http://www.amaderforum.

com

CHAPTER 4. PARTIAL DIFFERENTIAL EQUATIONS

• Partial Differential Equations

• Single Partial Differential Equations: Their Origin

o Boundary Conditions of a Typical Partial Differential Equation in Two
Dimensions
o Cauchy Problem and Characteristics
o Hyperbolic Equations
o Riemann's Method for Integrating the Most General 2nd Order Linear
Hyperbolic Equation

• System of Partial Differential Equations: How to Solve Maxwell's Equations

Using Linear Algebra
o Maxwell Wave Equation
o The Overdetermined System
o Maxwell Wave Equation (continued)
o Cylindrical Coordinates
o Spherical Coordinates

http://www.amaderforum.com
 http://www.amaderforum.com

Partial Differential Equations

Linear algebra is the mathematical guide of choice for implementing the principle of unit-
economy61 applied to partial differential equations. The present chapter considers two
kinds:

1. Single linear partial differential equations corresponding to the linear system

However, instead of merely exhibiting general solutions to such a system, we

shall take seriously the dictum which says that a differential equation is never
solved until ``boundary'' conditions have been imposed on its solution. As
identified in the ensuing section, these conditions are not arbitrary. Instead, they
fall into three archetypical classes determined by the nature of the physical system
which the partial differential equation conceptualizes.

1. Systems of pde's corresponding to an over-determined system

(61)

2.
The idea for solving it takes advantage of the fundamental subspaces62of [#!

StrangLinearAlgebra!#]. Let be a matrix having rank . Such a matrix,

we recall, has a vector which satisfies , or, to be more precise

(62)
3.

where is any non-zero scalar. Thus spans 's one-dimensional nullspace

4.
5. This expresses the fact that the columnes of are linearly dependent.
6. In addition we recall that the four rows of are linearly dependent also, a fact

which is expressed by the existence of a vector which satisfies

(63)
7.
and which therefore spans 's one-dimensional left nullspace

http://www.amaderforum.com
 http://www.amaderforum.com

8.
9.

In general there does not exist a solution to the over-determined system Eq.(6.1).
However, a solution obviously does exist if and only if satisfies

Under such a circumstance there are infinitely many solutions, each one differing from

any other merely by a multiple of the null vector . The most direct path towards these
solutions is via eigenvectors.

One of them is, of course, the vector in Eq.(6.2). The other three, which (for the

under consideration) are linearly independent, satisfy with , or, in

the interest of greater precision (which is needed in Section 6.2.3),

(64)

(65)

(66)

where, like , the 's are any non-zero scalars. Because of the simplicity of for

the under consideration one can find the eigenvectors , and hence their
eigenvalues, by a process of inspection. These vectors span the range of ,

and therefore determine those vectors for which there exists a solution to Eq.(6.1). Such
vectors belong to and thus have the form

These eigenvectors also serve to represent the corresponding solution,

http://www.amaderforum.com
 http://www.amaderforum.com

This, the fact that , and the linear independence of the imply that the scalars

satisfy the three equations

(67)

(68)

(69)

As expected, the contribution along the direction of the nullspace element is left
indeterminate. These ideas from linear algebra and their application to solving a system,
such as Eq.(6.1), can be extended to corresponding systems of partial differential
equations. The Maxwell field equations, which we shall analalyze using linear algebra, is

a premier example. In this extension the scalar entries of and get replaced by

differential operators, the vectors and by vector fields, the scalars and by scalar

fields, the eigenvalues by a second order wave operator, and the three Eqs.(6.7)- by
three inhomogeneous scalar wave equations corresponding to what in physics and
engineering are called

• transverse electric ( ),
• transverse magnetic ( ), and
• transverse electric magnetic ( ),

modes respectively.

http://www.amaderforum.com
 http://www.amaderforum.com

Single Partial Differential Equations: Their Origin

There are many phenomena in nature, which, even though occurring over finite regions
of space and time, can be described in terms of properties that prevail at each point of
space and time separately. This description originated with Newton, who with the aid of
his differential calculus showed us how to grasp a global phenomenon, for example, the
elliptic orbit of a planet, by means of a locally applied law, for example .

This manner of making nature comprehensible has been extended from the motion of
single point particles to the behavior of other forms of matter and energy, be it in the
form of gasses, fluids, light, heat, electricity, signals traveling along optical fibers and
neurons, or even gravitation.

This extension consists of formulating or stating a partial differential equation governing

the phenomenon, and then solving that differential equation for the purpose of predicting
measurable properties of the phenomenon.

There exist many partial differential equations, but from the view point of mathematics,
there are basically only three types of partial differential equations.

They are exemplified by

1. Laplaces equation

which governs electrostatic and magnetic fields as well as the velocity potential of
an incompressible fluid, by

2. the wave equation

for electromagnetic or sound vibrations, and by

http://www.amaderforum.com
 http://www.amaderforum.com

for the vibrations of a simple string, and by

3. the diffusion equation

for the temperature in three dimensional space and in time, or by

for the temperature along a uniform rod.

http://www.amaderforum.com
 http://www.amaderforum.com

Boundary Conditions of a Typical Partial Differential Equation in Two

Dimensions

For the purpose of simplicity, we shall start our consideration with partial differential
equations in only two variables and linear in the second derivatives. Such equations have
the general form

Such an equation is called a quasilinear second order partial differential equation. If the

expression where linear in , i.e., if

then the equation would be a linear p.d.e., but this need not be the case.

The equation has a nondenumerable infinity of solution. In order to single out a unique

solution, the to-be-found function must satisfy additional conditions. They are
usually specified at the boundary of the domain of the p.d.e.

In three dimensional space, this boundary is a surface, but in our two dimensional case,
we have a boundary line which can be specified by the parametrized curve

where is the arclength parameter

http://www.amaderforum.com
 http://www.amaderforum.com

The tangent to this curve has components

They satisfy

The normal to this boundary curve has components

We assume that points towards the interior of the domain where the solution is to be
found. If this is not the case, we reverse the signs of the components of it.

The additional conditions which the to-be-found solution is to satisfy are imposed at
this boundary curve, and they are conditions on the partial derivatives and the value of

the function evaluated at the curve.

The boundary curve accomodates three important types of boundary conditions.

1. Dirichlet conditions: is specified at each point of the boundary.

2. Neumann conditions: , the normal componet of the graident of

is specified at each point of the boundary.

3. Cauchy conditions: and are specified at each point of the boundary.

The parameter is usually a time parameter. Consequently, Cauchy conditions
are also called intial value conditions or initial value data or simply Cauchy data.

http://www.amaderforum.com
 http://www.amaderforum.com

There exists also the mixed Dirichlet-Neumann conditions. They are intermediate
between the Dirichlet and the Neumann boundary conditions, and they are given by

Here , , and are understood to be given on the boundary.

We recall that in the theory of ordinary second order differential equations, a unique
solution was obtained once the solution and its derivative were specified at a point. The
generalization of this condition to partial differential equations consists of the Cauchy
boundary conditions.

Consequently, we now inquire whether the solution of the partial differential equation is
uniquely determined by specifying Cauchy boundary conditions on the boundary

http://www.amaderforum.com
 http://www.amaderforum.com

Cauchy Problem and Characteristics

In order to compute the function at points off the boundary curve, we resort to
the Taylor series on two dimensions;

Here the derivatives are to be evaluated on the boundary.

The problem we are confronted with is this:

Determine all partial derivatives, starting with the first partials on up from the given
Cauchy boundary conditions, the given boundary, and the given partial differential
equation!

We shall do this first for the first derivatives.

From the Cauchy data we obtain two equations

(610)

From these we obtain the first partial derivatives of evaluates on the boundary

(611)

http://www.amaderforum.com
 http://www.amaderforum.com

The procurement of the second derivatives is more interesting. We differentiate the

(known) first derivatives along the boundary. Together with the given p.d.e. we have

The left hand sides of these three equations are known along the whole boundary. So are
the coefficients of the three unknown partial derivatives on the right hand side. One can
solve for these partial derivatives unless

or

If this determinant does not vanish, one can solve for the second derivatives evaluated on
the boundary. Differentiating along the boundary yields

http://www.amaderforum.com
 http://www.amaderforum.com

Subscripts refer to partial derivatives. The last equation was obtained differentiating the
given p.d.e. with respect to . The left hand side contains only lower order derivatives,
which are known on the boundary.

We see that one can solve for

on the boundary unless the determinant, the same one as before, vanishes. It is evident
that one can continue the process of solving for the other higher order derivatives,
provided the determinant of the system does not vanish. We are led to the conclusion that

one can expand in a Taylor series at every point of the boundary and that the
coefficients of the series are uniquely determined by the Cauchy boundary conditions on
the given boundary.

We must now examine the vanishing of the system determinant

(612)

at every point of the domain of the partial differential equation.

Depending on the coefficients , , and , this quadratic form determines two

characteristic curves, and , through each point

. We distinguish between three cases:

1. : elliptic type in which the two characteristics and are

complex conjugates of each other.

2. : hyperbolic type in which case for each the

characteristics and are real. They sre two curves intersecting at . As

one varies one obtains two distinct families.

http://www.amaderforum.com
 http://www.amaderforum.com

3. : parabolic type in which there is only one family of

characteristics.

These three cases imply three different types of differential equations. By utilizing the
characteristic, one can introduce new coordinates relative to which a differential equation
of each type assumes a standard normal form. Let the new coordinate surfaces be

Then the coordinate transformation

yields a normal form of the elliptic type,

By contrast the coordinate transformation

yields a normal form of the hyperbolic type,

(613)

Finally, the coordinate transformation

yields a normal form of the parabolic type,

We recognize that elliptic partial differential equations express an equilibrium or a static

potential phenomenon.

http://www.amaderforum.com
 http://www.amaderforum.com

By introducing the standard coordinates

in terms of which

one finds that

the wave equation of a general vibrating string. We, therefore, recognize that a
hyperbolic p.d. equation expresses the phenomenon of a propagating wave or
disturbance.

Finally, a parabolic p.d. equation expresses a diffusion process. In fact, the two
dimensional Laplace equation, the equation for a vibrating sting, and the heat
conduction equation are the simplest possible examples of elliptic, hyperoblic, and
parabolic equations.

http://www.amaderforum.com
 http://www.amaderforum.com

Hyperbolic Equations

The quadratic form, Eq.(6.12), determined by the coefficients , , and of the given
p.d.e. can be factored into two ordinary differential equation

These are the equations for the two families of characteristic curves of the given p.d.e.

Their significance, we recall, is this: if the boundary line coincides with one of them, then
specifying Cauchy data on it will not yield a unique solution. If, however, the boundary
line intersects each family only once, then the Cauchy data will yields a unique solution.

This point becomes particularly transparent if one introduces the curvilinear coordinates

and relative to which the given p.d.e. assumes its standard form, Eq.(>6.13). We
shall consider the hyperbolic case by assuming that

throughout the domain.

We shall demand the new coordinates and - the characteristic coordinates - have the
property that their isograms (``loci of points of constant values'') contain the

characteristic lines , i.e.,

for all . This implies that

where, as usual

http://www.amaderforum.com
 http://www.amaderforum.com

Substituting these equations into Eq.(6.12), the equation for the characteristic directions,
one obtains

(614)

An equation with the same coefficients is obtained for the other function . The

two solutions and are real valued functions. Their isograms, the
characteristics of the hyperbolic equation, give us the new curvilinear coordinate system

The partial derivatives of the given differential equation are now as follows

Here refers to additional terms involving only the first partial derivatives of .
Inserting these expressions into the given p.d. equation, one obtains

http://www.amaderforum.com
 http://www.amaderforum.com

(615)

It follows from Equation 6.14 that the coefficients of and vanish. Solving for
yields Equation > 6.13, the hyperbolic equation in normal form.

The coordinates and , whose surfaces contain the characteristic lines, are called the
characteristic coordinates or null coordinates of the hyperbolic equation.

These coordinates are important for at least two reasons. First of all, they are boundaries

across which a solution can be nonanalytic. If is one of the isograms (``locus

of points where has constant value'') of the solution to Eq.(6.14), then the first term of
the p.d. Eq.(6.15)

even if as . In other words, there are solutions to Eq.(6.15) for

which the first derivative has a discontinuity across the characteristic .

Similarly, there exist solutions to Eq.(6.15) whose first derivative has a discontinuity

across whenever satisfies Eq.(6.14) with replaced by .

Secondly, these coordinates depict the history of a moving disturbance. The simple string
illustrates the issue involved.

http://www.amaderforum.com
 http://www.amaderforum.com

Example: The Simple string The governing equation is

Its characteristic coordinates are the ``retarded'' and the ``advanced'' times

and its normal form is

The solution is

where and are any functions of and .

Next consider the initial value data at :

These equations imply

http://www.amaderforum.com
 http://www.amaderforum.com

Consider the intersection of the two families of characteristics with the boundary line
as in the figure below.

Figure 6.1: Characteristic coordinate lines and as determined by the wave equation
for a simple string.

Note that is constant along the characteristics (i.e., where constant), while is

constant along the characteristics. It follows that if is known on the boundary

segment , then is known along all the -characteristics intersecting .

Similarly, if is known along , then is known along all the -characteristics

intersecting . And this is precisely the case because the Cauchy data on

determine the values of both and on that segment.

Being the sum of the two functions, the solution to the wave equation is

(616)

http://www.amaderforum.com
 http://www.amaderforum.com

Thus one sees that any disturbance on a string consists of two parts: one propagating to

the right the other to the left. The propagation speeds are , the slopes of the
characteristics relative to the - coordinate system. The idiosyncratic aspect of the
simple string is that these two parts do not change their shape as they propagate along the
string.

A general linear hyperbolic system does not share this feature. However, what it does
share with a simple string is that its solution is uniquely determined in the common
region traversed by the two sets of characteristics which intersect . In fact, the

Cauchy data on determine a unique solution at every point in the region .

This is why it is called the domain of dependence of . To justify these claims it is
neccessary to construct this unique solution to a general second order linear hyperbolic
differential equation.

http://www.amaderforum.com
 http://www.amaderforum.com

Riemann's Method for Integrating the Most General 2nd Order Linear
Hyperbolic Equation

In its most general form a linear second order hyperbolic equation is

(617)

In compliance with standard practice one designates the characteristic coordinates by

and . The problem to be solved is this:

Given

(a) the differential Eq.(6.17) and

(b) the initial value data (=Cauchy conditions) and its normal derivative on
the given curve in Figure 6.2,

Find: the function which satisfies (a) and (b).

http://www.amaderforum.com
 http://www.amaderforum.com

Figure 6.2: Characteristic coordinates of a hyperbolic differetial equation in two

dimensions.

Riemann's method of solving this problem is a three-step process whose essence parallels
the Green's function method

http://www.amaderforum.com
 http://www.amaderforum.com

System of Partial Differential Equations: How to Solve Maxwell's

Equations Using Linear Algebra

The theme of the ensuing development is linear algebra, but the subject is an
overdetermined system of partial differential equations, namely, the Maxwell field
equations. The objective is to solve them via the method of eigenvectors and eigenvalues.
The benefit is that the task of solving the Maxwell system of p.d. equations is reduced to
solving a single inhomogeneous scalar equation64

where is a time and space dependent source. The impatient reader will find that once
this master equation, or its manifestation in another coordinate system, has been solved,
the electric and magnetic fields are entirely determined as in Tables -6.9.

The starting point of the development is Maxwell's equations. There is the set of four
functions, the density of charge

(637)

and the charge flux

(638)

which are usually given. These space and time dependent charge distributions give rise to

electric and magnetic fields, and . The relationship is captured by

means of Maxwell's gift to twentieth century science and technology,

(639)

http://www.amaderforum.com
 http://www.amaderforum.com

(640)

and

(641)

(642)

Maxwell's field equations65.

Exercise 62.1 (Charge Flux-Density of an Isolated Charge)

Microscopic observations show that charged matter is composed of discrete point
charges. On the other hand, macroscopic observations show that charged matter is the
carrier of an electric fluid which is continuous. Dirac delta functions provide the means to
grasp both attributes from a single perspective. This fact is highlighted by the following
problem.

Consider the current-charge density due to an isolated moving charge,

a) Show that this current-charge density satisfies

http://www.amaderforum.com
 http://www.amaderforum.com

Remark. The four-vector is the charge's four-velocity in spacetime. The

parameter is the ``wristwatch'' time (as measured by a comoving clock) attached to this
charge.

b) By taking advantage of the fact , evaluate the -integrals, and obtain

explicit expressions for the components and .

Answer:

where

http://www.amaderforum.com
 http://www.amaderforum.com

Maxwell Wave Equation

The first pair of Maxwell's equations, (6.39) and (6.40), imply that there exists a vector

potential and scalar potential from which one derives the electric and magnetic
fields,

(643)

(644)

Conversely, the existence of these potentials guarantees that the first pair of these
equations is satisfied automatically. By applying these potentials to the differential
expressions of the second pair of Maxwell's equations, (6.41-6.42) one obtains the
mapping

(645)

where

(646)

It follows that Maxwell's field equations reduce to Maxwell's four-component wave

equation,

http://www.amaderforum.com
 http://www.amaderforum.com

(647)

Maxwell's wave operator is the linch pin of his theory of electromagnetism. This is
because it has the following properties:

1. It is a linear map from the space of four-vector fields into itself, i.e.

at each point event .

2. The map is singular. This means that there exist nonzero vectors and such
that

and

3.

4. In particular, one has

1. the fact that

http://www.amaderforum.com
 http://www.amaderforum.com

for all three-times differentiable scalar fields . Thus

The null space of is therefore nontrivial and 1-dimensional at each

2. the fact that

for all 4-vectors . Thus

(649)

or

(650)

The left null space of is therefore also 1-dimensional at each .

In light of the singular nature of , the four-component Maxwell wave equation

(651)

http://www.amaderforum.com
 http://www.amaderforum.com

has no solution unless the source also satisfies

This is the linear algebra way of expressing

(652)

the differential law of charge conservation. Thus Maxwell's equations apply if and only if
the law of charge conservation holds. If charge conservation did not hold, then Maxwell's
equations would be silent. They would not have a solution. Such silence is a
mathematical way of expressing the fact that at its root theory is based on observation
and established knowledge, and that arbitrary hypotheses must not contaminate the
theoretical.

http://www.amaderforum.com
 http://www.amaderforum.com

The Overdetermined System

The linear algebra aspects of Maxwell's wave operator are illustrated by the following
problem from linear algebra:

Solve for , under the stipulation that

(653)

The fact that is singular and belongs to the range of makes the system
over-determined but consistent. This means that there are more equations than
there are unknowns.
One solves the problem in two steps.

Step I:

Let be the set of eigenvectors having non-zero eigenvalues. Whatever

is, the task of finding three vectors that satisfy

(654)

and

(655)

http://www.amaderforum.com
 http://www.amaderforum.com

Being spanned by the three eigenvectors with non-zero eigenvalues, the range
space of ,

is well-determined. However, the scalars are at this stage as yet undetermined.

Step II:
Continuing the development, recall that quite generally

(656)

and that if

then

(657)

It is appropriate to alert the reader that in the ensuing section the vectors and the

eigenvalues become differential operators which act on scalar fields and that the
three subscript labels will refer to the TE, TM, and TEM eletromagnetic66 vector
potentials respectively.

Equating (6.56) and (6.57), one finds that the linear independence of

implies the following equations for , , and :

http://www.amaderforum.com
 http://www.amaderforum.com

(658)

(659)

(660)

Consequently, the solution is

where is an indeterminate multiple of the null space vector .

If one represents the stated problem ( determines ) as an input-output

process, as in Figure 6.3,

Figure 6.3: The matrix defines an input-output process.

then its solution is represented by the inverse input-output process as in Figure 6.4.

Figure: The solution to defines an inverse input-output process.

http://www.amaderforum.com
 http://www.amaderforum.com

In general, the task of finding the eigenvectors of a 4 4 matrix can be a nontrivial task.
However, given the fact that the solution to

is already known, one finds that the associated constraints on the eigenvectors,

make the task quite easy, if not trivial.

http://www.amaderforum.com
 http://www.amaderforum.com

Maxwell Wave Equation (continued)

The above linear algebra two-step analysis of an overdetermined (but consistent) system
is an invaluable guide in solving Maxwell's wave equation

(661)

http://www.amaderforum.com
 http://www.amaderforum.com

Cylindrical Coordinates

The benefits of the linear algebra viewpoint applied to Maxwell's equations can be
extended by inspection from rectilinear cartesian to cylindrical coordinates. This is
because the four-dimensional coordinate system lends itself to being decomposed into
two orthogonal sets of coordinate surfaces. For cylindricals these are spanned by the

transverse coordinates in the transverse plane, and the longitudinal coordinates

in the longitudinal plane.

The transition from a rectilinear to a cylindrical coordinate frame is based on the

replacement of the following symbols:

(677)

(678)

and

(679)

Such a replacement yields the vector field components relative to an orthonormal (o.n.)
basis tangent to the coordinate lines. To emphasize this orthonormality, hats ( ) are
placed over the vector components.

This replacement is very powerful. It circumvents the necessity of having to repeat the
previous calculations that went into exhibiting the individual components of Maxwell's
, , and systems of equations. We shall again take advantage of the
power of this algorithm in the next section when we apply it to Maxwell's system relative
to spherical coordinates.

Applying it within the context of cylindrical coordinates, one finds that the source and the
vector potential four-vectors are as follows:

1. for a source

http://www.amaderforum.com
 http://www.amaderforum.com

(680)

2.

3. the solution to the Maxwell field equations has the form

(681)

4.

5.
6. for a source

(682)

7.

8. the solution to the Maxwell field equations has the form

(683)

9.

10. and
11. for a source

(684)

12.

13. the solution to the Maxwell field equations has the form

(685)

14.

15.

http://www.amaderforum.com
 http://www.amaderforum.com

Spherical Coordinates

One of the chief virtues of the linear algebra viewpoint applied to Maxwell's equations is
that it directs attention to the system's fundamental vector spaces and their properties.
The easiest way to identify them in a computational way happens when the underlying
coordinate system permits a 2+2 decomposition into what amounts to longitudinal and
transverse surfaces. Spherical coordinates provide a nontrivial example of this. There a

transverse surface is a sphere spanned by , while the longitudinal coordinates are

The distinguishing feature of spherical coordinates, as compared to rectilinear or

cylindrical coordinates, is that coordinate rectangles on successive transverse surfaces
(nested spheres) are not congruent. Instead, they have areas that scale with the square of
the radial distance from the origin. This scaling alters the representation of the divergence
of a vector field and hence the Maxwell wave operator. Nevertheless, the eigenvalue
method with its resulting TE-TM-TEM decomposition of the e.m. field readily
accommodates these alterations.

http://www.amaderforum.com