Wednesday, 6 May 2015

calculus - Various evaluations of the series suminftyn=0frac(1)n(2n+1)3




I recently ran into this series:
n=0(1)n(2n+1)3



Of course this is just a special case of the Beta Dirichlet Function , for s=3.



I had given the following solution:
1133+153=n=0(1)n(2n+1)3()=(1+153+193+)(133+173+1113+)=n=01(4n+1)3n=01(4n+3)3=1243ψ(2)(14)+1243ψ(2)(34)=1243[ψ(2)(114)ψ(2)(14)]=1243[2π3cotπ4csc2π4]=π3cotπ4csc2π443=π332



where I used polygamma identities and made use of the absolute convergence of the series at () in order to re-arrange the terms.



Any other approach using Fourier Series, or contour integration around a square, if that is possible?


Answer



Method by Fourier Series




Consider the function f(x)=x(1x), 0x1. It has Fourier sine series expansion



f(x)=8π3n=11(2n1)3sin(2n1)πx.



Setting x=12 results in



14=8π3n=1(1)n1(2n1)3,



or




π332=n=1(1)n1(2n1)3.



By reindexing the sum we can write



π332=n=0(1)n(2n+1)3.



Method by Contour Integration



Let g(z)=1(2z1)3. Then g has only one pole of order 3 at z=12. Let N be a positive integer, and consider the contour integral




12πiΓNπcscπzg(z)dz,



where ΓN is a positively oriented square with vertices at (N+12)(±1±i). The residue theorem gives



12πiΓNπcscπzg(z)dz=Nn=NResz=nπcscπzg(z)+Resz=12πcscπzg(z)=Nn=N(1)ng(n)+π316.



For |z|1, |g(z)||z|3. Thus, 12πiΓNπcscπzg(z)dz0asN.




Hence



0=n=(1)ng(n)+π316



that is,



π316=n=(1)n1(2n1)3.



Now




n=(1)n1(2n1)3=0n=(1)n1(2n1)3+n=1(1)n1(2n1)3=n=0(1)n(2n+1)3+n=1(1)n1(2n1)3=2n=0(1)n(2n+1)3.



Thus



π316=2n=0(1)n(2n+1)3.




Finally, we have



π332=n=0(1)n(2n+1)3.


No comments:

Post a Comment

real analysis - How to find limhrightarrow0fracsin(ha)h

How to find lim without lhopital rule? I know when I use lhopital I easy get $$ \lim_{h\rightarrow 0}...