Differential Equations II

Josué David Regalado López

November 27, 2023

1 First lecture
F (t, x, x(1) , . . . , x(n) ) = 0 (1)
F is a function defined on some domain of Rn+2 ; t is the argument of an unknown function
dk x
x = x(t) and x(k) (t) = k . n is called the order of (1).
dx
Observation 1 (Notation). We have
x′ = x(1)
x′′ = x(2)
x′′′ = x(3)

When n = 1, we have
F (t, x, x′ ) = 0 (2)
A solution of (2) is a C 1 function x : I −→ R; I is an interval of real numbers such that

F (t, x(t), x′ (t)) = 0, ∀t ∈ I

Under certain hypothesis we can reduce (2) to an equation of the form

x′ = f (t, x), f : Ω ⊆ R × R −→ R

where f is continuous and Ω is an open set. Moreover, x′ = f (t, x) is the normal form.

Example 1.
x′ = ax, a∈R
x(t) = eat , x′ (t) = aeat = ax(t)
x(t) = Ceat =⇒ x(t) = aCeat = ax(t), ∀t ∈ R
Let x̃(t) be other solution, then

d
x̃(t) · e−at = x̃′ (t)e−at + x̃(t)(−ae−at )


dt
= ax̃(t)e−at − ax̃(t)e−at
=0

So, we have
d
x̃(t) · e−at = 0 =⇒ ∃C ∈ R, such that x̃(t)e−at = C =⇒ x̃(t) = Ceat .


dt
1
Cauchy Condition
x(t0 ) = x0
Cauchy problem
x′ = f (t, x), x(t0 ) = x0
In the previous example,
x(t) = Ceat , x0 = x(t0 )

x(t) = Ceat =⇒ x(t0 ) = Ceat0 = x0

=⇒ C = x0 e−at0
=⇒ x(t) = x0 e−at0 eat
=⇒ x(t) = x0 ea(t−t0 )

There is not loss of generality if we suppose t0 = 0. If x(t) = x0 eat , then x̃(t) = x(t − t0 ) =
x0 ea(t−t0 ) . The constant a is a parameter of the family of solutions x(t) = x0 eat
Comportamiento asintótico

1. x0 > 0, a > 0, lim x(t) = +∞

t→+∞

2. x0 > 0, a = 0, lim x(t) = x0

t→+∞

3. x0 > 0, a < 0, lim x(t) = 0

t→+∞

First-order differential systems

x′i = fi (t, x1 , . . . , xn ), i = 1, . . . , n (3)

where fi : Ω ⊆ R × Rn −→ R, Ω is an open set and fi are continuous.
C1
A solution for (3) is a set {x1 (t), . . . , xn (t)}, xi : I −→ R such that

x′i (t) = fi (t, x1 (t), . . . , xn (t)), t ∈ R, ∀i = 1, . . . , n

A Cauchy problem for a system of the form x′i = f (t, x1 , . . . , xn ), with xi (t0 ) = xi,0 .

Geometrically a solution

x : I −→ R of x′ = f (t, x), x(t0 ) = x0

is a curve in the tx-plane passing through (t0 , x0 ) and x′ (t) = f (t, x).

For first-order systems

x′i = fi (t, x1 , . . . , xn ), i = 1, . . . , n
a solution {x1 (t), . . . , xn (t)}, t ∈ I determines a curve or trajectory

t ∈ I 7−→ (x1 (t), . . . , xn (t)) ∈ Rn

such that (x′1 (t), . . . , x′n (t)) = (f1 (t, x1 (t), . . . , xn (t)), . . . , fn (t, x1 (t), . . . , xn (t)))

For n-th order differential equation systems

x(n) = f (t, x, x(1) , . . . , x(n−1) )

2
where f : Ω ⊆ R × Rn −→ R.
x(t0 ) = x0,0
x′ (t0 ) = x0,1
x′′ (t0 ) = x0,2
..
.
x(n−1) (t0 ) = x0,n−1
Cn
A solution is a function x : I −→ R such that

x(n) (t) = f (t, x1 (t), . . . , xn (t)), ∀t ∈ I, x(k) (t0 ) = x0,k

If x1 := x, x2 := x′ ,..., xn := x(n−1) , then

x′1 = x′ = x2

x′2 = x′′ = x3
..
.
x′n = x(n) = f (t, x1 , . . . , xn )
with
x1 (t0 ) = x(t0 ) = x0,0
..
.
xn (t0 ) = x0,n−1
Example 2.
x′′ = −x, x(t0 ) = x0,0 , x′ (t0 ) = x0,1
x′′ + x = 0, characteristic equation r2 + 1 = 0 =⇒ r = ±i. So, we have

{cos(t), sin(t)} =⇒ x(t) = C1 cos(t) + C2 sin(t)

Now,
x1 = x
x2 = x ′
 ′ 
x′1 = x2
  
x1 0 1 x1
≡ =
x′2 = −x1 x2 −1 0 x2
x0,0 = x(0) = C1
On the other hand,

x′ (t) = −C1 sin(t) + C2 cos(t) =⇒ x′ (0) = C2 = x0,1

Therefore,
x(t) = C1 cos(t) + C2 sin(t) = x0,0 cos(t) + x0,1 sin(t)
Moreover,
x1 (t) = x(t) = x0,0 cos(t) + x0,1 sin(t)
x2 (t) = x′ (t) = −x0,0 sin(t) + x0,1 cos(t)
t ∈ R 7−→ (x0,0 cos(t) + x0,1 sin(t), −x0,0 sin(t) + x0,1 cos(t))

3
We can observe the following
x′1 = x2


x′2 = −x1
d 2
(x1 + x22 ) = 2x1 x′1 + 2x2 x′2 = 2x1 x2 − 2x1 x2 = 0
dt
This implies there exists a constant r2 such that x21 + x22 = r2

x1 (t) = r cos(θ(t)), x2 (t) = r sin(θ(t))

Elementary methods for solving ODEs

1.
x′ = f (t), x(t0 ) = x0 , f : I −→ R, t0 ∈ I, f is continuous
If x(t) is solution of this ODE, then

x′ (t) = f (t), ∀t ∈ J ⊆ I

ˆ t ˆ t ˆ t
′
x (τ ) dτ = f (τ ) dτ ⇐⇒ x(t) − x(t0 ) = f (τ ) dτ
t0 t0 t0
ˆ t
⇐⇒ x(t) = x0 + f (τ ) dτ
t0

2. Separable equations

x′ = f (t)h(x), (t, x) ∈ I × J ⊆ R2 , x(t0 ) = x0

where I and J are open sets. Also, let’s consider h(x) ̸= 0, ∀x ∈ J.

If x : I˜ ⊆ I −→ R is a solution, then

x′ (t) = f (t)h(x(t)), ∀t ∈ I˜ ⊆ I

x′ (t)
x′ (t) = f (t)h(x(t)) =⇒ = f (t)
h(x(t))
ˆ t ′ ˆ t
x (τ )
=⇒ dτ = f (τ ) dτ
t0 h(x(τ )) t0
ˆ x(t) ˆ t
dτ
=⇒ = f (τ ) dτ
x(t0 ) h(τ ) t0
ˆ x(t) ˆ t
dτ
=⇒ = f (τ ) dτ
x0 h(τ ) t0

Let H : J −→ R, ˆ x
dτ 1
H(x) := =⇒ H ′ (x) = ̸= 0
x0 h(τ ) h(x)
This implies that H is monotone and a diffeomorphism. In particular, H(J) is an open
interval and H −1 : H(J) −→ H is differentiable
ˆ x(t) ˆ t ˆ t 
dτ
H(x(t)) = = f (τ ) dτ =⇒ x(t) = H −1
f (τ ) dτ , t ∈ I˜
x0 h(τ ) t0 t0

4
Conversely, let  ˆ t 
a = inf t ∈ I : f (τ ) dτ ∈ H(J)
t0
 ˆ t 
b = sup t ∈ I : f (τ ) dτ ∈ H(J)
t0

Then, ˆ t 
−1
x(t) = H f (τ ) dτ , t ∈ (a, b)
t0

By chain rule, ˆ t 
′ −1 ′
x (t) = (H ) f (τ ) dτ · f (t)
t0

By Inverse Function Theorem,

1
x′ (t) =  ˆ t  · f (t)
′
H H −1 f (τ ) dτ
t0
1
= · f (t)
H ′ (x(t))
1
= · f (t)
1
h(x(t))
= h(x(t))f (t)

So, we obtain x′ (t) = h(x(t))f (t)

5
 Example 3.
  x 
x′ = x(1 − x) Logistic Differential Equation: x′ = ax 1 −
N
From this equation, we have
x′ = x(1 − x) = f (t, x) = f (x), x(0) = x0
Consider these cases: x ∈ (−∞, 1), x ∈ (0, 1), x ∈ (1, +∞). Also, x(t) ≡ 0, x(t) ≡ 1 are
solutions.
ˆ t0 ′ ˆ t
x′ x
= 1 =⇒ dτ = dτ
f (x) f (x)
ˆ0 x 0
dτ
=⇒ =t
x0 f (τ )
ˆ x(t)
dτ
=⇒ =t
x0 τ (1 − τ )
ˆ x(t) ˆ x(t)
dτ dτ
=⇒ + =t
x0 τ x0 1−τ
=⇒ ln |x(t)| − ln |x0 | − ln |1 − x(t)| + ln |1 − x0 | = t
x(t) x0
=⇒ ln − ln =t
1 − x(t) 1 − x0
x(t)
1 − x(t)
=⇒ ln x0 =t
1 − x0
x(t)
1 − x(t)
=⇒ x0 = et
1 − x0
 
t x0
e
1 − x0
=⇒ x(t) =  
x 0
1 + et
1 − x0
So, x(t) is solution of the LDE.

2 Second Lecture
2.1 Separable equations
x′ = f (t)g(x), (t, x) ∈ I × J ⊆ R2 , g(x) ̸= 0, ∀x ∈ J
′
x = f (at + bx + c), f : R −→ R continuous such that a + bf (u) ̸= 0, ∀u ∈ R.

If x : I −→ R is a solution, then the function

u(t) := at + bx(t) + c, t∈I
Then,
u′ (t) = a + bx′ (t) = a + bf (u(t))
which is separable.

6
Example 4.
x′ = et−x+1 + 1 =⇒ f (u) = eu + 1
where u(t) = t − x(t) + 1. Now,

u′ (t) = 1 − x′ (t) = 1 − et−x(t)+1 + 1 = −et−x(t)+1 = −eu(t)

Then,
ˆ t
u′ (t) u′ (τ )
= −1 =⇒ dτ = −(t − t0 )
eu(t) t0 eu(τ )
ˆ u(t)
=⇒ e−τ dτ = −(t − t0 )
u(t0 )

=⇒ −e−u(t) + e−u0 = −(t − t0 )

=⇒ −e−u(t) = −e−u0 − (t − t0 )
=⇒ e−u(t) = e−u0 + t − t0
=⇒ u(t) = − ln |e−u0 + t − t0 |
=⇒ x(t) = t + 1 + ln |e−u0 + t − t0 |

2.2 Homogeneous equations

• U ⊆ R2 is invariant under homotheties if

∀s ∈ R, s ̸= 0, (t, x) ∈ U =⇒ (st, sx) ∈ U

• U ⊆ R2 is invariant (respectively positive invariant, s > 0) under homotheties.

• U ⊆ R2 is invariant (respectively negative invariant, s < 0) under homotheties.

• f : U −→ R is homogeneous of grade n ∈ Z if

f (st, sx) = sn f (t, x), ∀s ̸= 0

An equation of the form

M (t, x) + N (t, x)x′ = 0
where M, N : U −→ R are continuous, homogeneous of the same grade, and N (t, x) ̸= 0,
∀(t, x) ∈ M is called a homogeneous ODE. Also, let’s consider U ∩ x-axis = ∅.

M (t, x)
M (t, x) + N (t, x)x′ = 0 =⇒ x′ = − =: F (t, x)
N (t, x)

Then,

M (st, sx)
F (st, sx) = −
N (st, sx)
sn M (t, x)
=− n , ∀s ̸= 0
s N (t, x)
M (t, x)
=−
N (t, x)
= F (t, x)

7
So, F is homogeneous of grade 0. In particular,
 
1 1
F (t, x) = F · t, · x
t t
 x
= F 1,
 x t
=h
t
where h = F (1, ·).

x(t)
If x : I −→ R is a solution, let u(t) = , t ∈ I. Then,
t
x(t) = tu(t) =⇒ x′ (t) = u(t) + tu′ (t)
=⇒ h(u(t)) = u(t) + tu′ (t)
=⇒ tu′ (t) = h(u(t)) − u(t)
h(u(t)) − u(t)
=⇒ u′ (t) =
t
h(u(t)) − u(t)
If h(u) − u ̸= 0, ∀u in its domain, then u′ (t) = is separable.
t
Example 5.
(t − x) + (t + x)x′ = 0
where M (t, x) = (t − x) and N (t, x) = (t + x). We have

(t, x) ∈ U = {(t, x) ∈ R2 : t > 0 and t + x > 0}

Then, s > 0 and (t, x) ∈ U . That implies that st > 0 and s(t+x) = st+sx > 0 =⇒ (st, sx) ∈ U .
Now,
M (st, sx) = st − sx = s(t − x) = sM (t, x)
N (st, sx) = st + sx = s(t + x) = sN (t, x)
N (t, x) ̸= 0
U ∩ x − axis = ∅
Moreover,
t−x x−t x/t − 1
x′ = − = = = h(x/t)
t+x x−t x/t + 1
Let h : (−1, +∞) −→ R,
u−1
h(u) =
u+1
x(t) h(u(t)) − u(t)
If x(t) : I −→ R is solution, then u(t) = satisfies u′ (t) = .
t t
−(1 + u2 )
2 2
u−1 −1 − u −(1 + u )
h(u) − u = −u= = ̸ 0 =⇒ u′ = 1 + u
=
u+1 u+1 u+1 t

8
 Therefore, we get
ˆ t ˆ t
u′ (1 + u) 1 u′ (τ )(1 + u(τ )) 1
2
= − =⇒ 2
dτ = − dτ
1+u t t 1 + u (τ ) t0 τ
ˆ 0t
u′ (τ )(1 + u(τ )) t
=⇒ dτ = − ln
t0 1 + u2 (τ ) t0
ˆ u(t)
1+τ t
=⇒ 2
dτ = − ln
u0 1+τ t0
ˆ u(t) ˆ u(t)
τ 1 t
=⇒ 2
dτ + 2
dτ = − ln
u0 1+τ u0 1+τ t0
 
x2 (t)
1  1 + 2 

x(t)
  
x0 t
=⇒ ln 
 t − arctan = − ln
2  + arctan

2 x t t0 t0
1 + 20
t0

2.3 Equations reducible to homogeneous

 
′ at + bx + c
x =f , f : (a, b) −→ R, continuous, dt + ex + f ̸= 0
dt + ex + f

at + bx + c = 0
dt + ex + f = 0
 
a b
If det ̸= 0, then ∃(x0 , y0 ) solution. Let s = t − t0 , y = x − x0 ,
d e
 
′ as + by
y =f
ds + ey
 
a b
If det = 0, then
d e
at + bx = −c
kdt + kex − f
Let u = at + bx, then  
u+c
f
ku + f
Now,  
′ ′ u+c
u = a + bx = a + bf
ku + f
Example 6.  
′ t+x
x = ln(t + x) − ln(t + x + 1) = ln
t+x+1
   
′ ′ t + x(t) u(t)
We have u(t) = t + x(t), u (t) = 1 + x (t) = 1 + ln = 1 + ln
t + x(t) + 1 u(t) + 1

9
2.4 Linear equations (Order 1)
x′ + p(t)x = q(t), t∈I
ˆ t 
Let’s multiply exp p(τ ) dτ in either sides of the equality.
t0
´t ´t ´t ´t
p(τ ) dτ p(τ ) dτ d  p(τ ) dτ

p(τ ) dτ
e t0
(x′ + p(t)x) = q(t)e t0
=⇒ x(t)e t0 = q(t)e t0
dt
Now, let’s integrate
´t ˆ t ´τ
p(τ ) dτ p(s) ds
x(t)e t0
− x0 = q(τ )e t0 dτ
t0
´t ˆ t ´τ
p(τ ) dτ p(s) ds
x(t)e t0
= x0 + q(τ )e t0 dτ
t0
´t ´t ˆ t ´τ
− t p(τ ) dτ − p(τ ) dτ p(s) ds
x(t) = x0 e 0 +e t0
q(τ )e t0
dτ
t0
´t ˆ t ´τ
− p(τ ) dτ
x(t) = x0 e t0
+ q(τ )e t p(s) ds dτ
t0

2.5 Bernoulli’s equations

x′ + p(t)x = q(t)xα , α ̸= 0, 1 t ∈ I
x′ x−α + p(t)x1−α = q(t)
x(t) : J −→ R is solution

x′ (t)x−α (t) + p(t)x1−α (t) = q(t), ∀t ∈ J

Let v(t) = x(t)1−α

v ′ (t)
v ′ (t) = (1 − α)x(t)−α · x′ (t) =⇒ = x−α (t) · x′ (t) (4)
1−α
v ′ (t)
=⇒ + p(t)v(t) = q(t) (5)
1−α

2.6 Ricatti’s equation

x′ = c + p(t)x + q(t)x2
If φ(t) is solution, then the change of variable y = x−φ(t) transforms the ODE into a Bernoulli.
Now

x = φ + y =⇒ x′ = φ′ + y ′
=⇒ φ′ + y ′ = c + p(t)(φ + y) + q(t)(φ + y)2
=⇒ φ′ + y ′ = c + p(t)φ + p(t)y + q(t)(φ2 + 2φy + y 2 )
=⇒ φ′ + y ′ = c + p(t)φ(t) + p(t)y + q(t)φ2 (t) + 2φ(t)yq(t) + q(t)y 2
=⇒ y ′ = p(t)y(t) + 2φ(t)y(t)q(t) + q(t)y 2
=⇒ y ′ = (p(t) + 2φ(t)q(t))y(t) + q(t)y 2 (t)

10
2.7 D’Alembert’s Reduction of order
F (t, x(k) , . . . , x(n) ) = 0, 0<k<n
Changes of variable:
y = x(k)
y ′ = x(k+1)
..
.
y (n−k) = x(k+n−k) = x(n)
This implies that F (t, y, . . . , y (n−k) ) = 0, y(t) solution. Then,
ˆ t ˆ t
(k) (k)
x (t) = y(t) =⇒ x (τ ) dτ = y(τ ) dτ
t0 t0
ˆ t
(k−1) (k−1)
=⇒ x (t) − x (t0 ) = y(τ ) dτ
t0
ˆ t
(k−1)
=⇒ x (t) = c1 + y(τ ) dτ
t0
ˆ t ˆ t ˆ s 1 
(k−1)
=⇒ x (τ ) dτ = c1 (t − t0 ) + y(τ ) dτ dτ1
t0 t0 t0
ˆ t ˆ s1
(k−2)
=⇒ x (t) = c1 (t − t0 ) + c2 + y(τ ) dτ dτ1
t0 t0
ˆ t ˆ s1 ˆ s2
(k−3) c1 (t − t0 )2
=⇒ x (t) = + c2 (t − t0 ) + c3 + y(τ ) dτ dτ1 dτ2
2 t0 t0 t0
k ˆ t ˆ s1 ˆ sk−1
X (t − t0 )i
=⇒ x(t) = ck−i + ... y(τ ) dτ dτ1 . . . dτk−1
i=0
i! t0 t0 t0

3 Third Lecture
3.1 Lagrange’s equations
x = tφ(x′ ) + ψ(x′ )
C1
where φ, ψ : I −→ R, φ(p) − p ̸= 0, ∀p ∈ I

If x : J −→ R is solution, then

x(t) = tφ(x′ (t)) + ψ(x′ (t)) =⇒ x′ (t) = φ(x′ (t)) + tφ′ (x′ (t))x′′ (t) + ψ ′ (x′ (t))x′′ (t)

Let p(t) := x′ (t), t ∈ J

p(t) = φ(p(t)) + tφ′ (p(t))p′ (t) + ψ ′ (p(t))p′ (t)

This implies that

p(t) − φ(p(t))
p(t) − φ(p(t)) = [tφ′ (p(t)) + ψ ′ (p(t))]p′ (t) =⇒ p′ (t) =
tφ′ (p(t))+ ψ ′ (p(t))

We see that p(t) − φ(p(t)) ̸= 0. So, [tφ′ (p(t)) + ψ ′ (p(t))]p′ (t) ̸= 0. Then, p′ (t) ̸= 0, ∀t ∈ J.

11
We write t = t(p), p ∈ p(J) and

dt 1
=
dp dp
dt
tφ′ (p(t)) + ψ ′ (p(t))
=
p(t) − φ(p(t))
φ′ (p(t)) ψ ′ (p(t))
= t+ , p ∈ p(J)
p(t) − φ(p(t)) p(t) − φ(p(t))

We obtain that t = A(p, c), p ∈ p(J), c ∈ R


t = A(p, c)
x = tφ(p) + ψ(p)

3.2 Clairaut’s Equations

x = tx′ + ψ(x′ )
Differentiating the above equality, we deduce that

x′ = x′ + tx′′ + ψ ′ (x′ )x′′

and thus
x′′ (t + ψ ′ (x′ )) = 0
We distinguish two types of solutions. The first type is defined by the equation x′′ = 0 Hence,

x = C1 t + C 2

where C1 and C2 are arbitrary constants. We see that C1 and C2 are not independent but are
related by the equality
C2 = ψ(C1 )
Therefore,
x = C1 t + ψ(C1 )
whereC1 is an arbitrary constant. This is the general solution of the Clairaut equation.

3.3 Reduction of order

We discuss here several classes of higher order ODEs that can be reduced to lower order ODEs
using elementary changes of variables. One first example is supplied by equations of the form

F (t, x(k) , . . . , x(n) ) = 0 (∗)

where 0 < k < n. Using the substitution y := x(k) , we reduce (∗) to

F (t, y, . . . , y (n−k) ) = 0

Exercise 1.
x′′ + sin(t)x′ + t = 0
Let y = x′ , y ′ = x′′ .
y ′ + sin(t)t + t = 0

12
3.4 Integral Inequalities
x′ = f (t, x), (t, x) ∈ Ω ⊆ R × R
x : I −→ R is solution if
ˆ t
′
x (t) = f (t, x(t)) ⇐⇒ x(t) = x0 + f (s, x(s)) ds
t0

x′ = ψ(t)x =⇒ x′ − ψ(t)x = 0
 ˆ t 
exp − ψ(τ ) dτ (x′ − ψ(t)x) = 0
t0
  ˆ t 
d
x(t) · exp − ψ(τ ) dτ =0
dx t0
 ˆ t  ˆ t 
x(t) exp − ψ(τ ) dτ − x0 = 0 =⇒ x(t) = x0 exp ψ(τ ) dτ
t0 t0
But also, ˆ t
x(t) = x0 + ψ(s)x(s) ds
t0
What if ˆ t
x(t) ≤ x0 + ψ(s)x(s) ds
t0
 ˆ t 
Then, x(t) ≤ x0 exp − ψ(τ ) dτ , if ψ ≥ 0
t0

Gronwall’s Inequality
Let x, φ, ψ : [a, b] −→ R, continuous functions such that ψ ≥ 0 and
ˆ t
x(t) ≤ φ(t) + ψ(s)x(s) ds
a
ˆ t
Let y(t) := ψ(s)x(s) ds
a

y ′ (t) = ψ(t)x(t) ≤ ψ(t)(φ(t) + y(t)) ≤ ψ(t)φ(t) + ψ(t)y(t)

=⇒ y ′ (t) − ψ(t)y(t) ≤ ψ(t)φ(t)

 ˆ t   ˆ t 
′
=⇒ exp − ψ(τ ) dτ [y (t) − ψ(t)y(t)] ≤ exp − ψ(τ ) dτ ψ(t)φ(t)
a a
  ˆ t   ˆ t 
d
=⇒ y(t) exp − ψ(τ ) dτ ≤ exp − ψ(τ ) dτ ψ(t)φ(t)
dx a a
 ˆ t  ˆ t  ˆ t 
=⇒ y(t) exp − ψ(τ ) dτ ≤ exp − ψ(τ ) dτ ψ(s)φ(s) ds
a a s

Corollary 1. If x, ψ : [a, b] −→ R are continuous functions such that

ˆ t
x(t) ≤ M + ψ(s)x(s) ds
a

Then ˆ t 
x(t) ≤ M exp ψ(s) ds
a

13
 Bihari Inequality

x : [a, b] −→ R, continuous
ψ : [a, b] −→ R≥0 , continuous
w : R −→ (0, +∞), continuous
w non-decreasing
ˆ t
x(t) ≤ C + ψ(s)w(x(s)) ds
a
ˆ t
Let y(t) = ψ(s)w(x(s)) ds.
a

y ′ (t) = ψ(t)w(x(t)) ≤ ψ(t)w(C + y(t))

y ′ (t)
=⇒ ≤ ψ(t)
w(C + y(t))
ˆ t ˆ t
y ′ (τ )
=⇒ dτ ≤ ψ(τ ) dτ
a w(C + y(τ )) a
ˆ C+y(t) ˆ t
dτ
=⇒ ≤ ψ(τ ) dτ
a w(τ ) a
ˆ u
ds 1
Now, we set Φ(u) := , Φ′ (u) = . Then,
c w(s) w(s)
ˆ t  ˆ t   ˆ t 
−1 −1
Φ(C+y(t)) ≤ ψ(s) ds =⇒ y(t)+C ≤ Φ C+ ψ(s) ds =⇒ x(t) ≤ Φ C+ ψ(s) ds
a a a

4 Fourth Lecture
4.1 Existence and Uniqueness for the Cauchy Problem
Vector notation
x = (x1 , . . . , xn ) ∈ Rn
x′ = (x′1 , . . . , x′n ) ∈ Rn
f : Ω ⊆ R × Rn −→ Rn
x′ = f (t, x), x(t0 ) = x0 (∗)
where x0 = (x1,0 , x2,0 , . . . , xn,0 ) is equivalente to the system

x′i = fi (t, x1 , . . . , xn ), xi (t0 ) = xi,0

where fi are the component functions of f . We are going to use the maximum on R × Rn

∥(t, x)∥ = max{|t|, |x1 |, . . . , |xn |}

A solution to (∗) is a curve x : I −→ R such that

1. (t, x(t)) ∈ Ω, ∀t ∈ I

2. x′ (t) exists, ∀t ∈ I

14
 3. x′ (t) = f (t, x(t)), ∀t ∈ I

Hypothesis:

• f : ∆ ⊆ R × Rn −→ Rn , where ∆ := {(t, x) : |t − t0 | ≤ a, ∥x − x0 ∥ ≤ b}

• f is continuous on ∆

• f is lipschitz on the second, ∃L > 0 such that

∥f (t, y) − f (t, x)∥ ≤ L∥x − x0 ∥, ∀(t, x), (t, y) ∈ ∆

Result: There is a unique solution x : [t0 − δ, t0 + δ] −→ Rn to the Cauchy problem

x′ = f (t, x), x(t0 ) = x0

 
b
where δ = min a, , M := sup ∥f (t, x)∥
M (t,x)∈∆

Proof.

Step 1: Elementary remark

ˆ t
n ′
x : I −→ R is a solution of x = f (t, x) iff x(t) = x0 + f (s, x(s)) ds
t0

Step 2: (∀t ∈ [t0 − δ, t0 + δ])

x0 (t) ≡ x0
ˆ t
xn+1 (t) = x0 + f (s, xn (s)) ds
t0

Having proved that xn (t) is well-defined, note that

ˆ t
∥xn+1 − x0 ∥ = f (s, xn (s)) ds
t0
ˆ t
≤ ∥f (s, xn (s))∥
t0
ˆ t
≤ M ds
t0
≤ |t − t0 |M
≤δ·M
b
≤ ·M
M
≤b

Step 3: (xn (t))n is uniformly convergent

∗ ∥x1 (t) − x0 (t)∥ = ∥x1 (t) − x0 ∥

ˆ t
= f (s, x0 ) ds
t0
≤ |t − t0 |M

15
 ˆ t
∗ ∥x2 (t) − x1 (t)∥ = [f (s, x1 (s)) − f (s, x0 (s))] ds
t0
ˆ t
≤ L∥x1 (s) − x0 (s)∥ ds
t0
ˆ t
≤L (s − t0 )M ds
t0
|t − t0 |2
≤L M
2

|t − t0 |3
∗ ∥x3 (t) − x2 (t)∥ ≤ L2 M
3·2
..
.

|t − t0 |n+1 Ln δ n+1 M
∗ ∥xn+1 (t) − xn (t)∥ ≤ Ln M≤
(n + 1)! (n + 1)!
X
(xn (t))n converges uniformly iff (xn+1 (t) − xn (t)) does. But,
n≥1

X X Ln δ n+1 M
∥xn+1 (t) − xn (t)∥ ≤
n≥1 n≥1
(n + 1)!

Notice that
Ln+1 δ n+2 M
(n + 2)! Lδ
n n+1 = −→ 0
L δ M n+2
(n + 1)!
X
Then, by Weierstrass M-test, (xn+1 (t) − xn (t)) converges uniformly, so (xn (t))n converges
n≥1
uniformly.

Step 4: x(t) := lim xn (t) is the solution of x′ = f (t, x), x(t0 ) = x0 . Indeed,
n→+∞
ˆ t
x(t) = lim xn+1 (t) = x0 + lim f (s, xn (s)) ds
n→+∞ n→+∞ t
ˆ t 0

= x0 + f (s, lim xn (s)) ds

t0 n→+∞
ˆ t
= x0 + f (s, x(s)) ds
t0

Step 5: x(t) is unique. Let y : [t0 − δ, t0 + δ] −→ Rn be a solution of x′ = f (t, x), x(t0 ) = x0 .

For t ∈ [t0 , t0 + δ]:

ˆ t
∥x(t) − y(t)∥ = [f (s, x(s) − f (s, y(s))] ds
t0
ˆ t
≤L ∥x(s) − y(s)∥
t0

16
By Gronwall’s Lemma: ˆ t 
∥x(t) − y(t)∥ ≤ 0 · exp L ds = 0
t0
This implies that x(t) = y(t), ∀t ∈ [t0 , t0 + δ] Now,
∆˜ = {(t, x) ∈ R × Rn : |t + t0 | ≤ a, ∥x − x0 ∥ ≤ b}
f˜ : ∆
˜ −→ Rn , f˜(t, x) = −f (−t, x)
x̃(t) : [−t0 − δ, −t0 + δ] −→ Rn , x̃(t) = x(−t)
x̃ (t) = −x̃ (t) = −f (−t, −x(−t)) = f˜(t, x̃(t))
′ ′

If ỹ(t) = y(−t), t ∈ [−t0 − δ, −t0 + δ]. We also have,

ỹ ′ (t) = f˜(t, ỹ(t))
By unicity on [−t0 , −t0 + δ],
x̃(t) = ỹ(t), on [−t0 , −t0 + δ]

=⇒ x(−t) = y(−y), on [−t0 , −t0 + δ]

=⇒ x(t) = y(t), on [−t0 , −t0 + δ]

Example 7.
x′ = tx = f (t, x), x(0) = 1
x0 (t) ≡ 1
ˆ t
x1 (t) = 1 + s · 1 ds
0
t2
=1+
2
ˆ t 
s2

x2 (t) = 1 + s· 1+ ds
0 2
t2 t4
=1+ +
2 4·2
2
t t4 t6
x3 (t) = 1 + + +
2 4·2 6·4·2
..
.
t2 t2n
xn (t) = 1 + + ... +
2 2n!
+∞ 2n
X t 2
x(t) = = et /2
n=1
2n!
Example 8.
x′ = x2/3
x(t) ≡ 0 is solution, satisfying x(0) = 0
ˆ ˆ
dx
= dt + c
x2/3
3x1/3 = t + c
 3  3
t+c t+c t c
x1/3 = =⇒ x = = +
3 3 3 3

17
5 Fifth Lecture
5.1 Existence and Uniqueness for the Cauchy Problem
• f : ∆ −→ Rn , where ∆ = {(t, x) : |t − t0 | ≤ a, ∥x − x0 ∥ ≤ b}

• f is continuous on ∆

• f is lipschitz on the second variable

 
b
• δ = min a, , there is a unique solution x : [t0 − δ, x0 + δ] −→ Rn to the Cauchy
M
Problem
x′ = f (t, x), x(t0 ) = x0

Example 9.
tx′ − x = 0, x(0) = 0
x(t) = mt is solution for every m ∈ R
Example 10.
x′ = sgn(t), x(0) = c
Suppose that x : [−ε, ε] −→ R is solution. Then, on (0, ε], x(t) = t+c1 ; on [−ε, 0), x(t) = −t+c2

 t + c1 , t ∈ (0, ε]
x(t) = c , t=0
−t + c2 , t ∈ [−ε, 0)


Since x is continuous, then c1 = c = c2 .

Example 11. Suppose that f : R × Rn −→ Rn is locally lipschitz on the second variable. For
every compact X ⊆ R × Rn , exists LK > 0 such that

∥f (t, x) − f (t, y)∥ ≤ LK ∥x − y∥, ∀(t, x), (t, y) ∈ X

and suppose that globally bounded, ∃M > 0 such that ∥f (t, x)∥ ≤ M , ∀(t, x) ∈ R × Rn . For
instance, f (t, x) = sin(tx). Then, there is a unique solution on R to the Cauchy Problem

x′ = f (t, x), x(t0 ) = x0 , ∀(t0 , x0 ) ∈ R × R

Solution. Consider the set

∆m = {(t, x) : |t − t0 | ≤ m, ∥x − x0 ∥ ≤ m}

Then, f |∆m satisfies the hypothesis of Picard’s theorem on ∆m. Therefore, there is a unique
solution
xm (t) : [t0 − δm , t0 + δm ] −→ Rn
to the Cauchy Problem
x′ = (f |∆m )(t, x), x(t0 ) = x0
 m
where δm = min m, = m, if m ≥ M .
M
Let x : R −→ Rn the function

x(t) = xm (t), if t ∈ [t0 − δm , t0 + δm ]

18
 x(n) = g(t, x1 , . . . , xn ), g : ∆ ⊆ R × Rn −→ Rn (∗)
Suppose g is continuous, lipschitz on the second variable, then there is a solution to (∗),
x : [t0 − δ, x0 + δ] −→ R
x(t0 ) = x1,0
x′ (t0 ) = x2,0
x(n−1) (t0 ) = xn,0
 
b
where δ = min a, , M := max {|x1 |, |x2 |, . . . , |g(t, x)|} , (t, x) ∈ ∆
M
Proof. Let G : ∆ ⊆ R × Rn −→ Rn be given by

G(t, x) = (x2 , x3 , . . . , g(t, x1 , . . . , xn )), x = (x1 , . . . , xn )

x1 := x =⇒ x′1 = x2
x2 := x′ =⇒ x′2 = x3
..
.
xn := x(n−1) =⇒ x′n = g(t, x1 , . . . , xn )
Note that, x = (x1 , . . . , xn ) and y = (y1 , . . . , yn )

∥G(t, x) − G(t, y)∥ = ∥(x2 , x3 , . . . , g(t, x)) − (y2 , y3 , . . . , g(t, y))∥

= ∥(x2 − y2 , x3 − y3 , . . . , g(t, x) − g(t, y))
≤ max{|x2 − y2 |, |x3 − y3 |, . . . , |g(t, x) − g(t, y)|}
≤ max{|x2 − y2 |, |x3 − y3 |, . . . , L∥x − y∥}
≤ max{1, L}∥x − y∥
 
b
Therefore, there is a unique solution x : [t0 − δ, t0 + δ] −→ R, δ = min a, , m =
m
max{G(t, x)} = max{|x2 |, |x3 |, . . . , |g(t, x)|} to x′ = G(t, x), x(t0 ) = x0 . The wanted solu-
tion is x(t) defined by x(t) = (x1 (t), x2 (t), . . . , xn (t))

5.2 Peano Theorem

Let f : ∆ −→ Rn be continuous, where ∆ = {(t, x) : |t − t0 | ≤ a, ∥x − x0 ∥ ≤ b} Then, there
is a solution to the Cauchy Problem

x′ = f (t, x), x(t0 ) = x0

Proof.
P = {t0 , t1 , . . . , tn }
t0 = t0
δ
t1 = t0 +
n
2δ
t2 = t0 +
n
tn = t0 + δ
Step 1: Construction of a polygonal functional

19
For n ∈ N, let φn : [t0 , t0 + δ] −→ Rn given by


 x0 , t = t0
φn (t0 ) + (t − t0 )f (t0 , φn (t0 )) , t0 ≤ t ≤ t1


φn (t) = ..


 .
 φ (t ) + (t − t )f (t , φ (t ))
n i−1 i−1 i−1 n i−1

Step 2: φn is well-defined

If t = t0 , ∥φn (t) − x0 ∥ = ∥x0 − x0 ∥ = 0.

If t0 ≤ t ≤ t1 ,
δ
∥φ( t) − x0 ∥ = ∥φn (t0 ) + (t − t0 )f (t0 , φn (t0 )) − x0 ∥ ≤ |t − t0 |∥f (t0 , φn (t0 ))∥ ≤ M ≤ δM ≤ b
n
Having proved that
iδ
∥φn (t) − x0 ∥ ≤ M, ti−1 ≤ t ≤ ti
n
Then, for ti ≤ t ≤ ti+1 , we have

∥φn (t) − x0 ∥ ≤ ∥φn (ti ) + (t − ti )f (ti , φn (ti )) − x0 ∥

≤ ∥φn (ti ) − x0 ∥ + |t − ti |∥f (ti , φn (ti ))∥
iδ δ
≤ M+ M
n n
(i + 1)
= δM ≤ b
n
Step 3: φn is uniformly continuous

∥φn (t) − φn (s)∥ ≤ M |t − s|, ∀s, t

For s, t ∈ [t0 , t1 ]

∥φn (t) − φn (s)∥ = ∥φ(t0 ) + (t − t0 )f (t0 , φ(t0 )) − φ(t0 ) − (s − t0 )f (t0 , φ(t0 ))∥
= |t − s|∥f (t0 , φ(t0 ))∥ ≤ M |t − s|

Similarly, s, t ∈ (ti , ti+1 ]

∥φn (t) − φn (s)∥ ≤ M |t − s|
For t ∈ (ti−1 , ti ] (i = 1, t ∈ [t0 , t1 ])

For s ∈ (tj−1 , tj ]

∥φn (t) − φn (s)∥ = ∥φn (t) − φn (ti ) + φn (ti ) − φn (ti+1 ) + . . . + φn (tj−1 ) − φn (s)∥
≤ ∥φn (t) − φn (ti )∥ + ∥φn (ti ) − φn (ti+1 )∥ + . . . + ∥φn (tj−1 ) − φn (s)∥
≤ (ti − t)M + (ti+1 − ti )M + . . . + (s − tj−1 )M
≤ M (ti − t + ti+1 − ti + . . . + s − tj−1 )
≤ M (s − t)
≤ M |t − s|

Step 4: (φn (t))n has a uniformly convergent subsequence.

20
φn (t) is uniformly bounded
∥φn (t) − x0 ∥ ≤ b =⇒ ∥φn (t)∥ ≤ b + ∥x0 ∥
(φn (t))n is equicontinuous
∥φn (t) − φn (s)∥ ≤ M |t − s|
By Arzela-Ascoli, there is a uniform subsequence
φnk (t) −→ φ(t), k → +∞
Step 5: φ(t) is solution. Let
φ′nk (t) − f (t, φnk (t)) , t ̸= tni k

gk (t) =
0 , t = tni k
gk (t) is Riemann-Integrable
ˆ t ˆ t
gk (s) ds = φ′nk (s) − f (s, φnk (s)) ds
t0 t0
ˆ t
= φnk (t) − φnk (t0 ) − f (s, φnk (s)) ds
t0

This implies that ˆ ˆ

t t
φnk (t) = x0 + f (s, φnk (s)) ds + gk (s) ds
t0 t0
ˆ t
Let us estimate gk (s) ds .
t0
ˆ t ˆ t
gk (s) ds ≤ ∥gk (s)∥ ds
t0 t0

t = tni k

0 ,
∥gk (t)∥ =
∥f (ti−1 , φnk (ti−1 )) − f (t, φnk (t))∥ , ti−1 ≤ t ≤ tni+1
nk k

Let ε > 0 be given. Since, f is uniformly continuous, there is β > 0 such that
∥(t, x) − (t, y)∥ ≤ β =⇒ ∥f (t, x) − f (t, y)∥ < ε
 
δ β
Let N ∈ N such that k ≥ N , then ≤ min β, Then, if k ≥ N
nk m
δ
|t − ti−1 | ≤ ≤β (ti−1 < t < ti+1 )
nk
δ β
and ∥φnk (ti−1 ) − φnk (t)∥ ≤ M |t − ti−1 | ≤ M ≤M = β Therefore, ∥gk (t)∥ ≤ ε, ∀k ≥ N
nk M
and ˆ t ˆ t
δ
gk (s) ds ≤ ∥gk (t)∥ ds ≤ |t − t0 |ε ≤ ε ≤ δε
t0 t0 nk
Finally, ˆ ˆ
t t
φnk (t) − x0 − f (s, φnk (s)) ds ≤ gk (s) ds ≤ δε
t0 t0
So, as ε → 0 ˆ t
φ(t) = x0 + f (s, φ(s)) ds
t0

21
6 Sixth Lecture
6.1 Global existence theorem
Setting: f : Ω ⊆ R × Rn −→ Rn

• Ω is open

• f is continuous

• f is lipschitz on the second variable:

∀K ⊆ Ω, compact, ∃LK > 0 such that ∥f (t, x)−f (t, y)∥ ≤ Lk ∥x−y∥, ∀(t, x), (t, y) ∈ K

Theorem 1. The Cauchy Problem

x′ = f (t, x), x(t0 ) = x0

has a solution x(t; t0 , x0 ) defined on some closed interval [t0 − δ, t0 + δ]

Proof. Let ∆ = {(t, x) ∈ Ω : |t − t0 | ≤ a,  ∥x −x0 ∥ ≤ b}. Then, ∆ is compact. Then,
b
f : ∆ −→ Rn is lipschitz on x. If δ = min a, where M = sup ∥f (t, x)∥ . Then, by
M (t,x)∈∆
the Picard’s Theorem on ∆, there is a solution x(t; t0 , x0 ) : [t0 − δ, t0 + δ] −→ Rn such that
x(t0 ) = x0 .

Theorem 2. If x : I −→ Rn , y : J −→ Rn are two solutions of x′ = f (t, x), satisfying that

x(t0 ) = y(t0 ), for t0 ∈ I ∩ J, then x(t) = y(t), ∀t ∈ I ∩ J.
Proof. Let (t1 , t2 ) = I∩J. Let prove that x(t) = y(t), ∀t ∈ [t0 , t2 ). Let T := sup {t ∈ [t0 , t2 ) : x(t) = y(t)}.
Claim 1. T = t2 . Suppose that T < t2 . Then, x(t) = y(t) on [t0 , T ). We also have that
x(T ) = y(T ) by continuity. Therefore x(t) = y(t) on [t0 , T ]
Proof. Let x1 := x(T ) = y(T ). Consider the following Cauchy Problem

x′ = f (t, x), x(T ) = x1 (6)

Let ∆ = {(t, x) : |t − T | ≤ a, ∥x − x0 ∥ ≤ b} ⊆ Ω. By Picard’s Theorem, there is δ > 0

and a function z : [T, T + δ] ⊆ [T, t2 ) solution of (6). But, x, y are also solutions of (6), then
x(t) = y(t) = z(t), ∀t ∈ [T, T + δ] contradiction. Therefore, T = t2 .

A solution φ : [a, b] −→ Rn of x′ = f (t, x) is right-extendible if there c > b and a solution

ψ : [a, c] −→ R such that
ψ|[a,b] = φ
Left-extendible is defined analogously.

A right-satured solution φ : I −→ Rn is one that is not right-extendible. A left-satured solution

φ : I −→ Rn is one that is not left-extendible.

Proposition 1. If φ : I −→ Rn is right-satured, then I is open on the right.

22