2

How do I find the formula for this sum

$S_n = \sum_{i=1}^n i(i-1)$

Wolfram alpha gives me the correct formula which is $ \frac{1}{3} (n-1)n(n+1)$, but I'm interested in how I get to that and how I can approach this kind of problem in general.

N. F. Taussig
  • 76,571

5 Answers5

4

We have that $$ S_n=\sum_{i=1}^n[i(i-1)]=\sum_{i=1}^n[i^2-i]=\sum_{i=1}^ni^2-\sum_{i=1}^ni. $$ Also, $$ \sum_{i=1}^ni=\frac{n(1+n)}2 $$ and $$ \sum_{i=1}^ni^2=\frac{n(n+1)(2n+1)}6 $$ (see, for example, this question). Hence, $$ S_n=\frac{n(n+1)(2n+1)}6-\frac{n(1+n)}2=\frac{n(n+1)(n-1)}3=\frac{n(n^2-1)}{3}. $$

Cm7F7Bb
  • 17,364
  • I've forgotten even basic high school math it seems! I was trying to somehow write $a_n = a_1 + 2(1 + 2 + ... + a_{n-1})$ and then find a formula for the sum but I was going the wrong way it seems. – Adrian Buzea Feb 23 '16 at 10:32
  • @Adrian It is really easier to split the sum into two parts. One part is just an arithmetic progression and another part is the sum of squares. The sum of squares has this nice formula. – Cm7F7Bb Feb 23 '16 at 10:38
3

$$\sum_{i=0}^n i(i-1)=2\sum_{i=0}^n\binom i2=2\binom {n+1}3=\frac {(n-1)n(n+1)}3\quad\blacksquare $$


NB: the above solution makes use of the fact that $$\sum_{i=0}^n\binom im=\sum_{i=m}^n\binom im=\sum_{i=m}^n\binom {i+1}{m+1}-\binom i{m+1}=\binom {n+1}{m+1}$$ with the summation result arrived at by telescoping and noting that $\binom {m-1}{m+1}=0$.

1

A systematic approach was discovered about 1690 by one of the Bernoulli's. For a sum with n terms, where each term is the same polynomial in the index variable, assume that the sum will be a polynomial in n, and the order will be one greater.

Then fit coefficients and factor the resultant polynomial to get the formula. I discovered this in high school on my own; I was inspired by the story of the young Gauss as given by E.T. Bell in "Men of Mathematics". The geometric methods given above are also neat.

1

If perhaps you forget the summation formulas, or can't derive them for some reason. There is a trick we can do. Notice that:

$$s_{n}-s_{n-1}=n(n-1)=n^2-n$$

This is what we call a recurrence relation. Notice that $s_{n}-s_{n-1}=\frac{s_n-s_{n-1}}{n-(n-1)}$ resembles the definition of the derivative of $s$, and in fact should be somewhat close to the derivative. So we can make an educated guess that if $s_{n}-s_{n-1}$ is a quadratic, then $s_{n}$ is an integral of some quadratic (a $3$rd degree polynomial):

$$s_n=an^3+bn^2+cn+d$$

Now what you can do is find:

$$s_n-s_{n-1}$$

This is:

$$s_n-s_{n-1}=a(3n^2-3n+1)+b(2n-1)+c=n^2-n$$

Now equate coefficients:

$$3a=1$$ $$-3a+2b=-1$$ $$a-b+c=0$$

To get:

$$a=\frac{1}{3}$$

$$b=0$$

$$c=\frac{-1}{3}$$

And with our base case $s_1=1^2-1=0$ we can find that $d=0$.

In conclusion:

$$s_n=\frac{n^3-n}{3}$$

0

This is much simpler. Let's use the notation $x^{\underline{k}} = x (x - 1) \dotsm (x - k + 1)$ for arbitrary $x$ and $k \in \mathbb{N}_0$.

Prove by induction on $n$ that:

$\begin{align} \sum_{0 \le k < n} k^{\underline{m}} = \frac{n^{\underline{m + 1}}}{m + 1} \end{align}$

Base: For $n = 0$, this is true, the left hand side is an empty sum and the right hand side is zero.

Induction: Asume it is true for $n$, check $n + 1$:

$\begin{align} \sum_{0 \le k < n + 1} k^{\underline{m}} &= \sum_{0 \le k < n} k^{\underline{m}} + n^{\underline{m}} \\ &= \frac{n^{\underline{m + 1}}}{m + 1} + n^{\underline{m}} \\ &= \frac{n^{\underline{m}} \cdot (n - m) + n^{\underline{m}} \cdot (m + 1)} {m + 1} \\ &= \frac{n^{\underline{m}} (n + 1)}{m + 1} \\ &= \frac{(n + 1)^{\underline{m + 1}}}{m + 1} \end{align}$

For your specific case:

$\begin{align} \sum_{1 \le i \le n} i (i - 1) &= \sum_{0 \le i < n + 1} i^{\underline{2}} \\ &= \frac{(n + 1)^{\underline{3}}}{3} \\ &= \frac{(n + 1) n (n - 1)}{3} \\ &= \frac{n^3 - n}{3} \end{align}$

vonbrand
  • 27,812