Literal-Clause Incidence Graph

• Literal-Clause Incidence Graph is a graph structure that represents 2-SAT and Max-2-SAT formulas as literal and clause gadgets, enabling nuanced complexity proofs.
• The graph preserves key properties like acyclicity and bounded degree, which are critical in reductions showing PSPACE-completeness and NP-completeness.
• Its design facilitates cycle-elimination and structural simplifications in graph games, linking logical formula structure to positional game complexity.

A literal-clause incidence graph is a special form of incidence graph designed to encode the variable–clause structure of a 2-SAT or Max-2-SAT formula for use as a reduction gadget in complexity proofs for graph games and constraint satisfaction problems. Its construction and properties enable precise control over graph-theoretic features such as acyclicity and degree, which are crucial in establishing computational hardness for classes of combinatorial games and satisfiability problems. The notion is central for demonstrating PSPACE-completeness and NP-completeness analogs under restrictive structural conditions.

1. Formal Definition

Given a quantifier-free 2-CNF formula $\varphi$ on variables $x_1,x_2,\ldots,x_n$ with clauses of the form $(\ell_u \lor \ell_v)$, where $\ell$ denotes a literal (either $x_i$ or $\neg x_i$), the literal-clause incidence graph $G_\varphi$ is constructed as follows (Hilaire et al., 12 Jan 2026):

• For each variable $x_i$, include two vertices: $x_i^+$ (the positive literal) and $x_i^-$ (the negated literal).
• For each clause $(\ell_u \lor \ell_v)$, insert a new degree-2 vertex and connect it by edges to $\ell_u$ and $\ell_v$, forming a path of length 2 between the corresponding literal vertices.

Thus, $G_\varphi$ consists of a bipartite-like structure with a "V" (path of length 2) for each clause, but with explicit separation between positive/negative literals and with every clause projected as a small gadget rather than as a simple edge.

The construction generalizes naturally to quantified (QBF) variants for complexity reductions, maintaining an explicit mapping from logical variables/literals to graph vertices.

2. Structural Properties

The literal-clause incidence graph $G_\varphi$ exhibits several key structural characteristics:

• Maximum Degree Constraints: Each literal vertex's degree is equal to its number of occurrences in the formula. Thus, restricting each literal to appear at most twice ensures that $G_\varphi$ has maximum degree 2.
• Acyclicity: $G_\varphi$ is acyclic if and only if the clause–literal connectivity of $\varphi$ (i.e., considering which literals co-occur in clauses) forms a forest structure.
• Component Structure: For formulas with bounded occurrences and no cycles, $G_\varphi$ is a disjoint union of paths.

These properties are directly exploited in reductions to restrict the combinatorial complexity of the underlying graph while retaining computational hardness (Hilaire et al., 12 Jan 2026).

3. Role in Complexity Reductions

The literal-clause incidence graph functions as a central gadget in hardness proofs for the Maker–Breaker Happy Vertex Game (SHVG) and related scoring games (Hilaire et al., 12 Jan 2026). Specifically:

• To prove SHVG is PSPACE-complete on trees, a reduction is performed from Quantified Max-2-SAT on formulas whose literal-clause incidence graph is acyclic. Each literal and clause becomes a gadget within a tree; quantifiers are mapped to Maker/Breaker alternation.
• To establish NP-hardness for SHVG on caterpillars, the reduction starts from Max-2-SAT where each literal appears at most twice, guaranteeing $G_\varphi$ has degree at most 2 and is acyclic. The resulting SHVG instance is a caterpillar.
• The construction ensures that even under severe structural restrictions (acyclicity, bounded degree), computational hardness remains, due to the encoding power of the literal-clause incidence graph.

The complexity landscape enabled by this gadget is encapsulated in the following results:

Problem Variant	Structure of $G_\varphi$	Hardness
Q-Max-2-SAT	acyclic	PSPACE-cmpl.
Max-2-SAT, $\deg\le2$	acyclic, union of paths	NP-cmpl.

These results directly yield the tree/caterpillar reductions for SHVG and demonstrate that Incidence, the scoring positional game, is also PSPACE-complete even when restricted to forests (Hilaire et al., 12 Jan 2026).

4. Cycle-Elimination Gadgets

To convert an arbitrary 2-CNF formula into one whose literal-clause incidence graph is acyclic without changing (by more than a constant additive amount) the maximal set of satisfiable clauses, the following technique is employed (Hilaire et al., 12 Jan 2026):

• Whenever $G_\varphi$ contains a cycle, insert a small 4-clause gadget that "breaks" the cycle and increases the total clause count by at most 4. Each such gadget reduces the number of potential cycle-edges by one. After a polynomial number of such steps, $G_\varphi$ becomes acyclic.
• For the bounded occurrence variant, a similar strategy combined with reductions from Max-2-SAT-3 suffices.

This cycle-elimination process is pivotal for showing that strong hardness persists even under highly restricted graph-theoretic conditions.

5. Significance in Graph Games and SAT

The introduction of the literal-clause incidence graph provides a robust tool for linking the structure of logical formulas to the topology of their associated graph games and constraint problems. Notably:

• It enables fine-grained reductions between logical satisfiability and positional games by encoding quantifiers, satisfaction of clauses, and player alternations as coloring or selection processes on graphs.
• It allows for premise-tight complexity results, such as showing SHVG remains PSPACE-complete even on trees and Incidence is PSPACE-complete when restricted to forests (Hilaire et al., 12 Jan 2026).
• Its use demonstrates that even pathwise (deg $\leq2$) or tree-based (acyclic) instances retain full computational power for key quantitative game-theoretic and SAT variants.

This suggests that in the landscape of algorithmic game theory and SAT, the literal-clause incidence graph acts as a minimal yet complete encoding bridge—preserving both logical structure and graph-theoretic simplicity, while enabling precise complexity-theoretic separations and parameterized algorithmics.