Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Consider the following polynomial in two variables :

$$ Q(k,x)=27x^6 - 144kx^4 + 80k^2x^3 + 240k^2x^2 - 192k^3x + (64k^4 - 128k^3) $$

Then for any integer $k \geq 5$, the polynomial $Q(k,.)$ (in one variable $x$) seems to be always positive (i.e, $Q(k,x) >0$ for any real number $x$). Prove or disprove.

share|cite|improve this question
The usual way to show a polynomial is positive is to express it as a sum of squares. Have you tried that? – Gerry Myerson Jan 28 '12 at 22:46
I tried and failed. – Ewan Delanoy Apr 29 '12 at 8:54
up vote 9 down vote accepted
  1. $x\mapsto Q(5, x)$ is positive. There is numerous ways to prove that, more or less algorithmically.

  2. Assume that there exists a $k$ such that $x\mapsto Q(k,x)$ is not positive, and consider $L$ be the infimum of all real numbers $k \geqslant 5$ such that $x\mapsto Q(k, x)$ is not positive. I claim that $Q(L,x)$ has a double real root, this is equivalent to say that $L$ is a root of the discriminant of $Q$ w.r.t. the variable $x$.

  3. The discriminant of $Q$ with respect to $x$ is (computed with Maple) : $$ \operatorname{disc}_x Q = 92162779488452608 k^{16} (k-4)^4 $$

  4. The real roots of this discriminant are 0 and 4, thus $L$ cannot be finite, and hence the claim.


  • Point 1 can be proved using Sturm sequences, and you can compute these sequences using Maple :

    sturm(subs(k=5, Q),x,-infinity, infinity);

This means that $Q(5, x)$ has no real roots. Since the leading coefficient is positive, the polynomial $Q(5, x)$ is positive whenever $x\in \mathbb R$.

  • To prove point 2 consider $x_0$ a real numbers such that $Q(L, x)$ is minimal. We have $Q(L, x_0) = 0$, because if $Q(L, x_0) > 0$ then $x\mapsto Q(L+\epsilon, x)$ is positive for $\epsilon > 0$ small enough, which contradict the definition of $L$. If $Q(L, x_0) < 0$, then $Q(L-\epsilon, x_0)$ is stille negative for $\epsilon > 0$ small enough, which is again a contradiction. So $x_0$ is a root of $Q(L,x)$ and of $\partial_x Q(L,x)$, thus it is a double root. The discriminant thing is usual.

  • The point 3 is Maple :

    factor(discrim(Q, x));
share|cite|improve this answer
@Lierre : perfect ! after seeing your solution I'm surprised and ashamed I didn't think of it earlier. – Ewan Delanoy Apr 29 '12 at 11:18

You can also use quantifier elimination algorithms. This gives you precisely for which $k$ the polynomial $Q(k,x)$ is positive. For example Mathematica implements such an algorithm :

In[1] := Resolve[ForAll[{x}, Q > 0]]
Out[1] := k < 0 || k > 4

I insist that you can trust this result, it is an exact algorithm not a numerical heuristic.

share|cite|improve this answer

You can try to prove your assumption using induction :

$1.$ Show that :

$Q(5,x)=27\cdot x^6-720 \cdot x^4+2000 \cdot x^3+6000\cdot x^2-24000\cdot x+24000 > 0$ ,for all $x$

$2.$ Suppose that :

$Q(k,x)=27x^6 - 144kx^4 + 80k^2x^3 + 240k^2x^2 - 192k^3x + (64k^4 - 128k^3) >0$ ,for all $x$

$3.$ Try to prove that : $Q(k+1,x) >0 $ , for all $x$ using assumption from step $2$

share|cite|improve this answer
Does it work??? – Gerry Myerson Jan 28 '12 at 22:44
@GerryMyerson,Probably...since $k$ is an integer ... – pedja Jan 29 '12 at 5:50
A main problem with the induction idea is that for "most" $x$ (i.e. large enough $x$), we do not have $Q(k+1,x)>Q(k,x)$ as excepted. It seems that you cannot solve the problem without studying the "relative position" of $k$ and $x$. – Ewan Delanoy Jan 29 '12 at 11:39

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.