Classification of Second Order Partial Differential Equations

Lesson: Classification of Second Order Partial Differential

Equations

Course Developer: Sada Nand Prasad

College/Department: Acharya Narendra Dev College

Institute of Lifelong Learning, University of Delhi 1

 Classification of Second Order Partial Differential Equations

Table of Contents:
Chapter : Classification of Second Order Partial Differential
Equations
 1. Learning Outcomes
 2. Introduction
 3. Second – Order Equation in Two Independent Variables
 4. Canonical Forms
 5. Equations with Constant Coefficients
 6. General Solutions
 7. Further Simplification
 8. The Cauchy Problem
 Summary
 Exercises
 Glossary
 References/ Further Reading
1. Learning Outcomes:
After studying this chapter, you will be able to
 classify linear second order PDEs into elliptic, parabolic and
hyperbolic types;
 reduce linear second order PDEs into canonical form;
 classify linear equation with constant coefficient and Euler equation;
 obtain the general solution of linear second order PDEs;
 further Simplification of the reduced linear second order PDEs by
introducing the new dependent variable;
 derive the Cauchy problem;

Institute of Lifelong Learning, University of Delhi 2

 Classification of Second Order Partial Differential Equations

2. Introduction:

In the last chapter we have derived the three fundamental equations of

mathematical physics, namely one and two - dimensional wave equation,
one - dimensional heat conduction equation/diffusion equation and
Laplace’s equation and mentioned that these equations are of hyperbolic,
parabolic and elliptic type. Here, in this chapter we will explain in detail,
about the classification of these second order partial differential equations
with variable and constant coefficients and will find the general solution
after reducing these equations to their respective canonical form.

To begin with, we have in this chapter described the second order partial
differential equations (PDEs) in two independent variables and classified
linear PDEs of second order into elliptic, parabolic and hyperbolic types.

3. Second – Order Partial Differential Equation in Two

Independent Variables :

It is seen that a large number of PDEs arising in the study of applied

mathematics with special reference to biological, physical and engineering
applications, can be treated as a particular case of the most general form
of a linear, second order PDE in two independents of the form
2 z 2 z 2 z z z
a 2 b  c 2  d  e  f z  g, or,
x x y y x y
a zx x  b zx y  c z y y  d zx  e z y  f z  g , (1)
where a, b, c, d, e, f, and g are functions of the independent variables x
and y and do not vanish simultaneously.

Value Addition: Do you know?

From coordinate geometry, we know that general equation of second
degree in two variables
a x2  b x y  c y 2  d x  e y  f  0,
represents, ellipse if b2 – 4 a c < 0,
parabola if b2 – 4 a c = 0 or, (2)
2
hyperbola if b – 4 a c > 0.

The classification of second order partial differential equations (3.1), is

suggested by the classification of the above equation (2), and based upon
the possibilities of transforming equation (1) to canonical form at any
point (x0, y0) by suitable coordinate transformation. Therefore, an
equation at any point (x0, y0) is called

elliptic if b2 (x0, y0)– 4 a(x0, y0) c(x0, y0) < 0,

parabolic if b2 (x0, y0)– 4 a(x0, y0) c(x0, y0) = 0 or, (3)
hyperbolic if b2 (x0, y0)– 4 a(x0, y0) c(x0, y0) > 0.
Institute of Lifelong Learning, University of Delhi 3
 Classification of Second Order Partial Differential Equations

Example 1. Consider the equation, zx x  x 2 z y y  0, , comparing this equation

with equation (1), we get, a = 1, b = 0, c = x2, and therefore, b2- 4ac <
0. Thus, the given equation is elliptic. Similarly, we can say that the
Laplace’s equation derived in the last chapter ux x  u y y  0, is elliptic.

Example 2. Consider the equation, zx x  2 zx y  z y y  0, , comparing this

equation with equation (1), we get, a = 1, b = 2, c = 1, and therefore,
b2- 4ac = 0. Thus, the given equation is parabolic. Similarly, we can say
that the one – dimensional heat equation / diffusion equation derived in
the last chapter Tt  K Tx x is hyperbolic.
Example 3. Consider the equation, z x x  x 2 z y y , comparing this equation
with equation (1), we get, a = 1, b = 0, c = x2, and therefore, b2- 4ac =
4x2 > 0. Thus, the given equation is hyperbolic. Similarly, we can say
that the one - dimensional wave equation derived in the last chapter
ut t  c 2 ux x is hyperbolic.

It may also be remarked here that an equation can be of mixed type,

depending upon its coefficients.

Example 4. The equation given by x zx x  z y y  x 2 , can be classified as

elliptic if, x > 0, parabolic, if x = 0, or hyperbolic if x < 0 as b2- 4ac = - 4
x.

Now, we shall show that by a suitable change in the independent

variables, we can reduce any equation of type (1) to one of the three
standard or canonical forms. Let us suppose we change the variables x
and y to u and v, respectively where
u = u (x, y), v = v(x, y). (4)

We will also assume that u and v are twice continuously differentiable and
in the region under consideration the Jacobian

ux u y
 0.
vx v y
Then for the system (4), we can determine x and y uniquely. Suppose x
and y are twice continuously differentiable functions of u and v. Then, we
have,

Institute of Lifelong Learning, University of Delhi 4

 Classification of Second Order Partial Differential Equations

zx  z  z . u  z . v  zu .ux  zv.vx
x u x v x
z y  z  z . u  z . v  zu .u y  zv.v y
y u y v y
zxx   z    z     . u   . v   z . u  z . v
2     

x2 x  x   u x v x   u x v x 


2  2 2  2
  z  u   2  z u v   z  v   z  u  z  2v
2 2
u 2  x  u.v x x v2  x  u x2 v x2
 zuu ux2  2 zuv uxvx  zvv vx2  zu u x2  zv vx2
z xy   z    z     . u   . v   z . u  z . v 
2     
x y x  y   u x v x   u y v y 

  z u u   z  u v  u v    z v v  z  2u  z  2v
2 2   2
u 2 x y u.v  x y y x  v2 x y u yx v yx
(5)
 zuu uxu y  zuv  uxv y  u yvx   zvv vxv y  zu uxy  zv vxy
 

z yy   z    z     . u   . v   z . u  z . v 
2     

y2 y  y   u y v y   u y v y 
2  2 2  2
  z  u   2  z u v   z  v   z  u  z  v
2 2 2
u 2  y  u.v y y v2  y  u y2 v y2
 zuu u y2  2 zuv u yv y  zvv v y 2  zu u y 2  zv v y 2

Substituting all these values in equation (1), we get,

a* zuu  b* zuv  c* zvv  d * zu  e* zv  f * z  g *

where

a*  a u x 2  b u x u y  c u y 2 ,
b*  2 a u x vx  b  u x v y  u y vx   2c u y v y ,
c*  a v x 2  b v x v y  c v y 2 ,
d *  a u xx  b u xy  c u yy  du x  eu y , (6)
e*  a vxx  b vxy  c v yy  dvx  ev y ,
f* f,
g *  g.
The classification of the second order PDE (1) depends on the coefficients
a(x, y), b(x, y), and c(x, y) at any point (x, y). So, we can write equation
(1) and equation (2) as

a zxx  b z xy  c z yy  h  x, y, z, z x , z y  (7)

and

Institute of Lifelong Learning, University of Delhi 5

 Classification of Second Order Partial Differential Equations

a* zuu  b* zuv  c* zvv  h* u, v, z, zu , zv  (8)

respectively.

4. Canonical Forms:
Now the problem is to determine u and v so that the equation (7) takes
the simplest form. We first suppose that a ≠ 0, b ≠ 0, c ≠ 0. We will
choose the new variables u and v such that the coefficients a* and c* in
equation (6) vanish.

So, we get,

a*  a u x 2  bu x u y  c u y 2  0,
c*  a vx 2  b vx v y  c v y 2  0,

Or, equivalently,

a wx 2  b wx wy  c wy 2  0, (9)

Dividing the above equation (9) throughout by wy 2 , we get,

2
w  w 
a x   b  x   c  0, (10)
 wy
   wy 

wx
If,   , then eqn. (10) can be rewritten as
wy

a  2  b   c  0, (11)

Equation (11) is a quadratic equation and its roots, are known as

characteristic equation, are ordinary differential equation in the xy – plane
along which u = constant and v = constant and the integrals of the roots
are known as characteristic curves.

Value Additions: Do you know?

The procedure to determine u and v to get equation (7) in simplest form

is simple when the discriminant b2 – 4ac of the quadratic eqn. (11) is
positive, negative or zero.

Institute of Lifelong Learning, University of Delhi 6

 Classification of Second Order Partial Differential Equations

Now we will discuss all these three cases one by one.

Case I.

If b2  4 a c  0, then the roots 1 , 2 of eqn. (11) are real and distinct. The
coefficients a* and c* in eqn. (6) will vanish if we choose the new variables
u and v such that
ux  1 u y , (12)
vy  2 vy (13)
Equations (12) and (13) will determine the form of u and v as a function
of x and y and to determine this use Lagrange’s auxiliary equation, that
we had already discussed in the previous chapter.
The Lagrange’s auxiliary equation of equation (12) is given by
dx dy du
 
1  1 0
du = 0 => u =constant and from the first two members, we have,
dy
 1  0, (14)
dx
Suppose 1  x, y  = constant is the solution of the above equation (14),
then the solution of equation (12) can be taken as,
u  1  x, y  (15)
Similarly, if 2  x, y  = constant is the solution of the equation
dy
 2  0, (16)
dx
then the solution of equation (13) can be taken as,
v  2  x, y  (17)

In general, we can also observe that

 
4 a c  b2  4 a c  b 2  ux v y  u y vx 
2

so that when a* and c* are zero, then

  u v  u y vx 
2
b2  b2  4 a c x y (18)

This follows that b2  0,  b2  4 a c  0  and therefore we can divide both sides
of the equation by it.
Making the substitution defined by equation (15) and equation (17),
equation (7) transform to the form
zuv    u, v, z, zu , zv  (19)

Equation (19) is the canonical or standard form in this case.

Institute of Lifelong Learning, University of Delhi 7

 Classification of Second Order Partial Differential Equations

Example 5: Reduce the equation in example 3, given by z x x  x 2 z y y to

canonical form.
From example 3, we know that the equation is hyperbolic and the
quadratic equation (11) becomes  2  x2  0     x. Also, the equations
(14) and (16) becomes
dy dy
 x  0 and  x  0 .
dx dx
Integrating both these equations, we get,

x2 x2 x2 x2
y  c1 and y   c2 => u  y  and v  y 
2 2 2 2
From equation (5), we get,
z x  x  zu  zv  ,
z y  zu  zv ,

 
z xx  x 2 zu u  2 zu v  zvv  zu  zv ,
z yy  zu u  2 zu v  zvv ,
Substituting these values in the equation z x x  x 2 z y y , we get,
x2  zuu  2 zu v  zvv   zu  zv  x 2  zuu  2 zu v  zvv   0,
1 1
2  u
zu v  z  zv   z u v   zu  zv  ,
4x 4 u  v 
This is the required canonical form of the hyperbolic equation z x x  x 2 z y y .

Case II.

When b2  4 a c  0, then the roots 1 , 2 of equation (11) are equal. We will

define the function u as we have defined in the earlier case and take v to
be the function of x and y, which is independent of u.

We have a* = 0 as in the last case. Also, since b2  4 a c  0, therefore, b* =

0. In the case, we also take c* ≠ 0, otherwise v would be a function of u.
Dividing equation (8) throughout by c*, we get,
zvv    u, v, z, zu , zv  (20)
Equation (20) is called the canonical form of the parabolic equation.

Example 6: Consider the equation given in example 2,

zx x  2 zx y  z y y  0, We want to reduce this parabolic equation to the canonical
form. For this equation the
quadratic equation (11) becomes
dy dy
 2  2  1 0    1,  1 . Also, the equation    0 becomes 1  0 . And
dx dx
integrating the equation, we get, x – y = constant and consequently,

Institute of Lifelong Learning, University of Delhi 8

 Classification of Second Order Partial Differential Equations

u = x – y. If we take v = x + y, which is a function of variables x and y

independent of u, then from equation (5), we get,
z x  zu  zv ,
z y   zu  z v ,
z xx  zu u  2 zu v  zvv ,
z xy   zu u  zvv ,
z yy  zuu  2 zuv  zvv ,
Substituting the values into the equation zx x  2 zx y  z y y  0, we get,

zu u  2 zu v  zvv  2 zu u  2 zvv  zuu  2 zuv  zvv  0,

 4 zvv  0, or zvv  0

which is the required canonical form of the equation zx x  2 zx y  z y y  0 .

Case III.

For elliptic type equation, we have b2  4 a c  0, in this case the roots 1 , 2

of equation (11) are complex. Proceeding, as in the case I, we find the
equation (7) reduces to the form (19) but the variables u and v complex
conjugate.

To get a real canonical form, we take u    i  , v    i  , so that

1 1
 u  v  ,   u  v  ,
2 2

Now, zu  z u  z u 
1
2
 z  i z 
Similarly, we have,

zv  z  v  z  v 
1
2
 z  i z 
 zu v   zv u 
1
4
 z  z 
Again, transforming the variables u and v to α and β, respectively, we get

z  z    ,  , z, z , z  (21)

Equation (21) is called the canonical form of the elliptic equation.

Institute of Lifelong Learning, University of Delhi 9

 Classification of Second Order Partial Differential Equations

Example 7: In this example we will reduce the elliptic equation in

example 1, given by zx x  x 2 z y y  0, to canonical form.
In this case the quadratic equation (11) becomes  2  x2  0     i x . Also,
the equations (14) and (16) becomes
dy dy
 i x  0 and  i x  0
dx dx
Integrating both these equations, we get,

x2 x2 x2 x2
y  i  c1 and y  i  c2 => u  y  i    i  and v  y  i    i  ,
2 2 2 2
2
x
   y, 
2
From equation (5), we get,
z x  z . x  z .  x  x z ,
z y  z . y  z .  y  z ,

z xx   z x  x   x z   z  x  z  . x   z  .  x   z  x 2 z  ,
x  x  
z yy   z y   z  ,
y

Substituting the values into the equation zx x  x 2 z y y  0, we get,

1
z  x 2 z   x 2 z   0,  z   z    z ,
x2
1
or , z   z    z
2
which is the required canonical form of the equation zx x  x 2 z y y  0 .
Now in the next section, we will discuss the canonical form of the second
order PDE with constant coefficients, in the similar manner we have
discussed for the second – order PDE with variable coefficients in this
section.

5. Equations with Constant Coefficients:

When the coefficients of the equation (1), are constant and real, then the
discriminant b2  4 a c is constant and therefore the equation will be of a
single type at all points.
The general form of the second order PDE with constant coefficients is
given by
a zxx  b zxy  c z yy  d zx  e z y  f z  g  x, y  , (22)
In particular, the equation given by
a zxx  b z xy  c z yy  0, (23)
is called the Euler equation.

Institute of Lifelong Learning, University of Delhi 10

 Classification of Second Order Partial Differential Equations

Now we will discuss all the three cases one by one as we have discussed
in the case of equations with variable coefficients.

Case I. (Hyperbolic Type)

When b2  4 a c  0, we get two distinct roots and integrating equation (14)

and equation (16), we get,

y   1 x  c1 , y   2 x  c2 , (24)

This implies,

u  y  1 x , v  y  2 x, (25)

where

 b  b2  4 a c
1,2  (26)
2a

Equation (24), are two families of straight lines.

By using equation (25), equation (22) becomes

zu v  d1 zu  e1 zv  f1 z  g1 u, v  , (27)

where d1, e1, f1 are constant.

If in equation (26), a = 0, from equation (11), equation (14) and equaiton

(16), we get,
2 2
b c  dx   dx   dx   dx 
a   2  0  b     c     0   b    c    0,
   dy   dy   dy   dy 

dx  dx 
=>  0, and  b  c    0,
dy  dy 

Integrating the above eqn. we get,

b
x  c1 and x    y  c2 ,
c

where c1 and c2 are constants.

This implies,

Institute of Lifelong Learning, University of Delhi 11

 Classification of Second Order Partial Differential Equations

b
u  x and v  x    y , (28)
c

Using equation (28), equation (22) reduces to the canonical form

zu v  d1 zu  e1 zv  f1 z  g1 u, v  , (29)

where d1 , e1 and f1 are constants.

For Euler equation given by equation (23), the canonical form is given by
zu v  0, (30)

We can get the general solution of the Euler equation by integrating

equation (30), and the solution is given by
z  1  u   2  v   1  y  1 x   2  y  2 x  , (31)
where 1 and 2 are arbitrary functions, and 1 and 2 are given by
equation (25).

Example 8: Consider the equation zx x  zx y  2 z y y  3 zx  6 z y  9  2 x  y  , to

reduce this equation to canonical form; first we will compare this equation
with the equation (22) and get, a = 1, b = 1, c = -2, d = - 3, e = - 6 and
g = 9 (2 x - y). This implies b2  4 a c 1 8  9  0, and consequently the
equation is hyperbolic.
Also, from equation (25) and equation (26), we get,

1  9
u  y  x, v  y  2 x, 1,2   1,  2
2

This gives,

z x   2 zu  zv ,
z y  zu  z v ,
z x x  4 zu u  4 zu v  zv v ,
z x y   2 zu u  zu v  zv v ,
z y y  zu u  2 zu v  zv v ,

Substituting these values into the equation,

zx x  zx y  2 z y y  3 zx  6 z y  9  2 x  y  ,

we get,

Institute of Lifelong Learning, University of Delhi 12

 Classification of Second Order Partial Differential Equations

4 zu u  4 zu v  zvv  2 zu u  zu v  zvv  2 zu u  4 zu v  2 zvv  6 zu  3 zv  6 zu  6 zv   9 u,

 zu v  zv  u,

which is the required canonical form.

Case II. (Parabolic Type)

When b2  4 a c  0, we get equal roots

b
1  2 
2a
and only one family of characteristic exists.

dy dy
Also,  1  0, =>   1 , and integration gives,
dx dx
y  1 x  c1 .
Let
 b 
u y  x
 2a 
and v  h x  k y (32)
where h and k are choose so that the transformation is not equal to zero.
Using equation (32), equation (22) reduces to the canonical form

zvv  d2 zu  e2 zv  f 2 z  g 2 u, v  , (33)

If b = 0, we have 4 ac = 0 => a = 0 or c = 0. Then we observe that the
equation is already in the canonical form. In the similar way if a = 0 or c
= 0, then b = 0, and the equation is already in the canonical form in the
parabolic equation, and the canonical form is given by,
zvv  0, (34)
Integrating the above equation, we get the general solution as
zv  1  u   u  v 1  u   2  u  (35)

Example 9: Let us reduce the equation zx x  2 zx y  z y y  0, to canonical

form, comparing this equation with the equation (22) we get, a = 1,
b = -2, c = 1, d =0, e =0 and g =0, this gives b2  4 a c  4  4  0,
and therefore the equation is parabolic.
From, equation (32), we have,
u  y  x, and v  y,
chosen so that the transformation is not equal to zero.
This gives,
z x  zu ,
z y  zu  z v ,
z x x  zu u ,

Institute of Lifelong Learning, University of Delhi 13

 Classification of Second Order Partial Differential Equations

z x y  zu u  zu v ,
z y y  zu u  2 zu v  zvv ,
Substituting these values, equation (22) reduces to the canonical form

zu u  2 zu u  2 zu v  zu u  2 zu v  zvv  0,
 zvv  0.

Case III. (Elliptic Type)

When b2  4 a c  0, the characteristics are complex conjugate and are given

by

y   1 x  c1 , y   2 x  c2 , (36)

where 1  A  i B, 2  A  i B, c1 and c2 are complex numbers and

b 4 a c  b2
A ,B
2a 2a

This implies,

u  y  1 x , v  y  2 x, (37)

 u  v    y   A  i B  x  y   A  i B  x   y  Ax,
1 1

2 2
(38)
1
  u  v   B x
2i

Using equation (38), equation (22) reduces to the canonical form

z  z  d2 z  e2 z  f3 z  g3  ,   , (39)

Example 10: To reduce the equation zx x  z y y  0, to canonical form, we

compare this equation with the equation (22) and we get, a = 1, b = 0,
c = 1, d =0, e =0 and g =0, this gives b2  4 a c   4  0,
and therefore the equation is elliptic.

From, equation (37), we have, u  x  i y , v  x  i y, and therefore from

equation (38), we get,
1 1
   u  v   x, and    u  v   y,
2 2i
This gives,

Institute of Lifelong Learning, University of Delhi 14

 Classification of Second Order Partial Differential Equations

z x  z ,
z y  z ,
z x x  z  ,
z y y  z  ,
Substituting these values, equation (22) reduces to the canonical form.
z   z   0.

6. General Solutions:
The general solution of the second order PDE is the most general possible
relation between the independent variables x, y and the dependent
variable z such that the value of z and the associated derivatives obtained
from it and substituted in the given equation reduce it to an identity.
Generally, it is difficult to determine the general solution of any given
equation but reduction to canonical form simplifies the equation and then
it becomes easier to get the general solution. If the canonical form is
simple, then we can determine the general solution immediately.

Example 11: In example 8 we have reduced the equation

zx x  zx y  2 z y y  3 zx  6 z y  9  2 x  y  , to canonical form; let us determine the
general solution of this equation.
Substituting t  zv , in the canonical form zu v  zv  u, we get,
tu  t  u,
I.F. = e 
du
 eu
 t . eu   u eu  f  v  ,
 t   u 1  e  u f  v  ,
 zv   u 1  e  u f  v  ,
 z  u, v    u  v  v  e u f  v   g  v  ,
 z  u , v    x  y  y  2 x 1  e 2 x  y  f  x  y   g  y  2 x  .

Example 12: In example 9, we reduced the parabolic equation

zx x  2 zx y  z y y  0, to canonical form zvv  0, in this example we will determine
the general solution of this equation.
Integrating the equation zvv  0, we get,
zv  f  u   z  v f  u   g  u  ,
 z  x, y   y f  y  x   g  y  x  .

Institute of Lifelong Learning, University of Delhi 15

 Classification of Second Order Partial Differential Equations

Example 13: In example 10, we have reduced the elliptic equation

zx x  z y y  0, to canonical form; let us determine the general solution of this
equation.
The canonical form is given by, z   z   0. We will first transform this
equation in the independent variable u and v and to get this we have,
u  x  i y , v  x  i y, This gives,
z x  zu  zv ,
z y  i zu  i zv ,
z x x  zu u  2 zu v  zv v ,
z y y   zu u  2 zu v  zv v ,

Substituting all these values in eqn. zx x  z y y  0, we get,

zu v  0  zu  f  u   z  f  u   g  v  ,
 z  x, y   f  x  i y   g  x  i y  .

7. Further Simplification:
We have already simplified the second order PDE with constant
coefficients by reducing them to canonical forms. Here we will further
simplify the reduced equations by introducing the new dependent variable

v  z e  a r  b s  , (40)

Where a and b are undetermined coefficients.

We have classified the second order PDE with constant coefficients as

hyperbolic:

zr s  a1 zr  a2 zs  a3 z  f1 (41)

zr r  zs s  a1 zr  a2 zs  a3 z  f1 (42)

parabolic:

zs s  b1 zr  b2 zs  b3 z  f 2 (45)

elliptic:
zr r  zs s  c1 zr  c2 zs  c3 z  f3 (46)

where r and s are new variables in the transformations

Institute of Lifelong Learning, University of Delhi 16
 Classification of Second Order Partial Differential Equations

r = r (x, y), s = s (x, y),

and the jacobian J ≠ 0.

From equation (40), we get,

zr   vr  a v  e a r  b s ,
zs   vs  b v  e a r  b s ,
zr r   vr r  2 a vr  a 2 v  e a r  b s ,
zr s   vr s  a vs  b vr  a b v  e a r  b s ,
zs s   vss  2 b vs  b 2 v  e a r  b s ,
Substituting these values into equation (41), we get,
vr s   b  a1  vr   a  a2  vs   a b  a1 b  a2 b  a3  v  f1 e  a r  b s  ,
If b = a1 and a = a2 we get,
vr s   a1a2  a3  v  g1 ,
where
g1  f1 e  a r  b s  .
Similarly by substituting these values into equation (42) – (44), we get
the corresponding canonical forms as
hyperbolic : vr s  h1 v  g1 ,
vr r  vs s  h1 v  g1 ,
(45)
parabolic : vs s  h2 v  g 2 ,
elliptic : vr r  vs s  h3 v  g3 ,

Example 14: Transform the equation zx x  z y y  3 zx  2 z y  z  0, to the form

vr s  c v, c  constant by introducing the new variable v  z e a r  b s  , where a and
b are undetermined coefficients.

Comparing this equation to equation (22), we get,

a = 1, b = 0, c = -1, d = 3, e = -2, f = 1 and g = 0, and

b2  4 a c  4  0, consequently the equation is hyperbolic.
Also, from equation (25) and equation (26), we get,

0 4
u  y  x, v  y  x, 1,2   1, 1
2

This gives,

Institute of Lifelong Learning, University of Delhi 17

 Classification of Second Order Partial Differential Equations

z x   zu  zv ,
z y  zu  zv ,
z x x  zu u  2 zu v  zv v ,
z y y  zu u  2 zu v  zv v ,

Substituting these values into the equation, zx y  z y y  3 zx  2 z y  z  0, we get,

zu u  2 zu v  zvv  zu u  2 zu v  zvv  3 zu  4 zv  2 zu  2 zv  z  0,
  4 zu v  5 zu  zv  z  0,

Substituting v  z e a r  b s  , we have,

zu  vr e a r  b s  v a e a r  b s  e a r  b s  vr  a v  ,

zu v  b e a r  b s  vr  a v   e a r  b s vr s  a v , 
zv  vs e a r  b s  b v e a r  b s ,

Substituting these values in the eqn.  4 zu v  5 zu  zv  z  0 , we get,

 
 4 ea r  b s b vr  b a v  vr s  a vs   5 e a r  b s  vr  a v   e a r  b s  vs  b v   v e a r  b s  0,
  4 vr s    4 a  1 vs    4 b  5  vr  4 a b v  5 a v  b v  v  0,

Putting – 4 a + 1 = 0 => a = ¼, and – 4 b – 5 = 0 => b = - 5/4 in the

above equation, we get,

v  1 
 4 vr s   0  vr s    v
4  16 

which is the required form.

8. The Cauchy Problem:

Consider a second – order PDE for the function z, in the independent
variables x and y, and suppose this equation can be solved explicitly for
zyy and hence it can be represented y the equation in the form

z y y  F  x, y, z, z x , z y , z x x , z x y  ,

If the initial conditions are described along the same curve in the xy –
plane, then this problem is called Cauchy problem.

Example 15: Consider the Euler equation

Institute of Lifelong Learning, University of Delhi 18

 Classification of Second Order Partial Differential Equations

a z x x  b z x y  c z y y  F  x, y , z , z x , z y  , (46)

where the coefficient a, b, c are function of x and y.

Let (x0, y0) be the point on the smooth curve in the xy – plane.

Let the parametric equation of this curve is

x0 = x0(λ), y0 = y0(λ) (47)

where λ is a parameter.

Suppose two functions f (λ) and g (λ) are prescribed along the same
curve and the Cauchy conditions are

z = f (λ) (48 a)

zn = g (λ) (48 b)

where n is the direction of the normal to the curve.

For every point on the curve, every value of z is specified by equation (48
a). The curve represented by equation (47) with condition (48 a) yields a
twisted curve in (x, y, z) space whose projection on the (x, y) is the curve
we have considered.

Hence the solution of the Cauchy problem is the surface called an integral
surface in the (x, y, z) space passing through the curve satisfying the
condition (48 a) which represent a tangent plane to the integral surface
along the curve.

If the function f (λ) is differentiable, then along the curve, we have,

dz  z d x  z d y d f
   , (49)
d  x d   y d  d 

z z d x z d y
   g, (50)
n  x d n  y d n

dx dy dy dx
 &  (51)
dn ds dn ds

Putting equation (51) into equation (50), we get,

z z d y z d x
   g, (52)
n x ds y ds

Institute of Lifelong Learning, University of Delhi 19

 Classification of Second Order Partial Differential Equations

dx dy
d d d x  d y
2 2

Since,   0.
dy ds d sd 

ds dx

Therefore we find zx and zy on the curve from the system of equations

(49) and (52). Since, zx and zy are known on the curve, we can find
higher derivatives by first differentiating zx and zy with respect to λ.

d   z  2 z d x 2 z d y
   (53)
d    x   x2 d   x  y d 

d z  2 z d x 2 z d y
    (54)
d    y   x  y d   y2 d 

The given equation was

2 z 2 z 2 z
a  b  c  F, (55)
 x2  x y  y2

The equations (53) – (55) can be solved for zxx, zxy and zyy if

dx dy
0
d d
dx dy
0  0.
d d
a b c

dx  dx dy  d y d y
c b  a   0,
d  d d  d d
2 2
dx dx d y d y
c  b  a   0,
d d d d
d y
2
 d x d yd  
a   b     c  0,
dx d   d x 2 
 

We know that
2
d y d y
a  b   c  0,
dx dx

Institute of Lifelong Learning, University of Delhi 20

 Classification of Second Order Partial Differential Equations

is called the characteristic equation. Therefore the necessary condition for

obtaining the second derivative is that the curve must not be
characteristic curve.

Summary:
In this chapter we have covered the following
1) Classification of linear second order PDEs into hyperbolic, parabolic
and elliptic types on the condition b2 - 4ac greater than, equal to or
less than zero.
2) We have discussed the canonical form of the second – order PDE
variable and constant coefficients.
3) We have determined the general solution after reducing the PDE
into its respective canonical forms.
4) We have further simplified the second order PDE with constant
coefficients by reducing them to canonical forms by introducing the
new dependent variable v  z e a r  b s  .
5) Finally, we introduced the cauchy’s problem.

Exercise:
1. Reduce the following equations in the canonical form after
classifying them.

i. y 2 zx x  x 2 z y y  0,

ii. x2 zx x  2 x y ux y  y 2 z y y  0,

iii. zx x  x 2 z y y  0,

iv. 4 zx x  5 zx y  z y y  zx  z y  2,

v. zx x  4 zx y  4 z y y  e y ,

vi. zx x  zx y  z y y  zx  0,

vii. zx x  4 zx y  4 z y y  0,

3
viii. 3 z x x  4 z x y  z y y  0,
4

ix. x zx x  z y y  x 2 ,

Institute of Lifelong Learning, University of Delhi 21

 Classification of Second Order Partial Differential Equations

x. zx x  zx y  x z y y  0,

2. Determine the general solution of the following equation.

i. 4 zx x  5 zx y  z y y  zx  z y  2,

3
ii. 3 z x x  4 zx y  z y y  0,
4

iii. y 2 zx x  x 2 z y y  0,

iv. x2 zx x  2 x y ux y  y 2 z y y  0,

v zx x  x 2 z y y  0,

3. Determine the general solution of the following equation.

i. zx x x x  2 zx x y y  zy y y y  0,

ii. z x x  x 2 z y y

4. Transform the equation 3 zx x  7 zx y  2 z y y  z y  z  0, to the form

vr s  c v, c  constant by introducing the new variable v  z e a r  b s  , where
a and b are undetermined coefficients.

Glossary:
C Cauchy Problem: Consider a second – order PDE for the function z,
in the independent variables x and y, and suppose this equation can
be solved explicitly for zyy and hence it can be represented y the
equation in the form z y y  F  x, y, z, z x , z y , z x x , z x y  , If the initial conditions
are described along the same curve in the xy – plane, then this
problem is called Cauchy problem.

D Discriminant : for the second order partial differential equation

a zx x  b zx y  c z y y  d zx  e z y  f z  g , (1)

the value b2  4ac is called the discriminant of the second order

partial differential equation.

E Elliptic: Partial differential equation of second order represented by

equation (1) is called elliptic iff b2  4ac  0 .

Institute of Lifelong Learning, University of Delhi 22

 Classification of Second Order Partial Differential Equations

H Hyperbolic: Partial differential equation of second order represented

by equation (1) is called hyperbolic iff b2  4ac  0 .

P Parabolic: Partial differential equation of second order represented

by equation (1) is called parabolic iff b2  4ac  0 .

References:
1. Tyn Myint-U, Lokenath Debnath: Linear Partial Differential
Equations for Scientists and Engineers, Springer
2. Sneddon, I.N., Elements of Partial Differential Equations,
McGraw-Hill, New York (1957)
3. Robert C. McOwen, Partial Differential Equations Methods
and Applications, Pearson
4. Epstein B., Partial Differential Equations, McGraw- Hill, New
York (1962).
5. Hadamard J., Lectures on Cauchy’s Problem in Linear Partial
Differential Equations, Dover Publications, New York (1952).

Institute of Lifelong Learning, University of Delhi 23