I would like to understand ${\bf nonnegative \ integral}$ solutions $(x,y,z)$ on the surface $$xyz-ax-ay-bz=d.$$ where $a,b,d$ are positive integers.

I can certainly prove that for a fixed $z$ there are finitely many non-negative solutions. But I'm having difficult time proving that it has finitely many solutions, even for some special triples $a,b,d$.


If $x=0$, then $-ay-bz=d$ where LHS is non-positive while RHS is positive. So, we have $x\not=0$. We have $y\not=0$ and $z\not=0$ similarly.

Now, we may suppose that $y\ge x\gt 0$ and $z\gt 0$.

If $\min(x,y,z)=z$, then $$1=\frac{a}{yz}+\frac{a}{xz}+\frac{b}{xy}+\frac{d}{xyz}\le \frac{a}{z^2}+\frac{a}{z^2}+\frac{b}{z^2}+\frac{d}{z^3}$$ $$\implies z^3-(2a+b)z-d\le 0$$ The number of $z$ satisfying this inequality is finite.

Consindering $$(zx-a)(zy-a)=bz^2+dz+a^2$$ for each $z$, we see that the number of $(x,y)$ is finite.

If $\min(x,y,z)=x$, then, similarly as above, the number of solutions is finite.

Example :

For your example where $a=9,b=4,d=37$, if $\min(x,y,z)=z$, we have $$z^3-22z-37\le 0\implies z\le 5$$

For each $z=1,2,\cdots, 5$, consider $$(zx-9)(zy-9)=4z^2+37z+81$$

For example, for $z=1$, we have $$(x-9)(y-9)=2\times 61$$ $$\implies (x-9,y-9)=(1,2\times 61),(2,61)$$ $$\implies (x,y)=(10,131),(11,70)$$ since we already supposed that $y\ge x$.

  • $\begingroup$ @mathlove.Nice.Thanks. $\endgroup$ – Jeff Apr 23 at 17:30

There will always be finitely many (positive) solutions.

Supposing $a, b, c, d$ are all positive integers, then the positive integer solutions to

$$xyz = ax + by + cz + d \tag{1}$$

are bounded, for the following reason:

Observe that $(1)$ is equivalent to

$$\begin{align} z & = {ax + by + d \over xy + c} \\ & \le {ax + by + d \over xy} \tag2 \\ & = \frac ay + \frac bx + \frac d{xy} \\ & \le a + b + d \tag3 \end{align}$$

$(2)$ because the new denominator is smaller (since $c \gt 0$), and $(3)$ similarly because $x, y \ge 1$.

By an identical argument you will find $1 \le x \le b + c + d$ and $1 \le y \le a + c + d$

Let $M = 3 \cdot \max \{a, b, c, d\}$ just for simplicity's sake; then all positive integer solutions to $(1)$ are bounded within the region $[1, M] \times [1, M] \times [1, M]$, within which there are only finitely many ($M^3$) integral points

  • $\begingroup$ You may have to do some further work beyond this for the case when one or more of $x, y, z$ are zero, but I suspect that will be a simpler problem/degenerate case $\endgroup$ – Rob Bland Apr 23 at 16:12
  • 1
    $\begingroup$ I believe it should be $z = \frac{ax+by+d}{xy\color{red}{-}c}$ $\endgroup$ – lisyarus Apr 23 at 16:13
  • $\begingroup$ Whoops, oh dear $\endgroup$ – Rob Bland Apr 23 at 17:32

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.