10

Is it true that the

$$X^5=5 Y (Y+1)+1$$

has no non trivial solution (except the $X=1,Y=-1$ and $X=1,Y=0$) in integers. I tried to solve it as a quadratic equation on $Y$. I got that the $$\sqrt{5} \sqrt{4 X^5+1}$$ should be integer. But no idea how to continue from here.

Then I tried with Fermat Theorem $X^5\equiv X \pmod 5$ with hope to get some contradiction but again nothing useful.

EDITED: Just an observation from the @stuart stevenson answer: $X=10k+1$

So we have the following solutions:

$$Y\to \frac{1}{2} \left(\pm\sqrt{80000 k^5+40000 k^4+8000 k^3+800 k^2+40 k+1}-1\right)$$

Or the $$80000 k^5+40000 k^4+8000 k^3+800 k^2+40 k+1 = L^2 = \\ 1 + 40 k (1 + 20 k (1 + 10 k (1 + 5 k (1 + 2 k))))$$

This can be generalized as a recurrent product:

$$P_0(k)=2k+1 \\ P_n(k)=1+2^{n-1} 5kP_{n-1}(k)$$

And we assume that the $$P_4(k)\ne L^2$$ for any $k$.

EDITED:

I would assume that the $$X^p=p Y (Y+1)+1$$

has no non-trivial solution for any prime $p>2$.

EDITED:

Just some observation. It seems there are few $k$ for which the $$X^p=k Y (Y+1)+1$$

has solutions. For example for the $p=5$ and $p=7$ the $k$-s for which solution exists are $-1$ and $4$. And seems it has no solution for other $k$-s in the range $|k|<100$.

For the $p=3$ the $k$-s for which there is solutions are $\{-27, -21, -14, -7, -3, -1, 0, 1, 4, 6, 13, 27, 31\}$ for the range $|k|<50$ and seems for other $k$-s in this range there is no solution

  • 2
    I think that it is related to $x^5-x \equiv 0$ mod $10$ because the 5 is already there and $Y (Y+1)$ is even. – Leonhard Euler Jan 15 '18 at 00:50
  • Not sure how to use this. This looks like something promising but can not find how to complete this. – Gevorg Hmayakyan Jan 15 '18 at 00:52
  • 1
    Well $x^5 = 10 \frac{Y(Y+1)}{2} + 1$. And we know that $ \frac{Y(Y+1)}{2}$ is integer because it's $Y \choose 2$. – Leonhard Euler Jan 15 '18 at 00:54
  • This is very interesting, I'm just not sure about the next steps. – Gevorg Hmayakyan Jan 15 '18 at 01:01
  • 1
    Write out the binomial expansion for $(a+b)^5$. This should help. – Leonhard Euler Jan 15 '18 at 01:03
  • I am not an expert in elliptic curves, but I think we can try to solve this using elliptic curves. Trying this now. – Gevorg Hmayakyan Jan 16 '18 at 15:55
  • 2
    @GevorgHmayakyan This is a curve of genus 2, so not Elliptic curve which has genus 1. One way to know this is via linear transform into $$(2y+1)^2 = 4(x^5-1)/5 + 1 \implies Y^2 = f(X)$$ This is a Hyperelliptic curve since $f(X)$ has 5 distinct roots. From the link, it has genus 2.

    For genus 2, Falting's theorem says that there are only finitely many rational points, so same for integer points. However there are no known good ways of finding all of them.

    – Yong Hao Ng Jan 17 '18 at 10:18
  • 3
    For integral points, there are known upper bounds for $|X|,|Y|$ as in here, so looks like it is technically solvable via a search. However the bound seems astronomical in general.

    Perhaps special methods for selected curves is required.

    – Yong Hao Ng Jan 17 '18 at 10:47

2 Answers2

3

0. Overview
Following this old post, we attempt to solve the problem in the Number Field $K= \mathbb Q(\sqrt 5)$, which reduces to solving a set of $5$ Thue equations. We then solve them via PARI/GP which gives us the full solutions $(x,y)\in \{(1,-1),(1,0)\}$. In the following, let $w=\sqrt 5$ and $\alpha = (w+1)/2$.

1. Properties of $K$
$K$ has class number $1$, hence it is a PID and has unique factorization. As $5\equiv 1 \pmod 4$, the ring of integers is $R = \mathbb Z[(w+1)/2] = \mathbb Z[\alpha]$. The fundamental unit is $\eta = (w+ 1)/2 = \alpha$ and roots of unity is $\pm 1$. We remark that $\eta^{-1} = (w-1)/2$, as $$ \eta\eta^{-1} = \frac{w+1}{2}\frac{w-1}{2}=\frac{w^2-1}{4}=1 $$

2. Solving the original problem in $R$
We start by factorizing the original problem in $R$: $$ \begin{align*} (wy+\frac{w+1}{2})(wy + \frac{w-1}{2}) &= w^2y^2 + w(\frac{w+1}{2}+\frac{w-1}{2})y+\frac{w+1}{2}\frac{w-1}{2} \\ &= w^2 y + w^2 y + \frac{w^2-1}{4}\\ &= 5y^2 + 5y + 1 \\ &= x^5 \end{align*} $$ Denote $A = wy + (w+1)/2$ and $B = wy + (w-1)/2$, noting that $A,B\in R$. Since we have $$ A - B = 1, $$ $(A)+(B)$ generates $R$ and we conclude that $A,B$ are coprime.

Let the unique factorization of $x$ into irreducibles be $$ x = \pm \eta^k I_1^{e_1} I_2^{e_2} \cdots I_m^{e_m} $$ (where $I_j$ are the irreducibles). Then $$ AB = x^5 = \pm \eta^{5k} I_1^{5e_1} I_2^{5e_2} \cdots I_m^{5e_m} $$ Since $A,B$ are coprime, each irreducible power divides exactly one of $A$ or $B$. By relabelling $\{I_j\}$, we may assume that exactly the first $s$ irreducibles divide $A$, therefore $A$ has the form $$ A = \pm \eta^l I_1^{5e_1}I_2^{5e_2} \cdots I_s^{5e_s} = \pm \eta^l (a+b\alpha)^5,\;\;\;\;l,a,b\in\mathbb Z $$ since product of irreducibles are in $R$ and hence of the form $a+b\alpha$. We can factor $-1$ and powers of $\eta^5$ into $(a+b\alpha)^5$, hence a simplified form would be $$ A = \eta^l (a+b\alpha)^5,\;\;\; l,a,b\in\mathbb Z,\;\;\;\; 0\leq l\leq 4 $$ Remark: we can factor $\eta^{-5}$ into $(a+b\alpha)^5$ too since $\eta^{-1} = (w-1)/2$, which is why we can assume $0\leq l \leq 4$.

3. Solution via Thue equations
For our original problem to have a solution, there must be $y,l,a,b\in \mathbb Z$ (and $0\leq l\leq 4$) such that the equation below holds: $$ \begin{align*} wy + \frac{w+1}{2} &= A = \eta^l (a+b\alpha)^5 \\ \frac{1}{2} + (y+\frac{1}{2})w &= \left(\frac{w+1}{2}\right)^l \left(a + b\frac{w+1}{2}\right)^5 \\ \end{align*} $$

For each $l$ in $[0,4]$, equating the rational constant gives us a degree $5$ Thue equation, which we can solve using computational tools (using PARI/GP in our case). For example, letting $l=0$ we obtain: $$ \begin{align*} \frac{1}{2} + (y+\frac{1}{2})w &= \left( a^5 + \frac{5 a^4 b}{2} + 15 a^3 b^2+20 a^2 b^3 + \frac{35 a b^4}{2} + \frac{11 b^5}{2} \right) \\ &+ w\left( \frac{5 a^4 b}{2} + 5 a^3 b^2+10 a^2 b^3+ \frac{15 a b^4}{2}+ \frac{5 b^5}{2} \right) \\ \implies 1 &= 2 a^5+5 a^4 b+30 a^3 b^2+40 a^2 b^3+35 a b^4+11 b^5\\ \end{align*} $$ We solve this on the online PARI/GP calculator, via the command:

thue(thueinit(2*a^5 + 5*a^4 + 30*a^3 + 40*a^2 + 35*a + 11,1),1)

(The first '1' means to assume nothing.) This returns
%1 = []
which tells us that there are no solutions.

For the 4 remaining equations corresponding to $1\leq l\leq 4$, they are: $$ \begin{align*} l &= 1:& 1 &= a^5+15 a^4 b+40 a^3 b^2+70 a^2 b^3+55 a b^4+18 b^5 \\ l &= 2:& 1 &= 3 a^5+20 a^4 b+70 a^3 b^2+110 a^2 b^3+90 a b^4+29 b^5 \\ l &= 3: & 1 &= 4 a^5+35 a^4 b+110 a^3 b^2+180 a^2 b^3+145 a b^4+47 b^5 \\ l &= 4: & 1 &= 7 a^5+55 a^4 b+180 a^3 b^2+290 a^2 b^3+235 a b^4+76 b^5 \end{align*} $$ For convenience, the corresponding PARI/GP commands are
thue(thueinit(a^5+15*a^4+40*a^3+70*a^2+55*a+18,1),1)
thue(thueinit(3*a^5+20*a^4+70*a^3+110*a^2+90*a+29,1),1)
thue(thueinit(4*a^5+35*a^4+110*a^3+180*a^2+145*a+47,1),1)
thue(thueinit(7*a^5+55*a^4+180*a^3+290*a^2+235*a+76,1),1)

PARI/GP tells us that the only solutions are $(a,b) = (1,0)$ for $l=1$ and $(a,b) = (1,-1)$ for $l=4$. Putting in these values, we get the values of $A$ as $$ \begin{align*} (l,a,b) &= (1,1,0) &\implies A &= \left(\frac{w+1}{2}\right)^1(1)^5 = \frac{1}{2} + \frac{w}{2} \\ (l,a,b) &= (4,1,-1) &\implies A &= \left(\frac{w+1}{2}\right)^4 \left(1 - \frac{w+1}{2}\right)^5 = \frac{1}{2} - \frac{w}{2} \end{align*} $$ Recalling that $A=(1/2)+(y+1/2)w$, this in turn gives $y= 0$ and $y=-1$ respectively. Putting $y =0$ and $y=-1$ into our original equation, we get $x^5=1$ for both and hence $x=1$. This concludes that the only solutions are $$ (x,y) \in \{(1,-1),(1,0)\} $$

Remark: Perhaps it may be possible to solve the Thue equations manually so that this proof does not rely on computational tools.

Yong Hao Ng
  • 4,795
  • 19
  • 26
1

So $x^5 = 5y(y+1)+1$. We know then that $x^5 = 10 \frac{y(y+1)}{2}+1$ and the fractional part is integer as I have already said.

Therefore the units part of $x^5$ is $1$. We know (it's trivial to prove or you can look it up) that $x^5 \equiv x$ mod $10$. Therefore $x$ has unit part $1$ aswell.

Therefore $x = 10k+1$ where $k$ is an integer greater than $0$.

Not intended as a full answer but to demonstrate my commented ideas.

  • $x^5=5y(y+1)+1$ I checked this up to $y=10^7$ by Python and there was no integer solution for x. – sirous Jan 17 '18 at 16:41
  • 3
    With all due respect, I don't think that this is particularly helpful. We already know that it is the case that it is true for all but trivial cases. It is proving it for all cases that is tricky. – Leonhard Euler Jan 17 '18 at 17:02