MathLabs

Worked solution: A rainbow-tree proof of Ringel's conjecture (2020)

Step 3 of 8: Splitting trees into three shapes to attack separately
In plain words

Trees can look wildly different -- a star with nn leaves, a long thin path, or something in between -- so no single technique handles them all. The authors show that every large enough tree TT falls into (at least) one of three convenient shapes: plenty of separated leaves to trim off (Case A), plenty of long thin "bare" stretches to work with (Case B), or, failing both, a tree so dominated by a few very high-degree vertices that trimming their leaves collapses it to something tiny (Case C).

T has at least δ6n non-neighbouring leavesT \text{ has at least } \delta^6 n \text{ non-neighbouring leaves}
Detailed analysis

Montgomery, Pokrovskiy and Sudakov (2021, Section 2, Lemma 3.5 "Case division") show that for a small constant δ>0\delta > 0, every sufficiently large (n+1)(n+1)-vertex tree falls into at least one of three cases: Case A, T has at least δ6n non-neighbouring leavesT \text{ has at least } \delta^6 n \text{ non-neighbouring leaves}, where a leaf is called non-neighbouring if it is not adjacent to another leaf; Case B, T has at least δn/800 vertex-disjoint bare paths of length δ−1T \text{ has at least } \delta n/800 \text{ vertex-disjoint bare paths of length } \delta^{-1}, where a bare path is one whose internal vertices all have degree 22 in TT; and Case C, removing leaves next to vertices of degree≥δ−4 leaves a tree with at most n/100 vertices\text{removing leaves next to vertices of degree} \ge \delta^{-4} \text{ leaves a tree with at most } n/100 \text{ vertices}. The three cases overlap, but the paper only ever needs to treat a tree as belonging to Case A or B (if it does), or else Case C.

Terms in this step
Leaf
A vertex of degree 11 in a tree.
Bare path
A path inside TT all of whose internal vertices have degree exactly 22 in TT (so the path is not interrupted by any branching).
Knowledge used in this step