2
$\begingroup$

Expression

Sum[DiscreteDelta[a - 1]^2, {a, -∞, ∞}]

evaluates to 1, as expected, but

Sum[DiscreteDelta[a - b]^2, {a, -∞, ∞}]

does not converge, according to Mathematica. Is there a way to cleanly circumvent this?

$\endgroup$
  • $\begingroup$ DiscreteDelta[] evaluates to 1 only if the argument is zero. MMA cannot decide, wether a-b equals zero, because properties of b aren't known! $\endgroup$ – Ulrich Neumann Jul 11 '18 at 13:59
  • $\begingroup$ @UlrichNeumann However, Sum[DiscreteDelta[a - b], {a, -[Infinity], [Infinity]}] does give zero. I think this specifically a problem with power. $\endgroup$ – Weather Report Jul 11 '18 at 14:09
  • $\begingroup$ I believe the result Sum[DiscreteDelta[a - b], {a, -[Infinity], [Infinity]}]==1 cannot be true for arbitrary b (If I understand the definition DiscreteDelta right) $\endgroup$ – Ulrich Neumann Jul 11 '18 at 16:31
  • $\begingroup$ OK, I see. Carl Woil also suggested that this is a bug. $\endgroup$ – Weather Report Jul 11 '18 at 20:48
3
$\begingroup$

You could use Assuming:

Assuming[
    b ∈ Integers,
    Sum[DiscreteDelta[a-b]^2,{a,-∞,∞}]
]

1

$\endgroup$
  • $\begingroup$ Any ideas why I don't need this when the DiscreteDelta is not squared? $\endgroup$ – Weather Report Jul 11 '18 at 14:41
  • $\begingroup$ @WeatherReport My guess is that the unsquared version is a bug. For instance, if b=.5 the answer is incorrect. $\endgroup$ – Carl Woll Jul 11 '18 at 14:44
  • $\begingroup$ Hmm. I was really looking for an abstract notion of delta-symbol. which would evaluate Sum[AbstractDelta[a-b]f[a],{a,allvalues}] to f[b]. Like a DiracDelta but which should not diverge when squared. I thought that DiscreteDelta serves that purpose. But you are saying its probably a bug:) Is there an abstract notion that I'm looking for? $\endgroup$ – Weather Report Jul 11 '18 at 14:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.