Differential equations Part 2

Homogeneous equations
These are equations which can be written as
dy y
=f .
dx x
Often the right hand side will be a quotient of polynomials in x and y, where each term has the
same degree, such as:

x3 − x2 y
.
xy 2 + 2y 3

If we divide top and bottom by x3 , we can write this as

y
1−
x .
y2 y3
+ 2
x2 x3
And this can be written as
y
1−
x
 y 2  y 3 .
+2
x x
To solve a homogeneous equation, make the substitution
y
u=
x
so that

y = ux.

Since by the product rule we have

d du
(ux) = x +u
dx dx
we can write the initial equation as
du
x + u = f (u)
dx
or as
du
x = f (u) − u
dx
which is a separable equation:
1 1
du = dx.
f (u) − u x

1
Example 1. The equation to be solved is
dy y 2 − 2x2
=
dx x2
which can be written as
dy y2
= 2 − 2.
dx x
This is neither linear nor separable, but it is homogeneous. So making the substitution u = y/x:

du
x + u = u2 − 2.
dx
Separating the variables:
1 1
du = dx
u2 −u−2 x
The left hand side will need to be expanded into partial fractions:
1 1
=
u2 −u−2 (u − 2)(u + 1)
1 1 1 1
= −
3u−2 3u+1
and so when we integrate both sides, we obtain:
1
(ln(u − 2) − ln(u + 1)) = ln(x) + C.
3
Now we need a bit of algebra to tame this result. First multiply by 3:

ln(u − 2) − ln(u + 1) = 3 ln x + C

and rewrite each side as a single logarithm:

 
u−2
ln = ln(Cx3 ).
u+1

Now get rid of the logs on both sides:

u−2
= Cx3
u+1
and rewrite u as y/x:
y
−2
x = Cx3 .
y
+1
x
Multiplying the fraction out by x produces:
y − 2x
= Cx3
y+x
and this can be further simplified to be solved for y. . .

2
Second order equations
A second order differential equation is one which includes a second derivative. We shall look
only at linear equations with constant coefficients; equations which look like this:
d2 y dy
a 2
+b + cy = f (x)
dx dx
where a, b and c are all constants. If f (x) = 0, then the equation is said to be homogeneous1 .

Example 2. Show that

y = Ae3x + Be−2x
is a solution to
d2 y dy
2
− − 6y = 0.
dx dx
Start by taking derivatives:
dy
= 3Ae3x − 2Be−2x
dx
d2 y
= 9Ae3x + 4Be−2x .
dx2
Now put these into the equation:
9Ae3x + 4Be−2x − (3Ae3x − 2Be−2x ) − 6(Ae3x + Be−2x )
and collect out all the exponential terms:
e3x (9A − 3A − 6A) + e−2x (4B + 2B − 6B)
and this last can be seen to be zero.

Solution of homogeneous equations

We can turn this last example into a means of general solution. Suppose we have an equation
d2 y dy
a +b + cy = 0.
dx2 dx
and a possible solution y = ekx . Then since y 0 = kekx and y 00 = k 2 ekx we have
a(k 2 ekx ) + b(kekx ) + c(ekx ) = 0.
Taking out the common factor ekx :
(ak 2 + bk + c)ekx = 0
which produces the quadratic equation
ak 2 + bk + c = 0.
This equation is called the characteristic equation, and solving it is pretty much all we need to
do. In fact if λ and µ are the solutions to the characteristic equation, then a general solution to
the linear differential equation is
y = Aeλx + Beµx .
1
We’ve already seen the word “homogeneous” in another context; this is one of the many problems with
mathematics: there aren’t enough words to go around, and so sometimes words take on two completely different
meanings.

3
Example 3. Solve the equation
d2 y dy
2
−5 + 6y = 0.
dx dx
The characteristic equation is

k 2 − 5k + 6 = 0

which can be solved (by factorization, quadratic formula, whatever you like), to obtain

k = 2, k = 3.

Then the general solution is

y = Ae2x + Be3x .

Example 4. Solve the initial value problem

d2 y dy
2
+ − 2y = 0, y(0) = 2, y 0 (0) = 5.
dx dx
First we solve the equation to obtain a general result, and then we use the initial values to
determine the constants A and B.
First, the characteristic equation is

k2 + k − 2 = 0

which has solutions k = 1, k = −2, and so the general solution is

y = Aex + Be−2x .

Now we use the initial conditions: y(0) = 2 gives the equation

A + B = 2.

The derivative of the general solution is

y 0 = Aex − 2Be2x

and using y 0 (0) = 5 gives the equation

A − 2B = 5.

We now have a set of linear equations for A and B:

A+ B =2
A − 2B = 5

for which the solutions are A = 3 and B = −1. So the solution of the original IVP is

y = 3ex − e−2x .

4
More on the characteristic equation
We know that quadratic equations don’t necessarily have two real and distinct solutions. Some-
times the solution may be equal; other times they might be complex numbers.
These different types of solutions are reflected in the solutions of the differential equation.

1. If the solutions are equal, so that

ak 2 + bk + c = a(k − λ)2
for some λ, then the general solution is
y = Aeλx + Bxeλx .
Note the extra factor of x.
2. If the solutions of the characteristic equation are complex conjugates λ ± µj then the
general solution is
y = eλx (A cos(µx) + B sin(µx)).

Example 5. Solve the equation

d2 y dy
−6 + 9y = 0.
dx2 dx
The characteristic equation is k 2 − 6k + 0 = 0, or (k − 3)2 = 0, so there is a repeated root of
k = 3. The solution is thus
y = Ae3x + Bxe3x .

Example 6. Solve the initial value problem

d2 y dy
−4 + 13y = 0, y(0) = 2, y 0 (0) = 1.
dx2 dx
The characteristic equation is k 2 − 4k + 13 = 0, which has solutions k = 2 ± 3j. The solution is
thus
y = e2x (A cos(3x) + B sin(3x)).
Now determine the values of A and B from the initial conditions. From y(0) = 2 we obtain:
2 = e0 (A cos(0) + B sin(0)) = A
and so A = 2. To apply the second condition, we have to differentiate the general solution,
which will involve a product rule:
y 0 = 2e2x (A cos(3x) + B sin(3x)) + e2x (−3A sin(3x) + 3B cos(3x))
= e2x ((2A + 3B) cos(3x) + (2B − 3A) sin(3x))
Then from y 0 (0) = 1 we obtain
1 = e0 ((2A + 3B) cos(0) + (2B − 3A) sin(0)) = 2A + 3B.
Thus
3B = 1 − 2A = −3
and so B = −1. The final solution is thus
y = e2x (2 cos(3x) − sin(3x)).

5
Solution of non-homogeneous equations
Now we look at what happens if we have a function f (x) other than zero on the right hand side,
such as
d2 y dy
2
−5 + 6y = x2 .
dx dx
Solving such an equation requires two steps:
1. First solve the homogeneous equation
d2 y dy
2
−5 + 6y = 0.
dx dx
Call this result the homogeneous solution, and denote it h(x).
2. Find any function which satisfies the non-homogeneous equation, and call this the partic-
ular solution, or p(x).
Have found these two functions, the solution of the equation is
y = h(x) + p(x).
We know how to find h(x)—just use the characteristic equation—but we don’t know how to find
p(x). The method we will use is called the method of undetermined coefficients. Basically we
“guess” at the form of the particular solution, and substitute it into the equation to determine
any unknown coefficients.

Example 7. Find a particular solution to

d2 y dy
2
−5 + 6y = 6x2 − 2x − 2.
dx dx
Since f (x) is a polynomial, our guess will be a polynomial
p(x) = ax2 + bx + c
for values a, b and c yet to be found. Since p(x) must be a solution of the differential equation,
it must satisfy
p00 − 5p0 + 6p = 6x2 − 2x − 2.
The derivatives of p(x) are
p0 = 2ax + b
p00 = 2a
and if we substitute these into the differential equation we obtain
2a − 5(2ax + b) + 6(ax2 + bx + c) = 6x2 − 2x − 2.
Collecting together all the coefficients of powers of x:
6ax2 + (6b − 10a)x + (6c − 5b + 2a) = x2
and so comparing coefficients we must have
6a = 6
6b − 10a = 2
6c − 5b + 2a = −2
which can be solved to obtain
a = 1, b = 2, c = 1
for the particular solution
p(x) = x2 + 2x + 1.

6
Guessing the particular solution
Guessing the particular solution depends not only on the form of f (x), but also on the form of
the homogeneous solution h(x). For example, consider the equation

d2 y dy
2
−5 + 6y = e4x .
dx dx
Here the homogeneous solution is

h(x) = Ae2x + Be3x

and for a particular solution we might try

p(x) = ae4x

using the form of f (x).

However, what happens if we tried to find a particular solution for

d2 y dy
−5 + 6y = e3x .
dx2 dx
We can’t try

p(x) = ae3x

as e3x is already part of the homogeneous solution, and will just collapse down to zero. Here we
need to use

p(x) = axe3x

instead.
And for the equation

d2 y dy
2
−6 + 9y = e3x .
dx dx
which we have seen to have homogeneous solution

h(x) = Ae3x + Bxe3x .

So we can’t use either xe3x or axe3x in the particular solution. We have to go up another power
of x and try

p(x) = ax2 e3x .

In general the particular solution will have a form similar to that of f (x), as the following table

7
shows.

Homogeneous solution f (x) Particular solution

Aeλx + Beµx ekx , k 6= λ, µ aekx

Aeλx + Beµx ekx , k = λ or µ axekx

Aeλx + Bxeλx ekx , k 6= λ aekx

Aeλx + Bxeλx ekx , k 6= λ ax2 ekx

Any Polynomial Polynomial of same degree

eλx (A cos(µx) + B sin(µx)), λ 6= 0 sin(kx) or cos(kx) a cos(kx) + b sin(kx)

A cos(µx) + B sin(µx) sin(kx) or cos(kx), k 6= µ a cos(kx) + b sin(kx)

A cos(µx) + B sin(µx) sin(kx) or cos(kx), k = µ ax cos(kx) + bx sin(kx)

A complete solution
So, suppose we have a second-order non-homogeneous initial value problem. How do we find
the solution? There will be several steps:

1. Find the homogeneous solution y = h(x).

2. Find a particular solution y = p(x).

3. Add them to obtain a general solution y = h(x) + p(x).

4. Substitute the initial values into y and its derivative to obtain the values of the coefficients.

Example 8. Solve the IVP

d2 y dy
2
−5 + 4y = 6e4x , y(0) = 2, y 0 (0) = −5.
dx dx
Applying the steps above:

1. The homogeneous solution requires us to solve the characteristic equation

k 2 − 5k + 4 = 0

which has solutions k = 1, 4. So the homogeneous solution is

h(x) = Aex + Be4x .

2. Since the right hand function f (x) is part of the homogeneous solution, we need to try

p(x) = axe4x

8
 for a particular solution. Then:

p0 (x) = 4axe4x + ae4x

p00 (x) = 16axe4x + 8ae4x .

Putting these into the equation p00 − 5p0 + 4p = 6e4x produces:

16axe4x + 8ae4x − 5(4axe4x + ae4x ) + 4(axe4x )

and this can be simplified to

(16ax + 8a − 20ax − 5a + 4ax)e4x = 3ae4x

Comparing this with f (x):

3ae4x = 6e4x

so that a = 2 and our particular solution is

p(x) = 2xe4x .

3. Now we have the general solution

y = Aex + Be4x + 2xe4x .

4. Now we can use the initial values. First y(0) = 2 produces

2 = A + B.

From the general solution we find that

y 0 = Aex + 4Be4x + 2e4x + 8xe4x

and so y 0 (0) = −5 produces

−5 = A + 4B + 2.

Solving these two equations for A and B:

A = 5, B = −3

and so the solution we require is

y = 5ex − 3e4x + 2xe4x .

Systems of equations
Differential equations often occur in systems, such as this:
dx
= 1.5x − xy
dt
dy
= −3y + xy.
dt
This particular system represents a predator-prey model, where x(t) and y(t) represent the
numbers of prey and predators respectively at time t.

9
 We will look only at linear systems, of which a 2 × 2 system has the form:
x0 = ax + by (1)
0
y = cx + dy (2)
where a, b, c and d are all constants. There are methods of solving such systems which start
from writing them in matrix form:
 0   
x a b x
=
y0 c d y
but we can in fact solve a 2 × 2 linear system without matrices. Start by differentiating each
equation in the system:
x00 = ax0 + by 0 (3)
00 0 0
y = cx + dy . (4)
Now substitute equation 2 into equation 3:
x00 = ax0 + b(cx + dy). (5)
However, from equation 1 we have
x0 − ax
y= (6)
b
which we can now substitute into equation 5:
x0 − ax
 
00 0
x = ax + b cx + d
b
0 0
= ax + bcx + d(x − ax)
= (a + d)x0 + (bc − ad)x.
This can be rewritten as a homogeneous second-order equation in x:
x00 − (a + d)x0 + (ad − bc)x = 0.

Example 9. Solve the linear system:

x0 = 4x + y
y 0 = 2x + 3y.
Going through all the above motions produces the equation
x00 − 7x0 + 10x = 0
for which the characteristic equation can be solved to produce k = 2, 5. Thus:
x = Ae2t + Be5t .
From the first equation in our system
y = x0 − 4x
= (2Ae2t + 5Be5t ) − 4(Ae2t + Be5t )
= −2Ae2t + Be5t .
The solution of a linear system is often expressed in matrix form:
     
x 1 2t 1 5t
=A e +B e .
y −2 1

10
Solution of differential equations using a power series
Solving a differential equation, as we have seen, usually requires an integration somewhere.
But as we have also seen, many functions can’t be integrated in terms if simple functions. So
producing a solution to a differential equation can also be done in terms of a power series.

Example 10. Take the first order linear equation

dy
− xy = 1, y(0) = 1.
dx
The usual method would be to multiply through by the integrating factor, which in this case is
2 /2
R
−x dx
e = e−x
and so obtain
d  −x2 /2  2
e y = e−x /2 .
dx
This is correct, but not very useful, because if we try to integrate:
Z
−x2 /2 2
e y = e−x /2 dx

we have a non-integrable function on the right.

To find a power series solution, suppose the solution is
y = a0 + a1 x + a2 x2 + a3 x3 + a4 x4 + · · ·
We are going to plug this into the initial value problem, and try to work out the ai values. First,
rewrite the problem as
dy
= xy + 1, y(0) = 1.
dx
From the power series, the left hand side is
a1 + 2a2 x + 3a3 x3 + 4a4 x3 + 5a5 x4 + · · ·
and the right hand side is
1 + a0 x + a1 x2 + a2 x3 + a3 x4 + a4 x5 + · · ·
The initial value tells us that a0 = 1. Now if we equate coefficients from the last two power
series, we get:
a1 = 1
2a2 = a0
3a3 = a1
4a4 = a2
5a5 = a3
and so on. Solving these one by one produces:
a1 = 1
a2 = 1/2
a3 = 1/3
a4 = 1/8
a5 = 1/15
for the series
x2 x3 x4 x5
y =1+x+ + + + + ···
2 3 8 15

11
Bessel Functions. These are the solutions to the differential equation
d2 y dy
x2 +x + (x2 − n2 )y = 0.
dx2 dx
This equation in fact gives rise to many different functions, which are used in such applications
as vibrating membranes, or heat conduction in a cylinder, and lots of others.
If n is an integer, then we can define
∞
X (−1)k  x n+2k
Jn (x) =
k!(n + k)! 2
k=0

(There are also definitions for Jn (x) when n is not an integer, but they are slightly more com-
plicated.) In this case the function Jn (x) is called a Bessel function of the first kind of order
n.

Airy functions. These are the solutions of the second order differential equation
d2 y
− xy = 0
dx2
and arise naturally in many problems involving optics. One such function is denoted Ai(x) and
can be defined as:
 
1 3 1·4 6 1·4·7 9
Ai(x) = Ai(0) 1 + x + x + x + ···
3! 6! 9!
 
0 2 4 2 · 5 7 2 · 5 · 8 10
+ Ai (0) x + x + x + x + ···
4! 7! 10!

where the values Ai(0) and Ai0 (0) would be given as initial conditions to the differential equation.

Exercises
1. Solve the second-order equations:
(a) y 00 − y 0 − 6y = 0 (b) y 00 + 2y 0 − 15y = 0 (c) y 00 − 10y 0 + 25y = 0

(d) y 00 + 8y 0 + 16y = 0 (e) y 00 + 2y 0 + 5y = 0 (f) y 00 − 6y 0 + 25y = 0

2. Solve the initial value problems:
(a) y 00 − y 0 − 6y = 0, y(0) = 10, y 0 (0) = −5

(b) y 00 + 2y 0 − 15y = 0, y(0) = 4, y 0 (0) = 4

(c) y 00 − 10y 0 + 25y = 0, y(0) = 3, y 0 (0) = −1

(d) y 00 + 8y 0 + 16y = 0, y(0) = 2, y 0 (0) = 3

(e) y 00 + 2y 0 + 5y = 0, y(0) = 1, y 0 (0) = −5

(f) y 00 − 6y 0 + 25y = 0, y(0) = −4, y 0 (0) = 0

3. For each of the following non-homogeneous equations, find a particular solution of the
form given:

12
 (a) y 00 − y 0 − 6y = 4e2x , p(x) = ae4x

(b) y 00 + 2y 0 − 15y = 4e3x , p(x) = axe3x

(c) y 00 − 10y 0 + 25y = 25x2 − 1, p(x) = ax2 + bx + c

(d) y 00 + 8y 0 + 16y = 10 sin(2x), p(x) = a cos(2x) + b sin(2x)

(e) y 00 + 2y 0 + 5y = 5e−5x , p(x) = ae−5x

(f) y 00 − 6y 0 + 25y = 17 cos(x), p(x) = a cos(x) + b sin(x)

4. Using the results of the previous questions, solve the initial value problems:
(a) y 00 − y 0 − 6y = 4e2x , y(0) = 4, y 0 (0) = 3

(b) y 00 + 2y 0 − 15y = 4e3x , y(0) = −1, y 0 (0) = 1/2

(c) y 00 − 10y 0 + 25y = 25x2 − 1, y(0) = 0, y 0 (0) = 1/5

(d) y 00 + 8y 0 + 16y = 10 sin(2x), y(0) = 1/5, y 0 (0) = −1/5

(e) y 00 + 2y 0 + 5y = 5e−5x , y(0) = 1, y 0 (0) = 0

(f) y 00 − 6y 0 + 25y = 17 cos(x), y(0) = 1/2, y 0 (0) = 0

5. Solve the following non-homogeneous IVP’s:
(a) y 00 − 3y 0 − 10y = 2e3x , y(0) = −1, y 0 (0) = −6

(b) y 00 − y 0 − 2y = −3e−x , y(0) = 1, y 0 (0) = −3

(c) y 00 − y 0 − 6y = 3x2 + 3x, y(0) = 5, y 0 (0) = −5

(d) y 00 − 6y 0 + 9y = 13 sin(2x), y(0) = 1, y 0 (0) = 0

(e) y 00 + 4y 0 + 4y = 4e−2x , y(0) = 3, y 0 (0) = −1

(f) y 00 + 2y 0 + 10y = 6e−x sin(3x), y(0) = 5, y 0 (0) = 0

6. Use power series to solve y 0 = x2 y. The equation can easily be solved by separating
variables. Find the solution and represent it as a power series. Compare the solutions.
(Hint: While solving via power series, expand the series to n = 12 )
7. Show that
     
x 1 −3t 4 2t
=A e +B e
y −1 1
is the general solution of the system

x0 = x + 4y
y 0 = x − 2y.

8. Show that
     
x 1 −2t 3 5t
=A e +B e
y −2 1
is the general solution of the system

x0 = 4x + 3y
y 0 = 2x − y.

13
Answers
1. (a) y = Ae3x + Be−2x , (b) y = Ae−5x + Be3x , (c) y = Ae5x + Bxe5x ,
(d) y = Ae−4x + Bxe−4x , (e) y = e−x (A cos(2x) + B sin(2x)),
(f) y = e3x (A cos(4x) + B sin(4x))

2. (a) y = 3e3x + 7e−2x , (b) y = e−5x + 3e3x , (c) y = 3e5x − 16xe5x , (d) y = 2e−4x + 11xe−4x ,
(e) y = e−x (cos(2x) − 2 sin(2x)), (f) y = e3x (3 sin(4x) − 4 cos(4x))

3. (a) p = −e2x , (b) p = (1/2)xe3x , (c) p = x2 + (4/5)x + 1/5

(d) p = (3/10) sin(2x) − (2/5) cos(2x) (e) p = (1/4)e−5x (f) p = (2/3) cos(x) − (1/6) sin(x)

4. (a) y = 3e3x + 2e−2x − e2x , (b) y = (1/2)xe3x − (3/4)e3x − (1/4)e−5x ,

(c) y = (1/5)(2xe5x − e5x + 5x2 + 4x + 1),
(d) y = (1/5)(e−4x (8x + 3) + (1/10)(−4 cos(2x) + 3 sin(2x)),
(e) y = (1/4)(4ex sin(2x) + 3ex cos(2x) + e−5x ),
(f) y = (1/6)(e3x sin(4x) − e3x cos(4x) + 4 cos(x) − sin(x))

5. (a) y = (1/5)e−2x − e5x − (1/5)e3x , (b) y = 2e−x − e2x + xe−x ,

(c) y = 4e−2x + (10/9)e3x − (1/2)x2 − (1/3)x − 1/9,
(d) y = (1/13)(e3x − 13xe3x + 12 cos(2x) + 5 sin(2x)), (e) y = 3e−2x + 5xe−2x + 2x2 e−2x ,
(f) y = e−x (5 cos(3x) + 2 sin(3x)) − xe−x cos(3x)

6. Direct solution
x3
y = Ce 3

Series solution
∞
x x2 x3 X x3 x3
a0 + a0 + a0 + a0 + · · · = a0 = a0 e 3
3 18 162 3n n!
n=0

14