General Randić Index

• General Randić Index is a graph invariant defined by extending the classical index with a flexible exponent α to capture branching and connectivity patterns in networks.
• Spectral and combinatorial methods yield rigorous bounds and extremal results, linking the index to regularity conditions and pivotal structural transitions in graphs.
• Recent research explores applications in random graph models, directed graphs, and Hamiltonicity criteria, unifying diverse indices from mathematical chemistry to network data analysis.

The General Randić Index is a fundamental graph invariant that encodes branching and degree interaction patterns in networks, with origins in chemical graph theory and extensions spanning discrete mathematics, mathematical chemistry, and network data analysis. The principal generalizations include the edge-based general Randić index and the vertex-based zeroth-order general Randić index. Modern research rigorously explores extremal values, bounds, spectral interpretations, and structural transitions of these indices both for arbitrary graphs and within classes defined by chromatic number, degree bounds, or topological structure.

1. Definitions and Principal Variants

Let $G=(V,E)$ be a simple undirected graph, with vertex degree $d(v)$ for $v\in V$. The classical Randić index is

$R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$

which quantifies branching and is extensively used in chemoinformatics. The general Randić index extends this to arbitrary real exponents $\alpha$,

$$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$

which interpolates between various structural invariants: e.g., $\alpha = -1/2$ yields the original index; $\alpha = -1$ recovers the (modified) second Zagreb index; positive $\alpha$ emphasizes hub-to-hub connectivity.

The zeroth-order general Randić index (sometimes referred to as the "vertex-based" general Randić index) is defined for real $\alpha\ne 0$ by

$d(v)$0

and unifies several well-known topological indices:

• $d(v)$1: twice the number of edges.
• $d(v)$2: first Zagreb index.
• $d(v)$3: forgotten index.
• $d(v)$4: vertex count.
• $d(v)$5: classical zeroth-order Randić index of Kier–Hall (Jamil et al., 2018, Wang et al., 2 Apr 2026).

The generalization to digraphs, notably for oriented graphs, is given by

$d(v)$6

where $d(v)$7 and $d(v)$8 denote out- and in-degree, respectively, and $d(v)$9 is a real parameter (Yang et al., 2022, Yang et al., 2021).

2. Spectral and Combinatorial Bounds

Spectral techniques play a central role in bounding both the general Randić index and its zeroth-order form. Fundamental inequalities due to the application of the power-mean (Hölder-type) inequalities yield

$v\in V$0

where $v\in V$1, $v\in V$2, and $v\in V$3 is the adjacency spectral radius (Elphick et al., 2015). These bounds become equalities precisely for regular (or semiregular bipartite) graphs.

Extremal structures under fixed degree constraints can be completely classified: for a graph with minimum degree $v\in V$4, maximum degree $v\in V$5 and order $v\in V$6, the O–Shi bound asserts that for $v\in V$7,

$v\in V$8

with equality if and only if the graph is $v\in V$9-biregular (Haslegrave, 2024). This generalizes for all $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$0 via a sharp minimization over $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$1 in the degree interval, and three "phase regimes" for minimality and maximality (variants of δ-regular, Δ-regular, or biregular graphs) as shown in Theorem 3 (Haslegrave, 2024).

3. Extremal Results and Structural Graph Theory Connections

Recent work provides explicit extremal formulas for the zeroth-order general Randić index in a range of graph classes.

• Chromatic and clique number bounds: For fixed chromatic number $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$2, the Turán graph $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$3 minimizes $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$4 for $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$5, and the pineapple graph $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$6 maximizes it for $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$7 (Jamil et al., 2019). Analogous bounds hold for fixed clique number and in connectivity- or bridge-constrained classes.
• Cut-edges (bridges): For connected graphs with given number $R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$8 of cut-edges, the extremal is the graph with all bridges as pendent edges at a single vertex on a cycle (Jamil et al., 2019).
• Connectivity and minimum degree: Extremal graphs are split graphs ($R(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^{-1/2},$9) for lower bounds and stars for upper bounds in the regime $\alpha$0.
• k-Generalized quasi-trees: Sharp minimum and maximum formulas for the zeroth-order index in k-generalized quasi-trees elucidate the interplay between quasi-vertex structure, convexity/concavity of $\alpha$1, and graph attachment patterns (Jamil et al., 2018).
• Orientation extremals in cacti and related classes: The extremal orientations maximizing the digraph zeroth-order index are characterized by sink–source configurations, with explicit formulas depending only on order, matching number, and cycle structure (Yang et al., 2022, Yang et al., 2021).

4. Asymptotic Behavior in Random Graph Models

Large-scale asymptotics in random graph models provide insight into typical index values:

• Random trees with bounded maximum degree: The general Randić index admits a law of large numbers, with the edge counts of given degree types converging to limiting proportions, and variance decaying as $\alpha$2. For tree distributions $\alpha$3, the index is asymptotically linear in $\alpha$4 with explicit (model-dependent) constants (Li et al., 2010).
• Erdős–Rényi and inhomogeneous models: The expectation and concentration of $\alpha$5 is established in both homogeneous and inhomogeneous models. In dense $\alpha$6, $\alpha$7; in inhomogeneous models, the leading term depends on both edge probability $\alpha$8 and heterogeneity of degree sequence via a kernel $\alpha$9 (Yuan, 2023).
• Empirical validation: Real-world networks display $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$0 values close to $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$1 in homogeneous-like cases, substantially deviating in networks with strong heterogeneity or community structure.

5. Hamiltonicity, Thresholds, and Related Indices

The zeroth-order general Randić index provides sharp sufficient conditions for Hamiltonian and $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$2-Hamiltonian properties. Theorems parametrize threshold values of $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$3 below or above which a graph must be (k-)Hamiltonian, depending on $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$4 and graph order. These bounds are tight, with extremal examples characterized precisely (Wang et al., 2 Apr 2026).

The index encompasses many classical topological indices used in mathematical chemistry, e.g., Zagreb indices and the forgotten index, as special cases of the general form. The transition between convexity and concavity (e.g., at $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$5 and $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$6) determines the extremal graph families for a given degree range or graph class.

6. Methodological and Proof Techniques

Proofs leverage several recurring technical ingredients:

• Edge-swapping and degree-transfer lemmas: Manipulations that reallocate degree mass (locally or globally) to increase or decrease the index per monotonicity of $$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$$7.
• Spectral methods: Application of spectral radii and power-mean inequalities to derive sharp lower and upper bounds.
• Convexity/concavity of the power function: Yielding majorization results which determine when minimum/maximum is achieved on regular, nearly regular, or extremal degree sequences.
• Generating function and analytic combinatorics: In random structures, providing exact or asymptotic evaluation of index distributions.
• Sink–source orientation dominance: For oriented graphs, the maximal digraph indices are always realized on configurations where each vertex is a pure source or sink (Yang et al., 2022, Yang et al., 2021).

7. Extensions and Open Directions

Theoretical advances suggest several active and emerging directions:

• Boundary cases in ultra-sparse random graphs and more complex random graph models, where classical asymptotic expansions fail (Yuan, 2023).
• Generalization to block-decomposable graphs, polymeric networks, and fractal-type constructions, with explicit formulas for Sierpiński graphs and related networks (Estrada-Moreno et al., 2015).
• Statistical hypothesis testing and statistical inference on network invariants, with potential for diagnostics and parameter estimation in real-world network data (Yuan, 2023).
• Unification of extremal and spectral approaches: The power-mean methodology provides a global template for recovering all classical combinatorial bounds as spectral specializations (Elphick et al., 2015, Haslegrave, 2024).

• (Jamil et al., 2019) Some Bounds on Zeroth-Order General Randić$$Index</li> <li>(<a href="/papers/2604.02254" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Wang et al., 2 Apr 2026</a>) Sufficient conditions for Hamiltonianity in terms of the Zeroth-order General Randić Index</li> <li>(<a href="/papers/1801.03885" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Jamil et al., 2018</a>) Zeroth-order general Randic index of$$R_\alpha(G) = \sum_{\{u,v\}\in E} (d(u)d(v))^\alpha,$8-generalized quasi trees
• (Yang et al., 2022) Maximum zeroth-order general Randić index of orientations of cacti
• (Yang et al., 2021) Maximum zeroth-order general Randić index of orientations of trees, unicyclic and bicyclic graphs with given matching number
• (Elphick et al., 2015) Bounds and power means for the general Randic index
• (Haslegrave, 2024) The extremal generalised Randić index for a given degree range
• (Li et al., 2010) The asymptotic values of the general Zagreb and Randić indices of trees with bounded maximum degree
• (Yuan, 2023) On the Randić index and its variants of network data
• (Estrada-Moreno et al., 2015) On the General Randić index of polymeric networks modelled by generalized Sierpiński graphs