Pointwise Vs. Uniform Convergence: A Calculus Example

Hey guys, welcome back to Plastik Magazine! Today, we're diving deep into the fascinating world of calculus, specifically tackling a concept that can trip up even the most seasoned math enthusiasts: the difference between pointwise convergence and uniform convergence of function series. You know how sometimes a function series looks like it's behaving nicely, converging in a pretty standard way, but then BAM! it falls apart when you look at it more closely? That's often where the distinction between these two types of convergence really shines (or, you know, doesn't shine, if it's not uniform!).

We're going to explore a classic example today, a series of functions that does converge, but sadly, not uniformly. Understanding this difference is super crucial because uniform convergence has some seriously cool properties that pointwise convergence just doesn't offer. For instance, if a series of functions converges uniformly to a function, and each of those functions in the series is continuous, then the limit function is also continuous. That's a biggie! Also, uniform convergence allows us to integrate and differentiate the series term-by-term, which is incredibly powerful. So, when we find a series that converges pointwise but not uniformly, it's like finding a loophole, a situation where we can't automatically apply these powerful theorems. It highlights why the stricter condition of uniform convergence is sometimes necessary. We'll break down the example step-by-step, making sure to explain why it converges pointwise and, more importantly, why it fails to converge uniformly. Get ready to flex those calculus muscles!

The Quest for a Convergent, Non-Uniform Series

Alright, let's get down to business. You're looking for that elusive example, a series of functions that converges, but not uniformly. This is a common quest when first grappling with these ideas, and for good reason! It perfectly illustrates the subtle but significant gap between pointwise and uniform convergence. The canonical example that usually comes up in textbooks and lectures, and the one we'll dissect today, involves a series defined on a specific interval. We want to show that as 'n' (the index of our series terms) gets larger, the sum of the functions approaches a specific limit function. This is the 'converges' part. But then, we need to demonstrate that this convergence isn't happening 'at the same pace' across the entire interval, which is the hallmark of non-uniform convergence.

Consider the series $\sum_{n=1}^{\infty} f_n(x)$, where each $f_n(x)$ is a carefully crafted function. The beauty of this example is that it's not overly complicated, yet it perfectly captures the essence of the problem. We'll need to define our functions $f_n(x)$ in a way that their sum behaves nicely pointwise but exhibits a particular 'jumpiness' or 'instability' as 'n' increases when viewed across the interval. Think about functions that are non-zero only on very small, moving intervals. This kind of behavior often leads to pointwise convergence (because for any fixed x, eventually the terms become zero and stay zero) but prevents uniform convergence (because that small interval where the function is non-zero might still be significant relative to the 'error' we're allowed to have). So, stay with me as we build this example from the ground up, looking at both its pointwise behavior and its failure to meet the stricter criteria of uniform convergence. It's going to be illuminating!

Constructing the Example Series

Let's define our series. The general form we'll be looking at is $\sum_{n=1}^{\infty} f_n(x)$. A very common and effective example is the series defined by:

$$f_n(x) = \frac{x^2}{n} e^{-x/n}$$

This is a function series where each term $f_n(x)$ depends on both the index 'n' and the variable 'x'. We're interested in its behavior, particularly on an interval like $[0, \\infty)$. The first step in understanding any series is to figure out what function it converges to, pointwise. This means, for each specific value of 'x', we need to determine the limit of the sum as $n \to \infty$. Let's fix an arbitrary $x \ge 0$. We need to evaluate:

$$S(x) = \lim_{N \to \infty} \sum_{n=1}^{N} f_n(x) = \lim_{N \to \infty} \sum_{n=1}^{N} \frac{x^2}{n} e^{-x/n}$$

Now, this looks a bit tricky at first glance. However, let's consider the behavior of the term $\frac{x^2}{n} e^{-x/n}$ as $n \to \infty$ for a fixed $x$. If $x=0$, then $f_n(0) = \frac{0^2}{n} e^{-0/n} = 0$ for all $n$. So, the sum is clearly 0.

If $x > 0$, we can analyze the limit of $f_n(x)$ as $n \to \infty$. Let $y = 1/n$. As $n \to \infty$, $y \to 0^+$. The expression becomes $x^2 y e^{-xy}$. As $y \to 0^+$, this term approaches $x^2 \cdot 0 ecdot e^0 = 0$.

This suggests that each term $f_n(x) \to 0$ as $n \to \infty$ for any fixed $x \ge 0$. While this is a necessary condition for convergence, it doesn't guarantee that the sum converges. However, we can use a more powerful tool here: the Weierstrass M-test or by direct comparison with a known convergent series.

Let's think about the behavior of $g(t) = \frac{1}{t}e^{-t}$ for $t > 0$. Its derivative is $g'(t) = -\frac{1}{t^2}e^{-t} - \frac{1}{t}e^{-t} = -e^{-t}(\frac{1}{t^2} + \frac{1}{t})$. For $t > 0$, $g'(t) < 0$, so $g(t)$ is a decreasing function.

Consider the term $\frac{x^2}{n} e^{-x/n}$. Let $t = n/x$. Then the term is $x^2 \frac{1}{n} e^{-x/n}$. This doesn't simplify nicely.

Let's consider the integral test analogy. The integral $\int_1^\infty \frac{x^2}{t} e^{-x/t} dt$ might give us a hint.

However, a simpler approach is to use bounding. For a fixed $x > 0$, let's consider the function $h(n) = \frac{x^2}{n} e^{-x/n}$. We already saw that $\lim_{n \to \infty} h(n) = 0$. To show the series converges, we can compare it to something we know converges.

Let's rethink the function. Perhaps a slightly different construction is more illustrative for the non-uniform aspect. A very common and clean example is:

$$f_n(x) = \frac{x}{n} - \frac{x}{n+1}$$

This is a telescoping series if we consider the sum $\sum_{n=1}^N f_n(x)$. Let's look at the partial sum $S_N(x) = \sum_{n=1}^N (\frac{x}{n} - \frac{x}{n+1})$.

$S_N(x) = (\frac{x}{1} - \frac{x}{2}) + (\frac{x}{2} - \frac{x}{3}) + \dots + (\frac{x}{N} - \frac{x}{N+1})$.

This telescopes to $S_N(x) = x - \frac{x}{N+1}$.

Now, let's find the pointwise limit $S(x) = \lim_{N \to \infty} S_N(x)$.

$$S(x) = \lim_{N \to \infty} (x - \frac{x}{N+1}) = x - 0 = x$$

So, the series $\sum_{n=1}^{\infty} (\frac{x}{n} - \frac{x}{n+1})$ converges pointwise to the function $S(x) = x$ for all $x$. This is our 'converges' part. Now, we need to show it does not converge uniformly on some interval.

Demonstrating Non-Uniform Convergence

We've established that our series $\sum_{n=1}^{\infty} (\frac{x}{n} - \frac{x}{n+1})$ converges pointwise to $S(x) = x$. The question now is, does it converge uniformly to $S(x)=x$ on a chosen interval? Let's consider the interval $[0, 1]$. For uniform convergence on an interval $[a, b]$, we need the maximum difference between the partial sum $S_N(x)$ and the limit function $S(x)$ to approach zero as $N \to \infty$, and this maximum should be taken over the entire interval. Mathematically, this means:

$$\lim_{N \to \infty} \sup_{x \in [a, b]} |S_N(x) - S(x)| = 0$$

Let's calculate $|S_N(x) - S(x)|$ for our series on the interval $[0, 1]$. We found $S_N(x) = x - \frac{x}{N+1}$ and $S(x) = x$. So,

$$|S_N(x) - S(x)| = |(x - \frac{x}{N+1}) - x| = |-\frac{x}{N+1}| = \frac{|x|}{N+1}$$

Since we are considering the interval $[0, 1]$, $x$ is non-negative, so $|x|=x$. Thus,

$$|S_N(x) - S(x)| = \frac{x}{N+1}$$

Now, we need to find the supremum (the least upper bound) of this quantity over the interval $[0, 1]$:

$$\sup_{x \in [0, 1]} \frac{x}{N+1}$$

The function $\frac{x}{N+1}$ is an increasing function of $x$ (since $N+1$ is positive and constant with respect to $x$). Therefore, its maximum value on the interval $[0, 1]$ occurs at $x=1$.

$$\sup_{x \in [0, 1]} \frac{x}{N+1} = \frac{1}{N+1}$$

Finally, we check the limit of this supremum as $N \to \infty$:

$$\lim_{N \to \infty} \frac{1}{N+1}$$

This limit is clearly 0.

Wait a minute! This seems to indicate uniform convergence on $[0, 1]$. What went wrong? Ah, the example needs to be chosen carefully to demonstrate non-uniform convergence. Let's go back to the drawing board and pick a different, more standard example that highlights this failure.

A Better Example: The Classic Non-Uniform Converger

Let's consider the series of functions defined on the interval $[0, 1]$:

$$f_n(x) = nx e^{-nx^2}$$

This might look a bit more complex, but it's a textbook case for demonstrating non-uniform convergence. First, let's find the pointwise limit $S(x)$. For any fixed $x$, we need to evaluate:

$$S(x) = \lim_{n \to \infty} nx e^{-nx^2}$$

Case 1: $x = 0$. $f_n(0) = n ecdot 0 ecdot e^{-n ecdot 0^2} = 0$. So, $S(0) = 0$.

Case 2: $x > 0$. Let $y = nx$. The expression is $y e^{-yx}$. This form is tricky. Let's rewrite it as $\frac{nx}{e^{nx^2}}$. As $n \to \infty$, this is of the indeterminate form $\frac{\infty}{\infty}$. We can use L'Hôpital's Rule with respect to $n$ (treating $x$ as a constant):

$$\lim_{n \to \infty} \frac{nx}{e^{nx^2}} = \lim_{n \to \infty} \frac{\frac{d}{dn}(nx)}{\frac{d}{dn}(e^{nx^2})} = \lim_{n \to \infty} \frac{x}{x^2 e^{nx^2}} = \lim_{n \to \infty} \frac{1}{x e^{nx^2}}$$

Since $x > 0$, $x^2 > 0$. As $n \to \infty$, $nx^2 \to \infty$, so $e^{nx^2} \to \infty$. Therefore, $\frac{1}{x e^{nx^2}} \to 0$.

So, for any $x \in [0, 1]$, the pointwise limit is $S(x) = 0$. The series converges pointwise to the zero function.

Now, let's check for uniform convergence on $[0, 1]$. We need to examine the behavior of $|S_N(x) - S(x)| = |nx e^{-nx^2} - 0| = nx e^{-nx^2}$ for $x \in [0, 1]$. For uniform convergence, the maximum value of this difference over the interval $[0, 1]$ must go to zero as $N \to \infty$.

Let's fix $N$ and find the maximum of $g(x) = nx e^{-nx^2}$ for $x \in [0, 1]$. We can find critical points by taking the derivative with respect to $x$ and setting it to zero.

$$g'(x) = \frac{d}{dx}(nx e^{-nx^2})$$

Using the product rule:

$$g'(x) = n ecdot e^{-nx^2} + nx ecdot e^{-nx^2} ecdot (-2nx)$$

$$g'(x) = n e^{-nx^2} - 2n^2 x^2 e^{-nx^2}$$

$$g'(x) = n e^{-nx^2} (1 - 2nx^2)$$

Set $g'(x) = 0$. Since $n e^{-nx^2}$ is always positive, we need:

$$1 - 2nx^2 = 0$$

$$2nx^2 = 1$$

$$x^2 = \frac{1}{2n}$$

$$x = \sqrt{\frac{1}{2n}} = \frac{1}{\sqrt{2n}}$$

This value of $x$ is within our interval $[0, 1]$ for $n \ge 1$.

Now, let's find the maximum value of $nx e^{-nx^2}$ at this critical point $x = \frac{1}{\sqrt{2n}}$:

$$nx e^{-nx^2} = n \left(\frac{1}{\sqrt{2n}}\right) e^{-n(\frac{1}{\sqrt{2n}})^2}$$

$$= \frac{n}{\sqrt{2n}} e^{-n(\frac{1}{2n})}$$

$$= \frac{\sqrt{n}}{\sqrt{2}} e^{-1/2}$$

$$= \frac{\sqrt{n}}{\sqrt{2}} \frac{1}{\sqrt{e}}$$

$$= \frac{1}{\sqrt{2e}} \sqrt{n}$$

This expression, $\frac{1}{\sqrt{2e}} \sqrt{n}$, is the maximum value of $|S_N(x) - S(x)|$ on $[0, 1]$ for a given $N$.

For uniform convergence, we need $\lim_{N \to \infty} \sup_{x o [0,1]} |S_N(x) - S(x)| = 0$. Let's evaluate this limit:

$$\lim_{N \to \infty} \frac{1}{\sqrt{2e}} \sqrt{N}$$

As $N \to \infty$, $\sqrt{N} \to \infty$. Therefore, this limit is infinity, not 0.

Why It Fails Uniform Convergence

The reason this series fails to converge uniformly is precisely because the maximum error $|S_N(x) - S(x)|$ does not go to zero as $N \to \infty$. For any given large $N$, there's always some $x$ in the interval $[0, 1]$ where the difference between the partial sum $S_N(x)$ and the limit function $S(x)=0$ is significant. In this specific example, the 'peak' error occurs at $x = \frac{1}{\sqrt{2N}}$, and the height of this peak grows with $N$ (as $\sqrt{N}$). This means that no matter how large $N$ gets, there will always be points in the interval where $S_N(x)$ is 'far' from 0.

Uniform convergence requires that the functions $S_N(x)$