An Algorithm for the Decomposition of Complete Graph into Minimum Number of Edge-disjoint Trees
Antika Sinha, Sanjoy Kumar Saha, Partha Basuchowdhuri
TL;DR
The paper tackles decomposing the edge set of a complete graph $K_n$ into the minimum number of edge-disjoint trees. Using Tutte and Nash-Williams results and the tree covering number $\tau$, arboricity $\alpha$, and spanning tree packing number $\sigma$, it proves the optimal count is $\tau(K_n)=\lceil \frac{n}{2} \rceil$ with a corresponding size pattern. It then introduces two linear-time constructions, DECK-E for even $n$ and DECK-O for odd $n$, that realize the decomposition in $O(m)$ time. Experiments on $K_6$ and $K_7$ illustrate the method and show strong alignment with known bounds for even orders, with the constructed decompositions offering practical benefits for load balancing, parallel computation, and network design.
Abstract
In this work, we study methodical decomposition of an undirected, unweighted complete graph ($K_n$ of order $n$, size $m$) into minimum number of edge-disjoint trees. We find that $x$, a positive integer, is minimum and $x=\lceil\frac{n}{2}\rceil$ as the edge set of $K_n$ is decomposed into edge-disjoint trees of size sequence $M = \{m_1,m_2,...,m_x\}$ where $m_i\le(n-1)$ and $Σ_{i=1}^{x} m_i$ = $\frac{n(n-1)}{2}$. For decomposing the edge set of $K_n$ into minimum number of edge-disjoint trees, our proposed algorithm takes total $O(m)$ time.
