6

Find Sum of $$\bigg\lfloor 1+\frac{1}{\sqrt{2}}+\frac{1}{\sqrt{3}}+\frac{1}{\sqrt{4}}+\cdots+\frac{1}{\sqrt{50}}\bigg\rfloor$$

$\bf{My\; Try::}$ Let $\displaystyle y=f(x) = \frac{1}{\sqrt{x}}\;,$ Then draw that graph in coordinate axis, We get

$$\displaystyle \int_{1}^{51}\frac{1}{\sqrt{x}}dx<\sum^{50}_{k=1}\frac{1}{\sqrt{k}}<1+\int_{1}^{50}\frac{1}{\sqrt{x}}dx$$

So we get $$\displaystyle 2\left(\sqrt{51}-1\right)<\sum^{50}_{k=1}\frac{1}{\sqrt{k}}<1+2(\sqrt{50}-1)$$

So we get $$12.28<\sum^{50}_{k=1}\frac{1}{\sqrt{k}}<13.14$$

But answer given is $12,$ How cai i solve avobe question, Help Required, Thanks

juantheron
  • 53,015
  • 1
    Your estimates are fine, but the upper limit is too crude an approximation at the low end. If you used the integral from 3,...,50 and added on the first three terms instead of just the first you would get 12.96. – almagest Jun 14 '16 at 18:36
  • The actual sum is 12.752, so perhaps you have more acurately bound them than your book has suggested. If the book has 12, (a perfect square) that suggests $2(\sqrt{49}-1).$ – Doug M Jun 14 '16 at 18:36

4 Answers4

5

By the same method, you find for $m>1$ $$ 2(\sqrt{51}-\sqrt m)<\sum_{k=m}^{50}\frac1{\sqrt k}<2(\sqrt{50}-\sqrt{m-1})$$ With $m=5$ (actually, $m=4$ would be enough, but is "harder to compute" as it involves more numerical square root computations), this gives us $$\begin{align}\sum_{k=1}^{50}\frac1{\sqrt k}&<1+\frac1{\sqrt 2}+\frac1{\sqrt 3}+\frac1{\sqrt 4}+2(\sqrt{50}-\sqrt 4)\\ &<1+0.71+0.58+0.5+14.15-4 \\ &=12.94\end{align}$$

2

$$ \sum_{k=1}^{50} \frac 1 {\sqrt{k}} = \int_1^{51} \frac 1 {\sqrt{x}} \, dx + \text{a little bit more} = 2(\sqrt{51} - 1) + \text{a little bit more}. $$ We just need to show that that "little bit" is small enough. A carefully drawn graph show show you why $$ \text{that little bit} < \left( 1 - \frac 1 {\sqrt 2} \right) + \left( \frac 1 {\sqrt 2} - \frac 1 {\sqrt 3} \right) + \left( \frac 1 {\sqrt 3} - \frac 1 {\sqrt 4} \right) + \cdots + \left( \frac 1 {\sqrt{50}} - \frac 1 {\sqrt{51}} \right) $$ and all of the terms cancel out except the first and the last, so $$ \text{that little bit} < 1 - \frac 1 {\sqrt{51}}. $$ That doesn't quite do it, so we refine the technique: $$ \frac 1 {\sqrt 1} = \int_1^2 \frac{dx}{\sqrt x} + \text{a little bit} = 2(\sqrt 2 - 1) + \text{a little bit}. $$ That little bit can be found numerically and it is less than $60\%$ of $1 - \dfrac 1 {\sqrt 2}$.

In all the latter terms, the $60\%$ would be replaced by something even smaller than $60\%$ (but always bigger than $50\%$ for reasons that should be obvious from looking at the graph).

Thus $$ \text{little bit} < 0.6\times \left( 1 - \frac 1 {\sqrt 2} \right) + \left( \frac 1 {\sqrt 2} - \frac 1 {\sqrt 3} \right) + \left( \frac 1 {\sqrt 3} - \frac 1 {\sqrt 4} \right) + \cdots + \left( \frac 1 {\sqrt{50}} - \frac 1 {\sqrt{51}} \right) $$ and that does it.

1

It's lots easier if you use centered integrals. A little algebra shows that $$\begin{align}\int_{k-\frac12}^{k+\frac12}\frac{dx}{\sqrt x}-\frac1{\sqrt k}&=\frac1{2\sqrt k\left(\sqrt{k+\frac12}+\sqrt{k-\frac12}\right)^2\left(\sqrt{k}+\sqrt{k-\frac12}\right)\left(\sqrt{k+\frac12}+\sqrt{k}\right)}\\ &\le\frac1{32\left(k-\frac12\right)^{5/2}}\end{align}$$ So evaluating the first error exactly and bounding the rest by the number of terms times the largest error, $$0\le\int_{\frac12}^{\frac{101}2}\frac{dx}{\sqrt x}-\sum_{k=1}^{50}\frac1{\sqrt k}\le0.03527618+49\cdot0.005681136=0.313651854$$ $$\int_{k-\frac12}^{k+\frac12}\frac{dx}{\sqrt x}=2\sqrt{\frac{101}2}-2\sqrt{\frac12}=12.798457$$ So $$12.484805\le\sum_{k=1}^{50}\frac1{\sqrt k}\le12.798457$$ So the sum of centered integrals was accurate enough to determine the floor of the sum. Of course, we had to accurately bound the error...

EDIT: There is a much less strenuous way to develop a useful error bound. Start with the Taylor series $$2\sqrt{k+\frac12}=2\sqrt k+2\frac{\left(\frac12\right)}{1!}\frac{\left(\frac12\right)}{k^{1/2}}+2\frac{\left(\frac12\right)\left(-\frac12\right)}{2!}\frac{\left(\frac12\right)^2}{k^{3/2}}+2\frac{\left(\frac12\right)\left(-\frac12\right)\left(-\frac32\right)}{3!}\frac{\left(\frac12\right)^3}{\xi_{2k}^{5/2}}$$ $$2\sqrt{k-\frac12}=2\sqrt k+2\frac{\left(\frac12\right)}{1!}\frac{\left(-\frac12\right)}{k^{1/2}}+2\frac{\left(\frac12\right)\left(-\frac12\right)}{2!}\frac{\left(-\frac12\right)^2}{k^{3/2}}+2\frac{\left(\frac12\right)\left(-\frac12\right)\left(-\frac32\right)}{3!}\frac{\left(-\frac12\right)^3}{\xi_{2k-1}^{5/2}}$$ With $\xi_{2k}>k+\frac12$ and $\xi_{2k-1}>k-\frac12$. Simplifying and subtracting, $$2\sqrt{k+\frac12}-2\sqrt{k-\frac12}=\frac1{\sqrt k}+\frac1{8\sqrt2}\left(\frac1{(2\xi_{2k-1})^{5/2}}+\frac1{(2\xi_{2k})^{5/2}}\right)$$ So that way our error estimate reads $$0\le\int_{\frac12}^{\frac{101}2}\frac{dx}{\sqrt x}-\sum_{k=1}^{50}\frac1{\sqrt k}=\frac1{8\sqrt2}\sum_{k=1}^{100}\frac1{(2\xi_k)^{5/2}}<\frac1{8\sqrt2}\sum_{k=1}^{100}\frac1{k^{5/2}}<\frac1{8\sqrt2}\sum_{k=1}^{\infty}\frac1{k^2}=\frac{\pi^2}{48\sqrt2}$$ So now we have $$12.798457-\frac{\pi^2}{48\sqrt2}=12.653064<\sum_{k=1}^{50}\frac1{\sqrt k}<12.798457$$ So this way we developed an analytical error bound of sufficient accuracy without any real numerical work, just a simple formula. This, along with evaluation of the primitive at the endpoints, was sufficient to establish that $$\left\lfloor\sum_{k=1}^{50}\frac1{\sqrt k}\right\rfloor=12$$

user5713492
  • 15,938
0

May be, you could have used $$\sum_{i=1}^n \frac 1 {\sqrt n}=H_n^{\left(\frac{1}{2}\right)}$$ where appear the generalized harmonic numbers and then have used their asymptotic expansions $$H_n^{\left(\frac{1}{2}\right)}=2 \sqrt{n}+\zeta \left(\frac{1}{2}\right)+\frac 1 {2\sqrt{n}}-\frac{1}{24n\sqrt{n}} +O\left(\frac{1}{n^{5/2}}\right)$$

For $n=50$, limiting to $O\left(\frac{1}{n^{1/2}}\right)$, this would give $\approx 12.6818$; limiting to $O\left(\frac{1}{n^{3/2}}\right)$, this would give $\approx 12.7525$; imiting to $O\left(\frac{1}{n^{5/2}}\right)$, this would give $\approx 12.7524$.