Hilbert Space multiplication Operator, shift operator

402 Views Asked by At

I have this problem and am not sure how to even approach it..

Hilbert space $l^2(\mathbb{Z})$ with orthonormal basis$ $$(e_n)$ and Hamiltonian operator $He_n=i(e_{n+1}-e_{n-1})$

a)I need to use Fourier series to find a multiplication operator which is unitary equivalent to H

b)Use a) to show that there is a no vector $h \in l^2(\mathbb{Z})$ , $h\neq 0$ such that $exp(-itH)h=h$ for all $t \in R$

1

There are 1 best solutions below

4
On BEST ANSWER

As pointed out by MaoWao, consider the following Fourier transform operator ${\cal F}: \ell^2(\mathbb{Z}) \rightarrow L^2(S^1)$

$$ {\cal F}[f(n)](k) = \hat{f}(k) = \frac{1}{\sqrt{2 \pi}} \sum_n e^{i kn} f(n) $$

with inverse $$ {\cal F}^{-1}[g(k)](n) = \frac{1}{\sqrt{2 \pi}} \int_0^{2 \pi} dk e^{-i kn} g(k) $$

For point a)

First compute the Fourier transform of the standard basis. Note that $e_n(m) = \delta_{n,m}$ (Kronecker delta). So

\begin{align} \hat{e_n}(k) &= \frac{1}{\sqrt{2 \pi}} \sum_m e^{i km} \delta_{n,m}\\ & = \frac{e^{ikn}}{\sqrt{2\pi}} \end{align}

Now use

$$ {\cal F} H{\cal F}^{-1} {\cal F} e_n = i ({\cal F}e_{n+1} - {\cal F}e_{n-1}) $$

which means \begin{align} {\cal F} H{\cal F}^{-1} e^{ikn} &= i (e^{ik(n+1)}- e^{ik(n-1)} ) \\ &= -2\sin(k) e^{ikn} \end{align}

Hence $H$ is unitarily equivalent to multiplication by $-2\sin(k)$. Note that the sign is conventional: a different choice of signs in the Fourier operator would have led to $2\sin(k)$.

For point b)

Now Fourier transform the equality $ \exp{itH}h = h$ to get

$$ \exp{ ( it {\cal F} H{\cal F}^{-1} )} \hat{h} = \hat{h} \tag{1} $$

Hence you need to find a non-zero function $\hat{h}(k)$ in $L^2(S^1)$ such that

$$ \exp{[-2it \sin(k)]} \, \hat{h}(k) = \hat{h}(k) $$

for all $k\in [0,2\pi)$ (and all t) which is impossible.

Note added:

In fact $\sin(k)$ is non-zero for all $k$ (in the interval) except $k=0,\pi$. Hence, in order to satisfy (1), the function $\hat{h}(k)$ should be zero for all $k$ except possibly $k=0,\pi$. This is clearly the zero function almost everywhere. To add some intuition you can reason as follows. If such a function existed, it would try to be "infinite" at $k=0$ (or $k=\pi$ or both), in order to have a finite (non-zero) norm. This is achieved by a (Dirac) delta function --which of course is not a function--. So, "in a sense", $\hat{h}(k) = \delta(k)$, is a solution of $\exp{itH}h=h$. If you Fourier transform back you get the function (vector) with components

$$ h(n) = \frac{1}{\sqrt{2 \pi}} $$

You can plug this in your equation defining $H$ and you will notice that it leads to $H h = 0$. So it would seem that $h$ is an eigenvector of $H$ with eigenvalue zero. However the vector with components $h(n) \propto 1$ is not normalizable, hence such $h$ does not belong to $\ell^2(\mathbb{Z})$.