2
$\begingroup$

Consider the linear constant coefficient differential operator $P$ on the Hilbert space $L^2([0,1]^2;\mathbb C^2)$ $$P= \begin{pmatrix} D_{z}+c & a \\ b & D_{z}+c \end{pmatrix}$$ where $D_z=-i \partial_z =- i(\partial_{x_1} -i \partial_{x_2}).$

Here, $a,b,c$ are just some complex numbers.

I wonder whether one can explicitly compute the spectrum of the self-adjoint operator $P^*P$, with periodic boundary conditions, then?

$\endgroup$
6
  • 2
    $\begingroup$ What do you mean by $\partial_{x_1}$ and $\partial_{x_2}$ on a space whose functions depend only on one real variable? $\endgroup$ – Jochen Glueck Jan 7 at 23:12
  • $\begingroup$ @JochenGlueck sorry, should have been $[0,1]^2.$ $\endgroup$ – Sascha Jan 7 at 23:18
  • 3
    $\begingroup$ You'd need to give some boundary conditions for this to be a well-defined question. $\endgroup$ – Michael Engelhardt Jan 8 at 0:46
  • $\begingroup$ @sorry, yes, periodic ones. $\endgroup$ – Sascha Jan 8 at 0:47
  • $\begingroup$ @MichaelEngelhardt thank you for your efforts $\endgroup$ – Sascha Jan 8 at 1:50
6
$\begingroup$

By noting that $-i\partial_{x_1} $ is diagonalized by $e^{ik_1 x_1} $ and $-i\partial_{x_2} $ by $e^{ik_2 x_2} $, the problem reduces to a $2\times 2$ diagonalization for each $(k_1,k_2)$-block. The resulting eigenvalues are (denoting $k=k_1-ik_2 $) $$ \frac{1}{2} (|a|^2 + |b|^2 ) +|k+c|^2 \pm\frac{1}{2} \sqrt{(|a|^2 -|b|^2 )^2 +4(|a|^2 + |b|^2 )|k+c|^2 +8\,\mbox{Re}\, (a^*b^*(k+c)^2)} $$ where $k_1 , k_2 \in 2\pi \mathbb{Z} $ in view of the boundary conditions.

$\endgroup$
3
  • $\begingroup$ But how do you know this is all of the spectrum? $\endgroup$ – Sascha Jan 8 at 4:43
  • $\begingroup$ To explain my question. I am a bit sceptic about the conjugation by just the one-dimensional Fourier basis. Perhaps, we are losing something here? $\endgroup$ – Sascha Jan 8 at 4:50
  • $\begingroup$ I am essentially doing a 2-dimensional Fourier transform, to the basis $e^{i(k_1 x_1 + k_2 x_2 )}$. That's a unitary transformation on $L^2 $. As is the final $2\times 2$ diagonalization. $\endgroup$ – Michael Engelhardt Jan 8 at 4:53

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.