0

I need help moving forward with this question:

Determine the number of solutions $(x, y, z, w)$ of integers with each of $x$, $y$, $z$, and $w$ between $-10$ and $10$ inclusive, using these equations: $x + y + z = w; \frac1x + \frac1y + \frac1z = \frac1w.$

So far, I have gotten rid of the denominators and substituted $w = x + y + z$ in and have $2xyz + x^2z + x^2y + y^2x + y^2x + z^2x + z^2y$ and I have no idea where to go next.

amWhy
  • 210,739
  • 2
    Hint: Can you find some obvious numerical solutions? That could help you guess what identity you want to prove, which would hopefully make the factorization of your expression much more obvious. – Calvin Lin Aug 01 '24 at 20:46
  • Can you include information about where you encountered this question, and credit the original source where you encountered it? – D.W. Aug 02 '24 at 05:57
  • My own numerical evaluation returns $1140$ solutions. https://colab.research.google.com/drive/10r9jxVJGiVwSMt_y0V1GtRFeTvtYcUSl?usp=sharing – ducbadatcs Aug 02 '24 at 08:16
  • In M2 R=QQ[x,y,z,w] I=ideal(x+y+z-w,y*z*w+x*z*w+x*y*w-x*y*z) primaryDecomposition I -- {ideal(z−w,x+y),ideal(y−w,x+z),ideal(y+z,x−w)} The union of three planes in ${\Bbb A}^4$ (or three lines in ${\Bbb P}^3$) – Jan-Magnus Økland Aug 02 '24 at 09:15
  • 1
    Regarding this post, I see a lack of organization in MSE that should never occur with good managers of this useful and interesting blog. There are two answers, one of which was wrong first and does not give the definitive answer requested, has 3 upvotes, but the other one that does give the correct answer has only 1 upvote. At least the O.P. should compensate for this obvious irregularity by qualifying this answer as the best (in reality it is the only valid answer given). – Ataulfo Aug 02 '24 at 21:48

2 Answers2

3

building off of the second equation we have :

$$ \frac{1}{x} + \frac{1}{y} + \frac{1}{z} = \frac{1}{x+y+z} \Rightarrow (yz+xz+xy)(x+y+z) = xyz $$ above was attained by multiplying both sides by $xyz(x+y+z)$ we have: $$ 2xyz + yz(y+z) + xz(x+z) + xy(x+y) = 0 $$

note that: $$ 2xyz + yz(y+z) + xz(x+z) + xy(x+y) = x(2yz + z^2 + y^2 + x(y+z)) + yz(y+z) = x((y+z)^2 + x(y+z)) + yz(y+z) = (y+z)(xy + xz + x^2 + yz) $$

so : $$ (y+z)(x^2 + xy + yz + xz) = 0 $$ Also note that : $$ x^2 + (y+z)x + yz = (x+y)(x+z) $$

so this simplifies to : $$ (x+y)(y+z)(x+z) = 0 $$ this means that the answer is : $$ A = \{(x, y, z, w) | (x+y)(y+z)(x+z) = 0, x, y, z\neq 0,-10\leq x, y, z \leq 10\} $$

there are many possible answers. including but not limited to : $$ (1, 1, -1, 1), (1, 2, -2, 1), (1, 3, -3, 1) ... $$

FXDMLR
  • 41
1

From your equations we have $$\begin{align}x+y+z &= w\\ w(xy+yz+zx) &= xyz \end{align}$$ Eliminating $w$, $$(x+y+z)(xy+yz+zx)=xyz$$ Compare with the well-known identity $(x+y+z)(xy+yz+zx)=xyz+(x+y)(y+z)(z+x)$ which is true for all values of $x,y,z$. It follows that our system is equivalent to the following equation:

$$\begin{align}(x+y)(y+z)(z+x)&=0 && \text{where }0\notin \{x,y,z\}\end{align}$$

Thus, $w$ is redundant (ie, always fixed after determining $x,y,z$) and any solution is just an ordered $3$-tuple of form $(x,y,z)$. Thus we have to construct triples with each element as a non-zero integer in $[-10,10]$ s.t. some two elements must add to $0$.

For this count, we first consider tuples of the form $(a,-a,b)$ in some order where $|b| \neq |a|$. First we choose the $\color{blue}{\text{position and value of the $b$ element}}$; then $\color{orange}{\text{count all posible permutations of $(a,-a)$}}$. Thus there are $\color{blue}{3\cdot 20}\cdot\color{orange}{9\cdot 2}=\color{magenta}{1080}$ such tuples. Then we consider tuples of form $(a,-a,a)$ in some order. We first $\color{blue}{\text{choose the value of $a$}}$ and then $\color{orange}{\text{choose the position of the negative sign}}$. Thus there are $\color{blue}{20}\cdot \color{orange}{3} = \color{magenta}{60}$ such tuples.

Thus, there are exactly $\color{magenta}{1080+60} = \color{teal}{1140}$ solutions.

Sahaj
  • 5,355
  • P.S.: This algebraic identity mentioned in my post is the crucial part of the solution which I’ve used in many of my other answers. To prove it one can consider writing $(a+b)(b+c)(c+a)$, a symmetric polynomial, in terms of the elementary symmetric polynomials via procedures described here. – Sahaj Aug 01 '24 at 23:12
  • 1
    I think this is better. Thanks – Ross Millikan Aug 02 '24 at 04:49
  • How are there 2 * 9 permutations of (a, -a). Shouldn’t it be 2 * 20? – user1366989 Aug 02 '24 at 12:28
  • @user1366989 first of all the condition is $|b|\neq |a|$. So there’s $18$ values that $a$ can take. So you would think there are $18\cdot 2$ permutations of $(a,-a)$ … but that is a double count, so there are exactly $18$ such permutations. – Sahaj Aug 02 '24 at 12:30
  • It is a double count since when we have $(-a,a)$ we permute it to $(a,-a)$ and to count it we multiply by $2$, but we already counted $(a,-a)$ since $a$ can be from $[-10,10]$ and a non-zero integer. – Sahaj Aug 02 '24 at 12:31