Sequences and Series of Functions

Understanding how sequences and series of functions converge is not merely a technical exercise in real analysis; it is the foundational framework that justifies the operations we perform in calculus, differential equations, and Fourier analysis. When can you safely differentiate or integrate an infinite sum term-by-term? The answer hinges on the crucial distinction between pointwise convergence and the stronger, more well-behaved uniform convergence.

Pointwise Convergence: The Natural but Flawed Starting Point

A sequence of functions $\{f_n\}$, all defined on a common set $E$, is said to converge pointwise to a function $f$ if, for each fixed point $x_0$ in $E$, the sequence of numbers $\{f_n(x_0)\}$ converges to $f(x_0)$. In precise epsilon-language: For every $x \in E$ and for every $\epsilon > 0$, there exists an integer $N$ (which can depend on both $\epsilon$ and $x$) such that for all $n > N$, we have $|f_n(x) - f(x)| < \epsilon$.

Consider the classic example on the interval $E = [0,1]$: $f_n(x) = x^n$. For a fixed $x$ in $[0,1)$, $x^n \to 0$, and for $x=1$, $f_n(1)=1 \to 1$. Thus, the pointwise limit function is the discontinuous function: $$f(x) = \begin{cases} 0 & \text{if } 0 \leq x < 1 \\ 1 & \text{if } x = 1 \end{cases}$$ Herein lies the central problem: each $f_n$ is continuous on $[0,1]$, but the pointwise limit $f$ is not. Pointwise convergence is too weak to preserve continuity.

Uniform Convergence: The Key to Preserving Properties

Uniform convergence strengthens the pointwise definition by removing the dependence of $N$ on the point $x$. A sequence $\{f_n\}$ converges uniformly on $E$ to a function $f$ if: For every $\epsilon > 0$, there exists an integer $N$ (depending only on $\epsilon$) such that for all $n > N$ and for all $x \in E$, we have $|f_n(x) - f(x)| < \epsilon$.

The difference is qualitative. Pointwise convergence asks that for each $x$, the graphs of $f_n$ eventually enter an $\epsilon$-tube around $f(x)$ at that point. Uniform convergence demands that for a sufficiently large $n$, the entire graph of $f_n$ lies within a uniform, $\epsilon$-sized band surrounding the graph of $f$. Returning to our example, $f_n(x) = x^n$ does not converge uniformly to its pointwise limit on $[0,1]$. For any $\epsilon < 1$ and any large $n$, there are points near $1$ (like $x = \sqrt[n]{1/2}$) where $f_n(x) = 1/2$, which is not within $\epsilon$ of the limit value $0$.

The power of uniform convergence is captured in its fundamental preservation theorems:

1. Preservation of Continuity: If a sequence of continuous functions $\{f_n\}$ converges uniformly to $f$ on $E$, then $f$ is continuous on $E$. The proof hinges on the uniform condition allowing us to control the variation in $f_n$ across all points.
2. Interchange of Limit and Integral (for Riemann integrals): If $\{f_n\}$ converges uniformly to $f$ on a closed interval $[a, b]$, and each $f_n$ is integrable on $[a, b]$, then $f$ is integrable on $[a, b]$ and

$$\lim_{n \to \infty} \int_a^b f_n(x) \, dx = \int_a^b \lim_{n \to \infty} f_n(x) \, dx = \int_a^b f(x) \, dx.$$ This theorem legitimizes term-by-term integration of uniformly convergent series.

It is critical to note that uniform convergence does not, by itself, guarantee that derivatives are preserved. For that, one needs uniform convergence of the sequence of derivatives plus pointwise convergence of the original functions at least at one point.

Series of Functions and the Weierstrass M-Test

A series of functions $\sum_{k=1}^\infty g_k(x)$ is defined by the sequence of its partial sums $f_n(x) = \sum_{k=1}^n g_k(x)$. The series converges pointwise or uniformly precisely when the sequence of partial sums does.

The most powerful and frequently applied tool for establishing uniform convergence of a function series is the Weierstrass M-Test. It states: Suppose $\{g_k\}$ is a sequence of functions defined on $E$, and $\{M_k\}$ is a sequence of nonnegative real numbers such that $|g_k(x)| \leq M_k$ for all $x \in E$ and all $k$. If the numerical series $\sum_{k=1}^\infty M_k$ converges, then the function series $\sum_{k=1}^\infty g_k(x)$ converges absolutely and uniformly on $E$.

The logic is a direct application of the Cauchy Criterion for uniform convergence. The test is supremely useful because it reduces a question about the uniform convergence of functions to a question about the convergence of a simple, positive-term number series. For example, consider $\sum_{n=1}^\infty \frac{\cos(nx)}{n^2}$ on the entire real line. Since $|\frac{\cos(nx)}{n^2}| \leq \frac{1}{n^2}$ for all $x$, and $\sum \frac{1}{n^2}$ converges (it's a p-series with $p=2>1$), the Weierstrass M-Test with $M_n = 1/n^2$ immediately guarantees the series converges uniformly (and absolutely) on $\mathbb{R}$ to a continuous limit function.

Common Pitfalls

1. Assuming pointwise convergence implies uniform convergence: This is the most frequent conceptual error. The example $f_n(x)=x^n$ on $[0,1]$ is a canonical counterexample. Always suspect that pointwise convergence on a closed interval may fail to be uniform if the limit function has a discontinuity that the approximating functions do not.
2. Misapplying the interchange theorems: It is tempting to automatically swap limits with integrals or derivatives. Remember: integration is generally safe under uniform convergence on a bounded interval, but differentiation requires the stricter condition of uniform convergence of the derivative series. Just because $\sum f_n$ converges uniformly does not mean $\sum f_n'$ does.
3. Incorrectly applying the Weierstrass M-Test: The test provides a sufficient, but not necessary, condition for uniform convergence. A series can converge uniformly without satisfying the conditions of the M-Test (e.g., a series where the supremum of $|g_k(x)|$ is not summable). Furthermore, the bounding condition $|g_k(x)| \leq M_k$ must hold for all $x$ in the domain. Finding a bound that works on a restricted domain (like a closed interval) is common, but a series that converges uniformly on $[0,1]$ may not do so on $(0, \infty)$.
4. Confusing the modes of convergence in proof writing: When writing an epsilon-proof, carefully note the order of quantifiers. For pointwise convergence, you choose $x$ first, then $\epsilon$, then find $N(x, \epsilon)$. For uniform convergence, you choose $\epsilon$ first, then find an $N(\epsilon)$ that works globally. Mixing up this logical structure will lead to an invalid argument.

Summary

• Pointwise convergence is defined point-by-point and is too weak to preserve fundamental analytical properties like continuity. The limiting function of continuous functions can be discontinuous.
• Uniform convergence, a global condition requiring the entire function graph to approach the limit within a uniform band, is strong enough to preserve continuity and allow the interchange of limit and integral for Riemann integrals.
• The Weierstrass M-Test is a highly practical sufficient condition for establishing uniform (and absolute) convergence of a function series by comparing it to a convergent series of non-negative real numbers that bound the functions.
• The central theme is that justifying operations on infinite sums or limits of functions (integration, differentiation, continuity) requires verifying a stronger form of convergence—typically uniform convergence—and not merely pointwise convergence.