Group Abstract Group Abstract

Message Boards Message Boards

0
|
4K Views
|
8 Replies
|
3 Total Likes
View groups...
Share
Share this post:

Infinite series works only to order k=20

Hi Here is a function g defined by an infinite series which works only for k<= 20. The series is convergent as we check it by another mean

Needs["PlotLegends`"];
g[j_] :=  (1/2)^0.5*Sin[Pi*0.5]/Pi*     
   NSum[Gamma[-k + 0.5, 0, 10]/k!*(j/2)^(2 k), {k, 0, 30}];
Omega = Table[g[j], {j, 0.1, 20, 0.1}];
ListLinePlot[{Omega}, PlotStyle -> Thick, PlotRange -> {-2, 2}]
POSTED BY: MEK MUS
8 Replies

$\sum _{k=0}^{\infty } -\frac{2^{-2 k} j^{2 k} \Gamma \left(\frac{1}{2}-k,10\right)}{\Gamma (1+k)}+\sum _{k=0}^{\infty } -\frac{2 j^{2 k} \Gamma (-2 k) \sin (k \pi )}{\sqrt{\pi }}$

Sum[#, {k, 0, Infinity}] & /@ (Gamma[-k + 1/2, 0, 10]/k!*(j/2)^(2 k) //
    FunctionExpand)
 (*Sum[-((j^(2*k)*Gamma[1/2 - k, 10])/(2^(2*k)*Gamma[1 + k])), {k, 0, Infinity}] + 
  Sum[(-2*j^(2*k)*Gamma[-2*k]*Sin[k*Pi])/Sqrt[Pi], {k, 0, Infinity}]*)

First series we can compute from identity $\Gamma (a,z)=\int_z^{\infty } t^{a-1} \exp (-t) \, dt$

Integrate[
 Sum[-((2^(-2 k) j^(2 k) (t^(a - 1) Exp[-t] /. a -> 1/2 - k))/
   Gamma[1 + k]), {k, 0, Infinity}], {t, 10, Infinity}, Assumptions -> j >= 0]

(*-(1/2) E^(-I j) Sqrt[\[Pi]] (Erfc[(20 - I j)/(2 Sqrt[10])] + 
   E^(2 I j) Erfc[(20 + I j)/(2 Sqrt[10])])*)

Second series we can compute like this:

L = Table[
  Limit[-((2 j^(2 k) Gamma[-2 k] Sin[k \[Pi]])/Sqrt[\[Pi]]), k -> m, 
   Assumptions -> j >= 0], {m, 0, 10}]
Sum[FindSequenceFunction[L, k] // FunctionExpand, {k, 0, Infinity}]

(*Sqrt[\[Pi]] Cos[j]*)

Regards M.I.

POSTED BY: Mariusz Iwaniuk

Looks like: $\sum _{k=0}^{\infty } \frac{\Gamma \left(-k+\frac{1}{2},0,10\right) \left(\frac{j}{2}\right)^{2 k}}{k!}=-\frac{1}{2} e^{-i j} \sqrt{\pi } \left(\text{erfc}\left(\frac{20-i j}{2 \sqrt{10}}\right)+e^{2 i j} \text{erfc}\left(\frac{20+i j}{2 \sqrt{10}}\right)\right)+\sqrt{\pi } \cos (j)$

g[j_] := (1/2)^(1/2)*Sin[Pi*(1/2)]/Pi*
   Sum[Gamma[-k + 1/2, 0, 10]/k!*(j/2)^(2 k), {k, 0, 50}];
Omega = Table[g[j], {j, 1/10, 20, 2/10}];
L = ListLinePlot[{Omega}, PlotStyle -> Thick, PlotRange -> {-2, 2}]

g1[j_] := (1/2)^(1/2)*
  Sin[Pi*(1/2)]/
   Pi*(-(1/2) E^(-I j)
      Sqrt[\[Pi]] (Erfc[(20 - I j)/(2 Sqrt[10])] + 
       E^(2 I j) Erfc[(20 + I j)/(2 Sqrt[10])]) + 
    Sqrt[\[Pi]]*Cos[j]); Omega1 = Table[g1[j], {j, 1/10, 20, 2/10}];
L2 = ListLinePlot[{Omega1}, PlotStyle -> {Red, Thick}, 
  PlotRange -> {-2, 2}]

L3 = ListLinePlot[{Omega - Omega1}, PlotStyle -> {Red, Thick}]
POSTED BY: Mariusz Iwaniuk

Hi,

Could you please share the proof of the convergence? I did quick asymptotic analysis on the last term of your sum and it appears the absolute value grows exponential. Both the asymptotic value and direct computation show the same trend.

enter image description here

POSTED BY: Shenghui Yang
Posted 2 years ago

Thank you M.I for your instantaneous help. I appreciate the the way you compute the second series. I however have some remarks about it.

  • First the sequence you get L[k] from the first 10 terms may not be unique so one has to check the equality of the sequence with the original one -2j^(2k) Gamma[-2k]Sin[kPi]/Sqrt[Pi]
  • Second the sequence you get start from zero Table[L[k],{k,0,10}]=0,Sqrt[[Pi]], -(1/2) j^2 Sqrt[[Pi]],...... I don't know why but that will not impact the final result you get . To prove the equality one has to make the shift k->k+1 first which result in the sequence (-1)^kj^(2k)Sqrt(Pi)/Gamma(1+2k) then apply the identity Gamma[a]Gamma[1-a]=Pi/Sin[Pi*a] to get the equality ( I do it analytically because I' not not good in mathematica)
POSTED BY: Updating Name

It's about the sum you give at the beginning : Looks like ...... If I ask Mathematica to perform the sum naively that is if I write

Sum[Gamma[-k - 1/2, 0, 10]/k!*(j/2)^(2 k), {k, 0, [Infinity]}] I get no answer . My question is how did you get the above result: -1/2 exp[-ij]......by using Mathematica software then how or have you copied it from some textbook

POSTED BY: MEK MUS

What do you mean by

Will you please give some hints about how you proceeded

?

POSTED BY: Mariusz Iwaniuk

Thanks for the answer I understand that the terms of the series k>=50 have insignificant contributions to the sum. On the other hand I didn't succeed in reproducing the series giving at the beginning for fixed j. Will you please give some hints about how you proceeded Thanks again

POSTED BY: MEK MUS

Hi
Please execute first

Needs["PlotLegends`"];
g[j_] :=  (1/2)^0.5*Sin[Pi*0.5]/Pi*     
   NSum[Gamma[-k + 0.5, 0, 10]/k!*(j/2)^(2 k), {k, 0, 30}];
Omega = Table[g[j], {j, 0.1, 20, 0.1}];
ListLinePlot[{Omega}, PlotStyle -> Thick, PlotRange -> {-2, 2}]

to see what's going on. Note that you made a print mistake it's (j/2)^(2 k and not (k/2)^(2 k. Now I said convergence because I plotted the same function but written in terms of reduced Bessel function and found no errors.
The aim is to make the plot with k=0 to infinity.
Thanks.

POSTED BY: MEK MUS
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard