Weierstrass's Non-Differentiable Function
Speaker: Jeff Calder
Date: December 12, 2014
Introduction
- 19th Century Belief: Continuous functions are differentiable at a large set of points.
- Karl Weierstrass (1872): Introduced the first example of a continuous function that is nowhere differentiable.
- Function Definition:
- ( W(x) = \sum_{n=0}^{\infty} a^n \cos(b^n x) )
Theorem 1 (Weierstrass 1872)
- Conditions:
- ( a \in (0, 1) )
- ( b > 1 ) (odd integer)
- ( ab > 1 + \frac{3}{2} )
- Conclusion: The function W is continuous and nowhere differentiable on ( \mathbb{R} ).
Proof of Continuity
- Convergence:
- Uses the geometric series ( \sum a^n ) which converges for ( a \in (0, 1) ).
- The Weierstrass M-test shows uniform convergence to ( W ) on ( \mathbb{R} ).
- Continuity:
- Each ( a^n \cos(b^n x) ) is continuous; thus, ( W ) is the uniform limit of continuous functions.
Motivation for Condition (1)
- Partial Sums:
- ( W_n(x) = \sum_{k=0}^{n} a^k \cos(b^k x) )
- ( W'n(x) = \sum{k=0}^{n} (ab)^k \sin(b^k x) )
- If ( ab < 1 ), ( W ) is differentiable.
- Godfrey Hardy (1916) showed ( ab \geq 1 ) is sufficient for nowhere differentiability.
The Weierstrass Function as a Fractal
- Visual Property:
- Repeating patterns at every scale.
- Early example of a fractal.
Cosine Properties
- Identity:
- ( |\cos(x) - \cos(y)| \leq |x - y| )
- Cosine Periodicity:
- ( \cos(n + x) = \cos(x) ) if n is even.
- ( \cos(n + x) = -\cos(x) ) if n is odd.
Proof of Theorem 1
- Key Steps:
- Bound |A|: Using triangle inequality and identity for cosine.
- Derive bounds using geometric series.
- Bound |B|:
- Use properties of ( x_m ) and cosine to establish non-negativity.
- Combine Bounds:
- Ensure the difference quotient tends to infinity for ( ab > 1 + \frac{3}{2} ).
Conclusion
- Non-Differentiability:
- ( \lim_{m \to \infty} \frac{W(x_m) - W(x_0)}{x_m - x_0} = +\infty )
- No Vertical Tangent or Infinite Derivative:
- ( \lim_{x \to x_0} \frac{W(x) - W(x_0)}{x - x_0} ) does not exist.
Visual Representation:
- Figure 1: Plots of the Weierstrass function for ( a = 0.5 ) and ( b = 3 ), illustrating self-similarity across scales.