DISCLAIMER: These lecture notes consist of a compilation of material

taken from numerous sources, and so should not be taken as original text.
Note that there could be some typos in the text.

1
MAT 5122-PARTIAL DIFFERENTIAL EQUATIONS (PDEs)
TOPIC 3: Second Order Partial
differential Equations

0.1 Introduction

The most general form of second order linear PDE is

Auxx + Buxy + Cuyy + DUx + Euy + F u = G, (1)

where x, y are the independent variables and u is the dependent variable, depending on x
and y. The coefficients A, B, C, D, E, F, G are functions of x and y. When the coefficients
A, B, C, D, E, F, G are all constants, the equation is called second order linear PDE with
constant coefficients. Second order linear PDE with constant coefficients are classified into
three types on the basis that B 2 − 4AC is > 0, = 0 or < 0. If

1. B 2 − 4ACis > 0, then it is called Hyperbolic.

2. B 2 − 4ACis = 0, then it is called Parabolic.
3. B 2 − 4ACis < 0, then it is called Elliptic.
Example 1. Classify each of the following PDEs:

(a) uxx + uyy = 0. Here, A = 1, B = 0, C = 1 ⇒ B 2 − 4AC = −4 < 0. So the equation is

Elliptic.
(b) uxx − ut = 0. Here,A = 1, B = 0 = C ⇒ B 2 − 4AC = 0. So the equation is parabolic.
(c) utt − uxx = 0. A = −1, B = 0, C = 1, ⇒ B 2 − 4AC = 4 > 0. The equation is
Hyperbolic.

1
(d) uxx +4uxy+5uyy +ux +2uy = 0. A = 1, B = 4, C = 5, ⇒ B 2 −4AC = 16−20 = −4 < 0.
The equation is Elliptic.
(e) uxx − 4uxy + 4uyy + 3ux + 4u = 0. A = 1, B = −4, C = 4, ⇒ B 2 − 4AC = 16 − 16 = 0.
The equation is parabolic
(f) uxx + 2uxy − 3uyy + 2ux + 6uy = 0. A = 1, B = 2, C = −3, ⇒ B 2 − 4AC = 16 > 0.
Hence the equation is Hyperbolic.

When at least one of the coefficients involve terms of x and/or y, then the equation is called
second order linear PDE with variable coefficients. These are also classified on the
basis that B 2 (x, y) − 4A(x, y)C(x, y) is 0 > 0, = 0 or < 0. If

1. B 2 (x, y) − 4A(x, y)C(x, y) > 0, the equation is Hyperbolic.

2. B 2 (x, y) − 4A(x, y)C(x, y) = 0, the equation is Parabolic.
3. B 2 (x, y) − 4A(x, y)C(x, y) < 0, the equation is Elliptic.
Example 2. Classify each of the PDEs:

(a) uxx +xuyy = 0. A = 1, B = 0, C = x, ⇒ B 2 −4AC = −4x. When x > 0, B 2 −4AC =

−4x < 0. So it is Elliptic in the region x > 0. When x < 0, B 2 − 4AC = −4x > 0,
so it is Hyperbolic in the region x < 0. When x = 0, B 2 − 4AC = −4x = 0, so it is
parabolic.
(b) x2 uxx + 2xyuxy + y 2 uyy = 0. B 2 − 4AC = (2xy)2 − 4x2 y 2 = 0 ∀(x, y) ∈ R2 . So it is
parabolic every where.
(c) uxx − 2(cos x)uxy − (sin2 x)uyy = 0. B 2 − 4AC = 4 cos2 x + 4 sin2 x = 4 > 0. Hence
Hyperbolic.

Note that a given PDE may be of one type at a specific point, and of another type at some
other point/s as shown in the case of the Tricomi equation in part (a) of the example above.

The three fundamental second order linear PDEs are the wave equation, the heat equation
and the Laplace’s equation. These are of hyperbolic, parabolic, and elliptic type, respectively.
These types arise in many applications such as the wave propagation, the time-dependent
diffusion processes, and the steady state or equilibrium processes. Thus, hyperbolic equations
model the transport of some physical quantity, such as fluids or waves. Parabolic problems
describe evolutionary phenomena that lead to a steady state described by an elliptic equation.
Elliptic equations are associated to a special state of a system, in principle corresponding to
the minimum of the energy.

2
0.2 Canonical forms

To reduce the equation

A(x, y)uxx + B(x, y)uxy + C(x, y)uyy = F (x, y, u, ux , uy ) (2)
to a canonical form, we transform the independent variables x and y to the new independent
variables ξ and η through the change of variables
ξ = ξ(x, y), η = η(x, y), (3)
where both ξ and η are twice continuously differentiable and that the Jacobian
 
∂(ξ, η) ξx ξy 
J= = 6= 0 (4)
∂(x, y) ηx ηy 
in the region under consideration. The non-vanishing of the Jacobian of the transformation
ensures that a one-to-one transformation exists between the new and old variables. If we
define w(ξ, η) = u(x(ξ, η), y(ξ, η)), then u(x, y) = w(ξ(x, y), η(x, y)) and, apply the chain rule
to compute the terms of equation (2) in terms of ξ and η as follows:
ux = wξ ξx + wη ηx (5)
uy = wξ ξy + wη ηy
uxx = wξξ ξx2 + 2wξη ξx ηx + wηη ηx2 + wξ ξxx + wη ηxx
uyy = wξξ ξy2 + 2wξη ξy ηy + wηη ηy2 + wξ ξyy + wη ηyy
uxy = wξξ ξx ξy + wξη (ξx ηy + ξy ηx ) + wηη ηx ηy + wξ ξxy + wη ηxy .
Substituting the these expressions into equation (2) we obtain the transformed PDE as
awξξ + bwξη + cwηη = f (ξ, η, w, wξ , wη ) (6)
where F becomes f and the new coefficients of the higher order terms a, b, and c are expressed
via the original coefficients and the change of variables formulas as follows:
a = Aξx2 + Bξx ξy + Cξy2 (7)
b = 2Aξx ηx + B(ξx ηy + ξy ηx ) + 2Cξy ηy
c = Aηx2 + Bηx ηy + Cηy2 .
Now, the form (6) is no simpler than that of the original PDE (2), but this is expected because
the choice of the new variables ξ and η has been arbitrary. We can construct transformations
which will make one, or possibly two of the coefficients of the leading second order terms of
equation (6) vanish, thus reducing the equation to a simpler form called canonical form.

3
By definition, a PDE is hyperbolic if the discriminant 4 = B 2 − 4AC > 0. It is easy to show
that the sign of the 4 is invariant under the change of coordinates. It follows that for a
hyperbolic PDE, we should have b2 − 4ac > 0. The simplest case of satisfying this condition
is a = c = 0. So we chose ξ and η such that a and c vanish and get
wξη = Ψ(ξ, η, w, wξ , wη ) (8)
where Ψ = fb . This form is called the first canonical form of the hyperbolic equation. Another
simple case is when b = 0 and c = −a. In this case we have:
wξξ − wηη = ψ(ξ, η, w, wξ , wη )
where ψ = fa . This is called the second canonical form of the hyperbolic equation.

A PDE is parabolic if 4 = 0, and so for a parabolic PDE we should have b− 4ac = 0. the
simplest case of satisfying this condition is a (or c = 0). In this case another necessary
requirement b = 0 will follow automatically. So if we try to chose new variables ξ and η
such that the coefficients a and b vanish, we get the following canonical form of the parabolic
equations:
wηη = ψ(ξ, η, w, wξ , wη ) (9)
where ψ = fc .

A PDE is elliptic if 4 < 0 and so for an elliptic PDE, we should have b2 − 4ac < 0. The
simplest case of satisfying this condition is b = 0 and c = a. So, if we try to chose the new
variables ξ and η such that b vanishes and c = a, we get the following canonical form of
elliptic equation:
wξξ + wηη = ψ(ξ, η, w, wψ , wη ) (10)
where ψ = fa .

0.2.1 Hyperbolic equations

From (7) and using the condition that a and c vanish in (6), we have
a = Aξx2 + Bξx ξy + Cξy2 = 0 (11)
c = Aηx2 + Bηx ηy + Cηy2 = 0.
Dividing the first equation in (11) by ξy2 and the second equation by ηy2 , we obtain
 2  
ξx ξx
A +B + C = 0, (12)
ξy ξy

4
  2  
ηx ηx
A +B + C = 0. (13)
ηy ηy
Equation (12) is a quadratic equation in ηηxy , whose roots are given by
√ √
−B − B 2 − 4AC −B + B 2 − 4AC
µ1 (x, y) = , µ2 (x, y) = .
2A 2A
The roots of (13) can also be found in an identical manner, so only two distinct roots are
possible between the two equations. Thus we may consider µ1 as the root of (12) and µ2 as
that of (13). That is,
√ √
ξx −B − B 2 − 4AC ηx −B + B 2 − 4AC
µ1 (x, y) = = , µ2 (x, y) = = (14)
ξy 2A ηy 2A
Equations (14) lead to the following two first-order differential equations
ξx − µ1 (x, y)ξy = 0, ηx − µ2 (x, y)ηy = 0. (15)
These are the equations that define the new coordinate variable ξ and η that are necessary to
make a = c = 0 in (6). Since the derivative of ξ along the coordinate line ξ(x, y) =constant,
dξ = 0, we have: dξ = ξx dx + ξy dy = 0. Hence the slope of such curves is given by
dy ξx
=− .
dx ξy
We have a similar result along the coordinate line η(x, y) = constant, i.i.,
dy ηx
=− .
dx ηy
Using these results, equation (12) (or (13)) can be written as
 2
dy dy
A −B + C = 0. (16)
dx dx
This is called the characteristic polynomial of the PDE (2) and its roots are given by
√ √
dy B + B 2 − 4AC dy B − B 2 − 4AC
= = λ1 (x, y), = = λ2 (x, y). (17)
dx 2A dx 2A
The required variables ξ and η are determined by the respective solutions of the two ordinary
differential equations in (17), known as the characteristic equations of the PDE (2). They

5
are ordinary differential equations for families of curves in the xy− plane along which ξ
and η are constant. These families of curves depend on the coefficients A, B, and C in the
original PDE (2). Integration of equations (17) gives two families of Curvilinear coordinates
ξ(x, y) = c1 and η(x, y) = c2 , called characteristic curves of the hyperbolic equation, or the
characteristics of the equation. The fact that 4 > 0 means that the characteristics are real
curves in the xy-plane.

When the coefficients A, B, and C are constants, the two families of characteristics reduces
to two distinct families of parallel straight lines, and the change of variables are given by the
following equations:
ξ = y − λ1 x, η = y − λ2 x. (18)
The Jacobian of the transformation defined by (18) is
 
−λ1 1
J =   = λ2 − λ1 .
−λ 1

For a Hyperbolic PDE B 2 − 4AC 6= 0, and so λ1 6= λ2 . Therefore, at no point can the

particular curves from each family share a common tangent line.
∂ 2u 2
2∂ u
Example 3. Show that the one-dimensional wave equation − c = 0 is hyperbolic,
∂t2 ∂x2
find an equivalent canonical form, and then obtain the general solution.

Solution: Clearly, the wave equation is a constant coefficient PDE with A = 1, B = 0,

C = −c2 . The discriminant is 4 = 4c2 > 0, and therefore, the PDE is hyperbolic. The roots
of the characteristic polynomial are given by:
√ √
B+ 4 B− 4
λ1 = =c λ2 = = −c.
2A 2A
So the characteristic equations are :
dx dx
= c, = −c.
dt dt
Integrating the ODEs we get
x = ct + k1 , x = −ct + k2
where k1 , k2 are the constants of integration. Hence the two families of characteristics for the
wave equation are given by x−ct = constant and x+ct = constant. Thus, the transformation
ξ = x − ct, η = x + ct

6
Figure 1: Pair of characteristic curves

7
reduces the wave equation to canonical form. Here, a = 0 = c, b = − 4 A
= −4c2 , hence, we
have wξη = 0. For the wave equation the characteristics are found to be straight lines with
negative and positive slopes as shown in figure 1 Integrating the canonical form with respect
to ξ we get Z
wη = 0dξ = h(η),

where h is an arbitrary function of η. Integrating with respect to η we have:

Z
w(ξ, η) = h(η)dη + f (ξ) = f (ξ) + g(η)

where f and g are arbitrary twice differentiable functions and g is the integral of the arbitrary
function h. Therefore, the general solution is given by
u(x, t) = f (x − ct) + g(x + ct).
Note that f is constant on ”wavefronts” x = ct + ξ that travel towards right, whereas g is
constant on ”wavefronts” x = −ct + η that travel towards left. Thus any general solution
can be expressed as the sum of two waves, one travelling to the right with constant velocity
c and the other travelling to the left with same velocity c.
Example 4. Given the small disturbance potential equation
∂ 2φ ∂ 2φ
(1 − M 2 ) + = 0.
∂x2 ∂y 2

(a) Show that, depending on the Mach number, M, the small disturbance potential equation
is elliptic, parabolic, or hyperbolic.
(b) Find the characteristic variables for the hyperbolic case and hence write the equation in
canonical form.

Solution: The given equation is of the form (2), where A = 1 − M 2 m B = 0, C = 1.

The discriminant 4 = B 2 − 4AC = −4(1 − M 2 ). Thus, the PDE is hyperbolic for M > 1,
elliptic for M < 1, and parabolic for M = 1 (along the sonic line). For the case of supersonic
flow (M > 1), the roots of the characteristic polynomial are given by
√ p
B+ 4 4(M 2 − 1) 1
λ1 = = 2
= −√
2A 2(1 − M ) M2 − 1
√ p
B− 4 4(M 2 − 1) 1
λ1 = =− 2
=√ .
2A 2(1 − M ) M2 − 1

8
Therefore, we have:
dy 1 dy 1
=√ , = −√ ,
dx M2 − 1 dx M2 − 1
and integrating the two ODEs, we obtain the characteristics of the wave equation,
x x
y=√ + c1 , y = −√ + c2 ,
M2 − 1 M2 − 1
where c1 and c2 are the constants of integration. so the two families of characteristics for
x x
the wave equation are given by y − √ = constant and y + √ = constant.
2
M −1 M2 − 1
Therefore, the transformation
x x
ξ=y−√ , η=y+√
M2 − 1 M2 − 1
reduces the wave equation to canonical form as follows:
φxx = wξξ ξx2 + 2wξη ξx ηx + wηη ηx2 + wξ ξxx + wη ηxx
1 2 1
= 2
wξξ − 2 wξη + 2 wηη
M −1 M −1 M −1
φyy = wξξ ξy2 + 2wξη ξy ηy + wηη ηy2 + wξ ξyy + wη ηyy
= wξξ + 2wξη + wηη .
Substituting the relations in the given PDE we obtain
wξη = 0.
ξ =constant and η =constant represent two families of straight lines with slopes, ± √M12 −1 .

0.2.2 Parabolic equation

For a parabolic PDE the discriminant 4 = 0. Recall that to reduce this PDE to canonical
form we need to choose the new variables ξ and η such that the coefficients a and b in (6)
vanish. Thus we get a = Aξx2 + Bξx ξy + Cξy2 = 0. Dividing the above equation throughout
by ξy2 we obtain
 2  
ξx ξx
A +B + C = 0. (19)
ξy ξy
Since the total differential along the line ξ(x, y) =constant is zero, i.e., dξ = 0, we have that
dy
dξ = ξx dx + ξy dy = 0 and so, the slope is given by dx = − ξξxy , and so equation (19) can be

9
written as  2  
dy dy
A −B + C = 0, (20)
dx dx
which is the characteristic polynomial of the PDE (2). For parabolic PDE, 4 = 0, hence the
characteristic polynomial (20) has only one root
dy B
= = λ(x, y). (∗)
dx 2A
Hence there is only one family of real characteristic curves. The required variable ξ is
determined by the ODE above, which is an ordinary differential equation for families of
curves in the xy− plane along which ξ =constant.

Next we determine the second transformation variable η, by setting b = 0 in (6). That is,
2Aξx ηx + B(ξx ηy + ξy ηx ) + 2Cξy ηy = 0
 
ξx ξx
2A ηx + B ηy + ηx + 2Cηy = 0
ξy ξy
    
B B
2A − ηx + B − ηy + ηx + 2Cηy = 0
2A 2A
B2
−Bηx − ηy + Bηx + 2Cηy = 0
2A
(B 2 − 4AC)ηy = 0.
Since 4 = 0, ηy could be an arbitrary function of (x, y) and so the transformation variable η
can be chosen arbitrarily, as long as the change of variables formulas define a non-degenerate
transformation. If A, B, and C are constants integration of (∗) yields
B B B
y= x + c1 , or y − = c1 ⇒ξ=y− .
2A 2A 2A
If we take η = x, we have:
   
ξx ξy  − B 1
J =
  = 2A  = −1 6= 0,

ηx ηy   1 0
and indeed
b = 2Aξx ηx + B(ξx ηy + ξy ηx ) + 2Cξy ηy = 2A(−B/2A) + B + 0 = 0.
In the two new variables equation (6) reduces to the following canonical form
wηη = Ψ(ξ, η, w, wξ , wη ).

10
If we choose c = 0 instead of a = 0, then equation (6) reduces to
wξξ = Ψ(ξ, η, w, wξ , wη ).
∂ 2u ∂u
Example 5. Show that the one-dimensional heat equation α 2 = is parabolic, choose
∂x ∂t
the appropriate characteristic variables, and write the equation in equivalent canonical form.

Solution: The heat equation is a constant coefficient PDE with A = α, B = 0, C = 0,

and 4 = 0. Therefore, the PDE is parabolic. The single root of the polynomial is λ =
dx
B/2A = 0. Therefore, from the characteristic equation we get: = 0, and integrating
dt
yields x = k where k is a constant of integration. Hence ξ = x. If we set η = t, then the PDE
is already expressed in canonical form and so no change of variables is needed to simplify the
structure. From (5) we have
ut = wξ ξt + wη ηt = wη
uxx = wξξ ξx2 + 2wξη ξx ηx + wηη ηx2 + wξ ξxx + wη ηxx = wξξ , and a = A = α.
Therefore, the canonical form of the heat equation is given by
1
αwξξ = wη or wξξ = wη .
α

0.2.3 Eliptic equations

In this case 4 = B 2 − 4AC < 0, and we have already deduced that to reduce this PDE to
canonical form we need to choose the new variables ξ and η such that b = 0 and a = c. Hence
from (6) we have:
A(ξx2 − ηx2 ) + B(ξx ξy − ηx ηy ) + C(ξy2 − ηy2 ) = 0 (21)
2Aξx ηx + B(ξx ηy + ξy ηx ) + 2Cξy ηy = 0.
The characteristic equations of the PDE (2) are
√
dy B − i 4AC − B 2
= λ1 = (22)
dx √2A
dy B + i 4AC − B 2
= λ2 = . (23)
dx 2A
The respective solutions of the ODEs (22) determine the complex variables
α(x, y) = ξ(x, y) + iη(x, y), β(x, y) = ξ(x, y) − iη(x, y).

11
Integration of (22) leads to the curves α(x, y) = c1 and β(x, y) = c2 , respectively, where c1 , c2
are complex constants of integration. Since α and β are complex functions the characteristic
curves of the elliptic equation are not real. The real and imaginary parts of α and β give the
required transformation variables ξ and η. Thus, we have
α+β α−β
ξ= , η= . (24)
2 2i
Hence equation (6) becomes
wξξ + wηη = ψ(ξ, η, w, wξ , wη .) (25)

where Ψ = fa .
Example 6. Show that the equation uxx +x2 uyy = 0 is elliptic everywhere except on the coor-
dinate axis x = 0, find the characteristic variables and hence write the equation in canonical
form.

Solution: Here, A = 1, B = 0, , C = x2 and the discriminant 4 = −4x2 < 0 for x 6= 0,

and therefore the PDE is elliptic. The roots of the characteristic polynomial are given by
√ √
B − i 4AC − B 2 B + i 4AC − B 2
λ1 = = −ix, λ2 = = ix.
2A 2A
Therefore, we have
dy dy
= −ix, = ix.
dx dx
Hence the characteristics of the elliptic equation are
x2 x2 x2 x2
y = −i + c1 , y=i + c2 ⇒ α = y + i , β =y−i ,
2 2 2 2
and so we have
x2
ξ = y, η= ⇒ uxx = x2 wηη + wη , uyy = wηη .
2
Substituting these relations in the PDE, we obtain
1
wξξ + wηη = − wη .
2η

12
Bibliography

[1] A. Salih, Classification of Partial Differential Equations and Canonical forms, De-
partment of Aerospace Engineering, Indian Institute of Space Science and Technology,
Thiruvananthapuram, 22 December 2014.
[2] Yeduda Pnchover and Jacob Rubinstein, An Introduction to Partial Differential Equa-
tions, New York, 2005.
[3] Ravi P. Agarwal and Donald O’Regan, Differential Equations, with Special Functions,
Fourier Series and Boundary Value Problems, Springer Science + Business Media,
2009.
[4] Jeffrey R. Chasnov, Introduction to Differential Equations. Lecture notes for MATH
2351/2352 , The Hong Kong University of Science and Technology, Department of
Mathematics.
[5] Marcel B. Finan, Lecture Notes in Mathematics Arkansas Tech University, Department
of Mathematics, 2018.
[6] StGoChap6.pdf, Partial differential equations, www.cns.gatech.edu .
[7] Ally Learn, First order partial differential equations, Ally Learn-Youtube.

13