2
$\begingroup$

I am having a problem with the FullSimplify and DiracDelta functions. When I run the following line in mathmatica:

Assuming[x >= 0 && y > 1, FullSimplify[DiracDelta[x]*DiracDelta[x - y]]]

the results is:

DiracDelta[x] DiracDelta[x - y]

while I expect 0. Any clue how to solve this problem?

| improve this question | |
$\endgroup$
  • 2
    $\begingroup$ This reply is totally confusing. No matter how you interpret the dirac function, the mathmatical equality above is correct. $\endgroup$ – A.J. Dec 9 '18 at 15:57
  • 2
    $\begingroup$ @user64494 But we're in Mathematica, not mathematics. $\endgroup$ – Michael E2 Dec 9 '18 at 17:14
  • 2
    $\begingroup$ @user64494 The result $\delta(x)=0$ for $x>0$ makes perfect sense. It means that $\delta[f]=0$ for all $f$ whose support is $x>0$. Just because you don't understand a concept does not mean that the concept itself makes no sense. Please stop making comments on every post that contain the words "Dirac delta", especially wrong comments. $\endgroup$ – AccidentalFourierTransform Dec 9 '18 at 18:42
  • 2
    $\begingroup$ @user64494 No, the support need not be compact (recall that $\delta$ is tempered). A Schwartz function is a smooth function such that $|P\partial_\alpha f|<\infty$ for all multi-indices $\alpha$ and all polynomials $P$. For example, a gaussian function is Schwartz, and its support is all of $\mathbb R$. Moreover, $\mathrm{sup}(f)\subset \{x>0\}$ does not mean that the support is not compact. For example, $\mathrm{sup}(f)=[1,2]$ is both compact and in $x>0$. These are traditional results; any reference should do. $\endgroup$ – AccidentalFourierTransform Dec 9 '18 at 18:59
  • 2
    $\begingroup$ @user64494 A "typical space of test functions" is not the only possibility. Note that I said tempered. If you don't know basic distribution theory, don't post comments trying to correct others. It is obnoxious. $\endgroup$ – AccidentalFourierTransform Dec 9 '18 at 19:14
1
$\begingroup$

Within Mathematica, DiracDelta makes only sense within Integrate (or integral transforms such as FourierTransform. And a product of DiracDelta makes only sense in a multidimensional setting:

Assuming[x >= 0 && y > 1,
 Integrate[DiracDelta[x] DiracDelta[x - y] f[x], {x, -2, 2}]
 ]

returns

enter image description here

If one uses two-dimensional integration and incorporates the assumptions into the integration, one obtains the desired results:

Integrate[ DiracDelta[x] DiracDelta[x - y] f[x, y], {x, 0, ∞}, {y, 1, ∞}]
Integrate[ DiracDelta[x] DiracDelta[x + y] f[x, y], {x, 0, ∞}, {y, 1, ∞}]

0

0

By the way, Mathematica seem to interpret functions in the integrant as functions that are compactly support in the interior of the integration domain:

Integrate[DiracDelta[x] DiracDelta[x - y] f[x, y], {x, 0, ∞}, {y, 0, ∞}]
Integrate[DiracDelta[x] DiracDelta[x - y] f[x, y], {x, -ϵ, ∞}, {y, -ϵ, ∞}, 
 Assumptions -> ϵ > 0]

This can lead to wrong results when integration function that do not vanish on the domain of integration:

Integrate[DiracDelta[x] DiracDelta[x - y] Cos[x + y], {x, 0, ∞}, {y, 0, ∞}]
Integrate[DiracDelta[x] DiracDelta[x - y] Cos[x + y], {x, -ϵ, ∞}, {y, -ϵ, ∞}, 
 Assumptions -> ϵ > 0]

0

1

Here, I would have expected that both evaluate to 1.

| improve this answer | |
$\endgroup$
  • $\begingroup$ I'm not sure I understand your last sentence. Surely $x=2=y$ satisfies $x=y$ and $x\ge0$ and $y>0$. Is there a typo somewhere? Or what am I missing? $\endgroup$ – AccidentalFourierTransform Dec 9 '18 at 16:09
  • $\begingroup$ Are you sure you are not making a mistake with the limits of the integral? With the correct limits these integrals will both evaluate to zero. $\endgroup$ – A.J. Dec 9 '18 at 16:09
  • $\begingroup$ Well of course one of the delta functions diverges for x=y=2, but the point is that the other delta function is zero already. $\endgroup$ – A.J. Dec 9 '18 at 16:11
  • $\begingroup$ The answer to the following question might be relevant, but it is just a trick, that makes the whole simplification more complicated. link $\endgroup$ – A.J. Dec 9 '18 at 16:15
  • $\begingroup$ @MichaelE2 Uh. Good point. Now I get what A.J. meant. $\endgroup$ – Henrik Schumacher Dec 9 '18 at 16:46

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.