1
$\begingroup$

According to Beyer (1987) the following progression is called harmonic series: $$\frac {1} {a_{1}},\frac {1} {a_{1}+d}, \frac {1} {a_{1}+2d},...$$

How it can be calculated the partial sum of the above mentioned sequence?

$\endgroup$
  • $\begingroup$ I don´t see that there exists a closed form for the partial sum. $\endgroup$ – callculus Nov 17 '18 at 16:03
  • $\begingroup$ @callculus what about approximation? $\endgroup$ – David Nov 17 '18 at 16:03
  • 1
    $\begingroup$ Search for "Harmonic numbers". Note however, that your sequence is more general, related to "Digamma function" $\endgroup$ – Yuriy S Nov 17 '18 at 16:06
  • $\begingroup$ @YuriyS Thank you for your comment. Can you please provide some material reagrdinf this problem? $\endgroup$ – David Nov 17 '18 at 16:10
3
$\begingroup$

The sum of terms for $n=0,1,...,m$ is $$\sum_{n=0}^m \frac{1}{d(a_1/d+n)} =\frac{1}{d} \left(\psi \left(m+1+\frac{a_1}{d}\right) -\psi \left(\frac{a_1}{d}\right) \right)$$

Where $\psi$ is the Digamma function.

Use the definition:

$$\psi(z)=-\gamma+\sum_{n=0}^\infty \frac{z-1}{(n+1)(n+z)}$$

$\endgroup$
  • $\begingroup$ Thank you for your answer. Could you please provide some information how it is derived? P.S. sorry, if my comment is weird, i don't have background in this topic. $\endgroup$ – David Nov 17 '18 at 16:23
  • $\begingroup$ @David, try using the definition I provided, and you will see that the formula is true $\endgroup$ – Yuriy S Nov 17 '18 at 16:25
  • $\begingroup$ Thank you. My aim is to find a closed-form expression for the finite series. Is it possible? $\endgroup$ – David Nov 17 '18 at 16:31
  • $\begingroup$ @David, I have written it in my answer above $\endgroup$ – Yuriy S Nov 17 '18 at 16:38
  • $\begingroup$ The Digamma function is expressed in terms of infinite sum, but my aim is to find closed-form expression. Thank you. $\endgroup$ – David Nov 17 '18 at 16:52
2
$\begingroup$

Yuriy S gave the only possible closed form for the summation.

For large $n$, for sure, you could use series expansions and get from $$S_n=\sum_{i=0}^n \frac{1}{a+i\,d}=\frac 1d \left(\psi \left(n+1+\frac{a}{d}\right)-\psi \left(\frac{a}{d}\right) \right)$$ $$S_n=\frac{\log(n)-\psi \left(\frac{a}{d}\right)}{d}+\frac{2 a+d}{2 d^2 n}-\frac{6 a^2+6 a d+d^2}{12 d^3 n^2}+\frac{2 a^3+3 a^2 d+a d^2}{6 d^4 n^3}+O\left(\frac{1}{n^4}\right)$$

Using $a=\pi$, $d=e$ and $n=10$, the exact result would be $\approx 1.03107$ while the above series expansion would give $\approx 1.03113$.

$\endgroup$
  • $\begingroup$ Please have a look. Thank you for your support. math.stackexchange.com/questions/3003278/… $\endgroup$ – David Nov 18 '18 at 8:56
  • $\begingroup$ @David. No idea about this other one. I am quite skeptical about even an approximation. Even the case $a=\alpha=0$ makes the problem more than difficult (at least to me). $\endgroup$ – Claude Leibovici Nov 18 '18 at 9:11
  • $\begingroup$ Do you mean both of them? $\endgroup$ – David Nov 18 '18 at 9:12
1
$\begingroup$

The Harmonic Series

The partial sum of the standard Harmonic Series is given by $$ H_n=\sum_{k=1}^n\frac1k\tag1 $$ This can be extended to a function that is analytic except at the negative integers $$ H(z)=\sum_{k=1}^\infty\left(\frac1k-\frac1{k+z}\right)\tag2 $$ The Euler-Maclaurin Sum Formula gives the asymptotic expansion $$ H_n=\log(n)+\gamma+\frac1{2n}-\frac1{12n^2}+\frac1{120n^4}-\frac1{252n^6}+\frac1{240n^8}-\frac1{132n^{10}}+O\!\left(\frac1{n^{12}}\right)\tag3 $$ where $\gamma=0.57721566490153286060651209$ is the Euler-Mascheroni Constant.


The Series in the Question $$ \begin{align} \sum_{k=1}^n\frac1{a+(k-1)d} &=\frac1d\sum_{k=1}^n\frac1{\frac ad+k-1}\\ &=\frac1d\sum_{k=1}^\infty\left(\frac1{\frac ad+k-1}-\frac1{\frac{a-d}d+k-1+n}\right)\\[3pt] &=\frac1d\left(H\!\left(n+\frac ad-1\right)-H\!\left(\frac ad-1\right)\right)\tag4 \end{align} $$ using the extension in $(2)$.

Formula $(2)$ from this answer allows us to compute $H\!\left(\frac ad-1\right)$ for rational $a$ and $d$, while formula $(3)$ above allows us to approximate $H\!\left(n+\frac ad-1\right)$ for large $n$.

$\endgroup$
  • $\begingroup$ Thank you for your answer. $\endgroup$ – David Nov 18 '18 at 20:34
  • $\begingroup$ Please have a look. math.stackexchange.com/questions/3003278/… $\endgroup$ – David Nov 18 '18 at 20:35
  • $\begingroup$ Could you, please, support in order to calculate or approximate the sums represented in the above mentioned link? $\endgroup$ – David Nov 18 '18 at 20:36

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.