- The paper introduces a deterministic O(n^6) algorithm that decides the existence of a perfect matching with exactly t red edges.
- It employs innovative techniques such as Vandermonde-weighted matrices and tight-cut decompositions to encode and analyze matchings.
- The work resolves a long-standing open problem by unifying algebraic and combinatorial methods, ultimately placing bipartite exact matching in P.
Deterministic Polynomial-Time Algorithm for Bipartite Exact Matching
Introduction and Problem Setting
The Exact Matching (EM) problem in bipartite graphs considers a bipartite graph G=(A∪B,E), an edge coloring ρ:E→{0,1} (assigning “red” or “blue” to each edge), and an integer target t. The task is to decide whether G admits a perfect matching containing exactly t red edges. Despite the availability of a randomized polynomial-time algorithm via the Schwartz–Zippel lemma since the late 1980s, the existence of a deterministic polynomial-time algorithm for the general case remained open, and the problem played a central role in delineating the boundary between the complexity classes P and RP.
The presented work demonstrates, constructively and unconditionally, that bipartite Exact Matching is in P. Specifically, it provides a deterministic O(n6) algorithm for general bipartite graphs using novel algebraic and combinatorial machinery, resolving a question open for over four decades (2604.01571).
Algebraic Framework: Vandermonde-Weighted Matrices
The core algorithmic reduction encodes matchings and the exact-red-count constraint via determinants. For G as above, define the matrix:
ρ:E→{0,1}0
with rows and columns indexed by ρ:E→{0,1}1 and ρ:E→{0,1}2, respectively.
Let ρ:E→{0,1}3. The coefficient ρ:E→{0,1}4 generates, by indexing matchings and tracking red edges, the sum:
ρ:E→{0,1}5
Thus, the existence of a perfect matching with ρ:E→{0,1}6 red edges is equivalent to ρ:E→{0,1}7.
This algebraic encoding enables precise combinatorial decision reductions to polynomial divisibility and nonvanishing conditions, allowing the algorithm to replace randomized identity testing with explicit interpolation and blockwise evaluation.
Structural Decomposition: Braces and Tight Cuts
Central to the method is the exploitation of structural decompositions in bipartite matching-covered graphs. The algorithm employs tight-cut decompositions, reducing ρ:E→{0,1}8 into block components known as braces. Each brace satisfies a strengthened Hall's condition: ρ:E→{0,1}9 for proper nonempty t0.
The proof inducts on the structure of braces using McCuaig’s brace decomposition and handles minimal braces, nonminimal braces (via superfluous-edge deletion), and exceptional families (biwheels, prisms, M\"obius ladders) with specialized arguments.
Crucially, for a brace, the exact-matching existence problem becomes deciding for each block (i.e., each brace) whether the associated fiber polynomial t1 is nonzero for some t2, with the top-level red-count check recursed via a knapsack-style DP corresponding to the block partition of t3.
The Main Theorem and the Affine-Slice Nonvanishing Conjecture (ASNC)
The algebraic reduction hinges on a nonvanishing conjecture established and proved in the paper:
Affine-Slice Nonvanishing (ASNC):
For every brace, any nonempty exact-t4 matching fiber yields a nonzero fiber polynomial t5.
Formally, given the fiber t6, we have:
- If t7, then t8 trivially.
- The nontrivial content is that for nonempty t9, G0.
The deterministic polynomial-time algorithm follows immediately: for each brace block,
- Evaluate G1 at G2 points,
- Interpolate to detect nonvanishing,
- Aggregate results via DP to answer the full instance.
The total complexity is G3 arithmetic operations for graphs with G4 vertices.
Inductive and Algebraic Proof Architecture
Structural Induction
- Base Cases: Small graphs (G5), width-2 path and cycle matchings, and McCuaig’s exceptional families are handled by explicit enumeration, jet operator arguments, and reduction to lower-dimensional cases.
- Superfluous Edge Deletion: For any nonminimal brace, a “superfluous” edge can be deleted without destroying the brace property. The authors provide a precise determinant identity capturing how the matching fiber polynomials change, reducing the problem to divisibility of G6 by powers of the relevant linear form.
- Narrow Extension Cases (including KA and G7): The elimination of repeated divisibility at distinct Vandermonde bases is performed by analyzing generalized Vandermonde matrices and column-replacement determinants.
- Exceptional Families: Specialized arguments and operator-based “jet” extraction reduce these instances to already-handled lower-order cases.
- Replacement Determinant Algebra: Determinant-based criteria for boundary minors of the matching polynomial track how local modifications (patches, edge insertions) influence polynomials encoding matching fibers.
- Generalized Vandermonde Elimination: The authors develop precise elimination theorems to control when a collection of functional constraints can simultaneously vanish, allowing for sharp divisibility and nonvanishing results.
- Hall-type Theorems and Projective Collapse: The proof uses deep connections between the combinatorics of perfect matchings and the ranks/minors of certain structured matrices to preclude cancellation or pathological vanishing across inductive steps. The “Two-extra Hall” theorem and circuit arguments are introduced to handle all remaining algebraic dependencies.
A substantial Lean 4 formalization supports the arguments, with all key algebraic constructions, combinatorial reductions, and kernel/circuit lemmas machine-verified. Eight graph-to-algebra bridges remain as explicit hypotheses, with the remaining proof fully checked.
Numerical and Complexity Consequences
- The deterministic G8 algorithm matches the complexity of classic bipartite matching algorithms up to a low-order polynomial overhead.
- Numerically, the algorithm is practical for G9 on modern hardware, with no reliance on randomized procedures or large-field arithmetic.
- The result resolves the complexity status of bipartite Exact Matching, strictly placing it in t0, and eliminates its role as a candidate for separating t1 and t2.
- Contradictory Claims Addressed: The historical belief that deterministic approaches would not be feasible for four decades is rebutted.
Implications and Future Directions
Theoretical Impact: The result unifies several disparate threads in algebraic, combinatorial, and algorithmic matching theory. The structural induction, algebraic elimination, and fine-grained analysis of dependency loci provide a template for attacking related matching enumeration and constrained combinatorial optimization problems.
Practical Impact: The approach removes randomness from a core matching subroutine with applications in computational biology, combinatorial optimization, network design, and parallel algorithms.
Further Developments: The tight-cut/blockwise methodology and the analysis of fiber polynomials suggest approaches for extending determinism further into constrained perfect matching-type problems in nonbipartite graphs, weighted settings, and parameterized regimes. The authors note that ASNC for non-brace graphs (for nontrivial convolution sums) remains open, providing a concrete path for continued research.
Conclusion
This work conclusively settles the complexity status of the bipartite Exact Matching problem by providing a deterministic polynomial-time algorithm with a careful integration of brace-structural decomposition and Vandermonde-algebraic analysis. The robust proof pipeline and formalization greatly strengthen the result’s foundation and set a high standard for rigor in deterministic combinatorial optimization. The methods developed are likely to influence a broad range of matching-theoretic and algebraic algorithmic questions.