Relative Convergence Rate

In[]:=
DecimalapproximationoftheMRBconstant
Out[]=
approximationconstantDecimalMRBofthe
In[]:=
m=WolframAlpha["MRB constant",{{"DecimalApproximation",1},"ComputableData"}]
Out[]=
0.1878596424620671202485179340542732300559030949001387861720046840894772315646602
In[]:=
Precision[m]
Out[]=
79.

Error using 10^n terms

In[]:=
p=Table[m-NSum[
n
(-1)
(
1/n
n
-1),{n,10^x},WorkingPrecision200,PrecisionGoal200,Method"AlternatingSigns"],{x,6,33}]//N//Short
Out[]//Short=
{-6.9078×
-6
10
,-8.05905×
-7
10
,-9.21034×
-8
10
,23,-3.68414×
-31
10
,-3.79927×
-32
10
}
In[]:=
c1=ActivateLimit
n
(-1)
n!
Derivative[n][DirichletEta][x]/.n1,x1,Unevaluated[Sum]
Out[]=
1
2
Log[2](-2EulerGamma+Log[2])
In[]:=
f1[n_]:=
n
(-1)
n!
Derivative[n][DirichletEta][n];

Error using just n terms of the first Crandall formula

In[]:=
(q=Table[m+(c1+Sum[f1[n],{n,2,x}]),{x,6,33}]//N)//Short
Out[]//Short=
{1.87833×
-8
10
,4.54955×
-10
10
,24,-3.21255×
-52
10
,-3.5388×
-54
10
}

An average ratio of the largest partial sums by each method

In[]:=
me=MantissaExponent;
In[]:=
s0=Table[me[q[[-n]]][[2]]/me[p[[-n]]][[2]]//N,{n,2,26}]//Mean
Out[]=
1.61795
In[]:=
GoldenRatio//N
Out[]=
1.61803
In[]:=
t1="For partial sums ranging from 10^6 to 10^33 terms, Crandall's formula converges ";​​t2="approximately a factor of the golden ratio faster per decade increase in the number of terms than direct summation.";​​Print[​​Style[t1,Red],​​Style[GoldenRatio/s0//PercentForm,Large],​​Style[t2,Blue]​​]
For partial sums ranging from 10^6 to 10^33 terms, Crandall's formula converges 100%approximately a factor of the golden ratio faster per decade increase in the number of terms than direct summation.
In[]:=
​