On the Number of Hamilton Cycles in Pseudo-Random Graphs

Michael Krivelevich

DOI: https://doi.org/10.37236/1177

2011-11-27

Abstract:We prove that if $G$ is an $(n,d,\lambda)$-graph (a $d$-regular graph on $n$ vertices, all of whose non-trivial eigenvalues are at most $\lambda)$ and the following conditions are satisfied: $\frac{d}{\lambda}\ge (\log n)^{1+\epsilon}$ for some constant $\epsilon>0$; $\log d\cdot \log\frac{d}{\lambda}\gg \log n$, then the number of Hamilton cycles in $G$ is $n!\left(\frac{d}{n}\right)^n(1+o(1))^n$.