7

Stewart wants me to prove stuff but I have no idea how to.

a) Show that a polynomial of degree 3 has at most three real roots.

b) Show that a polynomial of degree n has at most n real roots.

  • $n$-odd? ${}{}$ –  Apr 06 '12 at 23:21
  • This is the same idea as your other question. The Mean Value Theorem implies that between any two roots of a polynomial, there has to be a root of the derivative of the polynomial (between any two $0$s, there has to be a critical point). – Arturo Magidin Apr 07 '12 at 01:49
  • @Arturo I am confused, I thought it wasn't specfically roots unless it is Rolle's Theorem. –  Apr 07 '12 at 01:56
  • 1
    It's the same idea: between any two roots of the function, there has to be a critical point/root of the polynomial. Both the previous problem and this one revolve around the same idea: if you know how many roots the derivative can have, then you know how many roots the function can have. For example: since a degree 1 polynomial can have at most one root, then a degree 2 polynomial (whose derivative is degree 1) can have at most two roots (between any two roots of the degree 2 polynomial, there has to be one of the derivative). – Arturo Magidin Apr 07 '12 at 01:58
  • I understand that the derivative determines if the function can change direction so does that mean that the zeros can never exceed the degree? –  Apr 07 '12 at 02:04
  • Actually, for any field $\mathbb{F}$ - a polynomial of degree $n$ over $\mathbb{F}$ has at most $n$ roots – Belgi Aug 09 '12 at 10:25

3 Answers3

10

I'll do some lower dimensional cases. You can build them up.

A degree-one polynomial has at most 1 real root. This, I take for granted.

Now, we would like to say that a quadratic polynomial as at most 2 real roots. Suppose it has three. Then by the mean value theorem (in fact, just Rolle's Theorem), the derivative of our quadratic has at least 2 real roots. But this is a contradiction, as above I mentioned that a linear polynomial has at most 1 real root. So a quadratic has at most 2 real roots.

Do you see how this continues?

  • I don't follow. –  Apr 06 '12 at 23:35
  • I know Rolle's Theorem, being applied here it just means that there is a derivative that is equal to zero. I don't understand how that means there can only be 3. –  Apr 06 '12 at 23:39
  • @Jordan: Ah, so we know the derivative is equal to zero between the first and second root, and between the second and third root. Thus the derivative is equal to zero twice. But the derivative is a linear, one degree polynomial. And those can't have 2 roots. – davidlowryduda Apr 06 '12 at 23:41
  • How do we know there are 2 roots and not just one? –  Apr 06 '12 at 23:42
  • @Jordan: suppose the 3 assumed roots were at $a, b, c$ with $a < b < c$. Then Rolle guarantees a root between $a$ and $b$, and another between $b$ and $c$. – davidlowryduda Apr 06 '12 at 23:44
  • But only if a b and c are equal to eachother. –  Apr 06 '12 at 23:45
  • My books says that f(a)=f(b) –  Apr 06 '12 at 23:53
  • @Jordan: That's good. Since $a,b,c$ were assumed to be roots, we have that $f(a) = f(b) = f(c) = 0$. – davidlowryduda Apr 06 '12 at 23:55
  • I don't understand how we know that they will be equal to eachother. –  Apr 06 '12 at 23:57
  • @Jordan: What does it mean for $y$ to be a root of a polynomial $p(x)$? – davidlowryduda Apr 06 '12 at 23:59
  • I am not sure but I think it means that it intercepts the y axis. –  Apr 07 '12 at 00:01
  • @Jordan: No, that's not what it means. It would mean that $p(y) = 0$. For example, the polynomial $x-5$ has a root at $5$. The polynomial $(x-1)(x+1)$ has roots at $1$ and $-1$. – davidlowryduda Apr 07 '12 at 00:03
  • Okay so it means that y will make the function intercept the y axis? –  Apr 07 '12 at 00:04
  • @Jordan: no - it means that $y$ would make the polynomial intersect the x axis. We know the exact value when the polynomial's graph intersects the y axis - it happens when $x = 0$. – davidlowryduda Apr 07 '12 at 00:06
  • So a root is an x intercept, so if f is continuous and on an open interval a,b and if f(a)=f(b) thne there is a number where the derivative = 0. So this means on a closed interval if the function is continuous and that if the value of the function evaluated at the end points are equal that there is a derivative that equals 0? –  Apr 07 '12 at 00:09
  • @Jordan: Yes - that is the statement of Rolle's Theorem. – davidlowryduda Apr 07 '12 at 00:12
  • So how does this help find roots? –  Apr 07 '12 at 00:13
  • @Jordan: If the derivative at that point is 0, then the derivative has a root at that point. – davidlowryduda Apr 07 '12 at 01:00
  • I don't understand, what is the point of finding the roots of a derivative? Those are critical numbers. –  Apr 07 '12 at 01:02
  • @Alex Becker: I do not understand the downvote either. Perfectly valid approach. – André Nicolas Apr 07 '12 at 01:03
  • @Andre: That was on the deleted answer below. – davidlowryduda Apr 07 '12 at 01:03
  • @Jordan: The idea is that the derivative has one degree less, and we understand that degree. So we get a contradiction. Reread my answer, write some things down, draw a picture, and then ask any questions you have. – davidlowryduda Apr 07 '12 at 01:04
  • @mixed I do not see how this ties into The Mean Value Theorem, is there something else I need to apply here? I just do not see how it follows. –  Apr 07 '12 at 01:17
  • @Jordan: This uses exactly one fact: Rolle's Theorem (which is the Mean Value Theorem). We say our quadratic has 3 roots. Then Mean Value Theorem says the derivative (which is linear) has 2 roots. But a linear polynomial doesn't have 2 roots. Thus our quadratic does not have 3 roots. So we use only one theorem, and that's the Mean Value Theorem. – davidlowryduda Apr 07 '12 at 01:20
  • @mixed How do we know how many roots the polynomial has? –  Apr 07 '12 at 01:21
  • @Jordan: we don't know. We are trying to show it can't have 3 roots. So that analysis says that if there are 3 roots, then there is a contradiction. So there can't be 3 roots. Maybe there are 0, 1, or 2, but now we know there can't be 3. – davidlowryduda Apr 07 '12 at 01:23
  • @mixed I don't quite understand, we can know that the derivative has roots on an interval but how does that translate to roots on the function? Is this because of critical numbers? I don't quite understand why the function can't have two roots as well as the derivative. –  Apr 07 '12 at 01:24
4

According to Rolles theorem and the polynomial theorem any third degree polynomial sports at most 1 to 3 roots. Therefore, any polynomial of degree n possesses at least one root, whether real or complex. Moreover, a polynomial with a degree n bears at most n roots. To back this argument with Rolles theorem assume that there are three roots such that $a>b>c$. According to Rolles theorem there must be a number $m$ such that $f'(m) = 0$ between $a$ and $b$. Likewise there must be a value $n$ such that $f'(n) = 0$ between $b$ and $c$. This implies that $m$ and $n$ are minimums or maximums. Since there are two solutions for $f'(x)$ because a third degree polynomial's derivative is a degree two function. Therefore the third degree polynomial acquires at most three real roots. If the function is twice differentiated there must be such a "$c$" between $m$ and $n$ such that $f"(c) = 0$ thus revealing that indeed that a third degree function bears at least one root. Moreover if you took a quadratic function and analyze between $x$- intercepts there is either a max or min value in between the two thus asserting the fact that when the derivative equals to zero there is indeed two zeros per critical point. As a result, there must be two c.p s to have three real roots.

3

Part (a)

Let $f(x)=K_3x^3+K_2x^2+K_1x+K_0$ Since $f(x)$ is a polynomial, it is continuous and differentiable everywhere. Suppose $f(x)$ has four roots $a_1, a_2, a_3$ such that $f(a_1)=f(a_2)=f(a_3)=f(a_4)=0$. By Rolle's Theroem there exists; $b_1$ in $(a_1, a_2)$, $b_2$ in $(a_2, a_3)$, $b_3$ in $(a_3, a_4)$ such that $f'(b_1)=f'(b_2)=f'(b_3)=0$. So, $f'(x)=3K_3x^2+2K_2x+K_1$ is also a polynomial and thus continuous and differentiable everywhere, by Rolle's Theorem again there exists; $c_1$ in $(b_1, b_2)$ and $c_2$ in $(b_2, b_3)$ such that $f''(c_1)=f''(c_2)=0$. So, $f''(x)=6K_3x+K_2$ which is still continuous and differentiable everywhere, by Rolle's Theorem there should exist; $d_1$ in $(c_1, c_2)$ such that $f'''(x)=0$ however $f'''(x)=6K_3$ which is not zero. So $f'''(x)$ never equals zero and thus; $f''(x)$ can only have at most one root, $f'(x)$ can only have at most two roots, and $f(x)$ can only have at most three roots.

Part (b)

First, all polynomials are continuous everywhere and differentiable everywhere.

Let $f(x)$ be a polynomial of degree $n$, $f(x)=K_nx^n+K_{n-1}x^{n-1}+...+K_2x^2+K_1x+K_0$. Suppose that $f(x)$ has $n+1$ roots, $a_{n+1}, a_n, a_{n-1}, ... a_2, a_1$, such that $f(a_n)=f(a_{n-1})=...=f(a_2)=f(a_1)=0$. Then by Rolle's Theorem there exists some $b_n$ in $(a_{n+1}, a_n)$, $b_{n-1}$ in $(a_n, a_{n-1})$,..., $b_2$ in $(a_3, a_2)$, $b_1$ in $(a_2, a_1)$, such that $f'(b_n)=f'(b_{n-1})=...=f'(b_2)=f'(b_1)=0$. By applying Rolle's Theorem again there exists some $c_{n-1}$ in $(b_n, b_{n-1})$, $c_{n-2}$ in $(b_{n-1}, b_{n-2})$,..., $c_2$ in $(b_3, b_2)$, $c_1$ in $(b_2, b_1)$ such that $f''(c_{n-1})=f''(c_{n-2})=...=f''(c_2)=f''(c_1)=0$. If we keep applying Rolle's Theorem we find that eventually there exists two roots $z_2$ in $(y_3, y_2)$ and $z_1$ in $(y_2, y_1)$ such that $f^n(a_2)=f^n(a_1)=0$. Where $y_3, y_2$ and $y_1$ are the roots of $f^{n-1}(x)$. But $f^n(x)=n!K_n$ and so $f^n(x)>0$ for all $x$ and so the $(n-1)$th derivative of $f(x)$ can only have one root. Therefore the $(n-2)$th derivative can have at most 2 roots, the $(n-3)$th derivative can have at most 3 roots,..., the 3rd derivative can have at most $(n-3)$ roots, the 2nd derivative can have at most $(n-2)$ roots, the 1st derivative can have at most $(n-1)$ roots and hence $f(x)$, a polynomial of degree $n$, can have at most $n$ roots.

John Habert
  • 4,001
Simon
  • 31
  • 3