# Solving the nonlinear Diophantine equation $x^2-3x=2y^2$

How can I solve (find all the solutions) the nonlinear Diophantine equation $x^2-3x=2y^2$?

I included here what I had done so far. Thanks for your help.

Note: The equation above can be rewritten into $x^2-3x-2y^2=0$ which is quadratic in $x$. By quadratic formula we have the following solutions for $x$.

$$x=\frac{3\pm\sqrt{9+8y^2}}{2}$$

I want $x$ to be a positive integer so I will just consider: $$x=\frac{3+\sqrt{9+8y^2}}{2}$$

From here I don't know how to proceed but after trying out values for $1\leq y\leq 1000$ I only have the following $y$ that yields a positive integer $x$.

$$y=\{0,3,18,105,612\}.$$

Again, thanks for any help.

Multiply through by $4$ and complete the square, $(2x-3)^2 - 8 y^2 = 9,$ so that $(2x-3)^2 + y^2 \equiv 0 \pmod 3.$ It follows that both $(2x-3)$ and $y$ are divisible by $3,$ therefore $x$ as well. Write $x=3t, y=3v.$ So far $(2t-1)^2 - 8 v^2 = 1.$ Let $u=2t-1,$ so $u^2 - 8 v^2 = 1.$

...if we write $$x = \frac{3(1+u)}{2}$$ and $$y = 3v,$$ we get $$u^2 - 8 v^2 = 1.$$ This gives a sequence of pairs, $u=1,3,17,99,...$ and $v=0,1,6,35...$

These obey $$u_{n+1} = 3 u_n + 8 v_n,$$ $$v_{n+1} = u_n + 3 v_n.$$ By Cayley-Hamiton $$u_{n+2} = 6 u_{n+1} - u_n,$$ $$v_{n+2} = 6 v_{n+1} - v_n.$$ Then

$$x_{n+2} = 6 x_{n+1} - x_n - 6,$$ $$y_{n+2} = 6 y_{n+1} - y_n.$$ The (non-negative) $(x,y)$ pairs begin $$(3,0)$$ $$(6,3)$$ $$(27,18)$$ $$(150,105)$$ $$(867,612)$$ $$(5046,3567)$$

• Thanks Sir @WillJagy for the comprehensive answer. This is exactly the answer I am looking for. Thanks again. – Jr Antalan Sep 17 '15 at 21:48
• Good Day Sir @WillJagy. I just want to ask if you can suggest a reference in helping me solving equations of this type: $x^2-3x=2y^n$ where $n\geq 2$. This is a generalization of my question here on MSE. Thanks a lot. – Jr Antalan Sep 20 '15 at 0:13
• @JrAntalan, no particular reference. When $n=3,$ it might be (I'm not sure) that it separates into a small number of Mordell curves, in which case the answers are sometimes tabulated on websites. For $n \geq 4$ i have no idea, except to point out the great likelihood of finiteness. – Will Jagy Sep 20 '15 at 0:22
• ok Sir, thanks a lot. – Jr Antalan Sep 20 '15 at 0:30
• Sir, I am trying to revisit your answer and it was brilliant! But I am curious about your method on solving this non-homogeneous Diophantine Equation. Is there a general method on solving 2nd order non-homogeneous Diophantine Equations such as the given here? If so, can you recommend some readings on it Sir? I do not find any method on my search over the web. Thank you so much Sir, I will also send some message thru email Sir about this topic. Again thank you. – Jr Antalan Oct 24 '15 at 12:34

It is evident that equation.

$$x^2-3x=2y^2$$

This is a Pell equation. And we have to use solutions which are determined by the sequence. The next value from the previous one.

$$p_2=3p_1+4s_1$$

$$s_2=2p_1+3s_1$$

If we use the sequence where the first number. $(p_1 ;s_1) - (1 ; 1)$

Decisions will be.

$$x=6s^2$$

$$y=3ps$$

If we use the sequence where the first number. $(p_1 ;s_1) - (3 ; 2)$

Decisions will be.

$$x=3p^2$$

$$y=3ps$$

• You mean "solutions", not "decisions". – user236182 Sep 18 '15 at 6:23

I'll solve it in integers instead. $x^2-3x=2y^2$, by the quadratic formula, is equivalent to

$$x=\frac{3\pm\sqrt{9+8y^2}}{2}$$

So i.e. the problem is equivalent to solving $8y^2+9=m^2$ in integers.

$8y^2\equiv m^2\pmod{3}$, so $(y,m)=\left(3y_1,3m_1\right)$ for some $y_1,m_1\in\Bbb Z$, because $h^2\equiv \{0,1\}\pmod{3}$ for any integer $h$.

The problem is equivalent to $m_1^2-8y_1^2=1$ with $m_1,y_1\in\Bbb Z$. This is a Pell's Equation and has infinitely many solutions given exactly by $\pm\left(3+\sqrt{8}\right)^n=x_n+y_n\sqrt{8}$ for $n\ge 0$ (because $(m_1,y_1)=(3,1)$ is the minimal non-trivial solution), so for example $\pm\left(3+\sqrt{8}\right)^2=\pm17\pm6\sqrt{8}$ and $(m_1,y_1)=(\pm 17,\pm 6)$ is another solution.

In general, Pell's Equation $x^2-dy^2=1$ with $d$ not a square has infinitely many solutions; which are given exactly by $\pm\left(x_m+ y_m\sqrt{d}\right)^n=x_n+y_n\sqrt{d}$, $n\ge 0$, where $(x_m,y_m)$ is the minimal solution, i.e. the solution with $x_m,y_m$ positive integers and minimal value of $x_m+ y_m\sqrt{d}$.

• thanks for the answer @user236182 . I have a question from your answer since Im a bit confused. the expression $3\cdot [(3+\sqrt{8}^n)]_{\sqrt{8}}$ is always an integer? I am not familiar with the notation$_{\sqrt{8}}$ – Jr Antalan Sep 17 '15 at 14:09
• @JrAntalan I created the notation myself. I thought the example would make it clear, sorry for the confusion. – user236182 Sep 17 '15 at 14:10
• it is okay, thanks for the answer. an up vote for that, will consider it as a correct answer once I understand the whole thing. Thanks again. – Jr Antalan Sep 17 '15 at 14:12
• @JrAntalan Please tell me if anything is unclear about my answer. I've edited it. – user236182 Oct 25 '15 at 4:46