Re: Integrals of Fourier Series
- Subject: [mg3207] Re: [mg3179] Integrals of Fourier Series
- From: jpk at apex.mpe.FTA-Berlin.de (Jens-Peer Kuska)
- Date: 18 Feb 1996 01:11:02 -0600
- Approved: usenet@wri.com
- Distribution: local
- Newsgroups: wri.mathgroup
- Organization: Wolfram Research, Inc.
- Sender: daemon at wri.com
Dear George Oster,
the situation is not so simple as it seems.
a) there is no finite algorithm to deal with
products of infinite sums
b) to leave (Sum[_,_]*Sum[_,_]) unevaluatet you must
supply different a summation index.
The rest is easy
(* my version of Your series with a named index *)
In[1]:=
u[x_,n_]:=
Sum[A[n]*Sin[n*Pi*x/L],{n,1,Infinity}]
(* Do the derivative *)
In[2]:=
u2=D[u[x,n],{x,2}]
Out[2]=
2 2 n Pi x
n Pi A[n] Sin[------]
L
Sum[-(-----------------------),
2
L
{n, 1, Infinity}]
(* calculate the square and replace one summation index *)
In[3]:=
sqru=u2*(u2 /. n->k)
Out[3]=
2 2 k Pi x
k Pi A[k] Sin[------]
L
Sum[-(-----------------------),
2
L
{k, 1, Infinity}]
2 2 n Pi x
n Pi A[n] Sin[------]
L
Sum[-(-----------------------),
2
L
{n, 1, Infinity}]
(* Collect the product of two sum's in a double sum *)
In[10]:=
sqru2=sqru /. Literal[Sum[an_,iter1_]*Sum[ak_,iter2_]] :> Sum[an*ak,iter1,iter2]
Out[10]=
2 2 4 k Pi x n Pi x
k n Pi A[k] A[n] Sin[------] Sin[------]
L L
Sum[-------------------------------------------, {n, 1, Infinity},
4
L
{k, 1, Infinity}]
(* Change the integration rule to map into a sum *)
In[20]:=
Unprotect[Integrate];
Literal[Integrate[Sum[as_,iter__],range_]]:=Sum[Integrate[as,range],iter]
Protect[Integrate];
Integrate[sqru2,{x,0,L}]
Out[23]=
2 2 3
Sum[(k n Pi A[k] A[n] (k Sin[(k - n) Pi] + n Sin[(k - n) Pi] -
3 2 2
k Sin[(k + n) Pi] + n Sin[(k + n) Pi])) / (2 L (k - n )),
{n, 1, Infinity}, {k, 1, Infinity}]
There are one notice, for a general rule You have to change
Literal[Integrate[Sum[as_,iter__],x_Symbol]]:=
Sum[Integrate[as,range],iter] /; FreeQ[{iter},x]
Literal[Integrate[Sum[as_,iter__],interval_List]]:=
Sum[Integrate[as,range],iter] /; FreeQ[{iter},First[interval]]
For multidimensional integrals the pattern's become more complicated.
Hope that helps
Jens