ACM95b Notes
TABLE OF CONTENTS
1st Order ODEs ...........................................................................................................................................................................................4
Linear 1st Order ODEs ................................................................................................................................................................................4
General Strategy (Leibniz) .........................................................................................................................................4
Existence and Uniqueness .........................................................................................................................................5
Complex Case ............................................................................................................................................................5
nd
Giordon Stark
ACM95b Notes
Giordon Stark
ACM95b Notes
INTRODUCTION
These notes were taken during the 2010 Winter term of ACM95. At the time, Niles Pierce was the instructor and these notes come from the lectures he
held. This version of the notes is meant to be the finalized version, with (hopefully) no typos, and less spaces so that it can be read like a novel.
Similarly, Sean Machs textbook (found in the Underground) will do just as well. Originally, I had no purpose other than to electronic-ize all my work
(problem sets, for example). However, I realized that after a while, a lot of people started getting interested in having electronic notes. Getting some
inspiration from other students who had done similar things, I decided to consistently post all my notes on my ITS student account. Without further ado,
enjoy!
-
Giordon
Giordon Stark
ACM95b Notes
LECTURE 1
1ST ORDER ODES
= (, )
Pick an integral curve (1,2,3, for example) by providing initial condition (0) = 0 .
General Form
+ () = (),
= ,
(0) = 1
= ,
= 0
So,
(0 ) = 0
EXAMPLE
= ln|| = + 0 || = 0
(0) = 0 = 1 = 1
() + () = ()
(1)
Giordon Stark
ACM95b Notes
Equivalent to
() =
(2)
+ =
(1)
And then integrate as in example. How do we find ()? Compare (1) and (2).
+ =
(2)
Then solve
()
(0 ) = 1
1
0
() =
()() +
()
()
0
(0 ) = 0
Unique solution exists throughout an open interval containing 0 in which () and () are continuous (proof is
constructive).
Moral: just inspect ODE to identify interval of guaranteed existence and uniqueness. You may lose existence and
uniqueness at a singular point where () or () is not continous.
EXAMPLE
(0 ) = 0
So () = 0,
Then () =
() =
If 0 0: () =
0
, "all is well"
0
(0 ) 0 Solution DNE
If 0 = 0:
(0 ) = 0 infinite solutions not unique all integral curves pass through origin
COMPLEX CASE
+ () = ()
Giordon Stark
ACM95b Notes
() =
() ()
1
(0 )
() =
()() +
()
()
0
Analytic in since () 0.
EXAMPLE
+ = 0
= integer 0 () is entire
LECTURE 2
2 N D ORDER LINEAR IVPS
+ () + () = ()
Homogeneous if () = 0, nonhomogeneous otherwise. IVP requires two initial conditions (roughly, to fix two
constants of integration)
(0 ) = 0 ,
(0 ) = 0
If , , are continuous in an open interval, then a unique solution to the IVP through the interval (proof
Coddington)
EXAMPLE
(1 2 ) + (1 + ) = 1 ,
(0) = 1,
(0) = 1
Giordon Stark
= 0,
ACM95b Notes
1
,
1
1
1 +
SOLUTION PROPERTIES
SATISFIABILITY
1 1 (0 ) + 2 2 (0 ) = 0
1 (0 )
1 (0 )
2 (0 )
= 1 2 1 2
2 (0 )
GENERALITY
0 = (0 )
Giordon Stark
ACM95b Notes
2 (1 + 1 + 1 ) = 0
1 (2 + 2 + 2 ) = 0
(1 2 1 2 ) + (1 2 1 2 ) = () + () = 0
3 = 2 3
3 2
LINEAR INDEPENDENCE
(3)
. If (3) requires 1 = 2 = 0, then 1 and 2 are linearly independent on . Can show: 1 and 2 are
linearly independent on iff () 0 . (proof Coddington)
SUMMARY
nd
To find general solution to homogeneous 2 order linear ODE, find two linearly independent solutions with
() 0 in .
LECTURE 3
HOMOGENEOUS 2 N D ORDER LINEAR ODES WITH CONSTANT
COEFFICIENTS
+ + = 0
(1)
1 = 1 ,
2 = 2 ,
1 2
2 4
2
Giordon Stark
ACM95b Notes
1
1 1
2
= (2 1 ) (1 +2 ) 0
2 2
General Solution: = 1 1 + 2 2
0,
1 , 2 = ( )
1 + 2
= cos ,
2
= 2 0,
1 2
= sin
2
= ( 1 )2
( )
=0
=1
Also, note
So
And so 1 = 1 , 2 = 1
=
1
1 1
1
= 21 0
1 1 + 1
General Solution: = 1 1 + 2 1
8 + 16 = 0,
(0) = 1,
EXAMPLE
(0) = 5
( 2 8 + 16) = 0 1 = 2 = 4 1 = 4
Try 2 = () 4 :
+ +
21
1
= 0 + (8 + 8) = 0 = 0 or () = +
Giordon Stark
ACM95b Notes
1 + 2 0 = 1
41 + 2 = 5 1 = 1, 2 = 1
+ () + () = ()
()
p,q,h continuous on interval . Suppose is a solution to () and is any other solution to (). Then = 0.
So = 1 1 () + 2 2 () with 1 and 2 linearly independent solutions to homogeneous problem
()
Hence any solution to () may be written (i.e., general solution to ()) = 1 2 + 2 2 + where = 1 1 +
2 2 is the complementary solution and is any particular solution to (). How to find ?
EXAMPLE
2 = ,
(0) = 2,
( 2 2) = ( + 1)( 2)
(0) =
2
3
1 + 22 2 =
1
+ 2 2
3
1 ()2 () 1 ()2 ()
() =
() = (, )()
(1 , 2 )()
1 ()2 () 1 ()2 ()
(1 , 2 )()
LECTURE 4
10
Giordon Stark
ACM95b Notes
Define intuitively by
0,
( 0 ) =
,
0
,
= 0
( 0 ) = 1
( 0 )() = (0 )
1,
( 0 ) = ( 0 ) =
0,
So we have
( 0 )
= ( 0 )
> 0
< 0
nd
() = (, )()
0
11
(1)
Giordon Stark
ACM95b Notes
() = (, )( )
0
And by the sifting property () = (, ). So (, ) is the response of the solution of due to an impulse at
. Hence Greens function satisfies
(2)
[(, )] = ( )
(, )|= 0 = 0
(, )|= 0 = 0,
2
+ ()
+ () = ( )
2
Highest derivative balances ( ) so expect to jump because it represents an integral of an impulse, expect
+ ()
+ () = 1
= 1,
| = 0
So
2 ()
,
(1 , 2 )()
2 () =
1 ()
(1 , 2 )()
0,
(, ) = 1 ()2 () 1 ()2 ()
,
(1 , 2 )()
<
>
() = (, )() + (, )()
0
The first integral is non-zero as > ( is upper limit) while the second integral is zero as < ( is lower limit).
() =
12
1 ()2 () 1 ()2 ()
()
(1 , 2 )()
Giordon Stark
ACM95b Notes
EXAMPLE (REVISITED)
2 = ,
So
1 = ,
2 = 2 ,
0 = 0
2 = ( )
2
At =
(0, ) = 0,
+
|
=1
= 0,
So = 1 () + 2 () 2 with 1 () + 2 () 2 = 0 and 1 () + 22 () 2 = 1. So 2 () =
1 () = .
3
0,
(, ) = 1 1 22
+
,
3
3
2
3
and
<
>
() = (, )() + (, )() = 0
0
= + (1 1 + 2 2 = 0 for convenience)
3
LECTURE 5
IVPS WITH CONSTANT COEFFICIENTS
EXAMPLE
6 = 0,
(0) = 2,
(0) = 1
{} =
13
2 1
()
6
(i)
Giordon Stark
ACM95b Notes
It turns differential equation into algebraic expression. To solve IVP, need to find whose transform is (i). By
partial fractions
() = { ()} =
1
1
+
3 + 2
1
,
>
SOLUTION STRATEGY
t-space
= ,
(0) = 1,
(0) = 1
Subsidiary equation: ( 2 1) 1 =
Some algebra: =
2 1
Solution: () = + sinh
Denoted
() = ()
0
() = 1 {()}
UNIQUENESS
Two functions with the same transform cannot differ over any interval of positive length (Lerchs Theorem). Hence,
if () is a continuous function with transform (), then it is the only possible one. We can build lookup tables of
() () invert simple transforms.
14
Giordon Stark
ACM95b Notes
CONVOLUTION INTEGRALS
Suppose (), (), () have transforms (), (), () for > 0. If () = ()(), then () is the
convolution of () and ().
() = ( )() = ()( )
0
PROOF SKETCH
() = ()() =
() ()
0
= () ( +) ()
0
Let = so =
= () ( )
0
=
0
( )()
0
= ()
0
15
Giordon Stark
ACM95b Notes
So
() = ( )()
0
() =
EXAMPLE
1
,
Then () = , () =
() =
() =
() = = = 1
0
LECTURE 6
SHIFTING THEOREMS
S-SHIFTING
If { ()} = (), > , i.e. |()| , , then { ()} = ( ), > or
1 {( )} = ()
( ) =
EXAMPLE
()
() = [ ()] = { ()}
0
Note
16
= cos ,
2 + 2
> 0
Giordon Stark
ACM95b Notes
So if
() =
+ 1
,
( + 1)2 + 22
+ 1 > 0
Then () = cos 2.
T-SHIFTING
() =
Let = + , so
Note:
Because
() =
0
{( )} =
{1} =
EXAMPLE
1
,
() = (+) ()
0
( ) = ( )( ) = {( )( )}
> 0,
0,
() =
cos ,
17
> 0
{()} =
1
,
0 < < 2
> 2
> 0
,
() =?
Giordon Stark
ACM95b Notes
() = ( 2) cos = ( 2) cos( 2)
So
{()} = 2
+1
(0) = 0
(0) = 0,
() = ()(1 )
With () =
(+1)(+2)
() =
1
. Then
1
2
2
+
+ 1 + 2
1
+ 2 2 ,
( 1) 2( 2 1) 2 ,
lim () =
18
2 1
,
2
lim
1+ 2
1
> 1
1
1
Giordon Stark
ACM95b Notes
Jump in highest derivative. [ ] = 1 at = 1 (brackets denote jump notation) to match jump in forcing term
[] = 1 at = 1.
Let () = ( ). Delta function fails to satisfy previous sufficient conditions for existence of the Laplace
Transform (proceed formally).
{( )} = ( ) =
0
EXAMPLE
= + 3 + 2 = ( ),
2 + 3 + 2 =
(0) = 0,
(0) = 0
So () = () with
() =
So
1
1
+ 1 + 2
() = 1 {()} = 2
() = ( )( ) =
0,
0 <
()
2()
,
>
= (, )()
0
LECTURE 7
19
Giordon Stark
ACM95b Notes
() = ()
1
() =
()
2
|()| <
And piecewise continuous (well derive these later in the course). Consider a piecewise continuous () of
exponential order (recall |()| , ): apply Fourier formulae to a modified function.
-
() ()
() could blow up as +
() removes blow up as
Then
() = ()
()
= () (+ )
0
1
() =
() (+ )
2
() = () ,
0
() = >
We obtain the Laplace Transform. We can show () is analytic for () > (Weinberger, Dover 1965).
Analytically continue for () , changing variables
=
20
< < +
Giordon Stark
ACM95b Notes
1
()() =
()
2
Omit () in the future by remembering that () = 0 for < 0. This is the Mellin inversion formula
+
1
() =
()
2
If |()| , , then > is to the right of all singularities in () since () is analytic for () > .
Using
21
Giordon Stark
ACM95b Notes
If
() = 2 { (); }
=1
= = 0,
() 0 as by Jordans Lemma
On 1
| | = (()+) = ()
Likewise for 2 .
This leaves
lim () = 2 { (); }
( > 0) =
() =
1
So |()| ||2
So
1 {()}
=1
=1
EXAMPLE
1
2 + 2
|| 0 as || . Simple poles at 1 = , 2 = .
=
2
+ 2
2 = 2
() =
22
1
=
() = { (); }
2
sin
=
,
2
2
> 0
Giordon Stark
ACM95b Notes
LECTURE 8
NONLINEAR 1 S T ORDER ODES
General Form
Facts:
= (, ),
(1)
(0 ) = 0
= 0. Settle for local existence and uniqueness result near (0 , 0 ). Moral: simple inspection of ODE not sufficient
to judge solution behavior.
PICARDS THEOREM
Suppose that (, ) is a continuous function in the rectangle = {(, ): | 0 | , | 0 | } ( > 0,
> 0 constants) with |(, )| (, ) and Lipschitz constant such that
|(, 2 ) (, 1 )| |2 1 |
Whenever (, 1 ) and (, 2 ) lie in . Then there exists a unique solution () satisfying the IVP (1) on the interval
= {: | 0 | min(, )}. (Proof Coddington). The proof relies on the Picard iteration
0 () = 0
(2a)
+1 () = 0 + (, ()
0
23
(2b)
Giordon Stark
ACM95b Notes
Consider
= (, )
(3)
(0 ) = 0
(4)
Solve on interval 0 . Discretize the interval into mesh points = 0 + , = 0,1, , with step size
EULERS METHOD
Discretize the derivative
Or
+1
= ( , ),
= 0,1, , 1
+1 = + ( )
(5)
ALTERNATIVE DERIVATIONS
(6)
(+1 ) = () + , () ,
= 0,1, , 1
, () , ( )
OTHER POSSIBILITIES
IMPLICIT OR BACKWARD EULER
(+1 ) = ( ) + +1 , (+1 )
24
(7)
Giordon Stark
ACM95b Notes
, ( ) + +1 , (+1 )
2
(8)
TRAPEZOID RULE
(+1 ) = ( ) +
These methods are implicit: must solve implicit nonlinear equation to solve for +1 given .
EXAMPLE
2 +
= 2
,
+ 3 2
(0) = 1
a)
Explicit Euler
b) Implicit Euler
c)
Trapezoidal
On a sequence of meshes. The implicit methods were updated using a Picard iteration to solve the nonlinear
equation within each time step.
LECTURE 9
TRUNCATION ERROR
It is the degree to which the exact solution fails to satisfy the discretization. Consider the ODE
= (, )
With discretization
+1
= ( , )
Substitute Taylor series for analytical solution into discrete operator to find the truncation error:
25
(Explicit Euler)
Giordon Stark
So
ACM95b Notes
( ) + ( ) +
2
( ) + ( 3 ) ( )
2
, ( )
= ( ) , ( ) +
Or by Taylors Theorem
1
= ( ),
2
( ) + ( 2 )
2
( , +1 )
EXPLICIT EULER
1
= () = ()
2
IMPLICIT EULER
+1
= (+1 , +1 )
2
(+1 ) ( 3 )
2
1
(+1 , +1 ) = ( )
2
TRAPEZOID RULE
+1 1
= [( , ) + (+1 , +1 )] = ( 2 )
some math
Where ( ) was the exact solution and was the numerical (or approximate) solution. If
| | = 1 = 2 ,
= time step,
= # of steps
Then log| | = log 1 + log = log 2 log . Plot log(error) versus log(steps) and find slope of the line to
estimate .We observed:
explicit Euler, 1
implicit Euler, 1
Trapezoid rule, 2
26
Giordon Stark
ACM95b Notes
The order of accuracy of the truncation error matches the empirical behavior of the solution error in our example.
Can we predict this? Consider the explicit Euler method
+1 = + ( , )
(1)
( +1 )( )
, ( ) or
(+1 ) = ( ) + , ( ) +
(2)
Then by induction
|1 | (1 + )|0 | +
|2 | (1 + )2 |0 | + (1 + ) +
Note:
| | (1 + ) |0 | + [1 + (1 + ) + (1 + )2 + + (1 + )1 ]
1
= 1 + + 2 + + 1
1
(1 + ) 1
= 1 + (1 + ) + + (1 + )1
(1 + ) 1
( )
0 1 + ( 0 ) |0 |
(3)
Hence the solution error should decrease at the same rate as the truncation error as we refine the mesh (which we
observed in our numerical example).
LECTURE 10
EXAMPLE (ERROR CONTROL)
Solve = arctan , (0) = 0 using explicit Euler with | | . By our error bound
27
Giordon Stark
ACM95b Notes
| |
( 1) + |0 |
(4)
Set initial condition correctly so 0 = 0. We need to find valid values of , . The truncation error is
So
1
= ( )
2
Thus
arctan
(arctan ) =
=
,
2
1 +
1 + 2
. To find , note
(, 1 ) (, 2 ) =
For
| | |arctan |
1
2
1 +
2
(, )
In our case
(, )
=
1
1 + 2
So = 1 is a valid choice (maybe not the best, but it works). By (4) we then have
| |
( 1),
4
| |
( 1)
4
We now require
Or
4
( 1)
= 1, ,
REMOVING IMPLICITNESS
nd
Trapezoidal method gives 2 order accuracy but it is implicit requires a nonlinear iteration at each step.
+1 = +
[( , ) + (+1 , +1 )]
2
We can try approximating +1 by forward Euler step +1 + ( , ) to obtain the Improved Euler
Method.
Giordon Stark
+1 = +
ACM95b Notes
( , ) + +1 , + ( , )
2
1-STAGE SCHEME
1 = ( , )
+1 = + 1
st
= ()
2-STAGE SCHEME
1 = ( , )
2 = ( + , + 1 )
1 + 2
+1 = +
nd
= ( 2 )
4-STAGE SCHEME
1 = ( , )
2 = +
3 = +
, + 1
2
2
, + 2
2
2
4 = ( + , + 3 )
1 + 22 + 23 + 4
+1 = +
th
= ( 4 )
29
Giordon Stark
ACM95b Notes
2 = 3
=
1
st
= (, 1 , 2 , , )
So 1 order systems are a unifying framework for ODE theory. It is not immediately obvious that Runge-Kutta
st
methods should work for 1 order nonlinear systems but it can be shown rigorously that they do (Lambert,
1991). Whats different? We treat , , as vector functions. The order of accuracy remains the same as the scalar
case.
LECTURE 11
LINEAR STABILITY ANALYSIS
Consider model problem
= ,
(0) = 0 0,
() < 0
+1 = (1 + ) ,
= 0,1,
1.0
0.5
2.0
1.5
1.0
0.5
0.5
0.5
1.0
1.5
30
1.0
Rez
Giordon Stark
ACM95b Notes
4-STAGE RUNGE-KUTTA
+1 = () ,
|()| < 1
+1 = + +1 +1 =
So () =
BACKWARD EULER
1
< 1 or |1 | > 1.
1.0
0.5
1.0
0.5
0.5
1.0
1.5
2.0
Rez
0.5
1.0
1.5
The stability region if , < 0 then (0, ). Moral: In practice, we must limit the time step to ensure
sufficient accuracy.
LECTURE 12
LINEAR EQUATIONS WITH ANALYTIC COEFFICIENTS
Consider
+ () + () = 0, (0 ) = 0, (0 ) = 0
31
(1)
Giordon Stark
a)
ACM95b Notes
(2)
() = ( 0 )
=0
(3)
0 > 0
Then there exists a series solution to the IVP (1) of form (2) that converges at least throughout interval (3). (Proof
Coddington). Note: find interval of convergence for () and () series by
a)
Ratio test
EXAMPLE
(0 ) = 0 , (0 ) = 0 , 0 =
(1 + 2 ) + 2 + 4 2 = 0,
5
2
1
2
Note: () and () are continuous for < < so there exists a unique solution () for < < but
may not have power series about 0 that converges for all .
Recall: termwise differentiation, addition, and multiplication are valid inside circle of convergence.
32
Giordon Stark
ACM95b Notes
= 0, < <
(4)
() and () are analytic for all 0 : every point is an ordinary point. Note:
= 0 1,2 = , monotonic
+ = 0 1,2 = , oscillatory
= 0 + 1 + 2 + + + =
=0
= 1 + 22 + 33 + +
= 22 + + ( 1)
= ( + 2)( + 1)+2 ,
=0
So to satisfy (4)
+ = 1
=1
+ = ( 1) 2
=2
= 1
=1
= 22 + [( + 2)( + 1) +2 1 ] = 0
=1
( + 2)( + 1) +2 1 = 0,
= 1: 3 =
0
32
= 3: 5 =
2
=0
54
= 2: 4 =
By induction
3 =
1
,
[3 4][6 7] [3(3 + 1)]
3 +2 = 0,
General Solution
33
= 1,2,
1
43
0
,
[2 3][5 6] [(3 1)(3)]
3 +1 =
= 0,1,2
= 1,2,
= 1,2,
Giordon Stark
ACM95b Notes
= 0 1 +
=1
3
3+1
+ 1 +
= 0 1 () + 1 2 ()
(5)
=1
0 , 1 two constants to satisfy (0) = 0 , (0) = 0 . To see that 1 and 2 are linearly independent, check the
Wronskian at a single point:
1 (0) = 1,
1 (0) = 0,
2 (0) = 0
2 (0) = 1
So (1 , 2 )(0) = 1 0. Series solution (5) converges for < < since () and () are entire.
LECTURE 13
BEHAVIOR NEAR SINGULAR POINTS
Can sometimes hold greatest physical interest look in complex plane
(1)
(2)
+ () = 0
() = (0 )
()
34
Giordon Stark
ACM95b Notes
() = ,
=
0 < || <
(3)
Where () is analytic in || < with (0) 0. If solution can be written in form (3), then = 0 is a regular
singular point otherwise = 0 is an irregular singular point. What is the requirement on ()? = 0 is a regular
singular point () has a simple pole at = 0.
PROOF (FORWARD)
= + . (See
EXAMPLE
() =
1
() = 1 0
() =
1
1
()
=
() =
=0
() ()
+ 2 = 0
()
(4)
nd
At worst,
is a simple pole, 2 is a 2 order pole. = 0 is a regular singular point it is not ordinary and ()
and () are analytic at = 0. If = 0 is a regular singular point, (4) has at least one solution of form:
35
Giordon Stark
ACM95b Notes
() = (0 + 1 + 2 2 + ) = ,
=0
EULER EQUATIONS
2 + + = 0,
, constants
With = 0 a regular singular point. Attempt solution of form (5) with = 0 for 1, i.e.,
=
(6)
DISTINCT ROOTS (1 2 )
1 = 1 ,
2 = 2
( 1 , 2 ) = (2 1 ) 1 +2 1 0,
Linearly independent solution for < 0 or > 0 satisfy initial conditions in either interval.
REPEATED ROOTS (1 = 2 )
And also
But
|=1 = 0
[ ]|=1 = [ ]|=1 = 0
2 2 + 3 = 0
EXAMPLE
= 1 ||2 + 2 ||1 ,
36
< 0 or > 0
Giordon Stark
ACM95b Notes
LECTURE 14
THE METHOD OF FROBENIUS
Consider
2 + () + () = 0
(1)
() = ,
=0
() = ,
=0
|| < , > 0
And = 0 is a regular singular point. Note: if = = 0 for 1 this is just the Euler equation
2 + 0 + 0 = 0
(2)
(3)
= = (0 + 1 + )
=0
Where 0 0. Then
And
(4)
= ( + ) +1 = 1 [0 + ( + 1)1 + ]
(5)
=0
= ( + )( + 1) +2
(6)
=0
[( 1)0 + ] + (0 + 1 + ) [0 + ( + 1)1 + ] + (0 + 1 + ) = 0
(7)
() = ( 1) + 0 + 0 = 0
(8)
(9)
(8) is known as the indicial equation. Note that it is identical to that for corresponding Euler equation (3). Roots of
(8), 1 and 2 , determine qualitative behavior of solution near singularity. Setting coefficients of + to zero for
1 gives
( + ) + [( + ) + ] = 0
=0
37
Giordon Stark
ACM95b Notes
1 () = (1 )
=0
TROUBLE
a)
LECTURE 15
2 N D ORDER LINEAR BOUNDARY VALUE PROBLEM (BVP)
Consider
+ () + () = 0
(1)
With continuous () and () and general solution = 1 1 () + 2 2 (). Does the BVP with boundary
conditions () = , () = have a unique solution?
EXAMPLE
+ = 0,
( > 0)
= 0, = 1,
(1) = 0
2 = 0
1 = 0
Giordon Stark
ACM95b Notes
(0) = 1,
(1) = 0
2 = 1
1 sin 1 + cos 1 = 0
So =
sin
tan 1
(1) = 0
2 = 0
1 sin = 0
(0) = 1,
(1) = 0
2 = 1
1 sin 1 0
(0) = 1,
(1) = 1
2 = 1
1 sin 1 = 1
So there is no solution.
ii)
PROOF
The BVP consisting of (1) plus homogeneous boundary conditions () = , () = has a unique
solution if and only if the homogeneous boundary conditions are satisfied only by the trivial solution
= 0.
The nonhomogenous linear systems
1 1 () + 2 2 () =
1 1 () + 2 2 () =
Has a unique solution if and only if Det 1 ()2 () 1 ()2 () 0. If det 0, the homogeneous
system
39
1 1 () + 2 2 () = 0
Giordon Stark
ACM95b Notes
1 1 () + 2 2 () = 0
Has only the trivial solution (1 , 2 ) = (0,0). Alternatively, if det = 0, there is a nontrivial solution
(1 , 2 ) (0,0).
Whats going on? + = 0, (0) = 0, (1) = 0 is an eigenvalue problem. It has a nontrivial solution
(called an eigenfunction) if and only if is an eigenvalue.
EXAMPLE 1
+ = 0,
(0) = 0,
() = 0
= 1 sin + 2 cos
(0) = 0 2 = 0
() = 0 1 sin = 0 =
So we have eigenvalues
=
2 2
,
2
= 1,2,
= 1,2,
CASE 2: = 0
= 0 = 1 + 2
(0) = 0 1 = 0
() = 0 2 = 0
CASE 3: < 0
Consider = 0, = > 0
40
= 1,2,
Giordon Stark
ACM95b Notes
= 1 + 2 = 3 sinh + 4 cosh
(0) = 0 4 = 0,
() = 0 3 = 0
CASE 4: COMPLEX
We will later prove for a more general class of problems that there are no complex eigenvalues.
EXAMPLE 2
+ = 0,
(0) = 0,
() = 0
= 1 sin + 2 cos
(0) = 0 1 = 0
() = 0 2 sin = 0 =
So we have eigenvalues
=
2 2
,
2
= 1,2,
= 1,2,
CASE 2: = 0
= 1,2,
= 1 + 2
(0) = 0 2 = 0
() = 0 2 = 0
41
2 2
,
2
= 0,1,
Giordon Stark
ACM95b Notes
= cos
LECTURE 16
FOURIER SERIES
Represent a periodic function () with period 2 as an infinite series in terms of these eigenfunctions:
0
+ cos
+ sin
() =
2
(1)
=1
PERIODICITY
( + ) = (); periodic function with period . Smallest valid is the fundamental period.
sin
2
2
and cos
have fundamental period =
; all share period 2
ORTHOGONALITY PROPERTIES
Want to avoid redundancy in information: two real functions (), () are orthogonal on the interval [, ]
if their inner product vanishes, ie:
(, ) = ()() = 0
All sin and cos are orthogonal to each other for varying values of on = [, ], = 1,2,
sin
= , ,
sin
cos
1, =
Note that , =
(Kronecker-Delta).
0,
sin
cos
= 0
42
sin
= ,
Giordon Stark
ACM95b Notes
( )
( + )
1
sin
sin
= cos
cos
( )
( + )
sin
sin
1
=0
+
2
Similar methods can be used for the other two cases by using complex exponentials and rearranging. Now suppose
the Fourier Series (1) converges to the function () on the interval [, ]. How do you find the coefficients?
() cos
=
cos
+ cos
cos
+ sin
cos
=1
=1
The first and last integrals go to zero by orthgonality conditions, the middle integral becomes
+ cos
+ sin
() =
2
=1
=1
The last two integrals go to zero by periodicity, the first integral becomes 0 .
,
= () cos
= 0,1,2 ,
= () sin
,
= 1,2,
If the series converges to () on [, ] then it converges to the periodic extension of () outside the
interval.
EXAMPLE
() = 2 ,
[, ]
1
2
0 = 2 = 2 ,
43
1
2 sin = 0
4
1
= 2 cos = (1) 2 ,
= 1,2
Giordon Stark
ACM95b Notes
4
2
+ (1) 2 cos
() =
3
=1
Is the periodic extension of (). Note: it is also continuous. Also note coefficients decay by 2 . More smooth
A function is piecewise continuous on [, ] if the interval can be split into a finite number of subintervals in
which () is continuous with one-sided limits at the ends of each subinterval.
A function is piecewise smooth if it is piecewise continuous and () is continuous in each subinterval.
THEOREM
If () is piecewise smooth on [, ] then the Fourier series converges to the periodic extension of
+ ( )
2
A power series fails with discontinuities. Note; Fourier series converges even for a function containing
discontinuities (IE: it converges even if the function doesnt converge). Compare this to power series.
LECTURE 17
EXAMPLE
1,
() =
1,
On [, ].
< 0
> 0
1
1
0 = (1) + (+1) = 1 + 1 = 0,
44
1
1
= cos + cos = 0
Giordon Stark
ACM95b Notes
4
1
1
2
= sin + sin = sin = ,
0,
0
0
1
4
sin[(2 + 1)]
() =
2 + 1
odd
even
=0
Periodic extension of ()
,
= () cos
= 0,1,2 ,
,
= () cos
= () sin
,
= 1,2,
= 0,1,2
0
+ cos
2
=1
,
= () sin
sin
=1
= 1,2,
Now consider a function that is not necessarily even or odd on [, ] and suppose we are only interested in
the interval [0, ]. We have a choice of series representations.
45
Giordon Stark
ACM95b Notes
EXAMPLE
Fourier Cosine is smoother so faster convergence. Choose based on convenience (e.g., rate of convergence or
notational simplicity). How do eigenvalue problems and Fourier series arise in solving PDEs?
SOLVING PDES
EXAMPLE
Consider heat conduction along a perfectly insulated wire governed by the PDE
2
= 2 2 ,
0 < < ,
> 0
This is the heat equation, where is thermal diffusivity, with boundary conditions
(0, ) = 0,
46
(, ) = 0,
> 0
(1)
(2)
Giordon Stark
ACM95b Notes
(, 0) = (),
0 ,
= 0
(3)
In order for this to hold (left side depends on , right side depends on ) throughout the rectangle
= {, : (0, ), (0, )}
Both sides must be equal to a constant. Otherwise, equality would be violated by varying and holding
constant (or vice-versa). Hence
1
= 2 =
+ 2 = 0,
0 < <
> 0
LECTURE 18
47
Giordon Stark
ACM95b Notes
2
= 2 2
(1)
(0, ) = (, ) = 0,
(, 0) = (),
0 ,
Separation of variables
(2)
= 0 (initial condition)
+ 2 = 0,
> 0
(, ) = ()()
0
+ = 0,
(3)
With (0) = () = 0. This defines our eigenvalue problem. The eigenvalues are =
eigenfunctions () = sin
So
2 2
2
and corresponding
2 2 2
2 2 2
=
0
()
=
(, ) =
2 2 2
sin
= 1,2,3,
This satisfies PDE (1) and boundary conditions (2). To satisfy the initial condition (3), consider the infinite series
obtained by superposition of (valid because PDE is linear)
(, ) = (, )
(4)
=1
(, 0) = sin
=1
= ()
(5)
= () sin
Note: if working with a cosine series, would also have 0 . Check that the infinite series PDE (1) solves, differentiate
term-by-term:
2 2 2
2
=1
48
2 2 2
sin
,
2
2 2
2
2
=1
2 2 2
sin
Giordon Stark
ACM95b Notes
Hence the solution to the initial boundary value problem defined by (1), (2), (3) is given by (4) with coefficients (5);
assuming termwise differentiation of Fourier series is justified.
EXAMPLE
= ,
(0, ) = (, ) = 0,
() (, 0) = 1,
> 0
2
2
2
(1 cos ) = ,
cos =
= sin =
0
0,
0
(, ) =
=1,3,5,
4 2
sin
odd
even
Note: for > 0, continuous and partials are continuous based on physical concept of diffusive interaction (at
least in = {, : (0, ), (0, )}).
Separation of variables
(, ) =
=1
With coefficients
2 2 2
sin
(6)
= () sin
() sin
=1
(, )
=
() cos
=1
49
(7)
Giordon Stark
ACM95b Notes
=1
2 2
() sin
2
We differentiated a Sine Series when the physical boundary conditions justified differentiation. We differentiated a
Cosine Series when no boundary conditions were available or needed in order to justify differentiation.
By Theorem 4, termwise differentiation of (6) with respect to yields
equation IBVP.
2
2
Need to check conditions to make sure we can take . Obtain Fourier Cosine Series for by termwise
differentiation of (7) to get
(8)
=
() cos
=1
Because (, ) satisfies homogeneous boundary conditions, (0, ) = (, ) = 0 for > 0, which are precisely
the conditions required by Theorem 2 for differentiation of a Fourier Sine Series.
LECTURE 19
COMPLEX FORM OF FOURIER SERIES
Consider Fourier Series
0
+ cos
+ sin
() =
2
=1
With coefficients
(1a)
= () cos
Recall:
(1b)
= () sin
cos
1
= + ,
sin
1
=
1
0 1
+ ( ) + ( + )
() =
2 2
2
=1
=1
50
Giordon Stark
ACM95b Notes
1
( + )
2
=1
From (1)
() =
=
(2)
Hence
() cos
sin
=
2
1
()
=
2
USEFUL TERMINOLOGY
period
frequency =
angular frequency = 2 =
wavelength
wavenumber =
angular wavenumber = 2 =
Question of the Day: A man is at the beach with his dog. His dog starts running around from him at 2 . The man has a
whistle he blows to get the dog to come back. Each time the dog hears the whistle, he doubles his speed. Given that the
speed of sound is 300 , how many whistles does the dog hear?
The answer is 15. The dog will hear whistles 4,8,16,32,64,128,256,512. At this point, hes faster than the speed of sound,
so he starts to catch up to the old whistles: 4,8,16,32,64,128,256.
51
Giordon Stark
ACM95b Notes
with angular wavenumbers = , = , , . The spectrum is discrete with gap = between angular
FOURIER TRANSFORM
() =
Let =
1
=
()
2
and define () = 2 . So
() = ()
And
() =
=
Note: =
1
(),
2
1
=
()
2
() = ()
1
() =
()
2
The Fourier Transform () exists if () is absolutely integrable |()| < and piecewise continuous.
(e.g. if () 0 sufficiently fast as ).
52
Giordon Stark
ACM95b Notes
Note: if you cant read the cut-off portion of the bottom left, it says Fourier Transform.
LECTURE 20
Professor Note: Everything from this point on will appear in the final exam crucial.
() = ,
() =
53
> 0
Giordon Stark
ACM95b Notes
+ = +
2
4
() =
Let = +
Where
2
+
2
4
so = . So
2
4
() =
+ + + =0
1
() =
On 3 , 4 , =
: = =
2 2 +2
= =0
3
2 2
0 as . So
2
2
= 4
54
Giordon Stark
ACM95b Notes
(Interpreting the picture. As decreases: () increases spread while () decreases spread, and vice-versa.)
2
Small spread transforms to large spread and vice-versa. Also if () = then
() =
() = ()
() = () = 1
If we make a perfect spike, we get an equal amount of every component in the spectrum. All harmonics are equally
strong in a delta function!
1 {()()}
1
=
()()
2
1
1
=
() ( ) = ( )
()
2
2
55
Giordon Stark
ACM95b Notes
= ( )( )
Then
() = = = ()( )
DERIVATIVES
()}
()
()
+ ()
Thus { ()} = {()}. Similarly, if () has continuous derivatives through order that vanish as :
() () = () {()}
LINEARITY
2
= 2 2 ,
(, 0) = ()
< < ,
(, ) = 0
> 0
initial condition
boundary conditions
Professors Note: Sometimes the boundary conditions are left unstated in books, etc Fourier Transform the PDE in
space
=
(, ) = (, )
2
= ()2 (, ) = 2 (, )
2
(, ) = 2 2 (, )
56
Giordon Stark
ACM95b Notes
Solution is (, ) = (, 0)
2 2
where
(, 0) = {()} = ()
Hence (, ) = 1 (, 0}
1
2 2
2 2
4 2
(, ) = ()
4 2
2
4 2
( )2
4 2
Suppose we consider the diffusion of salt in a pipe from an initial point source () = ( 0 ) at = 0
(, ) =
4 2
( 0 )2
4 2
(, ) > 0 for > 0; diffusion at infinite speed. (Picture interpretation: start with something that looks like an
impulse at = 0 . As , the spread of the Gaussian becomes infinitely large.)
LECTURE 21
MORE GENERAL EIGENVALUE PROBLEMS
We have examined the eigenvalue problem + = 0, (0) = () = 0 with [] , i.e. [] + = 0,
which arises in the solution of the heat equation using separation of variables. The eigenvalues are =
nd
2 2
2
eigenfunctions = sin
for = 1,2, We now examine the general case of 2 order linear eigenvalue
57
with
Giordon Stark
ACM95b Notes
< <
(1)
(2)
() = 0
(3)
Where defines adjoint or dual operator and () is boundary terms from integration by parts. Adjoint
boundary conditions defined to make ()| = 0 for satisfying (2). Recall the inner product
(, ) ()()
(4)
()| = [( ) ( )|
(5)
() = 0,
(6)
Which vanish for any satisfying (2) if we choose the adjoint boundary conditions
() = 0
SELF-ADJOINTNESS
A problem is self-adjoint if = and the primal and adjoint boundary conditions are the same. This is equivalent
to requiring that (3) becomes
(, []) ([], ) = 0
(7)
For any () and () satisfying the same primal/adjoint boundary conditions. Comparing (1) and (4), this
requires 2 = , + = or simply
=
(8)
(9)
With () () and () (). This is the Sturm-Liouville Operator. In general, the second order linear
operator (1) can be transformed to the self-adjoint Sturm-Liouville form (9) using an integrating factor.
58
Giordon Stark
ACM95b Notes
(10)
< <
1 () + 2 () = 0
(11)
Where < < < . , , , are continuous with () > 0, () > 0 for . 1 , 2 , 1 , 2 are real
with
|1 | + |2 | > 0,
|1 | + |2 | > 0
i.e. they cant both be zero. For convenience, (10) may be written
[] + () = 0
(12)
()| = [()( )|
(13)
Where is given by (9). For this operator (9), we have already verified that = . Also, (8) simplifies the
boundary terms (5) to yield
Do these vanish for any () and () satisfying (11)? If 2 = 0 then () = () so () = 0. If 2 0, then
= . Hence (7) holds and the S-L problem (10),(11) is self-adjoint. Question to answer for next class: Why is this
important?
LECTURE 22
Note: This continues on from Lecture 21. From last time:
(, []) ([], ) = 0
[] + () = 0
(7)
(12)
ORTHOGONALITY OF EIGENFUNCTIONS
THEOREM
Eigenfunctions of a regular S-L problem (10),(11) corresponding to different eigenvalues are orthogonal
with respect to the weight function ().
() () () = 0
59
(14)
Giordon Stark
ACM95b Notes
PROOF
Letting = , = , (7) gives: ( , [ ]) ([ ], ) = 0. Using (12):
0 = ( , () ) ( () , ) = ( ) () () ()
, so (14) follows.
(15)
(, ) = ()()
Repeating the self-adjointness derivation using the definition shows that (7) remains valid.
THEOREM
The eigenvalues of a regular S-L problem (10),(11) are real.
PROOF
Consider possibly complex eigenvalue with eigenfunction (). By (7) and (12), using (15)
So
= 0
()()()
()|()|2 = 0
This is positive because () > 0 hence = 0 is real. Note also, the eigenfunctions can be
chosen to be real.
SIMPLICITY OF EIGENVALUES
THEOREM
Each eigenvalue of the regular S-L problem (10),(11) corresponds to one linearly independent
eigenfunction.
PROOF
For a given , consider two eigenfunctions () and () satisfying (10). The Wronskian is (, )() =
which, like () vanishes at = . So (), () are linearly dependent.
60
Giordon Stark
ACM95b Notes
SEQUENCE OF EIGENVALUES
The regular S-L problem (10),(11) has an infinite sequence of eigenvalues 1 < 2 < < < with as
. The eigenfunction has 1 zeros in < < (Proof: Birkhoff-Roots)
EXAMPLE
+ = 0,
1 = 1 = 1,
() = 1,
() = 0,
2 = 2 = 0,
2 2
2
() = 1
(0) = () = 0
, = sin
is orthogonal, = 1,2,
EIGENFUNCTION EXPANSIONS
These properties suggest the idea of generalizing Fourier Series to expansions of orthogonal eigenfunctions
() = () ,
=1
(16)
< <
=1
() ()()
()2 ()
ORTHONORMAL EIGENFUNCTIONS
()2 () = 1
() ()() =
61
(17a)
Giordon Stark
ACM95b Notes
(17b)
= () ()()
+ +( )
2
LECTURE 23
Note: this continues on from Lecture 22.
< <
(18)
Where is the Sturm-Liouville operator (9) and is a real constant. Furthermore, consider the homogeneous
separated boundary conditions (11).
Attempt solution by expanding in the eigenfunctions of the corresponding S-L eigenvalue problem (10) given by
(19)
[] + () = 0
With boundary conditions (11). Denote the eigenvalues 1 < 2 < (ordered) and orthonormal eigenfunctions
1 , 2 , Then let
(20)
() = ()
=1
=1
=1
[] = [ ()] = ()()
=1
=1
() () + () () = ()
62
(21)
(22)
Giordon Stark
ACM95b Notes
()()
,
()
() =
And expand
(recall () > 0)
(23)
()
= ()
()
=1
With
Or
()
()()
()
(25)
= () ()
() = [( ) ] () = 0
(26)
=1
Taking the inner product with (). Orthogonality (with respect to ()) gives
CASE 1
( ) = 0,
= 1,2,
If is not an eigenvalue of the S-L problem (19),(11) (i.e. for = 1,2, ) then
=
(27)
(28)
() =
=1
CASE 2
(29)
()
a)
If 0, then (27) has no solution and the nonhomogeneous problem (18),(11) has no solution
b) If = 0, then (27) shows that is arbitrary. Hence the nonhomogeneous problem (18),(11) has
infinitely many solutions
() = () +
63
=1
()
(30)
Giordon Stark
ACM95b Notes
Containing an arbitrary multiple of the eigenfunction (). From (25) we see that = 0 when
() () = 0
(31)
LECTURE 24
Note: this continues from the past two or so lectures.. yes, I realize its a long thing on S-L problems and solving
BVPs
() =
=1
()
(32)
Where are the eigenvalues of (19),(11) and are the corresponding orthonormal eigenfunctions. By (25)
= () ()
(33)
() = (, )()
(, ) =
64
=1
() ()
(34)
Giordon Stark
ACM95b Notes
Note: the Greens function does not exist if = for some = 1,2, What are the properties of the Greens
function? Consider letting () = ( ) so by (33):
() = (, )( ) = (, )
So (, ) is the response of the solution at to an impulse at . Hence, by (18) the Greens function satisfies
[(, )] + ()(, ) = ( ),
< <
(, ) = 0
(, ) = 0
(35)
(36a)
(36b)
RECIPROCITY
(, ) = (, )
The response at due to an impulse at is identical to the response at due to an impulse at . Physically, this is a
remarkable observation! Note: Reciprocity did not hold for the IVP Greens function. What is the behavior of
(, ) at = ? (35) and (9) give
() (, ) + [() + ()](, ) = ( )
2 =
to balance ( ). So
() (, )
+ [[() + ()](, ) = 1
And also
1
() (, ) =
()
(, )| = 0
65
(38)
(39)
Giordon Stark
ACM95b Notes
Find and for S-L eigenvalue problem (19),(11) and use (34) or equivalently solve (35),(36) using
eigenfunction expansion.
METHOD 2
Alternatively, solve (35),(36) by solving homogeneous problems in < and > and
matching solutions at = using (38) and (39).
<
(, ) = 2 ()2 (),
<
Likewise,
2 ()2 () 1 ()1 () = 0
2 ()2 () 1 ()1 () =
So
1 () =
2 ()
,
()()
1
()
2 () =
1 ()
()()
1 ()2 () ,
()()
2 ()1 ()
()() ,
LECTURE 25
EXAMPLE
66
Giordon Stark
ACM95b Notes
2
= (),
2
[]
(0) = 0,
() = 0
()
2
= ()
2
Solve
2
(, ) = ( )
2
(1a)
[] + = 0,
(1b)
Eigenvalues are =
2 2
2
And
(0) = () = 0
(, ) = () ()
. Let
(2)
=1
So
(3)
( ) = () ()
=1
sin
( ) = () sin
sin
=1
. Now
67
=1
=1
Giordon Stark
ACM95b Notes
=1
=1
2
2 2
2 sin
= sin
sin
Note: this came up during class, and I thought it was a good question. We cant just ignore the summations and
equate coefficients, we need to justify picking off individual terms which is why we talk about orthogonality and
inner products. Taking the inner product with respect to (), by orthogonality we have
2 2
sor =
2 2
sin
2 2
2
= sin
(4)
2
2
(, ) = 2 2 sin
sin
=1
Professor Pierce made a point of saying that we cant study the behavior of the solution from an infinite series of
sine terms. Also, to get a particular solution, we take the inner product with ().
METHOD 2:
(5)
(6a)
(6b)
(, ) = 0
(7)
1
=
=1
()
+
(8)
| = 0
For < : (, ) = 1 () satisfies (5) and (6a). For < : (, ) = 2 ()( ) satisfies (5) and
(6b). To satisfy (7) and (8), we must have both 2 ()( ) 1 () = 0 and 2 () 1 () = 1 This gives us
1 () =
and 2 () = , so
NOTES
a)
( )
,
(, ) =
( )
,
68
(9)
Giordon Stark
ACM95b Notes
b) (4) is the Fourier Sine series for (9) (i.e. the eigenfunction expansion).
LECTURE 26
SINGULAR STURM-LIOUVILLE EIGENVALUE PROBLEMS
Again consider
[] + () = 0,
< <
[] = (() ) + ()
(1)
(2)
A singular S-L problem arises if the regular assumptions on , , are valid for < < , but one or more of
the following apply:
0, 0, unbounded (at or )
Whatever the singularity, appropriate boundary conditions are chosen to ensure that the problem remains selfadjoint. Recall that integration by parts for the S-L operator yields
(, []) ([], ) = ()|
()| = [()( )|
(3)
(4)
Use separated boundary conditions such that any () and () satisfying the boundary conditions give () = 0
and () = 0. For example, if () 0 as , then need to be bounded. Then appropriate
boundary conditions are (), () bounded as . 1 () + 2 () = 0.
69
Giordon Stark
ACM95b Notes
Singular S-L problems remain self-adjoint with (, []) ([], ) = 0 for any () and () satisfying the
same primal/adjoint boundary conditions.
b) Therefore, by similar arguments to the regular case, square integrable eigenfunctions corresponding to
different eigenvalues, IE:
2 ()() <
Suppose at least one of the boundary conditions is of the regular S-L type (say at = 0). Consider
eigenvalue and eigenfunctions and . Since () 0 and = (). (, ) = 0, and are not
linearly independent solutions (, linearly dependent).
= 2
(3)
(, ) = 0
(4)
(, 0) = ()
(5)
Boundary conditions
Initial conditions
1 1
= +
(6)
+ = 0
(7)
() = 0
(8)
, bounded as 0
(9)
70
Giordon Stark
ACM95b Notes
+ + 2 = 0
0
0
2
+ 0
0
0 2
>0
the Method of Frobenius, the general solution is () = 1 0 () + 2 0 () where 0 (), 0 () are the Bessel
functions of first and second kind, both of order 0. Other examples of special functions defined only as solutions
to differential equations. We note that 0 (), 0 () bounded as 0 and |0 ()| as 0. Also 0 () has
infinitely many zeros for > 0 0 fails the conditions so set 2 = 0. The infinite sequence of eigenvalues is
defined by 0 = 0, = 1,2, with eigenfunctions () = 0 that are orthogonal with respect to
2
weight function . By (6), () = . By superposition,
(, ) = 0
=1
() = 0
=1
0 ()0
0 02
LECTURE 27
ACM FINAL REVIEW
Topics:
-
IVP: Euler equations, series solutions around regular and ordinary points, Frobenius method
BVP: Linear 2 order BVPs, eigenvalue problems, separation of variables for PDEs, adjoint operator and selfadjoint linear operators, S-L eigenvalue problems, Greens function for BVPs
nd
71
Giordon Stark
ACM95b Notes
= 0,
( + 1)2
0 < < ,
(0) = () = 0
= ( + )( + 1)(1 + )+ 2
( + )( + 1)(1 + )+ 2 +
(1 + )+ = 0
(1 + )2
( + )( + 1) + = 2 2 + 2 + + = 0
2 2 + = 0 =
() = (1 + )+ = (1 + )2 (1 + )
1
2
1
1
1
1 = (1 + )2 cos log(1 + ) + sin log(1 + )
4
4
1
1
1
2 = (1 + )2 cos log(1 + ) sin log(1 + )
4
4
= 1 1 + 2 2
1
1
1
1
= (1 + 2 )(1 + )2 cos log(1 + ) + 2 (1 + )2 sin log(1 + )
4
4
Let 1 = 1 + 2 so
(0) = 0 1 = 0,
=
2 2
1
+ ,
(log(1 + ))2 4
1
() = 0 log(1 + ) =
4
1
= (1 + )2 sin
log(1 + )
log(1 + )
72
Giordon Stark
ACM95b Notes
= 0
1
3
(0) = cos
3
lim () = 0,
[()] = ()
Taking transform:
2 = 0
Applying inversion:
=
() =
() =
()
3
2
= = 2 () = (0) 3
+
() =
3 =
3
2
2
3
3
cos + + sin +
3
3
2
By odd/even orthogonality
3
2 cos +
2
3
0
= 2 +
With on real line < < and our initial condition (, 0) = (). Take fourier transform
2
(, ) = ()2 + (, ) (, ) = () ( ) +
(, ) = ()
( + )2
(4 2 )
Consider
a)
2 + + = ,
(1) = (2) = 0
73
Giordon Stark
ACM95b Notes
() = is integrating factor, so
1
[ ] + = 0
c)
= 0 < 1
= 1
2
,
log 2
log
,
log 2
= sin
lim () <
2 = log
1
log() ,
= (, ) =
log() ,
74
0 <
< 1
= 1,2,