Zulip Chat Archive

Stream: Is there code for X?

Topic: Holder Spaces


Yongxi Lin (Aaron) (Aug 11 2025 at 12:05):

Let URnU\subset \mathbb{R}^n be an open set. Let kk be a positive integer and 0<γ10<\gamma\leq 1. Let Ck,γ(U)C^{k,\gamma}(U) be the set of k times continuously differentiable functions such that DkuD^ku is Hölder continuous with exponent γ\gamma defined on UU. I didn't see a proof that Ck,γ(U)C^{k,\gamma}(U) forms a Banach space in mathlib. I want to make a PR about this, and I want to confirm nobody is formalizing this right now before I proceed.

Yongxi Lin (Aaron) (Aug 11 2025 at 12:07):

The proof relies on the Arzela-Ascoli theorem, which we do have in Mathlib: https://leanprover-community.github.io/mathlib4_docs/Mathlib/Topology/ContinuousMap/Bounded/ArzelaAscoli.html#BoundedContinuousFunction.arzela_ascoli

Michael Rothgang (Aug 11 2025 at 12:17):

@Yury G. Kudryashov and I started formalising Moreira's generalisation of the Morse-Sard theorem. The analytical set-up is about Ck+(α)C^{k+(\alpha)} functions, for kk an integer and α(0,1]\alpha\in(0,1] (English definition, Lean version). This generalises Ck,αC^{k,\alpha} maps --- so it would be nice to prove e.g. most estimates about Holder functions for this more general class.

Michael Rothgang (Aug 11 2025 at 12:18):

The blueprint I linked above has proof sketches for all of them; if you want to prove the inverse function theorem, you use the Faa die Bruno formula (which is already in mathlib).

Michael Rothgang (Aug 11 2025 at 12:18):

So, it could be nice to keep this more general set-up in mind when proving C^k,\alpha is Banach.

Michael Rothgang (Aug 11 2025 at 12:20):

At the same time, Yury and I got both too busy to finish the project for now, so it'll take some time to get completed. (I still care about finishing it! I just can't be everywhere at once.) So, you shouldn't wait on that project --- but being compatible with these changes would be nice.

Yongxi Lin (Aaron) (Aug 11 2025 at 12:57):

@Michael Rothgang Thank you for your comments! I would definitely consider this generalization. Just to check I understand the definition of Ck+(α)C^{k+(\alpha)} correctly: a Ck,αC^{k,\alpha} function on UU is Ck+(α)C^{k+(\alpha)} on the pair (U,U)(U,U) (Edit: never mind I think you mentioned this in Lemma 5).

Wuyang (Aug 11 2025 at 19:09):

Proving Ck,γ(U)C^{k, \gamma}(U) is Banach would be a great addition to mathlib, especially if it aligns with the Ck+(α)C^{k+(\alpha)} framework for future compatibility.

You might also explore related formalizations in LeanFinder (www.leanfinder.org) to build on existing work.

Wuyang (Aug 11 2025 at 19:09):

@leanfinder Banach space proof for Hölder space $C^{k, \gamma}(U)$ and relation to $C^{k+(\alpha)}$ generalization in Lean mathlib.

leanfinder (Aug 11 2025 at 19:09):

Here’s what I found:

  1. The function f:EFf : E \to F is Cn+1C^{n+1} within ss at xx.
  2. There exists a neighborhood uu of xx in s{x}s \cup \{x\} such that:

  3. gg is Hölder continuous on YY with constant CgC_g and exponent rgr_g,

  4. ff is Hölder continuous on ss with constant CfC_f and exponent rfr_f,

then the composition gfg \circ f is Hölder continuous on ss with constant CgCfrgC_g \cdot C_f^{r_g} and exponent rgrfr_g \cdot r_f." (score: 0.662)

  1. The function f:EFf : E \to F is Cn+1C^{n+1} within ss at xx.
  2. There exists a neighborhood uu of xx in s{x}s \cup \{x\} such that:

  3. gg is Hölder continuous on tt with constant CgC_g and exponent rgr_g

  4. ff is Hölder continuous on all of XX with constant CfC_f and exponent rfr_f
  5. ff maps into tt (i.e., f(x)tf(x) \in t for all xXx \in X)

then the composition gfg \circ f is Hölder continuous on XX with constant CgCfrgC_g \cdot C_f^{r_g} and exponent rgrfr_g \cdot r_f." (score: 0.661)

  1. gg is Hölder continuous on tt with constant CgC_g and exponent rgr_g
  2. ff is Hölder continuous on ss with constant CfC_f and exponent rfr_f
  3. ff maps ss into tt (f(s)tf(s) \subseteq t)

then the composition gfg \circ f is Hölder co
[message truncated]


Last updated: Dec 20 2025 at 21:32 UTC