Generalized Turán problem for directed cycles
Andrzej Grzesik, Justyna Jaworska, Bartłomiej Kielak, Piotr Kuc, Tomasz Ślusarczyk
TL;DR
This work advances the generalized Turán problem for directed cycles by determining the order of growth of $ex(n, \overrightarrow{C_k}, \overrightarrow{C_\ell})$ for all $k,\ell \ge 3$, showing a dichotomy between $\Theta(n^k)$ and $\Theta(n^{k-1})$ depending on divisibility. In the dense regime, the leading term often comes from a balanced blow-up of $\overrightarrow{C_d}$, with $d$ the smallest divisor $>2$ of $k$ not dividing $\ell$, yielding $ex(n) = \frac{n}{k}(\frac{n}{d})^{k-1} + o(n^k)$ under suitable conditions; when these fail, a random bipartite orientation furnishes the extremal bound $ex(n) = \frac{2}{k}(\frac{n}{4})^k + o(n^k)$. The paper also analyzes small cycles ($k=3,4,5$) in depth, providing exact or near-exact results in several cases and highlighting diverse extremal constructions, including iterated blow-ups and more intricate configurations. Additionally, sparse-case results, extensions to directed graphs, and a discussion of cycle-type generalizations broaden the scope and open questions for future work.
Abstract
For integers $k, \ell \geq 3$, let $\mathrm{ex}(n, \overrightarrow{C_k}, \overrightarrow{C_\ell})$ denote the maximum number of directed cycles of length $k$ in any oriented graph on $n$ vertices which does not contain a directed cycle of length $\ell$. We establish the order of magnitude of $\mathrm{ex}(n, \overrightarrow{C_k}, \overrightarrow{C_\ell})$ for every $k$ and $\ell$ and determine its value up to a lower error term when $k \nmid \ell$ and $\ell$ is large enough. Additionally, we calculate the value of $\mathrm{ex}(n, \overrightarrow{C_k}, \overrightarrow{C_\ell})$ for some other specific pairs $(k, \ell)$ showing that a diverse class of extremal constructions can appear for small values of $\ell$.
