1

Find the polynomial $p(x)=x^2+px+q$ for which $\max\{\:|p(x)|\::\:x\in[-1,1]\:\}$ is minimal.

This is the 2nd exercise from a test I gave, and I didn't know how to resolve it. Any good explanations will be appreciated. Thanks!

Florin M.
  • 635
  • 4
    You gave an exam question that you didn't know the answer to? That seems a bit unfair. – Jemmy May 29 '13 at 07:08
  • 2
    Why did you edit back to the notation where $px$ is a term in the polynomial $p(x)$? It is an improvement to either choose a different name for the polymonial, or for the coefficients. – Jonas Meyer May 29 '13 at 07:20
  • Because this is the problem's statement I received at the test – Florin M. May 29 '13 at 07:22
  • 1
    @MFGFlay: The exact notation that appeared on the test is irrelevant here. A clearer presentation would be helpful. – Jonas Meyer May 29 '13 at 07:23
  • Dear Jonas Meyer, I understand that is a bit strange but I can assure this is the exactly statement of the problem. – Florin M. May 29 '13 at 07:23
  • 3
    @MFGFlay: It is not important what notation was used on the test. It is a mathematical problem, which could easily be presented more clearly. "Solve the equation $a^2 -a -5=0$" is the same problem as "Solve the equation $q^2-q-5=0$". – Jonas Meyer May 29 '13 at 07:26
  • Well, I read again the statement from my test paper, and I edit-it. The only difference was at the beginning of the phrase: Find the ... , instead of Find all ... . I can assure you that this is the statement – Florin M. May 29 '13 at 07:28
  • Aside from the poor notation, I have to agree with Adriano. – Jemmy May 29 '13 at 07:34
  • @Adriano Note, that a polynomial of the form $x^2 + px +q$ cannot be zero. – martini May 29 '13 at 07:37
  • 2
    @MFGFlay: What you don't realize is that no one else cares how the problem was written on your test. – dfeuer May 29 '13 at 07:37
  • Ok then, sorry I started this topic. I was curious on how this problem could be solve, I appreciate your responses! – Florin M. May 29 '13 at 07:38
  • 1
    MFGFlay, the problem is that you edited away someone else's improvement to the problem statement because you were under the false impression that the exact original wording mattered. – dfeuer May 29 '13 at 07:53

5 Answers5

2

Here's an informal argument that doesn't use calculus. Notice that $p(x)$ is congruent to $y = x^2$ (for example, simply complete the square). Now suppose that we chose our values for the coefficients $p,q$ carefully, and it resulted in producing the minimal value of $m$. Hence, we can think of the problem instead like this:

By changing the vertex of $y=x^2$, what is the minimal value of $m$ such that for all $x\in [-1,1], -m \le p(x) \le m$?

By symmetry, there are only two cases to consider (based on the location of the vertex).

Case 1: Suppose the vertex is at $(1,-m)$ and that the parabola extends to the top left and passes through the point $(-1,m)$. Using vertex form, we have $p(x)=(x-1)^2-m$ and plugging in the second point yields $m=(-1-1)^2-m \iff 2m=4 \iff m = 2$.

Case 2: Suppose that the vertex is at $(0, -m)$ and that the parabola extends to the top left and passes through the point $(-1,m)$. Using vertex form, we have $p(x)=x^2-m$ and plugging in the second point yields $m=(-1)^2-m \iff 2m = 1 \iff m = 1/2$.

Since $m=1/2$ is smaller, we conclude that $\boxed{p(x)=x^2-\dfrac{1}{2}}$ is the desired polynomial.

Adriano
  • 41,576
  • I don't think it's informal. It's something that a grade 10 student can work through. However, you might want to deal with your cases better. I believe that they should be Case 1: Vertex outside of $[-1,1]$, then the min value of $m$ is 2; Case 2: Vertex lies within $[-1,1]$, then the min value of $m$ is $\frac{1}{2}$.
  • – Calvin Lin May 29 '13 at 13:34
  • @CalvinLin Hahah, yeah that's what makes this informal. I don't really justify why it suffices to consider only these two cases. I figured this out by staring at the rectangle defined by ${(x,y) \mid -1 \le x \le 1, -m \le y \le m}$ on a piece of paper and sketching a bunch of parabolas to find the one that seemed to fit the best. Not the most rigorous proof. – Adriano May 29 '13 at 17:53
  • This is what I did. complete the square to $(x-h)^2 + k$, so the vertex is $(h,k)$. Then, if $h \geq 1$, we get that $f(-1) - f(1) = (-1-h)^2 - (1-h)^2 = 4h \geq 4$. This implies that $\max( |f(-1)|, |f(1)| ) \geq \frac{4}{2} = 2 $. If $0\leq h \leq 1 $, then $f(-1) - f(0) = (-1-h)^2 - h^2 = 1+2h \geq 1 $. This implies that $\max( |f(-1)|, |f(0)|) \geq \frac{1}{2} $. The other 'cases' are similar. Hence we are done. – Calvin Lin May 29 '13 at 17:57