Skip to main content
Post Undeleted by Christian Remling
deleted 926 characters in body
Source Link
Christian Remling
  • 24.2k
  • 2
  • 48
  • 83

Thanks to the discussionAs correctly pointed by Anthony in the comments (Anthony's argument), we already know that this holds for smooth functions $f(z)=\sum a_nz^n$ (equivalentlybelow, it holds for rapidly decaying $a_n$'s)the argument I attempted here originally did not hold water.

So it now suffices to show that an arbitrary $f(e^{ix})=e^{ig(x)}\in H^{1/2}$ can be approximated in $H^{1/2}$ by smooth functions $e^{ih_k(x)}$: we thenThese issues apparently have thatbeen studied quite extensively recently. See for example $\sum n|a_n|^2 = \lim_{k\to\infty} \sum n|a_n^{(k)}|^2$J. Bourgain, One cannot hear the winding number. In particular, and these latter sums are integers.

To doin the introduction to this paper, we recall that $H^{1/2}$ functions (in dimension $1$) can be characterized as those $L^2$ functions $u$Bourgain mentions that also satisfy $$ \int\!\!\!\int dxdy\, \frac{|u(y)-u(x)|^2}{|x-y|^2} < \infty . \quad\quad\quad\quad (1) $$ The $L^2$ condition is automatic here since our functions satisfy $|f|=1$, and (1) for $u=e^{ig}$ implies the same condition foranswer to your question is $g$ after some precautions associated withyes; the non-uniqueness of $g$. More specifically, let's assume that $|u^{-1}(-1)|=0$, and let's then take $|g(x)|<\pi$. Then $|u(y)-u(x)|$result is comparableattributed to $|g(x)-g(y)|$ on every set $A_n=\{ x: |g(x)\mp\pi|\ge 1/n \}$, and these exhaust our domain.

Thus $g\in H^{1/2}$ also, as claimed,Boutet de Monvel and if we now approximate $h_k\to g$ in $H^{1/2}$, then $e^{ih_k}\to e^{ig}$ also, by a similar argument: (1) for $u=e^{ih_k}-e^{ig}$ together with $\|u\|_2$ essentially computes the $H^{1/2}$ norm, but this is controlled by (1) for $u=h_k-g$Gabber.

Thanks to the discussion in the comments (Anthony's argument), we already know that this holds for smooth functions $f(z)=\sum a_nz^n$ (equivalently, it holds for rapidly decaying $a_n$'s).

So it now suffices to show that an arbitrary $f(e^{ix})=e^{ig(x)}\in H^{1/2}$ can be approximated in $H^{1/2}$ by smooth functions $e^{ih_k(x)}$: we then have that $\sum n|a_n|^2 = \lim_{k\to\infty} \sum n|a_n^{(k)}|^2$, and these latter sums are integers.

To do this, we recall that $H^{1/2}$ functions (in dimension $1$) can be characterized as those $L^2$ functions $u$ that also satisfy $$ \int\!\!\!\int dxdy\, \frac{|u(y)-u(x)|^2}{|x-y|^2} < \infty . \quad\quad\quad\quad (1) $$ The $L^2$ condition is automatic here since our functions satisfy $|f|=1$, and (1) for $u=e^{ig}$ implies the same condition for $g$ after some precautions associated with the non-uniqueness of $g$. More specifically, let's assume that $|u^{-1}(-1)|=0$, and let's then take $|g(x)|<\pi$. Then $|u(y)-u(x)|$ is comparable to $|g(x)-g(y)|$ on every set $A_n=\{ x: |g(x)\mp\pi|\ge 1/n \}$, and these exhaust our domain.

Thus $g\in H^{1/2}$ also, as claimed, and if we now approximate $h_k\to g$ in $H^{1/2}$, then $e^{ih_k}\to e^{ig}$ also, by a similar argument: (1) for $u=e^{ih_k}-e^{ig}$ together with $\|u\|_2$ essentially computes the $H^{1/2}$ norm, but this is controlled by (1) for $u=h_k-g$.

As correctly pointed by Anthony in the comments below, the argument I attempted here originally did not hold water.

These issues apparently have been studied quite extensively recently. See for example J. Bourgain, One cannot hear the winding number. In particular, in the introduction to this paper, Bourgain mentions that the answer to your question is yes; the result is attributed to Boutet de Monvel and Gabber.

Post Deleted by Christian Remling
Source Link
Christian Remling
  • 24.2k
  • 2
  • 48
  • 83

Thanks to the discussion in the comments (Anthony's argument), we already know that this holds for smooth functions $f(z)=\sum a_nz^n$ (equivalently, it holds for rapidly decaying $a_n$'s).

So it now suffices to show that an arbitrary $f(e^{ix})=e^{ig(x)}\in H^{1/2}$ can be approximated in $H^{1/2}$ by smooth functions $e^{ih_k(x)}$: we then have that $\sum n|a_n|^2 = \lim_{k\to\infty} \sum n|a_n^{(k)}|^2$, and these latter sums are integers.

To do this, we recall that $H^{1/2}$ functions (in dimension $1$) can be characterized as those $L^2$ functions $u$ that also satisfy $$ \int\!\!\!\int dxdy\, \frac{|u(y)-u(x)|^2}{|x-y|^2} < \infty . \quad\quad\quad\quad (1) $$ The $L^2$ condition is automatic here since our functions satisfy $|f|=1$, and (1) for $u=e^{ig}$ implies the same condition for $g$ after some precautions associated with the non-uniqueness of $g$. More specifically, let's assume that $|u^{-1}(-1)|=0$, and let's then take $|g(x)|<\pi$. Then $|u(y)-u(x)|$ is comparable to $|g(x)-g(y)|$ on every set $A_n=\{ x: |g(x)\mp\pi|\ge 1/n \}$, and these exhaust our domain.

Thus $g\in H^{1/2}$ also, as claimed, and if we now approximate $h_k\to g$ in $H^{1/2}$, then $e^{ih_k}\to e^{ig}$ also, by a similar argument: (1) for $u=e^{ih_k}-e^{ig}$ together with $\|u\|_2$ essentially computes the $H^{1/2}$ norm, but this is controlled by (1) for $u=h_k-g$.