Roman Bondage Number in Graph Theory

• Roman bondage number is defined as the minimum number of edges whose removal increases a graph’s Roman domination number, reflecting its edge vulnerability.
• It is studied using combinatorial optimization techniques and is NP-hard to compute, even for bipartite graphs, with bounds linked to degree and connectivity.
• Exact results exist for families like complete graphs, cycles, and trees, while open problems drive research in classification and approximation methods.

The Roman bondage number is an invariant in graph theory tailored to measure the vulnerability of the Roman domination number of a graph under edge removal. Given the Roman domination function’s role as a generalization of classical domination—allowing vertex labels in $\{0,1,2\}$ with the requirement that every $0$-labeled vertex has a $2$-labeled neighbor—the Roman bondage number quantifies the minimum number of edges whose deletion forces the Roman domination number to rise. This parameter lies at the intersection of combinatorial optimization, domination theory, and algorithmic complexity.

1. Fundamental Definitions and Properties

Let $G = (V, E)$ be a finite, simple, undirected graph. A Roman dominating function (RDF) on $G$ is a labeling $f : V \to \{0, 1, 2\}$ such that each vertex $v$ with $f(v) = 0$ has at least one neighbor $u$ with $f(u) = 2$. The weight of $0$0 is $0$1. The Roman domination number $0$2 is defined as the minimum weight of an RDF on $0$3.

The Roman bondage number $0$4 is defined as: $0$5 This is well-defined for graphs with maximum degree at least $0$6. For a minimum Roman bondage set $0$7, it always holds that $0$8 (Hu et al., 2011).

Key properties include:

• $0$9, where $2$0 is the standard domination number.
• If $2$1 is a Roman graph ($2$2), then $2$3, where $2$4 is the classical bondage number (Hu et al., 2011).
• The computation of $2$5 is NP-hard, even for bipartite graphs (Bahremandpour et al., 2012, Hu et al., 2011).

2. General Bounds and Extremal Results

The literature provides several sharp upper and lower bounds for the Roman bondage number:

Upper bounds (Hu et al., 2011, Bahremandpour et al., 2012, Samodivkin, 2014):

• $2$6, with $2$7 the maximum degree.
• $2$8, where $2$9 is the minimum degree and $G = (V, E)$0 the vertex connectivity.
• $G = (V, E)$1 for connected $G = (V, E)$2, with $G = (V, E)$3 the average degree (Samodivkin, 2014).
• If $G = (V, E)$4 is 2-cell embedded on a surface of Euler characteristic $G = (V, E)$5, then

$G = (V, E)$6

where $G = (V, E)$7 is the girth, $G = (V, E)$8.

• For $G = (V, E)$9 embedded on surfaces with $G$0, $G$1 for $G$2 (Samodivkin, 2014).

Lower bounds and extremal cases (Hu et al., 2011, Bahremandpour et al., 2012):

• For Roman graphs, $G$3.
• For vertex-Roman-domination-critical graphs ($G$4, $G$5), $G$6, with $G$7 the vertex cover number.

3. Exact Results for Key Graph Families

Roman bondage numbers have been explicitly computed for several graph families:

• Complete graphs $G$8: $G$9 (Hu et al., 2011).
• Paths $f : V \to \{0, 1, 2\}$0:

$f : V \to \{0, 1, 2\}$1

(Hu et al., 2011)

• Cycles $f : V \to \{0, 1, 2\}$2:

$f : V \to \{0, 1, 2\}$3

(Hu et al., 2011, Bahremandpour et al., 2012)

• Cartesian grids $f : V \to \{0, 1, 2\}$4: $f : V \to \{0, 1, 2\}$5 for $f : V \to \{0, 1, 2\}$6 (Hu et al., 2011, Bahremandpour et al., 2012).
• Complete $f : V \to \{0, 1, 2\}$7-partite graphs $f : V \to \{0, 1, 2\}$8 (Hu et al., 2011):
  • If all parts have size $f : V \to \{0, 1, 2\}$9: $v$0 for $v$1.
  • If $v$2, $v$3 where $v$4 is the number of singleton parts.
  • If $v$5, $v$6 for $v$7 parts of size 2.
  • If $v$8 and $v$9, $f(v) = 0$0.
• $f(v) = 0$1-regular graphs ($f(v) = 0$2) except $f(v) = 0$3: $f(v) = 0$4 (Hu et al., 2011).

For trees in the class $f(v) = 0$5 (Roman domination invariant under vertex deletion), $f(v) = 0$6 (Samodivkin, 2015).

4. Structural and Algorithmic Aspects

The Roman bondage number is structurally sensitive. In general, the removal of a single edge does not always suffice to increase $f(v) = 0$7; edge deletion often needs to target specific substructures or localities associated with the labeling under minimal RDFs.

For the class $f(v) = 0$8 (graphs where $f(v) = 0$9 for all $u$0), every vertex $u$1 receives $u$2 in any minimum RDF. Deleting all edges incident to any such vertex $u$3 increases $u$4 by at least one. Therefore, for $u$5,

$u$6

with equality for many trees (Samodivkin, 2015). Constructive characterizations for $u$7-trees involve iterated operations starting from $u$8 and preserving independent minimum dominating sets of certain private neighbor cardinality.

Algorithmically, the following complexity results are established:

• Computing $u$9 is NP-hard, even for bipartite graphs (Bahremandpour et al., 2012, Hu et al., 2011).
• The Roman bondage decision problem (is $f(u) = 2$0?) is also hard, via reductions from classic NP-complete problems (notably 3SAT).

5. Roman Bondage Number for Surfaces and Dense Graphs

For graphs with embeddings in surfaces of non-negative Euler characteristic ($f(u) = 2$1), the Roman bondage number is bounded above by $f(u) = 2$2. This is notable for planar ($f(u) = 2$3) and toroidal ($f(u) = 2$4) graphs, showing that relatively few edge removals can always increase the Roman domination number in such topologies (Samodivkin, 2014).

In dense graphs, families constructed via attachments of paths to every vertex yield Roman bondage numbers linear in degree and order. For example, the bound $f(u) = 2$5 is shown to be sharp on infinite subfamilies constructed by $f(u) = 2$6-attachments (Samodivkin, 2014).

6. Open Problems and Research Directions

The study of the Roman bondage number gives rise to several outstanding questions:

• Characterize all connected graphs with $f(u) = 2$7 (Bahremandpour et al., 2012).
• Determine classes for which $f(u) = 2$8, which remains open in general.
• Explore algorithmic approaches for approximation or parameterized computation of $f(u) = 2$9, motivated by established hardness results.
• Complete the classification of Roman bondage numbers for trees outside $0$00, unicyclic graphs, and specific extremal families.
• Strengthen bounds relating $0$01, $0$02, and other structural parameters (edge or vertex connectivity, girth, minimum or average degree).

These directions point toward deeper structural insights and finer algorithmic theory for edge vulnerability in domination-type invariants (Samodivkin, 2015, Bahremandpour et al., 2012, Samodivkin, 2014).