0

So, I was trying to find a general pattern/formula for the Dirichlet Regularization of a Divergent series. I hadn't faced any problems until I reached:

$$\sum_{n=0}^\infty n^x$$

So, let's make a function $\operatorname{Di}$, which basically gives the Dirichlet regularization for any summation put in (if valid). The value $x$ here is nothing but a constant, and plays a similar role as $a$ in $ax^2 + bx + c = 0$.

Now, if I try to do this calculation for:

$$\operatorname{Di}\left(\sum_{n=0}^\infty n^x\right)$$

Here's where I arrive at the problem. For even values of $x$, the pattern seems to be simple, it is just $\frac{-1}{n+1}$, the problem arrives for odd values of $x$.

For even values, the final regularized result seems to be:

$$\frac{-1}{3}, \frac{-1}{5}, \frac{-1}{7}, \ldots, \frac{-1}{x+1}$$

However, there does not seem to be some set pattern for odd values of $x$.

The values seem to be chaotic, and below is a list of results for all odd values $1$ to $20$:

x  |  Regularized Sum
---|----------------------
1  |  5/12
3  |  31/120
5  |  41/252
7  |  31/240
9  |  61/660
11 |  3421/32760
13 |  -1/84
15 |  4127/8160
17 |  43069/14364
19 |  174941/6600

I would like to know the pattern, a general formula based on the value of $x$ for this, and the reason why it isn't negative unlike the values for when $x$ is even, or when $x$ is $13$.

1 Answers1

1

Take a look at this post of the famous mathematician Terry Tao : https://terrytao.wordpress.com/2010/04/10/the-euler-maclaurin-formula-bernoulli-numbers-the-zeta-function-and-real-variable-analytic-continuation/

You will find that the denominators of the fractions you find agree with the formula

$$\frac{B_{s+1}}{s+1}$$

of this document (where the $B_n$ are the Bernoulli numbers) but there is no such agreement with the numerators.

Edit:

In fact, with your OEIS finding, the right formula :

$$\frac{1-B_{n+1}}{n+1}$$

is the good one ; we can check it on the first results (in red the successive Bernoulli numbers).

$$\begin{array}{|r|r|r|} \hline 1&\tfrac{5}{12}&\left(1-\left(\color{red}{+\tfrac{1}{6}}\right)\right)\tfrac{1}{2} \\ 3&\tfrac{31}{120}&\left(1-\left(\color{red}{-\tfrac{1}{30}}\right)\right)\tfrac{1}{4}\\ 5&\tfrac{41}{252}&\left(1-\left(\color{red}{+\tfrac{1}{42}}\right)\right)\tfrac{1}{6}\\ 7&\frac{31}{240}&\left(1-\left(\color{red}{-\tfrac{1}{30}}\right)\right)\tfrac{1}{8}\\ 9&\tfrac{61}{660}&\left(1-\left(+\tfrac{5}{66}\right)\right)\tfrac{1}{10}\\ 11&\tfrac{3421}{32760}&\left(1-\left(-\tfrac{691}{2730}\right)\right)\tfrac{1}{12}\\ 13&-\tfrac{1}{84}&\left(1-\left(+\tfrac{7}{6}\right)\right)\tfrac{1}{14}\\ 15&\tfrac{4127}{8160}&\left(1-\left(-\tfrac{3617}{510}\right)\right)\tfrac{1}{16}\\ \hline \end{array}$$

Jean Marie
  • 81,803
  • 1
    Thanks a lot for providing this information. So basically, there is a known pattern for the denominators, but none for the numerator? That seems rather odd, but I can understand. – Tsar Asterov XVII Dec 01 '22 at 12:38
  • Ok, I am confused. Does this only work for odd values of x? Because, the problem is, when I try to apply this calculation for equal values of x (or s here), it seems to suggest the denominator as 0, which doesn't make sense, as this means that the regularization value would be undefined, as you can't divide by 0, but it isn't, so it is quite confusing. I would appreciate some clarification with this issue. – Tsar Asterov XVII Dec 03 '22 at 14:53
  • In fact it is the Bernoulli numbers with an even index that I should have mentionned (all the odd-index Bernoulli numbers are $0$ except $B_1$) – Jean Marie Dec 03 '22 at 18:24
  • I can see what you mean, but a little bit of changing the equation works out. What I did find however is a link to an actual pattern for the numerators. https://oeis.org/A162173 I'm curious, is this correct? – Tsar Asterov XVII Dec 04 '22 at 13:09
  • Once more, OEIS site is very useful ! Yes, it's correct : in this way, your formula boils down plainly to $1-\frac{B_{n+1}}{n+1}$... – Jean Marie Dec 04 '22 at 13:33
  • So basically, the formula you have given in this comment is the entire formula? Numerator and Denominator? – Tsar Asterov XVII Dec 04 '22 at 13:43
  • Yes, exactly, I have checked. – Jean Marie Dec 04 '22 at 13:44
  • Jean Marie, thanks a lot for the help! I genuinely needed this a lot! Would you like to be credited in what I'm working on? – Tsar Asterov XVII Dec 04 '22 at 13:45
  • My contribution as been rather modest, don't you think ? – Jean Marie Dec 04 '22 at 13:58
  • Wait, Jean Marie, I have been trying out the formula, but the result doesn't seem to work out. For example, in $n^3$ the result should be 31/120, but according to the formula, the result is 121/120, so it doesn't seem to be quite right. What went wrong? – Tsar Asterov XVII Dec 04 '22 at 14:00
  • I'ù going to write it in my answer. – Jean Marie Dec 04 '22 at 14:03
  • Jean Marie, do you have the answer ready? I don't want to rush you, but I just want to know. I kind of need the formula, that's why. – Tsar Asterov XVII Dec 04 '22 at 15:28
  • Just done : there is a perfect coincidence... in particular for 31/120 – Jean Marie Dec 04 '22 at 15:39
  • Ok then, I'll be waiting. Again, thanks for the help. I genuinely appreciate it. – Tsar Asterov XVII Dec 04 '22 at 15:43
  • I have included the table as an Edit in my answer. – Jean Marie Dec 04 '22 at 15:44
  • 1
    Oh, I got it now, so it is supposed to be $\frac{1-B_(n+1)}{x+1}$. I got it now, thanks! – Tsar Asterov XVII Dec 04 '22 at 15:47
  • Sorry for the parentheses error !!! – Jean Marie Dec 04 '22 at 15:49
  • Thats fine haha. Thanks a lot comrade, you just saved me multiple sleepless nights. I can't thank you enough. I genuinely appreciate this community helping me out with my dumb questions. Again, thanks a lot, and I genuinely can't help but give you credit. Even the smallest contribution is a contribution. – Tsar Asterov XVII Dec 04 '22 at 15:55
  • It's a pleasure to help people like you ! – Jean Marie Dec 04 '22 at 15:57
  • 1
    The pleasure is mine to be able to get knowledge from smart folk like you. Now, I'll be on my way to add the formula and then continue onward! Guten Tag! It's people like you that make working with math a fun task and add to the math community. – Tsar Asterov XVII Dec 04 '22 at 15:59