Proof of alternating zeros of Bessel function

173 Views Asked by At

We will be dealing with the Bessel functions of the first kind. Notice that to prove that their zeroes are alternating, one can just prove that for every two zeroes for $J_\nu(x)$ there exists a zero between those for $J_{\nu+1}(x)$ where $\nu \in \mathbb{R}$.

First, let $\pi_{\nu,k}$ denote the $k$:th positive root for $J_\nu$. Then, since $J_{\nu}(\pi_{\nu,k}) = J_{\nu}(\pi_{\nu,k+1}) = 0$ and the Bessel functions of first kind are continous and differentiable on the whole reals, one can apply Rolle's theorem.

Rolle's theorem then states that $\exists \eta \in (\pi_{\nu,k}, \pi_{\nu,k+1}) : J_{v}'(\eta) = 0$.

We can then utilize the recurrence relation:

$$(x^{-\nu} J_\nu(x))' = -x^{-\nu}J_{\nu+1}(x)$$

Expanding on the LHS yields us:

$$-\nu x^{-\nu-1} J_\nu(x) + x^{-\nu} J_\nu'(x) = -x^{-\nu}J_{\nu+1}(x)$$

Letting $x = \eta$, we get that the second term in the LHS vanishes. All that's left is that we would like to expand the first term in the LHS such that it involves $J_\nu'(x)$ as well as $J_{\nu+1}(x)$. In that way, we can rearrange terms such that we'll be able to prove that $x = \eta$ is indeed a zero for $J_{\nu+1}$.

I've tried making use of various recurrence relations, but I can't get anywhere from here. Does anyone have any ideas on how to proceed? I also somehow have to show that $x = \eta$ is the only root in between, since Rolle's theorem only states that there exists at least one such $\eta : J'_{v}(\eta) = 0$. I have no idea how to do this aswell.

1

There are 1 best solutions below

5
On

The reverse direction is almost exactly the same, but instead of using the relation $(x^{-\nu}J_\nu(x))' = -x^{-\nu}J_{\nu+1}(x)$, we use its sibling $(x^\nu J_\nu(x))' = x^\nu J_{\nu-1}(x)$.

Then a similar argument to your original one shows that there needs to be at least one zero of $J_\nu$ between every two zeroes of $J_{\nu+1}$. Since we know that $\pi_{\nu,k}$ and $\pi_{\nu,k+1}$ are adjacent zeroes, there cannot be two zeroes of $J_{\nu+1}$ between these, as otherwise there should also be another zero of $J_\nu$ between these, which there isn't.

More formally

The question author's reasoning, together with the two identities $(x^{-\nu}J_\nu(x))' = -x^{-\nu}J_{\nu+1}(x)$ and $(x^\nu J_\nu(x))' = x^\nu J_{\nu-1}(x)$ give us two lemmas:

lemma 1 If $a\in\mathbb{R}_{>0}$, $b\in\mathbb{R}_{>0}$, $a < b$ and $J_\nu(a) = J_\nu(b) = 0$, then there exists $\eta\in\mathbb{R}_{>0}$ such that $a < \eta < b$ and $J_{\nu+1}(\eta) = 0$

lemma 2 If $a\in\mathbb{R}_{>0}$, $b\in\mathbb{R}_{>0}$, $a < b$ and $J_\nu(a) = J_\nu(b) = 0$, then there exists $\eta\in\mathbb{R}_{>0}$ such that $a < \eta < b$ and $J_{\nu-1}(\eta) = 0$

Now, we label the positive zeroes of $J_\nu$ as $\pi_{\nu,k}$, such that if $a>0$ and $J_\nu(a) = 0$ there exists an $l\in\mathbb{N}_+$ such that $a = \pi_{\nu,l}$. Furthermore, for $k\in\mathbb{N}_+$, $l\in\mathbb{N}_+$ we label such that $k<l$ if and only if $\pi_{\nu,k} <\pi_{\nu,l}$.

To prove there is exactly one zero of $J_{\nu+1}$ between two consecutive zeroes of $J_{\nu}$, assume we have $k\in\mathbb{N}_+$ and $\eta_1\in\mathbb{R}_{>0}$, $\eta_2\in\mathbb{R}_{>0}$ such that $\pi_{\nu,k}<\eta_1<\eta_2<\pi_{\nu,k+1}$ and $J_{\nu+1}(\eta_1) = J_{\nu+1}(\eta_2) = 0$. Then by lemma 2 we have a $\sigma\in\mathbb{R}_{>0}$ such that $J_{\nu}(\sigma) = 0$ and $\eta_1 < \sigma < \eta_2$.

However, then by definition of the sequence $\pi_{\nu,k}$ there should be an $n\in\mathbb{N}_+$ such that $\sigma = \pi_{\nu,n}$. Now, $\pi_{\nu,k} < \eta_1 < \sigma = \pi_{\nu,n}$ implies $k < n$. and $\sigma=\pi_{\nu,n} < \eta_2 < \pi_{\nu,k+1}$ implies $n < k+1$. But since $k\in\mathbb{N}_+$, this implies $n\notin\mathbb{N}_+$, giving a contradiction.

Thus, by contradiction, there is at most one zero of $J_{\nu+1}$ between two consecutive zeroes of $J_{\nu}$, and since lemma 1 shows there is at least one, there must be exactly one.

Proofsketch of lemmas 1 and 2

The easiest approach to showing lemmas 1 and 2 is to use note that because the bessel functions are continuous and differentiable, so is $x^{\nu}J_{\nu}(x)$, and to note that this is zero when $J_{\nu}(x)$ is zero. This implies that if a and b are zeroes of $J_{\nu}(x)$, we can use Rolle's theorem to find a zero of the entire combination $(x^{\nu}J_{\nu}(x))'$, which, through the above identities and the fact that a and b are both positive, immediately gives a zero of $J_{\nu-1}$.