Topological embeddings into random 2-complexes

TL;DR

This paper determines the threshold probabilities in a multi-parameter random 2-complex model for when all 2-dimensional complexes can be embedded into the random complex, linking topological embedding properties with hyperbolicity and fundamental group triviality.

Contribution

It establishes the precise multi-parameter threshold for topological embeddings of 2-complexes into random complexes, connecting geometric, topological, and probabilistic aspects.

Findings

Threshold for embedding all 2-complexes is p0 p1^3 p2^2 = 1/n.

Embedding is almost sure above the threshold with subdivided complexes.

Torus embeddings are almost surely impossible below the threshold.

Abstract

We consider 2-dimensional random simplicial complexes $Y$ in the multi-parameter model. We establish the multi-parameter threshold for the property that every 2-dimensional simplicial complex $S$ admits a topological embedding into $Y$ asymptotically almost surely. Namely, if in the procedure of the multi-parameter model, each $i$-dimensional simplex is taken independently with probability $p_i=p_i(n)$, from a set of $n$ vertices, then the threshold is $p_0 p_1^3 p_2^2 = \frac{1}{n}$. This threshold happens to coincide with the previously established thresholds for uniform hyperbolicity and triviality of the fundamental group. Our claim in one direction is in fact slightly stronger, namely, we show that if $p_0 p_1^3 p_2^2$ is sufficiently larger than $\frac{1}{n}$ then every $S$ has a fixed subdivision $S'$ which admits a simplicial embedding into $Y$ asymptotically almost surely. The main geometric result we prove to this end is that given $\epsilon>0$, there is a subdivision $S'$ of $S$ such that every subcomplex $T \subseteq S'$ has $\frac{f_0(T)}{f_1(T)}>\frac{1}{3}-\epsilon$ and $\frac{f_0(T)}{f_2(T)}>\frac{1}{2}-\epsilon$, where $f_i(T)$ denotes the number of simplices in $T$ of dimension $i$. In the other direction we show that if $p_0 p_1^3 p_2^2$ is sufficiently smaller than $\frac{1}{n}$, then asymptotically almost surely, the torus does not admit a topological embedding into $Y$. Here we use a result of Z. Gao which bounds the number of different triangulations of a surface.