Let $\mathbb{D}$ and $\mathbb{T}$ denote the open unit disk and unit circle in $\mathbb{C}$ respectively. We write $Hol(\mathbb{D})$ for the space of all holomorphic functions on $\mathbb{D}.$ The Hardy spaces on $\mathbb{D}$ are defined as: $$H^{p}:= \left\{ f\in Hol\left( \mathbb{D}\right) :\sup _{r < 1}\int ^{2\pi }_{0}\left| f\left( re^{i\theta}\right) \right| ^{p}d\theta < \infty \right\} \;\;\;\;(0<p<\infty), $$ $$H^{\infty }:= \left\{ f\in Hol\left( \mathbb{D}\right) :\sup_{z\in D}\left| f\left( z\right) \right| < \infty \right\}.$$ A function $g\in H^p(\mathbb{D})$ is outer if there exists a function $G:\mathbb{T}\longrightarrow [0,\infty)$ with $G\in L^1(\mathbb{T})$ such that $$g\left( z\right) =\alpha \text{exp}\left( \int ^{2\pi }_{0}\dfrac {e^{i\theta }+z}{e^{i\theta }-z}G\left( e^{i\theta }\right) \dfrac {d\theta }{2\pi }\right) \qquad(z\in \mathbb{D})$$ and $|\alpha|=1$. Let $\mathscr{P}[h]$ denote the closed subspace generated by the functions $z^n h(z),\; n=0,1,2,....$, ie $\mathscr{P}[h]$ consists of all $H^p$ functions that can be approximated by polynomial multiples of $h$. Note that $\mathscr{P}[1]=H^p$, since polynomials are dense in $H^p$.

I wanted to ask:

I know that if $h$ is not outer, then $\mathscr{P}[h]\neq H^p $. Will this also imply that $1\notin \mathscr{P}[h]$ with dist$(1,\mathscr{P}[h])>0.$? Will this distance be bounded below by some +real number(irrespective of given outer function h)?

  • 1
    $\begingroup$ The answer to (1) depends on what you mean by a "subspace generated by (...)". If you mean "closed subspace generated by (...)", as suggested by the remaining part of the quoted sentence, then yes, $\mathscr P[h]$ is closed. :-) $\endgroup$ – Mateusz Kwaśnicki Jul 31 '20 at 13:32
  • $\begingroup$ You have basically answered your own question already (and the answer is, yes, $1\notin P[h]$ if $h$ is not outer). This follows because $P[h]$ is invariant under multiplication by $z$, so if we had $1\in P[h]$, then this space would contain all polynomials. (And then the distance is positive automatically, as $P[h]$ is closed.) $\endgroup$ – Christian Remling Jul 31 '20 at 18:25
  • $\begingroup$ By the way, math.stackexchange.com is a better site to ask such questions. $\endgroup$ – Christian Remling Jul 31 '20 at 18:27
  • $\begingroup$ @ChristianRemling I still have a part of the question remaining which asks if the distance is bounded below by any positive real number... $\endgroup$ – user429197 Jul 31 '20 at 18:47

The answer to your final question is no if $p < \infty$ and yes if $p = \infty$.

Let $a > 0$ and let $h_a$ be a singular inner function given by $$h_a(z) = \exp(-a(1-z)/(1+z)). $$ Note that $|h_a(z)| \leqslant 1$ and $h_a(z) \to 1$ as $a \to 0^+$. Therefore, by the dominated convergence theorem, $h_a$ converges to $1$ in $H^p$ for every $p < \infty$ (at least when $p \geqslant 1$; I bet the same is true when $p < 1$, but I have never worked with these spaces and so I am not entirely sure). In particular, the distance between $\mathscr P[h_a]$ and $1$ converges to zero as $a \to 0^+$.

On the other hand, the distance in $H^\infty$ between $1$ and any function $f$ which is not outer is at least $1$, because either $f$ has a zero in the unit disk or some non-tangential boundary limit of $f$ is equal to $0$. Thus, the distance between $\mathscr{P}[h]$ and $1$ is either $0$ or $1$.

  • $\begingroup$ The $H^\infty$ norm is equal to the supremum norm inside the unit disk, so if $f$ takes values arbitrarily close to $0$ in the unit disk, then $\|f - 1\|_{H^\infty}$ is at least $1$. $\endgroup$ – Mateusz Kwaśnicki Jul 31 '20 at 21:36
  • $\begingroup$ Huh, your comment just disappeared... $\endgroup$ – Mateusz Kwaśnicki Jul 31 '20 at 21:36
  • $\begingroup$ Will there be any outer function for which this distance will be some k>0? $\endgroup$ – user429197 Aug 2 '20 at 7:31
  • $\begingroup$ There's nothing special about constant $1$, I think, just approximate any $H^p$ function by a polynomial $P_n$, and then this polynomial by an inner function $h_{a_n}$ multiplied by $P_n$. $\endgroup$ – Mateusz Kwaśnicki Aug 2 '20 at 20:26

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.