According to a result of Whitney any $C^r$-manifold, $r\geq 1$, is $C^r$-homeomorphic to a smooth embedded submanifold of some Euclidean space; see Theorem 1 in

Hassler Whitney, "Differentiable manifolds", Annals of Mathematics (2) 37 (1936), no. 3, pp. 645–680, JFM 62.1454.01, MR1503303, Zbl 0015.32001.

I expect that an analogous result should be true for $C^{r,\alpha}$-Hölder manifolds, $r\geq 1$, $\alpha\in[0,1]$, and I am wondering if there exists a reference for such a statement. Does perhaps anybody know such a reference?

Actually, I am not so much interested in the embedding, but in a refinement of the $C^{r,\alpha}$-Hölder structure to a smooth structure, i.e. in a $C^{r,\alpha}$-Hölder homeomorphism to a smooth manifold.

In particular, I am interested in the case in which the manifold to start with is itself a ($C^{r,\alpha}$-)embedded submanifold of some Euclidean space and in the case $\alpha=1$. Maybe it is easier to say something here?


1 Answer 1


Every $C^1$ manifold admits a compatible $C^\infty$ structure. You can find a proof in Hirsch's "Differential topology". It is actually quite easy and based on a fact that smoothing a $C^1$ diffeomorphism by convolution leads to a smooth diffeomorphism (because derivatives converge uniformly) on a slightly smaller domain.

In particular $C^{r,\alpha}$ manifolds, $r\geq 1$, admit a compatible $C^\infty$-structure and hence such a manifold can be embedded as a smooth submanifold of an Euclidean space. However, the embedding will only be $C^{r,\alpha}$ smooth since the original manifold is only $C^{r,\alpha}$.

  • $\begingroup$ The $C^{\infty}$ structure might be compatible with the $C^1$ structure, but why should it be compatible with the $C^{r,\alpha}$ structure? If it is not, the resulting embedding is not $C^{r,\alpha}$ smooth with respect to the $C^{r,\alpha}$ structure we started with, only with respect to a distinct $C^{r,\alpha}$ structure (that is induced by the $C^{\infty}$ structure). $\endgroup$ Oct 2, 2022 at 17:54
  • $\begingroup$ I have the impression that one needs e.g. a Hölder analogue of Theorem 2.7 in Hirsch's book: "Let $M$ and $N$ be $C^s$ manifolds $1\leq s \leq \infty$. Then $\mathrm{Diff}^s(M,N)$ is dense in $\mathrm{Diff}^r(M,N)$ in the strong $C^r$ topology, $1\leq r <s$." The proof of this theorem using convolutions seems to go through also in the Hölder case. Thanks for the reference. $\endgroup$ Oct 3, 2022 at 8:01

Your Answer

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

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