0

I want to prove the following formula by induction: $$\sum_{k=0}^n \begin{pmatrix}n \\ k\end{pmatrix} f^{(k)}g^{(n-k)}$$

The base case is trivial but I am not sure if my reasoning is valid for the inductive step. Here is what I have got so far:

$$(fg)^{(n+1)}=\frac{d}{dx}(fg)^{(n)}=\frac{d}{dx} \left(\sum_{k=0}^n \begin{pmatrix}n \\ k\end{pmatrix} f^{(k)}g^{(n-k)} \right) \\ = \sum_{k=0}^n \begin{pmatrix}n \\ k\end{pmatrix} f^{(k+1)}g^{(n-k)}+\sum_{k=0}^n \begin{pmatrix}n \\ k\end{pmatrix}f^{(k)}g^{(n+1-k)}\\ =\sum_{k=1}^{n+1} \begin{pmatrix}n \\ k-1\end{pmatrix}f^{(k)}g^{(n+1-k)}+\sum_{k=0}^n \begin{pmatrix}n \\ k\end{pmatrix}f^{(k)}g^{(n+1-k)} \\ = \sum_{k=0}^{n+1} \begin{pmatrix}n \\ k-1\end{pmatrix}f^{(k)}g^{(n+1-k)}-\color{red}{\begin{pmatrix}n \\ -1\end{pmatrix}f^{(0)}g^{(n+1)}}+\sum_{k=0}^{n+1} \begin{pmatrix}n \\ k\end{pmatrix}f^{(k)}g^{(n+1-k)}-\color{red}{\begin{pmatrix}n \\ n+1\end{pmatrix} f^{(n+1)}g^{(0)}}$$

My goal here was to change the starting point for the first sum from $k=1$ to $k=0$ and to change the end point for the second sum from $n$ to $n+1$. In order to compensate for that I subtracted the terms I added to the sums. I am just not sure about the binomial coefficients of the red parts. Wolfram alpha tells me that $\begin{pmatrix}n \\ n+1\end{pmatrix}=\begin{pmatrix}n \\ -1\end{pmatrix}=0$. Is that correct? Are binomial coefficients even defined for negative $k$? If this is valid then I think I can complete the proof:

$$\implies \sum_{k=0}^{n+1} \begin{pmatrix}n \\ k-1\end{pmatrix}f^{(k)}g^{(n+1-k)}+\sum_{k=0}^{n+1} \begin{pmatrix}n \\ k\end{pmatrix}f^{(k)}g^{(n+1-k)} \\ =\sum_{k=0}^{n+1} \left[ \begin{pmatrix}n \\ k\end{pmatrix}+\begin{pmatrix}n \\ k-1\end{pmatrix} \right]f^{(k)}g^{(n+1-k)} \\ =\sum_{k=0}^{n+1} \begin{pmatrix}n+1 \\ k\end{pmatrix}f^{(k)}g^{(n+1-k)} \space \space \space \square$$

qmd
  • 4,275
  • 3
    Ther are zero ways to pick $n+1$ obejcts from $n$ given objects, and there are zero ways to pick $-1$ objects from $n$ objects. Or, when expanding $(1+x)^n$, the coefficient of $x^{n+1}$ is $=0$, and the coefficient of $x^{-1}$ is also $=0$ – Hagen von Eitzen May 18 '19 at 10:50
  • 1
    Looking at it as the coefficient of $x^{n+1}$ and $x^{-1}$ makes so much sense! Thank you for your help! – qmd May 18 '19 at 10:55

2 Answers2

2

I have no idea what $\binom n{-1}$ is. But note that\begin{multline}\sum_{k=1}^{n+1}\binom n{k-1}f^{(k)}g^{(n+1-k)}+\sum_{k=0}^n\binom nkf^{(k)}g^{(n+1-k)}=\\=f^{(n+1)}g^{(0)}+\sum_{k=1}^n\binom n{k-1}f^{(k)}g^{(n+1-k)}+\sum_{k=1}^n\binom nkf^{(k)}g^{(n+1-k)}+f^{(0)}g^{(n+1)}\end{multline}Now, you can use these facts:

  • $\displaystyle\binom n{k-1}+\binom nk=\binom{n+1}k$ (as you did);
  • $\displaystyle f^{(n+1)}g^{(0)}=\binom{n+1}{n+1}f^{(n+1)}g^{(0)}$;
  • $\displaystyle f^{(0)}g^{(n+1)}=\binom{n+1}0f^{(0)}g^{(n+1)}$.
  • That looks like a more elegant way to do it and it evades the $\begin{pmatrix}n \ -1 \end{pmatrix}$ "problem". However, I am a bit confused now since two people have said that $\begin{pmatrix}n \ -1 \end{pmatrix}$ is zero and you are saying that it is not defined (or you are not sure)? – qmd May 18 '19 at 11:05
  • Actualyy, what I wrote is that I don't know what it means. – José Carlos Santos May 18 '19 at 11:08
  • I am sorry I misinterpreted that. Thank you for your help! – qmd May 18 '19 at 11:12
1

This is not an answer, but I don't have enough reputation to comment, so there you go. First off: the fact about combinatorial numbers that Wolfram Alpha told you is indeed valid, and as far as I could see your proof is correct.

If you want a quick intuition for $$\begin{pmatrix}n \\ n+1\end{pmatrix}=\begin{pmatrix}n \\ -1\end{pmatrix}=0$$ think of the Tartaglia triangle (aka Pascal triangle) written in terms of these numbers. The numbers in that triangle can be thought of as surrounded by zeroes for the sum of two contiguous numbers to equal the number below even at the sides of the triangle. In "more formal" terms, you can define $$\begin{pmatrix}n \\ n+1\end{pmatrix}=\begin{pmatrix}n \\ -1\end{pmatrix}=0$$ by imposing that the condition $$\begin{pmatrix}n \\ k\end{pmatrix}=\begin{pmatrix}n-1 \\ k-1\end{pmatrix} + \begin{pmatrix}n-1 \\ k\end{pmatrix}$$ holds.

TeicDaun
  • 202
  • 1
    I have to think about the pascal triangle part a bit but the second "more formal" argument makes perfect sense. Thank you! – qmd May 18 '19 at 10:58