Study material for Week-3 on

Partial Differential Equations

S. Sivaji Ganesh, IIT Bombay.

Distributed to students of MA 515, Autumn 2021.

 Study material on
Partial Differential Equations

S. Sivaji Ganesh
Department of Mathematics
Indian Institute of Technology Bombay

August 1, 2021
 Preface

This notes is intended for circulation to students of a course on Partial differential

equations.
Though the notes are proof-read many times, it could still have some misprints.
Please send your comments, corrections etc. to [email protected], where X=msg2sgspde

S. Sivaji Ganesh,
Mumbai.
August 1, 2021.

v
2.5. Cauchy problem for general nonlinear equation 79

Thus, the solution to Cauchy problem is given by

u(x1 , x2 , x3 ) = Z(T (x1 , x2 , x3 ), S1 (x1 , x2 , x3 ), S2 (x1 , x2 , x3 ))

= h(S1 (x1 , x2 , x3 ), S2 (x1 , x2 , x3 ))e kT (x1 ,x2 ,x3 )
 
x1 x2 k
=h , x .
x3 x3 3

Thus the solution is given by

 
x1 x2
u(x1 , x2 , x3 ) = x3k h , , (2.67)
x3 x3

defined for all (x1 , x2 , x3 ) ∈ R3 for which x3 6= 0. For each λ ∈ R such that
λ 6= 0, observe that the function u given by (2.67) satisfies

u(λx1 , λx2 , λx3 ) = λk u(x1 , x2 , x3 ).

whenever x3 6= 0. If the function u is known to be continuous, then the

above equation holds for (x1 , x2 , x3 ) ∈ R3 . Thus the solution to the Cauchy
problem is a homogeneous function of order k.

2.5 Cauchy problem for general nonlinear equation

In this section, we would like to extend the method of characteristics to
solve Cauchy problem for general nonlinear equations.
Let Ω5 be an open and connected subset of R5 , and let F : Ω5 → R
be an arbitrary function. We denote the argument of F by the 5-tuple
(x, y, z, p, q). Such an F gives rise to the most general form of a first order
PDE, which is
F (x, y, u, u x , uy ) = 0. (GE)
Observe that the equation (GE) reduces to a quasilinear equation (QL) if
F is of the form

F (x, y, z, p, q) = a(x, y, z) p + b (x, y, z)q − c(x, y, z)

for some functions a, b , c. Thus we may say that we are considering equa-
tions which are more general than quasilinear equations.
Our analysis of equation (GE) draws inspiration from that of (QL),
and frequent comparisons to it will be made throughout the presentation

August 1, 2021 Sivaji

80 Chapter 2. First order PDE

which helps in understanding the new difficulties and the process of gener-
alizing the ideas from the case of (QL). We do not wish to disturb the flow
of the analysis for general nonlinear equation with frequent comparisons
to quasilinear equation, and such comparisons are typeset in a smaller font
size throughout the rest of this section, along with justifications for some
of the assumptions that we make.
The experience with Cauchy problems for quasilinear equations (QL)
will be used in extending the ideas therein to general nonlinear equations
(GE). Recall that in the case of quasilinear equation (QL), the integral
surface was constructed using characteristic curves. The strategy was ex-
ecuted successfully by making some additional assumptions like transver-
sality condition, to overcome the difficulties in implementing the strategy.
We would like to pursue the same idea here also.
The analysis of general nonlinear equation (GE) is organized into the
following steps:

Step 1: Obtaining a system of ordinary differential equations for characteris-

tic strip.

Step 2: Finding an initial strip.

Step 3: Defining a candidate solution.

Step 4: Establishing that the candidate solution solves Cauchy problem.

Each of these steps will be carried out in a separate subsection, to demar-

cate the arguments involved in each of the steps. For quasilinear equations
(QL), Step 1 was carried out very easily owing to the explicit geometric
meaning of the equation, in Subsection 2.3.2, and there was no Step 2 as
Cauchy data was sufficient to find solutions to the characteristic system of
ODEs.

2.5.1 Hypothesis and Notations

The projections of Ω5 to xy-plane and xy z-space are denoted by Ω2 and Ω3

respectively.
Hypotheses on the function F
The function F : Ω5 → R is assumed to satisfy the following hypothesis:

(i) F is a continuously differentiable function. Equivalently, the partial

derivatives F x , Fy , F z , F p , Fq are continuous on Ω5 .

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 81

(ii) The partial derivatives F p and Fq satisfy

F p2 (x, y, z, p, q) + Fq2 (x, y, z, p, q) 6= 0 ∀ (x, y, z, p, q) ∈ Ω5 . (2.68)

Hypotheses on the Cauchy data

Let I be an interval in R, and the functions f , g , h be continuously
differentiable on I such that ( f 0 (s ))2 + (g 0 (s))2 6= 0 for all s ∈ I . Consider
the space curve Γ described parametrically by

Γ : x = f (s), y = g (s), z = h(s), s ∈ I. (2.69)

For a given space curve Γ , recall that Cauchy problem for the general
nonlinear equation (GE)

F (x, y, u, u x , uy ) = 0. (GE)

consists of finding a solution u to the equation (GE) satisfying the Cauchy

condition
u ( f (s ), g (s)) = h(s ), s ∈ I 0 ,
where I 0 is a subinterval of I .
In other words, Cauchy problem for (GE) consists of finding a function
u : D ⊆ Ω2 → R such that for each (x, y) ∈ D,

(i) (x, y, u(x, y), u x (x, y), uy (x, y)) ∈ Ω5 ,

(ii) F x, y, u(x, y), u x (x, y), uy (x, y) = 0,
(iii) and for each s ∈ I 0 , u( f (s), g (s)) = h(s).

Geometrically speaking, Cauchy problem consists of finding an integral

surface on which a part of the datum curve Γ lies.
Recall that in most of the examples presented in Section 2.4, we can nei-
ther expect a solution to be defined on whole of Ω2 nor the corresponding
integral surface to contain the entire datum curve Γ . Thus we can only
expect the existence of a local (w.r.t. datum curve) solution to the Cauchy
problem for the equation (GE).

2.5.2 Equations for Characteristic strips

Recall that for quasilinear equation (QL)

a(x, y, u) u x + b (x, y, u) uy = c(x, y, u), (QL)

August 1, 2021 Sivaji

82 Chapter 2. First order PDE

we observed that at a point (x, y, z) lying on an integral surface S : z =

u(x, y), the characteristic direction (a(x, y, z), b (x, y, z), c(x, y, z)) is a di-
rection in the tangent plane as (u x (x, y), uy (x, y), −1) is a normal direction
to the surface S. This observation led to the definition of a characteristic
system of ODE (chara.ODE) (see Definition 2.15).
Note that there is no such automatic choice of a characteristic direction
suggested by the general nonlinear equation (GE). We follow a geomet-
ric argument to choose a direction at each point of Ω3 which belongs to
the tangent space to a possible integral surface containing the point, and
the chosen direction plays the role that characteristic direction played for
quasilinear equations.
Let S : z = u(x, y) be a possible integral surface for the equation (GE)
where u is a continuously differentiable function, and P0 (x0 , y0 , z0 ) ∈ S.
Equation of the tangent plane to the integral surface S at the point P0 ∈ S
is of the form
z − z0 = p(x − x0 ) + q(y − y0 ),
where p and q satisfy F (x0 , y0 , z0 , p, q) = 0.
In fact, we know that p = u x (x0 , y0 ) and q = uy (x0 , y0 ) must hold. However, remember that
our goal is to find a direction at each of the points of Ω3 with the property that it belongs to the
tangent plane to every integral surface containg that point. This goal must be achieved using only
the information that can be extracted from the given nonlinear equation (GE).
Thus we get lots of possible tangent planes (to possible integral surfaces)
at the point P0 , which are as many as the number of pairs ( p, q) satisfying
the equation F (x0 , y0 , z0 , p, q) = 0. Possible tangent planes (to possible in-
tegral surfaces) at a point P0 (x0 , y0 , z0 ) ∈ S are given by the two equations

z − z0 = p(x − x0 ) + q(y − y0 ) (2.70a)

F (x0 , y0 , z0 , p, q) = 0 (2.70b)
The equation (2.70b) in the special case of quasilinear equation (QL) reads as

a(P0 ) p + b (P0 )q = c(P0 ).

Without loss of generality assume that b (P0 ) 6= 0. Then for each p ∈ R, q( p) is given by

a(P0 ) c(P0 )
q( p) = − p+ .
b (P0 ) b (P0 )

Thus equation of the tangent plane (2.70a) for the equation (QL) becomes
 
c(P0 ) a(P0 )
z − z0 = (y − y0 ) + p x − x0 − (y − y0 ) , (2.71)
b (P0 ) b (P0 )

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 83

indexed by a parameter p ∈ R.
For the general nonlinear equation (GE), the system of equations (2.70)
represents a one-parameter family of possible tangent planes at the point
P0 .
Motivated by the analysis for equation (QL), we were searching for a tangential direction that
belongs to a possible integral surface but ended up with a one-parameter family of possible tangent
planes in the case of the equation (GE). If we understand the relation (if any) between the tangential
direction and the one-parameter family of possible tangent planes in the special case of quasilinear
equation (QL), we hope to employ a related idea and find the tangential direction as intended. The
geometric idea of an envelope (see Appendix ??) connects the characteristic direction and the one-
parameter family of tangent planes in the case of quasilinear equation (QL), and we will pursue this
idea for the general equation (GE).
The geometric idea to find a characteristic vector field is to consider the
envelope of the one-parameter family of planes described by (2.70). This
idea is expected to succeed since the envelope and members of the fam-
ily (2.70) share a common tangent plane at the points of intersection (see
Lemma ??). In our situation the envelope is expected to be a cone since all
the planes pass through a fixed point P0 , a generator of the cone will be
chosen as the characteristic direction . In fact, the characteristic direction
for quasilinear equation (QL) is the envelope of the one-parameter family
of tangent planes (2.71) (See Example ?? for details).
We now find the envelope of the one-parameter family of planes (2.70a).
Assume that there exists a differentiable function q = q( p) such that

F (x0 , y0 , z0 , p, q( p)) = 0. (2.72)

Note that it is always possible to find such a function locally. Let p0 , q0 be such that F (x0 , y0 , z0 , p0 , q0 ) =
0 and Fq (x0 , y0 , z0 , p0 , q0 ) 6= 0. By implicit function theorem we can express q as a function of the
variable p in a neighbourhood of p0 .
Now the one-parameter family of planes in (2.70a) is given by

z − z0 = p (x − x0 ) + q( p) (y − y0 ) (2.73)
Differentiating the equations (2.72) and (2.73) w.r.t. p yields

F p (x0 , y0 , z0 , p, q( p)) + Fq (x0 , y0 , z0 , p, q( p)) q 0 ( p) = 0, (2.74a)

0 = (x − x0 ) + q 0 ( p) (y − y0 ). (2.74b)

August 1, 2021 Sivaji

84 Chapter 2. First order PDE

Eliminating q 0 ( p) from (2.74), we get

x − x0 y − y0
= ,
F p (P0 , p, q( p)) Fq (P0 , p, q( p))

where (P0 , p, q( p)) is used to denote (x0 , y0 , z0 , p, q( p)). Now using the
equation of planes (2.73) we get the following expression for the curve of
intersection C p of the family of planes and its envelope for each p.

x − x0 y − y0 z − z0
Cp : = =
F p (P0 , p, q( p)) Fq (P0 , p, q( p)) p F p (P0 , p, q( p)) + q( p)Fq (P0 , p, q( p))
(2.75)
The equation (2.75) represents a family of lines indexed by p passing through
the point P0 (x0 , y0 , z0 ) and having direction
€ Š
F p (P0 , p, q( p)), Fq (P0 , p, q( p)), pF p (P0 , p, q( p)) + q( p)Fq (P0 , p, q( p)) .
(2.76)
corresponding to each solution ( p, q( p)) of (2.72). Note that the 3-tuple in
(2.76) is non-zero by the hypothesis (2.68) on F .
For quasilinear equation (QL), the lines (2.75) do not depend on p, and is the line passing
through P0 having characteristic direction. For details, see Example ??.
Envelope of the family of planes (2.73) is, by definition, the union of the
lines (2.75) as p varies, and is a ‘cone’ having vertex at P0 in general (visu-
alize the way a family of lines passing through P0 looks like). Thus a cone
field is defined on the domain Ω3 , which reduces to a vector field when the
equation (GE) reduces to a quasilinear equation (QL). The cone is called
Monge Cone, and each of the lines C p (corresponding to each fixed p) is
a generator of the cone, and any of them may be taken as a characteristic
direction. Thus for every P (x, y, z) ∈ Ω3 and for every ( p, q) such that
F (P, p, q) = 0, we have a choice for a characteristic direction at P given by

€ Š
F p (P, p, q), Fq (P, p, q), p F p (P, p, q) + q Fq (P, p, q) . (2.77)

Inspired by the success in finding an integral surface for quasilinear equa-

tion (QL) using characteristic curves, we hope to repeat the same for gen-
eral nonlinear equations as well. Thus we are in search of curves in Ω3 hav-
ing tangential direction (2.77) at each of its points. Note here that p, q in
(2.77) depends on the point P (x, y, z) through the relation F (x, y, z, p, q) =
0.

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 85

Let (x(t ), y(t ), z(t )) (t ∈ J , J is an interval in R, and 0 ∈ J ) be a curve in

parametric form that passes through a point P0 (x0 , y0 , z0 ) at t = 0, having
the tangential direction (2.77) at each of its points P (x(t ), y(t ), z(t )). Thus
the following system of characteristic ODEs are satisfied along the curve:

dx
= F p (x, y, z, p, q) (2.78a)
dt
dy
= Fq (x, y, z, p, q) (2.78b)
dt
dz
= pF p (x, y, z, p, q) + q Fq (x, y, z, p, q). (2.78c)
dt
along with (x(0), y(0), z(0)) = (x0 , y0 , z0 ).
Note that we cannot solve the above system as the unknowns p, q ap-
pearing in it depend on (x(t ), y(t ), z(t )) via F (x(t ), y(t ), z(t ), p, q) = 0. In
other words, we also need to know p and q as functions of t , which was not
the case for a quasilinear equation (QL). Since x(t ), y(t ), z(t ) themselves
are unknown functions, there is no way to determine p(t ) and q(t ). This
is the first difficulty in extending the ideas from the analysis of quasilinear
equations (QL).
Thus we need to complete the system (2.78) by supplementing equa-
tions for the quantities p(t ), q(t ) along the curve (x(t ), y(t ), z(t )) such that
( p(t ), q(t ), −1) is the normal direction to a possible integral surface. Thus
we have to find five quantities x(t ), y(t ), z(t ), p(t ), q(t ) and not just the
three quantities x(t ), y(t ), z(t ). Geometrically speaking, we have to deter-
mine not only a curve but also a tangent plane to a possible integral surface
that contains this curve.

Definition 2.57 (Characteristic strip). Let J be an interval in R, and γ :

x = x(t ), y = y(t ), z = z(t ), t ∈ J be a curve having the tangential direction
(2.77) at each of its points, where p(t ), q(t ) satisfy F (x(t ), y(t ), z(t ), p(t ), q(t )) =
0. The 5-tuple (x(t ), y(t ), z(t ), p(t ), q(t )) is called a characteristic strip, and
the point (x(t ), y(t ), z(t )) is called support of the characteristic strip.

Strip can be thought of as a point (x(t ), y(t ), z(t )) along with an in-
finitesimal plane element passing through the point (x(t ), y(t ), z(t )) and
having the normal direction (to a possible integral surface) as ( p(t ), q(t ), −1),
see Figure 2.11 for an illustration.

August 1, 2021 Sivaji

86 Chapter 2. First order PDE

Figure 2.11. Characteristic strips: Characteristic curve with a tangent plane attached at each
of its points.

Derivation of equations for Characteristic strips

Assume that the function u defining the integral surface S : z = u(x, y) is

a twice continuously differentiable function for the purposes of deriving
the equations for p and q.

Derivation of equation satisfied by p(t )

Differentiating the equation F x, y, u(x, y), u x (x, y), uy (x, y) = 0 w.r.t.
x gives the equation

F x (ζ ) + F z (ζ )u x (x, y) + F p (ζ )u x x (x, y) + Fq (ζ )uy x (x, y) = 0, (2.79)

where we used the notation ζ := x, y, u(x, y), u x (x, y), uy (x, y) . Observe
that we used the assumption on u being a twice continuously differentiable
function.
Let (x(t ), y(t ), z(t )) be a point lying on the integral surface z = u(x, y).
We require the equation

p(t ) = u x (x(t ), y(t )) (2.80)

to hold. On differentiating the above equation w.r.t. the variable t , we get

p 0 (t ) = u x x (x(t ), y(t )) x 0 (t ) + uy x (x(t ), y(t )) y 0 (t ).

Since equations (2.78) hold, denoting ζ (t ) := (x(t ), y(t ), z(t ), p(t ), q(t )),

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 87

we get

p 0 (t ) = u x x (x(t ), y(t )) F p (ζ (t )) + uy x (x(t ), y(t )) Fq (ζ (t )) .

From the equation (2.79), we have

u x x (x(t ), y(t )) F p (ζ (t )) + uy x (x(t ), y(t )) Fq (ζ (t ))

= − (F x (ζ (t )) + p(t )F z (ζ (t ))) .

Thus we get
p 0 (t ) = − (F x (ζ (t )) + p(t )F z (ζ (t ))) .

Derivation of equation satisfied by q(t )

Differentiating the equation F x, y, u(x, y), u x (x, y), uy (x, y) = 0 w.r.t.
y gives the equation

Fy (ζ ) + F z (ζ )uy (x, y) + F p (ζ )u xy (x, y) + Fq (ζ )uyy (x, y) = 0, (2.81)

where we used the notation ζ := x, y, u(x, y), u x (x, y), uy (x, y) . Observe
that we used the assumption on u being a twice continuously differentiable
function.
Let (x(t ), y(t ), z(t )) be a point lying on the integral surface z = u(x, y).
We require the equation

q(t ) = uy (x(t ), y(t )) (2.82)

to hold. On differentiating the above equation w.r.t. the variable t , we get

q 0 (t ) = u xy (x(t ), y(t )) x 0 (t ) + uyy (x(t ), y(t )) y 0 (t ).

Since equations (2.78) hold, denoting ζ (t ) := (x(t ), y(t ), z(t ), p(t ), q(t )),
we get

q 0 (t ) = u xy (x(t ), y(t )) F p (ζ (t )) + uy y (x(t ), y(t )) Fq (ζ (t )) .

From the equation (2.81), we have

u xy (x(t ), y(t )) F p (ζ (t )) + uyy (x(t ), y(t )) Fq (ζ (t ))

= − Fy (ζ (t )) + q(t )F z (ζ (t )) .

Thus we get

q 0 (t ) = − Fy (ζ (t )) + q(t )F z (ζ (t )) .

August 1, 2021 Sivaji

88 Chapter 2. First order PDE

Thus we get the following system of ODEs for the characteristic strip.

dx
= F p (x, y, z, p, q) (2.83a)
dt
dy
= Fq (x, y, z, p, q) (2.83b)
dt
dz
= pF p (x, y, z, p, q) + q Fq (x, y, z, p, q) (2.83c)
dt
dp
= −F x (x, y, z, p, q) − pF z (x, y, z, p, q) (2.83d)
dt
dq
= −Fy (x, y, z, p, q) − q F z (x, y, z, p, q) (2.83e)
dt
and we denote this sytem of ODE (2.83) by (Chara.Strip.ODE).

2.5.3 Finding an initial strip

The second step in solving Cauchy problem is to find an IVP for the system
(Chara.Strip.ODE). The datum curve gives rise to the initial conditions for
x, y, z, and we need to find the initial conditions for p, q. In other words,
we need to determine an initial strip.
Let ζ (t ) := (x(t ), y(t ), z(t ), p(t ), q(t )) be a solution to the system of
ODE (2.83). It follows easily from the system (2.83) that along a character-
istic strip
d
F (x(t ), y(t ), z(t ), p(t ), q(t )) = 0.
dt
Thus the function t 7→ F (ζ (t )) is a constant function, and hence F (ζ (t )) =
F (ζ (0)). Therefore if we choose an initial strip ζ (0) = (x0 , y0 , z0 , p0 , q0 )
satisfying F (x0 , y0 , z0 , p0 , q0 ) = 0, then it follows that

F (x(t ), y(t ), z(t ), p(t ), q(t )) = 0 for all t ∈ J. (2.84)

Recall that the Cauchy data is given by

Γ : x = f (s), y = g (s ), z = h(s) s ∈ I. (2.69)

We are interested in passing a characteristic curve through every point of

Γ . Since the (Chara.Strip.ODE) is a coupled system involving p and q as
well, we first have to determine an initial strip having the datum curve as
its support.

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 89

Since the integral surface z = u(x, y) must contain a part of Γ on it, we

must have
h(s ) = u( f (s ), g (s))
for s belonging to a subinterval of I . Differentiating the above equation
w.r.t. s, we get

h 0 (s) = u x ( f (s ), g (s)) f 0 (s ) + uy ( f (s), g (s )) g 0 (s ).

Since p(s) = u x ( f (s), g (s)) and q(s) = uy ( f (s ), g (s)) (by definition of p

and q), the above equation becomes

h 0 (s ) = p(s) f 0 (s ) + q(s ) g 0 (s).

Thus, we need to solve for ( p, q) as functions of s , i.e., ( p, q) = ( p(s), q(s)),

from the system of transcendental equations

F ( f (s), g (s), h(s), p, q) = 0 (2.85a)

p f 0 (s) + q g 0 (s ) = h 0 (s ). (2.85b)

Implicit function theorem says that if we know a special solution ( p0 , q0 )

for some s = s0 of the system of equations (2.85), and if a certain Jacobian
condition is satisfied, then we can solve uniquely for p, q as differentiable
functions of s in a neighbourhood of s = s0 from the system of equations
(2.85).
Let ( p, q) = ( p0 , q0 ) be a solution to the system of equations (2.85) at
s = s0 . That is,

F ( f (s0 ), g (s0 ), h(s0 ), p0 , q0 ) = 0, (2.86a)

0 0 0
p0 f (s0 ) + q0 g (s0 ) = h (s0 ). (2.86b)

In order to put our problem in the set-up of implicit function theorem,

we define the function (φ1 , φ2 ) : I × R × R → R2 by

φ1 (s, p, q) = F ( f (s), g (s), h(s), p, q), (2.87a)

0 0 0
φ2 (s, p, q) = p f (s ) + q g (s) − h (s ). (2.87b)

Note that we are interested in simultaneous solutions to the system of

equations φ1 (s, p, q) = 0 and φ2 (s , p, q) = 0. From (2.86), we know that
(s0 , p0 , q0 ) is a particular solution. To apply implicit function theorem, we
need that the functions φ1 , φ2 are continuously differentiable functions of
the variables s, p, q, which is satisfied if we assume that f , g , h ∈ C 2 (I ).

August 1, 2021 Sivaji

90 Chapter 2. First order PDE

Figure 2.12. Initial strip at point of Datum curve.

Another hypothesis of implicit function theorem requires that the Jaco-

bian of φ1 , φ2 w.r.t. ( p, q) at the special solution (s0 , p0 , q0 ) of the system
φ1 (s, p, q) = 0, φ2 (s, p, q) = 0, denoted by ∆0 , satisfies
 
∂ (φ1 , φ2 )  F (ζ ) F (ζ ) 
 p 0 q 0 
∆0 := (s0 , p0 , q0 ) =   6= 0, (2.88)
∂ ( p, q)  f (s0 ) g (s0 ) 
0 0

where we used the notation ζ0 = ( f (s0 ), g (s0 ), h(s0 ), p0 , q0 ).

In view of the assumption (2.88), we can apply implicit function theo-
rem, and conclude that there exist unique functions p = p(s) and q = q(s)
which are continuously differentiable for s belonging to an interval con-
taining s0 (let us still denote it by I ) such that

φ1 (s, p(s), q(s)) = 0, φ2 (s, p(s), q(s)) = 0,

p(s0 ) = p0 , q(s0 ) = q0 .

Thus an initial strip ( f (s ), g (s), h(s), p(s), q(s)) has been determined, and
thus succesffully implementing Step 2 of our programme. Figure 2.12 il-
lustrates initial strips having the datum curve Γ as their supports.

2.5.4 Defining a candidate solution

Determining characteristic strips

For each of the initial strips whose supports are points on Γ , find a char-
acteristic strip such that its support intersects Γ , i.e., find solutions to the
system (Chara.Strip.ODE) given by equations (2.83) along with initial con-

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 91

Figure 2.13. Characteristic curve and Datum curve along with associated strips.

ditions

(x(0), y(0), z(0), p(0), q(0)) = ( f (s ), g (s), h(s), p(s), q(s)) .

Let

x = X (t , s), y = Y (t , s), z = Z(t , s), p = P (t , s), q = Q(t , s ). (2.89)

represent solutions to (Chara.Strip.ODE) given by equations (2.83) satis-

fying the initial conditions

X (0, s) = f (s), Y (0, s) = g (s), Z(0, s) = h(s), P (0, s) = p(s ), Q(0, s) = q(s ),
(2.90)
and these solutions are unique if we assume that F ∈ C (Ω5 ). 2

The progress made in finding a solution to the Cauchy problem is illus-

trated in Figure 2.13.

As explained on page 52, in a similar situation for the case of quasilinear

equation, we may assume without loss of generality that there exists an
interval J such that 0 ∈ J , and the functions in (2.89) are defined for (t , s ) ∈
J × I . Also, all these functions are continuously differentiable functions
on J × I , by differentiable dependence of solutions on parameters in the

August 1, 2021 Sivaji

92 Chapter 2. First order PDE

theory of ordinary differential equations.

In view of the equations (2.84), (2.90), we have for (t , s) ∈ J × I

F (X (t , s), Y (t , s), Z(t , s), P (t , s), Q(t , s )) = 0, (2.91a)

X (0, s ) = f (s), Y (0, s) = g (s), Z(0, s) = h(s), (2.91b)
P (0, s) = p(s), Q(0, s ) = q(s). (2.91c)

The arguments presented in Subsections 2.5.2, 2.5.3, 2.5.4 till now prove
the following result.

Proposition 2.58. Consider the Cauchy problem (GE) -(2.69).

(i) Let F be a twice continuuosly differentiable function on Ω5 .

(ii) Let S : z = u(x, y) be an integral surface corresponding to the general
nonlinear equation (GE), where u is a twice continuously differentiable
function.
(iii) Let the surface S contain a part of the datum curve Γ .
(iv) Assume that the functions f , g , h describing the datum curve Γ paramet-
rically are twice continuously differentiable functions.
(v) Let P0 denote the point ( f (s0 ), g (s0 ), h(s0 )) ∈ S ∩ Γ .
(vi) Let ( p0 , q0 ) ∈ R2 be such that the equation (2.86) is satisfied.
(vii) Assume that the Jacobian condition (2.88) is satisfied.

Then the surface S is the union of the supports of the characteristic strips through
the points of Γ , in a neighbourhood of the point P0 .

Note that the above proposition is a generalization of Theorem 2.29

to the case of a general nonlinear equation. A proof of the proposition
proceeds on the lines of the arguments presented on page 44, in view of
the equations (Chara.Strip.ODE), (2.80), (2.82). Writing down the details
of the proof is left as an exercise to the reader.
In view of Proposition 2.58, there is hope to find an integral surface
using the method of characteristics even for the general nonlinear equation.
Recall that we have used u is twice continuously differentiable in deriv-
ing the system (Chara.Strip.ODE) in Step 1, and twice differentiability of
the functions f , g , h are used to assert the existence of an initial strip in
Step 2. However, if the system of equations (Chara.Strip.ODE) and ini-
tial strip are assumed to be given, it is enough to assume that the functions
f , g , h are once continuously differentiable for proceeding further.

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 93

From now on assume that

(i) an initial strip ( f (s), g (s), h(s), p(s), q(s )) consisting of continuously
differentiable functions on an interval containing s0 is given, and
(ii) the functions X (t , s), Y (t , s), Z(t , s), P (t , s), Q(t , s ) solve the system
(Chara.Strip.ODE), and satisfy the initial conditions (2.91b)-(2.91c).

Application of inverse function theorem

Now we are interested in the invertibility of the function

Φ : J × I → Ω2
(t , s) 7−→ (X (t , s), Y (t , s )) .

near the point (0, s0 ) ∈ J × I . In order to apply inverse function theorem,

the Jacobian J (0, s0 ) is required to be non-zero, where J (0, s0 ) is given by
   
 X (0, s ) X (0, s )   F p (ζ0 ) f 0 (s0 ) 
 t 0 s 0   
J (0, s0 ) =  = ,
 Y (0, s ) Y (0, s )   F (ζ )) g (s ) 
0
t 0 s 0 q 0 0

where we used the notation ζ0 = ( f (s0 ), g (s0 ), h(s0 ), p0 , q0 ). Note that J (0, s0 )
is equal to ∆0 defined in (2.88), which was assumed to be non-zero.
By inverse function theorem, there exist

(i) an open subset E of J ×I containing the point (0, s0 ) , and an open sub-
set F of Ω2 containing the point (X (0, s0 ), Y (0, s0 )) i.e., ( f (s0 ), g (s0 )),
(ii) a continuously differentiable function Ψ : F → E such that

Ψ ◦ Φ = IE , Φ ◦ Ψ = IF ,

where IE and IF are the identity functions defined on E and F respec-

tively.

Denoting Ψ(x, y) = (T (x, y), S(x, y)), for (t , s) ∈ E, we have

t = T (x, y), s = S(x, y).

Recalling that Z(t , s) was expected to be the value of solution at the point
(X (t , s), Y (t , s)), we are now in a position to define a candidate solution to
the Cauchy problem (GE) -(2.69).

August 1, 2021 Sivaji

94 Chapter 2. First order PDE

Define the function u : F → R by

u(x, y) = Z (T (x, y), S(x, y)) . (2.92)

As the function u defined above is a composition of two continuously dif-

ferentiable functions, it follows that u ∈ C 1 (F ). This u is our candidate
solution, and in the next subsection we establish that u solves the Cauchy
problem.

2.5.5 Candidate solution is a solution

To show that u defined by (2.92) solves the Cauchy problem for the general
nonlinear equation (GE), we need to prove that
(i) the following identity holds for every (x, y) ∈ F :


F x, y, u(x, y), u x (x, y), uy (x, y) = 0,

(ii) and the Cauchy condition u( f (s ), g (s)) = h(s) is satisfied for s ∈ I 0

where I 0 is a subinterval of I .
Proof of (i). Proving (i) is equivalent to showing that the equation

F X (t , s), Y (t , s), Z(t , s), u x (X (t , s), Y (t , s)) , uy (X (t , s ), Y (t , s)) = 0

holds for every (t , s) ∈ E, which is a re-statement of (i) in terms of the

(t , s )-coordinates.
We recall the equation (2.91a) here for ready reference.

F (X (t , s), Y (t , s), Z(t , s), P (t , s), Q(t , s )) = 0. (2.91a)

If we can prove that for each (t , s ) ∈ E, the equalities

P (t , s) = u x (X (t , s ), Y (t , s)) , Q(t , s) = uy (X (t , s), Y (t , s)) (2.93)

hold, then we would have

F X (t , s), Y (t , s), Z(t , s), u x (X (t , s), Y (t , s)) , uy (X (t , s), Y (t , s))
= F (X (t , s), Y (t , s), Z(t , s ), P (t , s), Q(t , s))
= 0,

thus completing the proof of (i).

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 95

Proof of (2.93).
Let us turn to the proof of the equalities (2.93). The validity of equations
(2.93) is checked by proving that for each (t , s) ∈ E, the two pairs


(P (t , s), Q(t , s )) and u x (X (t , s ), Y (t , s)) , uy (X (t , s), Y (t , s))

are solutions to a system of linear non-homogeneous equations having a

unique solution.
The equation (2.92) expressed in the (t , s)-coordinates reads as

z(t , s) = u(X (t , s ), Y (t , s )), (t , s ) ∈ E. (2.94)

Differentiating the equation (2.94) w.r.t. s and t results in the system of

equations
‚ Œ ‚ Œ‚ Œ
z s (t , s ) X s (t , s) Y s (t , s) u x (X (t , s ), Y (t , s))
= . (2.95)
z t (t , s) X t (t , s ) Y t (t , s) uy (X (t , s), Y (t , s))

On the other hand, the pair (P (t , s), Q(t , s )) satisfies the system of equa-
tions
‚ Œ ‚ Œ‚ Œ
z s (t , s) X s (t , s ) Y s (t , s) P (t , s)
= . (2.96)
z t (t , s ) X t (t , s) Y t (t , s) Q(t , s)
The second equation of (2.96) follows from the equation (2.83c). We are
going to prove that the first equation of (2.96) holds. Towards this goal, let
us define

A(t , s ) := z s (t , s) − P (t , s)X s (t , s) − Q(t , s )Y s (t , s).

Note that A(0, s ) = h 0 (s ) − p(s ) f 0 (s ) − q(s)g 0 (s) = 0 for all s by the defini-
tion of an initial strip.
We now compute the derivative of A(t , s ) w.r.t. the variable t at a point
(t , s). In the following steps, we do not write the arguments of functions,
to reduce the clutter in the equations, as they can be easily deciphered.

August 1, 2021 Sivaji

96 Chapter 2. First order PDE

At = z s t − P t X s − Q t Y s − P X s t − QY s t
∂
= (z − P X t − QY t ) + P s X t + Q s Y t − P t X s − Q t Y s
∂s t
= P s F p + Q s Fq + X s (F x + P F z ) + Y s (Fy + QF z )
∂
= (F ) − F z (z s − P X s − QY s )
∂s
= −F z A

In this computation we used

∂
(F (X (t , s), Y (t , s), Z(t , s ), P (t , s), Q(t , s))) = 0,
∂s
which follows from the equation (2.91a). Thus we have shown that for ev-
ery fixed s ∈ I , the function A(t , s) is a solution to the initial value problem
corresponding to a linear first order ordinary differential equation given by

At + F z A = 0, A(0, s ) = 0.

Since F z is a continuous function, solution to the above initial problem

is unique. We noted that A(0, s) = 0, and thus the function A(t , s) must be
the zero function. In other words, A(t , s ) = 0 for all t and s. This shows
that the second equation of (2.96) also holds.
Note that the matrix appearing in the equations (2.95) and (2.96) is the
 X Y 
 s 
same, and has non-zero determinant given by ∆ :=  s  as this is the
 Xt Yt 
Jacobain corresponding to the change of coordinates (x, y) and (t , s ). This
observation allows us to conclude (2.93).
The equations (2.93) in terms of (x, y)-coordinates read as for each (x, y) ∈
F,
p(x, y) = u x (x, y), q(x, y) = uy (x, y).

Proof of (ii).
The I 0 defined by

I 0 := {s ∈ I : ( f (s), g (s )) ∈ F }

is a subinterval of I . Observe that T ( f (s), g (s)) = 0 and S ( f (s ), g (s)) = s

for s ∈ I 0 , in view of the definitions of the functions T and S.

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 97

Thus for s ∈ I 0 , we have

u ( f (s), g (s)) = Z (T ( f (s ), g (s )) , S ( f (s), g (s )))

= Z (0, s)
= h(s).

This shows that Cauchy condition u ( f (s), g (s)) = h(s) is satisfied for
all s ∈ I 0 , thus completing the proof of (ii).
Thus we have proved the existential assertion of the following local ex-
istence and uniqueness theorem. Proof of uniqueness is similar to the one
presented for Theorem 2.36

Theorem 2.59. Let Ω5 be an open and connected subset of R5 .

1. Let F : Ω5 → R be a twice continuously differentiable function.
2. Let the partial derivatives F p and Fq satisfy

F p2 (x, y, z, p, q) + Fq2 (x, y, z, p, q) 6= 0 ∀ (x, y, z, p, q) ∈ Ω5 .

3. Let Γ be a continuously differentiable parametrized curve: that is,

Γ : x = f (s), y = g (s), z = h(s) s ∈ I,

where I ⊆ R is an interval, and the functions f , g , h are continuously

differentiable on the interval I .
4. Assume that the system

F ( f (s), g (s), h(s), p(s ), q(s )) = 0 (2.97a)

p(s ) f 0 (s ) + q(s)g 0 (s) = h 0 (s). (2.97b)

admits a solution ( p(s ), q(s)) where p(s) and q(s ) are continuously dif-
ferentiable functions on the interval I such that the transversality condi-
tion holds at s = s0 :
 0 
 f (s0 ) F p ( f (s0 ), g (s0 ), h(s0 ), p0 , q0 ) 
 
∆= 0  6= 0.
 g (s ) F ( f (s ), g (s ), h(s ), p , q ) 
0 q 0 0 0 0 0

Then the Cauchy problem for the general nonlinear PDE

F (x, y, u, u x , uy ) = 0 (2.98)

with the Cauchy data Γ admits a solution defined on an open neighbourhood

August 1, 2021 Sivaji

98 Chapter 2. First order PDE

of the point ( f (s0 ), g (s0 )) ∈ Ω2 . Further the solution is locally unique.

Remark 2.60. (i) If the system (2.97) has no solutions for ( p(s), q(s )),
then the Cauchy problem for (2.98) does not have a solution.
(ii) If the system (2.97) has many solutions for ( p(s), q(s)), and the transver-
sality condition holds for any such solution, then the Cauchy problem
possesses at least as many solutions as the number of distinct solutions
( p(s), q(s )) to the system (2.97).

Example 2.61. Consider the Cauchy problem for the nonlinear equation

u x2 + uy2 = 1,

where the Cauchy data is given by u(x, y) = 0 for (x, y) such that x 2 + y 2 =
1.

Let us parametrize the given Cauchy data as

Γ : x = cos s, y = sin s, z = 0, s ∈ [0, 2π).

The system of ODEs for characteristic strips for the given equation is

dx dy dz dp dq
= 2 p, = 2q, = 2, = 0, = 0.
dt dt dt dt dt
To complete Γ into an initial strip, we need to find solutions ( p(s), q(s )) to
the system of equations

p(s)2 + q(s)2 = 1, p(s)(− sin s) + q(s )(cos s ) = 0.

The above system of equations has two solutions that are smooth functions
of s, namely,

( p1 (s), q1 (s)) := (cos s , sin s) and ( p2 (s), q2 (s)) := (− cos s, − sin s).

By taking the initial strip as (cos s , sin s, 0, cos s, sin s ), the solution of
the of ODEs for characteristic strips satisfying the initial conditions

x(0) = cos s , y(0) = sin s, z(0) = 0, p(0) = cos s, q(0) = sin s

is given by

Sivaji IIT Bombay

2.5. Cauchy problem for general nonlinear equation 99

X (t , s) = (2t + 1) cos s, Y (t , s ) = (2t + 1) sin s , Z(t , s) = 2t , (2.99a)

P (t , s ) = cos s, Q(t , s) = sin s. (2.99b)

From the equations (2.99a), we get

X 2 + Y 2 = (Z + 1)2 .

This implies that the function Z is given by

p
Z(t , s ) = −1 ± X 2 (t , s) + Y 2 (t , s).

The above equation suggests two possible solutions, given by

p
u1,2 (x, y) = −1 ± x 2 + y 2 .
p
Note that u1 (x, y) = −1 + x 2 + y 2 is the solution to Cauchy problem,
p
while the other function, namely, u2 (x, y) = −1 − x 2 + y 2 does not sat-
isfy the Cauchy data.

If we proceeded by taking the initial strip as (cos s, sin

ps, 0, − cos s, − sin s),
then we would have got the solution as u(x, y) = 1 − x 2 + y 2 .
Note that the given Cauchy problem is expected to have an even number
of solutions, since changing u to −u does not change the Cauchy problem.

2.5.6 General nonlinear equation in more than two variables

In this subsection, we discuss method of solution for Cauchy problem as-

sociated to a general nonlinear equation in d number of independent vari-
ables. It is instructive to compare the present case with that of d = 2 which
was presented in detail earlier. Since analysis is very similar to that of d = 2,
only main steps are presented here.
Consider a first order nonlinear equation in d -independent variables
x1 , x2 , · · · , xd given by

F (x1 , x2 , · · · , xd , u, u x1 , u x2 , · · · , u xd ) = 0, (d-GE)

where F is a C 2 function defined on a domain Ω2d +1 ⊆ R2d +1 . The argu-

ments of the function F are denoted by (x1 , x2 , · · · , xd , z, p1 , p2 , · · · , pd ).

August 1, 2021 Sivaji

Exercises 133

Exercises
Forming PDEs

2.1. Let f (x, y, a, b ) = (x − a)2 + (y − b )2 . Get a PDE by eliminating the

parameters a and b . (Answer: u x2 + uy2 = 4u.)

General solutions to quasilinear PDEs

2.2. Find general solutions to the following equations using change of

variables or Lagrange’s method.
(i) ut + c ux = 0
(ii) x u x + uy = 1
(iii) x u x + (y 2 + 1) uy = u
(iv) x uy − y u x = u
(v) x u u x + y u uy = u 2 − 1

Cauchy problem for quasilinear PDEs

You are given a list of Cauchy problems below. Γ stands for the da-
tum curve.
Solve each of the Cauchy problems, and also include answers to each
of the following aspects in the solution process.
1. What is the largest open set Ω on which a 2 + b 2 6= 0 and a, b , c ∈
C 1 (Ω).
2. Find the set Γ s , where Γ s denotes the set of all the points on Γ at
which transversality condition is satisfied.
3. Let Γ s2 denote the projection of Γ s to R2 .
(i) For Linear and Semilinear equations: Find base charac-
teristics through points of Γ s2 .
(ii) For Quasilinear equations: Find characteristic curves through
Γ s and then find base characteristics through points of Γ s2 .
(iii) Investigate if two distinct base characteristics intersect, and
find points of intersection. Identify the correspoinding points
on Γ .
(iv) Sketch a few of the base characteristics.
4. Find explcit formulae t = T (x, y), s = S(x, y), using x = X (t , s),
y = Y (t , s ). What are the domains of the functions S, T ?

August 1, 2021 Sivaji

134 Chapter 2. First order PDE

(i) Get an expression for solution u(x, y), and clearly mention
the domain on which it is defined.
(ii) Comment on the local/global nature of the solution ob-
tained above, w.r.t. datum curve, and also w.r.t. domain.
(iii) If the solution obtained was local w.r.t. datum curve or
local w.r.t. domain, then find reasons for the same.
5. In case finding explicit formulae as in the question 4 above is
impossible, then
(i) can you find a function φ : D ⊆ R3 → R such that the
equation
φ (X (t , s), Y (t , s), Z(t , s)) = 0
is satisfied for (t , s) in some neighbourhood containing (0, s)?
(ii) What are the points (x0 , y0 , z0 ) such that

∂φ
φ(x0 , y0 , z0 ) = 0, (x , y , z ) 6= 0 ?
∂z 0 0 0
Note that near such points as above, one can solve (by im-
plicit function theorem) for z as z = ψ(x, y) from φ(x, y, z) =
0. In other words, φ(x, y, z) = 0 is described locally as the
graph of ψ. That is, an integral surface is contained in the
set {φ = 0}, and φ(x, y, z) = 0 is called an implicit solution.
6. Find the set Γ f , where Γ f denotes the set of all the points on Γ at
which transversality condition is not satisfied. Is Γ f a curve, an
isolated point, or something else?
(i) What are your observations on the solutions that were ob-
tained in question 4(i) above in relation to the points of T f ?
(ii) If Γ f (or a part of it) is a curve, then find at least two solu-
tions if they exist. Otherwise give reasons for non-existence
of a solution.
2.3. The Cauchy conditions feature the variables x, y, t , we assume that
x, y, t ∈ R unless mentioned otherwise.
(i) u x = 1 such that u(0, y) = g (y).
(ii) u x + uy = 2 such that u(x, 0) = x 2 .
(iii) [30] u x + 2uy = 1 + u such that u = sin x on the line y = 3x + 1.
(Answer: u(x, y) = [sin(y − 2x + 1)] e 3x−y+1 − 1)
(iv) [24] x u x + uy = 1, u(x, 0) = e x .

Sivaji IIT Bombay

Exercises 135

(v) x u x + uy = 1, u(0, y) = e y .
(vi) [24] x u x + (y 2 + 1) uy = u, u(x, 0) = e x .
(vii) [24] x u x + (x + y) uy = u + 1 u(x, 0) = x 2 . (Ans. u(x, y) =
y y
x 2 exp(− x ) + exp( x ) − 1)
(viii) [44] x u x + y uy = u + 1, u(x, x) = x 2 . (Ans. No solution)
(ix) x u x + y uy = u + 1, u(x, x 2 ) = x 2 .
(x) u u x − u uy = u 2 + (x + y)2 , u(x, 0) = 1.
p y
(xi) x uy −y u x = u, u(x, 0) = h(x), x ≥ 0. (Ans. u = h( x 2 + y 2 )e arctan x )
(xii) [30] u x + 2x uy = 2x u, u(x, 0) = x 2 for all x ∈ R. (Ans. u =
(x 2 − y)e y )
(xiii) [30] u x + 2x uy = 2x u, u(0, y) = y 2 for all y ∈ R. (Ans. u =
2
(y − x 2 )2 e x )
(xiv) [30] u x + 2x uy = 2x u, u(x, 0) = x 2 for x ≥ 0 and u(0, y) = y 2
for y ≥ 0.
(xv) x u x − y uy = y such that u(1, y) = y, y ∈ R.
(xvi) [44] x u x + y uy = 2u such that u(x, 0) = x 2 , x > 0.
y

(xvii) [44] x u x + (x 2 + y)uy = 1 − x − x u such that u(1, y) = 0.
(xviii) u x + uy = u 2 , u(x, 0) = h(x).
(xix) x u x + uy = u 2 , u(x, 0) = h(x).
(xx) u x + uy = u 2 , u(x, −x) = x.
(xxi) x u u x + y u uy = u 2 − 1, u(x, x 2 ) = x 3 for x > 0. (Hint. Use a
change of variable v = u 2 − 1)
(xxii) [44] Solve x u u x + y u uy = x 2 + y 2 , x > 0, y > 0, with Cauchy
p p
data (i). u(x, 1) = x 2 + 1, (ii). u(x, x) = 2x.
2
(xxiii) (y + u)u x + y uy = x − y, u(x, 1) = 1 + x. (Ans. u = x − y + y )
(xxiv) uy = x u u x , u(x, 0) = x. (Ans. x = ue −y u )
(xxv) u u x + y uy = x, u(x, 1) = 2x.
2
2.4. Solve the equation u x + 3y 3 uy = 2 subject to the condition u(x, 0) =
x, x ∈ R. What happens if the Cauchy data is taken as u(x, 0) = 2x,
x ∈ R? Optional: Discuss whether the non-smoothness of character-
istic vector field affects the solution.
2.5. Solve the PDE u x + 3uy = u subject to Cauchy condition u = cos x
on the line y = αx. Find the value of α for which the method € 3x−y Šfails
y−αx

and interpret the result. (Answer: u(x, y) = exp 3−α cos 3−α )
2.6. [31](Picone) Let u be a continuously differentiable function in the

August 1, 2021 Sivaji

136 Chapter 2. First order PDE

closed unit disk B(0, 1) and is a solution of

a(x, y)u x + b (x, y)uy = −u in B(0, 1).

Let a(x, y)x + b (x, y)y > 0 on the unit circle. Prove that u vanishes
identically. (Hint: Show that max u ≤ 0 and min u ≥ 0.)
B(0,1) B(0,1)
2.7. [44] Let u be a continuously differentiable function on the closure
of D where D is given by

D = (x, y) ∈ R2 : |x| < 1, |y| < 1 .

Let u be a solution of

a(x, y)u x + b (x, y)uy = −u,

where a, b are positive differentiable functions in the entire plane.

Then
(a) Prove that every base characteristic curve passing through a point
in D intersects the boundary of D at exactly two points. (Hint:
Each base characteristic curve has a positive direction and it prop-
agates with a strictly positive speed in the square. Therefore, it
intersects the boundary of D exactly twice.)
(b) Show that if u is positive on the boundary of D, then it is pos-
itive at every point in D. (Hint: Suppose that u is positive on
the boundary of D and u ≤ 0 at some point in D. Consider the
base characteristic passing through this point. Consider what
happens to solution along this characteristic and get a contra-
diction.)
(c) Suppose that u attains a local minimum (maximum) at a point
(x0 , y0 ) ∈ D. Evaluate u(x0 , y0 ). (Hint: use your calculus)
(d) Denote by m the minimal value of u on the boundary of D.
Assume m > 0. Show that u(x, y) ≥ m for all (x, y) ∈ D. (Hint:
If this were not true, then u has a global minimum in D and
hence it is a local minimum. Use (c) and (b), and conclude.)
2.8. Using Lemma 2.49, conclude that the Cauchy problem considered in
Example 2.28 has no solution.
2.9. Consider the first order quasilinear PDE given by

u u x + uy = 1. (2.130)

Sivaji IIT Bombay

Exercises 137

Solve the Cauchy problems associated with (2.130) and each of the
following Cauchy data.

(i) (Jacobian is non-vanishing)

s
x = s, y = s, z = , 0 < s < 1.
2
(ii) (Jacobian is identically equal to zero, infinite number of solu-
tions to Cauchy problem)

s2
x= , y = s, z = s, 0 < s < 1.
2
Find all solutions and sketch some of them.
(iii) (Jacobian is identically equal to zero, no solutions to Cauchy
problem)

x = s 2 , y = 2s, z = s, 0 < s < 1.

Try solving this problem and see what happens.

(iv) (Jacobian is identically equal to zero at a few points)

x = s 2 , y = 2s, z = h(s), s ∈ R,
1 π

where (a). h(s) = 2 , (b). h(s) = sin 3 s .
Try solving this problem and see what happens.
(v) (Jacobian is identically equal to zero on a subset)

0 for 0 ≤ s ≤ 1,
x = 0, y = s, z =
(s − 1) for 1 ≤ s ≤ 3.
3

Try solving this problem and see what happens.

2.10. Solve the PDE uy + u u x = 0 subject to Cauchy condition

1
x = s 2 , y = 2s, z = , s ∈ R.
2
Is the solution unique?
2.11. Solve the PDE u t + f (u)u x = 0 subject to Cauchy condition u(x, 0) =
g (x), where f is a continuously differentiable function and g is a
continuous function. Also find an expression for u x (x, t ) and hence
find a condition that must be satisfied that ensures the existence of

August 1, 2021 Sivaji

138 Chapter 2. First order PDE

u x (x, t ) for all t > 0.

1
2.12. Solve the PDE u u t + u x = 0 subject to Cauchy condition u(x, 1) = x
for x > 1.
2.13. Solve the PDE u t + u u x = t subject to Cauchy condition u(x, 0) =
−2x. Find the domain in the upper half of the (x, t )-plane t ≥ 0
where the solution is valid.
2.14. Solve the PDE u t + u u x = 0 subject to Cauchy condition u(x, 0) = x
for 1 < x < 2. Find the domain in the upper half-plane t ≥ 0 where
the solution is valid.
2.15. Solve the PDE u t + u u x = 0 subject to Cauchy condition u(x, 0) =
−x for 1 < x < 2. Find the domain in the upper half-plane t ≥ 0
where the solution is valid.
2.16. According to Remark 2.17, two distinct base characteristic curves
may intersect. On the other hand, base characteristic curves are the
traces of solutions to the first two equations of the characteristic sys-
tem of ODEs (chara.ODE), whose right hand side functions a, b are
locally Lipschitz functions. Do these two statements contradict the
assertion of Cauchy-Lipschitz-Picard theorem? Justify your answer.
2.17. Suppose γ is a curve that lies on two integral surfaces corresponding
to the quasilinear PDE (QL). Can we conclude that γ is a character-
istic curve? If yes, give a proof. If not, explain where the proof of
Corollary 2.33 fails.
2.18. Deduce Corollary 2.33 from Theorem 2.29.
2.19. Explore the reasons behind the assertion of only local existence of so-
lutions to Cauchy problem in Theorem 2.36. Illustrate each of your
reasons with an example.
2.20. Assume that J ≡ 0, and Cauchy problem has no solution. Explain
why the procedure given in the proof of Lemma 2.49 to construct a
solution fails.

Nonlinear equations

2.21. Find if the following Cauchy problems have solutions, and find them
whenever they exist. In each case determine the domains on which
solutions exist and also comment on uniqueness of solutions. In
Cauchy conditions x ∈ R unless stated otherwise.
(i) u x uy = 1, u(x, 0) = log x, x > 0.
(ii) u x uy = 2, u(x, x) = 3x, 0 < x < 1. (Ans. u = y + 2x)

Sivaji IIT Bombay

Exercises 139

(iii) u x uy = u, u(x, 1) = x, 0 < x < 1.(Ans. u = xy)

(iv) u x2 + uy2 + 2(u x − x)(uy − y) − 2u = 0, u(x, 0) = 0, 0 < x < 1.
3 1
(Ans. u = 2xy − 2 y 2 , u = 2 y 2 )
p p
2x x
(v) u x3 − uy = 0, u(x, 0) = 2x x, 0 < x < 1. (Ans. u = p )
1−28y
1 2 y2
(vi) u x + u
2 y
= 1, u(0, y) = y , 0 < y < 1. (Ans. u = x + 1+2x )
2

u 2 +u 2 1−x 2 1−x 2
(vii) u = x u x + y uy + x 2 y , u(x, 0) = 2 . (Ans. u = ±y + 2 )
x x
(viii) 2x u x2 + y uy = u, u(x, 1) = 2 . (Ans. u = 2 )
(ix) u x2 + uy2 = 1, u(x, 0) = 0.
(x) u x2 + uy2 = n02 , u(x, 2x) = 1.
(xi) u x2 + uy2 = u, u(x, 0) = x 2 + 1.
(xii) u x2 +uy2 = u with the initial curve Γ given by the parametrization
(cos s, sin s , 1).
(xiii) u x2 + uy2 = u 2 , u(x, 0) = 1. (Ans. u = exp(±y))
(xiv) u x2 + uy2 = u 2 with the initial curve Γ given by the parametriza-
p
tion (cos s, sin s, 1).(Ans. u = exp(±(1 − x 2 + y 2 )))
(xv) u x2 + uy = 0, u(x, 0) = x.
(xvi) uy2 − u x = 0, u(x, 0) = f (x).
3
(xvii) u x4 + uy = 0, u(x, 0) = 4 x 4/3 , y > 0.
(xviii) uy + u 2 u x = 0, u(x, 0) = x.
y
(xix) u x2 − 3uy2 = u, u(x, 0) = x 2 . (Ans. u = (x + 2 )2 if q0 (s) = s,
y
and u = (x − 2 )2 if q0 (s) = −s)

2.22. Prove Proposition 2.58 .

2.23. Formulate a version of Lemma 2.49 for (GE) and prove it.
2.24. Write a short essay on the proofs of Theorems 2.36 and 2.59 high-
lighting the similarities and differences.

Conservation laws

2.25. Examine whether the weak solutions given in Example 2.76, and Ex-
ample 2.77 are entropy solutions.

August 1, 2021 Sivaji

 Appendix B

Envelopes

The idea of an envelope of a family of planes plays an important role in obtaining a solu-
tion to Cauchy problem for general nonlinear partial differential equations, see Section 2.5
for details.
In Section B.1, envelopes of one-parameter families of surfaces are studied. Envelopes
of families of curves are studied in Section B.2. In Section B.3, we discuss some of the mis-
conceptions that arise by expressing mathematical statements in plain language without
exercising caution.

B.1 Envelopes of families of surfaces

Consider the family of surfaces given by

Sλ : z = G(x, y; λ) (B.1)

where G is a differentiable function of three variables, and λ is the parameter. Differenti-

ating both sides of the equation (B.1) w.r.t. λ, we get

0 = Gλ (x, y; λ). (B.2)

For each fixed λ, let Cλ denote the curve of intersection of the surfaces given by (B.1)
and (B.2). That is, Cλ is described by

Cλ : {(x, y, z) : z = G(x, y, λ), Gλ (x, y, λ) = 0} . (B.3)

Definition B.1. The envelope E of the one-parameter family of surfaces (B.1) is defined as

E := ∪λ Cλ .

Example B.2. Let us compute the envelope of the family of tangent planes (2.71) for quasilinear

385
386 Appendix B. Envelopes

equation (QL) ( p ∈ R is a parameter)

 
c(P0 ) a(P0 )
z − z0 = (y − y0 ) + p x − x0 − (y − y0 ) . (B.4)
b (P0 ) b (P0 )

Differentiating the equation (B.4) w.r.t. p yields

a(P0 )
x − x0 − (y − y0 ) = 0. (B.5)
b (P0 )

Thus, for each p ∈ R the curve of intersection C p of the surfaces (B.4) and (B.5) is given by
 
c(P0 ) a(P0 )
C p = (x, y, z) : z − z0 = (y − y0 ), x − x0 − (y − y0 ) = 0
b (P0 ) b (P0 )
= {(x, y, z) : c(P0 )(y − y0 ) = b (P0 )(z − z0 ), b (P0 )(x − x0 ) = a(P0 )(y − y0 )}
 
x − x0 y − y0 z − z0
= (x, y, z) : = = .
a(P0 ) b (P0 ) c(P0 )

Note that C p is the same straight line for each p ∈ R, which is a line passing through P0 having
the direction (a(P0 ), b (P0 ), c(P0 )). Note that this is precisely the characteristic direction for the
quasilinear equation (QL).

Assuming that we can write λ = g (x, y) from the equation Gλ (x, y; λ) = 0, we may
express the envelope of Sλ by the equation

z = G(x, y, g (x, y)), (B.6)

by substituting for λ in the equation (B.1).

Lemma B.3. Let Sλ : z = G(x, y; λ) be a family of surfaces indexed by the parameter λ

where G is a differentiable function. Let E denote the envelope of the family Sλ described by
equation (B.6). Let Cλ be defined by equation (B.3). Then

(i) Assume that Cλ 6= ; for each λ. Then the envelope E of the family of surfaces Sλ intersects
every member of the family Sλ along Cλ .
(ii) The envelope E and Sλ touch each other i.e., at each point of E ∩ Sλ , which is nothing
but Cλ , the envelope E and Sλ have a common tangent plane.

Proof. Proof of (i): Let (x, y, z) ∈ E. By definition of the envelope, there exists a λ such
that (x, y, z) ∈ Cλ . Since Cλ is a subset of Sλ , (i) follows.
Proof of (ii): Proving (ii) is equivalent to proving that at each point of Cλ , the envelope
E and Sλ have the same normal direction. Let (x, y, z) ∈ Cλ . Normal direction to Sλ at
(x, y, z) is given by (Gx (x, y, g (x, y)), Gy (x, y, g (x, y)), −1), and normal direction to the
envelope E : z = G(x, y, g (x, y)) is given by

(Gx (x, y, g (x, y)) + Gλ (x, y, g (x, y))g x (x, y), Gy (x, y, g (x, y)) + Gλ (x, y, g (x, y))gy (x, y), −1)
= (Gx (x, y, g (x, y)), Gy (x, y, g (x, y)), −1)

Sivaji IIT Bombay

B.2. Envelopes of families of curves 387

since (x, y, z) ∈ Cλ implies Gλ (x, y, z) = 0.

Example B.4. Consider the PDE

u x2 + uy2 = 1,
which is of the form F (x, y, u, u x , uy ) = 0, where F (x, y, z, p, q) = p 2 + q 2 − 1. As discussed in
Subsection 2.5.2, the possible tangent planes (to possible integral surfaces for the given PDE ) at the
point P0 (0, 0, 0), are as many as the number of pairs ( p, q) satisfying the equation p 2 + q 2 − 1 = 0.
Thus the family of possible tangent planes at the point P0 (0, 0, 0) is given by the two equations

z = px + qy (B.7a)
p2 + q2 − 1 = 0 (B.7b)
p
From the equation (B.7b), we express q = ± 1 − p 2 . Thus we have the one-parameter family of
planes given by p
z = p x ± 1 − p2 y (B.8)
In order to find the envelope of the family of planes (B.8), we differentiate the equation (B.8) w.r.t.
p, which yields
p
0=x∓ p y (B.9)
1 − p2
From the equation (B.9), we get
1
p(x, y) = ± p . (B.10)
x2 + y2
Substituting the expression for p from (B.10) into (B.7a) gives the equation of the envelope as

z 2 = x 2 + y 2 , |z| > 1.

Thus the envelope lies on the double-sheeted cone with vertex at the origin. Figure B.1 depicts the
double-sheeted cone, and a few members of the family of planes.

B.2 Envelopes of families of curves

Consider the family of curves in the x y-plane given implicitly by the equation

γλ : F (x, y; λ) = 0 (B.11)

where F is a differentiable function of three variables, and λ is the parameter. Differenti-

ating both sides of the equation (B.11) w.r.t. λ, we get

Fλ (x, y; λ) = 0. (B.12)

For each fixed λ, let Pλ denote the points of intersection of the curves given by (B.11)

July 28, 2021 Sivaji

388 Appendix B. Envelopes

Figure B.1. Example 2.61: Cone, the envelope of possible tangent planes passing through the origin.

and (B.12). That is, Pλ is described by

Pλ : {(x, y) : F (x, y; λ) = 0, Fλ (x, y; λ) = 0} . (B.13)

Definition B.5. The envelope E of the one-parameter family of curves (B.11) is defined as

E := ∪λ Pλ .

Assuming that we can write λ = f (x, y) from the equation Fλ (x, y; λ) = 0, we may
express the envelope of γλ by the equation

F (x, y, f (x, y)) = 0, (B.14)

by substituting for λ in the equation (B.11).

We now state a result analogous to Lemma B.3 without proof, since it can be proved
by following arguments presented in the proof of Lemma B.3, and is left to the reader.

Lemma B.6. Let γλ : F (x, y; λ) = 0 be a family of curves indexed by the parameter λ

where F is a differentiable function. Let E denote the envelope of the family γλ described by
equation (B.14). Let Pλ be defined by equation (B.13). Then

(i) Assume that Pλ 6= ; for each λ. Then the envelope E of the family of curves γλ intersects
every member of the family γλ along Pλ .
(ii) The envelope E and γλ touch each other i.e., at each point of E ∩ γλ , which is nothing
but Pλ , the envelope E and γλ have a common tangent.

Sivaji IIT Bombay

B.3. Some misconceptions regarding envelopes 389

Example B.7. Let a > 0, and m denote a parameter defining the family of straight lines
a
L m : y − mx − =0
m
a
Differentiating the above equation w.r.t. m yieldsm 2 = x . Thus for x > 0, we can solve
Æa
for m as m = x . Now substituting the expression for m in the equation for L m , we
get the equation y 2 = 4a x. Thus, the envelope of the given family of lines is the parabola
y 2 = 4ax.

Example B.8. For G(x, y, λ) ≡ ax + b y + c + λ(a1 x + b1 y + c1 ) = 0, the envelope is the

point of intersection of the lines

ax + b y + c = 0, a1 x + b1 y + c1 = 0

B.3 Some misconceptions regarding envelopes

We cannot make blanket assertions like “Envelope of a one-parameter family of planes is
a cone” as they are simply not true. This was illustrated by Example B.2, in which the
envelope of a one-parameter family of planes turns out to be a straight line. This is also
the case in the next example.

Example B.9. For G(x, y, λ) ≡ a x + b y + c z + d + λ(a1 x + b1 y + c1 z + d1 ) = 0, the

envelope is the point of intersection of the planes

a x + b y + c z + d = 0, a1 x + b1 y + c1 z + d1 = 0.

Similarly, envelopes of a family of straight lines can be a parabola (Example B.7), and
can also be a point (Example B.8).
Thus envelope depends on the manner in which the parameter appears in the defini-
tions of families of surfaces or curves.
In the literature there are at least three ways of defining the notion of an envelope.
Their inter-relations may be found in a clearly written article by Kock [34]. Note that for
our purposes, the notion that we defined is good enough as it is used only in guessing a
characteristic direction in Section 2.5.

Exercises
p
2.1. Find the envelope of the family of lines given by y = m x ± a 1 + m 2 , where m is
a parameter, and a is fixed constant.

July 28, 2021 Sivaji

390 Appendix B. Envelopes

2.2. Find the envelope of the family of planes given by

Æ
(cos β)x + 1 − cos2 β − sin2 αy + (sin α)z = 0

where α is a constant and β is a real parameter. (Answer: (tan2 α)z 2 = x 2 + y 2 .)

2.3. Find the envelope of the family of surfaces
Æ
G(x, y, z, λ) ≡ λx + 1 − λ2 − µ2 y + µz = 0

where µ is a constant and λ is a real parameter. (Answer: (1−µ2 )(x 2 +y 2 )−µ2 z 2 =

0.)

Sivaji IIT Bombay