1
$\begingroup$

If $A$ is an unbounded, self-adjoint operator on a separable Hilbert space with spectral family $E_t$, then $\exp(iA)$, defined by functional calculus with function $f(t)=exp(it)$, is an unitary operator. My question is, what about the spectral family of $exp(iA)$? Is again the unbounded $E_t$? Must not be a bounded one? How to calculate it?

$\endgroup$
  • $\begingroup$ I guess $E_t = 1_{(-\infty,t]}(A)$? Are you familiar with the spectral measure induced by a spectral family, $E(B)=1_B(A)$ for any Borel set $B\subseteq \mathbb{C}$? $\endgroup$ – Jonas Dahlbæk May 24 '17 at 18:50
  • $\begingroup$ Yes $E_t=E (]-\infty,t])$. I familiar with spectral measure $\endgroup$ – dioxoid May 24 '17 at 19:10
0
$\begingroup$

For a closed, densely defined, normal operator $T$ in a Hilbert space $\mathcal{H}$, denote by $E^T(B)=1_B(T)$ the spectral measure induced by $T$, and if $B\subseteq \mathbb{C}$ is a Borel set and $x\in\mathcal{H}$, let $E^T_x(B):=\langle x, E^T(B) x\rangle$. Recall that $E^T$ is uniquely determined by the family of complex measures $(E^T_x)_{x\in\mathcal{H}}$. Now, if $f:\mathbb{C}\rightarrow \mathbb{C}$ is any measurable function, then $$ E_x^{f(T)}(B) = \langle x,1_B(f(T))x\rangle = \langle x,1_{f^{-1}(B)}(T)x\rangle = E_x^T(f^{-1}(B)). $$ It follows that $E^{f(T)}(B)=E^T(f^{-1}(B))$.

If $A$ is self-adjoint and $U=e^{iA}$, then the spectral measure of $U$ is supported in the unit circle $S\subseteq \mathbb{C}$, and is explicitly given by the formula $E^U(B)=E^A(f^{-1}(B))$, with $f(t)=e^{it}$.

$\endgroup$
  • $\begingroup$ So if $T=A=\int_RtdE_t$ is self adjoint and $U=e^{iA}=\int_Re^{it}dE_t$, then if B is a Borel subset of the circle T, $E^U(B)=E^A(-i\ln(B))$ and $U=\int_T\lambda dE_{-i\ln\lambda}$ ? $\endgroup$ – dioxoid May 24 '17 at 22:26
  • $\begingroup$ With a appropriate definition of the logarithm, $E^U(B)=E^A(-i\ln(B)))$, and $U=\int_S \lambda \, dE^U(\lambda) = \int_{\mathbb{R}} e^{i\lambda} \, dE^A_\lambda$ $\endgroup$ – Jonas Dahlbæk May 24 '17 at 22:30
  • $\begingroup$ Ok, many thanks $\endgroup$ – dioxoid May 24 '17 at 22:33

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.