2
$\begingroup$

Suppose that $f(x,y) \in \mathbb{Z}[x,y]$ is a homogeneous polynomial, or binary form, of degree $d$. The equation $$f(x,y) = h$$ for a given integer $h$ is known as Thue's equation (so named because Thue showed that this equation always has a finite number of solutions in integers $x,y$). There have been much work done subsequently to improve the bound on the number of solutions. As far as I know, the best general result is due to Stewart, who showed that the number of solutions is no more than $$2800\left(1 + \frac{1}{8 \varepsilon d}\right)d^{1 + \omega(g)},$$ where $\varepsilon$ is some positive constant and $g$ is a divisor of $h$ that is sufficiently large with respect to the discriminant of $f$ (and other invariants). One should interpret the above result as saying that when $h$ has a very large prime divisor or a large prime power divisor, the result is essentially that the number of solutions is $O(d^2)$. Indeed, in cases when $g$ can be chosen as 1, we even get the bound $O(d)$, which we know is best possible in general (due to an example of Bombieri and Schmidt, for example).

I am particularly interested in the case when $h$ is small, but the solution $(x_0, y_0)$ is large. It is not too hard to see that one of $x_0$ or $y_0$ can be made arbitrarily large in some cases. For instance, consider the equation $$(m^{kd} + 1)x^d - y^d = 1$$ which obviously has the solution $(1, m^k)$. Hence there are equations where $y$ can be arbitrarily large compared to $h$.

However, note that while $y$ is large with respect to 1, it is small compared to the coefficients. My question concerns this type of behaviour.

1) Suppose that a Thue equation $f(x,y) = h$ has a solution with $y$ (say) being much larger than $h^{1/d}$. Do we necessarily have that the coefficients of $f$ have to be comparably large?

2) Suppose we have a Thue equation $f(x,y) = h$ where $h$ is very small compared to the height $\lVert f \rVert$ of $f$, which is the largest of the absolute values of the coefficients of $f$. Is there a uniform bound (that is, independent of even the degree of $f$) that can be established for the number of solutions where at least one coordinate is very large compared to $h$ (but not necessarily to $\lVert f \rVert$)?

Thank you for any insights.

$\endgroup$
1
  • $\begingroup$ How do you define "much larger" for h^(1/d)? $\endgroup$
    – joro
    Dec 6, 2014 at 10:18

2 Answers 2

3
$\begingroup$

I'm not sure whether the second question is inadvertently misworded. As it is, the answer can easily be shown to be no. Fix $h$ and let $A$ be huge compared to $h$. Let $$ f(x,y)=A(x-y)(x-2y)\cdots (x-ny)+h y^n. $$ Then the Thue equation $f(x,y)=h$ has the $n$ solutions $(1,1),(2,1),\dots,(n,1)$. Here $h$ is very small compared to $\lVert f \rVert \gtrsim n! A$. So we cannot expect a bound on the number of solutions independent of the degree of $f$.

Added in response to the edited version of Question 2: The answer is still no. Take $n$ to be huge compared to $h^2$. Now in the above, the solutions $(\lfloor \sqrt{n} \rfloor,1),\dots,(n,1)$ all have a coordinate that is huge compared to $h$ and their number is roughly $n$.

$\endgroup$
1
  • $\begingroup$ I added a condition to eliminate this particular counter-example, which as you noted is quite obvious. $\endgroup$ Dec 7, 2014 at 1:21
1
$\begingroup$

I think you need $d>2$.

The answer to the first question is "no". For $F_n$ the Fibonacci numbers, $x,y=F_{2n},F_{2n+1}$ satisfy $x^2+xy-y^2+1=0$

Take the degree $3$ Thue equation $f(x,y)=x(x^2+xy-y^2)$.

For Fibonacci numbers as above take we have $f(x,y)=x$.

Take $h=x$. We have $x \gg x^{1/3}$ and all coefficients are units.


If you want $f$ irreducible, take $d=3$ and $f$ with real irrational root, then use Voloch's answer here


To the edited question about uniform bound.

I think again no.

In Siksek's answer replace $x-my$ by $x-m^{2^k}y$. You have the solutions $(m^{2^k},1)$ for fixed $h$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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