Let $f(x) = (1-x)^{\alpha} \log(1-x)$ be defined on $[0,1]$ with $\alpha > 0$ some real exponent.

Does such a function belong to the Sobolev (or Bessel potential) space $H^{\beta}((0,1))$ with $\beta < \alpha + 1/2$ ?

If $\alpha, \beta$ are non-negative integers, then $H^{\beta}((0,1)) = W^{\beta,2}((0,1))$ and using Leibniz rule, I find that

$$f^{(k)}(x) \approx (1-x)^{\alpha-k}\log(1-x) + (1-x)^{\alpha-k+1}$$

A primitive of $f^{(k)}(x)^2$ is found using integration by parts :

$$\int f^{(k)}(x)^2 dx \approx (1-x)^{2\alpha-2k+3} + (1-x)^{2\alpha-2k+1} \left( \log(1-x)^2 + \log(1-x) + 1\right)$$

so integrating on $(0,1)$, this will be finite given $2\alpha-2k+1 > 0$ and this must be true for $k = 0, 1, ..., \beta$. This is satisfied for $\beta < \alpha + 1/2$.

Is there an easy way to generalize this for real-valued $\alpha$ and $\beta$ ? Extension operators and Fourier transforms will be involved and it seems a bit tricky. Maybe there is some way using some embeddings or interpolation results ?

  • $\begingroup$ An idea, which does not directly use Besov spaces as in the book cited in the answer below, is to write $f(x) = \sum_{j \leq 0} \phi(2^{-j}(1-x)) f(x)$, where $\phi( \cdot )$ defines a dyadic partition of unity (as in the Littlewood-Paley decomposition). The sum is locally finite, so we can interchange derivative and infinite sum. Each term in the sum is smooth. After differentiating, we can square-integrate and estimate this. To get estimates on the $L^2$ norms of fractional derivatives, we use interpolation. $\endgroup$
    – Desura
    Nov 28, 2021 at 13:13

1 Answer 1


Yes, you can find the proof in Lemma 1 on page 44 of the book Runst, Thomas and Sickel, Winfried. Sobolev Spaces of Fractional Order, Nemytskij Operators, and Nonlinear Partial Differential Equations, Berlin, New York: De Gruyter, 2011. https://doi.org/10.1515/9783110812411

They do the case when the singularity is at $x=0$ and not at $x=1$, but you can reduce to that case with a change of variables.


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.