Hyper-Wiener Index¶

Compress a connected molecular graph's shortest-path distance distribution into a scalar by summing both distance and squared distance over unordered vertex pairs.

Version: v2 · 2026-08-30 · History
Domain-specific #: 2029
Origin domain: chemistry
Subdomain: chemical graph theory
Aliases: Hyper-Wiener number

Core Idea¶

The Hyper-Wiener Index is a distance-based graph invariant developed in chemical graph theory. For a finite connected simple graph G=(V,E), let d_G(u,v) be the shortest-path distance between vertices u and v. Its standard connected-graph definition is.

WW(G) = (1/2) sum_{{u,v} subset V, u!=v} [d_G(u,v) + d_G(u,v)^2],

where the summation is over unordered pairs of distinct vertices.^[1] This convention is load-bearing. The equivalent ordered-pair formula is.

WW(G) = (1/4) sum_{u in V} sum_{v in V} [d_G(u,v) + d_G(u,v)^2].

Writing one half in front of an ordered double sum would double the accepted value. The diagonal terms vanish but do not repair that factor-of-two error.

Randić introduced the descriptor for molecular trees as a generalization of the Wiener index. Klein, Lukovits, and Gutman then supplied the definition for cycle-containing molecular structures; the linear-plus-squared distance sum is now used as the definition for connected graphs.^[2]^[3]^[1] The quadratic term makes distant vertex pairs contribute disproportionately: a pair at distance k contributes k(k+1)/2 rather than the Wiener index's contribution k.

The output is one scalar. It is invariant under graph isomorphism because an isomorphism preserves all shortest-path distances, but it is not a complete graph identifier: different graphs can share the same distance aggregate. In chemical use, vertices and edges usually form a hydrogen-suppressed molecular graph, so the number summarizes topological compactness or extension of that representation. It does not directly encode three-dimensional conformation, bond energy, electron distribution, or causal molecular behavior.^[1]

Structural Signature¶

Sig role-phrases:

the connected graph G=(V,E) — the finite structural carrier, commonly a hydrogen-suppressed molecular graph
the unordered vertex-pair domain — every two-element subset {u,v} of V is counted exactly once
the shortest-path graph metric d_G — the minimum number of edges on a path between each pair
the linear distance term — preserves the ordinary Wiener contribution d_G(u,v)
the quadratic distance term — gives relatively greater weight to pairs separated by long graph paths
the half-sum normalization — converts the linear-plus-square pair expression into the accepted WW convention
the scalar aggregation — collapses the full pair-distance distribution to one nonnegative number
the isomorphism invariance — relabeling vertices preserves the pair distances and hence the result
the chemical-topology interpretation — uses the scalar as a molecular-graph descriptor while keeping representation limits explicit

Recognition test. Identify a finite connected graph, compute ordinary unweighted shortest-path distances, enumerate unordered distinct vertex pairs once, and apply one half to each d+d^2 contribution. Confirm that the claimed chemical meaning belongs to the declared molecular graph representation. If the graph is disconnected without an explicit extension, distances are geometric rather than graph-theoretic, ordered pairs receive the unordered-pair factor, or the value is treated as a unique molecular identifier, the signature fails.

What It Is Not¶

Not the Wiener index. W(G) is the sum of shortest-path distances over unordered pairs; WW(G) additionally uses squared distance and a half normalization.
Not any distance-based topological index. The exact d+d^2 functional and pair convention define this index.
Not a complete graph invariant. Equal Hyper-Wiener values do not prove isomorphism.
Not a three-dimensional molecular measurement. The usual input graph abstracts away coordinates, conformations, and many chemical attributes.
Not a causal property predictor. Correlation in a quantitative structure–property model does not show that WW causes the predicted property.
Not defined by the standard formula on disconnected graphs. Ordinary graph distance is infinite between components, so an extension must declare a different convention.
Not a generic network-security or hardware concept. A connected nonchemical graph can receive the same mathematical calculation, but an application claim requires separate evidence and does not redefine the chemical-graph abstraction.

Scope of Application¶

The Hyper-Wiener Index recurs primarily in mathematical chemistry and the graph theory that supports it.

Molecular descriptors. A molecular structure is converted to a graph and WW becomes one candidate feature in quantitative structure–property or structure–activity analysis. The model must specify atom and bond treatment; the scalar alone does not carry those choices.^[2]^[1]
Chemical graph families. Closed forms, recurrences, and extremal bounds compare benzenoid, phenylene, polyazulene, tree, and unicyclic families. These studies expose how branching, cycles, and long paths affect the descriptor.^[1]^[4]
Algorithm design. All-pairs shortest paths give a direct computation. Cut methods and Hosoya-polynomial methods exploit special graph structure to avoid enumerating every pair explicitly.^[1]
Graph-theoretic comparison. The formula is mathematically meaningful on any finite connected simple graph. That is literal reuse of the invariant, not evidence that every graph is a molecule or that chemical interpretations transfer automatically.

Weighted graphs, multigraphs, disconnected graphs, periodic infinite graphs, and uncertain molecular representations require declared extensions. A finite quotient or finite member of a graph family may have a Hyper-Wiener index; an infinite connected graph generally makes the unqualified total divergent or undefined.

Clarity¶

Three equivalent representations prevent convention drift. Let N_k be the number of unordered vertex pairs at distance k. Then

W(G) = sum_{k>=1} N_k k

and

WW(G) = (1/2) sum_{k>=1} N_k(k+k^2).

The same quantity may be written (W(G) + sum_{{u,v}} d_G(u,v)^2)/2. Because the pair set is unordered, no further division by two is needed. If a distance matrix includes both (u,v) and (v,u), its off-diagonal total must instead be divided by four.

The distance histogram also clarifies interpretation. WW is a deterministic functional of the histogram {N_k}; it forgets which pair realizes which distance and how those pairs overlap. Two graphs with the same distance distribution necessarily have the same value, and even different distributions can yield the same weighted total. “Compactness” is therefore a coarse topological reading, not a reconstruction guarantee.^[1]

Manages Complexity¶

The index reduces an O(|V|^2) family of pair relations to one scalar. That compression supports sorting graph families, testing formulas, adding a fixed-size feature to a statistical model, and comparing how structural modifications alter long-range separation.

The quadratic term makes a useful diagnostic distinction. Changing a remote branch or extending a chain affects WW more strongly than a linear-only distance sum, while a local edit may have widespread effects because it changes shortest paths for many pairs. A distance histogram lets an analyst localize the change by distance shell before accepting the scalar explanation.

Compression also creates risk. Size, branching, cycle placement, and many other changes can move the same scalar. Sound use retains the graph, the distance convention, and preferably companion descriptors rather than treating WW as a self-explaining label.

Abstract Reasoning¶

Normalize the pair convention first. Decide whether an implementation enumerates unordered pairs or an ordered distance matrix. Align the leading coefficient before comparing results.

Move from graph to distance distribution. Compute d_G(u,v), collect counts N_k, and evaluate the polynomial weight k(k+1)/2. This separates shortest-path correctness from final aggregation.

Use invariance as a check. Vertex relabeling must leave the histogram and WW unchanged. A change after relabeling exposes an implementation error, not chemical information.

Reason about edits globally. Adding or removing an edge can shorten many pairwise paths. Recompute affected distance shells rather than adding only a local edge contribution.

Audit representation before interpretation. Check which atoms are vertices, which bonds are edges, whether weights are ignored, and whether the graph is connected. Only then test correlations or compare chemical families.

Treat collisions as expected. If two structures share WW, use other graph invariants or domain measurements. Scalar equality is a prompt for further discrimination, not an identity certificate.

Knowledge Transfer¶

The mathematical mechanism transfers literally to any finite connected simple graph: shortest-path distance, unordered pairs, and the same normalized polynomial aggregation require no chemical substrate. Network graphs, combinatorial constructions, and software test graphs can therefore use WW as the same graph invariant.

The interpretation does not transfer automatically. In a molecular graph, path distance counts bond steps in a declared structural representation; in a communication graph it counts hops; in a social graph it may encode an uncertain relationship path. A larger value has no universal meaning across these domains because graph order, modeling choices, and what an edge represents all change.

The safe transfer protocol is: preserve the exact formula, state the graph semantics, verify connectedness and scale comparability, and validate any claimed relation to domain outcomes independently. Without those steps, only the words “distance” and “compactness” have traveled.

Examples¶

Canonical: compute `WW` for the four-vertex path¶

Let P_4 have vertices 1-2-3-4. Among its six unordered vertex pairs, three have distance 1, two have distance 2, and one has distance 3. Therefore

W(P_4) = 3(1)+2(2)+1(3)=10,

sum d^2 = 3(1)+2(4)+1(9)=20,

and WW(P_4)=(10+20)/2=15. An ordered double sum of d+d^2 would equal 60, so its coefficient must be 1/4, not 1/2.

Mapped back: P_4 is the connected graph; its six two-element subsets are the unordered vertex-pair domain; path lengths supply the shortest-path metric; 10 is the linear term total; 20 is the quadratic term total; division by two is the normalization; and 15 is the scalar aggregation.

Applied / In Practice: the six-cycle molecular skeleton¶

Model the carbon skeleton of a six-membered ring as C_6. It has six unordered pairs at distance 1, six at distance 2, and three opposite pairs at distance 3. Hence

W(C_6)=6(1)+6(2)+3(3)=27,

sum d^2=6(1)+6(4)+3(9)=57,

so WW(C_6)=(27+57)/2=42. The value characterizes this declared hydrogen-suppressed graph representation; it does not by itself distinguish bond orders, substituents, conformations, or measured properties.

Mapped back: the ring skeleton supplies the molecular graph; bond-step paths supply the graph metric; its distance shells implement the pair domain; the distance-three pairs receive greater relative weight through the quadratic term; 42 is the invariant scalar; and the explicit modeling caveat preserves the chemical-topology boundary.

Structural Tensions¶

T1: Unordered pairs versus ordered matrices. The same distances appear once in a pair set and twice off-diagonal in a matrix. Diagnostic: Does the coefficient match the enumeration convention, yielding WW(P_4)=15?

T2: Linear coverage versus quadratic emphasis. The Wiener term records all distances, while squaring magnifies long paths. Diagnostic: Is the analysis intentionally sensitive to extended structure, or is the quadratic term dominating an unrelated size effect?

T3: Compression versus discrimination. One scalar is convenient but many-to-one. Diagnostic: Could a competing graph or distance histogram produce the same weighted total, and what companion evidence distinguishes it?

T4: Graph topology versus chemical reality. A molecular graph is a selective representation. Diagnostic: Which atoms, bonds, weights, charges, stereochemistry, and conformations were omitted before interpreting the index?

T5: Connected definition versus broader graph classes. Standard shortest-path totals require finite distances. Diagnostic: Is the input connected, or has a nonstandard extension been declared and kept separate?

T6: Isomorphism invariance versus complete identification. Relabeling invariance is necessary for a graph descriptor but insufficient for uniqueness. Diagnostic: Is scalar equality being used only as similarity evidence, not as proof of isomorphism?

T7: Domain autonomy versus prime reduction. Aggregation and Metric explain the portable skeleton, yet they do not select the d+d^2 functional, half-sum convention, connected-graph scope, or molecular-descriptor diagnostics. Diagnostic: After naming those parents, does the exact distance-polynomial instrument still support distinct recognition, calculation, and failure checks? If yes, the domain node remains autonomous.

Structural–Framed Character¶

Criterion 1 — Vocabulary travels (0.25). Graph distance, pair aggregation, and invariance transfer broadly, but “Hyper-Wiener Index” retains one exact mathematical formula rather than becoming general-purpose vocabulary.

Criterion 2 — Evaluative weight (0.0). The value is a formal invariant; high or low is not intrinsically good, legitimate, or desirable.

Criterion 3 — Institutional origin (0.0). Scholarly convention fixes the name and normalization, but institutions do not confer the graph-theoretic value on an instance.

Criterion 4 — Human-practice boundedness (0.0). The invariant applies to mathematical graphs without requiring roles, norms, or recurring human institutions.

Criterion 5 — Import versus recognition (0.25). The formula is recognized literally on connected graphs, while empirical interpretation requires an explicit domain model.

The aggregate 0.1 yields structural under the documented rubric. Its objectivity does not make it a prime: the exact formula and chemical-graph lineage remain specialist.

Structural Core vs. Domain Accent¶

What is skeletal. Many pairwise relations are evaluated by a metric, transformed by a weighting rule, aggregated into a summary, and checked for invariance under relabeling. Aggregation, Metric, and Invariance express this transferable pattern.

What is domain-bound. The carrier is a finite connected graph, the relation is unweighted shortest-path distance, the functional is exactly half the unordered-pair sum of d+d^2, and the principal interpretation is a topological molecular descriptor. These constraints determine recognition and calculation.

Why this does not clear the prime bar. Replacing molecular or graph vertices with unrelated substrate elements does not preserve the identity unless the new domain first supplies a connected graph and deliberately imports the same formula. The broad insight is already captured by the parent primes; Hyper-Wiener Index remains a reusable chemical-graph instrument.

Hyper-Wiener Index instantiates prime:aggregation by strict subsumption. It deliberately collapses all pair distances into one scalar and thereby trades structural detail for tractability. Aggregation is broader because it does not determine a graph, distance, weight function, or normalization.

It also presupposes prime:metric through strict composition. Shortest-path distance on a connected graph supplies the pairwise values being aggregated. Metric alone does not select an all-pairs scalar or the Hyper-Wiener polynomial.

prime:invariance is a close consequence rather than a proposed parent: graph isomorphisms preserve shortest-path distances, so invariance follows once the formula is fixed. prime:isomorphism and prime:symmetry are useful reasoning neighbors but do not define the index. prime:fractal_geometry is neither an alias nor a parent; scale-dependent self-similarity is not part of the Hyper-Wiener signature.

Relationships to Other Abstractions¶

Current abstraction Hyper-Wiener Index Domain-specific

Parents (2) — more general patterns this builds on

Hyper-Wiener Index is a kind of Aggregation Prime

Hyper-Wiener Index instantiates prime:aggregation by strict subsumption.
Hyper-Wiener Index presupposes Metric Prime

Hyper-Wiener Index instantiates prime:aggregation by strict subsumption.

Hierarchy paths (2) — routes to 2 parentless roots

Hyper-Wiener Index → Aggregation → Micro Macro Linkage

Show alternative path (1)

Neighborhood in Abstraction Space¶

Hyper-Wiener Index sits in a sparse region of the domain-specific corpus (75^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Wiener index — 0.87
Diameter (graph theory) — 0.84
Randić index — 0.83
Graph Sphericity — 0.83
Zagreb indices — 0.82

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Wiener index. Sums d, without the same squared term. Tell: Does a distance-k pair contribute k or k(k+1)/2?
Randić connectivity index. Uses degree-based edge contributions rather than all-pairs shortest paths. Tell: Are vertex degrees or pair distances being aggregated?
Harary index. Sums reciprocal distances. Tell: Do long distances receive smaller reciprocal weight or larger quadratic weight?
Hosoya polynomial. Encodes the whole distance distribution as a generating polynomial and can be used to derive WW. Tell: Is the output a polynomial or one scalar?
Graph diameter. Retains only the largest pair distance. Tell: Are all unordered pairs contributing?
Distance matrix. Preserves pair-labeled distances instead of collapsing them. Tell: Can the output identify which vertices realize a distance?
Topological index in general. Names a broad descriptor family. Tell: Is the exact Hyper-Wiener functional specified?
Spatial index. A data structure for geometric lookup. Tell: Is the object a molecular graph invariant or a search structure?

References¶

[1] Gordon G. Cash, Sandi Klavžar, and Marko Petkovšek, “Three Methods for Calculation of the Hyper-Wiener Index of Molecular Graphs”, Journal of Chemical Information and Computer Sciences 42(3), 571–576, 2002; author-hosted manuscript. States the accepted connected-graph formula explicitly over unordered pairs and develops cut, Hosoya-polynomial, and interpolation methods. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[2] Milan Randić, “Novel molecular descriptor for structure—property studies”, Chemical Physics Letters 211(4–5), 478–483, 1993. Introduces the descriptor in molecular structure–property work, initially for tree structures. registry ↩a ↩b

[3] Douglas J. Klein, Ivan Lukovits, and Ivan Gutman, “On the definition of the hyper-Wiener index for cycle-containing structures”, Journal of Chemical Information and Computer Sciences 35(1), 50–52, 1995. Extends the definition from trees to molecular graphs containing cycles. registry ↩

[4] R. Xing, B. Zhou, and X. Qi, “Hyper-Wiener Index of Unicyclic Graphs”, MATCH Communications in Mathematical and in Computer Chemistry 66, 315–328, 2011. Uses the unordered-pair definition and studies extremal behavior on connected unicyclic graphs. registry ↩