_{\rm spire}$2 defined on labels rather than on structured vertex names. Given $G_{\rm spire}$3, the oracle returns the multiset of labels of neighbors of the unique $G_{\rm spire}$4 with $G_{\rm spire}$5, or $G_{\rm spire}$6 if $G_{\rm spire}$7 (Wocjan, 11 May 2026). The random injective relabelling removes explicit evidence of the product structure, the cluster decomposition, and the correspondence with the original graph $G_{\rm spire}$8.

The degree pattern is deliberately sparse in informational content. The only hint of structure is that apices have degree $G_{\rm spire}$9, while all other vertices have degree $d$0 (Wocjan, 11 May 2026). This means that local degree inspection reveals the apex set but does not reveal which apex corresponds to which base vertex, nor how the lifted layer encodes the base edges.

The lifted part of the graph is itself randomized. For each original edge $d$1, the associated bipartite block $d$2 is a $d$3 adjacency matrix with exactly $d$4 ones per row and column (Wocjan, 11 May 2026). Thus adjacency between clusters is regular but not canonically aligned. This randomization is the mechanism by which the local structure of the base graph is hidden.

The intended algorithmic consequence is explicit in the hidden-graph identification framework: the random $d$5-regular bipartite connections and random labelling hide the base graph’s structure from any classical algorithm, and any local walk must climb exponentially tall spires before seeing the lifted layer, requiring exponentially many queries (Wocjan, 11 May 2026). A plausible implication is that the hardness is not tied merely to graph size, but to the mismatch between local oracle access and the global spectral regularity inherited from the base graph.

3. Spectral reduction to the towered graph

The defining analytical fact about $d$6 is that, from the apex of any spire, the continuous-time walk is confined to a polynomial-dimensional invariant subspace (Wocjan, 11 May 2026). This is expressed by introducing the level states

$d$7

If $d$8 is the entry apex, then the Krylov subspace

$d$9

is exactly the span of the $G=(V,E)$0 level states (Wocjan, 11 May 2026).

On this subspace, the Hamiltonian $G=(V,E)$1 acts as the weighted adjacency of a simpler graph $G=(V,E)$2. The towered graph has vertices $G=(V,E)$3 for $G=(V,E)$4 and $G=(V,E)$5, vertical edges $G=(V,E)$6–$G=(V,E)$7 of weight

$G=(V,E)$8

and bottom edges $G=(V,E)$9–$n$0 of weight $n$1 whenever $n$2 (Wocjan, 11 May 2026).

In tensor form, with basis $n$3,

$n$4

where

$n$5

Because $n$6, one simultaneously diagonalizes $n$7, obtains eigenpairs $n$8, and block-diagonalizes $n$9 into $A\in\mathbb{R}^{n\times n}$0 independent tridiagonal blocks $A\in\mathbb{R}^{n\times n}$1 of size $A\in\mathbb{R}^{n\times n}$2 (Wocjan, 11 May 2026).

This reduction is the central structural feature of spired graphs. In $A\in\mathbb{R}^{n\times n}$3, each base vertex is blown up into an exponential gadget; in $A\in\mathbb{R}^{n\times n}$4, each base vertex simply sprouts a path of length $A\in\mathbb{R}^{n\times n}$5, preserving spectral fidelity within the Krylov subspace (Wocjan, 11 May 2026). The construction therefore separates graph-theoretic obfuscation from quantum-dynamical accessibility.

4. Tridiagonal blocks and Chebyshev secular equation

For each base-graph eigenvalue $A\in\mathbb{R}^{n\times n}$6, the corresponding block $A\in\mathbb{R}^{n\times n}$7 has matrix entries

$A\in\mathbb{R}^{n\times n}$8

The associated eigenvector amplitudes $A\in\mathbb{R}^{n\times n}$9 satisfy

$L\ge 1$0

(Wocjan, 11 May 2026).

The bulk solutions are expressed through Chebyshev polynomials of the second kind: $L\ge 1$1 with

$L\ge 1$2

The bottom boundary condition yields the secular equation

$L\ge 1$3

In the in-band regime $L\ge 1$4, one may write $L\ge 1$5, giving

$L\ge 1$6

(Wocjan, 11 May 2026).

This decomposition has two immediate consequences. First, the eigenvalues $L\ge 1$7 of each block are simple and are obtained by solving the secular equation. Second, the return amplitude from the entry apex back to itself is explicitly

$L\ge 1$8

where

$L\ge 1$9

is the “top weight” (Wocjan, 11 May 2026).

The paper’s spectral interpretation is that $G$00 may lie in $G$01, whereas $G$02 is spread into $G$03 bands corresponding to shifted and broadened versions of $G$04, analyzable block by block (Wocjan, 11 May 2026). This suggests that the spired construction preserves enough spectral information for identification while strongly disrupting direct combinatorial access.

5. Quantum hidden-graph identification

The algorithm proposed for hidden-graph identification is a continuous-time quantum walk on $G$05, followed by a single Hadamard test at a classically precomputed time $G$06; the algorithm returns the candidate whose predicted amplitude is closest to the measurement (Wocjan, 11 May 2026). The starting apex $G$07 lies in the Krylov subspace $G$08, and evolution under $G$09 from $G$10 is identical to evolution under $G$11 on $G$12 (Wocjan, 11 May 2026).

The spectral decomposition is operational rather than merely structural. The block-diagonal structure and closed-form Chebyshev solution allow fast classical preprocessing, described as $G$13 spectral data, to choose an optimal evolution time $G$14 (Wocjan, 11 May 2026). Efficient numerics enabled by this decomposition supply both $G$15 and the predicted amplitudes.

The main numerical study compares prism graphs $G$16 and Möbius ladders $G$17, each on $G$18 vertices. It supports a precise conjecture that $G$19 measurements at evolution time of order $G$20 suffice to distinguish the two families, and reports tests for $G$21, corresponding to $G$22 up to $G$23 (Wocjan, 11 May 2026).

The paper further conjectures, by analogy with the welded-trees lower bounds, that any classical algorithm requires queries exponential in $G$24 (Wocjan, 11 May 2026). Together with the measurement conjecture, this points to an exponential quantum speedup for identifying an obfuscated base graph. Since the classical lower bound is stated as a conjecture rather than a theorem, the separation remains conditional in the current formulation.

6. Scope, related notions, and terminological distinctions

The term “spired graph” should be distinguished from several similarly named concepts. In rectilinear planarity for independent-parallel SP-graphs, “spirality” is an integer defined from right and left turns along a spine in a rectilinear representation; it encodes allowable rectilinear shapes and supports a linear-time testing algorithm for PISP graphs (Didimo et al., 2021). That notion concerns orthogonal drawing and SPQ$G$25-tree composition, not oracle obfuscation, spectral reduction, or continuous-time quantum walks.

Likewise, “spider graphs” in the sense of abelian-spider theory are obtained by attaching 2-valent paths of finite length to designated vertices of a fixed connected finite graph, and are studied through Perron–Frobenius eigenvalues, cyclotomicity, and subfactor-theoretic constraints (Calegari et al., 2015). Their defining operation is path attachment at selected vertices; the spired-graph construction instead combines exponential cluster lifting, balanced spires, and random regular bipartite couplings (Wocjan, 11 May 2026).

Within its own setting, the spired graph occupies a specific position in quantum query complexity. It generalizes the welded-trees paradigm by replacing a single hard instance with a family indexed by hidden base graphs, while retaining the essential feature that classical local access is obstructed by exponentially large geometry and quantum evolution is compressible to a polynomial-dimensional invariant subspace (Wocjan, 11 May 2026). A plausible implication is that spired graphs provide a reusable template for black-box problems in which spectral information is deliberately preserved under severe combinatorial obfuscation.