Here is the question:

Suppose $P_0, P_1, P_2, \dots$ are polynomials orthonormal with respect to the inner product $$(f,g)=\int_a^b f(x)g(x)W(x)dx,$$ where $W(x) > 0$ is a weight function and $P_n$ is of degree $n$. Is it true that $P_n$ has $n$ distinct roots in $(a,b)$?

Clearly $P_0$ has no roots and since $(P_0,P_1)=0$, I know $P_1$ must cross the $x$-axis at least once otherwise the integral would not equal $0$, so $P_1$ has one root in $(a,b)$. However, I'm not sure how to prove this for an arbitrary $n$-value (if it is true). I would appreciate any advice.

  • 1
    $\begingroup$ The fact that $(P_0, P_k) = 0$ when $k\neq 0$ shows that all $P_k$ have a root in $(a,b)$, not just for $k = 1$. $\endgroup$ Oct 4 '11 at 3:04
  • 1
    $\begingroup$ That's true but how can I show that Pn has n distinct roots in (a,b)? $\endgroup$
    – Nick
    Oct 4 '11 at 3:10

This is true. The proof I know is beautiful but quite clever in my opinion.

Because $\deg P_k = k$ for all $k$, you get $\mathbb{R}_{n-1}[X] = \textrm{Span}(P_0, \ldots, P_{n-1})$. And because of orthogonality, if $\deg f < n$ then $(f, P_{n}) = 0$.

Now, let $r_1, \ldots, r_k$ be the roots of $P_n$ that are in $[a,b]$ and have odd multiplicity (if there are some). Denote $Q = (X-r_1) \ldots (X-r_n)$, and write $P_n = Q R$. All the roots of $R$ in $[a,b]$ have even multiplicity, so $R$ does not change sign in $[a,b]$. In order to conclude, we need to show $\deg Q = \deg P_n$. If we assumed $\deg Q < \deg P_n$, then we would have

$$(Q, P_n) = 0$$

So we'd get

$$\int_a^b Q^2 R w = 0$$

But the function $Q^2 R w$ does not change sign on $[a,b]$, so $Q^2 R w= 0$. This is obviously impossible, so $\deg Q = \deg P_n$, which concludes.

  • $\begingroup$ Thank you this is very helpful!! $\endgroup$
    – Nick
    Oct 4 '11 at 3:30
  • $\begingroup$ Very neat :) ${}$ $\endgroup$ Oct 4 '11 at 5:56
  • $\begingroup$ also, why do all the roots have to be distinct? $\endgroup$
    – Nick
    Oct 4 '11 at 6:21
  • $\begingroup$ @Nick : The roots of $Q$ are distinct by construction (you take each once each root that has odd multiplicity). $\endgroup$
    – Joel Cohen
    Oct 4 '11 at 7:24
  • $\begingroup$ I used a similar approach in the second part of this answer. $\endgroup$ Oct 4 '11 at 7:55

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.