Prob: Show that for any positive integer $a$, Diophantine equation $$x^3+x+a^2=y^2$$ has at least one solution $(x, y)$, where $x, y$ are positive integers.

Source: My teacher.

Attempt: First I tried $a=1$ and found the minimal solution $(x, y)=(72, 611)$, not a friendly one.

Now rewrite the equation as $$x(x^2+1)=(y-a)(y+a).$$ We hope that $x=b_{1}b_{2}, x^2+1=c_{1}c_{2}$ and $b_{2}c_{2}-b_{1}c_{1}=2a$. But how to determine these numbers? I got stuck here. One way promising is to relate to a Pell-type equation.

Another thought is to set $x$ to be some polynomial like $2t^2$ so that the left side can be factorized furthermore. Still little progress.

Please help.

  • $\begingroup$ The curve has the integer point $(0,a)$. At first glance this looks like an inflection point, but it's not. The tangent line intersects the curve at another point: $(1/4a^2,(1+8a^4)/8a^3)$. Perhaps these points can be used to generate another integer point? $\endgroup$ – B. Goddard Aug 24 '18 at 12:06
  • $\begingroup$ @B.Goddard Good point, but I don't think it can be very useful. We may consider it from the perspective of number theory rather than algebra. $\endgroup$ – J.Guldan Aug 24 '18 at 12:49

Your first idea is a great idea! When working this out, I had slightly different variables, but it is equivalent to your starting.

Suppose positive integers $b, c, d, e$ satisfy $y+a=bc$, $y-a=de$, $x=bd$, $x^2+1=ce$, $bc-de=2a$, $ce-(bd)^2 = 1$.

I propose the following lemma:

The sequence $k_n$ is defined as follows: $k \in \mathbb{N}$, $k_0 = 0$, $k_1=1$, $k_{n+2} = kk_{n+1} + k_{n}$ for $n \geq 0$. Then $k_{n-1} k_{n+1} - k_{n}^2 = (-1)^{n} \quad \forall n \geq 1$.


For $n=1$, this holds true. Now, assume true for $n$. Then for $n+1$,

$$\begin{aligned} k_{n} k_{n+2} - k_{n+1}^2 &= k_{n} (kk_{n+1} + k_{n}) - k_{n+1}^2\\ &= -k_{n+1} (k_{n+1} - kk_{n}) + k{n}^2\\ &= -(k_{n-1} k_{n+1} - k_{n}^2)\\ &= -(-1)^k\\ &= (-1)^{k+1} \end{aligned}$$

Thus, proven.

Now, if we let $c=k_3 = k^2+1$, $e=k_5=k^4+3k^2+1$, $bd=k_4 = k(k^2+2)$, then from the above lemma, we have $ce-(bd)^2=1$. Letting $d=2a, k=4a^2$, we get $b=2a(k^2+2)$ and $bc-de=2a(k^2+2)(k^2+1)-2a(k^4+3k^2+1) = 2a$. Thus, we have the result

$$x=bd=8a^2(8a^4+1)$$ $$y=4a(8a^4+1)(16a^4+1)-a$$

as a construction for all $a$, and does satisfy the solution you provided for $a=1$.

  • $\begingroup$ Very through answer indeed! $\endgroup$ – NoChance Jun 20 at 10:48

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.