3
$\begingroup$

The question and part of its answer is given as follows:

13. Let $K$ be an operator of a finite rank on a Hilbert space $H$. For $\varphi \in H$, $$ K\varphi = \sum_{j=1}^{n} \langle \varphi, \varphi_j\rangle\psi_j. $$ Suppose $\psi_j \in \operatorname{sp}\{ \varphi_1, \cdots, \varphi_n\}^{\perp}$ for $j = 1, \cdots, n$. Prove that $\mathrm{I} + \alpha K$ is invertible for any $\alpha$ and find its inverse.

Solution. Let $\alpha \in \mathbb{C}$, $K'\varphi = \sum_{j=1}^{n} \langle \varphi, \varphi_j\rangle (-\alpha \psi_j)$. By Theorem 7.1, \begin{align*} \text{$\mathrm{I} + \alpha K$ is invertible} &\quad \Leftrightarrow \quad \text{$\mathrm{I}-K'$ is invertible} \\ &\quad \Leftrightarrow \quad \det(\delta_{ij}-\langle(-\alpha\psi_j),\varphi_i\rangle)_{i,j=1}^{n}\neq 0. \end{align*} But $$\det(\delta_{ij}-\langle(-\alpha\psi_j),\varphi_i\rangle)_{i,j=1}^{n} = \det(\delta_{ij}+\alpha\langle\psi_j,\varphi_i\rangle)_{i,j=1}^{n} = 1 \neq 0$$ (because $\psi_j \in \operatorname{sp}\{ \varphi_1, \cdots, \varphi_n\}^{\perp}$ for $j = 1, \cdots, n$). Thus $\mathrm{I}+\alpha K$ is invertible.

A part of the theorem is given as follows:

7.1. Theorem. Suppose $K \in L(H)$ is of finite rank, say $$ Kx = \sum_{j=1}^{n} \langle x, \varphi_j\rangle\psi_j. $$ The operator $I-K$ is invertible if and only if $$\det(\delta_{ij}-\langle\psi_j,\varphi_i\rangle)_{i,j=1}^{n} \neq 0. $$ In this case, for every $y \in H$, $$ (I-K)^{-1}y = y - \frac{1}{\det(a_{ij})}\det\begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} & \langle y, \varphi_1\rangle \\ a_{21} & a_{22} & \cdots & a_{2n} & \langle y, \varphi_2\rangle \\ \vdots & \vdots & & \vdots & \vdots \\ a_{n1} & a_{n2} & \cdots & a_{nn} & \langle y, \varphi_n\rangle \\ \psi_1 & \psi_2 & \cdots & \psi_n & 0 \end{pmatrix} $$

Then the action of the inverse of the operator on $y$ is $y-I$, but how can this tell me what is the inverse of the operator $K$, could anyone clarify this for me please?

Thanks!

$\endgroup$

1 Answer 1

3
$\begingroup$

Your hypotheses on $\{\psi_1,\dotsc,\psi_n\}$ and $\{\varphi_1,\dotsc,\phi_n\}$ make an appeal to that theorem completely and utterly unnecessary. Since $\{\psi_1,\dotsc,\psi_n\} \subset \{\varphi_1,\dotsc,\phi_n\}^\perp$, it immediately follows that $K^2 = 0$. What, then, is $I^2 - \alpha^2 K^2$, and why should this suggest to you an explicit inverse for $I + \alpha K$?

$\endgroup$
6
  • $\begingroup$ you mean that it is not important for me to use this theorem at all? $\endgroup$
    – user426277
    Commented Dec 26, 2017 at 16:57
  • $\begingroup$ Your second statement is not clear for me at all (the statement containing $K^2 = 0$) could you explain it in details please? $\endgroup$
    – user426277
    Commented Dec 26, 2017 at 17:00
  • $\begingroup$ Frankly speaking I do not an answer to your last question ( why should this suggest to you an explicit inverse for ....?) :( ... could u please explain the answer for me ? $\endgroup$
    – user426277
    Commented Dec 26, 2017 at 17:02
  • 1
    $\begingroup$ (1) You can explicitly check that $K^2 \phi = 0$ for all $\phi \in H$ using the hypothesis that $\{\psi_1,\dotsc,\psi_n\} \subset \{\varphi_1,\dotsc,\varphi_n\}^\perp$. (2) How would you usually factorise $x^2-y^2$? $\endgroup$ Commented Dec 26, 2017 at 17:51
  • 1
    $\begingroup$ you mean that the inverse is $(I - \alpha K)$? $\endgroup$
    – user426277
    Commented Dec 26, 2017 at 19:13

You must log in to answer this question.