My intro to analysis book calls the following theorem The Zeros localization theorem:

Theorem: Let $p(x)=x^n + a_{n-1}x^{n-1} +\dots+a_1 +a_0,\ x \in \mathbb R$ , be a polynomial. Then all the zeros of $p$ (if there are any ) lie in the open interval $(-M,M),$ where $M= 1 + max\{ |a_{n-1}| ,\dots,|a_1|,|a_0| \}.$

Here is the problem which I have a solution of but which I DO NOT know if I am correct.

Problem: Using the Zeros Localization theorem and the Extreme Values theorem, prove that every polynomial of EVEN degree $n$ given by $$p(x)= a_nx^n +a_{n-1}x^{n-1}+\dots +a_1x +a_0,\ x \in \mathbb R,$$ where $a_n$ does not equal zero, has a minimum value on $\mathbb R$.

MY PROOF: Note that $$\left(\frac{1}{a_n}\right)p(x)=x^n+\frac{a_{n-1}}{a_n}x^{n-1} +\dots +\frac{a_1}{a_n}x+ \frac{a_0}{a_n}.$$ Hence by the Zeros localization theorem all zeros of $p(x)$ lie in the open interval $(-M,M)$ where $M= 1+ max \{ |\frac{a_{n-1}}{a_n}|,\dots,|\frac{a_1}{a_n}|,|\frac{a_0}{a_n}|\}$.

Also $\frac{1}{a_n}p(x)$ is continuous on $[-M,M]$, and hence by the Extreme Value theorem has a minimum on $[-M,M]$, that is, there exists a point $c \in [-M,M]$ such that $\frac{1}{a_n}p(c) \leq\frac{1}{a_n}p(x)$ for $x \in [-M,M]$. Since $p(x)$ is of even degree, limit of $p(x)$ as $x$ goes to plus or minus infinity equals plus infinity and hence $\frac{1}{a_n} p(c)$ is a minimum of $\frac{1}{a_n} p(x)$ on $\mathbb R ,$ and so $p(c)$ is a minimum value of $p(x)$ on $\mathbb R .$

Is my proof completely correct?


The property, as stated in the Problem, is false.

Here is a counterexample: $p(x)=-x^2$ is a polynomial of even degree ($a_2=-1 \neq 0$) but $p(x)$ does not have a minimum in $\mathbb{R}$.

Your proof must therefore be incorrect. I have detected at least two flaws:

  1. The existence of a minimum point $c$ for $\frac{1}{a_n}p(x)$ in $[-M,M]$ does not immediately imply that $\frac{1}{a_n}p(c)$ is a minimum of $\frac{1}{a_n}p(x)$ in $\mathbb{R}$.

  2. For $a_n < 0$ (as in the counterexample), if $\frac{1}{a_n}p(x)$ has a (local) minimum at $c$ then $p(x)$ has a (local) maximum at $c$.

  • $\begingroup$ Hi ,do you mean that the problem itself is incorrect ?? $\endgroup$ – herashefat May 2 '15 at 14:57
  • $\begingroup$ Yes. However, if you replace where $a_n$ does not equal zero by $a_n >0$ it becomes true. $\endgroup$ – P. Gomes May 2 '15 at 16:22
  • $\begingroup$ Thanks for your response , then how would a correct proof look like?? $\endgroup$ – herashefat May 2 '15 at 16:36

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.