Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Statement
For every finite tree on vertices, every root , and every finite simple host graph on vertices,
with the [[extremal_graph_theory/adamczewski_2026_erdos548/marked_cut_count|marked and rooted state counts]] defined earlier.
Proof
Induct on , with the assertion uniform in the root and host graph. For , the tree is a single edge. Every marked cut supplies a rooted copy of that edge, so and the assertion follows.
Now let . Connectivity implies that has a neighbor. There are two cases.
If has just one neighbor , delete and root the resulting tree at . Removing a leaf preserves connectivity and acyclicity, so is a tree on vertices. The induction hypothesis and Lemma 2 give
Otherwise has two distinct neighbors, say and . Write for the vertex set of the component of that contains . Deleting an edge of a tree separates it into two trees: an alternative path joining the edge's endpoints would have formed a cycle, and each of the two resulting components stays connected by the original unique paths. Thus while . Define
The second tree is the component containing , with attached by the edge . Both and have at least two vertices and are smaller than . Their union is , their intersection is , and there is no edge between their nonroot vertex sets. Writing gives , because only is counted twice.
Apply induction to both trees, with root , and add the inequalities:
By Lemma 1,
Substitution and subtraction of prove the desired bound. These two cases exhaust the possible degrees of the root, completing the induction.
Source and dependencies
A Counting Proof for Erdős Problem 548, preliminary exposition, equation
(2) on p. 2 and its proof in §4, pp. 3–4, in the
canonical PDF.
The pinned formal source uses tree_root_partition,
rooted_word_tree_bound_aux, and rooted_word_tree_bound; see the
source record.
The dependencies are Lemmas 1 and 2 and the elementary tree separation facts
spelled out in the proof. No asymptotic embedding theorem is imported.