JOMO KENYATTA UNIVERSITY OF AGRICULTURE AND

TECHNOLOGY
PARTIAL DIFFERENTIAL EQUATIONS NOTES

Course content
• Surfaces and curves in three dimensions. Simultaneous differential equations of the first order.
Methods of solution of

• Orthogonal trajectories of system of curves on a surface.

• Linear partial differential equations of the first order.

• Partial differential equations of the second order: Laplace, Poisson, heat and wave equations.
Methods of the solution by separation of the variables for Cartesian, spherical polar and cylin-
drical polar coordinates, and by Laplace and Fourier transform.

• Applications to engineering.

References
[1] Elements of PDE by Ian N. Sneddon

[2] Advanced Engineering Mathematics (10th ed.) by Erwin Kreyszig

[3] Schaum’s Outline Series: Theory and problems of PDE

LECTURE 1

1 Review of basic concepts

1.1 Partial differentials
Let u = u(x, y), where x = x(s, t) and y = y(s, t). The partial differential of u is defined by
∂u ∂u
∂u = ∂x + ∂y.
∂x ∂y
Therefore, the partial derivative of u with respect to s and t is defined by
∂u ∂u ∂x ∂u ∂y
= +
∂s ∂x ∂s ∂y ∂s
and
∂u ∂u ∂x ∂u ∂y
= + ,
∂t ∂x ∂t ∂y ∂t
respectively.

Example(s):

1
 ∂u ∂u
1. Given that u = x2 y + y 2 , x = 2st, and y = tes . Find and .
∂s ∂t

Solution
∂u ∂u ∂x ∂u ∂y
= + = (2xy)(2t) + (x2 + 2y)(tes ) = 2(2st · tes )(2t) + ((2st)2 + 2tes )(tes )
∂s ∂x ∂s ∂y ∂s
= 8st3 es + 4s2 t3 es + 2t2 e2s
∂u ∂u ∂x ∂u ∂y
= + = (2xy)(2s) + (x2 + 2y)(es ) = 2(2st · tes )(2s) + ((2st)2 + 2tes )(es )
∂t ∂x ∂t ∂y ∂t
= 8s2 t2 es + 4s2 t2 es + 2te2s

Exercise:

2. Given that u is a function of x and y. If x = r cos θ and y = r sin θ, show that

( ) ( ) ( ) ( )2
∂u 2 1 ∂u 2 ∂u 2 ∂u
(a) + 2 = +
∂r r ∂θ ∂x ∂y
2
∂ u 1 ∂u 1 ∂2u 2
∂ u ∂ u2
(b) + + = + 2
∂r2 r ∂r r2 ∂θ2 ∂x2 ∂y

1.2 Jacobian
Let F and G be functions of x, y and z and define a partial vector differential operator ∇
⃗ by

⃗ = î ∂ + ĵ ∂ + k̂ ∂
∇
∂x ∂y ∂z
Thus,
⃗ = î ∂F + ĵ ∂F + k̂ ∂F
∇F ⃗ = î ∂G + ĵ ∂G + k̂ ∂G
and ∇G
∂x ∂y ∂z ∂x ∂y ∂z
The cross product of ∇F
⃗ and ∇G
⃗ is defined by

î ĵ k̂
∇F × ∇G = Fx Fy Fz
⃗ ⃗ = (Fy Gz − Fz Gy )î + (Fz Gx − Fx Gz )ĵ + (Fx Gy − Fy Gx )k̂
Gx Gy Gz
Fy Fz F Fx F Fy
= î + z ĵ + x k̂
Gy Gz Gz Gx Gx Gy
∂(F, G) ∂(F, G) ∂(F, G)
= î + ĵ + k̂
∂(y, z) ∂(z, x) ∂(x, y)
= P î + Qĵ + Rk̂

The components of the above vector are called direction cosines or Jacobian of F and G.

2 Surfaces and curves in three dimensions

If the rectangular Cartesian coordinates (x, y, z) of a point P in three dimensions are connected by a
relation of the form

F (x, y, z) = 0 (1)

then the point P lies on the surface.

2
We call equation (1) the equation of a surface. For example, the equation x2 + y 2 + z 2 = a2 is the
equation of a sphere centre (0, 0, 0) and radius a. The total differential of equation (1) is given by
∂F ∂F ∂F
dF ≡ dx + dy + dz = 0
∂x ∂y ∂z
which can be interpreted in vector form as
( ) ( )
∂F ∂F ∂F ∂F ∂F ∂F
î + ĵ + k̂ · (dxî + dy ĵ + dz k̂) = 0 ⇒ , , · (dx, dy, dz) = 0
∂x ∂y ∂z ∂x ∂y ∂z
( )
∂F ∂F ∂F
Thus, , , are the direction ratios of the normal to the surface at point P while (dx, dy, dz)
∂x ∂y ∂z
are direction ratios of the tangent to the surface at the same point.

→ Note: a surface may be envisaged as being generated by a curve. A curve is obtained as the
intersection of any two surfaces. For instance, a conic is a curve obtained as the intersection of the
surface of a cone with a plane. Conics include hyperbola, parabola, and ellipse.

Example(s):
(a) Determine the direction cosines of the tangent at the point (x, y, z) to the conics, x2 y + xz = 0
and x2 + y 2 + z 2 = 0.

Solution
Here, F = x2 y + xz and G = x2 + y 2 + z 2 . So,
∂(F, G)
P = = Fy Gz − Fz Gy = 2x2 z − 2xy = 2x(xz − y)
∂(y, z)
∂(F, G)
Q = = Fz Gx − Fx Gz = 2x2 − 2z(2xy + z)
∂(z, x)
∂(F, G)
R = = Fx Gy − Fy Gx = 2y(2xy + z) − 2x3
∂(x, y)

(b) Show that the direction cosines of the tangent at the point (x, y, z) to the conics ax2 +by 2 +cz 2 = 1
and x + y + z = 1 are proportional to (by − cz, cz − ax, ax − by).

Solution
Here, F = ax2 + by 2 + cz 2 − 1 and G = x + y + z − 1. So,
∂(F, G)
P = = Fy Gz − Fz Gy = 2by − 2cz
∂(y, z)
∂(F, G)
Q = = Fz Gx − Fx Gz = 2cz − 2ax
∂(z, x)
∂(F, G)
R = = Fx Gy − Fy Gx = 2ax − 2by
∂(x, y)
Therefore, (P, Q, R) = 2(by − cz, cz − ax, ax − by) ⇒ (P, Q, R) ∝ (by − cz, cz − ax, ax −
by), where 2 is the constant of proportionality.

3
2.1 Simultaneous differential equations of the first order
Consider a curve in space which is the intersection of the surfaces F (x, y, z) = 0 and G(x, y, z) = 0.
It follows that
∂F ∂F ∂F ∂F ∂F ∂F
dF = 0 ⇒ dx + dy + dz = 0 ⇒ dx + dy = − dz · · · (i)
∂x ∂y ∂z ∂x ∂y ∂z
∂G ∂G ∂G ∂G ∂G ∂G
dG = 0 ⇒ dx + dy + dz = 0 ⇒ dx + dy = − dz · · · (ii)
∂x ∂y ∂z ∂x ∂y ∂z
In matrix form we have
   ∂F 
∂F ∂F  
 ∂x ∂y  
− dz 
  dx   ∂z 
  = 
   ∂G 
(2)
 ∂G ∂G  dy
− dz
∂x ∂y ∂z
 
∂F ∂F ∂F ∂F
 ∂x ∂y  ∂x ∂y
 
Let A = 

. Then, Det(A) = |A| =
 = Fx Gy − Fy Gx . Solving equation (2) using
 ∂G ∂G  ∂G ∂G
∂x ∂y ∂x ∂y
Cramer’s rule we obtain
∂F ∂F
− dz
∂z ∂y

∂G ∂G
− dz
∂z ∂y F y Gz − F z Gy dx dz
dx = = dz ⇒ =
|A| F x Gy − F y Gx Fy Gz − Fz Gy F x Gy − F y Gx
and
∂F ∂F
− dz
∂x ∂z
∂G ∂G
− dz Fz Gx − Fx Gz dy dz
dy = ∂x ∂z = dz ⇒ =
|A| Fx Gy − Fy Gx Fz Gx − Fx Gz Fx Gy − Fy Gx
Therefore,
dx dy dz dx dy dz
= = ⇒ = = (3)
Fy Gz − Fz Gy Fz Gx − Fx Gz Fx Gy − Fy Gx P Q R
Equations (3) are called symmetric equations and are the differential equations of the integral curves
(intersection of two surfaces).

Lecture 2

dx dy dz
2.2 Methods of solution of = =
P Q R
The reverse problem is that of finding the surfaces F (x, y, z) = 0 and G(x, y, z) = 0 from the equations
(3). We need to obtain two unique solutions (integral curves), say u(x, y, z) = c1 and v(x, y, z) = c2 ,
where c1 and c2 are arbitrary constants. There are 3 methods of solving equations (3):
1. Method of multipliers (intuition)

2. One variable absent (method of grouping)

3. Exact differential method

4
2.2.1 Method of Multipliers (method of intuition)
This method uses multipliers l, m and n (which are not necessarily constants) so that for a constant
k, equation (3) becomes

ldx + mdy + ndz

=k ⇒ ldx + mdy + ndz = k(lP + mQ + nR)
lP + mQ + nR

These multipliers are chosen such that the condition lP + mQ .+ nR = 0 is satisfied. Hence,

ldx + mdy + ndz = 0,

which on integration yields the first integral curve, u(x, y, z) = c1 . It may be possible to get another set
of multipliers l, m and n to obtain the second integral curve, v(x, y, z) = c2 . Therefore, the integral
surface generated by these curves is given by ϕ(u, v) = 0, where ϕ is an arbitrary function.

Example(s):
Find the integral curves and the integral surface of the following differential equations.
dx dy dz
(a) = =
x(y − z) y(z − x) z(x − y)

Solution
From the given equations we have P = x(y −z), Q = y(z −x) and R = z(x−y). Using the
multipliers l, m and n, we require that lP +mQ+nR = 0 · · · (∗) so that ldx+mdy+ndz = 0.
Now,
i) By intuition, condition (∗) is satisfied if we take l = 1, m = 1 and n = 1. Hence,
dx + dy + dz = 0. Integrating yields

x + y + z = c1
1 1 1
ii) Again by intuition, condition (∗) is satisfied if we take l = , m = and n = . Hence,
x y z
dx dy dz
+ + = 0. Integrating yields
x y z
ln x + ln y + ln z = ln c2 ⇒ xyz = c2

Therefore, the required integral curves are given by x + y + z = c1 and xyz = c2 , where
c1 and c2 are parameters. The integral surface generated by these curves is ϕ(c1 , c2 ) = 0.
That is,
ϕ(x + y + z, xyz) = 0

2.2.2 One variable absent

If one of the variables is absent in one of the equations, we can easily derive one of the solutions. Also,
it is possible to use the first solution to obtain a second solution by eliminating one of the variables
in one of the equations.

Example(s):
Find the integral curves of the following equations.
dx dy dz
(a) = =
x+z y z + y2

Solution
Since x is absent in the second and third equations, we take
dy dz dz 1
= ⇒ − z=y
y z + y2 dy y

5
 This is a linear ODE in z with integrating factor
∫ −1 1
= e− ln y =
dy
µ=e y
y
( )
1 d z
Multiplying through by we get = 1. Integrating yields
y dy y
z
= y + c1 ⇒ z = y 2 + c1 y
y
Using the first solution to eliminate z in the first equation, we have
dx dy dx dy dx 1
= ⇒ 2
= ⇒ − x = y + c1
x+z y x + y + c1 y y dy y
This is a linear ODE in x with integrating factor given by
∫ −1 1
= e− ln y =
dy
µ=e y
y
( )
1 d x c1
Multiplying through by yields =1+ . Integrating yields
y dy y y
x
= y + c1 ln y + c2 ⇒ x = c1 y ln y + c2 y + y 2
y

Therefore, the required integral curves are given by z = c1 y + y 2 and x = c1 y ln y + c2 y + y 2 ,

where c1 and c2 are parameters.
dx dy dz
(b) = = . [ans: x = y + c1 , z(2y + x − y) = −y 2 − (x − y)y + c2 ]
x+y x+y −(x + y + 2z)
dx dy dz y
(c) = = 2 . [ans: z = x2 + c1 x, = x + c1 ln x + c2 ]
x y+z x +z x

2.2.3 Exact differential method

It is much simpler to try to cast the given differential equations into a form which suggests their
solution. Note that adding/subtracting the numerators and denominators of any two “fractions”
doesn’t alter their value. E.g, we may write
dx dy dz ldx + mdy + ndz dx + dy dx − dy − 2dz
= = = = = , e.t.c.
P Q R lP + mQ + nR P +Q P − Q − 2R
So that if lP + mQ + nR = 0, then ldx + mdy + ndz = 0 is an exact differential. This method is best
illustrated in the following example.

Example(s):
Find the integral curves of the following equations:
dx dy dz
(a) = = .
y(x + y) + az x(x + y) − az z(x + y)

Solution
We have
dx dy dz dx + dy xdx − ydy
= = = =
y(x + y) + az x(x + y) − az z(x + y) (x + y)2 az(x + y)

dx + dy dz dx + dy dz
From = ⇒ = ⇒ ln(x + y) = ln z + ln c1
(x + y)2 z(x + y) x+y z
⇒ x + y = c1 z

6
 xdx − ydy dz
From = ⇒ xdx − ydy − adz = 0
az(x + y) z(x + y)

⇒ x2 − y 2 − 2az = c2

Therefore, the required integral curves are given by x + y = c1 z and x2 − y 2 − 2az = c2 ,

where c1 and c2 are parameters.

Exercise:
Use any method to find the integral curves of the following equations.
dx dy dz
(a) = = .
y 2 −xy x(z − 2y)

Solution
dx dy
From 2 = ⇒ xdx + ydy = 0
y −xy

⇒ x2 + y 2 = c1
dy dz
From = ⇒ zdy + ydz − 2ydy = 0 ⇒ d(zy) − 2ydy = 0
−xy x(z − 2y)

⇒ zy − y 2 = c2

Therefore, the integral curves of the given differential equations are the members of the
two-parameter family x2 + y 2 = c1 and zy − y 2 = c2 .
dx dy dz
(b) = = .
x2 −y −z
2 2 2xy 2xz

Solution
dy dz dy dz
From = ⇒ = ⇒ ln y = ln z + ln c1 ⇒ y = c1 z
2xy 2xz y z
xdx + ydy + zdz dy 2(xdx + ydy + zdz) dy
From = ⇒ =
x(x2 − − z ) + y(2xy) + z(2xz)
y2 2 2xy 2 2 2
(x + y + z ) y

⇒ ln(x2 + y 2 + z 2 ) = ln y + ln c2 ⇒ x2 + y 2 + z 2 = c2 y

Therefore, the integral curves of the given differential equations are the members of the
two-parameter family y = c1 z and x2 + y 2 + z 2 = c2 y.
dx dy dz
(c) = = .
xz − y yz − x 1 − z2

Solution
dx + dy dz dx + dy −dz
From = ⇒ =
xz − y + yz − x 1 − z2 x+y 1+z
⇒ (x + y)(1 + z) = c1
xdx − ydy zdz xdx − ydy zdz
From = ⇒ =
x(xz − y) − y(yz − x) z(1 − z 2 ) x −y
2 2 1 − z2

⇒ (x2 − y 2 )(1 − z 2 ) = c2

Therefore, the integral curves of the given differential equations are the members of the
two-parameter family (x + y)(1 + z) = c1 and (x2 − y 2 )(1 − z 2 ) = c2 .
dx dy dz
(d) = = . [ans: x2 + y 2 + z 2 = c1 , xyz = c2 ]
x(y 2 −z )
2 y(z − x )
2 2 z(x − y 2 )
2

7
 dx dy dz
(e) = = 2 . [ans: x2 − y 2 + z 2 = c1 , xy − z = c2 ]
y − xz x + yz x + y2
dx dy dz
(f) = = . [ans: x4 + y 4 + z 4 = c1 , xyz 2 = c2 ]
x(2y 4 − z 4 ) y(z 4 − 2x4 ) z(x4 − y 4 )
dx dy dz x−y
(g) = 2 = 2 . [ans: = c1 , xy + yz + xz = c2 ]
x2 − yz y − xz z − xy y−z
dx dy dz
(h) = = . [ans: xy = c1 , (x + y)(x + y + z) = c2 ]
x(x + y) −y(x + y) −(x − y)(2x + 2y + z)

Lecture 3

2.3 Orthogonal trajectories of a system of curves on a surface

Given a surface

F (x, y, z) = 0 (4)

and a family (or system) of curves on it, we need to find another family of curves each of which lies on
the surface (4) and cuts every curve of the original family at right angles. The new family of curves
is called the family of orthogonal trajectories on the surface (4).

The original family of curves may be thought of as the intersections of the surface (4) with the one-
parameter family of surfaces
G(x, y, z) = k
From F (x, y, z) = 0 and G(x, y, z) = k, the tangential direction (dx, dy, dz) to the given curve through
the point (x, y, z) on the surface (4) satisfies the equations

dF = 0 i.e, Fx dx + Fy dy + Fz dz = 0

and
dG = 0 i.e, Gx dx + Gy dy + Gz dz = 0
We solve the above two equations using Cramer’s rule to obtain
dx dy dz ∂(F, G) ∂(F, G) ∂(F, G)
= = , where P = , Q= , and R =
P Q R ∂(y, z) ∂(z, x) ∂(x, y)

The curve through (x, y, z) of the orthogonal family has tangential direction (dx′ , dy ′ , dz ′ )

which lies on the surface (4), so that

Fx dx′ + Fy dy ′ + Fz dz ′ = 0 (5)

and is perpendicular to the original system of curves. Also since the vectors (dx, dy, dz) and (dx′ , dy ′ , dz ′ )
are orthogonal, then

(dx, dy, dz) · (dx′ , dy ′ , dz ′ ) = 0 ⇒ dxdx′ + dydy ′ + dzdz ′ = 0

8
i.e.,

P dx′ + Qdy ′ + Rdz ′ = 0 (6)

Solving for dx′ , dy ′ and dz ′ between equations (5) and (6) using Cramer’s rule, we obtain

dx′ dy ′ dz ′
= =
RFy − QFz P Fz − RFx QFx − P Fy

This can be rewritten as:

dx′ dy ′ dz ′ Fy Fz b = Fz Fx , R
b = Fx Fy
= = , where Pb = , Q (7)
Pb b
Q b
R Q R R P P Q

The solution of equations (7) alongside with (4) gives the system of orthogonal trajectories.

Example(s):

(a) Find the orthogonal trajectories on the surface x2 + y 2 = αz 2 of its intersections with the family
of planes parallel to xy-plane (i.e, z = k)

Solution
Let F (x, y, z) = x2 + y 2 − αz 2 = 0 and G(x, y, z) = z = k. Then,

∂(F, G)
P = = Fy Gz − Fz Gy = 2y − (−2αz)(0) = 2y
∂(y, z)
∂(F, G)
Q = = Fz Gx − Fx Gz = (−2αz)(0) − 2x(1) = −2x
∂(z, x)
∂(F, G)
R = = Fx Gy − Fy Gx = 2x(0) − 2y(0) = 0
∂(x, y)

Fy Fz 2y −2αz
Pb = = = −4αxz
Q R −2x 0

b = Fz Fx −2αz 2x
Q = = −4αyz
R P 0 2y

b = Fx Fy 2x 2y
R = = −4x2 − 4y 2
P Q 2y −2x

The symmetric equations for the orthogonal trajectories are given by:
dx dy dz dx dy dz dx dy αzdz
= = ⇒ = = ⇒ = = 2
Pb b
Q b
R −4αxz −4αyz −4x − 4y 2
2 x y x + y2

dx dy dx dy
From = ⇒ = ⇒ ln x = ln y + ln c1 ⇒ x = c1 y, where c1 is a
x y x y
parameter.

xdx + ydy αzdz

From 2 2
= 2 ⇒ xdx + ydy = αzdz ⇒ x2 + y 2 = αz 2 + c2 ⇒
x +y x + y2
c2 = 0
∴ x2 + y 2 = αz 2 . Hence F ≡ 0 is necessarily one of the solutions. Therefore, the orthogonal
trajectories are the curves
x = c1 y and x2 + y 2 = αz 2

(b) Find the equations of the system of curves on the cylinder 2y = x2 orthogonal to its intersections
with the hyperboloids of the one-parameter system xy = z + c, where c is a parameter.

9
 Solution
Let F = 2y − x2 = 0 and G = xy − z = c. The partial derivatives are:
⇒ Fx = −2x, Fy = 2, Fz = 0, Gx = y, Gy = x, Gz = −1

∂(F, G)
P = = Fy Gz − Fz Gy = −2
∂(y, z)
∂(F, G)
Q = = Fz Gx − Fx Gz = −2x
∂(z, x)
∂(F, G)
R = = Fx Gy − Fy Gx = −2x2 − 2y = −2(x2 + y)
∂(x, y)

Fy Fz 2 0
Pb = = = −4(x2 + y)
Q R −2x −2(x2 + y)

b = Fz Fx 0 −2x
Q = = −4x(x2 + y)
R P −2(x + y) −2
2

b = Fx Fy −2x 2
R = = 4(x2 + 1)
P Q −2 −2x
dx dy dz
The symmetric equations for the orthogonal trajectories are = =
Pb b
Q b
R
dx dy dz dx dy −dz
⇒ = = ⇒ = = 2
−4(x + y)
2 −4x(x + y)
2 2
4(x + 1) x2+y 2
x(x + y) x +1
dx dy x2 c1
From = ⇒ xdx = dy ⇒ =y+ ⇒ x2 = 2y + c1
x2 + y x(x2 + y) 2 2
⇒ c1 = 0 ∴ x2 = 2y
dx −dz x2
From = . Substituting y = yields
x2 + y x2 + 1 2
dx −dz 2dx −dz 2(x2 + 1)dx
x2
= ⇒ = 2 ⇒ = −3dz
x2 + 2
x2 + 1 3x 2 x +1 x2
( ) ( )
1 1
3dz + 2 1 + 2 dx = 0. Integrating yields 3z + 2 x − = c2
x x
Therefore, the orthogonal trajectories are the curves
( )
1
3z + 2 x − = c1 , 2y = x2
x
Exercise:
(a) [Assignment 1 ] Find the orthogonal trajectories on surface of the sphere x2 + y 2 + z 2 = a2 of
its intersections with the paraboloids xy = cz, where c is a parameter. [ans:
a −x
2 2
x2 + y 2 + z 2 = a2 , = c2 ]
a2 − y 2
(b) Show that the orthogonal trajectories on the hyperboloid x2 + y 2 − z 2 = 1 of the conics in
which it is cut by the system of planes x + y = c are its curves of intersection with the surfaces
(x − y)z = k, where k is a parameter.
(c) Find the orthogonal trajectories on the conicoid (x + y)z = 1 of the conics in which it is cut by
the system of planes x − y + z = k, where k is a parameter. [ans: The orthogonal trajectories
1 1
are the curves x = z − 3 + + c2 , (x + y)z = 1]
6z 2z

Lecture 4

10