--- title: Hitting Geodesic Intervals in Graphs url: https://www.emergentmind.com/topics/hitting-geodesic-intervals type: topic --- # Hitting Geodesic Intervals in Graphs Searching arXiv for the primary graph-theoretic and related uses of “Hitting Geodesic Intervals” to ground the article in current preprints. Hitting Geodesic Intervals is a graph-theoretic shortest-path hitting problem in which the targets are geodesic intervals rather than arbitrary vertex sets. For vertices \(u,v\) in a graph \(G=(V,E)\), the geodesic interval \[ I_G[u,v]=\{x\in V : d_G(u,v)=d_G(u,x)+d_G(x,v)\} \] is the set of all vertices that lie on at least one shortest \(u\)-\(v\) path, equivalently the union of all shortest \(u\)-\(v\) paths. Given a graph \(G\), a set \(T\) of terminal pairs, and an integer \(k\), the decision problem asks whether there exists a set \(S\subseteq V\) with \(|S|\le k\) such that \(S\cap I_G[u,v]\neq \emptyset\) for every \(\{u,v\}\in T\). Introduced by Aravind and Saxena and substantially extended in "Hitting Geodesic Intervals in Structurally Restricted Graphs" [2509.01413], the problem now serves as a focal point for sharp contrasts between tractable and intractable shortest-path hitting regimes. ## 1. Definition and combinatorial meaning The central object is the geodesic interval \(I_G[u,v]\), defined by the distance identity \[ I_G[u,v]=\{x\in V : d_G(u,v)=d_G(u,x)+d_G(x,v)\}. \] This is the interval analogue of a shortest-path region: every vertex in \(I_G[u,v]\) lies on some shortest \(u\)-\(v\) path, and every shortest \(u\)-\(v\) path is contained in \(I_G[u,v]\) [2509.01413]. The Hitting Geodesic Intervals problem therefore asks for a small vertex set \(S\) that intersects each prescribed interval \(I_G[u,v]\). Because \(I_G[u,v]\) is the union of all shortest \(u\)-\(v\) paths, intersecting the interval means “hit at least one shortest path,” not “hit every shortest path individually.” This existential interpretation is fundamental: the requirement is over intervals, not over the family of shortest paths as separate objects [2509.01413]. The paper also considers a weighted analogue, but the main structural hardness statements are for the unweighted problem. In both variants, the central algorithmic question is how shortest-path geometry interacts with graph structure: the intervals are metric objects, but the hitting set is purely combinatorial. ## 2. Origins and the pre-2025 landscape Aravind and Saxena introduced the problem under the name Terminal Monitoring Set. Before the 2025 preprint, the known parameterized picture already contained a mix of positive and negative results: W[2]-completeness parameterized by \(k\); fixed-parameter tractability for \(k+\) cluster vertex deletion number and for \(k+\) neighborhood diversity; and, in the weighted setting, fixed-parameter tractability for feedback edge set number and vertex cover number [2509.01413]. The 2025 paper extends those results “in both negative and positive directions” and organizes them around structural graph parameters. Its main theme is that small perturbations of apparently simple graph classes do not collapse the problem, whereas certain separator-based or modular decompositions do. A main theme is a sharp contrast between closely related structural parameters: multiway cut of terminal vertices yields fixed-parameter tractability, while multicut of terminal pairs yields W[2]-completeness [2509.01413]. This places Hitting Geodesic Intervals alongside other shortest-path covering problems in parameterized complexity, but with a particularly delicate interaction between metric convexity and structural sparsity. The intervals are geodesic in the graph metric, yet even highly restricted graph classes retain enough shortest-path flexibility to encode hard constraint systems. ## 3. NP-completeness on highly restricted graph classes A striking contribution of the 2025 paper is that NP-completeness survives under extremely severe structural restrictions [2509.01413]. The constructions are reductions from 3-Coloring, and their common design principle is to encode one of three states inside a constant-size gadget while using geodesic intervals between gadgets to enforce consistency. The first theorem shows NP-completeness on graphs of vertex-deletion distance \(1\) to the class of disjoint unions of 5-vertex paths. Deleting one special vertex \(s^*\) leaves a disjoint union of copies of \(P_5\). For each original vertex \(v\), the reduction creates a path \[ v_1-v_1'-v_2-v_3'-v_3, \] adds \(s^*\), and uses local terminal pairs \(\{v_1,v_1'\}\) and \(\{v_3,v_3'\}\) to force exactly two selected vertices per gadget. Those two vertices encode one of three colors, while cross-gadget terminal pairs such as \(\{u_1,v_1\}\), \(\{u_2,v_2\}\), and \(\{u_3,v_3\}\) use shortest paths through \(s^*\) to forbid monochromatic adjacent vertices [2509.01413]. A modification of the same construction yields NP-completeness on graphs of vertex-deletion distance \(1\) to the class of paths. The gadgets are linked so that after deleting one vertex the graph becomes a single path, while the relevant geodesic intervals are preserved. This shows hardness already on graphs that are “one vertex away from a path” [2509.01413]. A second 3-coloring construction proves NP-completeness on graphs of vertex-deletion distance \(1\) to the class of disjoint unions of triangles. Here each original vertex is represented by a triangle \(C_v\) on \(v_1,v_2,v_3\), together with a universal vertex \(s^*\). Local triangle pairs force selecting exactly two triangle vertices, and the missing index determines the color. If adjacent original vertices omit the same index \(x\), then \[ I_H[u_x,v_x]=\{u_x,s^*,v_x\}, \] and no selected vertex can hit that interval. This construction improves the vertex-integrity upper bound of the hard instances to \(4\) [2509.01413]. Finally, replacing the universal vertex by a long path yields NP-completeness on graphs of bandwidth \(4\) and maximum degree \(5\). The role of the hub is simulated by local vertices \(s_v^*\) attached to each triangle and linked in a path; the instances remain equivalent because a size-\(k\) solution never needs to use any \(s_v^*\). The resulting graph has maximum degree \(5\) and an ordering witnessing bandwidth \(4\) [2509.01413]. These results are important because they rule out parameterized algorithms for many structural parameters if the solution size \(k\) is not part of the parameter. Hardness already at distance \(1\) from a path forest, distance \(1\) from a path, or vertex integrity \(4\) shows that shortest-path geometry alone does not enforce tractability. ## 4. Fixed-parameter tractability via modular and separator structure The positive results rely on two distinct mechanisms: modular decomposition and separator localization [2509.01413]. For parameter \(k+\) modular-width, the paper gives an \(O^*(2^{p+k})\)-time algorithm, where \(p\le \operatorname{mw}(G)\). The algorithm first branches over the set of modules intersected by the solution. After updating each terminal constraint \(J(u,v)\), a key structural fact is that if \(u\) and \(v\) lie in different modules, then the residual interval satisfies \[ J(u,v)\subseteq \{u,v\}. \] Hence the inter-module part reduces to a family of sets of size at most \(2\). The algorithm enumerates all minimal hitting sets of this size-2 family using Damaschke’s enumeration for minimal vertex covers of size at most \(k\), and then observes that every remaining nonempty family inside a module has a hitting set of size \(1\). This yields fixed-parameter tractability for \(k+\) modular-width [2509.01413]. The more general engine is a weighted theorem parameterized by \(k+p+q\). One is given a set \(Z\subseteq V\) with \(|Z|\le p\) such that every connected component of \(G-Z\) contains at most \(q\) terminal vertices. The key localization lemma states that for \(u,v\in V\setminus Z\), \[ I_G[u,v]\setminus\left(Z\cup \bigcup_{\{z,z'\}\in \binom Z2} I_G[z,z']\right)\subseteq V(C_u)\cup V(C_v), \] where \(C_u\) and \(C_v\) are the components of \(G-Z\) containing \(u\) and \(v\). Thus, after branching on which pairs in \(\binom Z2\) are hit, every interval from \(T\) intersects at most two components [2509.01413]. The algorithm then compresses vertices by \(T\)-equivalence: two vertices are equivalent if they interrupt exactly the same terminal pairs. Inside a component, equivalence is controlled by the pairs in \[ \binom{Z\cup (V(T)\cap V(C))}{2}, \] so each component meets at most \[ 2^{\binom{p+q}{2}} \] equivalence classes. Consequently each interval has bounded size \[ d=2^{\binom{p+q}{2}+1}. \] At that point the residual subfamily is a \(d\)-Hitting Set instance, and a sunflower reduction either rejects or replaces it by an equivalent family of at most \[ k^d\cdot d! \] sets. The remaining instance is then fixed-parameter tractable because its number of sets depends only on \(k,p,q\) [2509.01413]. Two corollaries are especially significant. First, weighted Hitting Geodesic Intervals is fixed-parameter tractable parameterized by \(k+\) vertex integrity. If \(\iota\) is the vertex integrity, a corresponding separator \(X\) can be found in time \[ O(\iota^{\iota+1} n), \] with \(|X|\le \iota\) and every component of \(G-X\) containing at most \(\iota\) vertices, hence at most \(\iota\) terminals [2509.01413]. Second, the weighted problem is fixed-parameter tractable parameterized by the minimum vertex multiway-cut size of the terminal set \(V(T)\), since then one may take \(q=1\) [2509.01413]. ## 5. Pairwise terminals, separator contrasts, and the current complexity map The 2025 paper identifies two hardness frontiers that clarify the problem’s parameterized status [2509.01413]. The first is the contrast between multiway cut and multicut. The weighted problem is fixed-parameter tractable when parameterized by the minimum vertex multiway-cut size of the terminal vertices \(V(T)\), but the unweighted problem is W[2]-complete parameterized by the minimum vertex multicut size of the terminal pairs \(T\). The hardness proof is a reduction from Hitting Set. Given a family \(\mathcal F\), the construction creates set vertices \(F_i\), element vertices \(u_j\), and a set \(W=\{w_i:i\in[h+3]\}\), with terminal pairs \[ T=\{\{F_i,w_j\}: i\in[m],\ j\in[h+3]\}. \] The set \(W\) is a vertex multicut for \(T\) of size \(h+3\), so the parameter is bounded. Because \(|S|\le h<|W|\), some \(w_p\) is unselected, and then each pair \(\{F_i,w_p\}\) forces the solution to contain an element adjacent to \(F_i\), recovering a hitting set for the original instance [2509.01413]. The second frontier concerns the special case \[ T=\binom Q2 \] for some \(Q\subseteq V\), called Pairwise Hitting Geodesic Intervals. This remains W[2]-complete parameterized by \(k\). The reduction augments the previous gadget by adding vertices \(F^*\), \(w^*\), and \(X=\{x_i:i\in[h+3]\}\), then sets \(Q=V(H)\setminus U\). Any solution of size \(h+2\) must contain both \(F^*\) and \(w^*\), because otherwise all \(h+3\) leaves on one side would need to be selected. Removing them leaves a size-\(h\) set that still hits all original \(\{F_i,w_j\}\) pairs, so the restriction to all pairs inside \(Q\) does not simplify the problem [2509.01413]. The resulting complexity picture is sharply stratified: | Parameter / setting | Status | Source | |---|---|---| | \(k\) | W[2]-complete | background in [2509.01413] | | \(k+\) modular-width | FPT | [2509.01413] | | \(k+\) vertex integrity (weighted) | FPT | [2509.01413] | | minimum vertex multiway-cut size of \(V(T)\) (weighted) | FPT | [2509.01413] | | minimum vertex multicut size of \(T\) | W[2]-complete | [2509.01413] | | \(T=\binom Q2\), parameter \(k\) | W[2]-complete | [2509.01413] | The paper explicitly notes that many “\(k+\) structural parameter” cases remain open, including treewidth, pathwidth, feedback vertex set, distance to path forest, and bandwidth in the plus-\(k\) regime. It also isolates an open separator problem: given \(G\), a terminal set \(R\), and integers \(p,q\), decide whether there exists \(Z\subseteq V\), \(|Z|\le p\), such that every component of \(G-Z\) contains at most \(q\) terminals. The paper obtains a nonuniform fixed-parameter consequence via graph minors, but asks for a uniform fixed-parameter algorithm [2509.01413]. ## 6. Other mathematical uses of “hitting geodesic intervals” The phrase “hitting geodesic intervals” also appears in several unrelated literatures, where it denotes different objects. In hyperbolic integral geometry, the relevant question is the invariant measure of \(\lambda\)-geodesic hyperplanes intersecting an ordinary geodesic segment. For a segment \(\ell(L)\) of hyperbolic length \(L\) starting at the origin in \(\mathbb H^d\), the paper "Seeing Through Hyperbolic Space: Visibility for \(\lambda\)-Geodesic Hyperplanes" proves \[ \nu_\lambda^o\bigl(\{H\in\mathrm{Hyp}_\lambda^o:\ H\cap \ell(L)\neq\varnothing\}\bigr) = \frac{\Gamma(\frac d2)}{2\sqrt{\pi}\Gamma(\frac{d+1}{2})}\,L, \] for every \(0\le \lambda\le 1\). The constant is independent of \(\lambda\), even though for \(\lambda>0\) a \(\lambda\)-geodesic hyperplane can intersect the segment in \(0\), \(1\), or \(2\) points [2602.20935]. In first-passage percolation, the closest analogue is not a graph geodesic interval but a local ambient-lattice window. "Geodesics in first-passage percolation cross any pattern" studies valid patterns \(\mathcal P\) and proves that, apart from an event with exponentially small probability, every long point-to-point geodesic encounters such patterns linearly often: \[ \mathbb P\Big(\exists \text{ a geodesic }\gamma \text{ from }0\text{ to }x \text{ such that }N_{\mathcal P}(\gamma)