Skip to content

Atom-Bond Connectivity Index

A degree-based graph descriptor that sums √[(d(u)+d(v)−2)/(d(u)d(v))] once over every edge to produce a label-independent scalar.

Version
v1 · 2026-09-28 · History
Domain-specific #
8060
Domain group
Natural Sciences
Origin domain
Chemistry & Materials Science
Subdomains
Chemical Graph Theory, Molecular Topological Indices → Chemistry & Materials Science
Aliases
ABC index, Atom–bond connectivity index

Core Idea

The atom–bond connectivity index (ABC index) is a degree-based topological index of a graph. For a graph G,

ABC(G) = Σ_{uv∈E(G)} √[(d(u)+d(v)−2)/(d(u)d(v))],

where the sum runs once over every edge and d(u) and d(v) are the degrees of its endpoints. Each bond or graph edge contributes a weight determined by the local degree pair; summing those weights produces one label-independent scalar for the entire graph.

Introduced by Ernesto Estrada, Luis Torres, Lissette Rodríguez, and Ivan Gutman in 1998, ABC was proposed as a molecular graph descriptor and related empirically to heats of formation for a studied alkane set. The index itself is an exact graph invariant once the carrier graph is fixed. Its chemical predictive meaning is not exact in the same sense: it depends on molecular representation, dataset, target property, and fitted model.

ABC differs from the Randić connectivity index and from similarly named variants. The atom–bond sum-connectivity index changes the edge formula's denominator; the Graovac–Ghorbani index uses distance-based vertex counts; exponential versions transform the local contribution. These can be useful descriptors without being the standard ABC index.

How would you explain it like I'm…

Stick-Score Molecule Number

Think of a molecule as balls joined by sticks. For each stick, look at how many sticks each of its two balls has, and use those two counts to give the stick a little score. Add up every stick's score and you get one number for the whole molecule, called the atom-bond connectivity index.

Molecule Connection Score

Chemists sometimes draw a molecule as a graph: dots for atoms and lines for bonds. The Atom-Bond Connectivity Index turns that whole drawing into one number. For every line, you look at how many lines touch each of its two end dots, which is called their degree. A formula turns those two degrees into a small score for that line, and you add up the scores for all the lines. The number doesn't depend on how you label the dots, and chemists have tried using it to help predict things about molecules, though how well that works depends on the molecules and the property.

Degree-Based Edge-Sum Index

The Atom-Bond Connectivity (ABC) index is a topological index: a single number computed from a graph's structure, often a molecular graph with atoms as vertices and bonds as edges. For each edge uv, take the degrees d(u) and d(v), the number of edges meeting each endpoint, and compute the square root of (d(u) + d(v) − 2) divided by d(u)·d(v). Summing this value over every edge once gives ABC(G). Because it depends only on the graph's structure, it is the same however the vertices are labeled. It was introduced in 1998 as a molecular descriptor and was related empirically to heats of formation for a set of alkanes. The index itself is exact once the graph is fixed, but its usefulness for predicting chemistry depends on how the molecule is represented, which data are used and which property is modeled. It is a different quantity from the Randić connectivity index and from similarly named variants.

 

The atom–bond connectivity (ABC) index is a degree-based topological index defined for a graph G as ABC(G) = Σ over edges uv of √[(d(u) + d(v) − 2)/(d(u)d(v))], where d(u) and d(v) are the endpoint degrees and each edge is counted once. Each edge thus receives a weight determined solely by its local degree pair, and summation yields a label-independent scalar invariant of the whole graph. Ernesto Estrada, Luis Torres, Lissette Rodríguez and Ivan Gutman introduced it in 1998 as a molecular graph descriptor, relating it empirically to heats of formation for a studied set of alkanes. Two senses of exactness must be kept apart: the index is an exact graph invariant once the carrier graph is fixed, whereas its chemical predictive meaning depends on the molecular representation, dataset, target property and fitted model. ABC is distinct from the Randić connectivity index and from variants with similar names: the atom–bond sum-connectivity index changes the denominator of the edge term, the Graovac–Ghorbani index uses distance-based vertex counts, and exponential versions transform the local contribution. Those variants can be useful descriptors but are not the standard ABC index.

Structural Signature

  • Finite graph supplies vertices and an edge set, often a hydrogen-suppressed molecular graph.
  • Vertex degree counts incident edges at each endpoint.
  • Edge-local weight applies the defining square-root expression to the endpoint degrees.
  • Edge aggregation sums one contribution for every edge.
  • Invariant output produces a scalar unchanged by relabeling vertices.
  • Empirical interpretation can relate that scalar to a property through an explicitly scoped model.

Change the local formula, use a distance count instead of degree, or sum over a different carrier and the descriptor changes. A QSPR relation is an application layered on the invariant, not part of its mathematical definition.

What It Is Not

The ABC index is not a direct experimental measurement of bond strength, enthalpy, toxicity, or biological activity. It is computed from a graph representation. It does not uniquely encode molecular structure: nonisomorphic graphs can share a scalar index, and stereochemistry, geometry, electronic structure, and environmental conditions are absent unless represented elsewhere.

It is not the Randić index, Zagreb index, ABS index, exponential ABC index, or Graovac–Ghorbani index. Algebraic resemblance and shared “connectivity” vocabulary do not make formulas interchangeable. It is also not primarily a branching count; the original work distinguished its behavior from a descriptor designed around branching.

Scope of Application

ABC is studied in chemical graph theory, mathematical chemistry, extremal graph theory, and quantitative structure–property modeling. Researchers derive bounds, identify extremal graphs, characterize ABC-minimal trees, and compare the index with molecular properties on defined datasets.

Literal application requires a specified graph convention. Whether hydrogens are explicit, how bond multiplicity is encoded, and what counts as adjacency can affect the carrier and hence the value. Predictive applications require validation appropriate to the chemical domain and endpoint. A high correlation in an alkane training set does not establish universal accuracy for other structures or properties, and descriptor studies do not replace experimental or mechanistic evidence.

Clarity

Atom–Bond Connectivity Index separates three things: molecular or graph representation, exact calculation of the invariant, and empirical interpretation of that number. A reproducible report states all three rather than calling the scalar a measured chemical property.

The formula provides a strong identity test. If the computation does not use endpoint degrees, the exact numerator and product denominator, and one sum over edges, it is a neighboring index even if authors call it ABC-related.

Manages Complexity

A molecular graph can contain many vertices, edges, branches, cycles, and local environments. ABC collapses all endpoint-degree pairs into one scalar through a fixed local-to-global aggregation. This supports mathematical comparison and compact model input.

The compression is deliberately lossy. It discards vertex labels, geometry, and much structural detail, and can create descriptor degeneracy. Keeping the underlying graph, degree-pair distribution, formula version, and intended property model visible allows the compact number to be used without claiming it is a complete molecular representation.

Abstract Reasoning

  1. Define the graph and its chemical or mathematical representation convention.
  2. Compute the degree of every vertex.
  3. For each edge uv, evaluate the standard ABC weight from d(u) and d(v).
  4. Sum each edge contribution once and verify invariance under relabeling.
  5. For bounds or extremal questions, relate degree constraints and graph class to the sum.
  6. For QSPR use, fit and validate a property model on a stated domain.
  7. Test whether another descriptor or full structural representation adds information ABC discards.

Knowledge Transfer

The computation transfers literally across finite graphs whenever degree and edge conventions are specified. Chemical interpretation transfers much more narrowly: a graph class, molecular representation, property, and validated model must match.

Its strict parent is Measurement in the encyclopedia's broad procedural sense: a fixed operation maps graph structure to a scalar descriptor. The edge does not imply experimental instrumentation or uncertainty-free chemical prediction. Ratio is not the parent because ABC aggregates transformed local ratios rather than reporting one named numerator per denominator as its output.

Examples

Canonical

For a path graph, the two end vertices have degree one and internal vertices degree two. Each edge's endpoint pair is inserted into the ABC term and the contributions are summed. Relabeling the path changes no degrees or adjacencies, so the result is invariant.

Mapped back: graph → path; degrees → one and two; local weight → standard square-root term; aggregation → all path edges; output → one invariant scalar; interpretation → none required.

Applied / In Practice

A QSPR study represents each alkane by a declared molecular graph, computes ABC, fits a relation to standard heat of formation, and evaluates it on an appropriate validation set. The regression supplies empirical meaning; the ABC formula itself remains unchanged.

Mapped back: graph → alkane representation; degree → carbon connectivity; local weight → ABC term; aggregation → molecular sum; output → descriptor; interpretation → scoped enthalpy model.

Structural Tensions

Descriptor simplicity versus structural discrimination. One scalar is cheap to compute but can map different graphs to the same value. Diagnostic: Does ABC distinguish the structures relevant to this task?

Mathematical invariant versus chemical predictor. The graph value is exact while property prediction is empirical. Diagnostic: Which dataset, endpoint, and validation warrant the interpretation?

Formula family versus identity integrity. Variants may improve a task but cease to be the standard index. Diagnostic: Which edge-local weight and carrier were actually used?

Structural–Framed Character

ABC is strongly structural as a graph invariant. Its graph input, degree operation, edge formula, and sum are formally testable. The chemical graph representation and choice of application introduce framing: the same physical molecule can be encoded under different conventions, and the relevance of a scalar depends on the target question.

Evaluation enters only in model use. The index is neither good nor bad in isolation; adequacy depends on discrimination, validation, and comparison with alternatives.

Structural Core vs. Domain Accent

The core is endpoint degrees → fixed edge weights → graph-wide sum. Chemical graph theory supplies atoms as vertices, bonds as edges, molecular representation conventions, and property-model uses. Remove that accent and ABC remains defined for general graphs. Change the formula and the named abstraction disappears.

This explains why the node is domain-specific despite mathematical portability: its established identity and interpretation arise in chemical graph theory, while Measurement supplies only the broader mapping-to-scalar genus.

This entry is a kind of Measurement.

  • Immediate parent — Measurement (subsumption). A fixed computational procedure maps graph structure to a scalar descriptor.
  • Graph invariant. Describes relabeling independence.
  • Aggregation. Combines local edge contributions.
  • Degree and adjacency. Supply the structural inputs.
  • QSPR modeling. Uses the index as a feature but is not part of its definition.

Relationships to Other Abstractions

Local relationship map for Atom-Bond Connectivity IndexParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Atom-BondConnectivity IndexDOMAINPrime abstraction: Measurement — is a kind ofMeasurementPRIME

Current abstraction Atom-Bond Connectivity Index Domain-specific

Parents (1) — more general patterns this builds on

  • Atom-Bond Connectivity Index is a kind of Measurement Prime

    Atom-Bond Connectivity Index is a strict kind of Measurement: A degree-based graph descriptor that sums √[(d(u)+d(v)−2)/(d(u)d(v))] once over every edge to produce a label-independent scalar.

Hierarchy path (1) — routes to 1 parentless root

Neighborhood in Abstraction Space

Atom-Bond Connectivity Index sits in a sparse region of the domain-specific corpus (88th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2551 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With

  • Randić connectivity index. Uses a different degree-based formula. Tell: What edge weight is summed?
  • Atom–bond sum-connectivity index. Replaces the degree-product denominator with a sum-based expression.
  • Graovac–Ghorbani index. Uses distance-based counts rather than endpoint degrees.
  • Experimental property measurement. Observes a physical quantity rather than computes a graph invariant.
  • Complete molecular representation. ABC cannot reconstruct every graph, geometry, or chemical behavior.

References

  • Estrada, Torres, Rodríguez, and Gutman, “An atom-bond connectivity index: Modelling the enthalpy of formation of alkanes” (1998): https://nopr.niscpr.res.in/handle/123456789/40308
  • Frozen Wikipedia discovery revision: https://en.wikipedia.org/wiki/Atom-bond_connectivity_index

The original study supports the formula, descriptor identity, and scoped alkane application. Later bounds and variants are retained as mathematical context, not universal chemical guarantees.