Work

Extremes of the two-sided Eulerian statistic

Year: 2026
Type: preprint
Abstract: Let T_n = des(π)+des(π^(-1)) for a uniform random permutation π of size n, and let M_n be the maximum of k_n independent copies of T_n. We prove that M_n, centred and scaled with the mean, the standard deviation and the Gaussian extreme-value constants, converges to the standard Gumbel law if and only if log k_n = o(√n). When log k_n/√n → c ∈ [0,∞) the limit is the Gumbel law shifted by -c²/10, and when √n ≪ log k_n ≪ n the normalised maximum tends to -∞. A corrected centring gives the standard Gumbel limit for log k_n = o(n^(2/3)), and a centring and scaling defined by the exact cumulant generating function give it for log k_n = o(n). The maxima of des(π_j) and of des(π_j^(-1)) are asymptotically independent, and when log k_n/√n → c each limit is the Gumbel law shifted by -c²/5. This answers two questions of Dörr and Heiny.
Language: English
Open Access status: green
OpenAlex ID: W7236665118
Cites:
Cited by:
Related to: 10