A threshold phenomenon for embeddings of Euclidean snowflakes and impossibility of dimension reduction

Abstract: Fix $0&lt;θ\leqslant 1$. We prove that if $1\leqslant p \leqslant 2/θ$, then the $θ$-snowflake of $\ell_2<sup>k$, namely, $\mathbb{R}<sup>k$ equipped with the metric $$((x,y)\in \mathbb{R}^k\times \mathbb{R}^k)\mapsto |x-y|<em>2<sup>θ$$, embeds with distortion $O(1)$ into $\ell_p<sup>m$ for some integer $m\lesssim</em>{p,θ}k$, which is optimal as $k\to \infty$, as seen by comparing dimensions. However, for $p$ larger than the sharp threshold $2/θ$ the following change in behavior occurs: If a $(1/\sqrt{k})$-dense subset of the Euclidean sphere $S<sup>{k-1}$ embeds into $\ell_p<sup>m$ with distortion $O(1)$, then necessarily $m\gtrsim_{p,θ}( k/\log k)<sup>{pθ/2}$, which grows super-linearly in $k$ as $pθ/2&gt;1$, and this dimension bound is optimal as $k\to \infty$ up to lower order factors. We deduce from this statement that if $2&lt;p&lt;\infty$, then there exist arbitrarily large $n$-point subsets of $\ell_p$ with the property that if they embed with distortion $O(1)$ into $\ell_p<sup>m$, then necessarily $m\gtrsim_p ((\log n)/(\log\log n)<sup>2)<sup>{p/2}$, thus demonstrating that the statement of the Johnson--Lindenstrauss dimension reduction lemma fails to hold for $\ell_p$