Differential Equations Assignment Help

Linear Partial Differential Equations:
Analysis and Numeric Assignment Help

For any Assignment related queries, call us at: – +1 678 648 4277
You can mail us at:- support@mathhomeworksolver.com or
reach us at:- https://www.mathhomeworksolver.com/
Problems
Problem 1: Distributions
This problem concerns distributions as defined in the notes:

continuous linear functionals f{φ} from test functions φ ∈ D,
where D is the set of infinitely differentiable functions with
compact support (i.e. φ = 0 outside some region with finite
diameter [differing for different φ], i.e. outside some finite
interval [a, b] in 1d).
In this part, you will consider the function

and its (weak) derivative, which is connected to something
called the Cauchy Principal Value.
(i) Show that f(x) defines a regular distribution, by showing that
f(x) is locally integrable for
all intervals [a, b].
(ii) Consider the 18.01 derivative of f(x), which gives

Suppose we
just set “f'(0) = 0” at the origin to define Show

that this g(x) is not locally
integrable, and hence does not define a distribution.
But the weak derivative f'{φ} must exist, so this means that we
have to do something different from the 18.01 derivative, and
moreover f'{φ} is not a regular distribution. What is it?
(iii) Write where

since this limit exists and equals f{φ} for all φ
from your proof in the previous part. Compute the
distributional derivative and show that
f'{φ} is precisely the Cauchy Principal Value (google the
definition, e.g. on Wikipedia) of the integral of g(x)φ(x).
(iv) Alternatively, show that
(which is a well-defined integral for all φ ∈ D).
(b) In class, we only looked explicitly at 1d distributions, but a

distribution in d dimensions Rd can obviously be defined
similarly, as maps f{φ} from smooth localized functions φ(x) to
numbers. Analogous to class, define the distributional gradient
'f by 'f{φ} = f{−'φ}.
Consider some finite volume V with a surface ∂V , and assume
∂V is differentiable so that at each point it has an outward-
pointing unit normal vector n, as shown in figure 1. Define a
“surface delta function” δ(∂V ){φ} = φ(x)dd−1x to give the surface
integral of the test function.
Suppose we have a regular distribution f{φ} defined by the
function where we may have a
discontinuity f2 − f1 =0 at ∂V .
(i) Show that the distributional gradient of f is
where the second term is a regular distribution given by the

ordinary 18.02 gradient of f1 and f2 (assumed to be
differentiable), while the first term is the singular distribution
You can use the integral identity that to help you integrate by
parts.
(ii) Defining '2f{φ} = f{'2φ}, derive a similar expression to the

above for '2f. Note that you should have one term from the
discontinuity f1 − f2, and another term from the discontinuity 'f1
−'f2. (Recall how we integrated '2 by parts in class some time ago.)
Problem 2: Green’s functions

Consider Green’s functions of the self-adjoint indefinite operator
Aˆ= −'2 − ω2 (κ> 0) over all space (Ω= R3 in 3d), with solutions that
→ 0 at infinity. (This is the multidimensional version of problem
2 from pset 5.) As in class, thanks to the translational and
rotational invariance of this problem, we can find G(x, x')= g(|x −
x'|) for some g(r) in spherical coordinates.
(a) Solve for g(r) in 3d, similar to the procedure in class.
(i) Similar to the case of Aˆ= −' in class, first solve for g(r) for r>
2
0, and write g(r) = limf→0+ ff(r) where ff(r)=0 for r ≤ E. [Hint:

although Wikipedia writes the spherical '2g(r) as 1/r2 (rg')', it
may be more convenient to write it equivalently as '2g =1/2
(rg)'', as in class, and to solve for h(r)= rg(r) first.
Hint: if you get sines and cosines from this differential equation,
it will probably be easier to use complex exponentials, e.g. eiωr,
instead.]
(ii) In the previous part, you should find two solutions, both of
which go to zero at infinity. To choose between them,
remember that this operator arose from a e−iωt time dependence.
Plug in this time dependence and impose an “outgoing wave”
boundary condition (also called a Sommerfield or radiation
boundary condition): require that waves be traveling outward
far away, not inward.
(iii) Then, evaluate Agˆ= Ag){q} = g{ ˆ= q(0) for an arbitrary

(smooth,
δ(x) in the distributional sense: (ˆAq}localized) test function
q(x) to solve for the unknown constants in g(r). [Hint: when
evaluating g{ Aq}, you may need to integrate by parts on the
radial-derivative term of '2q; don’t forget the boundary term(s).]
(b) Check that the ω → 0+ limit gives the answer from class.
Solutions
Problem 1:
(a) Solutions:
f(x) is bounded in every interval except intervals containing x

=0, so local integrability is trivial except for intervals containing
x =0. It is sufficient to consider integrals because any interval
[a, b] containing 0 can be broken up into and f(−x)= f(x)
so we only need to show that the latter is finite. But we can now
just do the integral explicitly:
since
as can easily be seen e.g. by L’Hôpital’s rule applied to
(ii) g(x) is not locally integrable for intervals containing the

origin. For example
Therefore, f'{φ} = f{−φ'} is a singular distribution.

(iii) We write f'{φ} = liml→0+ fl'{φ}, and integrate by parts in
fl'{φ} = fl{−φ'}:
In the last line, the first limit is precisely the Cauchy Principal
Value of g(x)φ(x)dx (CPV = remove a ball of radius E around
the singularity, do the integral, and then take the E → 0+ limit).
The second term vanishes because, since φ(x) is continuous and
infinitely differentiable, φ(E)−φ(−E) vanishes at least as fast as E
as E → 0, so its product with ln E vanishes in the limit as in part
(i)
(iv) Use f'{φ} = f{−φ'} = f{−[φ − φ(0)]'} = f'{φ − φ(0)}. Then

substitute φ − φ(0) into the previous part, and note that since
the integrand is now finite as x → 0 [since φ(x) − φ(0)
goes to zero as x → 0, at least proportionally to x or faster as in
the previous part], we can now just take the limit to write
without using the CPV.
(b) Solutions:
(i) Let V c denote the complement of V (the exterior region

outside V , i.e. V c = Rd\V ), and note that ∂V c = ∂V
(but with the outward-normal vector reversed in sign). We write:
as desired.
(ii) In this case, we will need to integrate by parts twice, but we

can just quote the results from the “notes on elliptic operators”
from class (where we integrated by parts twice with −V2
already), albeit keeping the boundary terms from ∂V that were
zero in the notes:
But the first term is δ(∂V )[f1(x) − f2(x)] {n·Vφ} =(n·V)δ(∂V )

[f2(x) − f1(x)] {φ} by the definition (note the sign change) of the
distributional derivative n·V (note that this is a scalar derivative
in the n direction, not a gradient vector). The second term is a
surface delta function weighted by (n ·Vf1 − n ·Vf2) ,the
discontinuity in the normal derivative. And the last terms are
just a regular distribution. So, we have
As noted in class, n ·V of a delta function is a “dipole” oriented
in the n direction, so the first term is a “dipole layer”.
Problem 2:
(a) We solve for g(r) in 3d as follows:
(i) For r> 0, −V2g − ω2g =0 and hence −ω2g = V2g =1/r(rg)'' =⇒ =
−ω2h where h(r)= rg(r)the solution to this is h(r)= cefor
some constants c and d, or
It is a little more tricky to determine whether we should use the

c or the d term than in class, since both decay at the same rate.
The ratio c/d will be determined by some kind of boundary
condition at inﬁnity, but what might this be? It is acceptable for
you to just punt on this here; since e±iωr are complex conjugates
of each other, your analysis will apply equally well to either one,
and you can arbitrarily pick one, say ceiωr/r, to analyze.
However, to see why there should be a sensible choice, recall

that this operator arose in pset 5 by assuming a time
dependence eiωtmultiplying the solution, in which case we are
looking at wave solutions
where the c term describes waves moving out towards r →∞,
while the d term describes
waves moving in from infinity. In wave problems, we typically
impose a boundary condition of outgoing waves at infinity, in
which case we must set d =0. (However, the choice would have
been reversed if we picked the opposite sign convention, e+iωt,
for the time dependence.)
(ii) Let’s focus on g(r)= ceiωr/r. As in class, the 1/r singularity is

no problem in 3d (it is cancelled by the Jacobian factor r2dr), so
g is a regular distribution. Given an arbitrary test function q(x),
we now evaluate
and hence c =1/4π . Thus,
assuming boundary conditions such that d =0. More generally,

since exactly the same result applies to the de−iωr/r term, we
obtain,
for c + d =1/4π , with the ratio c/d being set by the boundary
conditions at ∞. The value at x = x being irrelevant in the
distribution sense, e.g. we can assign it to zero, since this is a
regular distribution with a finite integral, similar to class.)
(b) The ω → 0 limit gives 1/4π|x − x'| as in class, by inspection.

Differential Equations Assignment Help

Uploaded by

Copyright:

Available Formats

Differential Equations Assignment Help

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Differential Equations Assignment Help

Uploaded by

Copyright:

Available Formats

Linear Partial Differential Equations:

Analysis and Numeric Assignment Help

This problem concerns distributions as defined in the notes:

In this part, you will consider the function

all intervals [a, b].

(ii) Consider the 18.01 derivative of f(x), which gives

just set “f'(0) = 0” at the origin to define Show

integrable, and hence does not define a distribution.

(iii) Write where

(b) In class, we only looked explicitly at 1d distributions, but a

(i) Show that the distributional gradient of f is

where the second term is a regular distribution given by the

(ii) Defining '2f{φ} = f{'2φ}, derive a similar expression to the

Problem 2: Green’s functions

(a) Solve for g(r) in 3d, similar to the procedure in class.

0, and write g(r) = limf→0+ ff(r) where ff(r)=0 for r ≤ E. [Hint:

(iii) Then, evaluate Agˆ= Ag){q} = g{ ˆ= q(0) for an arbitrary

f(x) is bounded in every interval except intervals containing x

as can easily be seen e.g. by L’Hôpital’s rule applied to

(ii) g(x) is not locally integrable for intervals containing the

Therefore, f'{φ} = f{−φ'} is a singular distribution.

(iv) Use f'{φ} = f{−φ'} = f{−[φ − φ(0)]'} = f'{φ − φ(0)}. Then

(i) Let V c denote the complement of V (the exterior region

(ii) In this case, we will need to integrate by parts twice, but we

But the first term is δ(∂V )[f1(x) − f2(x)] {n·Vφ} =(n·V)δ(∂V )

(a) We solve for g(r) in 3d as follows:

It is a little more tricky to determine whether we should use the

However, to see why there should be a sensible choice, recall

(ii) Let’s focus on g(r)= ceiωr/r. As in class, the 1/r singularity is

assuming boundary conditions such that d =0. More generally,

(b) The ω → 0 limit gives 1/4π|x − x'| as in class, by inspection.

You might also like