2

Can some one help me to find the sum of the following expression?

$$1^{\frac{1}{3}}+2^{\frac{1}{3}}+3^{\frac{1}{3}}+ . . . +(50)^{\frac{1}{3}}$$

Kumar
  • 2,259
  • 2
    I doubt there's a nice expression for that. Are you looking for something expressed in terms of special functions? –  May 20 '14 at 14:39
  • If not exact value at least approximate sum is fine to me. – Kumar May 20 '14 at 14:41
  • http://www.wolframalpha.com/input/?i=sum+of+n%5E%281%2F3%29 –  May 20 '14 at 14:43
  • You mean http://www.wolframalpha.com/input/?i=Sum%5Bi%5E%281%2F3%29%2C%7Bi%2C1%2C50%7D%5D&dataset= – rbm May 20 '14 at 14:46
  • Is the sum equal to integration of $x^{1/3}$ from 1 to 50. – Kumar May 20 '14 at 14:48
  • The approximate sum is 139.7179005695. You should really learn to program a computer, at least to the tiny extent needed to get the computer to solve this kind of problem. – MJD May 20 '14 at 14:50

8 Answers8

5

Of course a computer can evaluate it. You can also view this and similar sums in terms of a Riemann sum: $$50^{1 \over 3}*50*\bigg[{1 \over 50} \bigg({1 \over 50}\bigg)^{1/3} + {1 \over 50}\bigg({2 \over 50}\bigg)^{1 \over 3} + ... + {1 \over 50}\bigg({50 \over 50}\bigg)^{1 \over 3}\bigg]$$ This is approximately $$50^{1 \over 3}*50\int_0^1 x^{1 \over 3}\,dx$$ $$=50^{1 \over 3}*50*{3 \over 4}$$ $$= 138.1512...$$ This is reasonably close to the true value $139.7179...$

Zarrax
  • 44,950
4

Because $f(x)=x^{1/3}$ is increasing on $(0,\infty)$, you can get bounds on the sum from

$$\int_0^{50}x^{1/3}dx\lt\sum_{k=1}^{50}k^{1/3}\lt\int_1^{51}x^{1/3}dx$$

Evaluating the integrals gives

$$138.151\lt\sum_{k=1}^{50}k^{1/3}\lt141.09744$$

You can get a better pair of bounds from

$$1+\int_1^{50}x^{1/3}dx\lt1+\sum_{k=2}^{50}k^{1/3}\lt1+\int_{2}^{51}x^{1/3}dx$$

which gives

$$138.4\lt\sum_{k=1}^{50}k^{1/3}\lt140.96$$

You can also try the approximation

$$\sum_{k=1}^{50}k^{1/3}\approx1+\int_{1.5}^{50.5}x^{1/3}dx\approx139.708$$

but it's hard to tell how many digits of that answer you can trust.

Barry Cipra
  • 79,832
2

There is no closed form for it, however it can be approximated by $\displaystyle\int_1^{50}\sqrt[3]x~dx=\bigg[\dfrac34x\sqrt[3]x\bigg]_1^{50}\approx$ $\approx137.4$

Lucian
  • 48,334
  • 2
  • 83
  • 154
  • Could you briefly explain why this is an approximation? – user1729 May 20 '14 at 15:05
  • (+1 for neat low-tech method.) Hey, that rings a bell for me! (See p. 5 of the .pdf file "cal_2W1_.pdf" posted here.) – Dave L. Renfro May 20 '14 at 15:10
  • You can get a somewhat better approximation if you don't subtract the $x = 1$ term, see my answer. – Zarrax May 20 '14 at 15:13
  • 1
    @user1729: ... Now to actually address your question: in general, with certain conditions of monotony and convergence in place, $\sum_a^bf(n)\approx\int_a^bf(x)dx$. [edited ajf] – Lucian May 20 '14 at 15:47
  • @Lucian .... Could you perhaps give a reference? Will it be in a standard calculus text (if so, I cannot find it!). [edited ajf] – user1729 May 20 '14 at 15:52
  • 1
    @user1729: Think about the very definition of a definite integral as a Riemann sum. It should become pretty obvious. – Lucian May 20 '14 at 15:56
  • @Lucian: Ah, okay, I understand. Each rectangle has area $x^3$ because it has base length $1$. I have learned something today - thanks :-) – user1729 May 20 '14 at 16:01
  • @Lucian it would be more obvious with $a=1$ and $b=51$. – WimC May 20 '14 at 16:06
  • @user1729: Regarding your question Could you perhaps give a reference?, see the .pdf file I cited in an earlier comment here. It has labeled graphs, and all the basic details are included. – Dave L. Renfro May 20 '14 at 18:05
2

You can bound the sum for any upper limit b.

$\frac{3 b^{4/3}}{4}<\sum _{n=1}^b \sqrt[3]{n}< \frac{3}{4} \ \left(-1+\sqrt[3]{1+b}+b \sqrt[3]{1+b}\right)$

Plugging in b = 50 you would get

$ 138.151<\sum _{n=1}^{50} \sqrt[3]{n}<141.097 $

bobbym
  • 2,546
1

If you write : $$S_n = \sum_{i = 1}^n i^{1/3}$$ then notice that you can create the following difference equation : $$S_n = S_{n-1} + n^{1/3}$$

This has no "nice" solution but only in terms of the Riemann and Hurwitz zeta functions:

$$S_n = \zeta\biggl(-\frac13\biggr) - \zeta\biggl(-\frac13, n+1\biggr)$$

Using WA, you find that : $$S_{50} = \zeta\biggl(-\frac13\biggr) - \zeta\biggl(-\frac13, 51\biggr) \approx 140$$

MJD
  • 65,394
  • 39
  • 298
  • 580
user88595
  • 4,549
1

By the Euler-Maclaurin formula the sum $$ S = (1/2) 1^{1/3} + (2^{1/3} + 3^{1/3} + \cdots + 49^{1/3}) + (1/2) 50^{1/3} $$ can be approximated by the sum $$ I = \int_1^{50} x^{1/3} \: dx $$ and therefore your sum can be approximated by $$ I + (1/2) 1^{1/3} + (1/2) 50^{1/3} $$. Doing the integral, this is $$ (3/4) (50^{4/3} - 1^{4/3}) + (1/2) 1^{1/3} + (1/2) 50^{1/3} $$ or about 139.743. The true value is about 139.720, so this is already quite accurate.

This is actually nothing but the trapezoid rule, which works well because the function $f(x) = x^{1/3}$ is so smooth.

To get a more accurate formula for this sum, you can use the asymptotic form of the Euler-Maclaurin formula (under "asymptotic expansion of sums" in the Wikipedia article); I leave the details to someone else.

Michael Lugo
  • 22,354
0

You can get as much decimal expansion as you please at WA.

139.7179

0

$$6+3 \sqrt[3]{2}+3\ 2^{2/3}+3 \sqrt[3]{3}+2^{2/3} \sqrt[3]{3}+3^{2/3}+\sqrt[3]{2} 3^{2/3}+3 \sqrt[3]{5}+2^{2/3} \sqrt[3]{5}+3^{2/3} \sqrt[3]{5}+5^{2/3}+\sqrt[3]{2} 5^{2/3}+3 \sqrt[3]{6}+6^{2/3}+\sqrt[3]{7}+2^{2/3} \sqrt[3]{7}+7^{2/3}+\sqrt[3]{10}+\sqrt[3]{11}+2^{2/3} \sqrt[3]{11}+\sqrt[3]{13}+\sqrt[3]{14}+\sqrt[3]{15}+\sqrt[3]{17}+\sqrt[3]{19}+\sqrt[3]{21}+\sqrt[3]{22}+\sqrt[3]{23}+\sqrt[3]{26}+\sqrt[3]{29}+\sqrt[3]{30}+\sqrt[3]{31}+\sqrt[3]{33}+\sqrt[3]{34}+\sqrt[3]{35}+\sqrt[3]{37}+\sqrt[3]{38}+\sqrt[3]{39}+\sqrt[3]{41}+\sqrt[3]{42}+\sqrt[3]{43}+\sqrt[3]{46}+\sqrt[3]{47}$$

  • 2
    I hardly think this is helpful in any way. You could have used fewer keystrokes by simply replacing the "dots" in the OPs sum. – amWhy May 20 '14 at 15:35
  • 2
    @amWhy I am actually the only one in this thread who has answered the OPs question. Everyone else has given decimal approximations, which are completely useless (unless you're a physicist). I have simplified the OPs expression as much as possible and given an exact form. – user85798 May 20 '14 at 15:44
  • 2
    Your "answer" is more complicated than the original sum (and requires more computations), so, again I say, I hardly think this is an answer. – amWhy May 20 '14 at 15:47
  • 1
    @amWhy How exactly is my answer more complicated? The OPs expression has 50 terms, my answer has only 43 terms. – user85798 May 20 '14 at 16:22
  • 1
    Many terms with two factors, some of the form $a^{1/3}b^{2/3}$. As I said, the computational complexity of your un-answer greatly surpasses that of the OP's sum. I'm not going to argue with you anymore. – amWhy May 20 '14 at 16:24
  • 1
    @Oliver I do agree that this is an answer, but it's epistemologically useless with the respect to the cognitive processes of most human beings. It's as much an answer as saying the given sum is $\sum \limits_{n=1}^{50}\left(n^{1/3}\right)$. – Git Gud May 20 '14 at 16:29
  • @GitGud Lucian and Zarrax both gave approximations, not exact answers. – user85798 May 20 '14 at 16:32
  • @Oliver You're right. I didn't read them attentively enough. Edit: I've removed that part of my comment. – Git Gud May 20 '14 at 16:33