Embedding H ∞ ​ ( 𝔻 ) H^{\infty}(\mathbb{D}) into L ∞ ​ ( 𝕋 ) L^{\infty}(\mathbb{T}) : a proof without non-tangential limits

Mario P. Maletzki Β· 2025

Read Embedding H ∞ ​ ( 𝔻 ) H^{\infty}(\mathbb{D}) into L ∞ ​ ( 𝕋 ) L^{\infty}(\mathbb{T}) : a proof without non-tangential limits by Mario P. Maletzki (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-16489.txt.

Introduction

Mario Maletzki gives an accessible, self contained proof that the algebra of bounded analytic functions on the disk embeds isometrically into bounded functions on the circle. The classical route uses Fatou's theorem on non tangential limits of analytic functions. That theorem is heavy machinery for a result students meet early. Maletzki avoids it entirely in this note.

Only standard undergraduate complex and functional analysis are needed for the argument. The embedding preserves algebraic structure and norm exactly. Such boundary value results link interior function theory to boundary behavior. A lighter proof widens who can learn the result rigorously. The note is aimed at teaching as much as research, streamlining a staple of Hardy space theory.

Catalog record

CREATORMario P. Maletzki
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE13 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-16489

Browse the full library Β· About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Embedding H ∞ ​ ( 𝔻 ) H^{\infty}(\mathbb{D}) into L ∞ ​ ( 𝕋 ) L^{\infty}(\mathbb{T}) : a proof without non-tangential limits Mario P. Maletzki Abstract The purpose of this note is to show in an accessible and self-contained way the existence of an isometric algebra embedding from H ∞ ​ ( 𝔻 ) H^{\infty}(\mathbb{D}) into L ∞ ​ ( 𝕋 ) L^{\infty}(\mathbb{T}) , without appealing to Fatou’s classical theorem on non-tangential limits of analytic functions, and relying only on results from complex and functional analysis that are typically covered in a standard undergraduate course. 1 Introduction Beyond its intrinsic interest, the existence of an isometric embedding H^{\infty}(\mathbb{D})\hookrightarrow L^{\infty}(\mathbb{T}) has proven to be useful in the study of the classical Hardy space H^{\infty}(\mathbb{D}) of bounded analytic functions on \mathbb{D} , and for its applications (such as the identification of the Shilov boundary of H^{\infty}(\mathbb{D}) or the study interpolating sequences for this space in terms of Poisson integrals) the reader is referred to Garnett (1981) . For background and notation on functional and complex analysis see Conway (1990) and Rudin (1987) respectively, and for basic facts on the theory of Hardy spaces see for instance Duren (1970) . We first recall that the Poisson kernel in the unit disk is given by P_{r}(\theta):=\frac{1-r^{2}}{1-2r\cos(\theta)+r^{2}}=\frac{1-|re^{i\theta}|^{2}}{|1-re^{i\theta}|^{2}},\quad\mbox{ for }0\leq r<1, and that it is an approximate identity for L^{1}(\mathbb{T}) , which means that viewing the Poisson kernel as a net (P_{r})_{r}\subset L^{1}(\mathbb{T}) , we have that ||P_{r}\ast f-f||_{1}\rightarrow 0 as r\rightarrow 1 . The last fact follows from the following well-known properties: i) P_{r}(\sigma) is decreasing on 0<\sigma<\pi , ii) P_{r}(\sigma)\geq 0 for every \sigma , iii) P_{r}(\sigma)=P_{r}(-\sigma) for every 0<\sigma<\pi , iv) It holds \frac{1}{2\pi}\int_{-\pi}^{\pi}P_{r}(\sigma)\ d\sigma=1. v) For every \delta>0 , \lim_{r\rightarrow 1}\sup_{|t|\geq\delta}\{P_{r}(t)\}=0. Given a harmonic function u in a disk D(a,R) of the complex plane that is also continuous in \overline{D(a,R)} , we have that for each z=a+\rho e^{i\sigma}\in D(a,R) the following equality holds u(z)=\frac{1}{2\pi}\int_{-\pi}^{\pi}u(e^{it})P_{\frac{\rho}{R}}(\sigma-t)\ dt, (1) which is called the Poisson integral formula for harmonic functions . To derive this equality, we first prove that as a consequence of both Cauchy’s integral theorem and Cauchy’s integral formula, we have that ( 1 ) holds for analytic functions, and from there it will be straightforward that 1 holds for harmonic functions. Lemma 1 . If f is an analytic function on D(0,1+\varepsilon) for some \varepsilon>0 , then f(re^{i\sigma})=\frac{1}{2\pi}\int_{-\pi}^{\pi}f(e^{it})P_{r}(\sigma-t)\ dt\quad\mbox{ for every }\ re^{i\sigma}\in\mathbb{D} (2) Proof. By Cacuhy’s integral formula we have f(z)=\frac{1}{2\pi i}\int_{\partial\mathbb{D}}\frac{f(w)}{w-z}\ dw\quad\mbox{ for every }\ z\in\mathbb{D}, and since for any z\in\mathbb{D} the function w\mapsto\frac{f(w)}{w-1/\overline{z}} is holomorphic on an open set containing \overline{\mathbb{D}} , by Cauchy’s integral theorem we have \int_{\partial\mathbb{D}}\frac{f(w)}{w-1/\overline{z}}\ dw=0 , and thus f(z)=\frac{1}{2\pi i}\int_{\partial\mathbb{D}}f(w)\left(\frac{1}{w-z}-\frac{1}{w-1/\overline{z}}\right)\ dw. Now, if w\in\partial\mathbb{D} then \frac{1}{w}=\overline{w} , and computing \frac{1}{w-z}-\frac{1}{w-1/\overline{z}}=\frac{(w-1/\overline{z})-(w-z)}{w^{2}-w(1/\overline{z}+z)+z/\overline{z}}=\frac{z-1/\overline{z}}{w^{2}-w(1/\overline{z}+z)+z/\overline{z}}= =\frac{1}{w}\frac{|z|^{2}-1}{w\overline{z}-(1+|z|^{2})+z/w}=\frac{1}{w}\frac{1-|z|^{2}}{1-2\Re(z\overline{w})+|z|^{2}}, and we conclude f(re^{i\sigma})=\frac{1}{2\pi i}\int_{\partial\mathbb{D}}\frac{f(w)}{w}\frac{1-|z|^{2}}{1-2\Re(z\overline{w})+|z|^{2}}\ dw= =\frac{1}{2\pi i}\int_{-\pi}^{\pi}\frac{f(e^{it})}{e^{it}}\frac{1-r^{2}}{1-2r\cos(\sigma-t)+r^{2}}ie^{it}\ dw=\frac{1}{2\pi}\int_{-\pi}^{\pi}f(e^{it})P_{r}(\sigma-t)\ dt. ∎ Remark . From this we see that if f is holomorphic on \mathbb{D} and re^{i\sigma}\in\mathbb{D} , then for r<\rho<1 , the function f_{\rho}(z):=f(\rho z) is holomorphic on D(0,1/\rho) and therefore f(re^{i\sigma})=f_{\rho}((r/\rho)e^{i\sigma})=\frac{1}{2\pi}\int_{-\pi}^{\pi}f_{\rho}(e^{it})P_{r/\rho}(\sigma-t)\ dt. To simplify the notation, for a point z=re^{i\sigma}\in\mathbb{D} we will write P_{z}(t):=P_{r}(\sigma-t) , and for a function f defined on \mathbb{T} , we may consider the function defined on [-\pi,\pi] and write f(t) instead of f(e^{it}) . 2 Existence of isometric embedding We now proceed to show how the existence of the isometric embedding H^{\infty}(\mathbb{D})\hookrightarrow L^{\infty}(\mathbb{T}) can be deduced from Banach-Alaoglu’s theorem, which assures that any bounded sequence in the dual of a separable normed space has a w^{*} -convergent subsequence. Theorem 1 . Given f\in H^{\infty}(\mathbb{D}) , there is a function f^{*}\in L^{\infty}(\mathbb{T}) such that for every re^{i\sigma}\in\mathbb{D} it holds f(re^{i\sigma})=\frac{1}{2\pi}\int_{-\pi}^{\pi}f^{*}(e^{it})P_{r}(\sigma-t)\ dt. The mapping f\mapsto f^{*} defines an isometric algebra homomorphism from H^{\infty}(\mathbb{D}) into L^{\infty}(\mathbb{T}) . For the proof of the theorem we will need the following lemma: Lemma 2 . The linear space generated by \{P_{z}\}_{z\in\mathbb{D}} is dense in L^{1}(\mathbb{T}) . Proof. If \overline{\langle\{P_{z}\}_{z\in\mathbb{D}}\rangle}\neq L^{1}(\mathbb{T}) , by the Hahn-Banach theorem we would find a non-zero function f\in L^{\infty}(\mathbb{T}) such that (f\ast P_{r})(\sigma)=\frac{1}{2\pi}\int_{-\pi}^{\pi}f(e^{it})P_{z}(t)\ dt=0\quad\mbox{ for every }z=re^{i\sigma}\in\mathbb{D}, and therefore for a fixed 0<r<1 it is f\ast P_{r}\equiv 0 . Now, since L^{\infty}(\mathbb{T})\subset L^{1}(\mathbb{T}) and (P_{r})_{r} is an approximate identity for L^{1}(\mathbb{T}) , we have ||P_{r}\ast f-f||_{1}\rightarrow 0 as r\rightarrow 1 , and thus that f=0 , which is a contradiction. ∎ Proof of Theorem 1 . Let f_{r}(t):=f(rt)\in L^{\infty}(\mathbb{T}) for 0<r<1 . The Maximal Modulus Principle implies that ||f_{r}||\leq||f_{s}|| if 0<r\leq s<1 , and thus ||f_{r}||\leq\lim_{s\rightarrow 1}||f_{s}||=||f|| . Since L^{\infty}(\mathbb{T})\cong(L^{1}(\mathbb{T}))^{*} and L^{1}(\mathbb{T}) is separable, we have that \overline{B_{L^{\infty}(\mathbb{T})}(0,||f||)} is w^{*} -sequentially compact, and therefore for any sequence (f_{r_{n}})_{n} where (r_{n})\subset(0,1) with r_{n}\rightarrow 1 , there is a subsequence (f_{r_{n_{k}}})_{k} and a function f^{*}\in\overline{B_{L^{\infty}(\mathbb{T})}(0,||f||)} such that f_{r_{n_{k}}}\xrightarrow{w^{*}}f^{*} . For a fixed re^{i\sigma}\in\mathbb{D} , considering any r<r_{n_{k}}<1 , we have that f(re^{i\sigma})-\frac{1}{2\pi}\int_{-\pi}^{\pi}f^{*}(e^{it})P_{r}(\sigma-t)\ dt= \frac{1}{2\pi}\left(\int_{-\pi}^{\pi}f_{r_{n_{k}}}(e^{it})P_{\frac{r}{r_{n_{k}}}}(\sigma-t)\ dt-\int_{-\pi}^{\pi}f^{*}(e^{it})P_{r}(\sigma-t)\ dt\right)= \frac{1}{2\pi}\int_{-\pi}^{\pi}f_{r_{n_{k}}}(e^{it})(P_{\frac{r}{r_{n_{k}}}}-P_{r})(\sigma-t)\ dt-\frac{1}{2\pi}\int_{-\pi}^{\pi}(f_{r_{n_{k}}}-f^{*})(e^{it})P_{r}(\sigma-t)\ dt and thus \left|f(re^{i\sigma})-\frac{1}{2\pi}\int_{-\pi}^{\pi}f^{*}(e^{it})P_{r}(\sigma-t)\ dt\right|\leq \left|\frac{1}{2\pi}\int_{-\pi}^{\pi}f_{r_{n_{k}}}(e^{it})(P_{\frac{r}{r_{n_{k}}}}-P_{r})(\sigma-t)dt\right|+\left|\frac{1}{2\pi}\int_{-\pi}^{\pi}(f_{r_{n_{k}}}-f^{*})(e^{it})P_{r}(\sigma-t)dt\right| \leq||f||_{\infty}||P_{\frac{r}{r_{n_{k}}}}-P_{r})||_{1}+\left|\frac{1}{2\pi}\int_{-\pi}^{\pi}(f_{r_{n_{k}}}-f^{*})(e^{it})P_{r}(\sigma-t)\ dt\right|, and taking limits as k\rightarrow\infty , since r_{n_{k}}\rightarrow 1 and f_{r_{n_{k}}}\xrightarrow{w^{*}}f^{*} , we conclude that f(re^{i\sigma})=\frac{1}{2\pi}\int_{-\pi}^{\pi}f^{*}(e^{it})P_{r}(\sigma-t)\ dt. To see that f\mapsto f^{*} defines an isometric algebra homomorphism from H^{\infty}(\mathbb{D}) into L^{\infty}(\mathbb{T}) , we first check that i:H^{\infty}(\mathbb{D})\rightarrow L^{\infty}(\mathbb{T}) with i(f):=f^{*} is well-defined, that is, it does not depend on the sequence (r_{n})_{n} . Considering another sequence (r_{n}^{\prime})_{n}\subset(0,1) with r_{n}^{\prime}\rightarrow 1 , this would lead to another function g^{*}\in L^{\infty}(\mathbb{T}) such that f(re^{i\sigma})=\frac{1}{2\pi}\int_{-\pi}^{\pi}g^{*}(e^{it})P_{r}(\sigma-t)\ dt\quad\mbox{ for every }re^{i\sigma}\in\mathbb{D}, but then we would have for every z=re^{i\sigma}\in\mathbb{D} that (f^{*}-g^{*})(P_{z})=\frac{1}{2\pi}\int_{-\pi}^{\pi}(f^{*}-g^{*})(e^{it})P_{r}(\sigma-t)\ dt=f(z)-f(z)=0, and by the previous lemma we would conclude that f^{*}=g^{*} . That i is a linear homomorphism is straightforward, and since ||f^{*}|\leq||f|| , it is only left to see that ||f||\leq||f^{*}|| , but this is clear because |f(z)|\leq\frac{1}{2\pi}\int_{-\pi}^{\pi}|f^{*}(e^{it})|P_{r}(\sigma-t)\ dt\leq\frac{||f^{*}||}{2\pi}\int_{-\pi}^{\pi}P_{r}(\sigma-t)\ dt=||f^{*}||, and thus ||f||=\sup_{z\in\mathbb{D}}|f(z)|\leq||f^{*}|| ∎ 3 Extending the embedding to harmonic function Since in the proof of Theorem 1 we only used the fact that f is holomorphic on \mathbb{D} when we claimed that f(re^{i\sigma})=\frac{1}{2\pi}\int_{-\pi}^{\pi}f_{r_{n_{k}}}(e^{it})P_{r/{r_{n_{k}}}}(\sigma-t)\ dt, and the same equality holds for complex-valued harmonic function \mathbb{D} , it is straightforward that the isometric algebra embedding can actually be extended to h^{\infty}_{\mathbb{C}}(\mathbb{D}) , the Hardy space of bounded complex-valued harmonic functions on \mathbb{D} . Moreover, the isometric embedding h_{\mathbb{C}}^{\infty}(\mathbb{D})\hookrightarrow L^{\infty}(\mathbb{T}) turns out to be surjective: Theorem 2 . If f is a bounded complex-valued harmonic function on \mathbb{D} , then there is a function f^{*}\in L^{\infty}(\mathbb{T}) such that for every re^{i\sigma}\in\mathbb{D} it holds f(re^{i\sigma})=\frac{1}{2\pi}\int_{-\pi}^{\pi}f^{*}(e^{it})P_{r}(\sigma-t)\ dt. The mapping f\mapsto f^{*} defines an isometric algebra isomorphism from h^{\infty}(\mathbb{D}) onto L^{\infty}(\mathbb{T}) . To prove this, we first recall that for an integrable function f\in L^{1}(\mathbb{T}) its Poisson integral is defined by F(z):=\frac{1}{2\pi}\int_{-\pi}^{\pi}f(e^{it})P_{z}(t)\ dt, and that it satisfies the following: Proposition 1 . For any real-valued function f\in L^{1}(\mathbb{T}) , its Poisson integral F(z) is an harmonic function on \mathbb{D} . Proof. Consider the function g:[-\pi,\pi]\times\mathbb{D}\rightarrow\mathbb{C} defined by g(t,z):=\frac{1+e^{-it}z}{1-e^{-it}z} , which for each fixed t is holomorphic on \mathbb{D} , and define G(z):=\frac{1}{2\pi}\int_{-\pi}^{\pi}g(t,z)f(t)\ dt. Since f is real-valued, we have that \Re(G)(z)=\frac{1}{2\pi}\int_{-\pi}^{\pi}\Re(g(t,z))f(t)\ dt , and because g(t,re^{i\sigma})=\frac{1+re^{-i(\sigma-t)}}{1-re^{-i(\sigma-t)}}\frac{1-re^{-i(t-\sigma)}}{1-re^{-i(t-\sigma)}}= =\frac{1-re^{-i(t-\sigma)}+re^{-i(\sigma-t)}-r^{2}}{1-re^{-i(t-\sigma)}-re^{-i(\sigma-t)}-r^{2}}=\frac{(1-r^{2})+i2\sin(\sigma-t)}{1-2r\cos(\sigma-t)+r^{2}}, we have that \Re(G)(z)=\frac{1}{2\pi}\int_{-\pi}^{\pi}P_{z}(t)f(t)\ dt , so it is enough to prove that G(z) is holomorphic on \mathbb{D} to show that the Poisson integral of f is harmonic on \mathbb{D} . Now, considering a closed triangle \Delta\subset\mathbb{D} , the function g(t,z) is bounded on \partial\mathbb{D}\times\partial\Delta and f(t) is integrable in \partial\mathbb{D} , so the function g(t,z)f(t)\in L^{1}(\partial\mathbb{D}\times\partial\Delta)

Read the complete text