5

How do I find $\sum_ {k=0}^{\infty} \frac{k^3}{3^k}$ . I tried like derivative,like I did in other examples,but in this example that doesn't work... Can somebody help?

mrf
  • 43,639
Andjela
  • 163

4 Answers4

6

$$k^3=k(k-1)(k-2)+3k(k-1)+k.$$

Then by derivation,

$$s\left(\frac13\right):=\sum_{k=0}^\infty\frac1{3^k}=\frac1{1-\frac13},$$ $$s'\left(\frac13\right):=\sum_{k=0}^\infty\frac k{3^{k-1}}=\frac1{\left(1-\frac13\right)^2},$$ $$s''\left(\frac13\right):=\sum_{k=0}^\infty\frac{k(k-1)}{3^{k-2}}=\frac2{\left(1-\frac13\right)^3},$$ $$s'''\left(\frac13\right):=\sum_{k=0}^\infty\frac{k(k-1)(k-2)}{3^{k-3}}=\frac{3!}{\left(1-\frac13\right)^4},$$ and the sum is

$$\frac{3!}{3^3}\left(\frac32\right)^4+\frac{2\cdot3}{3^2}\left(\frac32\right)^3+\frac13\left(\frac32\right)^2= \frac{33}8.$$

  • Your result is wrong. – Enrico M. Jan 04 '16 at 12:12
  • Kim Peek II is right. The problem is differentiating $x^k$ three times gives $kx^{k-1},,k(k-1)x^{k-2},,k(k-1)x^{k-3}$ instead of $kx^k,,k(k-1)x^k,,k(k-1)x^k$. In other words, the derivatives are respectively $3,,9,,27$ times too big. The result is therefore $\tfrac{6}{27}\left(\frac{3}{2}\right)^4+\tfrac{6}{9}\left(\frac{3}{2}\right)^3+\tfrac{1}{3}\left(\frac{3}{2}\right)^2=4.125$, which is the answer Kim Peek II provides. – J.G. Jan 04 '16 at 12:33
  • @JosGibbons: thanks for the investigation. I am fixing. –  Jan 04 '16 at 13:12
3

Suppose you have $$s(x)=\sum x^k$$

Then $$t(x)=xs'(x)=\sum kx^k$$

apply a couple more times, with attention to limits.


So take the lower limit of the sum to be $k=0$ so that $s(x)=\frac 1{1-x}$ with $|x|\lt 1$.

Then $t(x)=xs'(x)$ gives $\sum kx^k$ with lower limit $k=1$, but since the term with $k=0$ is zero, we can take the lower limit as zero. This gives $t(x)=\frac x{(1-x)^2}$.

And $u(x)=xt'(x)$ sums $k^2x^k$ in the same way, with the same comment on limits with $u(x)=\frac {1+x}{(1-x)^3}$.

Now do the same for $v(x)=xu'(x)$.

Mark Bennet
  • 100,194
  • I tried but I don't know how. Can you explane me? – Andjela Jan 03 '16 at 22:20
  • @Andjela So $u(x)=xt'(x)=\sum k^2x^k$ and the sum for $s(x)$ is a geometric series, which you should know. – Mark Bennet Jan 03 '16 at 22:21
  • I know that , thank you ,but I still can't solve this and I am sorry but can you writte it down? – Andjela Jan 03 '16 at 22:27
  • @Andjela Maybe I'll write it tomorrow. But you'll learn a lot more if you work it out for yourself. $x=\frac 13$ for the example you have. You will need $v(x)=xu'(x)$. If you are preparing for an exam, what will you do if $k^4$ turns up? – Mark Bennet Jan 03 '16 at 22:30
  • I gonna understand If you help me,I promisse . Yes,for exam.And thank you anyway. – Andjela Jan 03 '16 at 22:32
1

Without differentiation of formal series:

$$S_0:=\sum_{k=0}^\infty\frac1{3^k}=\frac1{1-\frac13}.$$ By shifting of the index and noting that the sums can start at $0$ or $1$ indifferently:

$$S_1:=\sum_{k=0}^\infty\frac k{3^k}=\sum_{k=0}^\infty\frac{k+1}{3^{k+1}}=\frac{S_1+S_0}3,$$ $$S_2:=\sum_{k=0}^\infty\frac{k^2}{3^k}=\sum_{k=0}^\infty\frac{k^2+2k+1}{3^{k+1}}=\frac{S_2+2S_1+S_0}3,$$ $$S_3:=\sum_{k=0}^\infty\frac{k^3}{3^k}=\sum_{k=0}^\infty\frac{k^3+3k^2+3k+1}{3^{k+1}}=\frac{S_3+3S_2+3S_1+S_0}3.$$

This yields

$$S_0=\frac32, S_1=\frac34, S_2=\frac32, S_3=\frac83.$$


Generalizing, we have the simple recurrence

$$S_n(x):=\sum_{k=0}^\infty k^nx^k=\sum_{k=0}^\infty (k+1)^nx^{k+1}=x\sum_{j=0}^n\binom njS_j(x),$$ or

$$S_n(x)=\frac x{1-x}\sum_{j=0}^{n-1}\binom njS_j(x).$$

  • @BarryCipra: I know and I didn't want to clutter with this detail. I have reformatted to fix. –  Jan 04 '16 at 13:45
0

I will write here a curiosity I found in doing this exercise because I find it really amusing.

Let's test if the series converges:

$$\lim_{\Lambda\to +\infty} \int_0^{\Lambda}\ k^3\cdot 3^{-k}\ \text{d}k = \lim_{\Lambda\to +\infty} \left(- \frac{3^{-k}\cdot (6 + 6k\ln(3) +3k^2 \ln^2(3) + k^3\ln^3(3) )}{\ln^4(3)}\right)_1^{\Lambda}$$

substituting and calculating the limit we obtain

$$\lim_{\Lambda\to +\infty} \int_0^{\Lambda}\ k^3\cdot 3^{-k}\ \text{d}k = \frac{6}{\ln^4(3)} \approx 4.118 $$

The amusing fact is that that series does converge and its sum is

$$\sum_{k = 0}^{+\infty}\ \frac{k^3}{3^k} = \frac{33}{8} = 4.125$$

strictly similar to the integral above.

I'm not such an expert so I demand: is this a funny coincidence or is there some connection (in general)?

P.s. I obtained the sum value with Mathematica, but it's easily computing by hands too.

Enrico M.
  • 26,114
  • This would really be more appropriate as its own MSE question (perhaps with a link to the question that motivated it). Incidentally, in American English at least, it's more polite to "ask" than to "demand." – Barry Cipra Jan 04 '16 at 14:28
  • One of the most useless comments ever. – Enrico M. Jan 04 '16 at 16:01