1
$\begingroup$

The alternating harmonic series is 1-1/2+1/3-1/4+1/5-1/6+1/7-1/8+... I am trying to see if I can eliminate the last N/4 positive terms.

For instance, if the above were just a partial sum of the AHS (no "+..."), I would eliminate 1/5 and 1/7. When the number of terms grows to infinity, the number of terms I need to lop off has an infinite number of terms as well (N/4). Strangely enough, this series converges (and it's greater than zero) even though it belongs to the divergent series of all positive terms. But from a theoretical point of view I am a bit stumped because as N tends to infinity, I need to define a series of positive terms $\lim_{N\to \infty}\sum_{n=1}^\frac{N}{4} \frac{1}{(N-2n+1)}$ where none of the terms seem to exist because I cannot even define which would be the first term. What do you think?

(the terms of this series are as described for any N multiple of 4,

$\endgroup$
  • $\begingroup$ The terms of this series are as described for any N multiple of 4, zero otherwise. Note that the first term would be, for n=1, $\frac{1}{(N-1)}$ and the last term, for n=N/4, $\frac{1}{(N/2+1)}$ so both are infinitely far away in the AHS. $\endgroup$ – paul hawkey Aug 30 '16 at 19:34
2
$\begingroup$

By Riemann sums $$\lim_{M\to +\infty}\sum_{n=1}^{M}\frac{1}{4M-2n+1}=\int_{0}^{1}\frac{dx}{4-2x}=\color{red}{\log\sqrt{2}}$$ as well as $$ \sum_{1\leq n\leq\frac{N}{4}}\frac{1}{N-2n+1}=\log\sqrt{2}+O\left(\frac{1}{N}\right) $$ no matter what $N\pmod{4}$ is, as soon as $N$ is large enough.

$\endgroup$
  • $\begingroup$ Thank you! although I think is ln, not log. $\endgroup$ – paul hawkey Aug 30 '16 at 21:29
  • 1
    $\begingroup$ @paulhawkey Mathematicians often use $\log$ to denote the logarithm with base $e$. $\endgroup$ – Did Aug 30 '16 at 21:50
0
$\begingroup$

$\newcommand{\angles}[1]{\left\langle\,{#1}\,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{\mathrm{i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\Li}[1]{\,\mathrm{Li}_{#1}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$ \begin{align} &\color{#f00}{\lim_{N \to \infty} \sum_{n = 1}^{\left\lfloor\, N/4\, \right\rfloor}{1 \over N - 2n + 1}} = \lim_{N \to \infty} \sum_{n = 1}^{\left\lfloor\, N/4\, \right\rfloor}\int_{0}^{1}x^{N - 2n} \,\,\,\dd x = \lim_{N \to \infty}\,\,\int_{0}^{1}x^{N} \sum_{n = 1}^{\left\lfloor\, N/4\, \right\rfloor}\pars{x^{-2}}^{n} \,\,\,\dd x \\[5mm] & = \lim_{N \to \infty}\,\,\int_{0}^{1} {x^{N - 2\left\lfloor\, N/4\, \right\rfloor}\,\,\,\,\,\, -\,\,\, x^{N} \over 1 - x^{2}}\,\,\,\dd x = \half\,\lim_{N \to \infty}\,\,\int_{0}^{1} {x^{N/2 - \left\lfloor\, N/4\, \right\rfloor - 1/2}\,\,\,\,\,\, - \,\,\, x^{N/2 - 1/2} \over 1 - x}\,\,\,\dd x \\[5mm] & = \half\,\lim_{N \to \infty}\,\,\,\bracks{\Psi\pars{{N \over 2} + \half} - \Psi\pars{{N \over 2} - \left\lfloor\,{N \over 4}\, \right\rfloor + \half}\vphantom{\Huge A^{a}}}\quad\quad \pars{~\Psi:\ Digamma\ Function~} \end{align}


\begin{align} &\color{#f00}{\lim_{N \to \infty} \sum_{n = 1}^{\left\lfloor\, N/4\, \right\rfloor}{1 \over N - 2n + 1}} = \half\,\lim_{N \to \infty}\, \ln\pars{N/2 \over N/2 - \left\lfloor\, N/4\, \right\rfloor} \\[5mm] = &\ -\,\half\,\lim_{N \to \infty} \ln\pars{1 - {\left\lfloor\, N/4\, \right\rfloor \over N/2}} = -\,\half\,\ln\pars{1 - \half} = \color{#f00}{\half\,\ln\pars{2}} \approx 0.3466 \end{align}

Note that $\ds{\Psi\pars{z + \half} \sim \ln\pars{z} + {1 \over 24z^{2}} - {7 \over 960z^{4}} + {31 \over 8064z^{6}} + \cdots\quad}$ as $\ds{\verts{z} \to \infty}$ with $\ds{\,\verts{\mrm{arg}\pars{z}} < \pi}$.

$\endgroup$
  • $\begingroup$ @paulhawkey You're welcome. Glad it was helpful to you. $\endgroup$ – Felix Marin Aug 30 '16 at 21:23
  • $\begingroup$ As witnessed by the "miraculous" cancellations at the end, none of this détour is necessary, really. To invoke the digamma function in this context is (tiring and) to use a sledgehammer to kill a fly. $\endgroup$ – Did Aug 30 '16 at 21:49
  • $\begingroup$ @Did You are quite right. However, the 'easy one' was already ( or simultaneously ) worked out by $\texttt{Jack D'Aurizio}$. $\endgroup$ – Felix Marin Aug 30 '16 at 21:54
  • $\begingroup$ Since you are systematically flooding the site with these digamma-tricks and a few related ones (and with next to nothing else), @Jack's answer is not the reason you posted yours and you know it. So much for candor. $\endgroup$ – Did Aug 30 '16 at 22:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.