Re: Dirac Delta Function: Basics

*To*: mathgroup at smc.vnet.net*Subject*: [mg75185] Re: Dirac Delta Function: Basics*From*: dimitris <dimmechan at yahoo.com>*Date*: Fri, 20 Apr 2007 00:31:19 -0400 (EDT)*References*: <f07a53$4ci$1@smc.vnet.net>

Some important things to know: Currently, there are two distributions built into Mathematica: DiracDelta and UnitStep. In[28]:= (Information[#1, LongForm -> False] & ) /@ {UnitStep, DiracDelta}; "UnitStep[x] represents the unit step function, equal to 0 for x < 0 and 1 for x =E2=89=A5 0. UnitStep[x1, x2, ... ] represents the multidimensional unit step function which is 1 only if none of the xi are negative." "DiracDelta[x] represents the Dirac delta function =CE=B4(x). DiracDelta[x1, x2, ... ] represents the multidimensional Dirac delta unction =CE=B4(x1, x2, =E2=80=A6)." Both of them, like all generalized functions, are defined for real arguments only. In[33]:= ({DiracDelta[#1], UnitStep[#1]} & ) /@ {1, -5/3, 6 + I} Out[33]= {{0, 1}, {0, 0}, {DiracDelta[6 + I], UnitStep[6 + I]}} (*DiracDelta[0] stays uneveluated; it's no infinity*) Here is a plot of UnitStep[x] In[39]:= Show[Block[{$DisplayFunction = Identity}, (Plot[UnitStep[x], {x, #1[[1]], #1[[2]]}, Axes -> False, Frame -> True] & ) /@ Partition[Range[-2, 2], 2, 1]]] (*in this way we ommit the buggy vertical line appeared at x=0*) Obviously the following plot of the DiracDelta (generalized) function looks quite logical. In[43]:= Plot[DiracDelta[x], {x, -2, 2}, Axes -> False, Frame -> True] That is, the nonnumerical value at x=0 was not detected. In[44]:= Cases[%, {(x_)?NumericQ, (y_)?NumericQ} /; -0.002 < x < 0.002, Infinity] Out[44]= {{-0.0019572,0.},{-0.00125518,0.},{-0.000607188,0.},{0.000619756,0.}, {0.00192017,0.}} There are some identities of these distributions that Mathematica knows about. But they required the use of FunctionExpand. So, for example In[56]:= FunctionExpand /@ {UnitStep[(x - 1)*(x + 3)*(x - 5)], UnitStep[x]^2, DiracDelta[x*(x - 1)], UnitStep[-x^4 - 1], UnitStep[x^4 + 1]} Out[56]= {UnitStep[-5 + x] - UnitStep[-1 + x] + UnitStep[3 + x], UnitStep[x], DiracDelta[-1 + x] + DiracDelta[x], 0, 1} The generalization to more than one variable is also possible. The Dirac delta function and the Heaviside function are connected and Mathematica knows this relation In[57]:= Derivative[1][UnitStep][x] Derivative[-1][DiracDelta][x] Out[57]= DiracDelta[x] Out[58]= UnitStep[x] Integrate and D knows to deal with functions. In[67]:= Integrate[Cos[x]/(1 + x^2)^(1/2), {x, 0, Infinity}] Integrate[(Cos[x]*UnitStep[x])/(1 + x^2)^(1/2), {x, -Infinity, Infinity}] Out[67]= BesselK[0, 1] Out[68]= BesselK[0, 1] In[68]:= Integrate[UnitStep[Log[x] + 2], {x, 0, 2*Pi}] Out[68]= -(1/E^2) + 2*Pi In[70]:= Integrate[Log[x]*DiracDelta[x - 3], {x, -2, 5}] Out[70]= Log[3] (*At this point I want to inform you about the presence of the Integration of Piecewise Functions with Applications notebook of Maxim Rytin available from here: http://library.wolfram.com/infocenter/MathSource/5117/ which knows to deal even better that Integrate in relevant cases with distributions.*) In[76]:= (D[UnitStep[x], {x, #1}] & ) /@ Range[4] Out[76]= {DiracDelta[x], Derivative[1][DiracDelta][x], Derivative[2][DiracDelta] [x], Derivative[3][DiracDelta][x]} Currently only linear functionals are allowed. So the next integral stays unevaluated. In[81]:= Integrate[DiracDelta[x]^2, {x, -Infinity, Infinity}] Out[81]= Integrate[DiracDelta[x]^2, {x, -Infinity, Infinity}] A UnitStep distribution can appear inside NIntegrate. In[83]:= Integrate[UnitStep[x^12 - 6], {x, -3, 3}] {NIntegrate[UnitStep[x^12 - 6], {x, -3, 3}], N[%]} Out[83]= 6 - 2*6^(1/12) Out[84]= {3.6779266552520116, 3.6779266552520116} For DiracDelta this is not the case. In[86]:= NIntegrate[1 + DiracDelta[Exp[x] - 1/2], {x, -1, 1}] (*incorrect*) Out[86]= 2=2E In[89]:= Integrate[1 + DiracDelta[Exp[x] - 1/2], {x, -1, 1}] Out[89]= 4 Distributions like the Dirac distribution or the Heaviside distribution are often viewed as limits from continuous functions. But neither Limit nor Sum recognise it. In[90]:= Limit[e/(x^2 + e^2), e -> 0] Out[90]= 0 In[91]:= Sum[Exp[I*k*x], {k, -Infinity, Infinity}] Out[91]= 0 (*in a distributional sense we get 2Pi DiracDelta[x] Indeed In[94]:= Sqrt[2*Pi]*FourierTransform[1, x, s] Out[94]= 2*Pi*DiracDelta[s] *) Here is one example of such a sequence along with a visualization In[101]:= Needs["Graphics`"] delta[n_][x_] := n/(Sqrt[Pi]/E^((-n^2)*x^2)) frame[n_] := Plot[delta[n][x], {x, -2, 2}, PlotRange -> {-0.1, 6.1}, Frame -> True, PlotLabel -> delta["n"][x]", n ="NumberForm[n, {4, 2}, NumberPadding -> {" ", "0"}], ImageSize -> {800, 500}]; Animate[frame[n], {n, 1, 10, 0.25}] SelectionMove[EvaluationNotebook[], All, GeneratedCell] FrontEndTokenExecute["OpenCloseGroup"]; Pause[0.5]; FrontEndExecute[{FrontEnd`SelectionAnimate[200, AnimationDisplayTime - > 0.1, AnimationDirection -> ForwardBackward]}] And here is one example of application of the sequence In[105]:= Integrate[delta[n][x^4 - 1], {x, -Infinity, Infinity}, Assumptions -> n > 0] Limit[%, n -> Infinity] FunctionExpand[%] Out[105]= (n^(3/4)*(4*Gamma[9/8]*Hypergeometric1F1[1/8, 1/2, n^2] + n*Gamma[5/8]*Hypergeometric1F1[5/8, 3/2, n^2]))/(E^n^2*(2*Sqrt[Pi])) Out[106]= (Gamma[1/8] + 8*Gamma[9/8])/(4*Gamma[1/8]) Out[107]= 1/2 Indeed In[108]:= Integrate[DiracDelta[x^4 - 1], {x, -Infinity, Infinity}, Assumptions - > n > 0] Out[108]= 1/2 Be also aware that currently only the integral transforms will return distributions in their output for inputs that do not contain distributions. Functions like D and Integrate know how to deal with certain distributions, but their output will only contain them when their input does. So for example In[110]:= (Integrate[Cos[x]*Exp[(-I)*x*s], {x, -Infinity, Infinity}, GenerateConditions -> #1] & ) /@ {True, False} Integrate::idiv: Integral ofCos[x]/E^(I*s*x) does not converge on {- Infinity, \ Infinity}. Out[110]= {Integrate[Cos[x]/E^(I*s*x), {x, -Infinity, Infinity}, GenerateConditions -> True], Integrate[Cos[x]/E^(I*s*x), {x, -Infinity, Infinity}, GenerateConditions -> False]} whereas In[112]:= FourierTransform[Cos[x], x, s] Out[112]= Sqrt[Pi/2]*DiracDelta[-1 + s] + Sqrt[Pi/2]*DiracDelta[1 + s] I hope that these are enough for now! Regards Dimitris PS References 1) M. Trott's "Guidebook for Symbolics" 2) R. N. Bracewell's "The Fourier Transform and its applications" 3) Estrada and Kanwal's "Ditributional Approach to Asymptotics" =CE=9F/=CE=97 Gopinath Venkatesan =CE=AD=CE=B3=CF=81=CE=B1=CF=88=CE=B5: > Hello Friends > > I browsed some of the previous posts in this forum on Dirac Delta functio= n, and found some interesting distributions used to represent the delta fun= ction, at the below website: > > <http://mathworld.wolfram.com/DeltaFunction.html> > > Also when I tried using one of the distribution, Fourier series approxima= tion (shown below), I was getting very different solutions for different en= tries of upper limit of k, even at higher values of upper limits. > > DeltaF(x - a) = 1/(2*Pi) + > (1/Pi)*Sum[Cos[k*a]*Cos[k*x] + Sin[k*a]*Sin[k*x], > {k, 1, Infinity}] > > (here DeltaF is DiracDelta). > > Does anyone here can give me suggestions on using right distributions and= parameters. Thanks. > > Also Can I use the DiracDelta[] function available in Mathematica itself.= How does Mathematica calculate them. > > I am not integrating the equation containing DiracDelta[], so as seen fro= m the Mathematica examples, I think it is doing Laplace transformations. > > Any help/directions to resources are appreciated. Thanks, > > Gopinath > Graduate Student > University of Oklahoma