Re: question
- To: mathgroup at smc.vnet.net
- Subject: [mg75661] Re: question
- From: Peter Pein <petsie at dordos.net>
- Date: Tue, 8 May 2007 05:48:45 -0400 (EDT)
- References: <f1mrii$rb4$1@smc.vnet.net>
dimitris schrieb:
> Hi.
>
> I want to show that
>
> x/Pi - (1/2)*x*(StruveH[-1, x] + StruveH[1, x])
>
> is equal to zero.
>
> In[36]:=
> Plot[Chop[x/Pi - (1/2)*x*(StruveH[-1, x] + StruveH[1, x])], {x, -2,
> 2}, Axes -> False, Frame -> True]
>
> with Mathematica.
>
> Simplify, FullSimplify, FunctionExpand fail to do this task. Even
> replacing x by specific value does
> not yield better results.
>
> In[40]:=
> FullSimplify[3/Pi - (1/2)*3*(StruveH[-1, 3] + StruveH[1, 3])]
>
> Out[40]=
> 3/Pi - (3/2)*(StruveH[-1, 3] + StruveH[1, 3])
>
> On the contrary, in another CAS I took
>
> simplify(3/Pi - (1/2)*3*(StruveH(-1, 3) + StruveH(1, 3)));
> 0
>
> Any ideas?
>
> A quick way to show that x/Pi - (1/2)*x*(StruveH[-1, x] + StruveH[1,
> x]) is indeed
> zero is
>
> In[41]:=
> Integrate[x/Pi - (1/2)*x*(StruveH[-1, x] + StruveH[1, x]), x]
>
> Out[41]=
> 0
>
> but I would really appreciate any other suggestions.
>
> Thanks
>
>
Hi Dimitris,
I don't get it in the usual way (showing, that the derivative is identical to
zero _and_ expression at x==0 is zero (the latter is easy)).
My way is similar to your approach via integration:
ee = x/Pi - (1/2)*x*(StruveH[-1, x] + StruveH[1, x]);
InverseLaplaceTransform[LaplaceTransform[ee, x, s], s, x]
--> 0
and calculations like Series[ee, {x, 0, 50}] --> O[x]^51 are no proof.
Playing around with SeriesTerm[] leads to a funny result:
e2 = Sum[SeriesTerm[ee, {x, 0, k}]x^(k)*k!^(3/2), {k, 0, â??}]
gives two warnings "Sum does not converge" and the "result" x/Pi.
But really astonishing is:
e3 = Sum[SeriesTerm[ee, {x, 0, k}]*(x^k/k), {k, 0, Infinity}]
--> x/Pi - (1/(9*Pi))*(x^3*HypergeometricPFQ[{1}, {5/2, 5/2}, -(x^2/4)])
-(1/2)*StruveH[0, x]
ZeroQ[e3]
--> False
*but:*
FullSimplify[e3]
--> 0
I understood ZeroQ as if it tries transformations, which FullSimplify uses too??
But using SeriesTerm, we get a proof:
First FullSimplify the even terms, then the odd terms with k>=3 and at last
the coefficient with k==1:
ak = SeriesTerm[ee, {x, 0, k}];
FullSimplify[ak /. k -> 2*j, j \[Element] Integers && j >= 0]
--> 0
FullSimplify[ak /. k -> 2*j + 1, j \[Element] Integers && j > 0]
--> 0
and finally
FullSimplify[ak /. k -> 1]
--> 0
Cheers,
Peter