Evaluating $\int\prod_{i=1}^{n}(x-a_i)^{b_i}\mathrm dx$ using generalised Newton's binomial theorem

244 Views Asked by At

This post is in reference to this question which asks about the general antiderivative of the following function with the arbitrary constants $a_i,b_i\in \mathbb{R}\forall i\in[1,n]\cap\mathbb{Z^{+}}$.

$$\int\prod_{i=1}^{n}(x-a_i)^{b_1}\mathrm dx$$

It seems easy at first to generalise because this is practically a polynomial, but the source of ambiguity are the real exponents, because they seem to make applying the binomial theorem invalid.

I am thinking about using Taylor series for the $i$th term of the product, but again would we then center the series about the $a_i$? Also how to make sense of something like $(1+x)^{e}$? How to get a grip on it? Any hints are appreciated. Thanks.


Update $1$:

Here's what I've done with the case when $n=2$, using the generalised binomial theorem for complex valued index.

$$\begin{aligned}\int\prod_{i=1}^{2}(x-a_i)^{b_i}\mathrm dx&=\int \prod_{i=1}^{2}\sum_{k=0}^{\infty}(-1)^{k}{b_i\choose k}x^{b_i-k}a_i^{k}\mathrm dx\\ &=\int\left(\sum_{k=0}^{\infty}(-1)^{k}{b_1\choose k}x^{b_1-k}a_1^k\right)\left(\sum_{j=0}^{\infty}(-1)^{j}{b_2\choose j}x^{b_2-j}a_2^{j}\right)\mathrm dx\end{aligned}$$

Considering the $k$th term of the series with parameters $a_1,b_1$ with the corresponding series with parameters $a_2, b_2$ and integrating term by term.

$$\begin{aligned}\int{b_1\choose k}\frac{(-1)^{k}a_1^k}{x^k}\sum_{j=0}^{\infty}{b_2\choose j}\frac{(-1)^{j}a_2^j}{x^{j-(b_1+b_2)}}\mathrm dx&= \sum_{j=0}^{\infty}(-1)^{j+k}{b_1\choose k}{b_2\choose j}a_1^ka_2^j\int\frac{\mathrm dx}{x^{k+j-(b_1+b_2)}}\\&=\sum_{j=0}^{\infty}(-1)^{j+k}{b_1\choose k}{b_2\choose j}a_1^ka_2^j\left(\frac{x^{(b_1+b_2)-(j+k)+1}}{(b_1+b_2)-(j+k)+1}\right)\end{aligned}$$

The integral we started with is now simply the summation of this $k$th term with $k$ varying discretely from $k=0$ to $k=\infty$.

$$\bbox[5px,border:2px solid #C0A000]{\int\prod_{i=1}^{2}(x-a_i)^{b_i}\mathrm dx=\sum_{k=0}^{\infty}\sum_{j=0}^{\infty}(-1)^{j+k}{b_1\choose k}{b_2\choose j}a_1^ka_2^j\left(\frac{x^{(b_1+b_2)-(j+k)+1}}{(b_1+b_2)-(j+k)+1}\right)}$$


Similarly for the case when $n=3$

$$\begin{aligned}\int\left(\sum_{k=0}^{\infty}(-1)^{k}{b_1\choose k}x^{b_1-k}a_1^k\right)\left(\sum_{j=0}^{\infty}(-1)^{j}{b_2\choose j}x^{b_2-j}a_2^j\right)\left(\sum_{i=0}^{\infty}(-1)^{i}{b_3\choose i}x^{b_3-i}a_3^i\right)\mathrm dx\end{aligned}$$

Considering the $k$the term of the first sequence and the $j$th term of the second sequence with the third series.

$$\begin{aligned}(-1)^{j+k}{b_1\choose k}{b_2\choose j}a_1^ka_2^j\sum_{i=0}^{\infty}(-1)^{i}{b_3\choose i}a_3^i\int\frac{\mathrm dx}{x^{i+j+k-(b_1+b_2+b_3)}}\end{aligned}$$

Now how can this summation be written out as both $j$ and $k$ are varying parameters? Also is the approach correct, and whether it can be said that the general integral would be the sum with proper summation notation of the following expression $$\left((-1)^{\sum k_i}\prod_{i=1}^{n}{b_i\choose k_i}\prod_{i=1}^{n}a_i^{k_i}\right)\left(\frac{x^{\sum {b_i}-\sum k_i}+1}{\sum b_i-\sum {k_i}+1}\right)$$ where $k_i$ represents the value being taken on by the respective parameters or indexes of summation like the $i,j,k$ in the above cases?


Update $2$:

I think that the correct way to write this summation, since it is summing over all possible combinations of the indexes $k_i$, is as follows.

$$\bbox[5px,border:2px solid #C0A000]{\sum_{\sigma}\left((-1)^{\sum k_i}\prod_{i=1}^{n}{b_i\choose k_i}\prod_{i=1}^{n}a_i^{k_i}\right)\left(\frac{x^{\sum {b_i}-\sum k_i}+1}{\sum b_i-\sum {k_i}+1}\right), \sigma:=0\le k_i\lt\infty\forall i\in[1,n]\cap\mathbb{Z^{+}}}$$

Is this correct, could someone check my work? Thanks