-1

Let f be $L^1(R)$ and odd function. Then, for any positive $a < A$, there is $M>0$ such that $$ \left|{\int_{a}^A \frac{ \hat{f}(\alpha)}{\alpha} \ d\alpha}\right| \leq M $$

($\hat{f}$ is the fourier transform of transform of f)

holds. How can I show that such M exists?

Siminore
  • 35,136
artes75
  • 83

1 Answers1

0

If $f$ is odd and integrable, then $$ g(x) = \int_{0}^{x}f(y)dy $$ is even and bounded with limits at $\pm\infty$. Therefore, \begin{align} \int_{-R}^{R}e^{-isx}f(x)dx & = e^{-isx}g(x)|_{x=-R}^{R}+is\int_{-R}^{R}e^{-isx}g(x)dx \\ & = -2i\sin(sR)g(R)+is\int_{-R}^{R}e^{-isx}g(x)dx. \\ \frac{1}{2is}\int_{-R}^{R}e^{-isx}f(x)dx & = -\frac{\sin(sR)}{s}g(R)+\int_{0}^{R}\cos(sx)g(x)dx. \end{align} Integrate both sides over $[a,A]$ assuming $0\notin [a,A]$: $$ \int_{a}^{A}\frac{1}{2is}\int_{-R}^{R}e^{-isx}f(x)dxds \\ = -\int_{a}^{A}\frac{\sin(sR)}{s}dsg(R)+\int_{0}^{R}\frac{\sin(Ax)-\sin(ax)}{x} g(x)dx \\ = -\int_{aR}^{AR}\frac{\sin(u)}{u}du g(R)+\int_{0}^{R}\frac{\sin(Ax)-\sin(ax)}{x} g(x)dx $$ Now suppose that $0 \notin [a,A]$. The integral on the left converges as $R\rightarrow\infty$ because the integrand converges uniformly to $\sqrt{2\pi}\hat{f}(s)/2is$. The following converges to $0$ as $R\rightarrow\infty$: $$ \int_{aR}^{AR}\frac{\sin(u)}{u}du g(R) $$ This is because $\int_{0}^{x}\frac{\sin(u)}{u}du$ is uniformly bounded in $x$ and converges to $\pm\frac{\pi}{2}$ as $x\rightarrow \pm\infty$, and $g(x)$ is uniformly bounded. Therefore, the remaining integral on the right must also converge as $R\rightarrow\infty$. Hence, if $0\notin[a,A]$, $$ \sqrt{2\pi}\int_{a}^{A}\frac{1}{2is}\hat{f}(s)ds = \int_{0}^{\infty}\frac{\sin(Ax)-\sin(ax)}{x}g(x)dx. $$ The right side is guaranteed to exist as an improper integral. Integrating the right side by parts gives $$ \left.\int_{0}^{x}\frac{\sin(Ay)-\sin(ay)}{y}dyg(x)\right|_{x=0}^{\infty} - \int_{0}^{\infty}\int_{0}^{x}\frac{\sin(Ay)-\sin(ay)}{y}dy f(x)dx \\ = -\int_{0}^{\infty}\int_{0}^{x}\frac{\sin(Ay)-\sin(ay)}{y}dy f(x)dx $$ which is uniformly bounded by $$ 2\int_{0}^{\pi/2}\frac{\sin(y)}{y}dy\int_{0}^{\infty}|f(x)|dx = \int_{0}^{\pi/2}\frac{\sin(y)}{y}dy\int_{-\infty}^{\infty}|f(x)|dx. $$ So it appears to me that $$ \left|\int_{a}^{A}\frac{\hat{f}(s)}{s}ds\right|\le\sqrt{\frac{2}{\pi}}\int_{0}^{\pi/2}\frac{\sin(x)}{x}dx\int_{-\infty}^{\infty}|f(x)|dx. $$ So I think the best constant $M$ is $$ M = \sqrt{\frac{2}{\pi}}\int_{0}^{\pi/2}\frac{\sin(x)}{x}dx\int_{-\infty}^{\infty}|f(x)|dx $$

Disintegrating By Parts
  • 87,459
  • 5
  • 65
  • 149
  • Thank you for answering my question. But I have a question that why $\left.\int_{0}^{x}\frac{\sin(Ay)-\sin(ay)}{y}dyg(x)\right|{x=0}^{\infty}$ becomes 0 (below the sentence "The right side is guaranteed to exist as an improper integral. Integrating the right side by parts gives"). when x=0, it is clearly 0. but for x=$\infty$, g($\infty$) is non zero and since integral of sinx / x = pi/2, I think $\left.\int{0}^{x}\frac{\sin(Ay)-\sin(ay)}{y}dyg(x)\right|_{x=0}^{\infty}$ should be (1/A - 1/a)(pi/2)g(infty). Is there any part that I am missing? – artes75 Jun 09 '15 at 11:22
  • @artes75 : The reason is because $\int_{0}^{x}\frac{\sin(\rho y)}{y}dy=\int_{0}^{x}\frac{\sin(\rho y)}{\rho y}d(\rho y)=\int_{0}^{x\rho}\frac{\sin x'}{x'}dx'$, which tends to $\int_{0}^{\infty}\frac{\sin x'}{x'}dx'$ as $x\rightarrow\infty$, regardless of the value of $\rho > 0$. – Disintegrating By Parts Jun 09 '15 at 11:56
  • @artes75 : Also, keep in mind that $g(R)$ remains bounded as $R\rightarrow\infty$. So you have something that tends to $0$ times something that tends to $0$; so their product tends to $0$. Oh, I wrote that wrong because I have the wrong limits at substitution. The result is the same, though. I'll fix that now. – Disintegrating By Parts Jun 09 '15 at 12:06
  • @artes75 : fixed now. Thanks for catching that! I changed my mind about the limits of integration along the way and confused myself; but I should have known better because the integral has to be over more periods of $\sin$ as $R\rightarrow\infty$. – Disintegrating By Parts Jun 09 '15 at 12:27
  • Thank you for answer. I really appreciate your help. :) – artes75 Jun 09 '15 at 12:43
  • @artes75 You're welcome. – Disintegrating By Parts Jun 09 '15 at 12:43