5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 1

5 First order Partial Differential Equations

5.1 Transport Equation

In this section we will study the transport equations and its solutions. Consider a fluid flowing
through with velocity V in a thin straight tube whose cross section is A. Suppose fluid contain
a contamination whose concentration at position x at time t is u(x, t). Then at time t, the
amount of contaminant in a section of tube between x1 and x2 is
Z x2
u(x, t)Adx
x1

Similarly amount of contaminant that flows through a plane cross section at x during the
time between t1 and t2 is Z t2
u(x, t)AV dt
t1

Therefore the equation of continuity implies

Z x2 Z x2 Z t2 Z t2
u(x, t2 )Adx − u(x, t1 )Adx = u(x1 , t)AV dt u(x2 , t)AV dt
x1 x1 t1 t1

By fundamental theorem we see this equal to

Z x2 Z t2 Z t2 Z x2
∂t u(x, t)Adtdx = − ∂x u(x, t)AV dt
x1 t1 t1 x1

This is true for all intervals (x1 , x2 ) and times t1 , t2 if

∂t (uA) + ∂x (uAV ) = 0.

This is known as transport equation.

Initial value problem: Consider the problem of finding u(x, t) satisfying

ut + aux = 0, t ≥ 0, x ∈ R
u(x, 0) = h(x), x ∈ R

where a ∈ R is a constant.
We see this equation as directional derivative of a function u(x, t) in the direction (1, a).
So if we consider the function g : R → R2

g(t) = (x0 + at, t),

5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 2

then F (t) = u(g(t)) : R → R satisfies

F 0 (t) = ux a + ut = 0

Therefore F (t) = F (0) = u(g(0)) = u(x0 , 0). Therefore the solution u(x, t) is

u(x, t) = h(x − at)

Figure 1:

So we see that we can try to integrate along the directions along which the directional deriva-
tive is constant. Using this we can also solve non-homogeneous problem like

ut + aux = f (x, t), t ≥ 0, x ∈ R

u(x, 0) = h(x)

In this case, we write the parametric form of lines with slope a as

x(s) = x + as, t(s) = t + s

represents a line in R2 : s → (x(s), t(s)) passes through (x, t). Then define

F (s) = u(x(s), t(s))

5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 3

We immediately see that

F 0 (s) = ux x0 (s) + ut t0 (s) = aux + ut = f (x(s), t(s)) = f (x + as, t + s)

This implies
Z 0 Z t
F (0) − F (−t) = f (x + as, t + s)ds = f (x + (s − t)a, s)ds
−t 0

On the other hand

F (0) − F (−t) = u(x, t) − u(x(−t), t(−t))

= u(x, t) − u(x − at, 0)
= u(x, t) − h(x − at)

Therefore, Z t
u(x, t) = h(x − at) + f (x + (s − t)a, s)ds, x ∈ R, t ≥ 0.
0

5.2 Method of characteristics

Consider the partial differential equation in two variables x, y

aux + buy = c (5.1)

where a, b, c are continuous functions in x, y, u. Let u(x, y) be the solution and let z = u(x, y)
be the graph of u. That is, the level surface S : u(x, y) − z = 0. Let z0 = u(x0 , y0 ). The
normal to this surface at a point (x0 , y0 , z0 ) is N = (ux , uy , −1) at (x0 , y0 ).

The equation in (5.1) implies that the vector V0 = (a, b, c) is perpendicular to the normal N .
Hence V0 must lie on the tangent plane to the surface z = u(x, y). So our aim is to find the
surface z = u(x, y) knowing that the vector field {a(x, y, z), b(x, y, z), c(x, y, z)} lies on the
tangent place. Such surface is called Integral surface.

Cauchy Problem: Given a curve Γ in R3 , find a function u(x, y) satisfying (5.1) such that
the level surface z = u(x, y) contains Γ. In other words, find u(x, y) satisfying

aux + buy = c in U
(5.2)
u(x, y) = h(x, y) on γ

where U is open domain that contains the curve γ.

5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 4

Figure 2: Integral surface

Let us assume that the surface S is parametrizable and

S = {(x(s, t), y(s, t), z(s, t))}

with t ≥ 0 and s ∈ I ⊂ R.
Also let the initial space curve is Γ = {(x(s, 0), y(s, 0), z(s, 0))}. Therefore from the given
initial condition u(x, y) = h on γ implies z(s, 0) = h(x(s, 0), y(s, 0)) = h(s). We may assume
that x(s, 0) = f (s) and y(s, 0) = g(s).

Since (a, b, c) is perpendicular to the normal, it is proportional to the tangent vector. So it is

natural that it satisfies the system of equations with initial conditions

dx
= a(x(s, t), y(s, t)z(s, t)), x(s, 0) = f (s)
dt
dy
= b(x(s, t), y(s, t), z(s, t)), y(s, 0) = g(s)
dt
dz
= c(x(s, t), y(s, t), z(s, t)), z(s, 0) = h(s)
dt

We can solve this system of equations uniquely(Picard’s theorem) for all small t under the
assumption that (a, b, c) are C 1 functions. The curves so obtained are called characteristic
curves. We will find the surface as union of these curves obtained from the above ODE system.
Then at each point (x0 , y0 , z0 ) the tangent plane contains the vector V (x0 , y0 , z0 ) = (a, b, c).
In otherwords, smooth union of characteristic curves is the Integral surface.
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 5

Now to obtain the surface in the variables x, y we need to invert the map (x, y) → (s, t).
For this we use Inverse function theorem. The above map is invertible near t = 0 if the
Jacobian
 
 ∂x ∂y 
∂s ∂s
J =  6= 0 at t = 0.
 
 ∂x ∂y 
∂t ∂t

That is, b f 0 (s) − a g 0 (s) 6= 0.

Definition 5.2.1. Non-characteristic curve: A curve γ is called non-characteristic if b f 0 (s)−

a g 0 (s) 6= 0..

We proved the following

Theorem 5.2.2. Let γ be a non-characteristic curve and let a, b, c are C 1 functions. Then
the Cauchy problem in (5.2) has a solution in a neighbourhood of the initial curve γ.

Example 5.2.3. Solve the Initial value problem:

2ux + 3uy = 0, (x, y) ∈ R2 , u(x, 0) = g(x), x ∈ R.

The parametrization of initial curve γ = {(s, 0, g(s)), s ∈ R}. The Jacobian is

 
 1 0 
J =  6= 0.
 
 2 3 

So the solution exists near the initial curve γ. The characteristic equations are

dx
= 2, x(s, 0) = s
dt
dy
= 3, y(s, 0) = 0
dt
dz
= 0, z(s, 0) = g(s)
dt

The solution is
x(s, t) = 2t + s, y(s, t) = 3t, z(s, t) = g(s)

Inverting the variables we get

y 2
t= , and s = x − y
3 3
Hence u(x, y) = z = g(s) = g(x − 23 y).

Example 5.2.4. Characteristic problem:

Consider the problem ux + uy = 0, u = g on γ. Then show that
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 6

(a) Solution exists uniquely if γ is not parallel to y = x.

(b) If γ is parallel to y = x, then solution exists if and only if g is constant.

Solution: (a): If γ is parallel to y = x, then f 0 (s) = g 0 (s) = 1 and the Jacobian

 
 1 1 
J =  = 0.
 
 1 1 

(b) : Let us solve the problem with u(x, 0) = g(x). Then

dx
= 1, x(s, 0) = s
dt
dy
= 1, y(s, 0) = 0
dt
dz
= 0, z(s, 0) = g(s)
dt

Then the solution is x(s, t) = t + s, y(s, t) = t, z(s, t) = g(s). Hence

u(x, y) = z = g(x − y)

Therefore on y = x, u(x, x) = g(0). That is, solution of the problem exists if and only
u(x, x) = g(x)= constant. In this case note that g(x − y) solves the equation ux + uy = 0 for
any function g. If the initial condition is u(x, x) = 5 then 5 + (x − y)k , k ≥ 0 are all solutions.
So the characteristic problem does not admit solution for all initial values. If solution exists,
then there are infinitely many solutions.

Semi linear equations

The first order equation aux + buy = c is called semi linear equation if a = a(x, y), b = b(x, y)
and c = c(x, y, u).

Example 5.2.5. Solve the Cauchy problem

ux + 2uy = u2 , x ∈ R, y > 0 u(x, 0) = h(x).

A parametrization of initial curve is {(s, 0, h(s)), s ∈ R}. The characteristic equations are
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 7

dx
= 1, x(s, 0) = s
dt
dy
= 2, y(s, 0) = 0
dt
dz
= z 2 , z(s, 0) = g(s)
dt

The Jacobian J is    
 x y   1 2 
 s s  
J = = =6 0.

 x t yt   1 0 

Therefore
−1 1
y = 2t, x = t + s, and =t−
z h(s)
Inverting the variables, we get

y y h(s)
t= , s = x − , and z = .
2 2 1 − th(s)

The characteristic lines are y = 2x + 2s and the solution is

h(x − y2 )
u(x, y) = z = )
1 − y2 h(x − y2

y y
which is defined upto 1 − h(x − ) 6= 0. ///
2 2
Example 5.2.6. Characteristic problem:
Consider the initial problem

x2
 
2
ux + xuy = u , x, y ∈ R u x, = g(x).
2
c
Show that solution exists if and only if g(x) = 1−cx
2
A Parametrization of the initial curve is {(s, s2 , g(s)) : s ∈ R}. The Jacobian J is equal to
   
 x y   1 s 
 s s  
J = =  = 0.

 xt yt   1 s 

So the problem is characteristic problem. As earlier let us consider the initial condition

u(0, y) = h(y), y ∈ R

This space curve can be parametrized as {(0, s, h(s)); s ∈ R}. Then the The characteristic
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 8

equations are

dx
= 1, x(s, 0) = 0
dt
dy
= x, y(s, 0) = s
dt
dz
= z 2 , z(s, 0) = h(s)
dt

Solving this we get

x2 h(s)
t = x, y = + s, and z =
2 1 − th(s)
Inverting the variables we get

x2
h(y − 2 )
u(x, y) = z = 2 .
1 − xh(y − x2 )

x2 2 h(0)
Therefore on y = 2 , u(x, x2 ) = 1−xh(0) . Hence solution of the given problem exists if

c
g(x) = .
1 − xc

In this case there are infinitely many solutions exists, a family of them is

x2
c + f (y − 2 )
u(x, y) = x2
1 − x(c + f (y − 2 )

for any f with f (0) = 0. ///

5.3 Quasilinear Equations

The most important equation that appears in the study of fluid flows is Burgers equation

uux + ut = 0 (5.3)

This is seen as an idealistic case of 1-d Navier Stokes equations with no pressure gradient

∂v ∂v ∂2v
ρ + ρv −µ 2 =0
dt ∂x ∂x

where ρ is the density, v is velocity and µ is the viscosity of fluid. When the viscosity of the
fluid is assumed to be almost zero, this would lead to an equation of the form (5.3). In this
case the initial value problem looks like

uux + uy = 0, x, y ∈ R, u(x, 0) = h(x), x ∈ R

5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 9

The characteristic equations are

dx
= z, x(s, 0) = s
dt
dy
= 1, y(s, 0) = 0
dt
dz
= 0, z(s, 0) = h(s)
dt

Solving this we get

y = t, z = h(s) and x = h(s)t + s
dx
The characteristic lines are x = h(s)y + s and the speed is dy = h(s). Inverting the variables
we get
t = y, s = x − zy,

and the solution can be defined implicitly as

u(x, y) = z = h(x − zy) = h(x − uy)

Now let γ1 be the characteristic curve at s1 with starting speed as h(s1 ) and let γ2 be the
characteristic curve at s2 with starting speed h(s2 ). If these curves intersect, then

s1 + h(s1 )y = s2 + h(s2 )y

Therefore
s1 − s2
y=t=
h(s2 ) − h(s1 )
So y > 0 if s1 > s2 and h(s2 ) > h(s1 ). That is if h is decreasing function, then the character-
istic lines intersect at a point y > 0. This phenomena is called Gradient catastrope.

Figure 3:

That is if h0 (s0 ) < 0 at any s0 , then the solution does not exist ”globally”. On the other
hand if h0 (s0 ) ≥ 0, then solution exists globally. Given the original physical model, we can
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 10

interpret as: If the initial velocity of the fluid flow form a non-decreasing function of position,
then the fluid moves out in a smooth fashion. If the initial velocity is decreasing function,
then the fluid flow undergo a ”Shock” that correspond to collision of particles. That is the
integral surface folds itself.

Example 5.3.1. Solve the problem

uux + yuy = x, x, y ∈ R, u(x, 1) = 2x, x ∈ R

A parametrization of Γ is {(s, 1, 2s) : s ∈ R}. The Jacobain is

   
 x y   2s 1 
 s s  
J = =  = 1.

 x t yt   1 0 

The characteristic equations are

dx
= z, x(s, 0) = s
dt
dy
= y, y(s, 0) = 1
dt
dz
= x, z(s, 0) = 2s
dt

The solution of this problem is

y(s, t) = c(s)et , y(s, 0) = 1 =⇒ y = et

d d
(x + z) = x + z, and (x − z) = z − x
dt dt
Therefore
x + z = 3set , x − z = −set

Simplifying this we get

3 1 3 1
x = set − se−t , z = set + se−t , y = et
2 2 2 2

Hence
3y 2 + 1
u(x, y) = x
3y 2 − 1
We have the following important example where the initial condition is only continuous func-
tion.
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 11

Example 5.3.2. Find the solution of the initial value problem near the initial curve:



 1 x<0

uy + uux = 0, x ∈ R, y > 0, u(x, 0) = h(x) = 1 − x x ∈ (0, 1)


0

x>1

The characteristic lines are y(s, t) = t and

x(s, t) = s + th(s)



 s+t s≤0

= s + t(1 − s) s ∈ (0, 1),


s s≥1


For y < 1, the characteristic lines does not intersect. So given a point (x, y) with y < 1, we
can draw the backward through characteristics



 x−y x<y

s = x−y
1−y y≤x<1


x x≥1


The characteristic lines are show below.

Figure 4: local solution

The solution for y < 1 is computed as follows:

s = x − y, s ≤ 1 =⇒ x ≤ y

x−y
s = x − y + ys, s ∈ (0, 1) =⇒ s =
1−y
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 12

s = x, s ≥ 1 =⇒ x ≥ 1

Therefore, 
1

 x<y<1

u(x, y) = h(s) = 1 − x−y
1−y y≤x≤1


0 x≥1


At the point y = 1 we see that u is


1 x < 1
u(x, y) =
0 x > 1

Propagation of discontinuities

There are situations where the initial condition has a discontinuity. In this case we cannot
expect the solution to be classical (C 1 ) solution. One such problem is Riemann problem. To
understand the propagation of the discontinuity along the characteristic curve let us consider
the following problem:

Example 5.3.3. Consider the initial value problem

uy + uux = 0, x ∈ R, y > 0,

1 x < 0,
u(x, 0) = h(x) =
0 x > 0.

The characteristic curves are


y + s, s < 0,
x = h(s)y + s =
s s>0

The characteristic lines intersect for all x > 0.

If we choose the initial condition as

0 x < 0
h(x) =
1 x > 0

Then the characteristic lines are


s, s < 0,
x = h(s)y + s =
y + s s>0
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 13

Figure 5: propagation of discontinuity and shock

Figure 6: propagation of discontinuity and fan

The characteristic lines does not intersect. It is important to understand how one can define
the solution in this type of situations. We can use integration by parts to define the so called
generalized solutions or weak solutions.
We recall the following integration by parts formula: Let Ω ⊂ R2 be an open set. Then
Z Z Z
uvx1 dx = uvη1 ds − vux1 dx
Ω ∂Ω Ω

where η1 is the first component of the unit outward normal η̂ = (η1 , η2 ) to ∂Ω and the first
integral on the right hand side is the line integral. We will study the weak solutions in the
later section.

General solution

Theorem 5.3.4. Suppose there exist functions φ and ψ such that φ(x, y, z) and ψ(x, y, z) are
constant along
dx dy dz
= = . (5.4)
a b c
Then F (φ, ψ) = 0 is the general solution of aux + buy = c, where F is such that Fφ2 + Fψ2 6= 0.

Proof. Let x(t), y(t), z(t) be the solution of (5.4) and let φ(x(t), y(t), z(t)) = c1 , ψ(x(t), y(t), z(t)) =
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 14

c2 for some constants c1 and c2 . Then the differential of φ and ψ are zero. Therefore,

φx dx + φy dy + φz dz = 0
ψx dx + ψy dy + ψz dz = 0.

Therefore
a −b c
= = (5.5)
φy ψz − ψy φz φx ψz − φz ψx φx ψy − φy ψx
Since z = z(x, y), we have F = F (φ, ψ) = 0 is a function of x, y. Therefore its total derivative
is zero. That is
dF = Fx dx + Fy dy = 0 =⇒ Fx = 0, Fy = 0

But by chain rule,

Fx = Fφ (φx + φz zx ) + Fφ (ψx + ψz zx ) = 0

Fy = Fφ (φy + φz zy ) + Fφ (ψy + ψz zy ) = 0.

Since (Fφ )2 + (Fψ )2 6= 0, the above system has non-trivial solution. That is
 
 φ +φ z ψ +ψ z 
 x z x x z x
 = 0.


 φy + φz zy ψy + ψz zy 

Then again thank s to (5.5) this is equal to saying azx + bzy = c. ///

Example 5.3.5. Find the general solution of uux + yuy = x.

The equations in (5.4) imply

dx dy dz
= =
z y x
i.e.,
xdx − zdz = 0 =⇒ d(x2 − y 2 ) = 0

Therefore φ(x, y, z) = x2 − z 2 .
Also

xdy − ydx = xy − yz = yx − zy = ydz − zdy =⇒ (x + z)dy − y(dx + dz) = 0

y y
Therefore d( ) =⇒ ψ = .
x+z x=z
Therefore the general solution is

y
u2 = x2 + f ( ).
x+u

Here we point out that the general solution does not represent all solutions.
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 15

5.4 Nonlinear Equations

In this section we will study the nonlinear equation

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

We introduce two more variables p = ux , and q = uy . So we have the five variable function

F (x, y, z, p, q) = 0

Suppose F has the quasilinear form

F = a(x, y, z)p + b(x, y, z)q − c(x, y, z) = 0

Then

dF dx
=a =
dp dt
dF dy
=b =
dq dt
dF dF dz
p +q = ap + bq = c =
dp dq dt

Taking this as motivation, we write three equations

dx dy dz
= Fp , = Fq , = pFp + qFq (5.6)
dt dt dt

We nee equations satisfied by p and q as well. For this, we notice

dp d
= ux (x, y) = uxx x0 (t) + uxy y 0 (t) = Fp px + Fq qx
dt dt

But we don’t want px and qx . Differentiating the equation F = 0 with respect to x, we get

Fx + Fz zx + Fp px + Fq qx = 0

That is
Fp px + Fq qx = −Fx − Fz p

Therefore
dp dq
= −Fx − pFz , and = −Fy − Fz q (5.7)
dt dt
These five equations in (5.5) and (5.7) form characteristic strip. The initial condition u(γ) = h
gives
x(s, 0) = f (s), y(s, 0) = g(s), z(s, 0) = h(s)
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 16

The initial values for of p and q are obtained as follows:

u(f (s), g(s)) = h(s) =⇒ h0 (s) = ux f 0 (s) + uy g 0 (s) = pf 0 + qg 0

Therefore p0 and q0 must satisfy ”strip condition”

p0 f 0 (s) + q0 g 0 (s) = h0 (s) (5.8)

But such p0 and q0 may not be few. There could be infinitely many. Also

F (f (s), g(s), h(s), p0 (s), q0 (s)) = 0 (5.9)

be satisfied on the initial curve. So p0 , q0 are such that (5.8) and (5.9) holds.

Example 5.4.1. Solve the initial value problem

ux uy = u, x, y ∈ R, u(0, y) = y 2 .

F (p, q) = pq − z and parametrization of initial curve is {(0, s, s2 ) : s ∈ R}. Therefore

f (s) = 0, g(s) = s, h(s) = s2 . The initial values for p and q satisfy

p0 0 + q0 1 = 2s, p0 q0 − h = 0 =⇒ p0 2s − s2 = 0

Therefore p0 = 2s . The characteristic strip satisfy

dx
= Fp = q, x(s, 0) = 0
dt
dy
= Fq = p, y(s, 0) = s
dt
dz
= pFp + qFq = pq + pq = 2pq = 2z, z(s, 0) = s2
dt
dp s
= −Fx − Fz p = −p, p(s, 0) =
dt 2
dq
= q, q(s, 0) = 2s
dt

Therefore p(s, t) = 21 set , q(s, t) = 2set and z(s, t) = s2 e2t .

dx
= 2set =⇒ x = 2set − 2s
dt
dy 1 s
= p = set +
dt 2 2
Therefore
x
+ 2y = 2set
2
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 17

squaring this we get

x
( + y)2 = s2 e2t = z
4
x

2
Therefore u(x, y) = 4 +y .

Complete integral and Singular solutions

Let A ⊂ R2 be an open set which is parameter set. Let

" #
2 ua1 uxa1 uya1
(Da u, Dxa u) =
ua2 uxa2 uya2

Definition 5.4.2. A C 2 function u = u(x, a) is said to be a complete integral in Ω × A of

the equation
F (x, y, u, ux , uy ) = 0 in Ω (5.10)
2 u) = 2.
if u(x, a) solves (5.10) and rank of (Da u, Dxa

Example 5.4.3. Find a complete integral of ux uy = u

From the given equation F = pq − z and the characteristic equations are

dx dy dz dp dq
= q, = p, = 2z, =p = q.
dt dt dt dt dt

On integrating we get

p = c1 et , q = c2 et , z = c1 c2 e2t + c3 , x = c2 et + b, and y = c1 et + a

Therefore
u(x, y) = z = (x − b)(y − a) + c3

for u(x, y) to be a solution we need c3 = 0. Then we get

u(x, y) = xy + ab + (a, b) · (x, y)

Easy to check that " #

2 b+x 1 0
(Da u, Dxa u) =
a+y 0 1

whose rank is 2. Therefore u(x, y) is a complete integral.

There can be more than one complete integral as we see below

p
p = c1 et , q = c2 et =⇒ = a and pq = z
q
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 18

This implies
√
r
z
p = ± az q = ±
a
From the strip condition, we get

dz dx dy
=p +q
dt dt dt r
√ dx z dy
= ± az ±
dt a dt

Integrating this
√ √ y
2 z = ±( ax + √ ) + c3
a
Therefore
1 √ y
u(x, y) = b + ( ax + √ )2
2 a
is a complete integral if a > 0. (check!).

In the above example, in both complete integral, No choice of a and b will give trivial solution
u = 0. But u = 0 is also a solution. This motivates us to study

Singular solutions

In this section we study how to build more complicated solutions for nonlinear first order
PDEs. We will construct these new solutions as envelopes of complete integrals.

Definition 5.4.4. Suppose u(x, y, a1 , a2 ) be a C 1 function on U × A where U is open subset

of R2 and A be an open subset of R2 . If the equations

∂u
(x, y, a1 , a2 ) = 0, x ∈ U, a ∈ A, i = 1, 2...n
∂ai

can be solved for the parameter a and has solution a1 = φ(x, y), a2 = ψ(x, y). That is

∂u
(x, y, φ(x, y), ψ(x, y)) = 0, i = 1, 2, ..., n.
∂ai

Then we call the function v(x) = u(x, y, φ(x), ψ(x)), the envelope of the functions {u(., a)}a∈A .

Theorem 5.4.5. Singular solutions:

Suppose for each (a1 , a2 ) ∈ A as above that u = u(., a1 , a2 ) solves the partial differential
equation (5.10). Assume further that the envelope v, defined by

∂u
(x, y, a1 , a2 ) = 0, x ∈ U, a ∈ A, i = 1, 2...n
∂ai
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 19

v(x, y) = u(x, y, φ(x, y), ψ(x, y))

exists and is a C 1 function. Then v also solves (5.10).

Proof. We have v(x, y) = u(x, y, φ(x, y), ψ(x, y)) and so

vx = ux (x, y, φ, ψ) + ua1 (x, y, φ, ψ)φx + ua2 (x, y, φ, ψ)φx = ux (x, y, φ, ψ)

vy = uy (x, y, φ, ψ) + ua1 (x, y, φ, ψ)φy + ua2 (x, y, φ, ψ)φy = ux (x, y, φ, ψ)

Hence for each x ∈ U ,

F (x, y, u, ux , uy ) = F (x, y, u(x, y, φ, ψ), ux (x, y, φ, ψ), uy (x, y, φ, ψ)) = 0

///

Example 5.4.6. Find complete integral and singular solution of

u2x + u2y = 1 + 2u

Here F = p2 + q 2 − 1 − 2z and the characteristic equations are

dx dy dp dq
= 2p, = 2q, = 2p, = 2q
dt dt dt dt
p
solving this we get q = a and p2 + q 2 − 1 − 2z = 0, This implies

(1 + a2 )q 2 = 1 + 2z

Therefore r
1 + 2z 1 + 2z
p = ±a 2
, q=±
1+a 1 + a2
Now from the strip condition

dz dx dy
=p +q
dt dt
r dt
1 + 2z
=± (adx + dy)
1 + a2

Integrating
√ ax + y
1 + 2z = ± √ ±b
1 + a2
Therefore  2
1 ax + y 1
u(x, y, a, b) = √ +b −
2 1 + a2 2
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 20

is the complete integral. Now solving

ax + y
ua = 0, ub = 0 =⇒ √ + b = 0.
1 + a2

Therefore singular solution is u = − 21 .

To generate more solutions from the complete integral, we vary the above construction.
Choose an open set A0 ⊂ R and any C 1 function h : A0 → R, so that the graph of h
lies within A.

Definition 5.4.7. The general integral (depending on h) is the envelope v(x) of the functions

u(x, y, a1 ) = u(x, y, a1 , h(a1 )), (x, y) ∈ U, a ∈ A0

provided this envelope exists and is C 1 .

Example 5.4.8. Find a general integral of ux uy = u.

Earlier we found a complete integral as

u(x, y, a, b) = (x − a)(y − b) = xy + ab − ay − bx

Let h : R → R be h(a) = a. Then

u(x, y, a, a) = xy + a2 − ay − ax

x+y
ua = 2a − y − x = 0 =⇒ a = φ(x, y) =
2
Therefore

(x + y)2 x + y x+y 1
u(x, y, φ(x, y)) = xy + − y− x = − (x − y)2 .
4 2 2 4

5.5 Weak solutions*

In this section we will study the weak solutions of the Initial value problem

ut + (f (u))x = 0, t > 0, x ∈ R, (5.11)

u(x, 0) = h(x), x ∈ R. (5.12)
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 21

where f : R → R is a continuous function. Let v : R × [0, ∞) → R is a smooth function with

compact support. Multiplying (5.11) with v(x, t) and integrating by parts, we get
Z ∞Z ∞ Z ∞Z ∞
0= ut v dxdt + (f (u))x v dxdt
0 −∞ 0 −∞
Z ∞ Z ∞ Z ∞Z ∞ Z ∞
=− uvt dxdt − (f (u))vx dxdt + u(x, 0)v(x, 0) dx
0 −∞ 0 −∞ −∞

Motivated by this, we define

Definition 5.5.1. We call any bounded measurable function u(x, t) as a weak solution if
Z ∞Z ∞ Z ∞
(uvt + (f (u))vx ) dxdt = u(x, 0)v(x, 0) dx
0 −∞ −∞

for all v ∈ Cc∞ (R × [0, ∞)).

It is easy to see that all classical solutions are weak solutions. But not all weak solutions
are classical solutions. If u is a weak solution, then it need not be even continuous. Suppose
u(x, t) is a weak solution of (5.11) such that u(x, t) is discontinuous across a curve x = x(t),
but u(x, t) is smooth on either side of x(t). Let u− (x, t) be the limit of u approaching (x, t)
from the left and let u+ (x, t) be the limit of u approaching x(t) from right. We claim that
such a curve x(t) cannot be arbitrary.

Theorem 5.5.2. (Rankine-Hugoniot condition)

If u(x, t) is a weak solution of (5.11) above such that u(x, t) is discontinuous across the curve
x = x(t) but u is smooth on either side of curve D− and D+ . Then u(x, t) must satisfy the
condition
f (u− ) − f (u+ )
x0 (t) =
u− − u+
where u− is the limit of u from D− and u+ is limit of u from D+ .

Proof. Let D− = {(x, t) : 0 < t < ∞, −∞ < x < x(t)} and D+ − = {(x, t) : 0 < t <
∞, x(t) < x < ∞}. Then if u(x, 0) = 0, we have by divergence theorem on D− we get
ZZ ZZ Z
u− vn2 + f (u− )vn1 ds


[uvt + f (u)vx ] dxdt = − [ut + (f (u))x ] vdxdt +
D− D− x=x(t)
(5.13)
where n = (n1 , n2 ) is the outward unit normal to D− . Similarly,
ZZ ZZ Z
u+ vn2 + f (u+ )vn1 ds


[uvt + f (u)vx ] dxdt = − [ut + (f (u))x ] vdxdt −
D+ D+ x=x(t)
(5.14)
By assumption ut + (f (u))x = 0 in D− and D+ . Therefore from (5.11) and (5.13) we get
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 22

Figure 7: weak solution

Z Z
u− vn2 + f (u− )vn1 ds = u+ vn2 + f (u+ )vn1 ds



x=x(t) x=x(t)

holds for all v ∈ Cc∞ (R × [0, ∞)). Therefore

u− n2 + f (u− )n1 = u+ n2 + f (u+ )n1

This implies
n2 f (u− ) − f (u+ )
− =
n1 u− − u+
On the curve x = x(t),
dx 1 n2
= 0 =− .
dt x (t) n1
Hence the result. ///

Recall the example in 5.3.2. We defined the classical solution for t < 1(take y = t). Now
beyond the line y = 1, we assume that the solution is smooth on either side of a curve x = x(t).
We note that u− 0, u= 1. BY Rankine-Hugoniot condition,

f (u− ) − f (u+ ) 1
x0 (t) = − +
=
u −u 2
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 23

So x = x(t) = 2t , x(1) = 1. Therefore the equation of shock is x = t

2 + 1. Hence

1 x < 1+t
2
u(x, t) =
0 x > 1+t
2 .

We can even define weak solution in case the initial condition has discontinuity like in the
example 5.3.3. In the first case

1 t
x0 (t) = , x(0) = 0 =⇒ x =
2 2

The solution is 
1 t
x< 2
u(x, y) =
0 t
x> 2

In the second case, the wave on the right moves faster and the wave on the left moves
slower. So there is no shock. In this case one can define continuous solution which are called
rarefaction wave solution like



 0 x≤0

u1 (x, y) = xt 0 ≤ x ≤ t


1 x ≥ t


But even in this case one can introduce shock and define shock wave solution as

0 x < t
2
u2 (x, y) =
1 x > t
2

The solution u2 is NOT physically feasible solution. But the solution in u1 is more realistic
which is known as entropy solution.

5.6 Problems
1. Solve the given IVP and determine the values of x and y for which it exists;

(a) xux + uy = y, u(x, 0) = x2 .

(b) ux − 2uy = u, u(0, y) = y.
(c) y −1 ux + uy = u2 , u(x, 1) = x2

2. Solve the given IVP and determine the values of x,y and z for which it exists:

(a) xux + yuy + uz = u, u(x, y, 0) = h(x, y)

(b) ux + uy + zuz = u3 , u(x, y, 1) = h(x, y)
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 24

3. Solve the given IVP and determine the values of x and y for which it exists;

√
(a) ux + u2 uy = 1, u(x, 0) = 1 (b) ux + uuy = 0, u(x, 0) = x2 + 1

4. Find a general solution:

(a) (x + u)ux + (y + u)uy = 0 (b) (x2 + 3y 2 + 3u2 )ux − 2xyuy + 2xu = 0

√
5. Consider the equation ux + uy = u. Derive the general solution u(x, y) = (x + f (x −
y))2 /4. Observe that the trivial solution u(x, y) ≡ 0 is not covered by the general
solution.

6. Solve the IVP a(u)ux + uy = 0 with u(x, 0) = h(x), and show that solution become
singular for some y > 0 unless a(h(s)) is a nondecreasing function of s.
x2
7. Solve u2x + yux − u = 0 with initial condition u(x, 1) = 4 + 1.

8. Solve u = xux + yuy + (u2x + u2y )/2 with initial condition u(x, 0) = (1 − x2 /2)

9. Consider u = u2x + u2y with the initial condition u(x, 0) = ax2 . For what positive
constants a is there a solution? Is it unique? Find all solutions.

10. Show that family of spheres Sa given by (x − a)2 + y 2 + z 2 = 1 has as its envelope ζ the
unit cylinder y 2 + z 2 = 1.

11. Consider c2 (u2x + u2y ) = 1, where c = c(x, y). Derive the characteristic equations. In the
special case c = |x| with initial condition u(x, 0) = 0, find the solution to be
p
x2 + y 2 + y
u(x, y) = − log , for x > 0
x

12. Consider the eikonal equation in three dimensions u2x + u2y + u2z = 1

(a) Solve the initial value problem with u = k = constant on the plane αx + βy + z = 0
(b) Find a complete integral.

13. Find complete integral of

(a) uux uy = ux + uy (b) u2x + u2y = xu

14. Given the complete integral, find the singular solution of

(a) ux uy = z, u = xy + ab + (ab)(xy)
(b) u2x + u2y = 1 u = (xy)(cos a. sin a) + b
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 25

15. Consider uux + uy = 0 with the initial condition

(
0 x<0
u(x, 0) = h(x) =
u0 (x − 1) x > 0,

where u0 > 0. There is a weak solution u(x, y) that has a jump discontinuity along a
curve x = ξ(y). Find this curve and describe the weak solution.

16. Find a simple wave equation u(x, y) = u( xy ) for (G(u))x + uy = 0, when G(u) = u4 /4.
Use this to define a continuous weak solution of the problem for y > 0 that satisfies
(
0 x>0
h(x) =
−1 x < 0,

17. Consider uux + uy = 0 with

(
0 x>0
h(x) =
u0 x < 0,

where u0 < 0. In addition to the rarefaction solution described in the text, show that
there is a weak solution with a shock along x = u0 y/2.
dq dq
18. A reasonable model for low-density traffic is dρ ρx + ρt = 0 with dρ = c, where c is a
constant.

(a) Show that ρ is constant along the (characteristic) curves x = ct + x0 .

(b) If car is alone on the highway, what does ρ(x, 0) look like? What does ρ(x, t) look
like ?
(c) Explain why c represents the free speed of highway.

19. If ρmax denotes the maximum density of cars on a highway(i.e. under bumper-to-bumper
ρ
conditions), then a reasonable relation between q and ρ is given by G(ρ) = cρ(1 − ρmax )
where the constant c is the free speed of the highway (cf. Exercise 16). Suppose the
initial density is (
1
ρ(x, 0) = 2 ρmax x<0
ρmax x > 0,
Find the shock curve and describe the weak solution. Interpret your result for the traffic
flow.

20. Using G(ρ) as in Exercise above, describe the traffic flow after a long red light turns
5 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 26

green at t = 0; that is , the initial density is

(
ρmax x < 0
ρ(x, 0) =
0 x > 0,

In particular, find the density at the green light, ρ(0, t), while the light remains green.

21. Consider uux + ut = 0 with the initial condition


 2
 x<0
u(x, 0) = h(x) = 1 x ∈ (0, 1)

0 x > 1,


Describe the weak solution. Show that there is only one shock as t is large.

REFERENCES:

1. Mc Owen, Partial Differential equations: Methods and applications, Pearson

2. L.C. Evans, Partial Differential equations, AMS.