--- title: Triangle-Covered Problem Overview url: https://www.emergentmind.com/topics/triangle-covered-problem type: topic --- # Triangle-Covered Problem Overview Searching arXiv for recent and foundational papers on “triangle-covered” and related triangle covering formulations. In the arXiv literature, “triangle-covered problem” does not denote a single canonical problem. In the most direct graph-theoretic sense, a graph \(G\) is **triangle-covered** if every vertex of \(G\) belongs to at least one triangle, and the associated optimization problem asks for a minimum set of non-edges whose addition makes \(G\) triangle-covered [2509.11448]. Closely related usages concern covering all triangles by edges, arcs, or cliques; forcing every vertex of a \(3\)-uniform hypergraph to lie in a prescribed triangle-like configuration; and planar translation-covering problems in which triangles themselves are the covering objects [1806.08809] [1605.01816] [2307.01647] [2603.12495]. This suggests a family of triangle-centered covering problems rather than a single formalism. ## 1. Scope of the term The literature separates several non-equivalent notions according to what is being covered and what serves as a cover. In graph modification, the target property is local vertex participation in a triangle. In covering-and-packing theory, the target is the family of all triangles, which must be hit by edges, arcs, or cliques. In hypergraph covering thresholds, every vertex must lie in a copy of a fixed \(3\)-uniform “triangle.” In planar geometry, one seeks small convex regions or few homothetic unit triangles that cover all admissible triangles or arcs. | Formulation | Defining condition | Representative paper | |---|---|---| | Triangle-covered graph | every vertex of \(G\) belongs to at least one triangle | [2509.11448] | | Triangle edge cover | a subset of \(E(G)\) intersects each triangle of \(G\) | [1605.01816] | | Directed triangle cover | a set of arcs meets all directed triangles in a directed multigraph | [1806.08809] | | Triangle clique cover | a set of cliques covers all copies of \(K_3\) in \(G\) | [1709.01590] | | \(F\)-covering in \(3\)-graphs | every vertex lies in a copy of \(F\) | [2307.01647], [2212.03718] | | Planar triangle covering | translated or homothetic triangles cover prescribed shapes or curves | [2605.04111], [2603.12495], [2606.14625] | This multiplicity of meanings is structurally important. In some settings “triangle-covered” is a target graph property; in others it is a hitting-set problem; in still others it is an extremal threshold question. ## 2. Triangle-covered graphs as an edge-addition problem For a connected graph \(G\), a subset \[ F \subseteq \binom{V(G)}{2}\setminus E(G) \] is a **\(\Delta\)-completion set** if \(G+F\) is triangle-covered, and \(\Delta_G\) denotes the size of a minimum \(\Delta\)-completion set. The corresponding decision problem is: given a graph \(G\) and an integer \(t\ge 0\), does \(G\) have a \(\Delta\)-completion set of size at most \(t\)? The paper also uses the term **unsaturated vertex** for a vertex that does not belong to any triangle [2509.11448]. The structural theory begins with extremal connected triangle-covered graphs. If \[ n=3q+r,\qquad 0\le r<3, \] and \[ \epsilon_r=\frac{r(5-r)}{2}, \] then every connected triangle-covered graph \(G\) of order \(n>3\) satisfies \[ |E(G)| \ge 4q-1+\epsilon_r. \] Equivalently, \[ |E(G)|\ge \begin{cases} 4q-1,& n=3q,\\[2mm] 4q+1,& n=3q+1,\\[2mm] 4q+2,& n=3q+2. \end{cases} \] The bound is tight for every \(n\ge 3\), and for \(n\ge 8\) the extremal connected graphs are exactly the graphs in the family \(\mathcal G(n)\), obtained from small bags arranged in a tree-like decomposition [2509.11448]. The central algorithmic lemma states that when every component has at least three vertices, there exists a minimum \(\Delta\)-completion set \(F\) such that for every added edge \(uv\in F\), \[ N_G(u)\cap N_G(v)\neq \emptyset, \] equivalently \(\mathsf{dist}_G(u,v)=2\). In optimal solutions, every added edge may therefore be assumed to close a length-\(2\) path into a triangle. This yields an exact reduction to Set Cover: if \(U\) is the set of unsaturated vertices and \(uv\) ranges over non-edges with \(\mathsf{dist}(u,v)=2\), one defines \[ S_{uv}=\{\, w\in U : uw\in E \text{ and } vw\in E\,\}\cup (\{u,v\}\cap U), \] and then \(\Delta_G\) equals the minimum set cover size of \(\{S_{uv}\}\). Consequently, greedy Set Cover gives a polynomial-time \((\ln n+1)\)-approximation algorithm for graphs of order \(n\) whose components all have at least three vertices [2509.11448]. The complexity landscape is negative in general. The triangle-covered problem is \(\mathbb{NP}\)-complete, remains \(\mathbb{NP}\)-complete on connected bipartite graphs, and admits no polynomial-time constant-factor approximation unless \(P=NP\). At the same time, several classes admit exact analysis. For trees, \[ \Delta_{P_n}=\left\lceil \frac{n}{3}\right\rceil,\qquad \Delta_{S_n}=\left\lfloor \frac{n}{2}\right\rfloor, \] and \[ \min_{T\text{ tree of order }n}\Delta_T = \left\lceil \frac{n}{3}\right\rceil,\qquad \max_{T\text{ tree of order }n}\Delta_T = \left\lceil \frac{n}{2}\right\rceil. \] There is a linear-time exact algorithm for trees, and for chordal graphs one has \[ \Delta_G=\sum_{i=1}^m \Delta_{T_i'} \] after decomposition into the tree components \(T_i'\), giving an \(\mathcal O(n+m)\)-time algorithm. For random graphs \(\mathbb G(n,p)\), the threshold for being triangle-covered occurs at \(n^{-2/3}\) [2509.11448]. ## 3. Covering all triangles in graphs and directed graphs A different classical problem does not aim to make every vertex lie in a triangle; instead it asks for a minimum edge set meeting every triangle. For a simple graph \(G=(V,E)\), a **triangle cover** is a subset \(C\subseteq E\) that intersects every triangle of \(G\). The minimum size of such a set is \(\tau_t(G)\), and the maximum number of pairwise edge-disjoint triangles is \(\nu_t(G)\). Tuza’s conjecture asserts \[ \tau_t(G)\le 2\nu_t(G). \] One polynomial-time sufficient-condition theorem states that if \(G\) is irreducible and at least one of \[ \nu_t(G)/|\mathscr T_G| \ge 1/3,\qquad \nu_t(G)/|E| \ge 1/4,\qquad |E|/|\mathscr T_G| \ge 2 \] holds, then a triangle cover of size at most \(2\nu_t(G)\) can be found in polynomial time [1605.01816]. The same paper translates triangle covering to transversals in the triangle hypergraph \[ \mathcal H_G=(E,\mathscr T_G), \] where vertices are graph edges and hyperedges are triangles. In this translation, \[ \tau_t(G)=\tau(\mathcal H_G), \qquad \nu_t(G)=\nu(\mathcal H_G). \] Its main hypergraph tool is a feedback-vertex-set bound for linear \(3\)-uniform hypergraphs, \[ \tau_{c,\mathcal V}(\mathcal H)\le \frac{|\!|\mathcal H|\!|}{3}, \] together with exact solvability on acyclic hypergraphs. The point of the reduction is that many triangle-covering arguments become cycle-breaking arguments in linear \(3\)-uniform hypergraphs [1605.01816]. The directed analogue replaces edges by arcs and ordinary triangles by directed \(3\)-cycles. For a directed multigraph \(D\), the **directed triangle packing number** \(\nu_c(D)\) is the maximum size of a family of pairwise arc-disjoint directed triangles, and the **directed triangle covering number** \(\tau_c(D)\) is the minimum size of a set of arcs \(Y\subseteq E(D)\) such that \(D-Y\) has no directed triangle. The main theorem proves that if \(D\) has at least one directed triangle, then \[ \tau_c(D) < 2\nu_c(D). \] Equivalently, if \(D\) has at most \(t\) pairwise arc-disjoint directed triangles, then there exists a set of fewer than \(2t\) arcs meeting all directed triangles, except in the trivial case \(t=0\). The proof is an induction on \(|V(D)|\) using an auxiliary network and Menger’s theorem. The paper also formulates the stronger conjecture \[ \tau_c(D)\le \frac32 \nu_c(D), \] motivated by the rotational \(5\)-tournament \(T_5\), for which \[ \tau_c(T_5)=3,\qquad \nu_c(T_5)=2 \] [1806.08809]. A broader hypergraph generalization due to Aharoni and Zerbib asks for a minimum family of \((t-1)\)-subsets meeting every \(t\)-edge by inclusion. For \(H' = H^{(t-1)}\), the paper proves \[ \tau(H') \le \left(\frac t2 + 2\sqrt{t\ln t}\right)\tau^*(H'), \] giving a factor \(t/2+O(\sqrt{t\log t})\) approximation via LP rounding. This is not the classical graph triangle-cover problem itself, but it is an algorithmic generalization motivated by it [2008.07344]. An algorithmic variant of triangle covering by edges appears in exact triangle counting. A **cover-edge set** is an edge set \(S\subseteq E\) such that every triangle contains at least one edge in \(S\). If \(L(v)\) denotes BFS levels, then the horizontal edges \[ S=\{(u,v)\in E : L(u)=L(v)\} \] form a valid cover-edge set. Every triangle contains either one or three horizontal edges. This permits exact triangle counting by intersecting neighborhoods only over \(S\), rather than over all edges [2210.00389]. ## 4. Hypergraph covering thresholds for generalized and linear triangles In \(k\)-uniform hypergraphs, an **\(F\)-covering** means that every vertex lies in a copy of \(F\). For \(1\le i\le k-1\), \[ c_i(n,F)=\max\{\delta_i(G): G \text{ is an } n\text{-vertex } k\text{-graph with no }F\text{-covering}\}. \] Thus \(c_i(n,F)+1\) is the minimum \(i\)-degree condition forcing an \(F\)-covering [2307.01647] [2212.03718]. For the \(3\)-uniform **generalized triangle** \[ V(T)=\{v_1,v_2,v_3,v_4,v_5\},\qquad E(T)=\bigl\{\{v_1,v_2,v_3\},\{v_1,v_2,v_4\},\{v_3,v_4,v_5\}\bigr\}, \] the exact codegree threshold is \[ c_2(n,T)= \begin{cases} 1, & \text{when } n\in[5,10],\\[4pt] 2, & \text{when } n\ge 11 \text{ and } n-1\equiv 0 \pmod 3,\\[4pt] 1, & \text{when } n\ge 11 \text{ and } n-1\equiv 1,2 \pmod 3. \end{cases} \] For the degree threshold, \[ \binom{\lfloor n/2\rfloor}{2}\le c_1(n,T)\le \binom{\lfloor n/2\rfloor}{2}+\frac{n-3}{2}, \] so \[ c_1(n,T)=\frac14 n^2+O(n). \] The same paper distinguishes the three positions in which a vertex can lie in a copy of \(T\), denoted \(T_1,T_2,T_3\), and proves stronger sufficient minimum-degree conditions for forcing these specific positions [2307.01647]. For the \(3\)-uniform **linear triangle** \[ V(C_6^3)=\{v_1,v_2,v_3,v_4,v_5,v_6\},\qquad E(C_6^3)=\{v_1v_2v_3,\; v_3v_4v_5,\; v_5v_6v_1\}, \] the exact codegree threshold collapses to \[ c_2(n,C_6^3)=1 \qquad (n\ge 6). \] The asymptotic degree threshold is \[ \frac{3-2\sqrt{2}}{4}n^2-n < c_1(n,C_6^3) < \frac{3-2\sqrt{2}}{4}n^2+3n^{3/2}, \] hence \[ c_1(n,C_6^3)=\left(\frac{3-2\sqrt2}{4}+o(1)\right)n^2. \] The proof uses link-graph forbidden-configuration arguments for the exact codegree statement and a refined counting analysis for the degree statement [2212.03718]. A related extremal graph problem controls tetrahedron coverings in \(3\)-graphs. If \(\tau(\rho)\) denotes the asymptotically smallest possible maximum scaled triangle-degree among \(n\)-vertex graphs of edge density at least \(\rho\), then the paper proves an upper-bound construction \[ \tau(\rho)\le \begin{cases} 3\rho-\frac32,& \frac12\le \rho\le \frac{11}{18},\\[0.6em] \rho-\frac49,& \frac{11}{18}\le \rho\le \frac23, \end{cases} \] and conjectures that these bounds are tight. It further shows that this conjecture implies \[ c_1(K_4^{(3)})=\frac{19}{27}, \] and proves \[ \frac{19}{27}\le c_1(K_4^{(3)})\le \frac{19}{27}+7.4\times 10^{-9} \] [1901.09560]. ## 5. Geometric triangle covering in the plane In geometric set cover, triangles appear as covering objects rather than as target subgraphs. The abstract of “Improved Approximation Algorithms for Geometric Set Cover” states that the paper gives constant-factor approximation algorithms for covering by similar-sized fat triangles in \(\mathbb R^2\), and improved approximation guarantees for fat triangles of arbitrary size [0501045]. A different line studies exact thresholds for covering larger triangles by homothetic unit triangles. Let \(T_{n+d}\) denote a triangle with selected side length \(n+d\), where \(d\in(0,1)\). The paper proves that \(T_{n+d}\) can be covered by \(n^2+2p\) homothetic unit triangles if and only if \[ d\le \frac{p}{n+1},\qquad 1\le p\le n, \] and by \(n^2+2p+1\) homothetic unit triangles if and only if \[ d\le \frac{p}{n},\qquad 1\le p0\); - \(n^2+2\): exact threshold \(d\le \frac{1}{n+1}\); - \(n^2+3\): exact threshold \(d\le \frac{1}{n}\); - \(n^2+k\), \(4\le k\le 2n\): exact parity-dependent thresholds above [2605.04111]. The paper “Triangle covering problems and the Viterbo inequality in the plane” reformulates a symplectic-capacity question as a planar covering problem. For a convex polygon \(Q\subset \mathbb R^2\), a convex shape \(K\) is a **\(Q\)-cover** if every \(Q\)-normal triangle fits into \(K\). The main reduction proves that for polygonal \(Q\), \[ K\text{ is a \(Q\)-cover} \quad\Longleftrightarrow\quad c(K\times Q)\ge 1. \] In dimension \(2\), Viterbo’s inequality for lagrangian products becomes \[ \operatorname{area}(K)\operatorname{area}(Q)\ge \frac12\, c(K\times Q)^2, \] so for \(Q\)-covers it predicts \[ \operatorname{area}(K)\ge \frac{1}{2\,\operatorname{area}(Q)}. \] The paper proves this statement when \(Q\) is any quadrilateral, and explains the Haim–Kislev–Ostrover counterexample in the regular pentagon case [2603.12495]. Wetzel’s conjecture belongs to the same translation-covering tradition. The paper “Wetzel’s 30-60-90 Triangle Covers Unit Arcs” proves that a specific \(30^\circ\!-\!60^\circ\!-\!90^\circ\) triangle \(T\) covers every unit arc in the plane. More strongly, the homothetic copy \(T/1.0048\) still covers every unit arc, and its area \[ 0.260956\ldots \] is below \[ \pi/12 \approx 0.261799 \] [2606.14625]. ## 6. Clique covers, polygon covers, and geometric-triangle recognition The **triangle clique cover problem** asks for a set of cliques covering every copy of \(K_3\). Formally, a set \(C\) of cliques is a \(K_3\) clique cover of \(G\) if for every triangle \(S\subset V(G)\) there is a clique \(Q\in C\) with \(S\subseteq Q\); the minimum size is \(\theta_{K_3}(G)\). The paper proves the exact extremal bound \[ \theta_{K_3}(G)\le k_3(T(n,3)) \] for every graph \(G\) on \(n\) vertices, with equality if and only if \[ G\cong T(n,3). \] It also proves that the decision problem \(\mathrm{KCC}(t)\) is NP-complete for every fixed \(t\ge 1\), hence in particular for \(t=3\), while giving an exact polynomial-time algorithm for the weighted \(K_t\) clique cover problem on semichordal graphs [1709.01590]. A much harder exact covering problem arises for polygons. The decision problem \[ \text{“given a simple polygon }\mathcal P\text{ and integer }k,\text{ can }\mathcal P\text{ be covered by }k\text{ triangles?”} \] is \(\exists\mathbb{R}\)-complete. The paper obtains this as a consequence of the \(\exists\mathbb{R}\)-completeness of Minimum Convex Cover and shows that, for the constructed hard instances, if a cover exists then there also exists one consisting entirely of triangles. It further implies that, assuming the widespread belief that \(\mathsf{NP}\neq\exists\mathbb{R}\), the problem is not in \(\mathsf{NP}\) [2106.02335]. The term “triangle” also appears in representation problems. A **simple-triangle graph** is the intersection graph of a family of triangles spanned by a point on one horizontal line and an interval on another horizontal line. Recognition reduces to a **restricted 2-chain subgraph cover** problem on a bipartite graph \(G\) with a forbidden edge set \(F\): find a 2-chain subgraph cover such that one chain subgraph has no edges in \(F\). The paper gives a polynomial-time algorithm with running time \[ O(m\hat m + \min\{m^2,\hat m(\hat m+f)\}), \] thereby yielding a simpler recognition algorithm for simple-triangle graphs [1605.07322]. Taken together, these variants show that triangle-centered covering problems occupy several different complexity classes. Some admit exact linear-time algorithms on structured graph classes; some have logarithmic approximation algorithms but no constant-factor approximation on general graphs; some are NP-complete; and some are \(\exists\mathbb{R}\)-complete. A plausible implication is that “triangle-covered problem” is best understood as a cluster of local-density, hitting-set, translation-covering, and geometric-realization problems, linked by the common role of the triangle as the minimal nontrivial \(3\)-cycle or \(3\)-vertex simplex. Source: https://www.emergentmind.com/topics/triangle-covered-problem