[go: up one dir, main page]

About the Project
1 Algebraic and Analytic MethodsTopics of Discussion

§1.16 Distributions

Contents
  1. §1.16(i) Test Functions
  2. §1.16(ii) Derivatives of a Distribution
  3. §1.16(iii) Dirac Delta Distribution
  4. §1.16(iv) Heaviside Function
  5. §1.16(v) Tempered Distributions
  6. §1.16(vi) Distributions of Several Variables
  7. §1.16(vii) Fourier Transforms of Tempered Distributions
  8. §1.16(viii) Fourier Transforms of Special Distributions
  9. §1.16(ix) References for Section 1.16

§1.16(i) Test Functions

Let ϕ be a function defined on an open interval I=(a,b), which can be infinite. The closure of the set of points where ϕ0 is called the support of ϕ. If the support of ϕ is a compact set (§1.9(vii)), then ϕ is called a function of compact support. A test function is an infinitely differentiable function of compact support.

A sequence {ϕn} of test functions converges to a test function ϕ if the support of every ϕn is contained in a fixed compact set K and as n the sequence {ϕn(k)} converges uniformly on K to ϕ(k) for k=0,1,2,.

The linear space of all test functions with the above definition of convergence is called a test function space. We denote it by 𝒟(I).

A mapping Λ:𝒟(I) is a linear functional if

1.16.1 Λ(α1ϕ1+α2ϕ2)=α1Λ(ϕ1)+α2Λ(ϕ2),

where α1 and α2 are real or complex constants. Λ:𝒟(I) is called a distribution, or generalized function, if it is a continuous linear functional on 𝒟(I), that is, it is a linear functional and for every ϕnϕ in 𝒟(I),

1.16.2 limnΛ(ϕn)=Λ(ϕ).

From here on we write Λ,ϕ for Λ(ϕ). The space of all distributions will be denoted by 𝒟(I). A distribution Λ is called regular if there is a locally integrable function f on I (i.e., a function f on I which is absolutely Lebesgue integrable on every compact subset of I) such that

1.16.3 Λ,ϕ=If(x)ϕ(x)dx.

We denote a regular distribution by Λf, or simply f, where f is the function giving rise to the distribution. (If a distribution is not regular, it is called singular.) More generally, for α:[a,b][,] a nondecreasing function the corresponding Lebesgue–Stieltjes measure μα (see §1.4(v)) can be considered as a distribution:

1.16.3_5 μα,ϕ=Iϕ(x)dα(x).

Define

1.16.4 Λ1+Λ2,ϕ=Λ1,ϕ+Λ2,ϕ,
1.16.5 cΛ,ϕ=cΛ,ϕ=Λ,cϕ,

where c is a constant. More generally, if α(x) is an infinitely differentiable function, then

1.16.6 αΛ,ϕ=Λ,αϕ.

We say that a sequence of distributions {Λn} converges to a distribution Λ in 𝒟 if

1.16.7 limnΛn,ϕ=Λ,ϕ

for all ϕ𝒟(I).

§1.16(ii) Derivatives of a Distribution

The derivative Λ of a distribution is defined by

1.16.8 Λ,ϕ=Λ,ϕ,
ϕ𝒟(I).

Similarly

1.16.9 Λ(k),ϕ=(1)kΛ,ϕ(k),
k=1,2,.

If f is a locally integrable function then its distributional derivative is 𝐷f=Λf. In the situation of (1.16.3_5) we have

1.16.9_5 μα=𝐷α.

If the measure μα is absolutely continuous with density w (see §1.4(v)) then 𝐷α=Λw.

§1.16(iii) Dirac Delta Distribution

1.16.10 δ,ϕ =ϕ(0),
ϕ𝒟(I),
1.16.11 δx0,ϕ =ϕ(x0),
ϕ𝒟(I),
1.16.12 δx0(n),ϕ =(1)nϕ(n)(x0),
ϕ𝒟(I).

The Dirac delta distribution is singular. See also §1.17(i).

§1.16(iv) Heaviside Function

1.16.13 H(x) ={1,x>0,0,x0.
1.16.14 H(xx0) ={1,x>x0,0,xx0.
1.16.15 𝐷H =δ,
1.16.16 𝐷H(xx0) =δx0.

Since δx0 is the Lebesgue–Stieltjes measure μα corresponding to α(x)=H(xx0) (see §1.4(v)), formula (1.16.16) is a special case of (1.16.3_5), (1.16.9_5) for that choice of α.

Suppose f(x) is infinitely differentiable except at x0, where left and right derivatives of all orders exist, and

1.16.17 σn=f(n)(x0+)f(n)(x0).

Then

1.16.18 𝐷mf=f(m)+σ0δx0(m1)+σ1δx0(m2)++σm1δx0,
m=1,2,.

For α>1,

1.16.19 x+α=xαH(x)={xα,x>0,0,x0.

For α>0,

1.16.20 𝐷x+α=αx+α1.

For α<1 and α not an integer, define

1.16.21 x+α=1(α+1)n𝐷nx+α+n,

where n is an integer such that α+n>1. Similarly, we write

1.16.22 ln+x=H(x)lnx={lnx,x>0,0,x0,

and define

1.16.23 (1)nn!x+1n=𝐷(n+1)ln+x,
n=0,1,2,.

§1.16(v) Tempered Distributions

The space 𝒯() of test functions for tempered distributions consists of all infinitely-differentiable functions such that the function and all its derivatives are O(|x|N) as |x| for all N.

A sequence {ϕn} of functions in 𝒯 is said to converge to a function ϕ𝒯 as n if the sequence {ϕn(k)} converges uniformly to ϕ(k) on every finite interval and if the constants ck,N in the inequalities

1.16.24 |xNϕn(k)|ck,N

do not depend on n.

A tempered distribution is a continuous linear functional Λ on 𝒯. (See the definition of a distribution in §1.16(i).) The set of tempered distributions is denoted by 𝒯.

A sequence of tempered distributions Λn converges to Λ in 𝒯 if

1.16.25 limnΛn,ϕ=Λ,ϕ,

for all ϕ𝒯.

The derivatives of tempered distributions are defined in the same way as derivatives of distributions.

For a detailed discussion of tempered distributions see Lighthill (1958).

§1.16(vi) Distributions of Several Variables

Let 𝒟(n)=𝒟n be the set of all infinitely differentiable functions in n variables, ϕ(x1,x2,,xn), with compact support in n. If k=(k1,,kn) is a multi-index and x=(x1,,xn)n, then we write xk=x1k1xnkn and ϕ(k)(x)=kϕ/(x1k1xnkn). A sequence {ϕm} of functions in 𝒟n converges to a function ϕ𝒟n if the supports of ϕm lie in a fixed compact subset K of n and ϕm(k) converges uniformly to ϕ(k) in K for every multi-index k=(k1,k2,,kn). A distribution in n is a continuous linear functional on 𝒟n.

The partial derivatives of distributions in n can be defined as in §1.16(ii). A locally integrable function f(x)=f(x1,x2,,xn) gives rise to a distribution Λf defined by

1.16.26 Λf,ϕ=nf(x)ϕ(x)dx,
ϕ𝒟n.

The distributional derivative 𝐷kf of f is defined by

1.16.27 𝐷kf,ϕ=(1)|k|nf(x)ϕ(k)(x)dx,
ϕ𝒟n,

where k is a multi-index and |k|=k1+k2++kn.

For tempered distributions the space of test functions 𝒯n is the set of all infinitely-differentiable functions ϕ of n variables that satisfy

1.16.28 |xmϕ(k)(x)|cm,k,
xn.

Here m=(m1,m2,,mn) and k=(k1,k2,,kn) are multi-indices, and cm,k are constants. Tempered distributions are continuous linear functionals on this space of test functions. The space of tempered distributions is denoted by 𝒯n.

§1.16(vii) Fourier Transforms of Tempered Distributions

Suppose ϕ is a test function in 𝒯n. Then its Fourier transform is

1.16.29 (ϕ)(𝐱)=ϕ(𝐱)=1(2π)n/2nϕ(𝐭)ei𝐱𝐭d𝐭,

where 𝐱=(x1,x2,,xn) and 𝐱𝐭=x1t1++xntn. ϕ(𝐱) is also in 𝒯n.

Let

1.16.30 𝐃=(1ix1,1ix2,,1ixn).

For a multi-index 𝜶=(α1,α2,,αn), define

1.16.31 P(𝐱)=𝜶c𝜶𝐱𝜶=𝜶c𝜶x1α1xnαn,

and

1.16.32 P(𝐃)=𝜶c𝜶𝐃α=𝜶c𝜶(1ix1)α1(1ixn)αn.

Here 𝜶 ranges over a finite set of multi-indices, P(𝐱) is a multivariate polynomial, and P(𝐃) is a partial differential operator. Then

1.16.33 (P(𝐃)ϕ)(𝐱)=P(𝐱)ϕ(𝐱),

and

1.16.34 (Pϕ)(𝐱)=P(𝐃)ϕ(𝐱).

If u𝒯n is a tempered distribution, then its Fourier transform (u) is defined by

1.16.35 (u),ϕ=u,(ϕ),
ϕ𝒯n.

The Fourier transform (u) of a tempered distribution is again a tempered distribution, and

1.16.36 (P(𝐃)u),ϕ=P(u),ϕ=(u),Pϕ,
1.16.37 (Pu),ϕ=P(𝐃)(u),ϕ,

in which P(𝐱)=P(𝐱); compare (1.16.33) and (1.16.34). In (1.16.36) and (1.16.37) the derivatives in P(𝐃) are understood to be in the sense of distributions, as defined in §1.16(ii) and we also use the convention (1.16.6).

§1.16(viii) Fourier Transforms of Special Distributions

We use the notation of the previous subsection and take n=1 and u=δ in (1.16.35). We obtain

1.16.38 (δ),ϕ=δ,(ϕ)=δ,12πϕ(t)eixtdt=12πϕ(t)dt=12π1,ϕ,
ϕ𝒯.

As distributions, the last equation reads

1.16.39 (δ)=12π,

which is often written conventionally as

1.16.40 δ(t)eixtdt=1;

see also (1.17.2).

Since 2π(δ)=1, we have

1.16.41 (1),ϕ=2π((δ)),ϕ=2π(δ),(ϕ)=2πδ,((ϕ))=2πδ,ϕ=2πϕ(0),

in which ϕ(x)=ϕ(x). The second to last equality follows from the Fourier integral formula (1.17.8). Since the quantity on the extreme right of (1.16.41) is equal to 2πδ,ϕ, as distributions, the result in this equation can be stated as

1.16.42 (1)=2πδ,

and conventionally it is expressed as

1.16.43 12πeixtdt=δ(x);

see also (1.17.12).

It is easily verified that

1.16.44 sign(x)=2H(x)1,
x0,

and from (1.16.15) we find

1.16.45 sign=2H=2δ,

where H(x) is the Heaviside function defined in (1.16.13), and the derivatives are to be understood in the sense of distributions. Then

1.16.46 (sign)=(2H)=2(δ)=2π,

and from (1.16.36) with u=sign, P(𝐃)=𝐷, and P(x)=ix, we have also

1.16.47 (sign)=xi(sign).

Coupling (1.16.46) and (1.16.47) gives

1.16.48 (sign)=2πix,

that is

1.16.49 (sign),ϕ=i2πϕ(x)xdx.

The Fourier transform of H(x) now follows from (1.16.42) and (1.16.48). Indeed, we have

1.16.50 (H)=12(1+sign)=12[(1)+(sign)]=π2(δ+iπx),

that is

1.16.51 (H),ϕ=π2ϕ(0)+i2πϕ(x)xdx.

For more detailed discussions of the formulas in this section, see Kanwal (1983) and Debnath and Bhatta (2015).

§1.16(ix) References for Section 1.16

See Hildebrandt (1938) and Chihara (1978, Chapter II) for Stieltjes measures which are used in §18.39(iii); see also Shohat and Tamarkin (1970, Chapter II). Friedman (1990) gives an overview of generalized functions and their relation to distributions. See also Lighthill (1958), and Zemanian (1987).