Math 39100 K (38429)

- Lectures 01

Ethan Akin
Office: NAC 6/287
Phone: 650-5136
Email: [email protected]

Spring, 2019
Contents

First Order Differential Equations, B & D Chapter 2

Linear Growth
Exponential Growth
Existence and Uniqueness Theorem
Separable and Homogeneous Equations
Linear Equations
Exact Equations
Modeling with First Order Equations
Reduction of Order

Second Order Linear Equations, B & D Chapter 4

Second Order Linear Homogeneous Equations with
Constant Coefficients
Introduction
I Officially, the book is Boyce,DiPrima and Meade
Elementary Differential Equations and Boundary Value
Problems (11th Ed). However, it only exists as an eBook
or a looseleaf. I recommend that you get instead either
the 9th or 10th edition of Boyce and DiPrima Elementary
Differential Equations and Boundary Value Problems
I Keep up with the homework.
I Ask questions.
I The course information sheet with the term’s homework
assignments is posted on my site:
http : //math.sci.ccny .cuny .edu/peoplename = EthanAkin
I I will be posting there a pdf of the slides I am using here.
The first third of the course already up. I will post the
others as we get to them.
I Office: NAC 6/287. Hours: Tuesday, Thursday 10-11:00
am.
First Order Differential Equations - Introduction

Solving a differential equation means using given information

about the derivative to find the original function.

Recall the three notations for the derivative of a function f

with the equation y = f (x) indicating that x is the input or
independent variable and y is the output or dependent variable.
We write the derivative as y 0 = f 0 (x) or in Leibniz’s notation
dy
dx
.

A general first order ode (= ordinary differential equation) is

of the form y 0 = F (x, y ) or dy
dt
= F (t, y ).
The Integral as Antiderivative

y 0 = f (x) implies y =
R
f (x)dx + C .
Linear Growth
dy
dt
= m (Absolute Growth Rate y 0 is a constant).
dy = mdt, and so
Z
y = mdt = mt + C .

When t = 0, y0 = C .
dy
So = m implies y = mt + y0 .
dt
Exponential Growth
dy 0

dt
= ky (Relative Growth Rate yy is a constant).
dy =R kydt, and so
y = kydt = ky 2 /2 + C ??? NO
Z Z
dy
= kdt
y
.

ln |y | = kt + C ,

y = Ce kt different C.
When t = 0, y0 = C .
dy
So = ky implies y = y0 e kt .
dt
Let’s do the last bit slowly. Recall e A+B = e A e B .
ln |y | = kt + C implies |y | = e C · e kt .
Recall that e x is always positive and |y0 | = e C .

So if y0 > 0 then y0 = e C and y = y0 e kt .

If y0 < 0 then −y0 = |y0 | = e C and so

y = −|y | = −e C e kt = y0 e kt .
If y0 = 0 then y = 0 = y0 e kt is the solution. Our previous

sloppy procedure works fine, but it is worth looking at a

diagram.
The Fundamental Existence and Uniqueness
Theorem, Section 2.4
An initial value problem (= an IVP) consists of a differential
equation together with initial conditions.
In the first order case, this is a pair:

[ode]y 0 = F (x, y ), [ic]y (x0 ) = y0 .

Solving the ode involves an integration which yields a single

arbitrary constant. The solution with the arbitrary constant is
the general solution. That is, there are infinitely many
solutions obtained by making different choices of the
constants. However, the constant is determined by the initial
value. Theorem 2.4.2 in Section 2.4 says, as far as we are
concerned:
Theorem An IVP has one and only one solution. That is, the
solution exists and is unique.
Actually, some conditions are needed to make this work. It
suffices that F (x, y ) have continuous partial derivatives.
Example 3 of Section 2.4 illustrates what might otherwise
happen.
dy
IVP : = y 1/3 , y (0) = 0.
dt
Z
3
y −1/3 dy = y 2/3 = t − C .
2
Notice that t − C has to be positive and so t ≥ C (which is
why I used −C instead of C ). We obtain for any choice of C
two solutions:
(
0 t ≤ C,
y= 2
±[ 3 (t − C )]3/2 .
Any of these with C ≥ 0 satisfy the initial condition y (0) = 0.
Variables Separable, Section 2.2
dy
If dx
= A(x)B(y ), we can solve by separating the variables:
dy
= A(x)dx
B(y )
and integrating. We combine the constants of integration and
usually put the single constant on the x side. If we have initial
conditions, we determine the constant.
Section 2.2/13: y 0 = R2x/(y + Rx 2 y ), y (0) = −2.
Factor and separate: ydy = (2x/(1 + x 2 )dx.
y2
2
= ln(1 + x 2 ) + C and y (0) = −2 implies C = 2.
On a test stop here, But taking the square root, we get:
p
y = ± 4 + 2 ln(1 + x 2 )
and the initial condition requires the − sign for the unique
solution.
Homogeneous Equations, Section 2.2/34
A function F (x, y ) is called homogeneous of degree n if
F (λx, λy ) = λn F (x, y ). We will only be considering functions
homogeneous of degree 0 so that F (λx, λy ) = F (x, y ) and so
with λ = 1/x:
F (x, y ) = F (1, y /x).
So we use the change-of-variable:
y dy dz
z= , xz = y , and so =x + z.
x dx dx
So when F is homogeneous of degree zero the equation
dy
dx
= F (x, y ) becomes
dz dz
x + z = F (1, z), or x = F (1, z) − z,
dx dx
which is variables separable. Usually the integration uses
partial fractions.
 dy
Section 2.2/BD34; BDM29 : dx
= − 4x+3y
2x+y
.

dz 4 + 3z z 2 + 5z + 4
x = −z − =− .
dx 2+z 2+z
2+z dx
dz = − .
(z + 4)(z + 1) x
2+z A B
Partial fractions: (z+4)(z+1)
= (z+4)
+ (z+1)
.
2 + z = A(z + 1) + B(z + 4) and so (substituting
z = −1, −4) B = 1/3, A = 2/3.
Integrate to get 32 ln(z + 4) + 13 ln(z + 1) = − ln(x) + C .
Multiply by 3 and exponentiate:

(z + 4)2 (z + 1)1 = Cx −3 , or (y + 4x)2 (y + x) = C .

Exercises BD 5, 35, 37; BDM 3, 307
2.2/BD5;
R BDM3R : y 0 = (cos2 (x))(cos2 (2y )). So
sec (2y )dy = cos2 (x)dx.
2

We use that sec2 is the derivative of tan and that

cos(2x) = cos2 (x) − sin2 (x) = 2 cos2 (x) − 1.
1 x sin(2x)
tan(2y ) = + + C.
2 2 4
dy
2.2/35 : dx
= x+3y
x−y
. and so
dz 1+3z 1+2z+z 2
x dx = −z + 1−z
= 1−z
.
1−z A B
(z+1)2
= (z+1) + (z+1)2 . Set z = −1 to get B = 2 and z = 0 to

get A + B = 1 and so A = −1.

Integrate to get − ln(z + 1) − 2(z + 1)−1 = ln(x) + C .
2x
− ln(x + y ) − = C.
y +x
 dy x 2 −3y 2
2.2/BD37;BDM30 : dx
= 2xy
. and so
dz 1−3z 2 1−5z 2
x dx = −z + 2z
dz = 2z
.
Z Z
dx 2z 1
ln(x) + C = = 2
= − ln(1 − 5z 2 ).
x 1 − 5z 5
So
(x 2 − 5y 2 )x 3 = C .
Linear Equations, Section 2.1
A first order equation is linear if it can be written in the form
dy
dt
+ p(t)y = g (t). If g (t) = 0 then the equation is separable
with dy
y
= −p(t)dt. Notice that if p(t) is a constant then this
is exponential
R growth. Integrating and exponentiating we get
y = Ce − p(t)dt . When g (t) is not zero, the equation is not
separable and we use the integrating factor method.
I Set the equation in the form dydt
+ p(t)y = g (t), dividing
(both sides!) by the coefficient of dy
dt
if it is not 1.
R
I Compute the integrating factor µ(t) = e p(t)dt and
multiply through (both sides!).
I Check that the left side is [µ(t)y ]0 .
I Integrate and divide by µ to solve for y :
Z
y = [ µ(t)g (t)dt + C ]/µ(t).
It is sometimes not obvious that an equation is linear.
Example: x dy
dx
+ xy = x 3 + 2y .
Collect the y 0 terms and the y terms and divide to get a
coefficient of 1 on y 0 . x dy
dx
+ (x − 2)y = x 3 and so:

dy 2
+ (1 − )y = x 2 .
dx x
Z
2
µ = exp[ (1 − )dx] = e −2 ln(x)+x = x −2 e x ,
x

[x −2 e x y ]0 = e x .

x −2 e x y = e x + C , or y = x 2 + Cx 2 e −x .
Exercises 2.1/BD 2, 8, 19; BDM 2,

2.1/BD2;BDM2
R : y 0 − 2y = t 2 e 2t . Multiply by
µ = e −2dt = e −2t .

[e −2t y ]0 = e −2t y 0 − 2e −2t y = t 2 .

1 1
e −2t y = t 3 + C , and so y = ( t 3 + C )e 2t .
3 3
2.1/BD8 : (1 + t 2 )y 0 + 4ty = (1 + t 2 )−2 . Divide by (1 + t 2 ) to
get
y 0 + 1+t
4t 2 −3
2 y = (1 + t ) , with
µ = exp(2 ln(1 + t )) = (1 + t 2 )2 .
2

[(1 + t 2 )2 y ]0 = (1 + t 2 )−1 , with (1 + t 2 )2 y = arctan(t) + C .

y = (arctan(t) + C )(1 + t 2 )−2 .

2.1/BD19 : t 3 y 0 + 4t 2 y = e −t , y (−1) = 0, t < 0. Divide by
t 3 to get
4t −1 dt
R
y 0 + 4t −1 y = t −3 e −t , with µ = e = e 4 ln(t) = t 4 .

[t 4 y ]0 = te −t .
Integrate by parts to get t 4 y = −te −t − e −t + C .
When t = −1, y = 0, and so C = 0.
Hence,
y = −(t −3 + t −4 )e −t .
Exact Equations, Section 2.6 through Example 3
We begin with a review of the chain rule in several variables.If
F is a function of x and y and each of these is a function of t,
then for the composed function given by t 7→ F (x(t), y (t)),
we have
dF ∂F dx ∂F dy
= · + · .
dt ∂x dt ∂y dt
∂F ∂F
This is the dot product of the gradient of F ( =< ,
∂x ∂y
>)
with the velocity vector ( =< dx , dy >).
dt dt

About the gradient vector field, there are two things to

remember.
I The gradient of F points in the direction of greatest
increase of F .
I The gradient of F is perpendicular to the contour curves
defined by F (x, y ) = C .
The contour curve F (x, y ) = C defines y implicitly as a
function of x, with:
∂F ∂F dy
+ · = 0.
∂x ∂y dx
This is a differential equation whose solution is the family of
contour curves F = C . Multiplying by dx we see that it is of
the form
M(x, y )dx + N(x, y )dy = 0.
The total differential of F is dF = ∂F
∂x
dx + ∂F
∂y
dy . Think of it
as the effect on F of a little change in x and y . Given a
differential M(x, y )dx + N(x, y )dy we would like to find an F
so that it equals dF .
Such an F might not exist. When it does, the differential is
called exact.
Think of the vectorfield < M(x, y ), N(x, y ) > as a force field.
If x and y move along a path as a function of time, the
R t=b
integral W = t=a M(x(t), y (t)) dx dt
+ N(x(t), y (t)) dy
dt
dt is
called the line integral and its value is the work along the
path between the times a and b. The result usually depends
upon the choice of path.
But if the force is the gradient of a function F , then it is called
a conservative force with potential function F . In that case
Z t=b
dF
W = dt = F (x(b), y (b)) − F (x(a), y (a)).
t=a dt
The work is the potential difference between the two locations.
Happily there is a simple test for exactness. If M = ∂F
∂x
and
∂F
N = ∂y , then

∂M ∂ 2F ∂ 2F ∂N
= = = .
∂y ∂y ∂x ∂x∂y ∂x
Consider the diagram

F
∂x . & ∂y
M dx + N dy
∂y & . ∂x
=?

Example: (2xy 2 + 2y )dx + (2x 2 y + 2x + 2y )dy = 0.

∂(2xy 2 +2y ) 2 +2x+2y )

∂y
= 4xy + 2 = ∂(2x y ∂x .
R
F = 2xy 2 + 2y dx = x 2 y 2 + 2xy + H(y ).
∂F
∂y
= 2x 2 y + 2x + H 0 (y ) = 2x 2 y + 2x + 2y .
H 0 (y ) = 2y and so H(y ) = y 2 . Thus, F = x 2 y 2 + 2xy + y 2 .
Solution : x 2 y 2 + 2xy + y 2 = C .
Example :
(e xy (y cos(x) − sin(x)) + 1)dx + (e xy x cos(x) + cos(y ) + 2y )dy
= 0, y (0) = 0.
Exact. R Go up the y side :
F = (e xy x cos(x) + cos(y ) + 2y ) dy =
e xy cos(x) + sin(y ) + y 2 + H(x).
∂F
∂x
= e xy (y cos(x)−sin(x))+H 0 (x) = e xy (y cos(x)−sin(x))+1.
H 0 (x) = 1, and so F = e xy cos(x) + sin(y ) + y 2 + x. The
general solution is F = C .
Since y = 0 when x = 0, C = 1.
Example : (2x + y )dx = (x + 2y )dy . Rewrite as
(2x + y )dx − (x + 2y )dy = 0. Not exact. (It is
homogeneous). Let’s try anyway.
R
F = (2x + y ) dx = x 2 + xy + H(y ).
∂F
∂y
= x + H 0 (y ) = −x − 2y and so H 0 (y ) = −2x − 2y .
ERROR.
Example: (y ln(x) + xy )dx + (x ln(y ) + xy )dy = 0. Not exact
(It is variables separable).
R ln(x)+x
dx = − ln(yy)+y dy .
R
x

Split each integral: ln(x) dx = 21 (ln(x))2 + C

R
x
[u substitution with u = ln(x)].
2.6/BD17;BDM13
Rx : With Ry
ψ(x, y ) = x0 M(s, y0 ) ds + y0 N(x, t) dt,
the first term doesn’t depend on y . So
∂ψ
= N(x, y ).
∂y
Z y
∂ψ ∂N
= M(x, y0 ) + (x, t) dt.
∂x y0 ∂x
∂N
By assumption ∂x
(x, t) = ∂M
∂t
(x, t)
R y ∂M
and so the integral is y0 ∂t (x, t) dt = M(x, y ) − M(x, y0 ).

∂ψ
= M(x, y0 ) + M(x, y ) − M(x, y0 ) = M(x, y ).
∂x
Tank and Interest Problems, Section 2.3
Example: A 200 gallon tank initially contains 50 gallons of
water in which is dissolved 5 pounds of salt. At a rate of 5
gallons per minute a solution with a concentration of 1/2
pound of salt per gallon is poured into the tank. A well-stirred
solution is pumped out at a rate of 2 gallons per minute.
Compute the equation for the quantity S(t) of the salt in the
tank up to the time t ∗ the tank is filled. Then compute the
formula S(t) for the time t beyond t ∗ as the tank continues to
overflow.
Note the units. The volume V (t) in gallons and the quantity
S(t) in pounds. The net changes dV
dt
and dS
dt
are in gallons per
minute and in pounds per minute, respectively.
dV
dt
= Input − Output = 5 − 2 = 3. (Linear Growth)
So V = V0 + 3t = 50 + 3t, and the tank is filled when
V (t ∗ ) = 200 and so with t ∗ = 50 minutes.
dS
dt
= 5[gal/min] · 12 [lb/gal] − 2[gal/min]· S
50+3t
[lb/gal].
This is the linear equation
dS 2 5
+ S= .
dt 50 + 3t 2
2
= exp( 23 ln(50 + 3t)) = (50 + 3t)2/3 .
R
µ = exp( 50+3t
dt)
[(50 + 3t)2/3 S]0 = 52 (50 + 3t)2/3 .
(50 + 3t)2/3 S = 12 (50 + 3t)5/3 + C
When t = 0, S = 5. So C = 502/3 5 − 21 505/3 = −20(50)2/3 .
1 50
S = (50 + 3t) − 20( )2/3 .
2 50 + 3t
with S(50) = 20(5 − 4−2/3 ).
When the tank is overflowing the outflow is the same as the
inflow and the volume V is constant at 200.
dS
dt
= 5 · 12 − 5 · 200
S
with t ≥ t ∗ .
This equation is linear and variables separable
dS dt t
S−100
= − 40 . and so ln(S − 100) = − 40 + C.

S = 100 + Ce −t/40
with S(50) = 20(5 − 4−2/3 ) determining C .

C = −(100 − S(50))e 5/4 = −20 · 4−2/3 e 5/4 .

If we restarted the clock so that S = 20(5 − 4−2/3 ) at t = 0
then C = −20 · 4−2/3 .
Interest

For annual interest with rate r the interest on P dollars left for
one year is rP. Thus, P1 = P0 + rP0 .
For the next year the interest is rP1 and so
P2 = P1 + rP1 = P0 + rP0 + rP0 + r 2 P0 .
But notice that P + rP = (1 + r )P. So each year we multiply
by (1 + r ).
Beginning with P0 we have Pt = (1 + r )t P0 .
The units of the interest rate r are dollars of interest per dollar
of principal per year. So in a fraction 1/n of a year, the
interest on P is (r /n)P.
So compound interest with n periods per year yields Pt , after t
years, given by
r 1
Pt = (1 + )nt = ((1 + h) h )rt
n
r rt
where h = n
so that nt = n .
For continuous compounding we take the limit as n tends to
infinity or, equivalently, as h tends to zero. To recall what
happens to the limit, write
1
ln((1 + h) h ) = ln(1+h)−ln(1)
h
whose limit is the derivative of
ln(x) at x = 1 and so is 1.
1
Hence, limh→0 (1 + h) h = e 1 = e, and with continuous
compounding

Pt = P0 e rt .
The equation Pt = P0 e rt is exactly exponential growth with
rate r . This is also derived as follows
The change in the principal due to interest is dP = rPdt,
which is exponential growth with rate r .
The problems also feature a constant flow of k dollars per year
with k > 0, eg for money put into a bank account, or with
k < 0 for money paying off a loan. So the combined effect is

dP = (rP + k)dt.
2.3/ BD9; BDM7 : P0 = 8000 dollars, r = .1 per year and
P3 = 0.
dP dP
dt
= .1P − k, and so P−10k
= .1dt.
P = 10k + Ce .1t , with 8000 = 10k + C , 0 = 10k + Ce .3 .
So C = −8000/(e .3 − 1) and k = 800(e .3 /(e .3 − 1). So
3k − 8000 = 1257 dollars of interest.
Miscellaneous Problems,
Section 2.9/ BD 1-6, 8-11, 15, 29; BDM 1-10, 12,
22

1. Linear, 2. Separable, 3. Exact,

4. Linear (factor y), 5. Exact, 6. Linear,
8. Linear, 9. Exact,10. Separable (factor),
11. Exact, 15. Linear, 29. Homogeneous.
Reduction of Order Problems

In Calculus 201 falling body problems were introduced. With y

the height, the derivative dy
dt
is the velocity v and the second
d 2y dv
derivative dt 2 = dt is the acceleration a. Near the earth’s
surface, the acceleration due to gravity is a constant labeled
−g (Why − ?).
The motion due to gravity alone is given by the simple Second
2
Order Differential Equation ddt y2 = −g . The solution will
require two integrations yielding two arbitrary constants.
We convert the second order equation in y to a first order
2
equation in v , writing ddt y2 = dv
dt
= −g and so v = −gt + C
with C the initial velocity v0 . That is,
dy
dt
= v = v0 − gt. Integrating again, and observing that the
second constant of integration is the initial height, y0 , we get
 1
v = v0 − gt and y = y0 + v0 t − gt 2 .
2
In using these formulae, we observe that ground is when y = 0
and the maximum height occurs when v = 0.
There is an alternative procedure which uses the chain rule
d 2y dv dy dv dv
2
= = =v .
dt dt dt dy dy
d 2y dv
The falling body equation becomes dt 2
= v dy = −g .
Separating variables and integrating we obtain
1 2
2
v = −gy + C . Labeling the constant E we get:
1 2
v + gy = E .
2
This is Conservation of Energy.
Reduction of Order Problems,
Section 2.9/ BD 41, 49. 48; BDM 36

2.9/BD41 : t 2 y 00 = (y 0 )2 , t > 0 and so t 2 dv

dt
= v 2.
dv
v2
= dt
t2
and so v −1 = t −1 + C1
dy t
=v =
dt C1 t + 1
u−1
Z Z
t 1
y= dt = 2 du.
C1 t + 1 C1 u
1 1
and so y = C1
t − C12
ln(C1 t + 1) + C2 if C1 6= 0.

Otherwise, y = 12 t 2 + C2 .
2.9/BD49 : y 00 − 3y 2 = 0, y (0) = 2, y 0 (0) = 4.

dv
− 3y 2 = 0.
dt
WON’T WORK.
dv 1 2
v = 3y 2 , and so v = y 3 + C1 .
dy 2
√
With y = 2, v = 4, C1 = 0 and dy
dt
= v = 2y 3/2 .
√ √
−2y −1/2 = 2t + C2 . With t = 0, y = 2 and so C2 = − 2.
2
y= .
(t − 1)2
R √
6 0, t = √ 3 2 dy .
Notice that for C1 =
y +C1
2.9/BD48, BDM36 : y 0 y 00 = 2, y (0) = 1, y 0 (0) = 2.
√
Z Z
dy
vdv = 2dt, and so = v = 2 t + 1.
dt
So y = 34 (t + 1)3/2 − 31 .
Alternatively, v 2 dv
dy
= 2.

Z
dy
= v = (6y + 2)1/3 and so (6y + 2)−1/3 dy = t + C2 .
dt

So 41 (6y + 2)2/3 = t + 1.
Second Order Differential Equations

A general second order ode (= ordinary differential equation)

2
is of the form y 00 = F (x, y , y 0 ) or ddt y2 = F (t, y , dy
dt
).
For second order equations, we require two integrations to get
back to the original function. So the general solution has two
separate, arbitrary constants. An IVP in this case is of the
form:

y 00 = F (t, y , y 0 ), with y (t0 ) = y0 , y 0 (t0 ) = y00 .

Interpreting y 0 as the velocity and y 00 as the acceleration, we

thus determine the constants by specifying the initial position
and the initial velocity with t0 regarded as the initial time
(usually = 0).
The Fundamental Existence and Uniqueness Theorem says
that an IVP has a unique solution.
The General Solution

The general solution of the equation y 00 = F (t, y , y 0 ) is of the

form y (t, C1 , C2 ) with two arbitrary constants. Supposing that
these are solutions for every choice of C1 , C2 , how do we now
that we have all the solutions? The answer is that we have all
solutions if we can solve every IVP. To be precise:

THEOREM: If y 00 = F (t, y , y 0 ) is an equation to which the

Fundamental Theorem applies and y (t, C1 , C2 ) is a solution for
every C1 , C2 , then every solution is one of these provided that
for every pair of real numbers A, B, we can find C1 , C2 so that
y (t0 , C1 , C2 ) = A and y 0 (t0 , C1 , C2 ) = B.

PROOF: If z(t) is any solution then let A = z(t0 ) and

B = z 0 (t0 ), and choose C1 , C2 so that y (t, C1 , C2 ) solves this
IVP. Then y (t, C1 , C2 ) and z(t) solve the same IVP. This
means that z(t) = y (t, C1 , C2 ).
Linear Equations and Linear Operators
We will be studying exclusively second order, linear equations.
These are of the form

d 2y dy
2
+ p(t) + q(t)y = r (t), or y 00 + py 0 + qy = r .
dt dt
Just as in the first order linear case, if we see
Ay 00 + By 0 + Cy = D with A, B, C , D functions of t we can
obtain the above form by dividing by A. When A, B and C are
constants we often won’t bother.
The equation is called homogeneous when r = 0. It has
constant coefficients when p and q are constant functions of t.

These are studied by using linear operators. An operator is a

function L whose input and output are functions. For
example, L(y ) = yy 0 . If y = e x then L(y ) = e 2x and
L(sin)(x) = sin(x) cos(x).
An operator L is a linear operator when it satisfies linearity or
The Principle of Superposition.

L(Cy1 + y2 ) = C L(y1 ) + L(y2 ).

Given functions p, q we define L(y ) = y 00 + py 0 + qy . Observe
that:

C × (y100 + py10 + qy1 )

+ y200 + py20 + qy2
(Cy1 + y2 )00 + p(Cy1 + y2 )0 + q(Cy1 + y2 )

So the homogeneous equation y 00 + py 0 + qy = 0 can be

written as L(y ) = 0.
General Solution of the Homogeneous Equation, Section
3.2
If y1 and y2 are solutions of the homogeneous equation
L(y ) = 0 then by linearity C1 y1 + C2 y2 are solutions for every
choice of constants C1 and C2 , because
L(C1 y1 + C2 y2 ) = C1 L(y1 ) + C2 L(y2 ) = C1 0 + C2 0 = 0.
Notice there are two arbitrary constants. This is the general
solution provided we can solve every IVP. So given numbers
A, B we want to find C1 , C2 so that

C1 y1 (t0 ) + C2 y2 (t0 ) = A,
C1 y10 (t0 ) + C2 y20 (t0 ) = B.
Cramer’s Rule says that this has a solution provided the
coefficient determinant
 
y1 (t0 ) y2 (t0 )
y1 (t0 ) y20 (t0 ) is nonzero.
 0 
The Wronskian
Given two functions y1 , y2 the Wronskian is defined to be the
function:
 
y1 y2 
W (y1 , y2 ) =  0 0  = y1 y20 − y2 y10 .

y1 y2
Two functions are called linearly dependent when there are
constants C1 , C2 not both zero such that C1 y1 + C2 y2 ≡ 0 and
so when one of the two functions is a constant multiple of the
other. If y2 = Cy1 then W (y1 , y2 ) ≡ 0. The converse is almost
true:
y y 0 −y y 0
W /y12 = 1 2y 2 2 1 = ( yy21 )0 and so if W is identically zero the
1
ratio yy21 is some constant C and so y2 = Cy1 .
On the other hand, y1 (t) = t 3 and y2 (t) = |t 3 | have W ≡ 0
but are not linearly dependent. What is the problem? Hint:
look for a division by zero.
Now suppose the Wronskian vanishes at a single point t0 . If
y10 (t0 ) = y20 (t0 ) = 0 then we can pick C1 , C2 , not both zero,
such that with z = C1 y1 + C2 y2 is zero at t0 and of course
z 0 (t0 ) = C1 y10 (t0 ) + C2 y20 (t0 ) = 0. Otherwise, choose
C1 = y20 (t0 ) and C1 = −y10 (t0 ). Check that with
z = C1 y1 + C2 y2 we have

z(t0 ) = y1 (t0 )y20 (t0 ) − y2 (t0 )y10 (t0 ) = W (y1 , y2 )(t0 ) = 0,

z 0 (t0 ) = y10 (t0 )y20 (t0 ) − y20 (t0 )y10 (t0 ) = 0.

If y1 , y2 are solutions of the homogeneous equation

y 00 + py 0 + qy = 0 then z = C1 y1 + C2 y2 is a solution with
zero initial conditions. This means that z ≡ 0 and so y1 and
y2 are linearly dependent.
A pair y1 , y2 of solutions of the homogeneous equation
y 00 + py 0 + qy = 0 is called a fundamental pair if the
Wronskian does not vanish. The general solution is then
C1 y1 + C2 y2 because we can solve every IVP.
Abel’s Theorem
Abel’s Theorem shows that the Wronskian of two solutions
y1 , y2 of y 00 + py 0 + qy = 0 satisfies a first order linear ode:

W 0 = y1 y200 − y2 y100 + y10 y20 − y20 y10 = y1 y200 − y2 y100

q× y1 y2 − y2 y1 = 0
p× y1 y20 − y2 y10 = W
1× y1 y200 − y2 y100 = W0

The left side adds up to y1 0 − y2 0 = 0 and so 0 = W 0 + pW

or dW
dt
+ pW = 0.
R
This is variables separable with solution W = C × e − p(t)dt .
Since the exponential is always positive, W ≡ 0 if C = 0 and
otherwise W is never zero.
Homogeneous, Constant Coefficients, Sections 3.1, 3.3
The derivative itself is an example of a linear operator:
D(y ) = y 0 . For a linear operator L a function y is an
eigenvector with eigenvalue r when L(y ) = ry . For the
operator D, an eigenvector y satisfies y 0 = D(y ) = ry . This is
exponential growth with growth rate r and so the solutions are
constant multiples of y = e rt . Thus, such exponential
functions have a special role to play.
A second order, linear, homogeneous equation with constant
coefficients is of the form Ay 00 + By 0 + Cy = 0 with A, B, C
constants and A 6= 0.
We look for a solution of the form y = e rt . Substituting and
factoring we get
(Ar 2 + Br + C )e rt = 0.
Since the exponential is always positive, e rt is a solution
precisely when r satisfies the characteristic equation
Ar 2 + Br + C = 0.
The quadratic equation Ar 2 + Br + C = 0 has roots
√
−B ± B 2 − 4AC
r= .
2A
The nature of the two roots depends on the discriminant
B 2 − 4AC .

CASE 1: The simplest case is when B 2 − 4AC > 0. There are

then two distinct real roots r1 and r2 . This gives us two special
solutions y1 = e r1 t and y2 = e r2 t with the Wronskian.
 rt 
e1 e r2 t 
W =  r1 t
 = (r2 − r1 )e (r1 +r2 )t 6= 0.
r1 e r2 e r2 t 

The general solution is then C1 e r1 t + C2 e r2 t .