First-hand information for everyone
The authors show that a very sparse random graph typically contains every n‑vertex tree whose degrees are bounded by any fixed constant. Here the random graph model G(n,p) means we put n vertices and include each possible edge independently with probability p. The statement “with high probability” means the chance of the property goes to 1 as n grows. The paper proves there is an absolute constant C>1 so that, for every fixed maximum degree Δ, G(n, C·ln n / n) contains every n‑vertex tree with maximum degree at most Δ with high probability.
The work also treats cycle factors, which are ways to partition the vertices into disjoint cycles. The authors determine, up to a universal constant factor, how long the shortest cycle (the girth) must be so that G(n,p) contains all cycle factors with that girth. In particular they show that G(n, C·ln n / n) contains all cycle factors whose cycles all have length at least about 100·ln n / ln ln n, and they argue this bound is optimal up to constant factors in this sparse regime of p.
At a high level the proofs build a general framework for embedding many kinds of large structures into very sparse random graphs. A key tool is a “linking system.” Roughly, a linking system gives, for two lists of vertices a and b, a family of short vertex‑disjoint paths that can realize any pairing between a and b. The paper establishes how deep such linking systems need to be in G(n,p) when p is only slightly above the connectivity threshold p ≈ ln n / n. Getting the depth optimal allows the authors to glue pieces of trees and cycle factors together inside the random graph at very low density.
Why this matters: the result answers a question posed by Montgomery and removes a dependence on the tree degree that appeared in earlier work. It closes a long‑standing gap about how sparse a random graph can be while still being “universal” for all bounded‑degree trees. The cycle factor result similarly extends and sharpens earlier theorems about when random graphs can realize all large collections of cycles.
Important caveats: the constant C in the main tree result exists but is not given explicitly in the excerpt. The result requires Δ to be a fixed constant that does not grow with n. All statements are probabilistic and hold only “with high probability,” i.e., asymptotically as n→∞. The girth bounds for cycle factors are determined only up to absolute multiplicative constants, and the paper focuses on the sparse range p between C·ln n / n and p = n−1+o(1). The excerpt here may be partial; the full paper contains the technical constructions and proofs behind these claims.