Uniform Convergence · aiwiki.page

Uniform convergence is a form of convergence for sequences of functions in mathematical analysis. A sequence converges uniformly when its values approach a limiting function with an error that becomes arbitrarily small throughout the entire domain simultaneously. Unlike pointwise convergence, it requires a single index threshold that works at every point. This stronger condition supports important theorems concerning continuity, integration, and differentiation of limiting functions. (jirka.org)

Definition and quantifiers

Let $E$ be a nonempty set, and let $f_n:E\to\mathbb R$ be functions taking values in the real numbers. The sequence $(f_n)$ converges uniformly to $f:E\to\mathbb R$ if

\forall\varepsilon>0\;\exists N\in\mathbb N\; \forall n\ge N\;\forall x\in E,\qquad |f_n(x)-f(x)|<\varepsilon.

The essential requirement is that $N$ may depend on $\varepsilon$ , but not on $x$ . Pointwise convergence instead allows a separate threshold for each point. Uniform convergence therefore implies pointwise convergence, but the converse need not hold. The definition also applies to complex-valued functions, using the complex modulus. (jirka.org)

More generally, for functions into a metric space $(Y,d)$ , the error is measured by $d(f_n(x),f(x))$ . Thus the idea concerns uniform control of distances, rather than any particular algebraic structure of the values. (jirka.org)

Supremum formulation and examples

Define the maximum possible error, expressed through a supremum, by

e_n=\sup_{x\in E}|f_n(x)-f(x)|.

Uniform convergence is equivalent to $e_n\to0$ . The supremum need not be attained, and it can initially be infinite. For bounded functions, this formulation is convergence in the supremum norm, also called the uniform or infinity norm:

\|g\|_\infty=\sup_{x\in E}|g(x)|,\qquad \|f_n-f\|_\infty\longrightarrow0.

This replaces a family of point-dependent errors with one numerical error for each index. (jirka.org)

For example, the polynomials $f_n(x)=x^n$ converge pointwise on $[0,1]$ to

f(x)= \begin{cases} 0,&0\le x<1,\\ 1,&x=1. \end{cases}

Nevertheless, $\sup_{x\in[0,1]}|f_n(x)-f(x)|=1$ for every $n$ , so convergence is not uniform. On each smaller interval $[0,r]$ , where $0<r<1$ , the error is $r^n$ , which tends to zero. Uniform convergence therefore depends on the chosen domain. (jirka.org)

Uniform convergence does not require the functions themselves to be bounded: $f_n(x)=x+1/n$ converges uniformly to $x$ on $\mathbb R$ , because the error is exactly $1/n$ . This follows directly from the supremum criterion. (jirka.org)

Cauchy criterion and function spaces

For real- or complex-valued functions, uniform convergence is equivalent to the uniform Cauchy condition:

\forall\varepsilon>0\;\exists N\; \forall m,n\ge N\;\forall x\in E,\qquad |f_n(x)-f_m(x)|<\varepsilon.

This criterion establishes convergence without first identifying the limiting function. For metric-valued functions, sufficiency holds when the target is a complete metric space: completeness supplies pointwise limits, and the uniform Cauchy bound controls their errors simultaneously. (jirka.org)

In functional analysis, the bounded continuous scalar-valued functions on a topological space, equipped with the supremum norm, form a Banach space. Completeness follows by constructing pointwise limits of uniformly Cauchy sequences and showing that the resulting uniform limit remains continuous. (jirka.org)

Preservation of continuity and integration

The uniform limit of continuous functions is continuous. Its standard proof separates the difference $f(x)-f(x_0)$ into two approximation errors and one difference involving a fixed continuous function, then applies the triangle inequality. Compactness of the domain is not required. (math.ucdavis.edu)

Uniform convergence also permits an exchange of a limit and an integral on a bounded interval. If each $f_n$ is Riemann integrable on $[a,b]$ and $f_n\to f$ uniformly, then $f$ is Riemann integrable and

\int_a^b f(x)\,dx =\lim_{n\to\infty}\int_a^b f_n(x)\,dx.

The quantitative estimate is

\left|\int_a^b f_n-\int_a^b f\right| \le(b-a)\|f_n-f\|_\infty.

The finite interval length matters; uniform convergence alone does not justify an analogous assertion for arbitrary improper integrals. (jirilebl.github.io)

Differentiation requires additional conditions

Uniform convergence alone neither preserves differentiability nor guarantees convergence of derivatives. For instance,

f_n(x)=\sqrt{x^2+n^{-2}}

consists of differentiable functions and converges uniformly to $|x|$ , since the error is at most $1/n$ . The limit is not differentiable at zero. (jirilebl.github.io)

A sufficient theorem uses uniform convergence of the derivatives instead. If $f_n$ are continuously differentiable on a bounded closed interval, $f_n'\to g$ uniformly, and $f_n(x_0)$ converges at one point, then $f_n$ converges uniformly to a differentiable function $f$ with $f'=g$ . The fundamental theorem of calculus connects these conclusions. (users.math.msu.edu)

Series and compact convergence

A series $\sum u_n(x)$ converges uniformly when its partial sums do. The Weierstrass M-test supplies a useful sufficient condition: if $|u_n(x)|\le M_n$ everywhere and $\sum M_n<\infty$ , then the function series converges uniformly and absolutely. (math.ucdavis.edu)

A power series converges uniformly on every closed interval lying strictly inside its convergence interval, although not necessarily on the entire open interval. This is an example of locally uniform convergence, which, on real or complex open domains, is equivalent to uniform convergence on every compact subset. It retains local continuity properties without requiring one global error bound. (jirilebl.github.io)