1

I tried to use the Sturm Theorem and computed the Sturm sequence as the following: $$ P_0 = x^4-4x^2-8x+12 $$ $$ P_1 = 4x^3-8x-8 $$ $$ P_2 = 3x^3+2x^2-18 $$ $$ P_3 = 8x^2/3+8x-16 $$ $$ P_4 = 60-39x $$ $$ P_5 = something\ positive $$ where the sequence gives $V(-2)=2$ as $(+,-,-,-,+,+)$ and $V(2)=3$ as $(-,+,+,+,-,+)$, which gives $V(-2)-V(2)=-1$. It took me so much time evaluating the sequence, and it didn't even give me the correct answer somehow. I did redo the calculation, and it seemed to be correct. If someone can find the error in my calculation, please do. Otherwise, I am looking for another easier and simpler method to solve this type of question.

Ivan Neretin
  • 12,835
OriginK
  • 113
  • Alternative approach (assumes that you have a working knowledge of Calculus) : for this problem, abandon the Sturm Theorem in favor of exploration of the first and second derivatives. The equation $~f''(x) = 0,~$ is a quadratic equation. Therefore, you can routinely identify exactly where $~f''(x)~$ is positive, negative, or zero. Using this information, and using a manual computation of (for example) $~f'(-2),~f'(0),~$ and $~f'(2),~$ you can get an adequate picture of the behavior of $~f'(x),~$ in the interval $~[-2,2].~$ ...see next comment – user2661923 Sep 19 '23 at 13:47
  • Then, once you have dissected the behavior of $~f'(x)~$ in the interval $~[-2,2],~$ you can use this information, coupled with a manual computation of (for example) $~f(-2), ~f(0), ~$ and $~f(2),~$ to answer the basic question of : how many times does the function $~f(x)~$ cross the $~x$-axis in the interval $~[-2,2]~?$ – user2661923 Sep 19 '23 at 13:49
  • Re previous comments, I suggest that you take them with a grain of salt. Personally, I have never heard of the Sturm Theorem. However, I surmise, based on your posting, that the problem composer intend that you use this Theorem to answer the posted question, so that you could learn to apply the theory that you were just exposed to. The approach in my previous comments, while perhaps easier for you in this instance, constitutes re-inventing the wheel, which is not generally a good idea. ... see next comment – user2661923 Sep 19 '23 at 13:53
  • So a reasonable compromise is to use both methods, with the wheel re-invention approach leading to an answer that you are confident in. This allows you to sanity check your application of the Sturm Theorem, whatever that is. – user2661923 Sep 19 '23 at 13:54
  • I don't trust your $P_2$ -- it should have degree at most 2, not a cubic. – user10354138 Sep 19 '23 at 13:56

3 Answers3

0

I don't know the method you mentioned, but I solve these problems by (1) drawing a rough graph of the polynomial, especially its turning points, and (2) using intermediate value theorem.

First, we have $f(x) = x^4 - 4x^2 - 8x + 12$, so $f'(x) = 4x^3 - 8x - 8 = 4(x^3 - 2x - 2)$. Note that $f'(1) = -12 < 0$ and $f'(2) = 8 > 0$, so there is a turning point in $[1, 2]$. Next, I try to determine which of the two cases $f'(x)$ is:

enter image description here

To do so, I find the local maxima of $f'(x)$ by computing $f''(x) = 4(3x^2 - 2)$ and evaluating $f'\left(-\sqrt{\frac{2}{3}}\right) \approx f'(-0.8) \approx 4 \cdot (-0.512 + 1.6 - 2) < 0$, so it is the case on the left. This means that $f(x)$ only has a single turning point somewhere in $[1, 2]$, and we can draw this graph:

enter image description here

What this means is that we can split the real line into $(-\infty, 1.\ldots)$ and $(1.\ldots, \infty)$, and $f$ is decreasing in the first and increasing in the second. We can check whether there are any real roots in the two intervals using the intermediate value theorem!

Note that:

  • $f(-2) = 16 - 16 + 16 + 12 = 28 > 0$
  • $f(1) = 1 - 4 - 8 + 12 = 1 > 0$
  • $f(2) = 16 - 16 - 16 + 12 = -4 < 0$

The latter two equations show that $f(1.\ldots) < 0$, and also that the root on $(1.\ldots, \infty)$ is outside $[-2, 2]$. On the other hand, $f(-2)f(1.\ldots) < 0$ implies that there is a single real root in $[-2, 1.\ldots)$.

In the end, there are $\fbox{1}$ roots between $[-2, 2]$.

Gareth Ma
  • 3,725
0

With Sturm's theorem: Note that we could, at any stage, replace $P_i$ with $\tilde{P_i}=\lambda P_i$ where $\lambda>0$. But for the sake of conforming, \begin{align*} P_0(x)&=x^4-4x^2-8x+12\\ P_1(x)&=4x^3-8x-8\\ P_2(x)&=-\operatorname{rem}(P_0,P_1)=2x^2+6x-12\\ P_3(x)&=-\operatorname{rem}(P_1,P_2)=-52x+80\\ P_4(x)&=-\operatorname{rem}(P_2,P_3)=-P_2(\frac{80}{52})<0 \end{align*} So $V(-2)=3$ and $V(+2)=2$ (the sequences are $(+,-,-,+,-)$ and $(-,+,+,-,-)$), giving one root in $[-2,2]$.

Alternative method: if you know the discriminant of a polynomial, then a simple calculation shows $$ \operatorname{disc}P=\operatorname{Resultant}(P,P')=-339968 $$ where the first equality is from $P$ being a monic quartic and the second is explicit calculation of the determinant. So $P$ has two real roots and two nonreal roots. Hence the change of sign $P(2)<0<P(-2)$ gives the one real root in $[-2,2]$.

user10354138
  • 33,239
0

Another way:

The equation $x^4-4x^2-8x+12=0$ can be written as $$(x^2-2)^2=8(x-1).$$ From the table below and the intermediate-value theorem, we deduce that there are two roots of the equation, one in the interval $(1,2)$, the other in the interval $(2,3)$. $$ \begin{array}{c|c|c} x & (x^2-2)^2&8(x-1) \\ \hline 1 & 1&0\\ \hline 2 & 4&8 \\ \hline 3&49&16\\ \hline \end{array} $$ To show that there are exactly two real roots we need Descartes'rule of signs and graphical support: Signs of the non-zero term coefficients in order are $+1,-4,-8,+12$. There are two sign changes: One from $+1$ to $-4$, the other from $-8$ to $+12$. Therefore there are at most two positive roots. From graph, we see that there are no negative roots. Therefore our equation has two positive real roots and two complex conjugate roots.

Bob Dobbs
  • 10,988