Subsection 3.3.1 Definition
In this section, I am going to try to invert the product rule for derivatives to build a new rule for integration. First, recall the product rule, for \(f(x)\) and \(g(x)\) differentiable functions.
\begin{equation*}
\frac{d}{dx} f(x) g(x) = \frac{df}{dx} g(x) + f(x)
\frac{dg}{dx}
\end{equation*}
I’m simply going to integrate both sides of this equation. On the left, the integral of a derivative returns the original function. (I’ll simplify the notation from \(f(x)\) to just \(f\text{;}\) this is a very common notation shift that makes some of these more compicated integration rules a bit more succinct. It’s just up to you to remember that \(f\) and \(g\) are functions.)
\begin{align*}
\int \frac{d}{dx} fg dx \amp = \int \frac{df}{dx} g dx
+ \int f \frac{dg}{dx} dx\\
fg \amp = \int \frac{df}{dx} g dx + \int f
\frac{dg}{dx} dx
\end{align*}
This is a rule about integrals, but it’s not clear how to use it. I’m going to isolate one of the two integral pieces.
\begin{equation*}
\int \frac{df}{dx} g dx = fg - \int f \frac{dg}{dx} dx
\end{equation*}
This might be more useful, since it tells me how to transform certain integral. This is useful enough that there is a name for it.
Definition 3.3.1.
Let \(f\) and \(g\) be functions, with \(f\) integrable and \(g\) differentiable. The following equation for the integral of \(\frac{df}{dx} g\) is called integration by parts.
\begin{equation*}
\int \frac{df}{dx} g dx = fg - \int f \frac{dg}{dx} dx
\end{equation*}
So how do I actually use this? The left side of integration by parts is a product: \(\frac{df}{dx} g\text{.}\) Integration by parts applies to integals of product. However, I have to identify one part of the product as \(g\) and another part as \(\frac{df}{dx}\text{.}\) Then, I need to calculate \(\frac{dg}{dx}\text{,}\) a derivative, and \(f\text{,}\) an antiderivative. If I can do these two things for the choice I made, then I can transfer to the right side of integration by parts. The first part, \(fg\) is just a function, but the second part is a new integral, with integrand \(f
\frac{dg}{dx}\text{.}\) Maybe this is an easier integral. Like the substitution rule, integration by parts doesn’t solve the integral by itself; instead, it changes the form of the integral into something which is (hopefully) more approachable.
In applying the substitution rule, I encouraged careful labelling. The same is true here: I recommend labelling the terms and being particularly careful with \(\pm\) signs. In other literature, there are some concise version of integration by parts with slightly different notations (using \(df\) and \(dg\) for the integrals). I’m not going to use that notation, but for reference, here is what the rule looks like.
\begin{equation*}
\int g df = fg - \int f dg
\end{equation*}
If I have a definite integral, I can still use integration by parts. On the middle term (the term which is not an integral), I have to evalute on the bounds.
\begin{equation*}
\int_a^b \frac{df}{dx} g(x) dx = f(x)g(x) \Bigg|_a^b - \int_a^b
f(x) \frac{dg}{dx} dx
\end{equation*}