I’m interested in the problem of simultaneous rational approximation to $k\geq 2$ numbers $\alpha_1,…,\alpha_k$ in the generic case where the $\alpha_j$ are transcendental and algebraically independent. More precisely, what is the largest $m$ such that there exists an infinite sequence of positive integer denominators $q$ and corresponding integers $n_j$ (depending on q) such that

$|\alpha_j-n_j/q|<1/q^{m+1/k}\qquad (1\leq j\leq k)$

Alternatively, what is the largest $m$ such that the above holds for except on a set of $k$-tuples $(\alpha_1,\ldots,\alpha_k)$ of measure 0. The classical Dirichlet theorem shows $m\geq 1$ (with no exceptions). But is the generic value $m\geq 2$? I'm hoping that this is well known among number theorists; any pertinent references would be much appreciated.

  • 3
    $\begingroup$ FWIW, on MathOverflow, 'Title' is meant to describe your question and not your job title :) $\endgroup$ – Glorfindel Jul 7 '18 at 16:39

For any $m>1$ the set of such $k$-tuples $\alpha:=(\alpha_1,\dots,\alpha_k)$ have measure zero. For seeing this restrict onto $\alpha\in [0,1]^k$ and denote by $\Omega_q$ the set of $\alpha$'s in $[0,1]^k$ for which this $q$ works. The measure of $\Omega_q$ does not exceed, say, $2^k q^{k(1-m-1/k)}=2^k q^{-1-k(m-1)}$. Thus the sum of measures of $\Omega_q$'s is finite and by (easy part of) Borel -- Cantelli lemma almost any point $\alpha\in [0,1]^k$ belongs only to finitely many $\Omega_q$'s.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.