Pivot-Minor: Concepts & Applications

• Pivot-minor is a graph derived via sequences of edge pivots and vertex deletions, providing a reduction method that is local and invertible.
• It offers equivalent formulations in combinatorial, matrix, and matroid languages, linking binary matroid theory with graph structural properties.
• Pivot-minor theory influences rank-width analysis, obstruction theory, and computational problems, with applications in algorithm design and quantum information.

Searching arXiv for recent and foundational papers on pivot-minors to ground the article. A pivot-minor of a graph is a graph obtained by a sequence of edge pivots and vertex deletions. The pivot operation is a binary-field analogue of a minor-like reduction: it is local, invertible up to repeated pivots, and admits equivalent descriptions in combinatorial, matrix, and matroidal language. Across graph theory, binary matroid theory, width parameters, extremal combinatorics, algorithms, and quantum information, pivot-minors provide a containment relation that is weaker than induced subgraph containment and, in general, stronger than the restrictions imposed by ordinary minor theory. They play for rank-width and related parameters a role analogous to that of graph minors for tree-width (Oum, 2020, Dabrowski et al., 2023).

1. Definition and equivalent formulations

Let $G$ be a simple graph and let $uv\in E(G)$. Define

$$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$

The pivot of $G$ on the edge $uv$, denoted $G\pivot uv$ or $G\wedge uv$, is obtained by complementing all edges between $V_1$ and $V_2$, between $V_2$ and $uv\in E(G)$0, and between $uv\in E(G)$1 and $uv\in E(G)$2, and then swapping the labels of $uv\in E(G)$3 and $uv\in E(G)$4 (Oum, 2020, Campbell et al., 31 Jul 2025). An equivalent formulation is

$uv\in E(G)$5

where $uv\in E(G)$6 denotes local complementation at $uv\in E(G)$7 (Choi et al., 2015, Lee et al., 2021).

A graph $uv\in E(G)$8 is a pivot-minor of $uv\in E(G)$9 if it can be obtained from $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$0 by a finite sequence of pivot operations on edges and deletions of vertices. Equivalently, pivot-minors are exactly the induced subgraphs of graphs that are pivot-equivalent to $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$1 (Oum, 2020). In matrix language, pivoting on $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$2 is the principal pivot transform of the $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$3 principal submatrix indexed by $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$4 in the adjacency matrix over $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$5 (Oum, 2020, Kwon et al., 2012).

This formulation makes the relation to vertex-minors immediate. Since every edge pivot can be written as a sequence of three local complementations, every pivot-minor is a vertex-minor, but the converse need not hold (Choi et al., 2015). In the bipartite case, the two notions coincide (Kwon et al., 2012). A standard toy example is that pivoting the middle edge of the path $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$6 produces $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$7, and deleting one vertex then yields $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$8; thus $$V_1 = N_G(u)\setminus (N_G(v)\cup\{v\}),\qquad V_2 = N_G(v)\setminus (N_G(u)\cup\{u\}),\qquad V_3 = N_G(u)\cap N_G(v).$$9 is a pivot-minor of $G$0 (Kim et al., 2020).

2. Binary matroids and the fundamental-graph correspondence

The central structural interpretation of pivot-minors comes from binary matroid theory. Every binary matroid $G$1 can be represented, after choosing a basis $G$2, by a matrix $G$3 over $G$4, and the bipartite graph with biadjacency matrix $G$5 is called a fundamental graph of $G$6 (Campbell et al., 31 Jul 2025). In this setting, every pivot of the fundamental graph corresponds exactly to changing basis in the matroid, and every deletion or contraction of $G$7 corresponds to a vertex-deletion in some pivot of the fundamental graph. Thus pivot-minors of the fundamental graph correspond to minors of the matroid (Campbell et al., 31 Jul 2025).

This equivalence is the mechanism by which pivot-minor theory imports deep matroid structure. The announced Geelen–Gerards–Whittle theorem states that if $G$8 is a proper minor-closed class of binary matroids, then there exist integers $G$9 such that every $uv$0-connected member of $uv$1 is a $uv$2-perturbation of either a graphic matroid or a cographic matroid; here $uv$3-perturbation means that in two representations $uv$4 and $uv$5, the difference $uv$6 has $uv$7-rank at most $uv$8 (Campbell et al., 31 Jul 2025). Translated to bipartite graphs, this yields the hypothesis that for every fixed bipartite $uv$9 there are $G\pivot uv$0 so that every $G\pivot uv$1-rank-connected bipartite graph $G\pivot uv$2 with no pivot-minor isomorphic to $G\pivot uv$3 admits partitions $G\pivot uv$4 and $G\pivot uv$5 such that each block $G\pivot uv$6 is either a graphic fundamental graph or the bipartite complement of a graphic fundamental graph (Campbell et al., 31 Jul 2025).

The same correspondence also underlies several “prime graph” results. If $G\pivot uv$7 is a fundamental graph of a binary matroid $G\pivot uv$8, then pivots correspond to base changes and vertex deletions correspond to deletions or contractions. Consequently, theorems about prime graphs under pivot-minor reduction translate into $G\pivot uv$9-connected restrictions in binary matroids (Kim et al., 2022). In particular, corollaries for bipartite graphs recover and strengthen wheel-and-whirl type theorems for binary matroids (Kim et al., 2022).

3. Rank connectivity, rank-width, and obstruction theory

The natural connectivity function for pivot-minors is cut-rank. For $G\wedge uv$0,

$G\wedge uv$1

A graph is $G\wedge uv$2-rank-connected if whenever $G\wedge uv$3, one has $G\wedge uv$4 (Oum, 2020). In particular, $G\wedge uv$5-rank-connected coincides with connectivity, and $G\wedge uv$6-rank-connected is exactly primeness with respect to split decomposition (Oum, 2020).

Within this framework, Oum proved a chain theorem: every prime $G\wedge uv$7-rank-connected graph $G\wedge uv$8 with at least $G\wedge uv$9 vertices has a prime $V_1$0-rank-connected pivot-minor $V_1$1 with $V_1$2 (Oum, 2020). As a corollary, every excluded pivot-minor for the class of graphs of rank-width at most $V_1$3 has at most

$V_1$4

and the excluded pivot-minors for rank-width at most $V_1$5 have at most $V_1$6 vertices (Oum, 2020).

The relation to width parameters is especially tight. Every graph of rank-width at most $V_1$7 is a pivot-minor of some graph of tree-width at most $V_1$8, and if $V_1$9 then $V_2$0 is a pivot-minor of a graph of path-width at most $V_2$1 (Kwon et al., 2012). At width $V_2$2, the theory becomes exact: graphs of rank-width at most $V_2$3 are exactly vertex-minors of trees, graphs of linear rank-width at most $V_2$4 are exactly vertex-minors of paths, bipartite graphs of rank-width at most $V_2$5 are exactly pivot-minors of trees, and bipartite graphs of linear rank-width at most $V_2$6 are exactly pivot-minors of paths (Kwon et al., 2012).

Obstruction-size theory gives a complementary finiteness perspective. For a fixed finite field $V_2$7, every forbidden pivot-minor for the class of $V_2$8symmetric matrices of linear rank-width at most $V_2$9 has order at most $V_2$0, more precisely at most $V_2$1 when $V_2$2; as a corollary, every forbidden vertex-minor obstruction for graphs of linear rank-width at most $V_2$3 has order at most $V_2$4 (Kanté et al., 2014).

For tree exclusions, the pivot-minor analogue of the Robertson–Seymour path-width theorem is only partially valid. If $V_2$5 is a tree that is not a caterpillar, then the class of distance-hereditary graphs excluding $V_2$6 as a pivot-minor has unbounded linear rank-width. It is conjectured that boundedness holds when $V_2$7 is a caterpillar, and this is verified for caterpillars of order at most $V_2$8; in particular, every $V_2$9-pivot-minor-free graph has linear rank-width at most $uv\in E(G)$00, and every $uv\in E(G)$01-pivot-minor-free graph has linear rank-width at most $uv\in E(G)$02 (Dabrowski et al., 2020).

4. Degree-boundedness and sparse structure

A major recent theorem concerns sparse classes defined simultaneously by subgraph exclusion and pivot-minor exclusion. For every bipartite graph $uv\in E(G)$03, there is a function $uv\in E(G)$04 such that for every $uv\in E(G)$05, any graph $uv\in E(G)$06 with no $uv\in E(G)$07 subgraph and no pivot-minor isomorphic to $uv\in E(G)$08 has average degree at most $uv\in E(G)$09 (Campbell et al., 31 Jul 2025). The proof passes through an induced bipartite $uv\in E(G)$10-free subgraph of large average degree, then an induced $uv\in E(G)$11-rank-connected subgraph, and finally the bipartite matroid-minor structure hypothesis described above (Campbell et al., 31 Jul 2025).

The block decomposition reduces the problem to graphic fundamental graphs and their bipartite complements. In that setting, two degree lemmas drive the estimate: every $uv\in E(G)$12-free graphic fundamental graph has a vertex of degree at most $uv\in E(G)$13, and every $uv\in E(G)$14-free bipartite complement of a graphic fundamental graph has a vertex of degree at most $uv\in E(G)$15 (Campbell et al., 31 Jul 2025). Summing the resulting blockwise edge bounds yields the global average-degree bound.

A related theorem identifies a sharp local sparsity phenomenon for bipartite circle graphs. By a theorem of de Fraysseix, a bipartite graph is a circle graph if and only if it is the fundamental graph of some planar graph (Campbell et al., 31 Jul 2025). Using the same tree-cycle counting arguments, every $uv\in E(G)$16-free bipartite circle graph with $uv\in E(G)$17 has a vertex of degree at most $uv\in E(G)$18, and the paper gives tightness examples: $uv\in E(G)$19 when $uv\in E(G)$20, and the $uv\in E(G)$21-blow-up of $uv\in E(G)$22 when $uv\in E(G)$23 (Campbell et al., 31 Jul 2025).

These results place pivot-minor exclusion inside the sparse-structure program. The data explicitly note that degree-boundedness results help with colouring and sparse-structure questions, including the Erdős–Hajnal property, in pivot-minor-closed classes (Campbell et al., 31 Jul 2025).

5. Extremal, unavoidable, and prime-structure consequences

Pivot-minor exclusion has strong extremal consequences. Davies proved that every proper pivot-minor-closed class of graphs has the strong Erdős–Hajnal property: for every graph $uv\in E(G)$24, there exists $uv\in E(G)$25 such that every $uv\in E(G)$26-vertex graph with no pivot-minor isomorphic to $uv\in E(G)$27 contains disjoint sets $uv\in E(G)$28 with $uv\in E(G)$29 and $uv\in E(G)$30 complete or anticomplete to $uv\in E(G)$31 (Davies, 2023). An earlier special case established the same conclusion when the excluded pivot-minor is a fixed cycle $uv\in E(G)$32 (Kim et al., 2020).

Colouring results are known at least for cycles. For all positive integers $uv\in E(G)$33 and $uv\in E(G)$34, every graph with sufficiently large chromatic number contains either a clique of size $uv\in E(G)$35 or a pivot-minor isomorphic to a cycle of length $uv\in E(G)$36; equivalently, the class of $uv\in E(G)$37-pivot-minor-free graphs is $uv\in E(G)$38-bounded (Choi et al., 2015). The proofs use induced paths, parity-controlled pivoting along corridors, and “incomplete fan” configurations (Choi et al., 2015).

At the opposite structural extreme, large rank-depth forces unavoidable pivot-minors. For every positive integer $uv\in E(G)$39, every graph of sufficiently large rank-depth contains a pivot-minor isomorphic to a path on $uv\in E(G)$40 vertices or to $uv\in E(G)$41, where $uv\in E(G)$42 denotes two disjoint cliques of size $uv\in E(G)$43 joined by a half graph (Ahn et al., 17 Jul 2025). This answers an open problem raised by Kwon, McCarty, Oum, and Wollan in 2021 (Ahn et al., 17 Jul 2025).

Prime-structure theorems identify exceptional graphs in which pivot-minor reduction cannot preserve primeness in many ways. Every prime graph with at least four vertices has at least two non-pivotal vertices unless it is pivot-equivalent to a cycle, and it has at least three non-pivotal vertices if and only if it is not pivot-equivalent to a graph consisting of at least two internally-disjoint paths between two fixed distinct vertices having no common neighbors (Kim et al., 2022). In the bipartite case, the exceptional family reduces to even cycles and the bipartite members of that path-family (Kim et al., 2022).

6. Algorithms, complexity, and applications beyond structural graph theory

From the computational perspective, the general decision problem is hard. The Pivot-Minor problem—given graphs $uv\in E(G)$44 and $uv\in E(G)$45, decide whether $uv\in E(G)$46 contains a pivot-minor isomorphic to $uv\in E(G)$47—is NP-complete (Dabrowski et al., 2023). The reduction uses the correspondence between pivot-minors of fundamental graphs and minors of binary matroids: for a connected $uv\in E(G)$48-regular graph $uv\in E(G)$49 on $uv\in E(G)$50 vertices, $uv\in E(G)$51 has a Hamiltonian cycle if and only if a certain fundamental graph $uv\in E(G)$52 has a pivot-minor $uv\in E(G)$53 (Dabrowski et al., 2023).

For fixed $uv\in E(G)$54, however, several polynomial cases are known. There is a certifying polynomial-time algorithm for $uv\in E(G)$55-Pivot-Minor when $uv\in E(G)$56 is an induced subgraph of $uv\in E(G)$57, when $uv\in E(G)$58, or when $uv\in E(G)$59 except for $uv\in E(G)$60 (Dabrowski et al., 2023). The method either bounds the size of all induced-subgraph-minimal graphs containing $uv\in E(G)$61 as a pivot-minor, or determines those minimal obstructions explicitly (Dabrowski et al., 2023).

The topic also has a direct quantum-information interpretation. In graph-state language, measuring two adjacent qubits in the $uv\in E(G)$62-basis is equivalent, up to known Pauli-$uv\in E(G)$63 by-products, to pivoting on the corresponding edge and then deleting the measured vertices (Mhalla et al., 2012). This leads to the structural results that every graph on $uv\in E(G)$64 vertices is a pivot-minor of a planar graph of size $uv\in E(G)$65, and even a pivot-minor of a triangular grid of size $uv\in E(G)$66 (Mhalla et al., 2012). The same paper proves that measurements in the $uv\in E(G)$67-plane on graph states represented by triangular grids form a universal measurement-based model of quantum computation (Mhalla et al., 2012).

More broadly, pivot-minor exclusion and bounded rank-width are linked in the data to efficient model-checking of MSO logic and approximation schemes for many NP-hard problems (Campbell et al., 31 Jul 2025). This suggests that the pivot-minor relation is not only a structural analogue of minor containment over $uv\in E(G)$68, but also an algorithmically meaningful interface between dense graph structure, binary matroids, and width-based decomposition theory.