On some inequalities which are an generalization of F. Beukers' corresponding results

154 Views Asked by At

In 1981, F. Beukers proved the following theorem in his article On the generalized Ramanujan-Nagell equation I :

Theorem $\bf1$. $~$ Suppose $m \in \mathbb{Z}$, then for any integer $x$, $$ \left | \frac{x}{2^m} - \sqrt{2} \right | > 2^{-1.8m-43.9}. $$

By using Theorem $\bf 1$, he obtained:

Theorem $\bf 2$. $~$ Suppose $D \neq 0$ and $D \in \mathbb{Z}$. If the diophantine equation $$ x^2 - D = 2^n $$ has positive integer solution $(x,n)$, then $n < 435 + \frac{10 \log |D|}{\log 2}$.

Two questions come to my mind:

Firstly, is there an optimization of the upper bound of $n$ in Theorem $\bf 2$ ?

Secondly, if we make a small change in the above equation, namely, $x^3 - D = 2^n$, it may require inequalities like these to solve: $$ \left | \frac{x}{2^m} - \sqrt[3]{2} \right | > C_1(m), \tag{1} $$ and $$ \left | \frac{x}{2^m} - (\sqrt[3]{2})^2 \right | > C_2(m), \tag{2} $$ where $C_1(m)$ and $C_2(m)$ are some constants like in Theorem $\bf 1$.

Hence, my second question is: Are there some references on inequalities $(1)$ and $(2)$ ? If so, what are the best known lower bounds?


Supplements:

From his article Rational approximations to $\sqrt[3]{2}$ and other algebraic numbers, A. Baker, 1964, I have found something useful to part of my questions but not all. In his introduction, he states a theorem of Roth that

''Suppose that $\alpha$ is an algebraic number, not rational, and that $\kappa > 2$. The well-known theorem of Roth implies the existence of $$ c=c(\alpha,\kappa)>0 $$ such that $$ \left | \alpha - \frac{p}{q} \right | > \frac{c}{q^{\kappa}} $$ for all rationals $\frac{p}{q} (q>0)$, but the method of proof does not allow $c$ to be determined explicitly...''

Hence, this result maybe a good choice for the proof of the finite solution of a large kinds of diophantine equations. But it is not an effective method for determining a computable range. Thanks to the work of A. Baker, he gives a corollary which says that

''For all rationals $\frac{p}{q} (q>0)$ we have $$ \left | \sqrt[3]{2} - \frac{p}{q} \right | > \frac{c}{q^{2.955}} , \tag{3} $$ where $c = 10^{-6}$.''

This can be directly applied to inequality $(1)$. However, any references with respect to inequality $(2)$ is still unknown for me, let alone other further references. This above is all what I have known now.


Edit $1$:

It is a nice surprise when I found this article: Rational approximations to $\sqrt[3]{2}$ and other algebraic numbers revisited, P. M. Voutier, 2007. From it, I know a much better version of $(3)$ which was given by M. Bennett in 1997 is as follows: $$ \left | \sqrt[3]{2} - \frac{p}{q} \right | > \frac{0.25}{q^{2.45}}. $$ However, P. M. Voutier has obtained a wide range of such results ( Corollary $2.2$ ):

''For the values of $n$ given in Table $1$, we have $$ \left | \sqrt[3]{n} - \frac{p}{q} \right | > \frac{c_2}{|q|^{\kappa +1}}, $$ for all integers $p$ and $q$ with $q \neq 0$ where $c_2$ and $\kappa$ are the values corresponding to $n$ in Table $1$.''

The Table $1$ illustrates the values of $n$ from $2$ to $100$ ( not all continuous 99 numbers ) with corresponding constants $c_2$ and $\kappa$. As an example, if $n=4$, then $c_2 = 0.41$ and $\kappa = 1.4325$. This is a concrete example of $(2)$ if we take $p = x$ and $q = 2^m$.

1

There are 1 best solutions below

0
On BEST ANSWER

The upper bound in your Theorem 1 has been sharpened to $$ n < 5.55 \log |D|, $$ provided $(n,D) \neq (3,-1)$ or $(15,7)$, in a paper of Bauer et al "Applications of the hypergeometric method to the generalized Ramanujan-Nagell equation" (Ramanujan J 2002). Similar arguments allow one to prove reasonably good inequalities of the shape (1) or (2) with $C_1(m)$ or $C_2(m)$ roughly $2^{-1.61m}$. As far as I know, nothing along these lines has been published.