JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Uniform Convergence and the Sup Norm

Demand one N that works at every point at once. That single change of quantifier order gives uniform convergence, which we recast cleanly as convergence to zero in the sup norm.

Swap the order of the quantifiers

Pointwise convergence puts the for all x outside the there exists N: for each x, some N. Uniform convergence pulls the *for all x* inside, after the N is chosen. The definition reads: for every epsilon > 0 there exists N such that for all n >= N and all x in E, |f_n(x) - f(x)| < epsilon. One N must service the entire domain simultaneously.

Repackage with the sup norm

The clause *for all x, |f_n(x) - f(x)| < epsilon* is exactly a bound on the worst point. Define the sup norm ||g|| = sup over x in E of |g(x)|, the largest height the function reaches (using the supremum so it is defined even when no maximum is attained). Then uniform convergence is the crisp statement ||f_n - f|| -> 0: a single number sequence whose convergence to zero we already understand.

  1. Find the candidate limit f by computing the pointwise limit lim f_n(x) at each x.
  2. Form the error function e_n(x) = f_n(x) - f(x).
  3. Compute M_n = sup over x of |e_n(x)| — often by calculus, locating where the error is largest.
  4. Convergence is uniform iff M_n -> 0. If M_n stays bounded away from 0, convergence is merely pointwise.

Worked: a bump that fails to vanish

Compare two sequences with the same pointwise limit 0 on [0,1]. The recipe gives a one-line verdict. The first is x^n from the previous guide; the second is a moving spike. The estimate M_n decides each case.

Sequence A:  f_n(x) = x^n on [0,1).  Pointwise limit f = 0.
  M_n = sup_{0<=x<1} |x^n - 0| = sup x^n = 1   (approached as x -> 1).
  M_n = 1 does NOT go to 0  =>  NOT uniform.  (matches Guide 1)

Sequence B:  g_n(x) = n x (1 - x)^n on [0,1].  Pointwise limit g = 0
  (for fixed x in (0,1], (1-x)^n -> 0 beats the factor n; g_n(0)=0).
  Maximize: g_n'(x) = 0 gives the peak near x = 1/(n+1),
  height  g_n(1/(n+1)) = n * (1/(n+1)) * (n/(n+1))^n
                       ~ (n/(n+1)) * e^{-1}  ->  1/e  ~ 0.368.
  M_n -> 1/e  != 0  =>  NOT uniform: a bump of fixed height 1/e
  slides toward 0 but never flattens.

Sequence C:  h_n(x) = x/n on [0,1].  Pointwise limit 0.
  M_n = sup |x/n| = 1/n -> 0  =>  UNIFORM.
M_n -> 0 is uniform; a peak of fixed height is not.