Expansion of a real function on 2-sphere in spherical harmonics, so-called Laplace series, converges uniformly for continuously differentiable functions (see e.g. https://projecteuclid.org/euclid.bbms/1103408694).

Is there any explicit example of a real function on 2-sphere, which is merely continuous, and for which the corresponding Laplace series does not converge uniformly? I would expect that this is some "fractal beast", but I might be wrong...


1 Answer 1


One can just adapt an example from the circle to the sphere by lifting such a function from $\mathbb{S}^1$ to zonal spherical harmonics on $\mathbb{S}^2$.

It is then sufficient to give an example of a continuous function with Fourier series which does not converge pointwise, much less uniformly. On the circle, take the example from Katznelson's Introduction to Harmonic Analysis (part B in Theorem 2.1, Ch. 2),

$$f(t)=\sum\limits_{m}^\infty \frac{1}{n^2}\varphi_{\lambda_n}(\lambda_nt),$$

where $\varphi_n=\sigma_{n^2}(\psi_n,t)$ is a trigonometric polynomial of degree $n^2$, $\sigma_n(f,t)$ is the Fejér sum of $f$ to order $n$, and $\psi_n$ are a sequence of continuous functions satisfying $$\|\psi_n\|_{\infty}\leq 1,\ |S_n(\psi_n,0)|>\frac{1}{2}\|D_n\|_{L^1}>\frac{1}{10}\log{n},$$ where $S_n(f,t)$ is the $n$th partial sum of the F.S. of $f$ evaluated at $t$ and $D_n$ is the Dirichlet kernel. Taking $\lambda_{n}=2^{3^n}$ gives a continuous function with Fourier series diverging at $t=0$. Continuity is checked since the convergence of the series defining $f$ is uniform. Divergence of $f$ at $0$ comes since the partial sums of order $\lambda_n^2$ at $t=0$ grow faster than $\frac{K}{n^2}\log{\lambda_n}-3$ which goes to infinity as $n\rightarrow\infty$.

The functions $\psi_n$ above can be constructed by letting them satisfy $\psi_n(t)=\text{sgn}(D_n(t))$ except in small intervals around the discontinuities of $\text{sgn}(D_n(t))$ (as the preceding section in Katznelson details).

  • $\begingroup$ Great, thank you for the reference! So, beast it is, although not a fractal one... $\endgroup$ Oct 22, 2019 at 20:04
  • $\begingroup$ It might be interesting to plot (approximately) such a function and the partial sums of its Fourier Series. The $\lambda_n$ are so large however that it does not look very feasible. $\endgroup$ Oct 22, 2019 at 20:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.