🧮 Brain Teaser

The Uniform Limit of Differentiable Functions Need Not Be Differentiable — But Its Integral Remembers

Let fn:[0,1]→Rf_n : [0,1] \to \mathbb{R} be a sequence of continuously differentiable functions converging uniformly to ff.

(a) Must ff be differentiable? Give a brief justification or counterexample.

(b) Now suppose additionally that fn′f_n' converges uniformly to some function gg. Prove that ff is differentiable and f′=gf' = g.

The proof of (b) hides a beautiful trick: instead of differentiating, integrate.

uniform convergencedifferentiationFTCreal analysislimit interchange

Answer: Uniform Limit of Derivatives via Integration

Key Idea / Intuition

Uniform convergence alone is not enough to swap limits and derivatives — differentiability is a local property that requires controlling difference quotients, and uniform convergence only controls function values. However, integration is much more forgiving: uniform convergence does pass through integrals, and since a function is recovered from its derivative by the Fundamental Theorem of Calculus, we can bootstrap this to conclude differentiability when the derivatives also converge uniformly.


Formal Proof / Solution

Part (a): Counterexample

Uniform convergence does not preserve differentiability.

Counterexample: Let fn(x)=x2+1n.f_n(x) = \sqrt{x^2 + \tfrac{1}{n}}. Each fnf_n is smooth on [0,1][0,1], and ∣fn(x)−∣x∣∣=x2+1n−∣x∣=1/nx2+1/n+∣x∣≤1/n1/n=1n→0|f_n(x) - |x|| = \sqrt{x^2 + \tfrac{1}{n}} - |x| = \frac{1/n}{\sqrt{x^2+1/n}+|x|} \leq \frac{1/n}{\sqrt{1/n}} = \frac{1}{\sqrt{n}} \to 0 uniformly. So fn→f=∣⋅∣f_n \to f = |\cdot| uniformly on [0,1][0,1], but f(x)=xf(x) = x is not differentiable at x=0x = 0.


Part (b): Uniform convergence of derivatives implies differentiability of the limit

Setup. We know:

  • fn⇉ff_n \rightrightarrows f uniformly on [0,1][0,1],
  • fn′⇉gf_n' \rightrightarrows g uniformly on [0,1][0,1].

Step 1: Integrate fn′f_n'.

By the Fundamental Theorem of Calculus, for each nn and each x∈[0,1]x \in [0,1]: fn(x)−fn(0)=∫0xfn′(t) dt.f_n(x) - f_n(0) = \int_0^x f_n'(t)\, dt.

Step 2: Pass to the limit using uniform convergence.

Since fn′⇉gf_n' \rightrightarrows g uniformly, we can pass the limit inside the integral: ∫0xfn′(t) dt→∫0xg(t) dtuniformly in x.\int_0^x f_n'(t)\, dt \to \int_0^x g(t)\, dt \quad \text{uniformly in } x.

Since fn(x)→f(x)f_n(x) \to f(x) and fn(0)→f(0)f_n(0) \to f(0), taking n→∞n \to \infty gives: f(x)−f(0)=∫0xg(t) dt.f(x) - f(0) = \int_0^x g(t)\, dt.

Step 3: Conclude differentiability.

The right-hand side is the integral of a continuous function gg (uniform limit of continuous functions is continuous). By the Fundamental Theorem of Calculus: f′(x)=g(x)for all x∈[0,1].f'(x) = g(x) \quad \text{for all } x \in [0,1].

Thus ff is (continuously) differentiable and f′=gf' = g. ■\blacksquare


Summary of the Trick

| Question | Answer | |---|---| | fn⇉ff_n \rightrightarrows f implies ff differentiable? | No — counterexample: x2+1/n→∣x∣\sqrt{x^2+1/n} \to |x| | | fn⇉ff_n \rightrightarrows f and fn′⇉gf_n' \rightrightarrows g implies f′=gf' = g? | Yes — integrate fn′f_n', swap limit and integral, differentiate |

The moral: to prove a derivative statement, integrate first, then differentiate. This is a recurring theme in analysis — integration is the robust operation that tolerates limit-swapping, while differentiation is fragile.

Type: analysisEdit on GitHub ↗