I have been stuck on the following problem for a long time but I could not get the answer. Would you please help me? I was reading one paper [Paternian Salo Uhlmann: Tensor tomography on the surface] in that authors directly write that The space $L^{2}(S M)$ decomposes orthogonally as a direct sum $$ L^{2}(S M)=\bigoplus_{k \in \mathbb{Z}} H_{k} $$ where $H_{k}$ is the eigenspace of $-i V$ corresponding to the eigenvalue $k$. A function $u \in L^{2}(S M)$ has a Fourier series expansion $$ u=\sum_{k=-\infty}^{\infty} u_{k} $$ I wanted to justify that. Would you please help me? What I know is the following:

Theory: The standard Hilbert space theory that the complex Hilbert space $H$ is said to be a direct-sum of the subspaces $H_{m}, m \in \mathbb{Z}^{+}$and we write $H=\oplus_{m=0}^{\infty} H_{m}$, if

1.The subspaces $H_{m}$ are closed for every $m$.

2.The subspaces $H_{m}$ are pairwise orthogonal, i.e., if $h \in H_{m}$ and $g \in H_{m^{\prime}}$, then $\langle f, g\rangle=0$ whenever $m \neq m^{\prime}$

3.Every element $f \in H$ has a representation of the form $$ f=h_{0}+h_{1}+\cdots+h_{m}+\cdots $$ where $h_{m} \in H_{m}\left(m \in \mathbb{Z}^{+}\right)$and the sum converges in the norm of $H$.

The unit sphere bundle is defined as $S M:=\left\{(x, v) ; x \in M, v \in T_{x} M,|v|_{g}=1\right\}$.

The vertical vector field $V$ which differentiate with respect to direction variable $v$ i.e. $$ V u(x, v)=\left.\frac{d}{d s} u\left(\rho_{s}(x, v)\right)\right|_{s=0} $$ where $\rho_{s}(x, v)=\left(x, e^{i s} v\right)$.

For any $k \in \mathbb{Z}$ define $$ H_{k}=\left\{u \in L^{2}(S M):-i V u=k u\right\} $$

Any help or hint or reference will be highly appreciated. Thank you so much.


1 Answer 1


I asume $M$ is a surface and $SM$ is the unit circle bundle. The bundle $SM\to M$ is also a principal $S^1$-bundle and the obvious $S^1$ action on $SM$ that preserves the natural metric on $SM$. $\newcommand{\ii}{\boldsymbol{i}}$ For $p\in SM$ and $e^{\ii t}\in S^1$ denote by $e^{\ii t}p$ the action of $e^{\ii t}$ on $p$. For $\newcommand{\bR}{\mathbb{R}}$ $t\in \bR$ define $$ U_t: L^2(SM)\to L^2(SM),\;\;U_t f(p) =f(e^{\ii t}p). $$ Since the $S^1$-action preserves the metric on $SM$ the operator $U_t$ is unitary. We now have a one-parameter group of unitary operators. Clearly if $f$ is continuous, then $$ \lim_{t\to 0} \Vert U_t f-f\Vert=0,\;\;\Vert-\Vert:=\Vert-\Vert_{L^2}. $$ If $f$ is in $L^2$, then for any $\newcommand{\ve}{{\varepsilon}}$ $\ve>0$ we can find a continuous function $f_\ve$ such that $$ \Vert f-f_\ve\Vert<\ve. $$ We have $$ \Vert U_t f-U_tf_\ve\Vert= \Vert f-f_\ve\Vert<\ve,\;\;\forall t. $$ Hence $$ \Vert U_t f-f\Vert \leq \Vert U_t f-U_t f_\ve\Vert+\Vert U_t f_\ve-f_\ve\Vert<\ve +\Vert U_t f_\ve-f_\ve\Vert $$ We deduce that $$ \limsup_{t\to 0}\Vert U_t f-f\Vert\leq \ve,\;\;\forall \ve>0, $$ and thus $$ \lim_{t\to 0} \Vert U_t f-f\Vert=0,\;\;\forall f\in L^2. $$ This shows that $(U_t)_{t\in \bR}$ is a strongly continuous one-parameter group of unitary operators. Invoking Stone's theorem we deduce that $U_t$ has the form $$ U_t=e^{\ii t A}, $$ where the generator $A$ is a closed selfadjoint operator. In this case the generator is $-\ii V$ so $U_t= e^{t V}$. Note that $U_{2\pi}=1$. Now use the spectral theorem for (unbounded) selfadjoint operators to prove the claims in the paper you mentioned.


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.