Counterexample: Differentiation Under The Integral Sign

Hey Plastik Magazine readers! Ever wondered when you can swap a derivative and an integral? It sounds like a neat trick, but sometimes things go sideways. Let's dive into a juicy counterexample that'll keep you on your toes. Today, we're exploring a classic problem in real analysis: finding a case where differentiation under the integral sign fails. This is super important in various fields, from physics to engineering, where integrals pop up all the time. So, buckle up, and let's get started!

Setting the Stage

Okay, so what's the big idea? We're talking about situations where you have an integral like this:

$$F(x) = \int_a^b f(x, t) dt$$

And you're wondering if you can just do this:

$$F'(x) = \int_a^b \frac{\partial}{\partial x} f(x, t) dt$$

Seems simple, right? Just swap the derivative and the integral. But hold on! This only works under certain conditions. One common condition is that $\frac{\partial}{\partial x} f(x, t)$ is continuous and bounded. But what happens if these conditions aren't met? That's where our counterexample comes in. We need to find a function $f(x,t)$ where the usual rules don't apply.

We're going to look at a specific integral that'll show us why we need to be careful. It involves the indicator function, which is a bit of a mathematical firecracker. Let's get into the nitty-gritty!

The Counterexample

Consider the integral:

$$F(x) = \int_{\mathbb{R}} 1_{[0, x]}(t) dt$$

Where $1_{[0, x]}(t)$ is the indicator function, which is 1 if $t$ is in the interval $[0, x]$ and 0 otherwise. In other words:

$$1_{[0, x]}(t) = \begin{cases} 1, & \text{if } 0 \leq t \leq x \\ 0, & \text{otherwise} \end{cases}$$

This integral is actually quite simple. It's just the length of the interval $[0, x]$, so:

$$F(x) = x$$

For $x > 0$. Clearly, $F'(x) = 1$ for $x > 0$.

Now, let's try to differentiate under the integral sign. The partial derivative of the indicator function with respect to $x$ is:

$$\frac{\partial}{\partial x} 1_{[0, x]}(t) = \delta(x - t)$$

Where $\delta$ is the Dirac delta function. If we blindly apply differentiation under the integral sign, we get:

$$\int_{\mathbb{R}} \frac{\partial}{\partial x} 1_{[0, x]}(t) dt = \int_{\mathbb{R}} \delta(x - t) dt = 1$$

This seems to work, right? But let's look closer. The problem is that the partial derivative $\frac{\partial}{\partial x} 1_{[0, x]}(t)$ doesn't exist in the classical sense for $t = x$. The indicator function has a jump discontinuity at $t = x$, which makes differentiation tricky.

Why It Matters

So, what's the big deal? Why can't we just swap the derivative and integral? The issue here is that the conditions for differentiation under the integral sign aren't met. Specifically, the partial derivative of the indicator function with respect to $x$ isn't well-behaved. It's not continuous, and it's not bounded.

Think of it like this: the indicator function is like a switch that flips on at $t = 0$ and flips off at $t = x$. When you differentiate it with respect to $x$, you're essentially looking at how quickly that switch flips. But at the exact point where the switch flips, things get weird. That's where the Dirac delta function comes in, representing an infinitely fast change.

The Moral of the Story

The main takeaway here is that you can't always blindly swap derivatives and integrals. You need to be careful and check that the conditions for differentiation under the integral sign are met. Otherwise, you might end up with a wrong answer.

In this case, the indicator function provided a simple yet powerful counterexample. It showed us that even though the integral $F(x)$ is well-defined and differentiable, the naive application of differentiation under the integral sign can lead to incorrect results.

Key Considerations

• Continuity: Is the function $f(x, t)$ and its partial derivative $\frac{\partial}{\partial x} f(x, t)$ continuous?
• Boundedness: Is the partial derivative $\frac{\partial}{\partial x} f(x, t)$ bounded by some integrable function?
• Uniform Convergence: Does the integral converge uniformly?

If these conditions are met, you're usually good to go. But if not, watch out!

Another Perspective

Let's consider another way to look at this. The function $F(x) = \int_{\mathbb{R}} 1_{[0, x]}(t) dt$ can be rewritten as:

$$F(x) = \int_0^x dt$$

This is a simple Riemann integral, and its derivative is clearly 1. The problem arises when we try to apply the Leibniz rule (which is a general form of differentiation under the integral sign) directly to the indicator function.

The Leibniz rule states that if:

$$F(x) = \int_{a(x)}^{b(x)} f(x, t) dt$$

Then:

$$F'(x) = f(x, b(x)) \cdot b'(x) - f(x, a(x)) \cdot a'(x) + \int_{a(x)}^{b(x)} \frac{\partial}{\partial x} f(x, t) dt$$

In our case, $a(x) = 0$, $b(x) = x$, and $f(x, t) = 1$. So, $a'(x) = 0$, $b'(x) = 1$, and $f(x, b(x)) = 1$, $f(x, a(x)) = 1$. Applying the Leibniz rule gives:

$$F'(x) = 1 \cdot 1 - 0 + \int_0^x 0 dt = 1$$

Which is correct. However, when we try to differentiate the indicator function directly, we run into trouble because the partial derivative doesn't exist in the classical sense.

Conclusion

Alright, folks! We've journeyed through a fascinating counterexample that shows us why differentiation under the integral sign isn't always a free pass. It's a reminder to always check the conditions and be mindful of the functions you're working with. Keep exploring, keep questioning, and never stop learning! Until next time, stay curious!

So, next time you're tempted to swap a derivative and an integral, remember this example and proceed with caution. You might just save yourself from a mathematical mishap! Keep your eyes peeled for more exciting topics in real analysis. Peace out!