Differential Equations (MA 1150)

Sukumar

Lecture 5

April 25, 2020

Overview

Bernoulli Equation

Orthogonal Trajectories

Existence and uniqueness of a solution

Linear ODE
Non-linear ODE

Picard’s Iteration
 Section 1

Bernoulli Equation
Transforming Non-Linear into Separable ODE

A non-linear differential equation of the form

y 0 + p(x)y = f (x)y r .

where r ∈ R \ {0, 1} is said to be a Bernoulli Equation.

For r = 0, 1, it is linear ODE.
If y1 is a non-zero solution of y 0 + p(x)y = 0, then putting y = u(x)y1 in
ODE, we get

u 0 y1 + uy10 + puy1 = fu r y1r

u0
=⇒ = f (x)y1 (x)r −1
ur
On integrating, we get

u −r +1
Z
= f (x)(y1 (x))r −1 dx + c.
−r + 1
Transforming Non-Linear into Separable ODE

A non-linear differential equation of the form

y 0 + p(x)y = f (x)y r .

where r ∈ R \ {0, 1} is said to be a Bernoulli Equation.

For r = 0, 1, it is linear ODE.
If y1 is a non-zero solution of y 0 + p(x)y = 0, then putting y = u(x)y1 in
ODE, we get

u 0 y1 + uy10 + puy1 = fu r y1r

=⇒ u 0 y1 = fu r y1r
u0
=⇒ = f (x)y1 (x)r −1
ur
On integrating, we get

u −r +1
Z
= f (x)(y1 (x))r −1 dx + c.
−r + 1
Example of Bernoulli Equation

Consider a non-linear ODE

y 0 + y = xy 2 .
Set y = u(x)e −x , where y1 = e −x is a solution of homogeneous part.
Putting y = u(x)e −x in ODE, we get
u 0 e −x − ue −x + ue −x = u 2 e −2x x
=⇒ u 0 e −x = u 2 e −2x x
u0
=⇒ = xe −x
u2

On integrating, we get
−1
= −(1 + x)e −x + c
u
1
=⇒ u=
(1 + x)e −x − c
Thus
1
y = u(x)e −x =
(1 + x) − ce x
Example (Bernoulli Equation)

Consider a (Bernoulli) ODE:

x2
xy 0 − 2y = .
y6
Rewrite it as
2 x
y0 − y = 6.
x y
The solution to homogeneous part is y1 = x 2 .

Setting y = u(x)y1 , we get

u 0 y1 = x(uy1 )−6
=⇒ u 6 u 0 = x −13  
1 7 1 1 7 1 −12
=⇒ u = − x −12 + c. =⇒ y = − x + c y17
7 12 7 12
y
Since u = and y1 = x 2 , we get
y1
 
7 14 1 −12
y = 7x − x +c .
12
 Section 2

Orthogonal Trajectories
Curves in Rn

Definition. A curve is a continuous map f : I 7→ Rn , where I ⊆ R is an interval.

Definition. The image set C = f (I ) ⊆ Rn is called the trace of the curve.
Physically, a curve describes the motion of a particle in n-space, and the trace
is the trajectory of the particle.

Orthogonal Trajectories Two curves C1 and C2 are said to be orthogonal at a

point of intersection (x0 , y0 ) if they have perpendicular tangents at (x0 , y0 ).

A curve is said to be an orthogonal trajectory of a given family of curves if it is

orthogonal to every curve in the family.
One-Parameter Families of Curves

Example. For each value of the parameter c ∈ R, the equation y = cx 2 defines

a curve in R2 .

Question. Find a differential equation for the family of curves defined by

y = cx 2 .

y0
Differentiating y = cx 2 w.r.t. x gives y 0 = 2cx. From this, we get c = .
2x
Now eliminate the parameter using the original equation for the family of
curves and finally obtain
xy 0
y= .
2
Orthogonal trajectory of a given one parameter families of curves

Step 1. Find a differential equation

y 0 = f (x, y )

for the given family.

Step 2. Solve the differential equation

1
y0 = −
f (x, y )

to find the orthogonal trajectories.

2y
Example. Notice that differential equation y 0 = represents one parameter
x
families of curves defined by y = cx 2 .

x
Solve y 0 = − . Then y =??.
2y
 Section 3

Existence and uniqueness of a solution

Subsection 1

Linear ODE
Existence and Uniqueness Theorem for Linear Nonhomogeneous First Order
Equations

Theorem: Suppose p(x) and f (x) are continuous functions on an open interval
(a, b), and let y1 be any nontrivial solution of the complementary equation
y 0 + p(x)y = 0
on (a, b). Then:
(i) The general solution of the non-homogeneous equation
y 0 + p(x)y = f (x) (1)
on interval (a, b) is
 Z 
y = y1 (x) c + f (x)/y1 (x) dx . (2)

(ii) The initial value problem y 0 + p(x)y = f (x), y (x0 ) = y0 has the
following unique solution on (a, b),
 Z x 
y0 f (t)
y = y1 (x) + dt .
y1 (x0 ) x0 y1 (t)
Proof of Existence and Uniqueness Theorem

Proof of (i). We shall show that if y is a solution of (1) on (a, b) then y is of

the form (2) for some constant c.
Suppose y = uy1 is a solution of (1) on (a, b) for some u.
We know that y1 has no zeros on (a, b) (why?), so the function u = y /y1 is
defined on (a, b). Moreover, since

y 0 = −py + f and y10 = −py1 ,

y1 y 0 − y10 y
u0 =
y12

y1 (−py + f ) − (−py1 )y f
= = .
y12 y1

Integrating u 0 = f /y1 , we get

 Z 
u = c + f (x)/y1 (x) dx .
Proof of Existence and Uniqueness Theorem

R
Proof of (ii). We have seen in the proof of (i) that f (x)/y1 (x) dx in (2) is an
arbitrary antiderivative of f /y1 .
Now we may choose the antiderivative that equals zero when x = x0 , and so
 Z x0 
f (t)
y (x0 ) = y1 (x0 ) c + dt = cy1 (x0 ),
x0 y1 (t)

y0
we see that y (x0 ) = y0 if and only if c = .
y1 (x0 )
Therefore unique solution on (a, b) is
 Z x 
y0 f (t)
y = y1 (x) + dt .
y1 (x0 ) x0 y1 (t)
Existence and Uniqueness Theorem for 1st order linear ODE

Problems:
For the following IVP:
(a) y 0 = sin x y (x0 ) = y0 .
(b) y 0 = x sin(1/x) y (x0 ) = y0 for all x > 0.
(c) y 0 = x + y y (x0 ) = y0 .
Using Existence and Uniqueness theorem, Justify whether the following IVP
has
I at least one solution on some interval containing x0 ,
I has a unique solution on some interval containing x0 .
 Subsection 2

Non-linear ODE
Existence and Uniqueness of solutions of non-linear ODE

Consider a non-linear ODE: y 0 = f (x, y ).

Theorem Let D = (a, b) × (c, d) be an open rectangle containing the point

(x0 , y0 ) and consider the IVP

y 0 = f (x, y ), where y (x0 ) = y0 .

(a) (Existence) Assume f (x, y ) is continuous on D. Then IVP has at least one
solution on some interval (a1 , b1 ) ⊂ (a, b) containing x0 .
∂f
(b) (Uniqueness) If both f (x, y ) and ∂y are continuous on D, then IVP has a
0 0
unique solution on some interval (a , b ) ⊂ (a, b) containing x0 .

Proof. We will not discuss the proof of this.

Linear vs Non-Linear ODE

(1) Note the theorem says that for non-linear ODE, the solution and the
interval where the solution exists, depends on the choice of our initial
condition.

(2) The solution of a non-linear ODE obtained using a particular method may
not be a general solution.

Example 1 For non-linear ODE y 0 = 2xy 2 , our solution y = − x 2 +C

1
does not
give the solution y ≡ 0 for any value of C .
Example 2 The circle x 2 + y 2 = C is an implicit solution of yy 0 = x.
For C = −1, it does not give any solution to ODE, since the curve
x 2 + y 2 = −1 is empty.

The above example shows that unlike linear ODE’s, not every value of C will
give an actual solution.
Example

Consider the IVP

x2 − y2
y0 = , where y (x0 ) = y0 .
1 + x2 + y2
x 2 −y 2
If f (x, y ) = 1+x 2 +y 2 , then

∂f −2y (1 + x 2 )
=??? = .
∂y (1 + x 2 + y 2 )2

∂f
Since f (x, y ) and ∂y are continuous for all (x, y ) ∈ R2 , by existence and
uniqueness theorem, for any (x0 , y0 ) ∈ R2 , IVP has a unique solution on some
open interval containing x0 .
Example

Consider the IVP

x2 − y2
y0 = , where y (x0 ) = y0 . (3)
x2 + y2
x 2 −y 2
Let f (x, y ) = x 2 +y 2 , then

∂f −4x 2 y
=??? = 2 .
∂y (x + y 2 )2
∂f
Notice that f (x, y ) and ∂y are continuous for all (x, y ) ∈ R2 \ (0, 0).
Assume that (x0 , y0 ) 6= (0, 0).

There is an open rectangle R containing (x0 , y0 ) but not containing (0, 0).
∂f
f (x, y ) and ∂y are continuous on R.

By existence and uniqueness theorem, (3) has a unique solution on some open
interval containing x0 .
Example

Consider the IVP

x +y
y0 = , where y (x0 ) = y0 . (4)
x −y
x+y
Let f (x, y ) = x−y , then
∂f 2x
= .
∂y (x − y )2
∂f
Note that f (x, y ) and ∂y are continuous for all (x, y ) ∈ R2 except on the line
y = x.
Assume that x0 6= y0 .

There is an open rectangle R containing (x0 , y0 ) that does not intersect with
the line y = x.
∂f
f (x, y ) and ∂y are continuous on R.

By existence and uniqueness theorem, (4) has a unique solution on some open
interval containing x0 .
Example

Consider the IVP

10 2/5
y0 = xy , where y (x0 ) = y0 . (5)
3
Here
10 2/5 ∂f 4
f (x, y ) = xy , and = xy −3/5 .
3 ∂y 3

I Since f (x, y ) is continuous for all (x, y ) ∈ R2 , IVP (5) has atleast one
solution for all (x0 , y0 ) ∈ R2 .
∂f
I If y 6= 0, then f (x, y ) and ∂y both are continuous for all (x, y ) ∈ R2 .
I If y =
6 0, there is an open rectangle R containing (x0 , y0 ) s.t. f (x, y ) and
∂f
∂y are continuous on R.

I Hence IVP (5) has a unique solution on some open interval containing x0 .
Example

Consider the IVP

10 2/5
y0 = xy , where y (0) = 0. (6)
3
Here ∂f
∂y = 34 xy −3/5 is not continuous if y = 0.

I IVP (6) may have more than one solution on every open interval
containing x0 .
I If y ≡ 0 is one solution of IVP (6).

Let us find a nonzero solution of ODE (6).

y0 10 5 5
= x =⇒ y 3/5 = (x 2 + C ).
y 2/5 3 3 3

This implies that

y (x) = (x 2 + C )5/3 .
Example continued

Note that y (x) = (x 2 + C )5/3 is defined for all (x, y ) ∈ R2 and

5 2 10 3/5
y0 = (x + C )2/3 (2x) = xy , for all(x, y ) ∈ R2 .
3 3
Thus y (x) is a solution on R for all C .

y (0) = 0 =⇒ C = 0.

Thus the IVP

10 2/5
y0 = xy , where y (0) = 0. (7)
3
10
has two solutions, y1 ≡ 0 and y2 (x) = x 3 .
Example

Consider the IVP

10 2/5
y0 = xy , where y (0) = −1. (8)
3
∂f
Here f (x, y ) and ∂y = 43 xy −3/5 are continuous in an open rectangle containing
(0, 1). Hence the IVP has a unique solution on some open interval containing
x0 = 0.
Let us find the unique solution and its interval of validity.
Let y 6= 0 be the solution of y 0 = 10
3 xy
2/5
. Then y (x) = (x 2 + C )5/3 and
y (0) = −1 implies that C = −1.
I Thus y (x) = (x 2 − 1)5/3 is a solution of (8) on (−∞, ∞) (Existence
part).
I If y0 6= 0, then by IVP, y 0 = 10
3 xy
2/5
, where y (x0 ) = y0 . has a unique
solution on some open interval around x0 .
I Let us check that y (x) is the unique solution to the IVP on the interval
(1, 1).
Example

I Suppose there is another solution w (x) to the IVP. Then by the existence
and uniqueness theorem, y (x) = w (x) for all x in a neighbourhood (a, b)
around 0. Choose a and b so that this interval is the largest.
I If < a, then since both y (x) and w (x) are continuous, we get
y (a) = w (a) = A.
I Now we apply the existence and uniqueness theorem to the IVP with the
condition y (a) = A. This will show that y (x) = w (x) in a neighbourhood
around a, which contradicts the assumption that (a, b) was the largest
interval on which y and w agreed.

If we take any interval (a, b) with a << 1 < b then, we can define another
solution (
(x 2 − 1)5/3 , if − 1 ≤ x ≤ 1,
y1 (x) =
0, if |x| > 1.
 Section 4

Picard’s Iteration
Picard’s Iteration Method
Picard’s Iteration Method

Assume that f (x, y ) is a continuous function on R2 . Consider the IVP:

y 0 = f (x, y ), y (x0 ) = y0 .

This is equivalent to the integral equation

Z x
y (x) = y0 + f (s, y (s)) ds
x0

A solution to the IVP ODE is equivalent to a solution to the integral equation

and vice versa.
Motivation for the Picard’s Iteration Method

Taylor series solution to the IVP ODE.

Define
y1 (x) = y0 ,

y2 (x) = y1 (x) + y 0 (x0 )(x − x0 ) = y0 + f (x0 , y0 )(x − x0 ).

Similarly,
y 00 (x0 )
y3 (x) = y2 (x) + (x − x0 )2 .
2
The (n + 1)th iterate is

y (n) (x0 )
yn+1 (x) = yn (x) + (x − x0 )n .
n!
Picard’s Iteration Method

Define
φ1 (x) = y0 .

Z x Z x
φ2 (x) = φ1 (x) + f (s, φ1 (s)) ds = y0 + f (s, y0 ) ds
x0 x0

Next, Z x
φ3 (x) = y0 + f (s, φ2 (s)) ds.
x0

Similarly, we define the (n + 1)th iterate as below

Z x
φn+1 (x) = y0 + f (s, φn (s)) ds.
x0
Picard’s Iteration Method

Note. Each φn satisfies the initial condition φn (x0 ) = y0 .

Suppose for some n, φn+1 = φn . Then

Z t
φn+1 (t) = φn (t) = y0 + f (s, φn (s)) ds
0

and this implies

d
(φn (t)) = f (t, φn (t))
dt

is a solution of the given IVP.

Picard’s Iteration Method

In general, the sequence {φn } may not terminate. In fact, all the φn may not
even be defined outside a small region in the domain.
∂f
However, it is possible to show that, if f (x, y ) and
is continuous in some
∂y
open rectangle (hence continuous and bounded in a smaller closed rectangle),
the sequence converges to a function

φ(t) = lim φn (t)

n→∞

which will be the unique solution to the given IVP.

Picard’s Iteration Method with Example

Example. Solve the IVP:

y 0 = xy ; y (−1) = 1.

Let φ1 (x) = 1 since φ1 (−1) = 1. Then,

x
1 x2
Z
φ2 (x) = 1 + sφ1 (s) ds = + .
−1 2 2

x x
1 s2 5 x2 x4
Z Z  
φ3 (x) = 1 + sφ2 (s) ds = 1 + s + ds = + +
−1 −1 2 2 8 4 8

Similarly compute higher φi (x) for i ≥ 4.

Picard’s Iteration Method - Example

Example. Solve the IVP:

y 0 = 2t(1 + y ); y (0) = 0.

The corresponding integral equation is

Z t
φ(t) = 2s(1 + φ(s)) ds
0

Let φ1 (t) = 0 since y (0) = 0. Then,

Z t
φ2 (t) = 2s ds = t 2 ,
0

t
t4
Z
φ3 (t) = 2s(1 + s 2 ) ds = t 2 + ,
0 2
t
s4 t4 t6
Z
φ4 (t) = 2s(1 + s 2 + ) ds = t 2 + + .
0 2 2 6
Picard’s Iteration Method - Example

Example (continued...) We claim

t4 t6 t 2(n−1)
φn (t) = t 2 + + + ··· + .
2 6 (n − 1)

Use induction to prove this:

Z t
φn+1 (t) = 2s(1 + φn (s)) ds
0

t
s4 s6 s 2(n−1)
Z  
2
= 2s 1 + s + + + ··· + ds,
0 2 6 (n − 1)

t4 t6 t 2n
= t2 + + + ··· + .
2 6 n!

P∞ t 2k
Hence φn+1 (t) is the n-th partial sum of the series k=1 .
k!
Picard’s Iteration Method - Example

What about convergence!!!

Applying the ratio test, we get

 2k+2
t2

 t k! 
 (k + 1)! t 2k  = k + 1 → 0
·


for all t as k → ∞. Thus,

∞ 2k
X t 2
lim φn (t) = = e t − 1.
k→∞ k!
k=1
Uniqueness (Not in course)

Let’s quickly see how to get uniqueness. Suppose φ and ψ are solutions of

y 0 = f (x, y ), y (0) = 0.

Thus, both these satisfy the integral equation as well. Then,

Z t
φ(t) − ψ(t) = (f (s, φ(s)) − f (s, ψ(s))) ds.
0

Thus
Z t
|φ(t) − ψ(t)| ≤ |f (s, φ(s)) − f (s, ψ(s)| ds. (9)
0

∂f
Since f and both are continuous on some smaller rectangle R, there is a
∂y
constant M, such that

|f (s, φ(s)) − f (s, ψ(s)| ≤ M |φ(s) − ψ(s)| . (10)

Uniqueness (Not in course)

Let Z t
W (t) = |φ(s) − ψ(s)| ds.
0

Clearly, W (0) = 0, W (t) ≥ 0. Also, W 0 = |φ(t) − ψ(t)|.

Now using (9) and (10), we get

W 0 (t) − MW (t) ≤ 0.

Thus,
e −Mt W (t) ≤ 0.
 

Integrate from 0 to t and use W (0) = 0 to conclude W (t) ≤ 0.

Thus
W (t) ≡ 0,
and so W 0 (t) ≡ 0. Thus φ(t) ≡ ψ(t).