Corrádi-Hajnal Theorem

Updated 2 February 2026

Corrádi-Hajnal Theorem is a fundamental result defining how vertex-disjoint cycles, particularly triangles, can be packed in graphs under prescribed local or global conditions.
It employs combinatorial methods including regularization, structural decomposition, and rotation arguments to achieve sharp extremal constructions.
The theorem has inspired extensions to multipartite, random, and colored graphs, influencing both theoretical insights and algorithmic applications in graph theory.

The Corrádi-Hajnal Theorem is a cornerstone result in extremal graph theory concerning the packing of vertex-disjoint cycles, specifically triangles, in graphs with prescribed minimum degree, and has inspired a spectrum of generalizations framed by edge-density, multipartite extensions, color constraints, and resilience under random perturbations. A central theme is the transition from local degree conditions to global density thresholds, reflecting a broader movement in modern combinatorics from structural to probabilistic and algorithmic paradigms.

1. Classical Statement and Extremal Constructions

For $k \ge 1$ , the original Corrádi-Hajnal Theorem asserts: Any graph $G$ of order $n \ge 3k$ with minimum degree $\delta(G) \ge 2k$ contains $k$ vertex disjoint cycles (in the classical form, triangle factors when $k$ is fixed and $n = 3k$ ) (Rautenbach et al., 2014). The minimum degree condition is tight, as demonstrated by constructing $K_{2k-1}$ joined to an independent set, which has minimum degree $2k-1$ but fails to pack $k$ triangles.

In density versions, the minimum degree restriction is lifted in favor of a global edge-count lower bound. The main theorem proven in "A Density Version of the Corrádi-Hajnal Theorem" (Rautenbach et al., 2014) states: For $k \ge 1$ and $n \ge 3k$ , every graph $G$ with

$e(G) > \max\left\{ \binom{2k-1}{2} + (2k-1)(n - (2k-1)),\, \binom{3k-1}{2} + (n - (3k-1)) \right\}$

contains $k$ vertex-disjoint cycles. Each term in the maximum corresponds to an extremal construction:

The first term is realized by a clique of size $2k-1$ joined to all other vertices;
The second term is realized by a clique of size $3k-1$ and an independent set connected minimally to the clique.

These constructions achieve equality in the bound without admitting $k$ disjoint cycles, thus confirming sharpness.

2. Proof Architecture and Key Combinatorial Tools

A canonical density-version proof proceeds through stages:

Regularization/Cleaning: Remove vertices forming sparse sets so the remainder has strong quasi-regularity.
Structural Decomposition/Stability: Either the graph is near-extremal (edit-distance to bounding construction is small), or minimum degree lifts sufficiently for the original theorem to apply directly.
Handling Extremal Cases: If close to an extremal structure, combinatorial matching and absorption techniques are employed to extract the required cycles.
Defect-Sum/Rotation Arguments: Specialized manipulations (rotations, defect sums) boost local densities to achieve additional cycle packings, avoiding loss of covering.

The tightness of the result is ensured by these extremal examples; any improvement in the coefficients would invalidate the theorem.

3. Relation to Degree and Ore-Type Packing Theorems

The density version subsumes the degree-based Corrádi-Hajnal threshold: For sufficiently large $n$ , any graph below the required edge bound must admit a low-degree vertex, recovering the minimum degree case. Similarly, Ore-type packings generalize minimum degree conditions to sums over pairs of nonadjacent vertices, as in the theorem: $d(x) + d(y) \ge 4k-1$ for all nonadjacent $x,y$ ensures $k$ vertex-disjoint cycles for $n \ge 3k$ .

This landscape has been further sharpened: precise characterizations are given for graphs with degree or Ore-degree just below the critical threshold, involving independence number restrictions and, for very small $k$ , enumeration of exception classes (Kierstead et al., 2016).

4. Extensions to Multipartite, Random, and Colored Settings

Multipartite Analogues

The tripartite version (for balanced tripartite graphs with $N$ vertices in each class and each vertex adjacent to at least $(2/3)N$ vertices of each other class) holds except for a uniquely characterized extremal configuration $T_3(N/3)$ where no perfect packing exists (Magyar et al., 2016).

Resilience in Random Graphs

In sparse random graphs, a local resilience result asserts that for $p \gg (\log n / n)^{1/2}$ , any subgraph of $G(n, p)$ with minimum degree at least $(2/3 + o(1))np$ contains a triangle packing covering all but $O(p^{-2})$ vertices, with this threshold and error term essentially optimal. The proof relies on a sparse regularity lemma and specialized embedding results for balanced three-partite graphs within the random setting (Balogh et al., 2010).

Edge-Colored and Rainbow Extensions

Recent results generalize Corrádi-Hajnal to edge-colored graphs; the minimum color degree $\delta^c(G) \geq (5/6 + \varepsilon) n$ guarantees perfect rainbow triangle tiling, but this bound cannot be improved below $5n/7$ (Lo et al., 2024). Anti-Ramsey variants quantify the minimal number of colors in $K_n$ needed to force $t$ rainbow-disjoint triangles; extremal examples and piecewise bounds are described, with characterizations extending to multipartite and hypergraph cases (Jinghua et al., 5 Oct 2025, Hou et al., 2023).

5. Algorithmic and General Density Versions

Algorithmic advances produce fixed-parameter tractable methods for finding $K_r$ -factors given minimum degree bounds, applying color-coding and stability arguments (Gan et al., 2023). The density paradigm has been pushed further into hypergraph domains, yielding matching Turán-type formulas valid for all $t \leq c_F n$ in nondegenerate $r$ -graphs, with rainbow and general $F$ -packing results described via join constructions (Hou et al., 2023, Allen et al., 2014). The extremal graphs consist of a clique on $t$ vertices joined to an extremal $F$ -free subgraph in the remaining $n-t$ vertices, giving precise edge bounds for the absence of $t+1$ disjoint $F$ -subgraphs.

6. Open Problems and Directions

Open questions concern exact thresholds for more general cycle lengths and hypergraph factors, further reducing computational complexity in algorithmic variants, and extending the defect/rotation techniques to richer combinatorial structures. Some conjectures propose new color-degree thresholds for perfect rainbow tilings in edge-colored graphs and hypergraphs (Lo et al., 2024). There is ongoing interest in classifying the sharp threshold figures for Ore-degree and degree profiles, and in understanding phase transitions in the extremal structure for large $t$ in density Turán problems.

7. Comparative Overview of Regimes

Regime	Key Constraint(s)	Packing Guarantee	Extremal Configuration
Classical (degree)	$\delta(G) \geq 2k$	$G$ contains $k$ disjoint cycles	Clique-plus-independent set
Density (edges)	$e(G) > \max\{\ldots\}$ as above	$G$ contains $k$ disjoint cycles	Clique+joined, large clique
Ore-type	$d(x)+d(y)\geq 4k-1 \,\, (\forall x\neq y, xy\notin E)$	$G$ contains $k$ disjoint cycles	Join constructions
Tripartite	Classwise $\deg(v;V_j)\geq (2/3)N$	Perfect triangle packing or unique exception	$T_3(N/3)$
Random graphs	Minimum degree, $p \gg (\log n / n)^{1/2}$	Packing covers all but $O(p^{-2})$ vertices	Error term optimal
Edge-colored	Minimum color degree $\geq (5/6+\varepsilon)n$	Spanning rainbow triangle tiling	Bound cannot go below $5n/7$

The table emphasizes the logical progression from local degree conditions to global and coloring constraints, mapping precise combinatorial thresholds to corresponding extremal structures and limitations proven in the cited literature.

The Corrádi-Hajnal Theorem and its density/transversal generalizations fundamentally determine the interplay between local connectivity and global packing structure, underpinning the threshold phenomena in covering problems across graph theory and its algorithmic, probabilistic, and colored extensions (Rautenbach et al., 2014).