Partial Fraction Decomposition

Click on an arrow to get a description of the connection!

Click on an arrow to get a description of the connection!

We know by partial fraction decomposition that every rational function \(f(x)\) possesses a representation as the sum of some polynomial \(r\) and a strict proper rational function \(\frac{p(x)}{q(x)}\), i.e \[f(x) = r(x) + \frac{p(x)}{q(x)},\quad \deg(p)<\deg(q).\] By using linearity of the integral, we can integrate each summand separately. Thus we restrict our discussion here to strict proper rational functions. Furthermore by partial fraction decomposition we can write every strict proper rational function as a sum of partial fractions \(\frac{A}{(x-x_{0})^{s}}\), with \(x_{0}\) being a root of \(q\) and \(A\) being some constant and \(s\geq 1\).

Altogether we have that it is enough to know the antiderivative of the partial fractions in order to integrate arbitrary rational functions.
Let us first restrict to the case \(x_{0}\in\mathbb{R}\) and \(s=1\). Then we have \[\begin{aligned} \int \frac{A}{x-x_{0}}\, dx = A \log|x-x_{0}|+\text{const}.,\qquad x_{0}\in\mathbb{R}. \end{aligned}\] It is clear that this equality is only valid in the case \(x_{0}\in\mathbb{R}\) since the antiderivative is a real valued function.
Now consider \(x_{0}\in\mathbb{C}\setminus\mathbb{R}\). Because of \(q\) having only real coefficients we know that also \(\overline{x_{0}}\) is a root of \(q\) thus the partial fraction decomposition at least has the terms \(\frac{A}{x-x_{0}}\) and \(\frac{B}{x-\overline{x_{0}}}\) included in the sum. Then setting \(x_{0}:=\alpha+i\beta\) \[\begin{aligned} \int \frac{A}{x-x_{0}}+\frac{B}{x-\overline{x_{0}}}\, dx &= \int \frac{\overbrace{(A+B)}^{=:a}(x-\alpha)+i\overbrace{(A-B)}^{=:b}\beta}{\beta^{2}+(x-\alpha)^{2}}\, dx. \end{aligned}\] Two integrals arise in this case. For the first one the antiderivative can be calculated using the substitution \(t=\frac{x-\alpha}{\beta}\), \(dt = \frac{dx}{\beta}\) \[\begin{aligned} \int \frac{a(x-\alpha)}{\beta^{2}+(x-\alpha)^{2}}\, dx &= a\int \frac{\frac{x-\alpha}{\beta}}{1+(\frac{x-\alpha}{\beta})^{2}}\, \frac{dx}{\beta}=a\int \frac{t}{1+t^{2}}\,dt\\ & = \frac{a}{2}\log(1+t^{2})+\text{const}. = \frac{a}2\log\left(1+\left(\frac{x-\alpha}{\beta}\right)^{2}\right)+\text{const}. \end{aligned}\] For the second term we have again using the same substitution as above \[\begin{aligned} \int \frac{b\beta}{\beta^{2}+(x-\alpha)^{2}}\, dx & = b\int \frac{1}{1+(\frac{x-\alpha}{\beta})^{2}}\, \frac{dx}{\beta}=b\int \frac{1}{1+t^{2}}\,dt\\ & = b \arctan(t) + \text{const} = b \arctan\left(\frac{x-\alpha}{\beta}\right) + \text{const}. \end{aligned}\] Altogether we have for \(x_{0}\in\mathbb{C}\setminus\mathbb{R}\) and \(s=1\) \[\begin{aligned} \int \frac{A}{x-x_{0}}+\frac{B}{x-\overline{x_{0}}}\, dx = \frac{A+B}2\log\left(1+\left(\frac{x-\alpha}{\beta}\right)^{2}\right)+i(A-B)\arctan\left(\frac{x-\alpha}{\beta}\right) + \text{const}. \end{aligned}\] Now we want to investigate the case \(s>1\). Here it is not relevant if \(x_{0}\) is real or complex. We have \[\begin{aligned} \int \frac{A}{(x-x_{0})^{s}}\, dx = \frac{A}{1-s}(x-x_{0})^{1-s}+\text{const}. \end{aligned}\]