Co-Matching-Free Fragment of Monadic Stability

• The topic defines co-matching-free fragment as graph classes with bounded co-matching indices that enable existential-positive logical encoding over sparse structures.
• It leverages combinatorial tools like the subflip operation and game-theoretic subflipper rank to characterize structural and logical tameness in graphs.
• The theory unifies existential-positive FO transductions, structural sparsity, and the collapse of positive MSO to FO, advancing algorithmic sparsification techniques.

The co-matching-free fragment of monadic stability is a rigorously delineated subclass of graph classes for which model-theoretic and combinatorial tameness properties align with the logical sparsification paradigm. Specifically, it connects classes precluding large co-matchings with the possibility of existential-positive logical encoding by sparse structures. The fragment is characterized by a suite of combinatorial and logical properties, operations, and game-theoretic characterizations, and its theory achieves notable unification of existential-positive FO transductions, structural sparsity, and expressive collapse of positive MSO to FO.

1. Foundational Definitions

Four interrelated notions underpin the subject: monadic stability, co-matching-free graph classes, nowhere dense classes, and existential-positive FO transductions. Graphs are treated as finite relational structures with adjacency predicate $E$, with the reflexive setting marked by $\epsilon$ superscripts.

• Monadic Stability: By the Baldwin–Shelah definition, a graph class $\mathcal{C}$ is monadically stable if no first-order (FO) transduction yields all half-graphs from $\mathcal{C}$. Equivalently, for binary FO formulas $\varphi(\bar x, \bar y)$, the absence of the monadic order property ($G \models \varphi(\bar{a}_i, \bar{a}_j) \iff i < j$ for all $m$) characterizes stability.
• Co-Matching-Free Classes: A co-matching of order $t$ is a bipartite graph on $\{a_1, \ldots, a_t\}$, $\{b_1, \ldots, b_t\}$ with $a_i b_j \in E \iff i \neq j$. A class is co-matching-free if all its graphs have bounded co-matching index: no semi-induced subgraph is a co-matching of order $t$.
• Nowhere Dense Classes: Following Nešetřil–Ossona de Mendez, $\mathcal{C}$ is nowhere dense if for any $r$, there is a bound on the size of complete graphs as depth-$r$ minors in $\mathcal{C}$. This is equivalent to being biclique-free and monadically dependent (Pouzet, Adler–Adler).
• Existential-Positive FO Transductions: An existential-positive FO formula (notation: $\exists^+$–formula) is built by conjunction, disjunction, and existential quantification without negations or universals. An $\exists^+$–transduction colors vertices, applies an $\exists^+$–formula to define new adjacencies, then extracts induced subgraphs and forgets colors.

2. Existential-Positive Sparsification Conjecture

The central conjecture extends the classical sparsification paradigm for monadic stability to the co-matching-free context, strengthening both the structural and logical requirements.

• Conjecture Statement: For reflexive graphs, the following are equivalent:
  1. The class $\mathcal{C}$ is co-matching-free and monadically stable.
  2. $\mathcal{C}$ is an $\exists^+$–transduction of a nowhere dense class (of reflexive graphs).

• Implication: Every $\exists^+$–transduction of a nowhere dense class is co-matching-free and monadically dependent. For any non-trivial transduction-closed property $P$, being semi-ladder-free+$P$ is equivalent to being an $\exists^+$–transduction of a biclique-free+$P$ class.

This posits a precise correspondence between the absence of large co-matchings and the ability to encode dense structures by existential-positive logic over sparse graphs.

3. The Subflip Operation and Its Properties

The subflip operation provides a combinatorial refinement well-suited to the co-matching-free setting.

• Definition: Given $G$ with partition $P = \{P_1, \dots, P_k\}$, a $k$-flip complements edges between specified part pairs. The subflip restricts this: only biclique-inducing part pairs $(P_i, P_j)$ are complemented, yielding $G \ominus P$ as a subgraph of $G$.
• Key Properties:
  • Hereditary with respect to induced subgraphs (partition refinements required).
  • Aggregation: common refinement of $k$-subflip partitions aggregates to a $k^2$-subflip, yielding a subflip contained in both.
  • Approximation: In absence of semi-induced co-matchings of size $t$, every $k$-flip is simulated by a $k t^k$-subflip with distortion $O(t)$.

This suggests the subflip operation retains the separation power of full flips for co-matching-free classes, yet is strictly “one-way” in edge deletion, preserving subgraph relations.

4. Combinatorial and Game-Theoretic Characterizations

Replacing classical “flip” notions with subflips permits direct characterizations paralleling the traditional theory for monadic stability.

• Subflip-Flatness: A class is $r$-subflip-flat if, for any sufficiently large vertex set, there exists a $k$-subflip and subset $A$ such that all pairs in $A$ are at distance $>r$ post-subflip.
  • Theorem: $\mathcal{C}$ is co-matching-free and monadically stable if and only if it is subflip-flat.
• Subflipper Game: Iterative game between Subflipper (choosing $k$-subflips) and Localizer (choosing $r$-neighborhoods) defines subflipper-rank (minimum rounds to singleton).
  • Theorem: For every $r$, existence of $k$ giving Subflipper a winning strategy in all $G\in\mathcal{C}$ equivalently characterizes the fragment.

Failure of subflip-flatness or bounded subflipper-rank correlates strictly with the presence of arbitrarily large co-matchings, precluding monadic stability in the fragment.

5. Canonical Sparsification: Verification for Special Cases

All known sparsifiable special cases—bounded shrub-depth, clique-width, twin-width, merge-width—exhibit canonical existential-positive sparsification.

<table> <thead> <tr><th>Input Class</th><th>Induced Subgraph Witness</th><th>Host Sparse Class</th></tr> </thead> <tbody> <tr><td>Shrub-depth</td><td>Bounded tree-depth</td><td>Subflip-flat class</td></tr> <tr><td>Linear clique-width</td><td>Bounded path-width</td><td>Subflip-flat class</td></tr> <tr><td>Clique-width</td><td>Bounded tree-width</td><td>Subflip-flat class</td></tr> <tr><td>Twin-width</td><td>Bounded sparse twin-width</td><td>Subflip-flat class</td></tr> <tr><td>Merge-width</td><td>Bounded expansion</td><td>Subflip-flat class</td></tr> </tbody> </table>

Each input graph yields an induced subgraph $G^*$ in a sparse class, via an $\exists^+$–transduction “Sparsify.” Recovery transductions reconstruct the original graph from $G^*$. Thus, sparsity extraction and logical encoding are strictly existential-positive—as opposed to requiring the full FO formalism.

6. Collapse of Existential-Positive MSO to FO

On relational structures—including graphs—existential-positive MSO logic collapses to existential-positive FO logic.

• Mechanism: Any $\exists Y\,\psi(\bar x, Y)$ with positive $\psi$ satisfies $G \models \exists Y\,\psi(\bar a, Y)$ iff $G \models \psi(\bar a, Y=V(G))$, exploiting monotonicity of positive formulas. Universal second-order quantifiers collapse to checking $Y=\emptyset$. Iteratively, the MSO quantifiers may be eliminated.

A plausible implication is that, for sparsification or logical encoding purposes, positive MSO transductions yield no additional expressiveness beyond existential-positive FO.

7. Synthesis and Conceptual Implications

The co-matching-free fragment of monadic stability admits an especially robust, clean theory of canonical sparsification. Existential-positive sparsification conjecture anchors the fragment, mapping it precisely to the image of existential-positive FO transductions from sparse classes. The novel subflip operation undergirds combinatorial and algorithmic characterizations, while positive monadic second-order logic provably collapses in expressive power to FO. These findings unify multiple strands of structural graph theory, logic, and sparsification under a model-theoretic and algorithmic lens (Mählmann et al., 22 Jan 2026).