Estimates for Dirichlet Eigenvalues of the Schrodinger operator with the Kronig-Penney Model

By the definition of A_{1,n}(n^{2}), we have

Since \dfrac{1}{k(2n+k)}=\dfrac{1}{2n}\bigl(\dfrac{1}{k}-\dfrac{1} {2n+k}\bigr), we see that

Using the obvious equality C_{-k}=C_{k} and C_{0}=0, we obtain the following equality for the second term on the right side of (Proof):

Now let us estimate the first term on the right side of (Proof). Since q_{-k}=\overline{q_{k}} and \cos kx=\dfrac{1}{2}(e^{ikx}+e^{-ikx}), we have C_{k}^{2}=\dfrac{1}{4}(2\left\vert q_{k}\right\vert ^{2}+q_{k} ^{2}+q_{-k}^{2}). Therefore, the first term on the right side of (Proof) can be written in the form

where

First, let us prove that

It is clear that Q(\pi,n)=0, Q(0,n)=0, and that the derivative of Q(x,n) with respect to x is Q^{\prime}(x,n)=q(x)e^{i2nx}\,-q_{-2n}. Using the integration by parts, we see that

for n+m\neq0. Therefore, the Fourier decomposition of Q(x,n) by the orthonormal basis \{e^{i2mx}/\sqrt{\pi}:m\in\mathbb{Z\}} of L_{2}[0,\pi] has the form

Using this decomposition in the integral for D_{1}(n,q), we obtain the proof of the first equality of (Proof).

Now, let us prove the second equality of (Proof). It is clear that

G(0,n)=0 and that G^{\prime}(x,n)=q(x)e^{i2nx}\,+\dfrac{i\pi}{2} e^{ix}q_{-2n}. Therefore,

Thus, the Fourier decomposition of G(x,n) by the orthonormal basis \{e^{i(2m+1)x}/\sqrt{\pi}:m\in\mathbb{Z}\} of L_{2}[0,\pi] has the form

Using this decomposition of G(x,n) in the integral for D_{2}(n,q), we obtain the proof of the second equality of (Proof). The third equality of (Proof) follows from (Proof).

Now, we consider I_{2}. Using the definition of p(x) and the substitution t=-x, we obtain

Therefore, instead of q(x), using p(x) and repeating the proof of (Proof), we obtain that

Thus, the proof of (1) follows from (Proof)-(Proof). The lemma is proved.

To prove (2), we estimate the left and right sides of this equality separately. First, estimate the right-side. Let Q_{k}= {\displaystyle\int_{0}^{\pi}} Q(x)\sin kxdx. Then, Q_{k}=\pi C_{k}/k. Therefore, we have the decomposition

of Q(x). Using the equality (Q^{2}(x))^{\prime}=2q(x)Q(x), integration by parts, the decomposition of Q(x), and then the equality \sin kx\sin 2nx=\dfrac{1}{2}(\cos(2n-k)x-\cos(2n+k)x), we obtain

Now, using the equalities Q_{k}=\pi C_{k}/k, C_{-k}=C_{k}, and doing the substitution k\rightarrow-k in the second summation of the last equality, we obtain

Thus, the right side of (2) has the form

since C_{0}=0.

Now, we prove that the left side of (2) also can be written in this form. Doing the substitution k\rightarrow-k, and then using the equality \dfrac{1}{k(2n-k)}=\dfrac{1}{2n}\bigl(\dfrac{1}{k}+\dfrac{1}{2n-k}\bigr), we see that

On the other hand, doing the substitution k\rightarrow2n-k in the second summation of the last equality, we obtain

which means that (2) holds. The lemma is proved.

It follows from (?) and the definition of S_{2}(n,q) that

Now using (?), we obtain

Now, let us prove (3) for j=3. By (?), we have

Making the substitution s=k+l, we obtain

Now using (?), we obtain

that is, the proof of (3) for j=3. Arguing as above, we conclude that

Therefore, (?) implies that

The lemma is proved.

It follows from (?), (?) and (?) that \lambda _{n}=n^{2}+O\bigl(\dfrac{1}{n}\bigr). Using this, we easily get the estimate

By virtue of this estimate

for j=1,2,\ldots. Therefore, it follows from (?)-(?) that

where a_{1}(n^{2})=A_{1,n}(n^{2})-B_{1,n}(n^{2}), equivalently

(see Lemma 1, Lemma 2 and (?)), and

(see (?) and (?)). On the other hand, we have

Moreover, arguing as in the proof of (Proof), we see that a_{3}\left( n^{2}\right) =O\bigl(\dfrac{1}{n^{3}}\bigr). Thus, the proof of the theorem follows from (Proof)-(Proof).

(a) By Lemma 4, letting m tend to infinity in equation (?), we obtain

Now, we prove that the root of (?) lying in the interval I_{n} is an eigenvalue of the operator L. To do this, we first give the following relation:

where a_{1}(\lambda) is defined in (?), a_{1}(n^{2})=A_{1,n} (n^{2})-B_{1,n}(n^{2}), A_{1,n}(n^{2}) and B_{1,n}(n^{2}) are given in (?).

The equation

has one root in the interval I_{n}=[n^{2}-M,n^{2}+M] and

Estimating |C_{2n}-A_{1,n}(n^{2})+B_{1,n}(n^{2})| by (?), we have

for all \lambda from the boundary of I_{n}, for n=1. Now we define

Estimating the summands of |a_{1}(\lambda)-a_{1}(n^{2})| and |a_{k} (\lambda)| by considering the greatest summands of them, for n=1, we obtain

for k\geq2. Therefore, it follows by the geometric series formula that

for n=1. Hence

for all \lambda from the boundary of I_{n}, for n=1. Similar estimations can be obtained also for n\geq2. Therefore |G_{n}(\lambda)-F_{n} (\lambda)|<|F_{n}(\lambda)| holds for all \lambda from the boundary of I_{n} and by Rouche’s theorem, G_{n}(\lambda) has one root in the set I_{n}, for n\geq1. Hence, L(q) has one eigenvalue (counting multiplicity) in I_{n}, which is the root of (?). On the other hand, equation (?) has exactly one root (counting multiplicity) in I_{n}. Thus, \lambda\in I_{n} is an eigenvalue of L(q) if and only if, it is the root of (?).

First, we prove (4) by using the mean-value theorem. To do this, we estimate |g_{n}^{\prime}(\lambda)|=|\dfrac{d}{d\lambda}g_{n}(\lambda)|. By (?), we have

By estimating the summands of the term |\dfrac{d}{d\lambda}a_{r,k,n} (\lambda)|, we write

for k\geq1. Thus, by the geometric series formula, we obtain

Hence,

Since the inequality

holds for all x,y\in I_{n}, (4) holds by the mean value theorem and equation (?) has a unique solution \rho_{n} in I_{n}, by the contraction mapping theorem. Now let us prove (4). By (?), we have K_{n}(x)=x-n^{2}-g_{n}(x) and by the definition of \rho_{n}, we write K_{n}(\rho_{n})=0. Therefore by (?) and (?), we obtain

For the estimation of the right-hand side of (Proof), we have

Using the estimations for \sum\limits_{k=s+1}^{\infty}|a_{k}(\lambda_{n} )| and \sum\limits_{k=1}^{s}|a_{k}(\lambda_{n})-a_{r,k,n}(\lambda_{n})|, by considering the greatest summands of them, we obtain

Thus, by (Proof) and (Proof), we have

To apply the mean value theorem, we estimate |K_{n}^{\prime}(\lambda)|:

By the mean value formula, (Proof), (Proof), (Proof) and (Proof), we obtain

and

which completes the proof.

Without loss of generality, we can take x_{n,0}=n^{2}. By (?), (4), and (?), we have

Therefore, it is enough to estimate |x_{n,0}-\rho_{n}|. By definitions of \rho_{n} and x_{n,0}, we obtain

and by the mean value theorem, there exists x\in\lbrack n^{2}-M,n^{2}+M] such that

The last two equalities imply that

Hence by (Proof), (?), and (Proof), we have

and

The theorem is proved.