My professor told me that

Suppose that $f \in L^p(-\pi, \pi)$ (i.e. $f$ is 2$\pi$-periodic and $\|f\|_{L^p} < \infty$). If $1<p<\infty$, then the Fourier series of $f$ converges to $f$ in $L^p$.

If $p = 2$, then this statement is easy to understand since $e^{inx}, \, n = 0, \pm1,\pm 2,\dots$ form an orthonormal basis of $L^2$. Also, $L^2$ is Hilbert. Thus, $\sum\langle e^{inx}, f(x)\rangle e^{inx} = f(x)$ in $L^2$ sense (i.e. $\|\left(\sum\langle e^{inx}, f(x)\rangle e^{inx}\right) - f(x)\|_{L^2} = 0$).

But how to understand the above statement if $p \ne 2$? Then the space $L^p$ with $1<p<2$ or $2<p < \infty$is not even Hilbert. Hope someone can give a nice interpretation.

  • $\begingroup$ Fourier series here given by the same formula: $c_n=\int e^{-inx}f(x)dx$ and the statement concludes, that $f_n=\sum_{-n}^n c_n e^{inx}$ convergers to $f$ in $L^p$. $\endgroup$ – Boris Bilich May 15 at 14:00
  • $\begingroup$ They are Banach spaces so the usual norm convergence applies $\endgroup$ – Conrad May 15 at 14:10
  • $\begingroup$ You can interpret the result in $L^p$-norm and almost everywhere (pointwise convergence outside of a zero-measure set). But the difficulty of the result is different for this two means of convergence. $\endgroup$ – Adrián González-Pérez May 16 at 10:53

There are two main ways of understanding the convergence of Fourier series in $L^p$.

  1. In $L^p$-norm, for $1 < p < \infty$. The proof of this result is -I think- due originally to Riesz in the 30's and although nontrivial is typically covered in graduate courses in Fourier analysis. The key point is to see that the Fourier multipliers associated with an interval $[-N,N]$ is bounded in $L^p$. You can read a proof of this in stard texts like [Do] or [Ka]. Norm convergence fails in $L^1$ and $L^\infty$ by an easy application of the uniform bound principle.

  2. Almost everywhere for $1 < p < \infty$. This is the (more difficult) Carleson-Hunt theorem [Ca, Hu]. It was proven by Carleson in the case of $L^2$ and by Hunt for general $p$. It requires to prove a maximal bound for the Dirichlet kernel. This is note generally covered in graduate courses and is more of an advance topic. Almost everywhere convergence fails in $p = 1$.

[Ca]: Carleson, Lennart, On convergence and growth of partial sums of Fourier series, Acta Math. 116, 135-157 (1966). ZBL0144.06402.

[Do]: Duoandikoetxea, Javier, Fourier analysis. Transl. from the Spanish and revised by David Cruz-Uribe, Graduate Studies in Mathematics. 29. Providence, RI: American Mathematical Society (AMS). xviii, 222 p. (2001). ZBL0969.42001.

[Hu]: Carleson, Lennart, On convergence and growth of partial sums of Fourier series, Acta Math. 116, 135-157 (1966). ZBL0144.06402.

[Ka]: Katznelson, Yitzhak, An introduction to harmonic analysis. 2nd corr. ed, Dover Books on Advanced Mathematics. New York: Dover Publications, Inc. XIV, 264 p. $ 4.00 (1976). ZBL0352.43001.


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.