Partial Differential Equations I: Basics and Separable ...

18

Partial Differential Equations I: Basics and Separable Solutions

We now turn our attention to differential equations in which the "unknown function to be determined" -- which we will usually denote by u -- depends on two or more variables. Hence the derivatives are partial derivatives with respect to the various variables.

(By the way, it may be a good idea to quickly review the A Brief Review of Elementary Ordinary Differential Equations, Appendex A of these notes. We will be using some of the material discussed there.)

18.1 Intro and Examples

Simple Examples

If we have a horizontally stretched string vibrating up and down, let

u(x, t) = the vertical position at time t of the bit of string at horizontal position x ,

and make some almost reasonable assumptions regarding the string, the universe and the laws of physics, then we can show that u(x, t) satisfies

2u t2

-

c2

2u x2

=

0

where c is some positive constant dependent on the physical properties of the stretched string. This equation is called the one-dimensional wave equation (with no external forces).

If, instead, we have a uniform one-dimensional heat conducting rod along the X?axis and let

u(x, t) = the temperature at time t of the bit of rod at horizontal position x ,

then, after applying suitable assumptions about heat flow, etc., we get the one-dimensional heat

equation

u t

-

2u x2

=

f (x, t)

.

Here, is a positive constant dependent on the material properties of the rod, and f (x, t)

describes the thermal contributions due to heat sources and/or sinks in the rod.

3/8/2014

Chapter & Page: 18?2

PDEs I: Basics and Separable Solutions

The physicists in the class, of course, are also well acquainted with Schr?dinger's equation

i

h?

u t

+

h? 2 2u 2m x2

=

V (x)u

where h? and m are positive constants and V (x) is some potential energy function.

Similar problems involving two- and three-dimensional objects lead to similar partial differential equations with 2 replacing 2u/x2 .1 Thus, we have the multi-dimensional wave equation,

2u t2

-

c22u

=

0

,

and the multi-dimensional heat equation,

u - 2u = f (x, t) . t

If, in either of these, we reach an "equilibrium" or "steady-state" (i.e., u/t = 0 ), then these equations reduce to either Laplace's equation

or Poisson's equation

2u = 0 2u = f (x) .

Basic Terminology

The order of a partial differential equation is the order of the highest derivative explicitly appearing. In practice, most partial differential equations of interest are second order (a few are first order and a very few are fourth order). We will concentrate on second-order "linear" equations.2

A second-order partial differential equation (in variables x1 , x2 , ..., xn ) is said to be linear if it can be written as

jk

a

j

k

2u xk x

j

+

l

bl

u xl

+ cu =

f

.

(18.1)

where f , c and the a jk's and bl's are constants or are functions of the variables (but not u ). This equation is said to be homogeneous if (and only if) f 0 . We will concentrate our initial attention on homogeneous equations because they are a little easier to deal with and because we must solve a homogeneous equation anyway to get the complete solution to any nonhomogeneous linear equation.

For convenience, we'll occasionally let L denote the differential operator defined by the left side of equation (18.1),

2

L = jk a jk xk x j + l bl xl + c .

1 Remember: 2u is the divergence of the gradient of u . In Cartesian coordinates,

2u

=

2u x2

+

2u y2

+ ???

.

2 Take MA 506 or MA 526 to learn how to reduce first-order, linear partial differential equations to first-order, linear ordinary differential equations.

Intro and Examples

Chapter & Page: 18?3

That is, for any sufficiently differentiable function w ,

L[w] =

jk

a

j

k

2w xk x

j

+

l

bl

w xl

+ cw

.

Using this, equation (18.1) can be written more succinctly as

L[u] = f ,

and, if it is homogeneous, as

L[u] = 0 .

You should be aware that second-order linear partial differential equations are often classified as being "hyperbolic", "parabolic" or "elliptic". Crudely speaking:

Hyperbolic partial differential equations are partial differential equations like the wave equation,

2u t2

-

c22u

=

0

.

Parabolic partial differential equations are partial differential equations like the heat equation, u - 2u = 0 . t

Elliptic partial differential equations are partial differential equations like Laplace's equation, 2u = 0 .

For an intelligent discussion of the "classification of second-order partial differential equations", take a true partial differential equation course (MA 506 or MA 526-626).

Linearity/Principle of Superposition

Letting

L=

jk

a

j

k

2 xk

x

j

+

l

bl

xl

+c

,

it is easily verified that, given a bunch of constants -- c1 , c2 , c3 , . . . -- and a corresponding bunch of sufficiently differentiable functions -- u1 , u2 , u3 , . . . -- then

L[c1u1 + c2u2 + c3u3 + ? ? ? ] = c1 L[u1] + c2 L[u2] + c3 L[u3] + ? ? ? .

Thus, L is a linear differential operator. In particular, if u1 , u2 , u3 , . . . are all solutions to the homogeneous equation

L[u] = 0 ,

then L[c1u1 + c2u2 + c3u3 + ? ? ? ] = c1 L[u1] + c2 L[u2] + c3 L[u3] + ? ? ? = c1 ? 0 + c2 ? 0 + c3 ? 0 + ? ? ? =0 .

version: 3/8/2014

Chapter & Page: 18?4

PDEs I: Basics and Separable Solutions

This gives us

Theorem 18.1 (principle of superposition) Any linear combination of solutions to a homogeneous linear partial differential equation is also a solution to that homogeneous partial differential equation.

We will use this often, even with linear combinations involving infinitely many terms (and, at times, slop over issues of the convergence of the resulting infinite series).

At this point we should spend a few seconds to observe that

L[0] =

jk

a

j

k

2 0 xk x

j

+

l

bl

0 xl

+ c?0 = 0

.

So the constant function u = 0 is a solution to every homogeneous linear partial differential equation. This not-so-exciting solution is often called the trivial solution. Our main interest, of course, will be in the nontrivial solutions.

General Solutions

In general, we cannot find "general solutions" (i.e., relatively simple formulas describing all possible solutions) to second-order partial differential equations.3 The one notable exception is

with the one-dimensional wave equation

2u t2

-

c2

2u x2

=

0

.

Using a clever change of variables, it can be shown that this has the general solution

u(x, t) = f (x - ct) + g(x + ct)

(18.2)

where f and g are arbitrary sufficiently differentiable functions of a single variable.4

?Exercise 18.1: Verify that, if f (s) and g(s) are any two twice-differentiable functions of one variable, then u(x, t) = f (x - ct) + g(x + ct)

satisfies

2u t2

-

c2

2u x2

=

0

.

Observe that, at any given time t , the graph of

y = f (x - ct)

3 General solutions to first-order linear partial differential equations can often be found. 4 Letting = x + ct and = x - ct the wave equation simplifies to

2u =0 .

Integrating twice then gives you u = f () + g( ) , which is formula (18.2) after the change of variables.

Separation of Variables for Partial Differential Equations (Part I)

Chapter & Page: 18?5

is just the graph of y = f (x) shifted to the right by ct . Thus, the f (x + ct) part of formula (18.2) can be viewed as a "fixed shape" traveling to the right with speed c . Likewise, the g(x + ct) part of formula (18.2) can be viewed as a "fixed shape" traveling to the left with speed c . Unsurprisingly, these are generally known as traveling waves.

?Exercise 18.2: Illustrate these last few statements by sketching

u(x, t) = f (x - ct)

as a function of x at t = 0 , t = 1 , t = 2 and t = 3 , using, say, c = 2 and either

f (s) = arctan(s) or

1 f (s) =

0

if -1 < s < 0 .

otherwise

?Exercise 18.3: Let f (s) be any twice-differentiable function of one variable and let n be any unit vector. Verify that u(x, t) = f (n ? x - ct)

is a solution to the three-dimensional wave equation

2u t2

-

c22u

=

0

.

(This is a plane wave solution -- f (n ? x - ct) remains constant on planes perpendicular to n and traveling with speed c in the direction of n .)

18.2 Separation of Variables for Partial Differential Equations (Part I)

Separable Functions

A function of N variables

u(x1, x2, . . . , xN )

is separable if and only if it can be written as a product of two functions of different variables,

u(x1, x2, . . . , xN ) = g(x1, . . . , xk) h(xk+1, . . . , xN ) .

It is completely separable if and only if it can be written as a product of N functions, each of which is a function of just one variable,

u(x1, x2, . . . , xN ) = g1(x1) g2(x2) g(x3) ? ? ? gN (xN ) .

!Example 18.1: The following functions are all separable: u(x, t) = e-6t sin(x)

version: 3/8/2014

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download