---
title: Nash-Williams' Triangle Decomposition Conjecture
url: https://www.emergentmind.com/topics/nash-williams-conjecture
type: topic
---

# Nash-Williams' Triangle Decomposition Conjecture

Nash-Williams' Conjecture is a central result in extremal design theory concerning triangle decompositions of graphs. It asserts a precise minimum degree threshold for when a triangle-divisible graph can be edge-partitioned into triangles, and its full resolution [2606.11178] closes a decades-old question by identifying the exact barrier.

## 1. Statement and Motivation of Nash-Williams' Conjecture

A graph $G$ on $n$ vertices is **triangle-divisible** if:
- Every vertex $v\in V(G)$ has even degree: $d_G(v)\equiv 0\pmod2$,
- The total number of edges $e(G)$ is divisible by 3: $3 \mid e(G)$,

These constraints are plainly necessary for $G$ to have a $K_3$-decomposition, that is, an edge-partition into triangles.

**Nash-Williams' Conjecture (1970).**  
Let $G$ be a triangle-divisible graph on $n$ vertices. If the minimum degree satisfies
$$
\delta(G) \ge \tfrac{3n}{4}
$$
and $n$ is sufficiently large, then $G$ admits an exact $K_3$-decomposition (an edge-partition into triangles).

This bound is tight: Graham's extremal construction (the join of two $n/2$-vertex graphs, each $n/4$-regular, joined by all cross-edges) demonstrates that $3n/4$ cannot be reduced in general.

The conjecture sits at the intersection of extremal combinatorics, design theory, and the theory of Steiner triple systems; it is the graph-theoretic analogue of the existence of $(n,3,2)$-Steiner systems and generalizes the divisible complete graph case solved by Kirkman.

## 2. Fractional Relaxation: The Fractional Nash-Williams' Conjecture

A **fractional $K_3$-decomposition** is a function $w$ from triangles of $G$ to $[0,1]$ so that
$$
\sum_{T\ni e} w(T) = 1 \quad \forall\, e\in E(G).
$$
Define the **fractional threshold**
$$
\delta^*_{K_3} = \inf\left\{c: \text{every $G$ on $n$ vertices with }\delta(G)\geq cn\text{ has a fractional }K_3\text{–decomposition}\right\}.
$$
Extremal examples show $\delta^*_{K_3}\geq 3/4$ is necessary.

The **Fractional Nash-Williams' Conjecture** posits that this lower bound is tight:  
If $G$ is an $n$-vertex graph with $\delta(G)\geq 3n/4$, then $G$ admits a fractional $K_3$-decomposition.

[2606.11178] provides the first proof of this conjecture, establishing $\delta^*_{K_3}=3/4$.

### Proof Outline for the Fractional Result

- By Farkas’ Lemma, failure of a fractional triangle decomposition leads to a certificate in the form of an assignment of real “charges” $\phi(e)$ to the edges such that every triangle has nonnegative total charge but $\sum_e\phi(e)<0$.
- The key technique is **discharging in the dual**: positive and negative edges discharge according to whether triangles are “acute” or “obtuse”.
- The argument proceeds with a series of technical reductions: removing contributions outside common neighborhoods, introducing edge “values”, proving the induced filtered subgraph is $K_5$-free, and bounding its total weight via a graph Lagrangian.
- These rules force every negative edge’s final charge to be nonnegative, contradicting the assumed certificate.

This result implies, through the absorption and reduction machinery of Barber–Kühn–Lo–Osthus, that the asymptotic Nash-Williams theorem holds for any $\delta(G)\geq (3/4+\varepsilon)n$.

## 3. Fractional Stability and Extremality

Proving mere existence of fractional decompositions is insufficient for a robust conversion to an exact decomposition. **Fractional stability** concerns the structural characterization of near-extremal graphs which lack a fractional decomposition.

**Fractional Stability Theorem:**  
For any small $\sigma>0$, there exists $\varepsilon>0$ such that if $G$ has
$$
\delta(G)\geq (\tfrac{3}{4}-\varepsilon)n
$$
but fails to have a fractional $K_3$-decomposition, then $G$ is $\sigma$-extremal in the following sense:
- At least $(1-\sigma)n$ vertices have degree $\leq (\frac{3}{4}+\sigma)n$,
- $\mathrm{MaxCut}(G)\geq (1-\sigma)n^2/4$.

The proof extends the discharging method, tracks charge propagation, and iteratively shows that any negative edge must live in a section of the graph nearly isomorphic to the extremal join construction.

## 4. Conversion: From Fractional to Exact Decomposition via Absorption

To move from fractional to exact decompositions, the **absorption method** is used:

1. **Absorber construction:** In non-extremal graphs with $\delta(G)\geq (3/4-\varepsilon)n$, one constructs a bounded absorber subgraph $A$ such that for every small balanced leftover $L$ compatible with divisibility, $A\cup L$ has a perfect triangle decomposition.
2. **Random-greedy packing:** Most edges are packed greedily into triangles following the fractional decomposition, leaving a sparse leftover.
3. **Final absorption:** Pre-selected absorbers are then used to cover all remaining edges exactly.

Postle–Delcourt’s absorber-based “black-box” theorem allows this to be implemented even for graphs with partitioned vertex sets and is robust to block-structured extremal configurations.

Applied to each of the finitely many structural cases, this method establishes the final step:  
For all sufficiently large $n$, every triangle-divisible $G$ with $\delta(G)\geq 3n/4$ has a $K_3$-decomposition ([2606.11178]).

## 5. Interplay, Tightness, and Broader Significance

The resolution of Nash-Williams’ conjecture is structured in three central layers:
- The fractional threshold is exactly $3/4$.
- Any graph near this threshold which fails to admit a decomposition must closely mimic the construction of two $n/4$-regular halves, joined.
- The absorption method, paired with the robust fractional structure, effects the transition from fractional to exact decomposition.

This result settles the minimum-degree barrier for decomposing large triangle-divisible graphs into triangles, providing the tight bound for all $n$.

### Table: Central Results and Techniques

| Result/Theorem                     | Essential Content                                             | Reference        |
|------------------------------------|--------------------------------------------------------------|------------------|
| Nash-Williams' Conjecture (1970)   | $\delta(G)\geq 3n/4$ forces triangle decomposition for large $n$ | [2606.11178]     |
| Fractional Nash-Williams Conjecture| $\delta^*_{K_3}=3/4$ (sharp bound for fractional)            | [2606.11178]     |
| Fractional Stability Theorem       | Structural characterization of near-extremal non-fractional  | [2606.11178]     |
| Absorption method for exact case   | Random-greedy + absorber yields full $K_3$-decomposition     | [2606.11178]     |

## 6. Connections and Impact

Triangle decomposition thresholds are deeply connected to design theory, regularity/absorption paradigms, and extremal combinatorics. Matching the fractional threshold to the integral threshold demonstrates the efficacy of robust absorption techniques in bridging fractional relaxations and integral packings.

The Nash-Williams conjecture now serves as a template for higher-order decompositional problems—see, for example, current work on $K_4$-decompositions, hypergraph generalizations, and analogues in design theory ([2510.07783], [2512.04071], [2605.21961]). The new technical tools for fractional stability and absorber construction have broad applicability across these domains.

The proof’s layered approach—fractional result, extremal characterization, and absorber-fueled conversion—is expected to inform parallel work in hypergraph and multipartite settings, where divisibility and stability phenomena are likewise paramount.

Source: https://www.emergentmind.com/topics/nash-williams-conjecture