First Order Differential Equations (continued)

In the first week, we introduced some terminology regarding ordinary differential

equations in general and then focused on first-order differential equations.

We have stated the existence uniqueness theorem for first-order initial value problems
written in the canonical form:
dy
= f (t, y), y(t0 ) = y0
dt
No general method exists to find solutions to a given differential equation, even
when we know a solution exists. However, we can find solutions for some classes
of differential equations. Last week, we discussed the case when the equation is
separable; such equations can be written in differential form as follows:
M (t) dt = N (y) dy
Note that M depends only on t, and N depends only on y. Now, we are going to
discuss another class of first-order differential equations. They are called homoge-
neous equations. Such equations can be transformed into separable equations, for
which we know how to find solutions.

Homogeneous differential equations (ordinary, first-order)

An ordinary, first-order, homogeneous differential equation is an equation of the
following form:
dy y 
=F
dt t
We can transform such equations into a separable equation by making the substi-
tution
y
v=
t
It follows that
y = tv =⇒ y ′ = v + tv ′
Therefore,
dy y  dv dv dt
=F =⇒ v+t = F (v) =⇒ =
dt t dt F (v) − v t
Note that this is now a separable equation where
1 1
M (t) = and N (v) =
t F (v) − v

1
Example: Solve the IVP:
t2 y ′ = y 2 + 2ty y(1) = −1
The equation is not separable and cannot be separated like we did in the first week.
However, it is homogeneous. To see that, divide all terms by t2 to get
 y 2 y
′
y = +2
t t
Putting v = y/t as above, we get
v + tv ′ = v 2 + 2v =⇒ tv ′ = v 2 + v
This equation can be separated now:
Z Z
dv dt dv dt
= =⇒ =
v2 + v t v2 + v t
We apply the partial fraction decomposition to the left side:
1 1 A B
=⇒ = + =⇒ A = 1, B = −1
v2 + v v(v + 1) v v+1
Thus, after the integration, we get
v
ln = ln |t| + C
v+1
The ln trick: Remember that
domain(ln x) = (0, ∞) and range(ln x) = (−∞, ∞)
Therefore, any real number C can be written as C = ln K for some positive number
K > 0. Therefore, we can replace any real number C with ln C where C > 0.

We apply the trick to the solution above to get

v
ln = ln C|t| where C > 0
v+1
Hence,
v
= C|t| where C > 0
v+1
We can get rid of the absolute value if we allow C to take any nonzero value:
v
= Ct, C ̸= 0
v+1

2
Putting v = y/t again, we get
y Ct2
= Ct =⇒ y = Cty + Ct2 =⇒ y= , C ̸= 0
y+t 1 − Ct
CAUTION: Always remember that along the process of separation, we may lose
some constant solutions. Remembering the equation before separation
tv ′ = v 2 + v = v(v + 1)
we see that
v=0 =⇒ y = 0 and v = −1 =⇒ y = −t
are solutions, too. Therefore, the general solution is

Ct2
y= , y = −t
1 − Ct
Note that, the general solution above contains the solution y(t) = 0 (when C = 0).

Inserting the initial condition we get

C
= −1 =⇒ C −1=C =⇒ −1 = 0 absurdity!
1−C
Therefore,
Ct2
y=
1 − Ct
is not a solution to the IVP as it does not satisfy the initial condition whatever
C is. But the solution y(t) = −t satisfies the initial condition!

Thus, there is only one solution to the IVP:

y(t) = −t
Definition: A polynomial P (t, y) in two variables is called an homogeneous poly-
nomial if all the terms of P have the same degree.

Exercise: Determine if the given polynomial is homogeneous or not:

1. P (t, y) = 3t2 − ty + 2y 2
2. P (t, y) = t3 − ty 2 + 5y 3
3. P (t, y) = t3 − ty 2 + 5y 3 − 1
4. P (t, y) = π

3
Relation between homogeneous differential equations and homogeneous
polynomials: Again, consider a first-order differential equation written in its canon-
ical form:
y ′ = f (t, y)
If f is a ratio of homogeneous polynomials of the same degree, then the differential
equation can be easily transformed into a homogeneous differential equation. More
precisely, let P (t, y) and Q(t, y) be two homogeneous polynomials of the same degree,
say d, and let f = P/Q. Then,
P (t, y)
y ′ = f (t, y) =
Q(t, y)
Now, one can divide every term on the right side by td where d = deg P = deg Q.
4y − 3t
Example: Solve y ′ = .
2t − y
Referring to the notations above, P (t, y) = 4y − 3t and Q(t, y) = 2t − y are homoge-
neous polynomials of degree d = 1. Therefore, dividing every term on the right by
td = t we get a homogeneous differential equation of the form y ′ = F (y/t):
4y − 3t 4y/t − 3
y′ = =⇒ y′ = = F (y/t)
2t − y 2 − y/t
Substituting v = y/t =⇒ y ′ = v + tv ′ , we get a separable equation in v:

′ 4v − 3 ′ v 2 + 2v − 3 (2 − v)dv dt
v + tv = =⇒ tv = =⇒ 2
=
2−v 2−v v + 2v − 3 t
To integrate the left side, we apply partial fraction decomposition:
(2 − v) A B 5 1
= + =⇒ A=− , B=
v 2 + 2v − 3 v + 3 v − 1 4 4
Hence, after integrating, we get
ln |v − 1| − ln |v + 3|5 = C + ln |t|4
Again, let’s use the ln-trick to replace C by ln C where C > 0. Applying the
logarithm laws carefully, it follows that
v−1
5
= Ct4 , C > 0.
(v + 3)
Now, substitute v = y/t back and then multiply/divide the left side by t5 to get
y/t − 1 ⋆ yt4 − t5 y−t
5
= Ct4 =⇒ 5
= Ct4 =⇒ = C, C>0
(y/t + 3) (y + 3t) (y + 3t)5

4
 t5
For ⋆ we multiplied the left-hand side by 5 (check the calculation). Letting C
t
retake negative values, we can remove the absolute value:
y − t = C(y + 3t)5 , C ̸= 0
Again, we check the equation before separation to see if any other solutions are lost
along the separation:
v 2 + 2v − 3 (v + 3)(v − 1)
tv ′ = =
2−v 2−v
Thus,
v=1 =⇒ y = t and v = −3 =⇒ y = −3t
are solutions, too. Hence, the general solution is
y − t = C(y + 3)5 , y = t, y = −3t, C ̸= 0
Observe that y = t is the solution when C = 0. Thus, one can write the general
solution as
y − t = C(y + 3)5 , y = t, C ∈ R

Exact Equations and Integrating Factors

Several integration methods apply to various classes of problems for first-order equa-
tions. We have previously discussed the class of separable and homogeneous equa-
tions (the latter can be transformed into separable equations). Here, we consider
a class of equations known as exact equations for which there is also a well-
defined method of solution.

Again, let’s consider a first-order equation written in canonical form:

dy
= f (y, t)
dt
Remember that we call such an equation separable if it can be written in the form
dy
M (t) + N (y)
= 0.
dt
Then, if one knows the antiderivatives of M and N , one can integrate easily to get
the general solution in the form of a family of implicit solutions.
Z Z
M (t) dt + N (y) dy = C

5
Remember that this process needs some care, as some constant solutions may
be lost during separation.

Definition: Now, consider a first-order equation in one of the following equivalent

forms:
dy
⋆ M (t, y) + N (t, y) =0 OR M (t, y) dt + N (t, y) dy = 0 ⋆
dt
Then, it is said to be an exact equation if the vector field F = ⟨M, N ⟩ is conser-
vative. In other words, if F(t, y) is the gradient of some function Ψ(t, y):
F = ∇Ψ =⇒ ⟨M, N ⟩ = ⟨Ψt , Ψy ⟩ =⇒ M = Ψt , N = Ψy for some Ψ
Here Ψ is called an energy function or a potential function for F.

It is easy to integrate exact equations. Replacing M by Ψt and N by Ψy , we can

write ⋆ as follows:

dy d
Ψt + Ψy = 0 =⇒ Ψ(t, y) = 0, (y = y(t) is the solution we are looking for)
dt dt
Hence, the general solution can be written as a family of implicit solutions:
Ψ(t, y) = C, C∈R
CAUTION: Note that
d ∂
Ψ(t, y) ̸= Ψ(t, y)
dt ∂t
For the left-hand side we need to use the Chain Rule for multivariable func-
tions where the unknown function y = y(t) depends on t:

t y

t
For the right-hand side however, we consider t and y as independent variables, i.e.
y is treated as a constant when partially differentiating Ψ with respect to t.

6
Example: All separable equations are exact. To see this, let’s take a separable
equation:
dy
M (t) + N (y) = 0
dt
It is easy to find a potential for F = ⟨M (t), N (y)⟩:
Z Z
Ψ(t, y) = M (t) dt + N (y) dy .

Then, ∇Ψ = ⟨M, N ⟩ = F, isn’t it? Therefore, F is a conservative vector field.

Furthermore, the general solution is just as above:
Z Z
Ψ(t, y) = C =⇒ M (t) dt + N (y) dy = C

Before we discuss examples of (nonseparable) exact equations, we need a quick re-

view of the basic theory of conservative vector fields.

Theorem: If F = ⟨M, N ⟩ is conservative, and My and Nt are continuous, then

My = Nt
Proof: This follows from the Clairaut’s Theorem which states that if a function
has continuous second partial derivatives, then the mixed derivatives are
equal. Now, assume that F is conservative. Then by the definition, there is an
energy function Ψ for F:
⋆
F = ∇Ψ =⇒ M = Ψt and N = Ψy =⇒ My = Ψty = Ψyt = Nt
where ⋆ follows from the Clairaut’s Theorem since Mt and Ny corresponds to the
second partial derivatives of the energy function.

Question: Is the opposite of the implication stated in the theorem above true? In
other words, does F = ⟨M, N ⟩ always have a potential function when My = Nt ?

Answer: The answer is no, in general. In other words, My = Nt does not always
imply that F is conservative. However, the following is true:

Theorem: If My = Nt on a simply-connected region R ⊂ R2 then F is conser-

vative on R.

7
Proof: We omit the proof because we don’t want to go all over MAT120 again! Just
a reminder, though a simply-connected region in R2 is a region without holes!
For instance, a disk t2 + y 2 < 5 is simply-connected as it does not have any hole in
it, whereas R2 − {(0, 0)} is not a simply-connected region.

This much review is enough to return to the study of Exact Equations. Let’s start
with an example.

Example: The following is a nonlinear, nonseparable, but an exact equation:

y cos t + 2tey + sin(t) + t2 ey − 1 y ′ = 0

Referring to the form given in the definition of exact equations above, we take
M = y cos t + 2tey and N = sin(t) + t2 ey − 1 =⇒ My = cos t + 2tey = Nt

Note that the last equality above is everywhere on R2 , a simply-connected region.

Therefore, by the previous theorem, F = ⟨M, N ⟩ is a conservative vector field.

We need to find a potential function Ψ of F to solve the equation. That is,

⃝
F = ∇Ψ =⇒ M = Ψt and N = Ψy .
We can use these to recover a potential function:
Z
y ⋆
Ψt = M = y cos t + 2te =⇒ Ψ = (y cos t + 2tey ) dt = y sin t + t2 ey + C(y).

So, we have
⋆ ⃝
Ψy = sin t + t2 ey + C ′ (y) and Ψy = N = sin t + t2 ey − 1 =⇒ C ′ (y) = −1
Choosing C(y) = −y, we get a potential function:
Ψ = y sin t + t2 − y.
Hence, Ψ = C gives the general solution as a family of implicit solutions:
y sin t + t2 ey − y = C
Example: Is the following equation exact?
t y
+ y′ = 0
(t2 + y 2 )3/2 (t2 + y 2 )3/2

8
Remember that the definition of exactness is given when the differential equation is
in the following form:
M + N y ′ = 0 or equivalently M dt + N dy = 0
In the discussion above, we noted that for an equation to be exact, the equality
My = Nt is necessary. It may not be sufficient, though, when the domain we
consider is not simply-connected.

Here, My = Nt holds:
t y ty
M= , N= =⇒ My = −3 = Nt
(t2 + y 2 )3/2 (t2 + y 2 )3/2 (t2 + y 2 )5/2
Here, the domain of consideration is not simply-connected as M and N is defined
on R − {(0, 0)}, which is not simply connected.

However, we can try the above standard procedure to find a potential Ψ for F. It
may not work, but what else would we lose except time if we try?
Z Z Z Z
t 1 −3/2 1
Ψ(t, y) = Ψt dt = M dt = 3/2
dt = u du = − √ + C(y)
2 2
(t + y ) 2 u

We used the substitution u = t2 + y 2 =⇒ du = 2t dt. Then

∂ 1 y ′ y
Ψy = Ψ = −p = 2 + C (y) = N (t, y) =
∂y t2 + y 2 (t + y 2 )3/2 (t2 + y 2 )3/2
The procedure works even if the domain is simply-connected! Note that C ′ (y) = 0
gives a potential by choosing C(y) = 0. Hence,
1
F = ⟨M, N ⟩ = ∇Ψ where Ψ = − p
t2 + y 2
Remember that whenever there is such a potential Ψ for a differential equation
M + N y ′ = 0, the general solution is given by an implicit family of solutions Ψ = C.
Thus, for this example, the general solution is
1 p
p = C =⇒ t2 + y 2 = C =⇒ t2 + y 2 = C
t2 + y 2
These curves are sometimes called integral curves. For this example, they are
just a family of circles centered at the origin. In other words, whatever the initial
condition is, the solution goes along the circles (as long as the initial condition is not
(0, 0)).

9
Keep in mind: In this example, we have seen that even though the domain is
not simply-connected, we pursued the fact that My = Nt to find a potential func-
tion. Sometimes, there is no potential function even if My = Nt when the domain is
simply-connected.

However, if My = Nt , one can always consider the differential equation M + N y ′ = 0,

on a simply-connected part R on which My = Nt . Then, the differential equation
is exact on R. In any case, finding the potential function explicitly is another task,
even if it is known that there is a potential.

Example: The following differential equation is not exact:

(et sin y + 3y) − (3t − et sin y)y ′ = 0
Here, M = et sin y + 3y, and N = et sin y − 3t. Then My = 3 + et cosy, and
Nt = et sin y − 3. Since My ̸= Nt , the differential equation is not exact. It is neither
separable nor homogeneous, too. Therefore, we don’t know any method of solving it!

Integrating Factors

Non-exact equations, as in the previous example, may be multiplied by a function

µ = µ(t, y) called an integrating factor, after which the equation becomes exact:
M + N y′ = 0 =⇒ µM + µN y ′ = 0
Now, this equation is exact if
⋆
(µM )y = (µN )t =⇒ µy M + µMy = µt N + µNt
We want to find such a function µ. Notice that ⋆ is a partial differential equa-
tion because the unknown function µ depends on two variables! It looks like we
made this more complicated. We may consider some special cases though, for which
⋆ is easier to solve.

For instance, if µ = µ(t) depends only on t, then µy = 0, so ⋆ becomes:

dµ My − Nt
= µ
dt N
Similarly, if µ = µ(y) depends only on y, then µt = 0, so ⋆ becomes:
dµ Nt − My
= µ
dy M

10
Example: y + (2t − yey )y ′ = 0

This is not a separable equation. Putting M = y, and N = 2t−yey , we have My = 1,

and Nt = 2. Since My ̸= Nt , it is not an exact differential equation.
1. Multiplying the equation by an integrating factor µ = µ(t) makes the equation
exact if µ satisfies the equation
dµ ⋆ My − Nt dµ 1
= µ =⇒ =− µ
dt N dt 2t − yey
Now, ⋆ looks hopeless since we don’t know y = y(t), which is what we want to
solve for.

2. Multiplying the equation by an integrating factor µ = µ(y) makes the equation

exact if µ satisfies the equation
dµ ⋆ Nt − My dµ µ dµ dy
= µ =⇒ = =⇒ = = ln |µ| = C + ln |y|
dy M dy y µ y
Remember, we aim to solve for y = y(t), and any nonzero integrating factor
would work. Thus, we can choose C = 0, and µ = y.
Now, let’s return to the equation we want to solve and multiply it by µ = y:
⋆
y + (2t − yey )y ′ = 0 =⇒ y 2 + (2ty − y 2 ey )y ′ = 0
Exercise: Check that ⋆ is exact. To solve this, we need to find a potential function
Ψ where ∇Ψ = F = ⟨M, N ⟩, and the new M and N are the previous ones above,
multiplied by µ = y. That is
Z Z
2 2 y
Ψt = M = y , Ψy = N = 2ty − y e =⇒ Ψ = Ψt dt = y 2 dt = ty 2 + C(y)

Thus,
N = Ψy =⇒ 2ty + C ′ (y) = 2ty − y 2 ey =⇒ C ′ (y) = −y 2 ey
Exercise: Apply integration by parts two times to get
C(y) = −y 2 ey + 2yey − 2ey
Hence, the general solution is

Ψ=C =⇒ ty 2 − y 2 ey + 2yey − 2ey = C

11