Let $U$ be an open set in $\mathbb{R}^d$. A continuous function $f\colon U \to \mathbb{R}$ is strongly nowhere differentiable if and only if for each $γ\in(0,1]$ and for each unit speed $C^{1,γ}$ curve $c\colon [a,b] \to U$, the composition $f\circ c \colon [a,b] \to \mathbb{R}$ is nowhere differentiable on $(a,b)$. For bounded $U$, let $\overline U$ be the closure of $U$ and $C(\overline U)$ be the Banach space of continuous real-valued functions on $\overline U$ with the sup norm. Theorem. In the sense of the Baire category theorem, almost every $f\in C(\overline U)$ is strongly nowhere differentiable on $U$.
- Paper ID: 2510.13061
- Title: Continuous Nowhere Differentiable Multivariate Functions
- Authors: Maria Girardi, Ralph Howard
- Classification: math.CA (Classical Analysis), math.FA (Functional Analysis)
- Publication Date: October 15, 2025 (arXiv preprint)
- Paper Link: https://arxiv.org/abs/2510.13061v1
Let U be an open set in Rd. A continuous function f:U→R is strongly nowhere differentiable if and only if for every γ∈(0,1] and every unit-speed C1,γ curve c:[a,b]→U, the composition f∘c:[a,b]→R is nowhere differentiable on (a,b). For bounded U, let U denote the closure of U and C(U) denote the Banach space of continuous real-valued functions on U (equipped with the supremum norm). Main Theorem: In the sense of Baire category theorem, almost every function f in C(U) is strongly nowhere differentiable on U.
- Problem to be Addressed: This paper investigates the existence and universality of multivariate continuous nowhere differentiable functions. In the univariate case, it is known that continuous nowhere differentiable functions not only exist but are "typical" in a certain sense. This paper extends this result to the multivariate setting.
- Importance of the Problem:
- From the perspective of mathematical analysis, this is an important generalization of classical results
- It reveals deep structural properties of multivariate function spaces
- It provides new perspectives for understanding differentiability of functions
- Limitations of Existing Methods:
- The univariate result (Banach-Mazurkiewicz theorem) cannot be directly extended to the multivariate case
- The concept of differentiability for multivariate functions is more complex, requiring new characterization methods
- Simple component-wise constructions (such as f(x1,…,xd)=g1(x1)+⋯+gd(xd)), while nowhere differentiable, may be differentiable everywhere on certain submanifolds
- Research Motivation: To establish the concept of "strongly nowhere differentiable" in the multivariate setting and prove the universality of such functions in function spaces.
- Introduction of the "Strongly Nowhere Differentiable" Concept: Characterization of multivariate function non-differentiability through test curves
- Proof of the Main Theorem: In the Baire category sense, almost all continuous functions are strongly nowhere differentiable
- Construction of Auxiliary Functions: Establishment of the existence of "exactly C0,α" functions
- Provision of a Complete Theoretical Framework: Including rigorous definitions of core concepts such as test curves and strong nowhere differentiability
Study the differentiability properties of multivariate continuous functions f:U→R (where U⊆Rd is an open set), particularly establishing the concept of "strongly nowhere differentiable" and proving its universality.
- Test Curve:
- C1 function c:[a,b]→Rd
- Unit speed: ∥c′(s)∥=1 for all s∈[a,b]
- C1,γ condition: there exists a constant ρ such that ∥c′(s2)−c′(s1)∥≤ρ∣s2−s1∣γ
- Strong Nowhere Differentiability:
A function f:U→R is strongly nowhere differentiable if and only if for every test curve c:[a,b]→U, the composition f∘c is nowhere differentiable on (a,b).
- Exactly C0,α Function:
A function f is exactly C0,α at point x0 if:
limsupx→x0∣x−x0∣α∣f(x)−f(x0)∣<∞
and for any β>α:
limsupx→x0∣x−x0∣β∣f(x)−f(x0)∣=∞
- Auxiliary Function Construction:
- Construction of exactly C0,α functions fj:R→R
- Using the Katzourakis sawtooth function method:
Φ(x)=∑k=0∞b−kαϕ(bkx)
where ϕ(x)=dist(x,2Z) is the sawtooth function
- Multivariate Function Construction:
For different α1,…,αd∈(1+γ1,1), define:
f(x1,…,xd)=∑j=1dfj(xj)
- Test Curve Method: Characterization of multivariate function differentiability through behavior restricted to special curves
- Precise Control of Hölder Continuity: Utilization of exactly C0,α conditions to ensure non-differentiability
- Compactness Argument: Application of the Arzelà-Ascoli theorem to establish compactness of test curve sets
- Baire Category Method: Transformation of the problem into proving that the set of "bad" functions is nowhere dense
This is a pure theoretical mathematics paper with no numerical experiments. Results are established through rigorous mathematical proofs.
- Auxiliary Lemmas:
- Proof of existence of exactly C0,α functions (Theorem 2.3)
- Establishment of properties of composite functions (Lemma 2.6)
- Proof that sums preserve exactly C0,α properties (Lemma 2.5)
- Compactness Results:
- Proof of compactness of test curve set Cnγ(U) (Lemma 3.2)
- Nowhere Dense Properties:
- Proof that the set of "bad" functions Fnγ(U) is nowhere dense (Lemma 3.4)
Let U be a bounded open set in Rd and C(U) be the Banach space of continuous functions on U. Then the set of strongly nowhere differentiable functions in C(U) contains a dense Gδ subset.
- Theorem 2.7: For γ∈(0,1] and α1,…,αd∈(1+γ1,1), the constructed function f(x1,…,xd)=∑j=1dfj(xj) satisfies: for any C1,γ test curve c, f∘c is nowhere differentiable.
- Lemma 3.4: The set Fnγ(U) (functions satisfying certain Lipschitz conditions) is a closed nowhere dense subset.
The proof employs the classical Baire category theorem method:
- Express the set of functions with "bad" properties as a countable union of closed nowhere dense sets
- Apply the Baire category theorem; its complement is a dense Gδ set
- Prove through compactness arguments and proof by contradiction that each set is indeed nowhere dense
- 19th Century Pioneering Work:
- Riemann's example: R(x)=∑n=1∞n−2sin(πn2x)
- Weierstrass function: W(x)=∑n=1∞ancos(bnπx)
- 20th Century Breakthrough (1930s):
- Banach and Mazurkiewicz proved the universality of nowhere differentiable functions in the univariate case
- Modern Developments:
- Hardy's precise analysis of the Weierstrass function
- Berezhnŏi and Katzourakis's construction of exactly Hölder functions
- Conceptual Innovation: First introduction of the "strongly nowhere differentiable" concept
- Methodological Innovation: Use of test curve method to handle the multivariate case
- Technical Innovation: Combination of exactly Hölder continuity and Baire category method
- Existence: Strongly nowhere differentiable multivariate continuous functions exist
- Universality: In the Baire category sense, such functions are "typical"
- Characterization: Multivariate function non-differentiability can be effectively characterized through test curves
- Test Curve Restrictions: Requirement that curves be C1,γ prevents handling of all C1 curves
- Technical Assumptions: The technical condition α>1+γ1 is required
- Construction Method: Concrete construction depends on the special separated-variables form
The paper proposes two important open problems:
- Conjecture 4.1: Does there exist a continuous function whose restriction to any C1 unit-speed curve is nowhere differentiable?
- Conjecture 4.2: Are such functions typical in the Baire category sense?
- Significant Theoretical Contribution: First extension of the classical Banach-Mazurkiewicz theorem to the multivariate case
- Methodological Innovation: The test curve method provides a new tool for studying multivariate functions
- Rigorous Proof: Technical treatment is meticulous and logic is clear
- Clear Concepts: The definition of strong nowhere differentiability is natural and meaningful
- Technical Limitations: The C1,γ requirement on test curves limits the generality of results
- Special Construction: Concrete examples depend on the special separated-variables form
- Open Problems: The most natural conjecture (regarding all C1 curves) remains unsolved
- Theoretical Value: Provides important structural results for multivariate real analysis
- Methodological Contribution: The test curve method may have applications in other problems
- Inspirational Value: Provides clear directions for further research
- Mathematical Analysis Research: Provides tools for understanding function space structures
- Geometric Analysis: May have applications in studying function properties on manifolds
- Harmonic Analysis: Related to function regularity theory
In proving properties of exactly C0,α functions, the key estimate is:
∣f(c(s))−f(c(0))∣≤K∣u(s)−u(0)∣α≤Kρα∣s∣α(1+γ)
where the condition α(1+γ)>1 ensures differentiability.
Using the Arzelà-Ascoli theorem, the key is to prove:
- Uniform boundedness of the test curve set
- Equicontinuity of derivatives (via Hölder condition)
This transforms the infinite-dimensional problem into a finite-dimensional compactness problem.
This paper makes important contributions to the field of multivariate real analysis. While there is room for technical improvement, its core ideas and methods possess significant theoretical value and inspirational significance.