Why is a non-convergent series simplified to zero?

Sum[Sin[n*x]*Sin[n] /. x -> 1, {n, 1, Infinity}] ==
Sum[Sin[n*x]*Sin[n], {n, 1, Infinity}] /. x -> 1


Sum::div: Sum does not converge. >>

Sum::div: Sum does not converge. >>

Sum[Sin[n]^2, {n, 1, Infinity}] == 0

Can anyone explain this? I do not understand why Sum[Sin[n*x]*Sin[n], {n, 1, Infinity}] is automatically simplified to 0 when the series does not converge (unless Mod[x, Pi] == 0 in which case the sum really is zero).

-
although this sum is wrong the partial sum given by Mathematica are correct. I have tested this sum with wolfram alpha and it gave me that it diverges so I assume that at v9 they have corrected it ... – Spawn1701D Apr 21 '13 at 5:16
No, same response from v9. – fairflow Apr 21 '13 at 5:29

The reason is you hit a distribution.

partialSum = Sum[Sin[n*x]*Sin[n], {n, 1, bigN}];
(*  1/4 Csc[(1 + x)/2] Sin[1/2 (-1 - 2 bigN - x - 2 bigN x)]
-1/4 Csc[(1 - x)/2] Sin[1/2 (-1 - 2 bigN + x + 2 bigN x)] *)


and you can immediately see from here (7th from the top) that in the limit bigN->Infinity you get the sum of two Dirac functions centered at -1 and +1.

So we can conclude that

Sum[Sin[n*x]*Sin[n], {n, 1, Infinity}] = -Pi/2 DiracDelta[x+1] + Pi/2 DiracDelta[x-1]


Another way to check this (not a full proof) is to make the sum convergent :

convSum =  Sum[Sin[n*x]*Sin[n] Exp[-eps n], {n, 1, Infinity}, Assumptions :> {eps > 0}];

result = FullSimplify[ComplexExpand[convSum, TargetFunctions -> {Re, Im}]]
(* (Sin[1] Sin[x] Sinh[eps])/
(1 + Cos[2] + Cos[2 x] - 4 Cos[1] Cos[x] Cosh[eps] + Cosh[2 eps]) *)

Plot3D[result, {x, -2, 2}, {eps, 0, 1}, PlotRange -> All]


Limit[result, eps -> 0]
(* 0 *)


but :

Limit[result /. x -> -1, eps -> 0]
(* -Infinity *)

Limit[result /. x -> 1, eps -> 0]
(* Infinity *)

-