Pathological Beauty: Constructing Continuous but Nowhere Differentiable Functions: An Overview

       When most of us first meet calculus, we’re told that a function is “nice” if it’s smooth — if its graph doesn’t have corners, jumps, or wild oscillations. We imagine curves that can be followed with the stroke of a pencil: circles, parabolas, sine waves. Differentiability, the existence of a tangent line at every point, feels almost automatic once continuity is in place.

But mathematics has a way of defying intuition. In 1872, Karl Weierstrass shocked the mathematical world by constructing a function that is continuous everywhere but differentiable nowhere. A function you can draw without lifting your pencil — yet at no point can you lay down a tangent line.

These so-called pathological functions became objects of fascination. At first, they seemed like monsters, counterexamples to the clean world analysts hoped to live in. But over time, they were revealed to be not monsters at all — just misunderstood creatures. They arise naturally in probability theory (as models of Brownian motion), in dynamical systems, in fractals, and in many corners of modern mathematics.

My recent paper, Functional Equations and Combinatorial Constructions of Pathological Functions in Real Analysis, takes a fresh look at how such functions can be constructed. The goal was to find approaches that are rigorous yet elementary, requiring no heavy machinery from harmonic analysis, but still capturing the strange beauty of these wild functions.Check out complete preprint

---

Two Roads to Pathology

The paper explores two complementary approaches:

1. The Functional Equation Approach
   This begins with a generalization of the Takagi function, a classical nowhere differentiable curve built from repeating sawtooth patterns. The key insight is that these functions satisfy a simple fixed-point equation involving scaling and self-similarity.

   Using just the Banach fixed-point theorem and elementary analysis, one can show a sharp dichotomy:

   • If the scaling parameters satisfy αβ < 1, the function is not only continuous but also Lipschitz continuous (it behaves as nicely as one could hope).

   • If αβ ≥ 1, the function becomes nowhere differentiable — pathological in the strongest sense.

   What’s striking is how clean this threshold is: at exactly the point where scaling “outpaces” smoothness, differentiability collapses everywhere.

2. The Combinatorial (Binary Expansion) Approach
   The second method uses binary digits to build functions. Each number in [0,1] has a binary expansion — a string of 0s and 1s. By weighting each digit with terms like $1/n^p$ and summing them, we get continuous functions.

   The challenge lies in the arithmetic of binary expansions: when you shift a number slightly, its digits can flip in complicated ways. By carefully analyzing this, the paper shows that for any $p > 1$, these functions are continuous but nowhere differentiable.

   This construction highlights a different source of irregularity: not self-similarity, but the chaotic structure hidden inside number representations.

---

Why It Matters

Both methods avoid the sophisticated Fourier-analytic proofs required for classical Weierstrass functions. They show that pathology doesn’t need exotic tools — it can emerge from the simplest mechanisms: self-similarity and digit expansions.

The paper also touches on Hölder regularity — a way of measuring “how continuous” a function is, even when it lacks derivatives. For certain cosine-series functions, the paper recovers Hardy’s classical result: they are nowhere differentiable when parameters cross a critical threshold, but with a precise Hölder exponent describing their roughness.

---

A Personal Note

I’m 17, and this paper is part of my independent exploration into real analysis. For me, mathematics is not just about solving problems but about seeing how beauty hides in apparent chaos. Pathological functions may look like the rebels of analysis, but they carry deep lessons: that smoothness is fragile, that structure can be hidden in simplicity, and that wildness can be rigorously understood.

Working on this paper taught me two things:

1. Even classical problems can be revisited with fresh, elementary methods.

2. Research is as much about finding the right perspective as it is about grinding through proofs.

---

Where Next?

Pathological functions are just the beginning. There are natural questions waiting:

• What happens in higher dimensions — can we construct nowhere differentiable surfaces using similar methods?

• What is the fractal geometry of these graphs — their Hausdorff dimension, their measure-theoretic properties?

• What if we introduce randomness into these constructions — do we get functions resembling Brownian paths?

These questions tie into probability, geometry, and even physics. The monsters of the 19th century have become the tools of the 21st.

---

Conclusion

Continuous nowhere differentiable functions remind us that mathematics is full of surprises. A function can be smooth in one sense (continuous) yet violently irregular in another (non-differentiable). And behind these paradoxes lie simple constructions, accessible even to a high school student with curiosity and persistence.

The beauty of mathematics is that it allows us to embrace such paradoxes, not as failures of intuition, but as gateways to deeper understanding.

---

👉 That’s the story I wanted to share in my first blog post: not just the technical theorems, but the wonder of discovering order in apparent chaos.

---