The system says $$x+y+z=0$$ $$xy +xz+yz=-1$$ $$xyz=-1$$ Find $$x^8+y^8+z^8$$

With the first equation I squared and I found that $$x^2+y^2+z^2 =2$$ trying with

$$(x + y + z)^3 = x^3 + y^3 + z^3 + 3 x^2 y + 3 x y^2 + 3 x^2 z++ 3 y^2 z + 3 x z^2 + 3 y z^2 + 6 x y z$$ taking advantage of the fact that there is an $xyz=-1$ in the equation, but I'm not getting anywhere, someone less myopic than me.how to solve it?


Edit : Will there be any university way to solve this problem , they posed it to a high school friend and told him it was just manipulations of remarkable products. His answers I understand to some extent but I don't think my friend understands all of it.

  • 6
    $\begingroup$ Hint: $\,x,y,z\,$ are the roots of $\,t^3-t+1=0\,$ by Vieta's relations, so $\,x^3=x-1\,$ and $\dots$ $\endgroup$
    – dxiv
    Sep 9 at 22:34
  • 3
    $\begingroup$ Similarly, but alternatively, you could figure out the polynomial (using Vieta) that has $x^2, y^2, z^2$ as roots. If you feed the resulting polynomial into this same method, you'll get a polynomial with $x^4, y^4, z^4$ as roots. Then after one more application of this method... $\endgroup$ Sep 9 at 22:44
  • 2
    $\begingroup$ We can also use Girard -newton identities . $\endgroup$
    – Sukhoi234
    Sep 9 at 22:46
  • 3
    $\begingroup$ To continue your approach, you may find it handy to expand $(x + y + z)(xy + yz + xz)$. $\endgroup$ Sep 9 at 23:13
  • 1
    $\begingroup$ @Nabla - You are propably aware of this: There is an explicit formula for the set of zeros of $f(t)=t^3-t+1$: You get an explicit formula for $x,y,z$ and may calculate any power $x^n+y^m+z^k$. $\endgroup$
    – hm2020
    Sep 10 at 8:23

The following is an elementary, entirely self-contained solution which does not assume knowledge of Vieta's relations, Newton's identities, or anything other than simple algebra. (This is not the best or recommended way to solve it - see the other answers and comments for that - but it answers OP's edit asking for a more basic solution.)

  • From $x+y+z=0$ it follows that:

$$y+z = -x \tag{1}$$

  • From $xy +xz+yz=-1$ and $(1)$ it follows that: $$-1=x(y+z)+yz=-x^2 + yz \;\;\implies\;\; yz = x^2-1 \tag{2}$$

  • From $xyz=-1$ and $(2)$ it follows that:

$$-1=xyz=x(x^2-1)=x^3-x \;\;\implies\;\; x^3=x-1 \tag{3}$$

Multiplying $(3)$ by $x$ gives:

$$x^4=x^2-x \tag{4}$$

Squaring $(4)$ and using that $x^3=x-1$ per $(3)\,$:

$$ \begin{align} x^8 &= x^4-2x^3+x^2 \\ &=x \cdot x^3-2x^3+x^2 \\ &=x(x-1)-2(x-1)+x^2 \\ &=2x^2-3x+2 \tag{5} \end{align} $$

Repeating the steps $(1)\dots(5)$ for the other variables, it follows by symmetry that:

$$ \begin{align} y^8 = 2y^2-3y+2 \tag{6} \\ z^8 = 2z^2-3z+2 \tag{7} \end{align} $$

Adding $(5)+(6)+(7)\,$ and using that $x+y+z=0$:

$$ \require{cancel} \begin{align} x^8+y^8+z^8 &= 2(x^2+y^2+z^2) - \cancel{3(x+y+z)} + 6 \\ &= 2\left(\bcancel{(x+y+z)^2}-2(xy+yz+zx)\right)+6 \\ &= -4 \cdot (-1) + 6 \\ &= 10 \end{align} $$

  • 1
    $\begingroup$ I like this approach which is simpler and more in spirit of modern algebra ($x$ is a root of a third degree polynomial so any rational function of $x$ can be written as a quadratic polynomial in $x$). +1 $\endgroup$ Sep 11 at 13:12

If you're not familiar to Girard-Newton identities, a recurrence relation may help:

\begin{align} \qquad a^n+b^n+c^n & = (a+b+c)(a^{n-1}+b^{n-1}+c^{n-1}) \\ & \quad-(bc+ca+ab)(a^{n-2}+b^{n-2}+c^{n-2})+abc(a^{n-3}+b^{n-3}+c^{n-3}) \end{align}

By successive iterations, you can raise up to the desired order.

Using SymmetricReduction[x^8+y^8+z^8,{x,y,z}] in Wolfram Alpha, the result is here.

  • $\begingroup$ Is that supposed to be $$a^n + b^n + c^n = (a + b + c)(a^{n-1} + b^{n-1} + c^{n-1}) - (bc + ca + ab)(a^{n-2} + b^{n-2} + c^{n-2}) + abc(a^{n-3} + b^{n-3} + c^{n-3}),$$ @NgChungTak? $\endgroup$ Sep 11 at 9:44
  • 1
    $\begingroup$ @ArnieBebita-Dris Corrected with thanks. $\endgroup$ Sep 11 at 10:52

Denoting with $e_k(x,y,z), k\geq 0$ the elementary symmetric polynomials \begin{align*} e_1(x,y,z)&=x+y+z=0\\ e_2(x,y,z)&=xy+xz+yz=-1\tag{1}\\ e_3(x,y,z)&=xyz=-1 \end{align*} and with $p_k(x,y,z), k\geq 0$ the $k$-th power sum \begin{align*} p_k(x,y,z)=x^k+y^k+z^k\tag{2} \end{align*} we recall Newtons identities admit a generating function representation of the power sums as \begin{align*} \color{blue}{\sum_{k=1}^\infty(-1)^{k-1}p_k\frac{t^k}{k}=\ln(1+e_1t+e_2t^2+e_3t^3+\cdots)}\tag{3} \end{align*}

We obtain from (1) - (3) \begin{align*} \color{blue}{\left.x^8+y^8+z^8\right|_{{{e_1=0\ \ }\atop{e_2=-1}}\atop{e_3=-1}}} &=(-8)[t^8]\ln\left(1-t^2-t^3\right)\\ &=(-8)[t^8]\ln\left(1-t^2(1+t)\right)\\ &=8[t^8]\sum_{j=1}^\infty\frac{t^{2j}(1+t)^j}{j}\tag{4.1}\\ &=8\sum_{j=1}^{4}\frac{1}{j}[t^{8-2j}](1+t)^j\tag{4.2}\\ &=8\sum_{j=1}^4\frac{1}{j}\binom{j}{8-2j}\tag{4.3}\\ &=8\left(\frac{1}{3}\binom{3}{2}+\frac{1}{4}\binom{4}{0}\right)\tag{4.4}\\ &\,\,\color{blue}{=10} \end{align*}


  • In (4.1) we expand the logarithmic series.

  • In (4.2) we apply the rule $[t^{p-q}]A(t)=[t^p]t^qA(t)$. We also set the upper limit of the series to $4$ since other terms do not contribute.

  • In (4.3) we select the coefficient of $t^{8-2j}$.

  • In (4.4) we use $\binom{p}{q}=0$ if $0<p<q$ are positive integers.


Note that $x$, $y$, $z$ are the roots of the polynomial $\lambda^3 - \lambda + 1$, so they are the eigenvalues of the companion matrix $$A=\left(\begin{matrix} 0 & 0 & -1 \\ 1 & 0 & 1\\ 0 & 1 & 0 \end{matrix} \right)$$ Conclude that the eigenvalues $A^{8}$ are $x^8$, $y^8$, $y^8$. Calculate $A^8$ by squaring three times ( @Theo Bendit:'s idea in the comments). We get $$A^8=\left(\begin{matrix} 2 & -2 & 3 \\ -3 & 4 & -5\\ 2 & -3 & 4 \end{matrix} \right)$$ so the sum $x^8+y^8+z^8$ equals the trace of $A^8$, that is $10$.

  • $\begingroup$ , you can clarify from here " We conclude that the eigenvalues...." $\endgroup$
    – Nabla
    Sep 10 at 21:17
  • $\begingroup$ @Nabla: The matrix $A$ can be diagonalized, with $x$, $y$, $z$ on the diagonal. So $A^8$ can also be diagonalized, with $x^8$, $\ldots$ on the diagonal. This is true for companion matrices $A$ of a polynomial with simple roots. $\endgroup$
    – orangeskid
    Sep 10 at 21:22
  • $\begingroup$ I think it would read better like this: "... $x$, $y$, $z$ are the roots of $\lambda^3 - \lambda + 1$, so they are the eigenvalues of the companion matrix..." $\endgroup$
    – NickD
    Sep 11 at 18:52

$x, y$ and $z$ are roots of the polynomial $x^3-x+1=0$

In the usual notation ,we have $$\Sigma x^2=(\Sigma x)^2-2\Sigma xy=0-2(-1)=2$$ Then since the relation $x^3=x-1$ holds for all three roots, summing these gives:

$$\Sigma x^3=\Sigma x - \Sigma 1=0-3=-3$$

Likewise, $$\Sigma x^4=\Sigma x^2-\Sigma x=2$$ $$\Sigma x^5=\Sigma x^3-\Sigma x^2=-3-2=-5$$ $$\Sigma x^6=\Sigma x^4-\Sigma x^3=2--3=5$$ And finally, $$\Sigma x^8=\Sigma x^6-\Sigma x^5=5--5=10$$

  • $\begingroup$ hi , would you be so kind as to clarify a little what you have done? $\endgroup$
    – Nabla
    Sep 10 at 21:14
  • $\begingroup$ @ Nabla where are you stuck? $\endgroup$ Sep 10 at 23:16

This is an expansion of my comment. Recall Vieta's formulas (in this case, for cubics):

If $p, q, r$ are the roots of $t^3 + bt^2 + ct + d$, where $a \neq 0$, then \begin{align*} p + q + r &= -b \\ pq + qr + pr &= c \\ pqr &= -d. \end{align*}

This follows easily by just expanding $(t - p)(t - q)(t - r)$ and equating coefficients with $t^3 + bt^2 + ct + d$.

Using these formulas, we can see, as others have commented, that the numbers $x, y, z$ in your question are just roots of the polynomial $t^3 - t + 1$. We can also use Vieta's formulas, and a little computation along the lines of what you've attempted already, to obtain a polynomial whose roots are precisely the squares of the roots of the original polynomial.

Let's say that $t^3 + bt^2 + ct + d$ has roots $p, q, r$ as above. Let $t^3 + b't^2 + c't + d'$ have roots $p^2, q^2, r^2$. We wish to compute $b', c' d'$ vie Vieta's formulas, i.e. to compute $p^2 + q^2 + r^2$, $p^2 q^2 + q^2 r^2 + p^2 r^2$, and $p^2 q^2 r^2$ in terms of $a, b, c$.

There's a quick and easy one: $p^2 q^2 r^2 = (pqr)^2 = d^2$. Thus $d' = -d^2$.

To compute $p^2 + q^2 + r^2$, consider $$b^2 = (p + q + r)^2 = p^2 + q^2 + r^2 + 2(pq + qr + rp) = -b' + 2c,$$ thus $b' = 2c - b^2$.

Lastly, we compute \begin{align*} c^2 &= (pq + qr + pr)^2 \\ &= p^2 q^2 + q^2 r^2 + p^2r^2 + 2p^2qr + 2pq^2r + 2pqr^2 \\ &= c' + 2pqr(p + q + r) \\ &= c' + 2bd, \end{align*} so $c' = c^2 - 2bd$.

Thus, our transformation which squares roots takes a monic polynomial $t^3 + bt^2 + ct + d$ and transforms it into the monic polynomial $$t^3 + (2c - b^2)t^2 + (c^2 - 2bd)t - d^2.$$

Let's apply it to $t^3 - t + 1$ to get a polynomial whose roots are $x^2, y^2, z^2$. We obtain the polynomial: $$t^3 + (2(-1) - 0^2)t^2 + ((-1)^2 - 2(0)(1))t - 1^2 = t^3 - 2t^2 + t - 1.$$ Apply again to get a polynomial whose roots are $x^4, y^4, z^4$: $$t^3 + (2(1) - (-2)^2)t^2 + (1^2 - 2(-2)(-1))t - (-1)^2 = t^3 - 2t^2 - 3t - 1.$$ We could apply this transformation one final time to get the polynomial whose roots are $x^8, y^8, z^8$, but really we just need the coefficient of $t^2$. So, partially applying the above formula, we get: $$t^3 + (2(-3) - (-2)^2)t^2 + \underline{\hspace{12pt}} t + \underline{\hspace{12pt}} = t^3 - 10t^2 + \underline{\hspace{12pt}} t + \underline{\hspace{12pt}}.$$ So, by Vieta, $$x^8 + y^8 + z^8 = -(-10) = 10.$$


You got already that $$x^2+y^2+z^2=2.$$ Also, we have: $$\sum_{cyc}x^2y^2=\left(\sum_{cyc}xy\right)^2-2xyz(x+y+z)=1$$ and $$x^2y^2z^2=1.$$ Thus, $$\sum_{cyc}x^8=\left(\sum_{cyc}x^4\right)^2-2\sum_{cyc}x^4y^4=$$ $$=\left(\left(\sum_{cyc}x^2\right)^2-2\sum_{cyc}x^2y^2\right)^2-2\left(\left(\sum_{cyc}x^2y^2\right)^2-2x^2y^2z^2\sum_{cyc}x^2\right)=$$ $$=(4-2)^2-2(1-2\cdot2)=10.$$


Question: "@hm2020 ,hi , I can't relate the equation you are proposing to the problem, could you clarify? – Nabla"

Answer: There is an explicit formula (you find it on wikipedia under the name "the Cardano formula", or in algebra books) for the roots of $f(t):=t^3-t+1$: Using the formula it follows the following number is a root:

$$x:=\sqrt[3]{-\frac{1}{2}+\sqrt{\frac{23}{108}}}+\sqrt[3]{-\frac{1}{2}-\sqrt{\frac{23}{108}} }$$

It follows you may (by hand, in principle) use polynomial division to calculate a decomposition $f(t)=(t-x)p(t)$ where $p(t)$ is a degree 2 polynomial and there is an explicit formula for the roots of $p(t)$. Hence you may give explicit formulas for all the roots $x,y,z$ of $f(t)$ and also calculate all powers $x^m+y^n+z^k$ for all integers $m,n,k$. This is a "brute force" method but it still gives explicit formulas. You get the following:


where $p(t)=t^2+xt+(x^2-1)$. The polynomial $p(t)$ has roots

$$y:= \frac{1}{2}(-x +\sqrt{4-3x^2})\text{ and } z:=\frac{1}{2}(-x-\sqrt{4-3x^2}).$$

The above formulas give all roots $x,y,z$ of $f(t)$ and you get

$$F1\text{ }x^m+y^n+z^k:= (\sqrt[3]{-\frac{1}{2}+\sqrt{\frac{23}{108}}}+\sqrt[3]{-\frac{1}{2}-\sqrt{\frac{23}{108}} })^m+$$

$$(\frac{1}{2}(-x +\sqrt{4-3x^2}))^n+(\frac{1}{2}(-x-\sqrt{4-3x^2}))^k.$$

The formula $F1$ gives an explicit formula for all $m,n,k$, in particular for $m=n=k=8$.

You must check the "Cardano formula": It says that a solution to $g(t):=t^3+pt+q=0$ is the following:

$$x:=\sqrt[3]{-\frac{q}{2}+\sqrt{\frac{q^2}{4}+\frac{p^3}{27}}}+\sqrt[3]{-\frac{q}{2}-\sqrt{\frac{q^2}{4}+\frac{p^3}{27}} }.$$

The above formulas give all solutions to the equation $g(t)=0$. The above calculation give explicit formulas for numbers $x(p,q), y(p,q), z(p,q)$ depending on $p,q$ solving $g(t)=0$.

  • 1
    $\begingroup$ @Nabla - you should check that the formula $F1$ agrees with the other formulas given above. $\endgroup$
    – hm2020
    Sep 11 at 10:08
  • 1
    $\begingroup$ @Nabla - if the above result saying that $x^8+y^8+z^8=10$ is correct, you get the same result with $F1$ - this you must check. $\endgroup$
    – hm2020
    Sep 11 at 10:36
  • $\begingroup$ Thank you all very much, even though I have a little knowledge in higher algebra I must admit that some things at the beginning were out of my hands, after the explanations I was much clearer, I did not imagine that the system had such a variety of answers for the same reason I raised it in a general way, expecting a typical substitution here and some other algebraic manipulation there. Then I was forced to edit and ask for a simpler solution. You learn a lot in passing, much appreciated. $\endgroup$
    – Nabla
    Sep 12 at 2:22
  • $\begingroup$ @Nabla -the above method is a "brute force" method, still it gives "all solutions". $\endgroup$
    – hm2020
    Sep 12 at 11:28
  • $\begingroup$ @Nabla - You should check the "Cardano formula" in a book - wikipedia tends not to be reliable. $\endgroup$
    – hm2020
    Sep 12 at 12:33

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.