0

$$\lim_{n\to \infty} \sum_{i=1}^n \frac{1}{\frac{an}{b-a}+i},\ a,b\in\Bbb R\setminus\{0\},\ a\ne b$$

I know for a fact that the solution can be found via Laurent Series if that hint helps. I inserted a sample of the above with $a=1$ and $b=2$ on Wolfram and concluded that the result, as expected, is:

$$\ln\left|\frac{b}{a}\right|$$

The overall issue here is how to end up to the very above result!

Note: I'm a high senior school student.

Thank you!

  • Well, it's hard not to conclude if it says Laurent Series in the bottom left corner on the page in the thread. – PinkyWay Jun 04 '20 at 16:03
  • I thought it was a typo of mine, but there is $i$ indeed. – PinkyWay Jun 04 '20 at 16:04
  • Yes, gentlemen, for some reason an editor has massively messed with my question. I will repost the Sum here once again – dimpap5555 Jun 04 '20 at 18:55
  • $$\lim_{n\to \infty} \sum_{i=1}^n \frac{1}{\frac{an}{b-a}+i},\ a,b\in\Bbb R\setminus{0},\ a\ne b$$ As edited above, this is the actual sum. Thank you to the editor who messed with my post! Pretty polite of them! – dimpap5555 Jun 04 '20 at 19:16

2 Answers2

3

Use a Riemann sum, i.e. note that equal-width strips gives $\int_0^1f(x)dx=\lim_{n\to\infty}\frac1n\sum_{i=1}^nf(i/n)$ so$$\begin{align}\lim_{n\to\infty}\sum_{i=1}^{n}\frac{1}{\frac{an}{b-a}+i}&=\lim_{n\to\infty}\frac{1}{n}\sum_{i=1}^{n}\frac{1}{\frac{a}{b-a}+\frac{i}{n}}\\&=\int_{0}^{1}\frac{dx}{\frac{a}{b-a}+x}\\&=\left[\ln\left|\frac{a}{b-a}+x\right|\right]_{0}^{1}\\&=\ln\left|\frac{b}{b-a}\right|-\ln\left|\frac{a}{b-a}\right|\\&=\ln\left|\frac{b}{a}\right|.\end{align}$$

J.G.
  • 115,835
  • Thank you for your answer, however, I am mainly looking into doing the opposite, aka proving that through the Riemann approximation the generalized answer is $$\ln\left|\frac{b}{a}\right|$$ – dimpap5555 Jun 06 '20 at 14:29
  • @dimpap5555 How is that the opposite? I proved the sum has that value. – J.G. Jun 06 '20 at 15:28
  • If you were to pay attention to my saying, I am saying, I wish to find another way to prove the sum without using the Riemann approximation. What I wish to do is to prove that the integral of 1/x for a to b is ln(b/a) since the sum we are evaluating comes from the Reimann expression for f(x)= 1/x – dimpap5555 Jun 07 '20 at 12:20
  • 1
    @dimpap5555 Sorry, your previous word choice confused me. If you just want to know how to integrate $1/x$, see here & here, but it depends somewhat on which definitions you know of $e$, $\ln x$, $\exp x$ etc. – J.G. Jun 07 '20 at 12:46
  • I seek no integration or anything that has to do with integral calculus. I just want to solve the sum with conventional ways, while forgetting calculus exists. Think it like I am Riemann and wish to prove that the area from a to b for a function f(x)=1/x is ln(b/a)... since the sum comes from: $$ \lim_{n\to \infty} \sum_{i=1}^n f(x_{i})Δx, f(x)=\frac{1}{x}, Δx = \frac{b-a}{n}, x_{i}=a+Δx*i $$ – dimpap5555 Jun 08 '20 at 22:16
  • @dimpap5555 Well, that's different again from what you said you wanted. If you want a proof sums of $O(1/n)$ positive terms is $O(\log n)$, that's doable without calculus, since $\tfrac12\le\sum_{k=2^{n-1}}^{2^n-1}\tfrac1k\le1$. But if you want to know the sum uses a natural logarithm, we have to use calculus, because calculus is needed just to define $e$. – J.G. Jun 09 '20 at 08:05
  • (Well, technically $e$ can be defined with a limit, but I doubt that'd help here.) – J.G. Jun 09 '20 at 08:31
1

Before you changed your question you were asking to evaluate

\begin{equation}\label{Q} \lim_{n\to \infty} \sum_{i=1}^n \frac{1}{\frac{an}{b-a}+i} \tag{Q} \end{equation}

which is not the same as the series $\sum_{n=1}^\infty \frac{1}{\frac{an}{b-a}+i}$. The latter one clearly diverges by confrontation with $\sum_n \frac{1}{n}$.

The short answer for \ref{Q} above is that it is a riemann sum approximation of the integral $\int_{a}^{b} \frac{1}{x} dx$, but i guess you are willing to solve this only by elmentary methods.

Assume for simplicity $0<a<b$, then as you can check

\begin{equation}\label{in} \frac{b-a}{b} = 1-\frac{a}{b} < \ln(b/a) < \frac{b}{a}-1 = \frac{b-a}{a} \tag{1} \end{equation}

Now let $\displaystyle{x_i= a+(b-a)\frac{i}{n}}$. Note that $x_{i+1}-x_{i}=(b-a)/n$ and that the terms of the sum in your limit can be written as $$ \frac{1}{\frac{an}{b-a}+i} = \frac{b-a}{n} \frac{1}{a\frac{n-i}{n}+b\frac{i}{n}} = \frac{x_{i}-{x_{i-1}}}{x_i}$$

Now you can write, as follows from \ref{in},

$$\ln(x_{i+1}/x_{i}) <\frac{x_{i+1}-x_{i}}{x_{i}} = \frac{x_{i}-{x_{i-1}}}{x_i} < \ln(x_i/x_{i-1})$$

and get the following bound on the sums you are trying to get the limit of

$$\sum_{i=1}^n \frac{x_{i}-x_{i-1}}{x_{i}} < \sum_{i=1}^n \ln (x_{i}/x_{i-1}) = \ln(b/a)$$ $$\sum_{i=1}^n \frac{x_{i}-x_{i-1}}{x_{i}} > \sum_{i=1}^{n} \ln (x_{i+1}/x_{i}) = \ln(x_{n+1}/x_{1})=\ln\left(\frac{b+(b-a)/n}{{a+(b-a)/n}}\right)$$

At this point you only need to point out that $\ln\left(\frac{b+(b-a)/n}{{a+(b-a)/n}}\right) \rightarrow \ln(b/a)$ as $n\rightarrow \infty$, and use the "Squeeze theorem".

You can deduce what happens in the other cases from this. The cases in which $ab<0$ require some care.

Dunnò000
  • 683
  • Thank you for your answer, it has been something rather helpful. As you may see by yourself, I had my post changed again to the original sum, since it was messed with by an editor without my consent! I know it can be solved via the Definite Integral Interpretation, however as you also guessed,I am looking deeper into it, basically trying to prove that the integral derives from that. Even though you are descriptive enough, I have a question and most specifically, how do you end up to method #1. Why is the solution used, when it's the very result we are looking for? – dimpap5555 Jun 04 '20 at 20:07
  • To add up, if I am not mistaken are you using the following property: $$ \ln\frac{x+1}{x} <\frac{1}{x} < ln\frac{x}{x-1} $$ to basically prove method #1? – dimpap5555 Jun 04 '20 at 20:33
  • If you follow the computation it is all elementry algebra and the (1) inequality. You just get upper and lower bound for the n-th term of your sequence and note that both the upper and lower bounds have the same limit, in fact one is constant.

    Edit: yes the (1) inequality is just a rewriting of $$

    – Dunnò000 Jun 04 '20 at 20:42
  • Of course, I understand that however what I do not understand is why you use $\ln(b/a)$ as something known beforehand in (1) when it is the result we seek to find. Why in other words the result is used to prove... the result? – dimpap5555 Jun 04 '20 at 20:46
  • The (1) inequality is just a rewriting of $\frac{x}{x+1} < \log(1+x) < x$ for $x>0$ – Dunnò000 Jun 04 '20 at 21:07
  • And how do you end up to the last inequality $$\ln(x_{i+1}/x_{i}) <\frac{x_{i+1}-x_{i}}{x_{i}} = \frac{x_{i}-{x_{i-1}}}{x_i} < \ln(x_i/x_{i-1})$$ where the logrithms have taken the place of the left and right arguments? – dimpap5555 Jun 06 '20 at 14:31
  • $\ln(x_{i+1}/x_i) < \frac{x_{i+1}-x_i}{x_i}$ is an instance of $\ln(b/a) < \frac{b-a}{a}$, whereas $\frac{x_{i} -x_{i-1}}{x_i} < \ln(x_{i}/x_{i-1})$ is an instance of $\frac{b-a}{b} < \ln(b/a)$ – Dunnò000 Jun 06 '20 at 15:29
  • Gotcha! Thank you very much for your help! – dimpap5555 Jun 07 '20 at 12:23