Beurling's theorem for \(\mathrm{SL} (2, {\mathbb{R}})\) (Q884984)

The authors prove Beurling's theorem for the full group \(\text{SL}(2, {\mathbb R})\). Beurling's theorem. Let \(f \in L^1({\mathbb R})\). If \[ \int_{{\mathbb R}}\int_{{\mathbb R}}|f(x)\widehat{f}(y)|e^{|xy|}\,dx\, dy < \infty \] where \(\widehat{f}=\int_{{\mathbb R}}f(t)e^{-i\lambda t}\), then \(f=0\) almost everywhere. This theorem is the master theorem in the quantitative uncertainty principle as all the other theorems of this genre follow from it. Let \(G=\text{SL}(2, {\mathbb R})\). Let \(G=\text{KAN}\) be an Iwasawa decomposition of \(G\) where \(K\) is the maximal compact subgroup \(SO(2)\) of \(G\). Let \(\widehat{K}= \{ e_n\mid n \in {\mathbb Z} \}\) be the set of continuous characters of \(K\), given by \(e_{n}(k_{\theta})=e^{in\theta}\) for \(k_{\theta}=\left(\begin{smallmatrix} \cos \theta & \sin \theta \\ -\sin \theta & \cos \theta \end{smallmatrix}\right) \in K\). Let \(M\) be \(\{ \pm I \}\), where \(I\) is the \(2 \times 2\) identity matrix. The unitary dual of \(M\) is \(\widehat{M}=\{ \sigma_{+}, \sigma_{-} \}\) of which \(\sigma_{+}\) is the trivial representation of \(M\) and \(\sigma_{-}\) is the only nontrivial unitary irreducible representation of \(M\). Let \({\mathbb Z}^{\sigma_+}\) (respectively \({\mathbb Z}^{\sigma_-}\)) be the set of even (respectively odd) integers. For \(\sigma \in \widehat{M}\) and \(\lambda \in {\mathbb C}\), let (\(\pi_{\sigma, \lambda}, H_{\sigma}\)) be the principal series representation of \(G\) where \(H_{\sigma}\) is the subspace of \(L^2(K)\) generated by the orthonormal set \(\{ e_{n}\mid n \in {\mathbb Z}^\sigma \}\). The representation \(\pi_{\sigma, \lambda}\) is normalized so that it is unitary if and only if \(\lambda \in i {\mathbb R}\). For every \(k \in {\mathbb Z}^*\), the set of non-zero integers, there is a discrete series representation \(\pi_{k}\) which occurs as a subrepresentation of \(\pi_{\sigma,|k|}\) so that \(k \in {\mathbb Z} \setminus {\mathbb Z}^\sigma\). For \(m ,n \in {\mathbb Z}^{\sigma}\) and \(k \in {\mathbb Z}\setminus {\mathbb Z}^{\sigma}\), let \(\Phi^{m,n}_{\sigma, \lambda} (x) = \langle \pi_{\sigma, \lambda}(x)e_{m}, e_{n} \rangle \) and \(\Psi^{m,n}_{k}(x) = \langle \pi_{k}(x)e^{k}_{m}, e^{k}_{n} \rangle_{k}\) be the matrix coefficients of the principal series and discrete series representations respectively, where \(\{ e_{n}^{k} \}\) are the renormalized basis and \(\langle\;,\;\rangle_{k}\) is the renormalized inner product of \(\pi_{k}\). In particular \(\Phi^{0,0}_{\sigma_+, \lambda}\) is clearly the elementary spherical function which we denote by \(\varphi_{\lambda}\). For a function \(f \in L^1(G)\), let \(\widehat{f}(\sigma, \lambda)\) and \(\widehat{f}(k)\) denote its principal and discrete Fourier transforms at \(\pi_{\sigma, \lambda}\) and \(\pi_{k}\) respectively. Precisely \[ \widehat{f}(\sigma, \lambda) = \int_{G} f(x) \pi_{\sigma, \lambda}(x^{-1}) \,dx \] and \[ \widehat{f}(k) = \int_{G} f(x) \pi_{k}(x^{-1}) \,dx. \] The Plancherel measure on the unitary principal series representations is given by \(d\mu(\sigma, \lambda) = \mu(\sigma, \lambda) d\lambda\) where \[ \mu(\sigma_{+}, i\xi) =\left(\frac{\xi}{4\pi}\right)\tanh \biggl(\frac{\xi \pi}{2}\biggr) \quad \text{and}\quad \mu(\sigma_{-}, i\xi)= \biggl(\frac{\xi}{4\pi}\biggr) \coth \biggl(\frac{\xi \pi}{2}\biggr) \quad \text{for}\quad \xi \in {\mathbb R}. \] The Plancherel measure on the discrete series representations is given by \(\mu(\pi_{k})= \frac{|k|}{2\pi}\) for \(k \in {\mathbb Z}^*\). Main theorem. Let \(f \in L^2(G)\). If \[ \int_{G}\int_{i{\mathbb R}} |f(x)|\,\|\widehat{f}(\sigma, \lambda)\|_{2} \varphi_{|\lambda|}(x) \,dx\,d\mu(\sigma, \lambda) < \infty \] for all \(\sigma \in \widehat{M}\) and \[ \sum_{k \in {\mathbb Z}^*}\frac{|k|}{2\pi}\int_{G}|f(x)|\,\|\widehat{f}(k)\|_2 \varphi_{|k|}(x)\,dx < \infty, \] then \(f=0\) almost everywhere. Here \(\|\cdot\|_2\) is the Hilbert-Schmidt norms. The basic tool of the proof of the main theorem is the Abel transform. The main theorem implies the other quantitative uncertainty principles. We recall that \(G\) has a Cartan decomposition \(G=K\overline{A^{+}}K\), \(x=k_{1}a_{t}k_{2}\) where \(k_{1}, k_{2} \in K\), \(t \geq 0\). Let \(\tau(x)=\tau(k_{1}a_{t}k_{2})=|t|\) for \(x=k_{1}a_{t}k_{2} \in G\). 1. Hardy's theorem: Let \(f : G \rightarrow {\mathbb C}\) be a complex-valued measurable function and assume that (1) \(|f(x)| \leq C e^{-\alpha \tau(x)^2}\) for all \(x \in G\); (2) \(\|\widehat{f}(\sigma, \lambda)\|_2 \leq Ce^{-\beta |\lambda|^2}\) for all \(\sigma \in \widehat{M}\) and \(\lambda \in i {\mathbb R}\) where \(\alpha, \beta\) are positive constants. If \(\alpha \beta > \frac{1}{4}\) then \(f=0\) almost everywhere. 2. Morgan's theorem: Let \(f : G \rightarrow {\mathbb C}\) be measurable and assume that: (1) \(|f(x)| \leq C e^{-\alpha \tau(x)^p}\) for all \(x \in G\); (2) \(\|\widehat{f}(\sigma, \lambda)\|_2 \leq C e^{-\beta |\lambda|^q}\) for all \(\sigma \in \widehat{M}\) and \(\lambda \in i{\mathbb R}\), where \(\alpha, \beta\) are positive constants, \(1 <p< \infty\) and \(\frac{1}{p}+\frac{1}{q} =1\). If \((\alpha p)^{1/p}(\beta q)^{1/q} > 1\), then \(f=0\) almost everywhere. 3. Cowling-Price theorem: Let \(f : G \rightarrow {\mathbb C}\) be measurable and assume that for positive constants \(\alpha\) and \(\beta\) we have: (1) \(e_{\alpha}f \in L^p(G)\); (2) \(e_{\beta}\|\widehat{f}(\sigma, \lambda)\|_{2} \in L^{q}(i{\mathbb R}, d\mu(\sigma, \lambda))\) where \(1 \leq p, q \leq \infty\), \(e_{\alpha}(x)= e^{\alpha \tau(x)^2}\) and \(e_{\beta}(\lambda)=e^{\beta |\lambda|^2}\). If \(\alpha \beta > \frac{1}{4}\), then \(f=0\) almost everywhere. 4. Gelfand-Shilov theorem: Let \(f \in L^2(G)\). Suppose \(f\) satisfies: (1) \(\int_{G}|f(x)|e^{\frac{(\alpha \tau(x))^p}{p}}\,dx < \infty\); (2) \(\int_{i{\mathbb R}}||\widehat{f}(\sigma, \lambda)||_2 e^{\frac{(\beta|\lambda|)^q} {q}} \,d\mu(\sigma, \lambda) < \infty\) where \(1 < p, q < \infty\), \(\frac{1}{p}+\frac{1}{q}=1\), \(\alpha \beta \geq 1\). Then \(f=0\) almost everywhere.