MA 102 (Ordinary Differential Equations)

IIT Guwahati
Tutorial Sheet No. 8 Date: March 16, 2020

First order ODEs, orthogonal trajectories, IVPs, Picard’s Theorem

(1) Find the orthogonal trajectories of the following families of curves.

i) y = cex . ii) x2 = 2y − 1 + ce−2y .

dy
Solutions: i) Let y = cex → (1). Differenting with respect to x we have, dx = y → (2).
This is the differential equation of which the given curves are a family of solutions. Thus
the orthogonal trajectories are the solutions of y = − dx
dy
and are given by 2x + y 2 = c.
−2y dy
2
 Let x  = 2y − 1 + ce
ii) Differentiating with respect to x, x = (1 − ce−2y ) dx ⇒c=
dy
1 − x dx
dy
e2y . Therefore the given curves satisfy x2 = 2y − x dx
dy
⇒ dx x
= 2y−x 2 . Thus the

dx x x2 c
orthogonal trajectories are the solutions of dy
= x2 −2y
. The solution is y = 4
+ x2
.
dy
(2) Let a family of curves be the integral curves of a differential equation dx = f (x, y). Let a
second family of curves have the property that at each point of intersection P (x, y) with
the first family, the angle between the respective tangents is always equal to α. If α 6= π2 ,
then each family of curves is said to be a family of oblique trajectories of the other.
(a) Show that the curves of the second family are solutions of the differential equation
dy f (x,y)+tan α
dx
= 1−f (x,y)tan α
.
(b) Use part (a) to find the family of oblique trajectories with α = π4 of all circles x2 +y 2 =
c2 .

Solutions: (a)

Let θ1 and θ2 be the angles made by the tangents to the first and second families of curves
at P with the positive direction of the x-axis respectively. If α be the angle between the
two tangents at P, then, assuming the geometry in the figure above, α = θ2 − θ1 , or
equivalently,
tan α + tan θ1
θ2 = α + θ1 ⇒ tanθ2 = .
1 − tan αtan θ1
2

dy
Let dx
= g(x, y) be the differential equation of the second family of curves. Then
dy tan α + f (x, y)
= (1).
dx 1 − tan αf (x, y)

Remark: If instead of the figure above, the following figure is assumed,

then the differential equation to be solved to find the required oblique trajectories is
dy f (x, y) − tan α
= . (2)
dx 1 + f (x, y)tan α
However, due to the constant of integration involved in the solution of either equation (1)
or (2), all possible trajectories get covered.
dy
(b) The differential equation of the family of circles x2 + y 2 = c2 is dx = − xy . Therefore
the required family of oblique trajectories form integral curves of the ODE
dy 1 − xy dy y−x
= x ⇒ = .
dx 1+ y dx y+x
p
Solving this homogeneous equation, the required family is tan−1 xy + ln(C x2 + y 2 ) = 0.

(3) Solve (x − y 2 )dx + 2xydy = 0.

dy y 1
Solution Rearranging the terms gives the Bernoulli equation dx = 2x
− 2y
. Substituting
2 dv v 2
v = y results in dx − x = −1. The solution is y = −xln|x| + Cx.
dy
(4) The nonlinear equation = P (x) + Q(x)y + R(x)y 2 is referred to as the Ricatti equation.
dx
Observe that if R(x) = 0, then it is a linear equation and if P (x) = 0, then it is a Bernoulli
equation. Show that if y1 is a known particular solution of the Ricatti equation, then it
has a family of solutions y = y1 + u where u is the solution of
du
− (Q + 2y1 R)u = Ru2 .
dx
R
e (Q+2y1 R)dx
Use this fact to establish that y = y1 − R R (Q+2y R)dx is a family of solutions of
e 1 Rdx − C
the given Ricatti equation.
dy
Solution: Since y1 is a solution of dx = P (x) + Q(x)y + R(x)y 2 → (1) and u is a solution
of du
dx
= (Q+2y1 R)u+Ru2 , → (2) we have dx d
(y1 +u) = P (x)+Q(x)(y1 +u)+R(x)(y1 +u)2 .
 3

Hence y1 + u is a solution of the given Ricatti equation. Setting w = u1 in (2) gives the
linear equation dw
dx
+ (Q + 2y1 R)w = −R. Therefore,
R
Z R R
− [Q(x)+2y1 R(x)]dx
w = −e e [Q(x)+2y1 R(x)]dx R(x)dx + Ce− [Q(x)+2y1 R(x)]dx .

Thus a family of solutions of the given Ricatti equation is

R
1 e [Q(x)+2y1 R(x)]dx
y = y1 + = y1 − R R [Q(x)+2y R(x)]dx .
w e 1 R(x)dx − C

(5) Use the above result to solve the following Ricatti equations where a particular solution is
given for each case.
dy 4 y 2
(i) = − 2 − + y 2 , (y1 = ).
dx x x x
dy 2 2
(ii) = sec x + ytanx − y , (y1 = tanx).
dx
Solutions: i) Here P (x) = − x42 , Q(x) = − x1 , R(x) = 1. As y1 = x2 is a solution of the
equation, a general solution is given by y = x2 + u where u is a solution of
du du 3
− (Q(x) + 2y1 R(x))u = R(x)u2 ⇒ − u = u2 .
dx dx x
dv
This is a Bernoulli equation which converts to the linear equation dx + x3 v = −1 by setting
3
v = u1 . Solving it gives v = − x4 + xC3 ⇒ u = 4C−x
4x
4 . Thus the general solution of the given
2 4x3
Ricatti equation is y = x + 4C−x4 .
1
ii) Solution is y = tanx − tanx+C sec x
.

(6) Discuss the existence and uniqueness of a solution of the following initial value problems
(IVP) in the region R : |x| ≤ 1 |y| ≤ 1.
dy dy p
(a) = 3y 2/3 , y(0) = 0; (b) = |y|, y(0) = 0;
dx dx
dy
(c) = x2 + y 2 , y(0) = 0.
dx
Solution: (a) f (x, y) = 3y 2/3 ∈ C(R), hence IVP has a solution. But, f does not satisfy
a Lipschitz condition on R, since
|f (x, y) − f (x, 0)|  3y 2/3 
 
3
= =
|y − 0| y  |y 1/3 |
is unbounded in every neighborhood of the origin. Thus, the IVP has a solution but it
may not be unique. Note that y1 (x) = 0 and y2 (x) = x3 are two different solutions.
p
(b) f (x, y) = |y| ∈ C(R), hence IVP has a solution. But, the solution is may not be
unique as for y > 0 √
|f (x, y) − f (x, 0)| y 1
= = 1/2
|y − 0| y y
is unbounded
 2 in every neighborhood of the origin. Note that y1 (x) = 0 and
x /4, x ≥ 0,
y2 (x) = 2 are two different solutions.
−x /4, x < 0.

(c) Note that f, ∂f

∂y
∈ C(R). Clearly, f satisfies Lipschitz condition w. r. t. y and
(0, 0) ∈ R. By Picard’s theorem IVP has a unique solution which is certain to exists in
|x| ≤ 1/2.

(7) Show that the equation |y 0 (x)| + |y(x)| + 1 = 0 has no real solutions.
4

Solution: Suppose φ(x) is a real solution on some interval I. Then |φ0 (x)| + |φ(x)| + 1 =
0, ∀x ∈ I. But, |φ0 (x)|, |φ(x)| ≥ 0. So, |φ0 (x)| + |φ(x)| + 1 ≥ 1, ∀x ∈ I, which leads to a
contradiction.

(8) Find particular solutions of the following.

(a) xy 0 + 3y = sin
x2
x
, x 6= 0, y(π/2) = 1. 
2, 0 ≤ x < 1,
(b) y 0 + y = f (x), y(0) = 0, where f (x) =
0, x ≥ 1.

Solution: (a) I.F. µ(x) = x3 . The general solution is x3 y(x) = − cos x + C. The initial
3 3
condition y(π/2) = 1 =⇒ C = π8 . The particular solution is yx3 + cos x = π8 .

(b) Split the problem into two IVPs: I. For 0 ≤ x < 1, y 0 + y = 2, y(0) = 0;
II. For x ≥ 1, y 0 + y = 0, y(1) = k, where limx→1− φ1 (x) = k, and φ1 (x) is the solution of
I. The solution of IVP I is φ1 (x) = 2 − 2e−x . Now,

lim φ1 (x) = lim (2 − 2e−x ) = 2 − 2e−1 = k.

x→1− x→1−

This value k = 2 − 2e−1 would be the initial condition for IVP II. The solution of IVP II
is φ2 (x) = (2e − 2)e−x for x ≥ 1. The solution of the given IVP is

φ1 (x) = 2 − 2e−x , 0 ≤ x < 1,


y(x) =
φ2 (x) = (2e − 2)e−x , x ≥ 1.

(9) A point P is dragged along the xy plane by a string P T of length a. If T starts at the
origin and moves along the positive y axis and if P starts at (a, 0), assuming that the
string is always tangent to the curve traced by P, what is the path of P ?

Solution: The differential of the path is

√
dy a2 − x 2
=− .
dx x

Separating variables and integrating and using the condition y(a) = 0 yields

√
√
 
a+ a2 − x 2
y(x) = a log − a2 − x 2 .
x

dy tany
(10) Verify that the initial value problem = , y(1) = π/2 has infinitely many solutions.
dx x
Explain why this does not contradict Picard’s Theorem.
dy
Solution: The general solution of dx = tanx y is y = sin−1 (Cx). Now y(1) = π2 ⇒ C = 1.
−1
But y = nπ +sin x is also a solution satisfying y(1) = π2 . this does not contradict Picard’s
Theorem as tanx y is discontinuous in any domain of the xy-plane containing the line y = π2 .

(11) Use Picard’s method of successive approximation to find y1 , y2 , y3 , y4 . Determine the limit
of the sequence {yn (x)} as n → ∞.

(i) y 0 = −y, y(0) = 1.

(ii) y 0 = x + y, y(0) = 1.
 5
Rx
Solutions: i) Picard’s n-thR iteration is yn (x) = 1 + 0 f (t, yn−1 (t))dt, n = 1, 2, . . . . Since
x
f (x, y) = −y, yn (x) = 1 − 0 yn−1 (t)dt, n = 1, 2, . . . Let y0 (x) = 1. Then,
Z 1
y1 (x) = 1 − dt = 1 − x
0
Z x
x2
y2 (x) = 1 + (−1 + t)dt = 1 − x + .
2
Z0 x 2
t x2 x3
y3 (x) = 1 + (−1 + t − )dt = 1 − x + −
0 2 2 3!
Z x 2 3
t t x2 x3 x4
y4 (x) = 1 + (−1 + t − + )dt = 1 − x + − + .
0 2 3! 2! 3! 4!
Pn (−x)k
Thus, yn (x) = k=0 k! and y(x) = limn→∞ yn (x) = e−x is a solution.
Rx
ii) Here f (x, y) = x + y and y(0) = 1. Therefore yn (x) = 1 + 0 (t + yn−1 (t))dt, n = 1, 2, . . .
Let y0 (x) = 1. Then,
Z 1
x2
y1 (x) = 1 + (1 + t)dt = 1 + x +
2
Z0 x 2
t x3
y2 (x) = 1 + (1 + 2t + )dt = 1 + x + x2 + .
2 6
Z0 x 3
t x3 x4
y3 (x) = 1 + (1 + 2t + t2 + )dt = 1 + x + x2 + +
0 3! 3 4!
Z x 3 4
t t x3 x4 x5
y4 (x) = 1 + (1 + 2t + t2 + + )dt = 1 + x + x2 + + + .
0 3 4! 3 12 5!
2 2 3
Therefore we have, y1 (x) = 2(1 + x) + x2 − (1 + x), y2 (x) = 2(1 + x + x2 ) − (1 + x) + x3! ,
2 3 4 2 3 4 5
y3 (x) = 2(1 + x + x2! + x ) − (1 + x) + x4! and y4 (x) = 2(1 + x + x2! + x3! + x4! ) − (1 + x) + x5! .
3! 
k xn+1
Hence yn (x) = 2 nk=0 xk! + (n+1)!
P
− (1 + x), and the solution is
y(x) = lim yn (x) = 2ex − (1 + x).
n→∞