betaFeedback

WorksMathematics

The Assouad Dimension of Graphs: An Affirmative Answer to Fraser's Question

Abstract

This article primarily investigates the Assouad dimension of graphs of continuous functions. Initially, the Assouad spectrum of the graph of a Hölder and count-Hölder continuous function has been derived. It has been shown that the Assouad dimension of the graph of a continuous function f:[a,b]→Rf:[a,b]\to \mathbb{R} with the count-Hölder condition is at least 1+sup⁡{dim⁡‾BLf(y):y∈Rf[a,b]} 1 + \sup \left\{ \overline{\dim}_{\mathrm{B}}L_f(y) : y \in \mathscr{R}_f[a,b]\right\} where dim⁡‾B\overline{\dim}_{\mathrm{B}} denotes the upper box dimension, Lf(y)L_f(y) is the level set of ff at level yy, and Rf[a,b]\mathscr{R}_f[a,b] is the range of ff on [a,b][a,b]. As a consequence, non-trivial lower bounds are obtained for the Assouad dimension of the graphs of classical functions such as the Weierstrass and Takagi functions. We have also employed a unified projection solely in the direction of the yy-axis to establish lower and upper bounds for the Assouad dimension. On this basis, we have shown that the lower bound is sharp in terms of level sets, namely the upper box dimension cannot be replaced by the Assouad dimension. This provides a partial positive answer to Question 17.11.1 posed by Fraser in~\cite{Fraser2021}, and to Question 2.2 posed by Yu in~\cite{Yu2020}. Building on this result, the analysis has been further extended to a broader class of Weierstrass-type functions and some self-affine functions like fractal interpolation functions. Moreover, we have discussed certain circumstances under which the Assouad dimension can reach 2 with several examples constructed. A supplementary explanation of why Brownian motion has the full Assouad dimension has also been provided based on the preceding work of Howroyd and Yu \cite{Howroyd2019}.

Introduction and main results

A significant area of investigation concerns the relationship between the analytic regularity of a continuous function f:[a,b]→Rf:[a,b]\to\mathbb{R} and the geometric regularity exhibited by its graph:

Gr⁡(f,[a,b]):={(t,f(t))∈R2:t∈[a,b]}.\operatorname{Gr}(f,[a,b]) := \left\{(t,f(t))\in\mathbb{R}^{2}:t\in[a,b]\right\}.

In particular, one often examines how assumptions regarding the smoothness or regularity of ff impact various notions of fractal complexity associated with Gr⁡(f,[a,b])\operatorname{Gr}(f,[a,b]), such as different metric dimensions.

A classical and illustrative example is the Weierstrass function Wλ,β:R→RW_{\lambda,\beta}:\mathbb{R}\to\mathbb{R} given by

Wλ,β(x)=∑n=0∞λncos⁡(2πβnx),(1)W_{\lambda,\beta}(x)=\sum_{n=0}^{\infty}\lambda^{n}\cos(2\pi\beta^{n}x), \tag*{(1)}

where β>1\beta>1 and λ∈(1β,1)\lambda\in\left(\frac{1}{\beta},1\right). This function has been a subject of ongoing interest in the fractal geometry community, which is continuous everywhere but differentiable nowhere. This counterintuitive behavior challenges conventional concepts of differentiability and regularity in analysis. Historically important, the Weierstrass function was among the first known examples to exhibit self-affine and scale-invariant irregularity. Its graph, exhibiting intricate details at every scale, serves as a model for exploring geometric complexity. Notably, its graph has inspired extensive research aimed at determining various fractal dimensions such as the Hausdorff, packing and box dimensions. While the upper box dimension of Gr⁡(Wλ,β,[a,b])\operatorname{Gr}(W_{\lambda,\beta},[a,b]) has been investigated since the early 20th century [21], the precise Hausdorff dimension was only rigorously established in 2018 by Shen [30], who showed that Gr⁡(Wλ,β,[a,b])\operatorname{Gr}(W_{\lambda,\beta},[a,b]) has the identical Hausdorff and box dimensions 2+log⁡λlog⁡β2+\frac{\log\lambda}{\log\beta} provided β\beta is an integer no less than 2. More literature on its research course can be found in [9, 23, 24, 29].

Another fundamental example of continuous but nowhere differentiable functions is the Takagi function, which also has emerged as a significant object in fractal geometry due to its inherent self-similarity and irregularity. It serves as a canonical model for constructing and analyzing fractal sets and curves, offering valuable insights into dimension theory and the complex geometry of function graphs. Owing to its recursive structure and scaling behavior, the Takagi function also plays a key role in the study of multifractal spectra, singular measures, and functions with prescribed local oscillations, rendering it central to various aspects of the modern fractal analysis. The Takagi function is defined by

Tλ,β(x)=∑n=0∞λndist⁡(βnx,Z),(2)T_{\lambda,\beta}(x)=\sum_{n=0}^{\infty}\lambda^n\operatorname{dist}\left(\beta^n x,\mathbb{Z}\right), \tag*{(2)}

where β>1\beta>1, λ∈[1β,1)\lambda\in\left[\frac{1}{\beta},1\right), and dist⁡(t,Z)\operatorname{dist}(t,\mathbb{Z}) denotes the distance from t∈Rt\in\mathbb{R} to the nearest integer, which is the tent map with period 11 and defined on [0,1][0,1] as

dist⁡(t,Z)={t,t∈[0,12],1−t,t∈[12,1].\operatorname{dist}(t,\mathbb{Z})= \begin{cases} t, & t\in\left[0,\frac{1}{2}\right],\\ 1-t, & t\in\left[\frac{1}{2},1\right]. \end{cases}

Historically, the term “Takagi function” referred to the specific case T12,2T_{\frac{1}{2},2}, but it is now commonly used to refer more generally to functions of the form Tλ,βT_{\lambda,\beta}. There has been significant interest in investigating the Hausdorff and box dimensions of the graphs of such functions. For the box dimension, it is known from [8 Theorem 2.4] and [26 Section 2] that the upper box dimension of the graph of Tλ,βT_{\lambda,\beta} is 2+log⁡λlog⁡β2+\frac{\log\lambda}{\log\beta}. In contrast, determining the Hausdorff dimension of the graph is more challenging. For recent developments and open problems in this area, we refer to [7, 30] and the references therein.

In Figure 1, (A) and (B) display the graphs of the Weierstrass and Takagi functions on [0,1][0,1] with some suitable parameters λ\lambda and β\beta.

Graphs of the Weierstrass and Takagi functions

Figure 1. (A) The graph of W12,5(x)W_{\frac{1}{2},5}(x) on [0,1][0,1]. (B) The graph of T23,2(x)T_{\frac{2}{3},2}(x) on [0,1][0,1]. Graphs of the Weierstrass and Takagi functions.

This work focuses on two advanced notions of fractal dimensions: the Assouad dimension, denoted dim⁡A\dim_{\mathrm{A}}, introduced by Assouad in [4], and the Assouad spectrum, denoted dim⁡Aθ\dim_{\mathrm{A}}^{\theta}, developed by Fraser and Yu in [19]. Both concepts describe the scaling behavior of coverings across a space, with the spectrum refining the Assouad dimension by incorporating a parameter θ∈(0,1)\theta\in(0,1) that reflects the relationship between fine and coarse scales. In recent years, an increasing number of scholars have been attracted to the research on the Assouad dimension of certain special sets, such as popcorn graphs [10], self-affine sets [17], Kakeya sets [32] and so on.

One outstanding problem, posed by Fraser as Question 17.11.1 in [15], concerns the exact value of the Assouad dimension for the graphs of classical functions such as the Weierstrass and Takagi functions. Up to now, this question still remains open. To our knowledge, the only partial progress regarding the Weierstrass function is due to Chrontsios-Garitsis and Tyson [11], who established an upper bound for its Assouad spectrum. However, for the Takagi-type function, several authors have contributed estimates and partial results on the Assouad dimension; see, for instance, Yu [34], Anttila, Bárány, and Käenmäki [2], as well as the work of Chrontsios-Garitsis and Tyson [11].

Motivated by these developments, we analyze how certain regularity conditions—specifically the Hölder and count-Hölder continuity (introduced later in § 2.3)—affect the lower and upper bounds for the Assouad dimension and the Assouad spectrum of Gr⁡(f,[a,b])\operatorname{Gr}(f,[a,b]) for a continuous function ff on any closed interval [a,b]⊂R[a,b] \subset\mathbb{R}. Our findings provide a constructive response to the open questions raised by Fraser in [15] and by Yu in [34], and the principal contributions of this paper are encapsulated in the upcoming main theorems.

For any continuous function f ⁣:[a,b]→Rf\colon[a,b] \to\mathbb{R}, our first main result establishes the relationship between the Assouad spectrum of its graph on [a,b][a,b], the order α\alpha corresponding to its Hölder or count-Hölder continuity, and the upper box dimension (denoted dim⁡‾B\overline{\dim}_{\mathrm{B}}) of its level set

Lf(y):={x∈[a,b]:f(x)=y}L_f(y) := \{x \in[a,b] : f(x)=y\}

at level y∈Rf[a,b]y \in\mathcal{R}_f[a,b] where Rf[a,b]\mathcal{R}_f[a,b] signifies the range of ff on [a,b][a,b], presented below.

Theorem 1.1. Let ff be a continuous function on [a,b][a,b] and α∈(0,1)\alpha\in(0,1).

(1) If ff is α\alpha-order Hölder continuous on [a,b][a,b], then

dim⁡AθGr⁡(f,[a,b])≤{1+1−α1−θ,θ∈(0,α),2,θ∈[α,1).\dim_{\mathrm{A}}^{\theta}\operatorname{Gr}(f,[a,b]) \le \begin{cases} 1+\dfrac{1-\alpha}{1-\theta}, & \theta\in(0,\alpha),\\ 2, & \theta\in[\alpha,1). \end{cases}

(2) If ff is α\alpha-order count-Hölder continuous on [a,b][a,b], then

dim⁡AθGr⁡(f,[a,b])≥{sup⁡y∈Rf[a,b]{dim⁡‾BLf(y)}+1−α1−θ,θ∈(0,α),sup⁡y∈Rf[a,b]{dim⁡‾BLf(y)}+1,θ∈[α,1).\dim_{\mathrm{A}}^{\theta}\operatorname{Gr}(f,[a,b]) \ge \begin{cases} \sup_{y\in\mathcal{R}_f[a,b]}\{\overline{\dim}_{\mathrm{B}}L_f(y)\}+\dfrac{1-\alpha}{1-\theta}, & \theta\in(0,\alpha),\\ \sup_{y\in\mathcal{R}_f[a,b]}\{\overline{\dim}_{\mathrm{B}}L_f(y)\}+1, & \theta\in[\alpha,1). \end{cases}

For the sake of intuitiveness, the Assouad spectrum of the graph of any α\alpha-order Hölder and count-Hölder continuous function has been visually shown in Figure 2 (A)–(D) where we assume the supremum of the upper box dimensions of all the level sets is 0.80.8, and α\alpha is taken as 0.20.2, 0.40.4, 0.60.6 and 0.80.8, respectively. Here, the upper line represents the upper bound of the spectrum, the lower line stands for the lower bound of the spectrum, and the shaded area between these two lines means the possible range of the spectrum. Observe that the gap between the upper and lower bounds is exactly 1−sup⁡y∈Rf[a,b]{dim⁡‾BLf(y)}1-\sup_{y\in R_f[a,b]}\{\overline{\dim}_{\mathrm{B}}L_f(y)\} (it is 0.20.2 under the above assumption).

The Assouad spectrum of the graph of an $\alpha$-order Hölder and count-Hölder continuous function for $\alpha = 0.2$, $0.4$, $0.6$, and $0.8$

Figure 2. The Assouad spectrum of the graph of any α\alpha-order Hölder and count-Hölder continuous function for some different values of α\alpha.

The next main result follows directly from Theorem 1.1, which estimates the bound for the Assouad dimension of the graph of a function with the Hölder and count-Hölder continuity.

Theorem 1.2. If ff is an α\alpha-order Hölder and count-Hölder continuous function on [a,b][a,b] for α∈(0,1)\alpha\in(0,1), then

sup⁡y∈Rf[a,b]{dim⁡‾BLf(y)}+1≤dim⁡qAGr⁡(f,[a,b])≤dim⁡AGr⁡(f,[a,b])≤2.\sup_{y\in R_f[a,b]}\left\{\overline{\dim}_{\mathrm{B}}L_f(y)\right\}+1\leq\dim_{\mathrm{qA}}\operatorname{Gr}(f,[a,b])\leq\dim_{\mathrm{A}}\operatorname{Gr}(f,[a,b])\leq2.

We emphasize that Theorem 1.2 still remains valid under the weaker assumption of count-Hölder continuity alone because dim⁡AGr⁡(f,[a,b])≤2\dim_{\mathrm{A}}\operatorname{Gr}(f,[a,b])\le2 is trivial for any continuous function f:[a,b]→Rf:[a,b]\to\mathbb{R}. In addition, Theorem 1.2 seems to provide a relatively optimal lower bound, the supremum of the upper box dimensions of all level sets plus 1, for the Assouad dimension of graphs of count-Hölder continuous functions. However, it might become harder to acquire the same upper bound.

In the present paper, we denote by (un)(u_n) the sequence of real numbers unu_n (if necessary to stress that nn ranges over a specific set such as the set of positive integers N+\mathbb{N}_+, we may also denote it as (un)N+(u_n)_{\mathbb{N}_+}). For α∈(0,1)\alpha\in(0,1), we now define a set of binary sequences by

Λα:={(rn,Rn) | 0<rn<Rn≤rnα<1, rn↘0, Rn↘0 and Rnrn↗∞ as n→∞}.\Lambda_\alpha:=\left\{(r_n,R_n)\ \middle|\ 0<r_n<R_n\le r_n^\alpha<1,\ r_n\searrow0,\ R_n\searrow0\text{ and }\frac{R_n}{r_n}\nearrow\infty\text{ as }n\to\infty\right\}.

The undermentioned theorem explicitly reveals that the lower bound given in Theorem 1.2 is sharp in terms of level sets. More precisely, the upper box dimension cannot be replaced by the Assouad dimension unless we add some strict constraints to level sets.

Theorem 1.3. Let ff be an α\alpha-order count-Hölder continuous function on [a,b][a,b] for α∈(0,1)\alpha\in(0,1). If the following set

\mathcal{Y}:=\{y\in\mathcal{R}_f[a,b]\ \middle|\ \text{there is a binary sequence }(r_n,R_n)\in\Lambda_\alpha\text{ such that,}
\text{for all }n\in\mathbb{N},\ \sup_{x\in L_f(y)}\#\{i\in\mathbb{N}\ \middle|\ \Delta_{x,n,i}\cap L_f(y)\ne\varnothing\}\ge(\frac{R_n}{r_n})^{\dim_{\mathrm{A}}L_f(y)} \}

is non-empty, then

dim⁡AGr⁡(f,[a,b])≥sup⁡y∈Y{dim⁡ALf(y)}+1.\dim_{\mathrm{A}}\operatorname{Gr}(f,[a,b])\ge\sup_{y\in\mathcal{Y}}\{\dim_{\mathrm{A}}L_f(y)\}+1.

Here Δx,n,i=[x−Rn2+(i−1)rn,x−Rn2+irn]∩[x−Rn2,x+Rn2]\Delta_{x,n,i}=\left[x-\frac{R_n}{2}+(i-1)r_n,x-\frac{R_n}{2}+ir_n\right]\cap\left[x-\frac{R_n}{2},x+\frac{R_n}{2}\right].

Very recently, one of the most impressive findings for us is a general slicing theorem for weak tangents, which is an immediate consequence of Furstenberg’s dimension conservation result [20 Theorem 1.6] and has been presented by Anttila and Rutar in [3 Proposition A] where they have also given a short elementary proof. Here we restate this result for the completeness of our discussion.

References

  1. [1]R. Anttila, Pointwise Assouad dimension for measures, Proc. Roy. Soc. Edinb. A 153 (2023), 2053–2078.DOI
  2. [2]R. Anttila, B. Bárány, and A. Kӓenmäki, Slices of the Takagi function, Ergod. Theory Dyn. Syst. 44 (2024), 2361–2398.DOI
  3. [3]R. Anttila and A. Rutar, Fibre stability for dominated self-affine sets, (2024), arXiv:2412.06579.arxiv.org/abs/2412.06579
  4. [4]P. Assouad, Plongements lipschitziens dans R^n, Bull. Soc. Math. Fr. 111 (1983), 429–448.
  5. [5]A. Banaji and J. M. Fraser, Assouad type dimensions of infinitely generated self-conformal sets, Nonlinearity 37 (2024), 045004.DOI
  6. [6]K. Barański, On the dimension of graphs of Weierstrass-type functions with rapidly growing frequencies, Nonlinearity 25 (2012), 193–209.DOI
  7. [7]K. Barański, B. Bárány, and J. Romanowska, On the dimension of the graph of the classical Weierstrass function, Adv. Math. 265 (2014), 32–59.DOI
  8. [8]K. Barański, Dimension of the Graphs of the Weierstrass-Type Functions, Cham: Springer (2015).DOI
  9. [9]A. S. Besicovitch and H. D. Ursell, Sets of fractional dimensions V: on dimensional numbers of some continuous curves, J. Lond. Math. Soc. 32 (1937), 142–153.DOI
  10. [10]H. P. Chen, J. M. Fraser, and H. Yu, Dimensions of the popcorn graph, Proc. Amer. Math. Soc. 150 (2022), 4729–4742.arxiv.org/abs/2007.08407
  11. [11]E. K. Chrοntsios-Garitsis and J. T. Tyson, On the Assouad spectrum of Hölder and Sobolev graphs, Math. Proc. Cambridge Philos. Soc. 180 (2026), 105–131.
  12. [12]K. J. Falconer, Fractal Geometry: Mathematical Foundations and Applications, 3rd ed., Chichester: John Wiley Sons Ltd. (2014).
  13. [13]T. Y. Feng and J. M. Fraser, The Assouad spectrum and dimension of typical graphs, (2026), arXiv:2601.05439.arxiv.org/abs/2601.05439
  14. [14]J. M. Fraser, Assouad type dimensions and homogeneity of fractals, Trans. Amer. Math. Soc. 366 (2014), 6687–6733.arxiv.org/abs/1301.2934
  15. [15]J. M. Fraser, Assouad Dimension and Fractal Geometry, Cambridge: Cambridge Univ. Press (2021).
  16. [16]J. M. Fraser, K. E. Hare, K. Hare, S. Troscheit, and H. Yu, The Assouad spectrum and the quasi-Assouad dimension: a tale of two spectra, Ann. Acad. Sci. Fenn. Math. 44 (2019), 379–387.arxiv.org/abs/1804.09607
  17. [17]J. M. Fraser and A. Rutar, Assouad-type dimensions of overlapping self-affine sets, Ann. Fenn. Math. 49 (2024), 3–21.DOI
  18. [18]J. M. Fraser and L. Stuart, The Assouad spectrum of Kleinian limit sets and Patterson–Sullivan measure, Geometriae Dedicata 217 (2023), 32.arxiv.org/abs/2203.04931
  19. [19]J. M. Fraser and H. Yu, New dimension spectra: finer information on scaling and homogeneity, Adv. Math. 329 (2018), 273–328.arxiv.org/abs/1610.02334
  20. [20]H. Furstenberg, Ergodic fractal measures and dimension conservation, Ergod. Theory Dyn. Syst. 28 (2008), 405–422.DOI
  21. [21]G. H. Hardy, Weierstrass's non-differentiable function, Trans. Amer. Math. Soc. 17 (1916), 301–325.DOI
  22. [22]D. C. Howroyd and H. Yu, Assouad dimension of random processes, Proc. Edinburgh Math. Soc. 62 (2019), 281–290.arxiv.org/abs/1707.02507
  23. [23]B. R. Hunt, The Hausdorff dimension of graphs of Weierstrass functions, Proc. Amer. Math. Soc. 126 (1998), 791–800.DOI
  24. [24]B. R. Hunt and J. A. Yorke, Smooth dynamics on Weierstrass nowhere differentiable curves, Trans. Amer. Math. Soc. 325 (1991), 141–154.DOI
  25. [25]L. Jiang, Assouad dimension of the Takagi function, (2025), arXiv:2502.01140.arxiv.org/abs/2502.01140
  26. [26]J. Kaplan, J. Mallet-Paret, and J. Yorke, The Lyapunov dimension of a nowhere differentiable attracting torus, Ergod. Theory Dyn. Syst. 4 (1984), 261–281.DOI
  27. [27]P. Lévy, Sur les intégrales dont les éléments sont des variables aléatoires indépendantes, Ann. Scuola. Norm.-Sci. 3 (1934), 337–366.
  28. [28]F. Lü and L. Xi, Quasi-Assouad dimension of fractals, J. Fractal Geom. 3 (2016), 187–215.DOI
  29. [29]H. J. Ren and W. X. Shen, A Dichotomy for the Weierstrass-type functions, Invent. Math. 226 (2021), 1057–1100.DOI
  30. [30]W. X. Shen, Hausdorff dimension of the graphs of the classical Weierstrass functions, Math. Z. 289 (2018), 223–266.arxiv.org/abs/1505.03986
  31. [31]S. J. Taylor, The Hausdorff α-dimensional measure of Brownian paths in n-space, Math. Proc. Cambridge Philos. Soc. 49 (1953), 31–39.DOI
  32. [32]H. Wang and J. Zahl, The Assouad dimension of Kakeya sets in R³, Invent. Math. 241 (2025), 153–206.DOI
  33. [33]B. Y. Yu and Y. S. Liang, On the Katugampola fractional integral and dimensional analysis of the fractal basin boundary for a random dynamical system, Physica D 468 (2024), 134289.DOI
  34. [34]H. Yu, Weak tangent and level sets of Takagi functions, Monatsh. Math. 192 (2020), 249–264.DOI

Paper details

vlnc:W_nnwd6ud4
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 8

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents