---
title: Vertex Cover Irredundance
url: https://www.emergentmind.com/topics/vertex-cover-irredundance
type: topic
---

# Vertex Cover Irredundance

Searching arXiv for recent and foundational papers on vertex cover irredundance and extension variants.
Vertex cover irredundance is the study of vertex subsets that cover edges in a graph in a nonredundant manner, in the sense that every chosen vertex is justified by at least one edge that would become uncovered if that vertex were removed. In its classical form, this is the private-edge characterization of minimal vertex covers. More recent work places the notion into a universal irredundance framework based on blocking sets, where edges serve as the natural blockers for vertex cover, yielding lower and upper vertex cover irredundance parameters and connecting them to domination-theoretic chains and TAR reconfiguration invariants [2509.18486]. A complementary algorithmic line studies irredundance through extension problems: given a prescribed set \(U\), one asks whether \(U\) can be embedded into a minimal vertex cover, thereby treating irredundance as a feasibility constraint rather than merely a structural property [1810.04629].

## 1. Definitions and core characterization

Let \(G=(V,E)\) be a graph. A set \(S \subseteq V\) is a vertex cover if every edge has at least one endpoint in \(S\). A vertex cover is inclusion-wise minimal if no proper subset of it is a vertex cover. The decisive characterization is the private-edge criterion:
\[
S \text{ is a minimal vertex cover } \Longleftrightarrow \forall v \in S,\ \exists u \in N_G(v)\text{ with }u \notin S.
\]
Equivalently, each \(v \in S\) has an incident edge \(vu\) such that \(u \notin S\), so that edge is covered uniquely by \(v\). In the language of irredundance, every vertex in the cover is irredundant because it contributes a private edge [1810.04629].

The universal blocking-set formulation abstracts this phenomenon. For vertex cover, the natural blocking family is the set of edges,
\[
B_{\tau}(G)=\{e \in E(G)\},
\]
viewing each edge as a two-element subset of \(V(G)\). A set \(S\) is then vertex-cover irredundant if every \(v \in S\) has a private blocker, namely an edge \(e\) with \(e \cap S=\{v\}\). This reproduces the classical private-edge notion exactly [2509.18486].

Two numerical parameters arise. The lower vertex cover irredundance number is
\[
\operatorname{ir}_{\tau}(G):=\vcir(G),
\]
the minimum size of a maximal vertex-cover-irredundant set, and the upper vertex cover irredundance number is
\[
\operatorname{IR}_{\tau}(G):=\VCIR(G),
\]
the maximum size of a maximal vertex-cover-irredundant set [2509.18486].

A central fact is that the lower parameter collapses to the ordinary vertex cover number:
\[
\operatorname{ir}_{\tau}(G)=\tau(G)
\]
for all graphs [2509.18486]. Thus the lower irredundance notion does not introduce a new minimization invariant for vertex cover, whereas the upper parameter does.

## 2. Relation to minimal vertex cover and independent set duality

The complementarity between vertex covers and independent sets is fundamental. A set \(S\) is an independent set if \(S \cap N_G(S)=\emptyset\), and an independent set is maximal if no proper superset remains independent. The complement of a maximal independent set is a minimal vertex cover; consequently, extension and irredundance statements for one side can be translated to the other [1810.04629].

This duality yields the identity
\[
\alpha(G)+\tau(G)=|V(G)|,
\]
and under the extension formalism it gives:
\[
(G,U) \text{ is a yes-instance of Ext VC iff } (G,V\setminus U) \text{ is a yes-instance of Ext IS}
\]
[1810.04629].

In the irredundance setting, duality clarifies the distinction between minimality and maximality. Minimal vertex cover is an inclusion condition certified locally by private edges; maximal independent set is the complementary inclusion condition certified by domination. The extension problem exploits this correspondence to transform minimal-cover feasibility into independent-set feasibility on a derived subgraph. This makes irredundance a bridge between covering and domination phenomena rather than a purely local covering condition [1810.04629].

The more recent universal theory further integrates vertex cover irredundance with domination irredundance. The Extended Domination Chain states
\[
\operatorname{dir}(G) \le \gamma(G) \le \underline{\alpha}(G) \le \alpha(G) \le \Gamma(G) \le \operatorname{DIR}(G) \le \operatorname{VCIR}(G),
\]
placing upper vertex cover irredundance at the end of the chain [2509.18486]. This inclusion is explained by the fact that a domination-irredundant witness can induce a private edge, so every DIr-set is also VCIr.

## 3. Extension viewpoint and local witness structures

The extension problem for vertex cover is defined as follows. Given a graph \(G=(V,E)\) and a prescribed set \(U \subseteq V\), Ext VC asks whether there exists a minimal vertex cover \(S\) with \(U \subseteq S\) [1810.04629]. This formulation isolates the irredundance burden to the requirement that all vertices of \(U\) must acquire private edges in some completion of \(U\) to a full minimal cover.

A precise structural characterization is given in terms of the closed neighborhood \(N_G[U]\). The following are equivalent:

1. \((G,U)\) is a yes-instance of Ext VC.
2. \((G[N_G[U]],N_G[U]\setminus U)\) is a yes-instance of Ext IS.
3. There exists an independent dominating set \(S' \subseteq N_G[U]\setminus U\) of \(G[N_G[U]]\).

This is Theorem 2.3 in the extension paper [1810.04629]. The condition translates irredundance into the existence of witness vertices outside \(U\) that are independent and dominate \(U\). If such an \(S'\) exists, then taking the complement inside \(N_G[U]\) yields a minimal vertex cover whose vertices in \(U\) each obtain a private edge into \(S'\).

This equivalence is especially revealing conceptually. It shows that private-edge feasibility for a prescribed subset is neither purely local nor merely a matter of covering all incident edges. The witnesses assigned to distinct vertices of \(U\) must satisfy independence and domination constraints simultaneously. Conflicts arise when several vertices of \(U\) compete for witnesses in incompatible ways. This suggests that vertex cover irredundance is best viewed as a constrained witness-assignment problem on \(N_G[U]\), rather than as a simple property of \(U\) itself.

The paper also notes a structural simplification: edges inside \(U\) cannot serve as private edges for Ext VC, so for the purpose of extension one may assume \(U\) is either a clique or an independent set [1810.04629]. This observation separates feasibility from internal adjacencies of \(U\), emphasizing again that the essential obstructions lie in the outside witness structure.

## 4. Complexity landscape

Despite the tractability of ordinary Vertex Cover on many graph classes, enforcing irredundance through extension is often computationally difficult. Ext VC is NP-complete in cubic bipartite graphs [1810.04629]. Since classical Vertex Cover is polynomial on bipartite graphs, this establishes a sharp separation between optimization of cover size and feasibility of minimality under preassignment.

Hardness persists under stronger restrictions. Ext IS is NP-complete on planar bipartite subcubic graphs, and by complement duality this extends to Ext VC [1810.04629]. These results indicate that the difficulty comes from global consistency of private-edge witnesses rather than from dense or topologically unconstrained structure.

The parameterized picture with parameter \(|U|\) is similarly split. On general graphs, Ext VC is W[1]-complete, even on bipartite graphs [1810.04629]. In contrast, on planar graphs the problem is fixed-parameter tractable with respect to \(|U|\), because the relevant induced graph \(G[N_G[U]]\) has bounded outerplanarity and hence bounded treewidth [1810.04629].

The exact-exponential complexity of planar instances is near-optimal under ETH. Ext VC can be solved in time \(O(2^{O(\sqrt{n})})\) on planar graphs via treewidth-based dynamic programming that tracks both cover membership and private-edge status [1810.04629]. The same paper proves that planar Ext VC on graphs of order \(n\) has no \(2^{o(\sqrt{n})}\)-time algorithm unless ETH fails [1810.04629]. For bipartite subcubic instances, there is no \(2^{o(n+m)}\)-time algorithm unless ETH fails [1810.04629].

These results delineate a recurring phenomenon: minimality constraints can convert a classically easy covering problem into a structurally richer problem governed by domination, reconfiguration-style locality, and witness compatibility.

## 5. Algorithms, reductions, and graph classes

The extension paper develops irredundance-aware reduction and branching rules for bounded-degree instances [1810.04629]. Several reductions directly encode private-edge necessities:

- If \(U=\emptyset\), answer yes.
- If some \(u \in U\) has degree \(0\), answer no, because no private edge is possible.
- If \(x \notin U\) has degree \(0\), delete \(x\).
- If \(u,u' \in U\) with \(uu' \in E\), delete the edge \(uu'\), since it cannot be private.
- If \(u \in U\) has degree \(1\) with neighbor \(x\), then \(ux\) must be private for \(u\).
- If \(x \notin U\) satisfies \(N(x)\subseteq U\), then \(x\) can witness private edges for all its neighbors.

These rules are sound and express a common principle: irredundance can often be forced or ruled out by local degree structure [1810.04629].

For maximum degree \(\Delta\), a simple recursive algorithm that guesses a private neighbor for each vertex of \(U\) runs in \(O^*(\Delta^{|U|})\) time [1810.04629]. On subcubic graphs, more refined branching yields an exact algorithm in time \(O^*(1.26^n)\) [1810.04629]. The same framework gives a linear kernel of size \((\Delta+1)|U|\) parameterized by \(|U|\) [1810.04629].

Positive polynomial-time results occur in graph classes where the witness structure can be optimized through independent domination. Ext VC is polynomial-time decidable in chordal and in circular-arc graphs [1810.04629]. The proof reduces the problem to weighted minimum independent dominating set on \(G[N_G[U]]\), with weight \(1\) on \(U\) and \(0\) outside it; the instance is feasible precisely when the optimum weight is \(0\) [1810.04629].

Trees admit a finer forbidden-structure characterization. For a tree \(T\) and an independent set \(U \subseteq V(T)\), feasibility of Ext VC is equivalent to the absence of a particular induced edge-full black-white subtree belonging to a class \(\mathcal{T}\) constructed inductively [1810.04629]. This gives a purely combinatorial obstruction to irredundance and leads to linear-time algorithms on forests.

A separate but related algorithmic point from the universal irredundance framework is that recognition of a VCIr-set is straightforward: one verifies that the set is a vertex cover and that every chosen vertex has an incident edge to a vertex outside the set that is uniquely covered by it [2509.18486]. This is a local test, but maximality and optimization of VCIr-sets remain structurally more involved.

## 6. Numerical parameters, inequalities, and representative graph families

The universal theory gives the vertex cover interlace chain
\[
\operatorname{ir}_{\tau}(G) \le \tau(G) \le \overline{\tau}(G) \le \operatorname{IR}_{\tau}(G),
\]
specializing the general inequality
\[
\xir(G) \le X(G) \le \overline{X}(G) \le \XIR(G)
\]
to the vertex cover setting [2509.18486]. Since \(\operatorname{ir}_{\tau}(G)=\tau(G)\), the lower end collapses, but the upper inequality can be strict.

The upper parameter \(\operatorname{IR}_{\tau}(G)\) is bounded below by domination:
\[
\operatorname{IR}_{\tau}(G) \ge |V(G)|-\gamma(G),
\]
with \(\operatorname{VCIR}(G)=n-1\) iff \(\gamma(G)=1\) [2509.18486]. This highlights the domination-theoretic content of upper irredundance: large VCIr-sets are forced when small dominating sets exist.

Several standard graph families illustrate the behavior precisely.

| Graph family | Parameter values |
|---|---|
| \(K_n\) | \(\tau=\operatorname{ir}_{\tau}=\overline{\tau}=\operatorname{IR}_{\tau}=n-1\) |
| \(K_{1,m}\) | \(\tau=\operatorname{ir}_{\tau}=1\), \(\operatorname{IR}_{\tau}=m\) |
| \(K_{a,b}\), \(1 \le a \le b\), \(a\ge 2\) | \(\tau=\operatorname{ir}_{\tau}=a\), \(\overline{\tau}=b\), \(\operatorname{IR}_{\tau}=a+b-2\) |
| \(P_n\) | \(\tau(P_n)=n/2\), \(\operatorname{IR}_{\tau}(P_n)=(2n-1)/3\) |
| \(C_n\), \(n\ge 4\) | \(\tau(C_n)=n/2\), \(\operatorname{IR}_{\tau}(C_n)=(2n-2)/3\) |

These identities are all stated in the universal irredundance paper [2509.18486].

The complete bipartite example is especially instructive. In \(K_{a,b}\) with \(a \ge 2\), the upper irredundance exceeds the upper vertex cover number, showing that maximal irredundant cover-like sets can be substantially larger than minimal vertex covers [2509.18486]. This confirms that \(\operatorname{IR}_{\tau}\) captures a different extremal principle: unique coverage, not merely inclusion-wise minimality.

For disconnected graphs, isolated vertices are never part of a VCIr-set [2509.18486]. This is consistent with the private-edge criterion: an isolated vertex cannot justify its presence by covering any edge.

## 7. Optimization, reconfiguration, and generalizations

The extension framework also introduces the price of extension, which measures how closely a prescribed set can be fitted into a minimal vertex cover. For vertex cover, the relevant optimization problem is Max Ext VC: among all minimal vertex covers \(S\), maximize \(|S \cap U|\) [1810.04629]. Its dual on the independent-set side is Min Ext IS, and the optimum values satisfy
\[
\operatorname{opt}_{\text{Max Ext VC}}(G,U)+\operatorname{opt}_{\text{Min Ext IS}}(G,V\setminus U)=|V|.
\]
This turns irredundance from a decision constraint into a quantitative notion [1810.04629].

Approximability depends strongly on graph class. Max Ext VC is as hard as MaxIS to approximate on general graphs, even when \(U\) is an independent set [1810.04629]. On bipartite graphs there is a polynomial-time \(2\)-approximation, obtained by choosing the bipartition side containing more vertices of \(U\), since each side is a minimal vertex cover [1810.04629]. On graphs of maximum degree \(\Delta\), a \(\Delta\)-approximation is obtained via coloring of \(G[N_G(U)\setminus U]\) and selection of a large neighborhood-dominating color class [1810.04629]. Chordal and circular-arc graphs admit exact optimization through weighted independent domination [1810.04629].

The universal framework contributes a reconfiguration perspective. For vertex cover irredundance, the TAR graph \(\mathrm{Cir}(G)\) has as vertices all VCIr-sets, with adjacency defined by symmetric difference of size one [2509.18486]. A specialized isomorphism theorem states that if \(G\) and \(G'\) have no isolated vertices and \(\mathrm{Cir}(G)\cong \mathrm{Cir}(G')\), then \(|V(G)|=|V(G')|\) and, after relabeling, the two graphs have exactly the same VCIr-sets [2509.18486]. This makes the family of irredundant vertex covers itself recoverable from the TAR structure.

An alternate universal account uses closure operators. For vertex cover,
\[
\varphi_{VC,G}(A)=A \cup \{\,v \in V(G): N(v)\subseteq A\,\}
\]
is a VC-compliant closure operator, and the associated blocking family generated by this closure has edge set as its generator family [2509.18486]. Thus the closure-theoretic and private-edge viewpoints coincide.

Finally, the extension paper generalizes beyond \(K_2\)-cover. For a fixed biconnected graph \(H\), Ext \(H\)-cover is NP-complete, and there is a characterization of minimal \(H\)-covers extending \(U\) in terms of copies \(H_u\) intersecting \(U\) in exactly one vertex and inducing an \(H\)-free graph outside \(U\) [1810.04629]. A plausible implication is that the private-edge phenomenon for vertex cover is the simplest instance of a broader irredundance principle: each prescribed element must be certified by a witness structure that is uniquely attributable to it.

Vertex cover irredundance therefore occupies two closely related positions in contemporary graph theory. Structurally, it is the private-edge anatomy of minimal vertex covers. Universally, it is a blocking-set irredundance parameter whose lower value equals \(\tau(G)\), whose upper value extends the domination chain, and whose TAR graph encodes strong invariants [2509.18486]. Algorithmically, it is the source of a family of extension and optimization problems whose complexity sharply departs from that of ordinary Vertex Cover, even on graph classes where covering itself is easy [1810.04629].

Source: https://www.emergentmind.com/topics/vertex-cover-irredundance