Landau's function for one million billions (original) (raw)
Abstract: Let mathfrakSn{\mathfrak S}_nmathfrakSn denote the symmetric group with nnn letters, and g(n)g(n)g(n) the maximal order of an element of mathfrakSn{\mathfrak S}_nmathfrakSn. If the standard factorization of MMM into primes is M=q1al1q2al2...qkalkM=q_1^{\al_1}q_2^{\al_2}... q_k^{\al_k}M=q1al1q2al2...qkalk, we define ell(M)\ell(M)ell(M) to be q1al1+q2al2+...+qkalkq_1^{\al_1}+q_2^{\al_2}+... +q_k^{\al_k}q1al1+q2al2+...+qkalk; one century ago, E. Landau proved that g(n)=maxell(M)lenMg(n)=\max_{\ell(M)\le n} Mg(n)=maxell(M)lenM and that, when nnn goes to infinity, logg(n)simsqrtnlog(n)\log g(n) \sim \sqrt{n\log(n)}logg(n)simsqrtnlog(n). There exists a basic algorithm to compute g(n)g(n)g(n) for 1lenleN1 \le n \le N1lenleN; its running time is co(N3/2/sqrtlogN)\co(N^{3/2}/\sqrt{\log N})co(N3/2/sqrtlogN) and the needed memory is co(N)\co(N)co(N); it allows computing g(n)g(n)g(n) up to, say, one million. We describe an algorithm to calculate g(n)g(n)g(n) for nnn up to 101510^{15}1015. The main idea is to use the so-called {\it ell\ellell-superchampion numbers}. Similar numbers, the {\it superior highly composite numbers}, were introduced by S. Ramanujan to study large values of the divisor function tau(n)=sumddvn1\tau(n)=\sum_{d\dv n} 1tau(n)=sumddvn1.
Submission history
From: Fatine Latif [view email] [via CCSD proxy]
[v1] Fri, 14 Mar 2008 13:37:11 UTC (74 KB)