This article has been edited for a bounty.

$C$ MRB, the MRB constant, is defined at http://mathworld.wolfram.com/MRBConstant.html .

There is an excellent 56 page paper whose author has passed away. You can find it in Google Scholar "MRB constant," Better yet, use the following link http://web.archive.org/web/20130430193005/http://www.perfscipress.com/papers/UniversalTOC25.pdf. You find a cached copy there.

Just before the author, Richerd Crandall, died I wrote him about a possible small error. What I'm worried about is formula 44 on page 29 and below. When I naively worked formula 44 it needed a negative sign in front of it. Crandall did write me back admitting to a typo, but he died before he had a chance to correct it. Is there anyone out there competent enough to check, correct and prove the corrected formulas for me? Thank you. I will use the proofs often and try to get the formulas published more.


part1 Here is how I worked formula 44 and got -B:

(*define the Dirichlet eta function*)

eta[s_] := (1 - 2^(1 - s)) Zeta[s];

(*define the higher derivatives of the eta(0)*)

a[i_] := Derivative[i][eta][0];

(*Define c:*)

c[j_] := Sum[Binomial[j, d](-1)^dd^(j - d), {d, 1, j}]

(*formula (44)*)

N[Sum[c[m]/m!*a[m], {m, 1, 40}], 100]

  • $\begingroup$ Would it be possible for you to type in the formula from the paper, and indicate the suspected typo? I don't care to download a 56(-page?) paper just to check one formula. $\endgroup$ – Gerry Myerson Nov 2 '13 at 23:36
  • $\begingroup$ @Gerry Myerson , I added the formula $\endgroup$ – Marvin Ray Burns Nov 3 '13 at 14:23
  • $\begingroup$ Thanks. Not enough info there to tell --- in particular, you'd have to know what $\eta$ stands for. Sorry --- I hope someone else will download the paper, and give it a try. $\endgroup$ – Gerry Myerson Nov 3 '13 at 22:10
  • $\begingroup$ That doesn't tell me what $\eta$ stands for. $\endgroup$ – Gerry Myerson Nov 4 '13 at 2:39
  • 3
    $\begingroup$ The context indicates it is the Dirichlet eta function. $\endgroup$ – anon Nov 5 '13 at 16:03

The first equality can be reproduced by formally stating the double sum: write down the expanded exponantial series for each term in one row, sum then column-wise to get the derivatives of the $\eta()$ and sum then the $\eta()$-expressions. Here the $\eta(s)$ is the alternating $\zeta(s)$ and $\eta^{(m)}(s)$ the $m$'th derivative.

$$ \begin{array} {rclll} -\exp( {\log(1) \over 1})+1 & = & -{\log(1) \over 1} &-{\log(1)^2 \over 1^2 2!} &-{\log(1)^3 \over 1^3 3!} & - \cdots \\ +\exp( {\log(2) \over 2})-1 & = &+{\log(2) \over 2} &+{\log(2)^2 \over 2^2 2!} &+{\log(2)^3 \over 2^3 3!} & + \cdots \\ -\exp( {\log(3) \over 3})+1 & = &-{\log(3) \over 3} &-{\log(3)^2 \over 3^2 2!} &-{\log(3)^3 \over 3^3 3!} & - \cdots \\ \vdots \qquad & \vdots & \quad\vdots & \quad\vdots& \quad\vdots & \ddots \\ \hline \\ B \qquad & = & {\eta^{(1)}(1) \over 1!} &- {\eta^{(2)}(2) \over 2!} &+ {\eta^{(3)}(3) \over 3!} & - \cdots \end{array}$$

I've not yet the expansion into the formula with the derivatives of $\eta()$ at zero; perhaps I can supply that tomorrow.

Numerical check

In Pari/GP I get for the original sum $$ B = \sum_{k=1}^\infty (-1)^k( \exp( \log(k)/k)-1)$$

                           B=sumalt(k=1,(-1)^k * (exp( log(k)/k) -1)      

well converging the value


as well for

                         B = -sum(k=1,6,(-1)^k*aetad(k,k)/k!)          

(but to fewer digits).
My user-defined function aetad(s,d) gives the d'th derivative of $\eta()$ at $s$ and I implemented that aeta() using the $\eta()$ / $\zeta()$ conversion formula

              aeta(s) = if(s==1,return(log(2));return( (1-2/2^s)*zeta(s)  )
                      z= sum(k=0,d,  (-1)^k*binomial(d,k)*aeta(s +(d/2-k)*h1));
                         return( z/h1^d); }

I have also another implementation of the $\eta()$ and its derivatives which allows to access higher derivatives (in the above version the 10'th or even 20'th derivative were impractical). This version is also independent of the software-included zeta-function but uses Pari/GP's sumalt() procedure which allows to evaluate alternating (moderately) divergent series. My approach computes the taylor-expansion for the $\eta()$ function by default centered around zero but can easily be generalized to any center:

fmt(200,12)  \\ user defined; set internal precision to 200 digits, display to 12 digits
default(seriesexpansion,24)          \\ set expansion-limit for power series
m_coeffs = vectorv(24);              \\ vector for 24 aeta-taylor-series
    tay_aeta = sumalt(k=0,(-1)^k/(1+k)^(x+m));  \\ taylorseries for aeta around m
    m_coeffs[1+m]= polcoeffs(tay_aeta,24);      \\ store leading 24 taylorcoefficients
  m_coeffs = Mat(m_coeffs);}        \\ make a matrix from the list of taylor series

After that, the desired coefficients (the derivatives of the $\eta(m)$ ) nicely divided by the factorials, occur on the diagonal of the coefficients-matrix, and we only need sum them with alternating sign:

   B = sum (k=1,23, -(-1)^k*m_coeffs[1+k,1+k])
   \\ 0.1878596424620671202485179340542732... the leading correct digits

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.