Equicontinuity and the Arzelà–Ascoli theorem

Continuity allows each function to demand its own neighborhood size near a point. Equicontinuity requires a neighborhood size that works for every member of the family at that point. On a compact metric domain, this pointwise family condition implies a common uniform version across the domain. The radius can then depend on the output tolerance but not on the function or the location. A shared Lipschitz bound is a convenient stronger hypothesis: divide the desired value tolerance by the common slope bound.

Individual continuity is far weaker. The continuous functions sin(nx) remain uniformly bounded, yet near zero they can change by one over arbitrarily small distances as n increases. No common continuity radius works. Likewise, xⁿ on the unit interval develops a steep boundary layer near one. These examples explain why amplitude control alone cannot yield uniform subsequential compactness: an infinite sequence can continually spend its bounded amplitude on finer spatial features.

For real-valued continuous functions on a compact metric space, a family is relatively compact in the supremum norm exactly when it is equicontinuous and uniformly bounded. Relative compactness means its closure is compact: a sequence drawn from the family has a uniformly convergent subsequence, but the limit need not belong to the original family unless it is closed. Pointwise boundedness can replace uniform boundedness in the equicontinuous compact-domain formulation because finite covering then supplies a common bound.

Keep the domain and codomain qualifications visible. On a noncompact domain, features can drift toward infinity and defeat a global uniform subsequence while every bounded region behaves well. For functions valued in an infinite-dimensional normed space, mere boundedness of values is not the same as pointwise relative compactness; the scalar theorem cannot be transferred unchanged. In this lesson the values are real and the domain compact, so bounded scalar sequences and finite nets provide the necessary ingredients.

Choose a countable dense set of points in the compact metric domain. Boundedness at the first point gives a subsequence whose values converge there; repeat at each new point and select a diagonal subsequence converging at every chosen point. Pointwise convergence on a dense set is not enough by itself. Equicontinuity fills the gap: a finite fine net controls the difference between each arbitrary point and a nearby sampled point simultaneously for all functions.

This turns convergence on finitely many net points into a uniform Cauchy estimate. Completeness of the continuous-function space then supplies a uniform continuous limit. In the quadratic lab, restrict a to a bounded interval and inputs to a fixed bounded interval. Both heights and slopes have common bounds, making this family equicontinuous. Predict these bounds from a and the domain before moving the controls; the diagram illustrates the hypotheses but does not perform diagonal extraction for arbitrary function families.