Partial Differential Equations PDF

Directorate of Distance Education
Guru Jambheshwar University of Science & Technology
Hisar (Haryana)
M. Sc. (Mathematics)
MAL-632
Partial Differential Equations
1
Contents
Chapters Page No.
Lesson 1: Preliminaries 3-20
Lesson 2:Solution of Linear Partial 21-51
Differential Equations
Lesson 3:Green’s Function 52- 77
Lesson 4:Solution of Wave Equation 78-102
Lesson 5 :Other Techniques to
R epresent Solution 103-126
2
Lesson 1: Preliminaries
Written by: Dr. Sunita Rani
Vetted by: Prof. Sarva Jit Singh
Structure
1.1 Introduction
1.2 Notations
1.3 Inequalities
1.4 Calculus
1.5 Function Analysis Concepts
1.6 Definition and classification of Partial Differential Equations
1.7 Definition and classification of system of Partial Differential
Equations
1.8 Solutions of Partial Differential Equations

1.9 Suggested References
1.1 Introduction:
The reader must be familiar with the several notations and certain
functional analysis concepts before studying the partial differential
equations. Here, first we give brief description and then define partial
differential equations.
1.2 Notations
(a) Geometric notations
(i) R n = n-dimensional real Euclidean space
(ii) R1 = R = real line

(iii) ei = unit vector in the ith direction
3
= (0, 0, 0, …1, … 0)
(iv) A point x in Rn is x = ( x1, x 2 ,... xn )
(v) {
R +n = x = ( x1 , x 2 , ... xn ) ∈ R n xn > 0 }
= open upper half-space.
(vi) A point in Rn+1 will be denoted as
( x , t ) = ( x1 , x 2 , ..., xn , t )
where t is time variable.
(vii) U, V, W denote open subsets of Rn. We write
V⊂⊂U if V ⊂ V ⊂ U and V is compact
i.e. V is compactly contained in U.

(viii) ∂U = boundary of U
U = closure of U = U ∪ ∂ U
(ix) U T = U × ( 0, T ]
(x) ΓT = U T − U T
= parabolic boundary of U T
(xi) {
B 0 ( x, r ) = y ∈ R n x − y < r }
= open ball in Rn with centre x and radius r > 0
(xii) {
B ( x ,r ) = y ∈ R n x − y ≤ r }
=closed ball in Rn with centre x and radius r > 0
(xiii) α (n ) = volume of unit ball B (0, 1) in Rn
r n /2
=
⎛n ⎞
Γ ⎜ + 1⎟
⎝2 ⎠
nα (n ) = surface area of unit sphere B (0, 1) in Rn
(xiv) If a, b ∈ Rn s.t .
4
a = (a1, a 2 , ..., an ) and b = (b1, b2 , ..., bn ) then
n
a.b = ∑a b
i =1
i i and
1/2
⎛ n ⎞
a = ⎜ ∑ ai2 ⎟
⎝ i =1 ⎠
(xv) denotes n dimensional complex plane.
(b) Notations for functions
(i) If u : U → R , we write
u ( x ) = u ( x1 , x 2 , ..., xn ) where x ∈ U .
(ii) u is smooth if u is infinitely differentiable.

(iii) If u, v are two functions, we write
u ≡ v if u, v agree for all arguments
u :=v means u is equal to v.

(iv) The support of a function u is defined as the set of points where the
function is not zero and is denoted by spt u.
{
spt u = x ∈ X f ( x ) ≠ 0 }
In other words, spt u is the closure of the set u where u does not
vanish.
(v) The sign function is defined by
⎧1 if x > 0
⎪
sgn x = ⎨0 if x = 0
⎪
⎩ −1 if x < 0
u + = max ( u, 0 )
u − = − min ( u, 0 )
u = u+ − u−
5
|u |= u + + u −
(vi) If u : U → R m
u ( x ) = ( u1 ( x ) ,......., u m ( x ) ) ( x ∈U )
where u = ( u1, u 2 ,..., u m )
(vii) The symbol
∫ fdS
Σ
denotes the integral of f over (n - 1) dimensional surface Σ in Rn.

(viii) The symbol
∫ f dl
C
denotes line integral of f over the curve C in Rn.

(ix) The symbol
∫ f dx
V
denote the volume integral of S over V ∈ R n and x ∈ V is an
arbitrary point.
(x) Averages:
1
∫
B ( x ,r )
f dy =
α (n ) r n ∫
B ( x ,r )
f dy
= average of f over ball B (x, r)

1
∫
∂B (n , r )
f dS =
nα (n ) r n −1 ∫
∂B (n , r )
f dS
= average of f over surface of ball B (x, r)

(xi) A function u : U → R is called Lipschitz continuous if
u ( x ) − u (y ) ≤ C x − y
for some constant C and all x, y ∈ U .We denote
6
u ( x ) − u (y )
Lip [u] = sup
x , y∈U x −y
x ≠y
(xii) The convolution of functions f, g is denoted by f * g.
(c) Notations for derivatives
Let u :U → R , x ∈ U
∂u ( x ) u ( x + hei ) − u ( x )
(i) = lt
∂x i h →0 h
∂u
provided that the limit exists.We denote by u xi
∂xi
∂ 2u
Similarly by u xi x j
∂x i ∂x j
∂ 3u
by u xi x j xk etc.
∂x i ∂x j ∂xk
(ii) Multi-index Notation
(a) A vector α of the form α = (α1 , ..., αn ) where each αi is a non-
negative integer, is called a multi-index of order
α = α1 + α 2 + ... + αn
(b) Given multi-index α , define
∂ u (x )
α
D u (x ) =
α
∂x1α1 ...∂xnαn
(c) If k is a non-negative integer
{
Dku (x ) = Dαu (x ) , α = k }
the set of all partial derivatives of order k
e.g. for k = 1
7
(
Du = u x1 , ..., u xn )
= grad(u)
for k = 2
⎛ ∂ 2u ∂ 2u ∂ 2u ⎞
⎜ ... ⎟
∂x12 ∂x1∂x 2 ∂x1∂xn ⎟
D 2u = ⎜ 2
⎜ ∂u ∂ 2u ∂ 2u ⎟
⎜⎜ ... ⎟
⎝ ∂xn ∂x1 ∂xn ∂x 2 ∂xn2 ⎟⎠
is called Hessian Matrix.

1/2
⎧⎪ 2⎫
⎪
(d) Dk u (x ) = ⎨ ∑ Dα u ⎬
⎩⎪ α =k ⎭⎪
n
(iii) Δu = ∑ u xi xi
i =1
= Laplacian of u
= trace of Hessian Matrix.
(iv) Let x , y ∈ R n i.e. x = (x1, …, xn), y = (y1, …, yn)
Then we write
(
D x u = u x1 , ..., u xn )
(
Dy u = uy1 , ..., uyn )
the subscript x or y denotes the variable w.r.t. differentiation is
being taken
(d) Function Spaces
(i) C (U ) = {u : U → R u is continuous}
(ii) C (U ) = {u ∈ C ( u ) u is uniformly continuous
on bounded subsets of U}
8
(iii) C k (U ) = {u : U → R u is k times continuous differentiable }
(iv) {
C k (U ) = u : C k (U ) D α u is uniformly continuous
on bounded subsets of U, for all α ≤ k }
(v) C ∞ (U ) = {u : U → R|u is infinitly differentiable}
(vi) CC (U ) means C (U) has compact support.
Similarly , CCK (U ) means C k (U ) has compact support.
(vii) The function u : U → R is Lebsegue measurable over Lp if
u L p (U )
<∞
The function u : U → R is Lebsegue measurable over L∞ if
u L∞ (U )
<∞
where
1/ p
⎛ p ⎞
u L p (U )
= ⎜ ∫ u dx ⎟ 1≤ p ≤ ∞
⎝U ⎠
u L∞ (U )
= ess supU |u |
(viii) {
L p (U ) = u : U → R u is Lebsegue measurable over L p }
L∞ (U ) = u : U → R{ u is Lebsegue measurable over L∞ }
(ix) Du L p (U )
= Du
L p (U )
Similarly D 2u = D 2u
L p (U ) L p (U )
(x) If u : U → R m is a vector
where u = ( u1, u 2 , ..., u m )
then
9
D k u = {D α u, α = k }
Similarly other operators follow.

Big Oh (O) order
We say
f = O ( g ) as x → xO provided there exists a constant C such that
f (x ) ≤ C g (x )
for all x sufficiently close to xO

Little Oh (o) order
We say
f = o ( g ) as x → x o provided
f (x )
lt →0
x → xo g (x )
1.3 Inequalities
There are some fundamental inequalities
(a) Cauchy’s Inequality
a 2 b2
ab ≤ + (a , b ∈ R )
2 2
(b) Holder’s Inequality
1 1
Let 1 ≤ p, q ≤ ∞ ; + =1
p q
u ∈ Lp ( u ) ,v ∈ Lq ( u )
∫ uv dx
U
≤ u L p (U )
v Lq (U )
10
(c) Minkowski’s Inequality
Let 1 ≤ p ≤ ∞, and u, v ∈ L p (U )
Then u + v LP (U )
≤ u L p (U )
+ v LP (U )
(d) Cauchy Schwartz Inequality
x .y ≤ x y ( x, y ∈ R ) n
1.4 Calculus
(a) Boundaries
Let U ⊂ R n be open and bounded, k = {1, 2, ..., }

Definitions: (i) The boundary ∂U is Ck if for each point x 0 ∈ ∂U there
exists r > 0 and a Ck function ϒ : R n −1 → R such that
{
U ∩ B ( x °, r ) = x ∈ B ( x °, r ) xn > ϒ ( x1,..., xn −1 ) }
Similarly, ∂U is C ∞ if ∂U is C k (k = 1, 2,...)
Also , ∂U is analytic if ϒ is analytic.
(ii) If U is C 1 , then along ∂U , the outward unit normal at any point
x 0 ∈ ∂U is denoted by ν ( x 0 ) = (ν 1,...,ν n ) .
(iii) Let u ∈ C 1 (U ) then normal derivative of u is denoted by
∂u
= ν . Du
∂ν
(b) Gauss- Green Theorem
Let U be a bounded open subset of Rn and ∂U bes C 1. u :U → R n
also u ∈ C 1 (U ) then
11
∫u xi dx = ∫ uv
i
dS (i = 1, 2, ..., n )
U ∂U
(c) Integration by parts formula
Let u, v ∈ C 1 (U ) then
∫u v dx = − ∫ uv xi dx + ∫ uv ν
i
xi dS
U U ∂U
Proof. By Gauss -Green’s theorem
⌠
⎮ ( uv )xi dx = ∫ ( uv ) ν i dS
⌡ ∂U
U
⌠
or ⎮ u xi v dx + ⌠
⎮ uv xi dx = ∫ ( uv )ν i dS
⌡ ⌡ ∂U
U
U
⌠
or ⎮ u xi v dx = −⌠
⎮ uv xi + ∫ ( uv ) v i dS
⌡ ⌡ ∂U
U
U
(d) Green’s formulas
Let u, v ∈ C 2 (U ) then
⌠ ∂u
(i)
U
∫ Δu dx = ⎮⌡ ∂ν dS
∂U
Proof.
U
∫ Δudx = ∫ (u ) xi
xi
dx
Integrating by parts , taking the 2nd function as unity
= ∫u
∂U
xi v i dS
∂u
= ⌠
⎮ dS
⌡ ∂v
∂U
Hence the result.
12
⌠ ∂v
(ii) ∫ Du.Dv dx = − ∫ u Δv dx + ⎮⌡ ∂ν u dS
U U
∂U
Proof. ∫ Du.Dv dx = − ∫ u Δv dx + ∫ u Dv.ν dS

U U ∂U
(integrating by parts)
∂v
= − ∫ u Δv dx + ⌠
⎮u dS
U
⌡ ∂ν
∂U
⌠⎛ ∂v ∂u ⎞
(iii) ∫ (u Δv − v Δu ) dx = ⎮⌡ ⎜⎝ u ∂ν
U
−v ⎟ dS
∂ν ⎠
∂U
⌠ ∂v
Proof. ∫ u Δv dx = − ∫ Du .Dv dx + ⎮⌡ ∂ν udS
U U
∂U
⌠ ∂u
Similarly, ∫ v Δu dx = − ∫ Du. Dv dx + ⎮⌡ ∂ν udS
U U
∂U
Subtracting, we get the result.
(e) Conversion of n-dimensional integrals into integral

over sphere
(i) Coarea formula
Let u : R n → R be Lipschitz continuous and assume that for a. e . r ∈ R ,
the level set
{x ∈ R n
u (x ) = r }
is a smooth and n –1 dimensional surface in R n . Suppose also
f : R n → R is smooth and summable. Then

∞
⌠⎛ ⎞
∫ f Du dx = ⎮ ⎜ ∫ f dS ⎟ dr
⎮ ⎜ {u =r } ⎟
Rn ⌡⎝ ⎠
−∞
13
Cor. Taking u( x ) = x − x 0
Let f : R n → R be continuous and summable then

∞
⌠⎛ ⎞
∫ f dx = ⎮ ⎜ ∫ fdS ⎟ dr
⎮⎜ ⎟
Rn ⌡ ⎝ ∂B ( x0 ,r ) ⎠
0
for each point x 0 ∈ R n . or we can say
d ⎛ ⎞
⎜ ∫ f dx ⎟ = ∫ f dS
dr ⎜ B ( x0 ,r ) ⎟
⎝ ⎠ ∂B ( x 0 ,r )
for each r > 0.
(f) To construct smooth approximations to given

functions
Def If U ⊂ Rn is open, given ∈> 0 . We define
U ∈ : = {x ∈ U dist ( x, ∂U ) >∈ }
Def. Standard Mollifier
Let η ∈C ∞ ( R n ) such that
⎧ ⎛ 1 ⎫
⎪⎪c exp ⎜ 2 if x < 1⎪⎪
η (x ):= ⎨ ⎜ x −1
⎝ ⎬
⎪ ⎪
⎩⎪0 if x ≥ 1⎭⎪
the constant c is chosen so that ∫ η dx = 1

Rn
Def. We define
1 ⎛x⎞
η∈ ( x ) : = η⎜ ⎟
∈n ⎝ ∈ ⎠
for every ∈> 0 .
Properties. (i) The functions η∈ are C ∞ since η ( x ) are C ∞ .
14
1 ⌠ ⎛x⎞
(ii) ∫η ∈ dx = ⎮ η ⎜ ⎟ dx
∈n ⌡ ⎝ ∈ ⎠
Rn n
R
= ∫ η ( x ) dx
n
(by definition of n-tuple integral)
R
= 1.
(g) Mollification of a function
If f : U → R is locally integrable
We define the mollification of f
f ∈ : = η∈ * f in U ∈
= ∫ η∈ ( x − y ) f ( y ) dy
U
= ∫ η∈ ( y ) f ( x − y ) dy (by definition)
B ( 0,∈)
Properties. (i) f ∈ ∈C ∞ (U ∈ )
(ii) f ∈ → f almost everywhere. ( a.e. ) as ∈→ 0
(iii) If f ∈ C (U ) then f ∈ → f uniformly on compact subset of U.
almost everywhere.
1.5 Function Analysis Concepts

(i) LP space. Assume U to be a open subset of R n and 1 ≤ p ≤ ∞ . If
f : U → R is measurable, we define
⎧⎛ p ⎞
1/ p
f LP U : = ⎨⎝ U∫
⎪⎪⎜ f dx ⎟ if 1 ≤ p<∞
( ) ⎠
⎪
⎪⎩ess supU f if p = ∞
Transformation from Ball B ( x , r ) to unit Ball B ( 0, 1)
15
Let B ( x , r ) be a ball with centre x and radius r and B (0, 1) be an
arbitrary point of B ( x , r ) and z be an arbitrary point of B (0, 1) then
relation between y and z is

y = x + rz .
1.6 Definition and classification of Partial

Differential Equation
Many physical, geometric and probabilistic problems can be modelled by
partial differential equations. In this section, we define the partial
differential equation, system of partial differential equations, their
classifications and the classical and weak solutions etc
Defination: Partial Differential Equation
A partial differential equation is an equation involving an unknown
function of two or more variables and its partial derivatives i.e. Let U be
an open subset of Rn4. An expression of the type
F ( D k u ( x ) , D k −1 u ( x ) , ..., Du ( x ) , u ( x ) , x ) = 0 ( x ∈U ) (1)
is called a kth-order partial differential equation, where

k k −1 x
F : Rn × Rn × ...R n × R × U → R is given and
u : U → R is the unknown function.
Dku (x ) ∈ R ,
nk
Note that
D k −1u ( x ) ∈ R n k −1 ,
Du ( x ) ∈ R n ,
u (x ) ∈ R
Exp. Let
θ = θ ( x , y, z ) where ( x , y, z ) ∈ R 3
16
then
⎛ ∂θ ∂θ ∂θ ∂ 2θ ∂ 2θ ⎞
f ⎜ , , , , ⎟=0
⎝ ∂x ∂y ∂x ∂y ∂x ∂y ⎠
2 2
defines a 2nd order Partial Differential Equation over R3, θ is the unknown
function and f is prescribed

Classification of Partial Differential Equations
Partial Differential Equation can be classified into four categories

(a) Linear (b) Semi-linear (c) Quasi-linear (d) Non-linear.
(a) Linear Partial Differential Equation: A Partial Differential Equation of
kth order is called linear if it has the form
∑
α
a ( x ) Dαu = f ( x )
≤k
0 (2)
for given functions a 0 ( α ≤ k ) , f .
i.e. coefficient of derivatives are only functions of x.
Exp. (i) ut + b.Du = 0
where b ∈ R n is a constant.
(ii) Δu = 0
(b) Semi-linear Partial Differential Equation: A Partial Differential Equation

is called semi-linear if it is of the form
∑ aα ( x ) D α u + a ( D 0
k −1
u,..., Du, u, x ) = 0 (3)
α=k
i.e. coefficient of highest order derivative is a function of x only.
Exp. φ ( x ) Δu − ux uy = 0
(c) Quasi-linear Partial Differential Equation: A Partial Differential

Equation is called quasi-linear if it is of the form
∑ (a D
α=k
a
k −1
)
u,..., Du, u, x D α u + a 0 ( D k −1 u,..., Du, u, x ) = 0 (4)
17
i.e. coefficient of highest order derivative are lower order derivative and
function of x but not same order derivatives.
Exp. u xx u x + uyy uy + u = 0
is quasi linear partial differential equation.

(d) Non-linear Partial Differential Equation: A Partial Differential Equation
is non-linear in the highest order derivatives.
Exp. u xx uyy + b Du = 0 , b ∈ R n
is non-linear.
1.7 Definition and classification of system of

Partial Differential Equation
Definition: System of Partial Differential Equations
An expression of the form
F ( D k u ( x ) , D k −1 u ( x ) ,..., D u ( x ) , x ) = 0, x ∈U (5)
is called a kth order system of partial differential equations in u where

k k −1 n
F : R mn × R mn × ...R m × R m × U → R m
is given.and
u = ( u1, u 2 ,..., u m ) be the unknown function s.t u :U → R m .
Exp. Navier ‘s equations of equilibrium in linear elasticity.
μ Δ u + ( λ + μ ) D div u = 0
form a system of partial differential equation in u = ( u1, u 2 , u 3 ) .
Classification of System of Partial Differential Equations
Note: System of partial differential equations are classified in the same

way as partial differential equations are classified.
Examples of Linear Partial Differential Equations
18
There are some well-known linear equations
(i) Laplace’s equation
Δu = 0 or ∑ui
x ,x =0
(ii) Linear transport equation
ut + b . Du = 0 , b ∈ R n
(
Du = u x1 ,..., u xn )
(iii) Heat (Diffusion) equation
ut − Δu = 0
(iv) Wave equation
utt − Δu = 0
These will be studied in detail later on.
1.8 Solutions of Partial Differential Equation

Solution. An expression of u which satisfies the given PDE is called a
solution of the Partial Differential Equation.

Well posed problem. A given problem in Partial Differential Equation is
well posed if
(i) the problem has a solution
(ii) solution is unique
(iii) solution depends continuously on the data given problem.
Classical Solution. If a solution of a given problem satisfies the above
three conditions i.e. the solution of kth order partial differential equation
exists, is unique and is at least k times differentiable, then the solution
is called classical solution. Solution of wave equation, Lalpace equation
etc. are classical solutions.
19
Weak Solution. If a solution of a given problem exists and is unique but
does not satisfy the conditions of differentiability, such solution is called
weak solution.
Exp. The gas conservation equation
ut + F ( u )n = 0
models a shock wave in particular situation. So solutions exists, is

unique, but not continuous. Such solution is known as weak solution.
Remark. There are several physical phenomenon in which the problem
has a unique solution, but does not satisfy the condition of
differentiability. In such cases, we cannot claim that we are not able to
find the solution rather such solutions are called weak solutions.

1. L.C. Evans, “Partial Differential Equations,” American
Mathematical Society, Rhade.
2. Duchateau and D.W. Zachmann, “Partial Differential Equations,”
Schaum Outline Series, McGraw Hill Series.
20
Lesson 2
Solution of Linear Partial Differential Equations
Structure
2.1 Introduction
2.2 Transport Equation
2.2.1 Homogeneous problem
2.2.2 Initial value problem
2.2.3 Non-homogeneous problem
2.3 Laplace equation
2.3.1 Fundamental solution
2.3.2 Mean value formula
2.4 Poisson’s equation
2.5 Properties of harmonic functions
2.5.1 Strong Maximum principle
2.5.2 Regularity property
2.5.3 Estimates of derivatives
2.5.4 Liouville’ s theorem
2.5.5 Analyticity property
2.5.6 Harnack’s Inequality
2.1 Introduction
In this lesson, we shall consider the solution of single
linear equations ,namely, Transport equation,Laplace equation
21
and Poisson Equation. Also,we discuss the properties of harmonic
functions, such as Strong Maximum principle , estimates of derivatives,
Harnack’s Inequality etc.
2.2 Transport equation
2.2.1 Homogeneous problem
The simplest partial differential equation is the transport equation

with constant coefficient, which is
ut + b.Du = 0 in R n × ( 0, ∞ ) (1)
where b is a fixed vector in R n ,i.e.
b = (b1, b2 , ..., bn ) and
u : R n × [0, ∞ ) → R
is the unknown function such that u = u ( x , t ) .
x = ( x1 ......xn ) ∈ R n
denotes a spatial variable and t ≥ 0 is the time variable.To solve (1), we
observe the L.H.S. of equation (1) carefully, we find that it denotes the dot
(
product of u x1 , u x2 , ..., u xn , ut ) with (b1,..., bn , 1) . So, L.H.S. of equation
(1) tells that the derivative of u in the direction of (b, 1) is zero in R n +1
dimensional space. So ,if ( x, t ) ∈ R n × ( 0, ∞ ) , we define the parametric
equation of line in the direction (b, 1)
z ( s ) : = u ( x + sb, t + s ) (2)
where s ∈ R , is the parameter.
Or z ( s ) = u (θ , φ )
where θ = x + sb, φ = t + s
22
Differentiating w.r.t. s
z& ( s ) = uθ b + uφ .1
=0 (using eqn 1)
⇒ z ( s ) = constant for each s, i.e., u is constant on the line through (x, t)
in the direction of (b, 1).
Hence, we conclude that if we know the value of u ( x , t ) at any point on
each such line, we know u ( x , t ) everywhere in R n × [0, ∞ ) .
2.2.2 Initial value problem
Consider the initial value problem
ut + b.Du = 0 in R n × ( 0, ∞ ) , (3a)
u = g on R n × {t = 0} (3b)
where b ∈ R n and g is the prescribed function.
Solution. As above, L.H.S. of eq. (3a) represents the directional
derivative of u in the direction of (b, 1) ∈ R n × ( 0, ∞ ) . Hence, the
parametric equation of line through (x, t) in the direction of (b, 1) is given

by
z ( s ) : = u ( x + sb, t + s )
where s is the parameter.
Also, z& ( s ) = 0 (using 3a)
Therefore, z (s) is constant on this line i.e. u (x, t) is constant on the line
through (x, t) in the direction of (b, 1). This line touches the plane
R n × {t = 0} for s = −t i.e. the point ( x − tb, 0 )
where
u ( x , t ) = g ( x − tb ) ( using 3b)
23
Since u is constant on this line, so
u ( x , t ) = g ( x − tb ) for all x ∈ R n , t ≥ 0 (4)
is required solution of initial value problem.
Note. If the function g (x) is C 1 then Eq. (4) gives the classical solution of
problem.
2.2.3 Non-homogenous problem
Consider the non-homogeneous case of transport equation
ut + b.Du = f (x,t) in R n × ( 0, ∞ ) (5)
with initial condition
u = g on R n × {t = 0} (6)
Solution. Fix a point ( x , t ) ∈ R n +1 , as discussed before, the equation of
line passing through ( x , t ) in the direction of (b, 1) is given by
z ( s ) = u ( x + sb, t + s ) (7)
where s is the parameter.

Differentiating w.r.t. s
z& ( s ) = uθ (θ , φ ) .b + uφ (θ , φ )
where θ = x + sb, φ = t + s
z& ( s ) = f ( x + sb, t + s ) (using 5)
Integrating w.r.t. s from -t to 0

0 0
∫ z& (s ) ds =
−t
∫ f ( x + sb, t + s ) ds
−t
0
z ( s ) −t = ∫ f ( x + sb, t + s ) ds
0
−t
Substitute t + s = ψ , ds = dψ
24
t
z ( 0 ) − z ( −t ) = ∫ f ( x + b (ψ − t ) , ψ ) dψ
0
t
u ( x , t ) − u ( x − bt , 0 ) = ∫ f ( x + b (s − t ) , s ) ds
0
(using 7)
t
⇒ u ( x , t ) = u ( x − bt , 0 ) + ∫ f ( x + b (s − t ) , s ) ds
0
t
= g ( x − bt ) + ∫ f ( x + b ( x − t ) , s ) ds (8)
0
Equation (8) gives the solution for each x ∈ R n and t ≥ 0 . It is the
required solution of initial value problem for non-homogeneous

Transport equation.
2.3 Laplace’s Equation

Let U be an open subset of R n and u : U → R , then the equation
Δu = 0 x ∈U
or
n
∑u
i =1
xi xi =0
defines the Laplace’s equation in u.
Harmonic function
A C2 function satisfying the Laplace’s equation is called Harmonic

function.
Physical occurrence
We get the Laplace’s equation in several physical phenomenon

such as irrotational flow of incompressible fluid, diffusion problem
,conduction problem etc.
25
Consider the equation

n
Δu = ∑ u xi xi = 0 (1)
i =1
where, u : U → R, U ⊂ R n , x ∈ U .
Eq. (1) is a linear partial differential equation. To solve (1), we rotate the
original coordinate system Ox1 … xn to Ox1′x 2′ ...xn′ at angle θ about O (Fig.
1).
Let lij = cos ( x i' , x j )
We have x i ′ = ∑ li j x j
j
Similarly, x i = ∑ l j i x ′j
j
Table of direction cosines

x1 x2 xn
x1′ l11 l12 … l1n
x 2′ l21 l22 … l2n
xn ′ ln1 ln2 lnn
u = u ( x1, x 2 , ..., xn )
∂u ∂u ∂x1′ ∂u ∂x 2′ ∂u ∂xn ′
= + + ... +
∂xi ∂x1′ ∂x i ∂x 2′ ∂x i ∂xn ′ ∂x i
∂u ∂u ∂u
= l1i + l 2i + ... + lni
∂x1′ ∂x 2′ ∂xn′
∂ 2u ⎡ ∂ 2u ∂ 2u ∂ 2u ⎤
= l1i ⎢ ′2 l1 j + ′ l 2 j + ... + ln j ⎥
∂xi ∂x j ⎣ ∂x1 ∂x1 ∂x 2′ ∂x1′ ∂xn′ ⎦
26
⎡ ∂ 2u ∂ 2u ∂ 2u ⎤
+l 2i ⎢ l1 j + l + ... +
2 2j
lnj ⎥
⎣ ∂x1′ ∂x 2′ ∂x 2′ ∂x 2′ ∂xn′ ⎦
⎡ ∂ 2u ∂ 2u ∂ 2u ⎤
+ lni ⎢ l1 j + l 2 j + ... + l
2 nj ⎥
⎣ ∂x1′ ∂xn′ ∂x 2′ ∂xn′ ∂xn′ ⎦
Taking i = j
∂ 2u ∂ 2u ∂ 2u ∂ 2u ∂ 2u
+ ... + = + + ... +
dx12 ∂xn2 ∂x1′2 ∂x 2′2 ∂xn′2
This shows that Laplace’s equation remains invariant w.r.t. the
transformation of coordinate axes. To find a solution of eq. (1) in Rn, we

seek a radial solution, i.e.
u ( x ) : = V (r )
where
r = x = x12 + x 22 + ... + xn2 (2)
∂r 1 x
= ( x12 + ... + xn2 )
−1/2
2x i = i x ≠0
∂xi 2 r
xi x j δ ij V ′ (r ) x i x j
u xi x j = V ′′ (r ) + V ′ (r ) −
r2 r r3
Taking i = j
xi xi δ ii V ′ (r ) x i xi
u xi x j = V ′′ (r ) + V ′ ( r ) −
r2 r r3
Δu = V ′′ (r ) +
(n − 1) V ′
(r )
r
n −1
V ′′ + V′ = 0 (3)
r
V ′′ n −1
=−
V′ r
Integrating w.r.t. r
log V ′ = − (n − 1) log r + log a
27
where log a is a constant, or
a
V′ =
r n −1
Again integrating
⎧a ′ log r + b n =2
⎪
V (r ) = ⎨ a ′
⎪⎩ r n −2 + b n ≥3
where a ′ and b are constants.
Hence if r > 0, the solution of Laplace’s equation (1) is
⎧a ′ log x + b x =2
⎪
u (x ) = ⎨ a′
⎪ x n −2 + b x ≥3
⎩
Without loss of generality we take b = 0. To find a ′ , we normalize the
solution, i.e.
∫ u ( x ) dx = 1
Rn
which yields a ′ .Thus
⎧ 1
⎪− 2π log x n =2
⎪
u (x ) = ⎨ (4)
1
⎪ n ≥3
⎪⎩n (n − 2) α (n ) x
n −2
for each x ∈ R n , x ≠ 0 .
The solution (4) is known as Fundamental solution of Laplace equation.
We denote it by Φ ( x ) .
⎧ 1
⎪− 2π log x n =2
⎪
Φ (x ) = ⎨
1
⎪ n ≥3 x ≠0
⎪⎩n (n − 2) α (n ) x
n −2
28
Remark. The solution (4) is not valid for x = 0, since u (x) is singular for
x = 0.
2.3.2 Mean value theorem
Theorem. If u ∈ C 2 (U ) is harmonic then
u (x ) = ∫ u ds = ∫ u dy (1)
∂B ( x ,r ) B ( x ,r )
for each ball B ( x , r ) ⊂ U .
Proof.
φ (r ) : = ∫ u ( y ) ds ( y ) (2)
∂B ( x ,r )
Shifting the integral to unit ball, if z is an arbitrary point of unit ball then
φ (r ) : = ∫ u ( x + rz ) ds ( z )
∂B ( x ,r )
Differentiating w.r.t. r
φ ′ (r ) = ∫ Du ( x + rz ) .z ds ( z )
∂B (0,1)
y−x
= ∫ Du ( y ) . ds ( y )
∂B ( x ,r )
r
= ∫ Du ( y ) .v ds ( y )
∂B ( x ,r )
where v is unit outward normal to ∂B ( x , r ) .
∂u
φ ′ (r ) = ∫ ds(y )
∂B ( x ,r )
∂ν
=
1 ⌠ ∂u ds ( y )
⎮
nα ( x ) r n −1 ⌡ ∂v
∂B ( x ,r )
r
=
n ∫
B ( x ,r )
Δu dy (by Green’s formula)
29
=0 since u is harmonic.
Hence φ (r ) is independent of r.
So φ (r ) = limit ∫ u(y ) ds(y )

t →0 ∂B ( x ,t )
= u (x ) (3)
From (2) and (3)
u (x ) = ∫ u(y ) ds(y ) (4)

∂B ( x ,r )
For the second equality,

r
⌠⎛ ⎞
∫ u dy =
⎮ ⎜ ∂B (∫x ,t )
⎮ ⎜ u ds ⎟ dt
⎟
(using coarea formula)
B ( x ,r ) ⌡⎝ ⎠
0
r
= ∫ u ( x ) n α (n ) t n −1 dt (using 4)
0
= u ( x ) α (n ) r n
1
⇒ u (x ) = ∫ u dy
α (n ) r n B ( x ,r )
= ∫
B ( x ,r )
u dy (5)
Combining (4) and (5), we get the result.

Note. The above formula is known as mean value theorem. Converse of
this result is also true.
Converse of Mean Value Theorem
If u ∈ C 2 (U ) satisfies the relation
u (x ) = ∫ u ds
∂B ( x ,r )
for each ball B ( x , r ) ⊂ U . Then, u is harmonic in U.
30
Proof. Suppose that u is not harmonic, so
Δu ≠ 0 .
Hence there exists a ball B ( x , r ) ⊂ U such that Δu > 0 within B ( x , r )
Preceeding ,as above,

r
φ ′ (r ) = ∫ Δu dy > 0
n B ( x ,r )
Also φ (r ) = u ( x )
φ ′ (r ) = 0
which is contrary to φ ′ (r ) > 0 . So u is harmonic.
2.4 Poisson’s equation

We observe that the solution of Laplace’s equation Δu = 0 , Φ ( x ) is
harmonic for x ≠ 0 . Shifting the origin to a new point y, the Laplace
equation remains unchanged. So Φ ( x − y ) is harmonic for x ≠ y .
If f : R n → R
is harmonic;then Φ ( x − y ) f ( y ) is harmonic for each y ∈ R n and x ≠ y .
If we take the sum of all different points y over R n ,then,
∫ Φ ( x − y ) f (y ) dy
Rn
is harmonic.No, since Δu ( x ) = ∫ Δ Φ ( x − y ) f (y ) dy
x
Rn
is not valid near the singularity x = y. We must proceed more carefully.
Theorem. Suppose f ∈ Cc2 ( R n ) i.e. f is twice differentiable with compact
support. Let
u (x ) = ∫ Φ ( x − y ) f (y ) dy
Rn
31
⎧ 1
⎪− 2π ∫n log x − y f ( y ) dy n =2
⎪ R
= ⎨ ⌠ f (y ) (1)
1
⎪ ⎮ dy n ≥3
⎪ n (n − 2 ) α (n ) ⌡ x − y
n −2
⎩ Rn
is a solution of Poisson’s equation

Δu = − f in Rn (2)
Proof. To show that, u (x) represented by eq. (1), is a solution of eq. (2),
we need to prove
(i) u ∈ C 2 (R n )
(ii) −Δu = f in R n
(i) We have
u (x ) = ∫ Φ ( x − y ) f (y ) dy
Rn
By change of variable x – y = z
= u (x ) = ∫ Φ ( z ) f ( x − z ) dz
Rn
= ∫ Φ (y ) f ( x − y ) dy (3)
Rn
Let us calculate u xi . By definition
u ( x + hei ) − u ( x ) ⌠ ⎡ f ( x + hei − y ) − f ( x − y ) ⎤⎦
= ⎮ Φ (y ) ⎣ dy
h ⌡n h
R
where h ≠ 0 and ei = ( 0, 0, ..., 1, ..., 0 ) ( in ith place)
Taking the limit h → 0 , since f ∈ Cc2 ( R n ) so
f ( x + hei − y ) − f ( x − y ) ∂f
lt → (x − y )
h →0 h ∂xi
Hence
32
∂u ( x ) ⌠ ∂f
= ⎮ Φ (y ) ( x − y ) dy (i = 1,2,..., n )
∂x i ⌡ ∂x i
R′
<∞
Similarly
∂ 2u ( x )⌠ ∂2 f ( x − y )
= ⎮ Φ (y ) dy (i , j = 1,2,...,n )
∂xi ∂x j ⌡ ∂x i ∂x j
n
R
is continuous.So u ∈ C 2 ( R n ) .
(ii) By part (i)
Δu ( x ) = ∫ Φ (y ) Δ f ( x − y ) dy
x
Rn
Since Φ ( y ) is singular at y = 0 ,so we include it in small ball B ( 0, ∈) ,
where ∈> 0 .
Hence
Δu ( x ) = ∫ Φ ( y ) Δ x f ( x − y ) dy + ∫ Φ ( y ) Δ x f ( x − y ) dy
B ( 0, ∈) R n − B ( 0,∈)
= I∈ + J ∈ (4)
where
I∈ : = ∫ Φ ( y ) Δ x f ( x − y ) dy (5)
B ( 0, ∈)
J∈ : = ∫ Φ ( y ) Δ x f ( x − y ) dy (6)
R n −B ( 0,∈)
I∈ = ∫ Φ ( y ) Δ x f ( x − y ) dy
B ( 0, ∈)
≤ Φ (y ) Δ x f ( x − y ) ∫ dy
B ( 0, ∈)
33
⎧C log ∈.∈2 n =2
⎪
I∈ ≤⎨ 1 (7)
⎪C n −2 .∈ n ≥3
n
⎩ ∈
since f ∈ CC2 ( R n ) , so f is bounded.
Now
J∈ = ∫ Φ ( y ) Δ x f ( x − y ) dy
R n − B ( 0, ∈)
⎛ ∂ ∂ ⎞
= ∫ φ ( y ) Δy f ( x − y ) dy ⎜Q =− , Δ x = Δy ⎟
R n − B ( 0, ∈) ⎝ ∂x ∂y ⎠
Integrating by parts
⌠ ∂f
J∈ = − ⎮ D Φ ( y ) . Dy f ( x − y ) dy + ⌠
⎮ Φ (y ) ( x − y ) ds (y )
⎮ ⌡ ∂v
⌡ ∂B ( 0,∈)
R n − B ( 0,∈)
=: K ∈ + L∈ (8)
L∈ = ⌠ Φ ( y ) ∂f ( x − y ) ds ( y )
⎮
⌡ ∂v
∂B ( 0,∈)
≤ Df Φ (y ) ∫ ds ( y )
( )
L∞ R n
∂B ( 0,∈)
⎧C ∈.log ∈ for n = 2
⎪
≤⎨ 1 n −1
(9)
⎪⎩C ∈n −2 .∈ for n ≥ 3
K∈ = − ∫ D Φ ( y ) Dy f ( x − y ) dy
R − B ( 0,∈)
n
⌠ ∂Φ
K∈ = ⎮ ΔΦ ( y ) f ( x − y ) dy − ⌠
⎮ f (x − y ) ds ( y )
⎮ ⌡ ∂v
⌡ ∂B ( 0,∈)
R n − B ( 0,∈)
34
⌠ ∂Φ
= − ⎮ f (x − y ) ds (since ΔΦ = 0 ) (10)
⌡ ∂v
∂B ( 0,∈)
∂φ
But = D Φ .v
∂ν
1 y −y
= − . (since normal is in opposite direction)
n α (n ) y n
y
1 ∈2
= on ∂B ( 0, ∈)
n α (n ) ∈n +1
1
=
n α (n ) ∈n −1
Substituting in equation (10)

1
K∈ = − ∫ f ( x − y ) ds ( y )
n α ( x ) ∈n −1 ∂β ( 0,∈)
= − ∫
∂B (0,∈)
f (x − y ) ds(y )
Shifting the centre of B ( 0,∈) to B ( x ,∈)
K∈ = − ∫
∂B (0,∈)
f (y ) ds(y )
= − f (x ) as ∈→ 0 (11)
Using eq. (4), (7), (8), (9) and (11), and taking the limit as ∈→ 0
Δu ( x ) = − f ( x )
Hence the result.
2.5 Properties of Harmonic Functions
35
2.5.1 Strong Maximum principle
Suppose U ⊂ R n is open and bounded. Let u ∈ C 2 (U ) ∩ C (U ) is
harmonic within U. Also U is connected and there exists a point x 0 ∈ U
such that
u ( x 0 ) = max u
U
then u is constant within U.
Proof. Let u ( x 0 ) = max = M

U
Take 0 < r < dist ( x 0 , ∂U ) ,
since u is harmonic,, for the ball B ( x 0 , r )
M = u ( x0 ) = ∫ u dy (By Mean Value theorem)

B ( x 0 ,r )
≤M.
Equality holds only if u = M within B ( x 0 , r ) .
Hence u ( y ) = M for all y ∈ B ( x 0 , r ) .
To show that this result holds for the set U, consider the set
{
X = x ∈U u (x ) = M }
We prove that X is both open and closed.
X is closed since if x is the limit point of set X, then ∃ a sequence {x n }

in X such that {xn } → x
Since u is continuous so {u ( xn )} → u ( x )
So u (x ) = M
⇒ x∈X
⇒ X is closed.
36
To show that X is open, i.e. X is neighbourhood of each of its points .Let
x ∈ X , there exists a ball B ( x , r ) ⊂ U such that
u (x ) = ∫ u dy
B ( x ,r )
So x ∈B ( x ,r ) ⊂ X
Hence X is open.
But U is connected. The only set which is both open and closed in U is
U. So U = X. Hence u ( x ) = M ∀ x ∈ U .So u is constant in U.
Maximum Principle
Suppose U ⊂ R n is open and bounded. Let u ∈ C 2 ( u ) ∩ C (U ) is
harmonic within U then

max u max u
=
U ∂U
Proof. Suppose there exists a point x 0 ∈ U such that u ( x 0 ) = max u = M

∂U
So u ( y ) ≤ u ( x 0 ) for some y and suppose x 0 ∉ ∂U since U is harmonic, so
by mean value theorem, there exists a ball B ( x 0 , r ) ⊂ U such that
u ( x0 ) = ∫ u dS ( y )
∂B ( x 0 , r )
1
M ≤ u (y ) ∫ dS ( y )
nα (n ) r n −1 ∂B ( x 0 ,r )
≤ u (y )
Maximum value is less than u (y ) which is a contradiction. Hence
x 0 ∈ ∂U . Hence the result.
Cor. If U is connected and u ∈ C 2 (U ) ∩ C (U ) satisfies
37
Δu = 0 in U
u = g on ∂U
where g ≥ 0 then u is positive everywhere in U.
Uniqueness of solution
There exists a unique solution of C 2 (U ) ∩ C (U ) of the boundary
value problem
−Δu = f in U
u = g on ∂U (1)
where g ∈ C ( ∂U ) , f ∈ C (U )
Proof. Let u and u be two solutions of problem (1) then

−Δu = f in U
u = g on ∂U
−Δu = f in U
u = g on ∂U
Let w : = ± (u − u )
Δw = 0 in U
w = 0 on ∂U
i.e. w is harmonic in U and w attains maximum value on boundary

which is zero. If U is connected then w is constant. So w = 0 in U
So u = u in U.
2.5.2 Regularity property
Regularity property states that if u ∈ C 2(U ) is harmonic then
u ∈ C ∞ (U ) i.e. Harmonic functions are regular function or smooth
functions.
38
Theorem. If u ∈ C (U ) satisfies the mean value property
u (x ) = ∫ u dy = ∫ u dS (1)
B ( x ,r ) ∂B ( x ,r )
for every ball B ( x , r ) ⊂ U then u ∈ C ∞ (U ) .
Proof. Consider the set
U ∈ = {x ∈ U |dist ( x , ∂u ) >∈} and η be the standard mollifier.
u∈ : ≡ η∈ ∗ u in U ∈ (2)
We first show that u ∈ ∈ C ∞ (U ∈ )
Fix x ∈ U ∈ where x = ( x1, x 2 ,..., xn ) .
Let h be very small such that x + hei ∈ U ∈
u ∈ ( x ) = η∈ * u
1 ⌠ ⎛x −y ⎞
= ⎮η ⎜ ⎟ u ( y ) dy (3)
∈n ⌡ ⎝ ∈ ⎠
U∈
1 ⌠ ⎛ x − y + hei ⎞
u∈ ( x + hei ) = ⎮η ⎟ u ( y ) dy (4)
∈n ⌡ ⎜⎝ ∈ ⎠
U∈
Subtracting (3) and (4)
⌠ ⎡ ⎛ x − y + hei ⎞ ⎛ x − y ⎞⎤
⎮ ⎢ η⎜ ⎟ −η ⎜ ⎟⎥
u ( x + hei ) − u ( x ) 1
∈ ∈
∈ ⎝ ∈ ⎠ ⎥ u y dy
= n ⎮⎢ ⎝ ⎠
( )
h ∈ ⎮⎢ h ⎥
⎮⎢ ⎥⎦
⌡⎣
U∈
Taking the limit as h → 0
x −y
∂η (
1 ⌠
)
∂u∈ ∈ ⌠ ∂η (x − y )
= n +1 ⎮ u ( y ) dy = ⎮ ∈ u ( y ) dy
∂xi ∈ ⎮ ∂xi ⌡ ∂x i
⌡ U∈
U∈
since η ∈ C ∞ ( R n )
39
∂u∈
so exists.
∂x i
Similarly D α u∈ exists for each multiindex α
So u ∈ ∈ C ∞ (U ∈ )
We now show that u ≡ u∈ on U ∈
Let x ∈ U ∈ then
u∈ ( x ) = ∫ ηε ( x − y ) u ( y ) dy
U
⌠ 1 ⎛ x −y ⎞
= ⎮ n η⎜ ⎟ u ( y ) dy (by definition)
⌡ ∈ ⎝ ∈ ⎠
B ( x ,∈)
1 ⌠ ⎛r ⎞⎛ ⎞
= n ⎮ η ⎜ ⎟ ⎜ ∫ u ( y ) dS ⎟ dr (using the cor. of coarea formula)
∈ ⎮ ⎝ ∈ ⎠ ⎜ ∂B ( x ,r ) ⎟
⌡ ⎝ ⎠
0
∈
1 ⌠ ⎛r ⎞
= n ⎮ η ⎜ ⎟ nα (n ) r n −1u ( x ) dr (by 1)
∈ ⌡ ⎝∈⎠
0
∈
nα (n ) u ( x ) ⌠ ⎛ r ⎞ n −1
= ⎮η ⎜ ⎟r dr
∈n ⌡ ⎝∈⎠
0
u (x ) ⌠ ⎛ y ⎞
= ⎮ η ⎜ ⎟ dy
∈n ⌡ ⎝ ∈ ⎠
B ( 0,∈)
= u (x ) ∫ η∈ ( y ) dy (by definition)
B ( 0,∈)
= u (x)
So u ∈ ≡ u in U ∈ and so u ∈ C ∞ (U ∈ ) for each ∈ > 0 .
Remark. The above property holds for each ∈ > 0. It may happen u may
not be smooth or even continuous upto to ∂U .
40
2.5.3 Estimates of derivatives
Theorem. Assume that u is harmonic in U then

Ck
D α u ( x0 ) ≤ u (1)
r n +k L1⎛⎜ B ( x0 ,r ) ⎞⎟
⎝ ⎠
for each ball B ( x 0 , r ) ⊂ U and each multiindex α of order k where
1
C0 =
α (n )
(2 nk )
n +1 k
Ck = k = 1, 2, … (2)
α (n )
Proof. We prove it by induction .For k = 0 , α = 0 .
1
To show u ( x 0 ) ≤ u
r α (n )
n L1 ( B ( x ,r ) )
By mean value theorem
u ( x0 ) = ∫ u(y ) dy for each ball B ( x 0 , r ) ⊂ U

B ( x 0 ,r )
1
u ( x0 ) = ∫ u ( y ) dy
α (n ) r n B ( x 0 ,r )
1
u ( x0 ) ≤ u (3)
α (n ) r n (
L1 B ( x0 ,r ) )
C0
D 0u ( x 0 ) ≤ u L1 ( B ( x 0 ,r ) )
rn
Hence the result.
k=1
To show
C1
D u ( x0 ) ≤ u L1 B x ,r
r n +1 ( ( 0 ))
2n +1 n
where C1 =
α (n )
41
Consider
∂2 ∂2
Δu xi =
∂x12
u ( )
xi + ..... +
∂xn2
u xi ( )
∂
= ( Δu ) = 0
∂xi
So u xi is harmonic. By Mean Value theorem
u xi ( x 0 ) = ∫ u xi dx
r
B (x0 , )
2
1
=
⎛r ⎞
n ∫ u xi dx
α (n ) ⎜ ⎟ ⎛ r⎞
B ⎜ x0 , ⎟
⎝ 2⎠
⎝2⎠
1
=
⎛r ⎞
n ∫ u vi dS (By Guass -Green Theorem)
α (n ) ⎜ ⎟ ⎛ r⎞
B ⎜ x0 , ⎟
⎝ 2⎠
⎝2⎠
2n
=
r ⎛
∫ r⎞
u v i dS
∂B ⎜ x 0 , ⎟
⎝ 2⎠
2n
≤ u ⎛ ⎛ r ⎞⎞
L∞ ⎜⎜ ∂B ⎜ x0 , ⎟ ⎟⎟
(4)
r ⎝ ⎝ 2⎠⎠
⎛ r⎞ ⎛ r⎞
If x ∈ ∂B ⎜ x 0 , ⎟ then B ⎜ x , ⎟ ⊂ B ( x 0 , r ) ⊂ U
⎝ 2⎠ ⎝ 2⎠
By eq. (3)
2n
u (x ) ≤ u
α (n ) r n
⎛ ⎛ r ⎞⎞
L1 ⎜ B ⎜ x , ⎟ ⎟
⎝ ⎝ 2 ⎠⎠
2n
≤ u L1 B x ,r
α (n ) r n ( ( 0 ))
42
Hence
n
1 ⎛2⎞
u ≤ u (5)
α (n ) ⎜⎝ r ⎟⎠
⎛ ⎛ r ⎞⎞ L1 ( B ( x 0 ,r ) )
L∞ ⎜ ∂B ⎜ x 0 , ⎟ ⎟
⎝ ⎝ 2 ⎠⎠
Combining (4) and (5)
2n +1 n
u xi ( x 0 ) ≤ u
α (n ) r n +1 L1 ( B ( x 0 ,r ) )
or
C1
D α u ( x0 ) ≤ u
r n +1 (
L1 B ( x 0 ,r ) )
Hence result is true for k = 1.
Assume that result is true for each multiindex of order less than or equal
to k – 1 for all balls in U.
Fix B ( x 0 , r ) ⊂ U and α be multiindex with α = k
Dα u = (D β u ) for some i = {1, 2, …, n)

xi
where β = k − 1
⎛ r ⎞
Consider the ball B ⎜ x 0 , ⎟ ,
⎝ k⎠
D α u ( x0 ) = (D β u )
xi
Proceeding as in eq. (4)
kn
≤ D βu ⎛ ⎛ r ⎞⎞ (6)
r L∞ ⎜ ∂B ⎜ x 0 , ⎟ ⎟
⎝ ⎝ k ⎠⎠
⎛ r ⎞
If x ∈ ∂B ⎜ x 0 , ⎟ then
⎝ k⎠
⎛ k −1 ⎞
B ⎜ x, r ⎟ ⊂ B ( x0 ,r ) ⊂ U
⎝ k ⎠
⎛ k −1 ⎞
Also by assumption, in the ball B ⎜ x , r⎟
⎝ k ⎠
43
k −1
⎡⎣2n +1 n (k − 1) ⎤⎦
D u ( x0 )
β
≤ n +k −1
u ⎛ ⎛ k −1 ⎞ ⎞
L1 ⎜ B ⎜ x ,
⎛ k −1 ⎞
r ⎟⎟
α (n ) ⎜ ⎝ ⎝ k ⎠⎠
r⎟
⎝ k ⎠
k −1
⎡⎣2n +1 n (k − 1) ⎤⎦
≤ u L1 B x ,r (7)
n +k −1 ( ( 0 ))
⎛ k −1 ⎞
α (n ) ⎜ r⎟
⎝ k ⎠
Combining eq. (6) and (7)

k −1
kn ⎡⎣2 n (k − 1) ⎤⎦
n +1
D α u ( x0 ) ≤ u L1 B x ,r
r n +k −1 ( ( 0 ))
⎛ k −1 ⎞
α (n ) ⎜ r⎟
⎝ k ⎠
(2 nk ) ⎛ k ⎞n 1
n +1 k
= u L1 B x ,r
α (n ) r n +k ⎜⎝ k − 1 ⎟⎠ 2n +1 ( ( 0 ))
(2 nk )
n +1 k
≤ u
α (n ) r n +k L1 ( B ( x 0 ,r ) )
n
1⎡ k ⎤
Since ⎢ ⎥ <1 for ∀ k ≥ 2
2 ⎢⎣ 2 (k − 1) ⎥⎦
Hence result holds for α = k .
2.5.4 Liouville’s theorem
Suppose u : R n → R is harmonic and bounded. Then u is constant.
Proof. Fix x 0 ∈ R n , r > 0 then by mean value theorem
Du ( x 0 ) = u xi ( x 0 ) = ∫ u xi dx
r
B ( x0 , )
2
2n
=
α (n ) r n ⎛
∫ r⎞
u ν dS (By Guass Green’s theorem)
∂B ⎜ x 0 , ⎟
⎝ 2⎠
44
2n
≤ u ⎛ ⎛ r ⎞⎞
L∞ ⎜ ∂B ⎜ x 0 , ⎟ ⎟
r ⎝ ⎝ 2 ⎠⎠
⎛ r⎞ ⎛ r⎞
If x ∈ ∂B ⎜ x 0 , ⎟ then B ⎜ x , ⎟ ⊂ B ( x 0 , r )
⎝ 2⎠ ⎝ 2⎠
n
1 ⎛2⎞
u (x ) ≤ u
α (n ) ⎜⎝ r ⎟⎠ L1 ( B ( x 0 ,r ) )
n
2n ⎛2⎞ 1
Hence u xi ( x 0 ) ≤ ⎜ ⎟ u L1 B n ,r
r ⎝ r ⎠ α (n ) ( ( 0 ))
2n +1 n
= n +1 u L1 B x ,r
r α (n ) ( ( 0 ))
n 2n +1
≤ u
r ( )
L∞ R n
→ 0 as r → ∞
Hence Du = 0 so u is constant.
Representation formula. For n ≥ 3
Let f ∈ CC2 ( R n ) , then every bounded solution of Poisson’s equation
−Δu = f in R n (1)
has the form
u (x ) = ∫ Φ ( x − y ) f (y ) dy + C ( x ∈ R )
n
Rn
where C is a constant and Φ ( x ) is the solution of Laplace equation.
Proof. For n ≥ 3
1
Φ (x ) = → 0 as x → ∞
n α (n ) (n − 2 ) x
n −2
⇒ Φ ( x ) is bounded.
Let u be a solution of eq. (1) so
u= ∫ Φ ( x − y ) f (y ) dy
Rn
45
is bounded. Since f ∈ C 2 ( R n ) and Φ ( x ) is bounded for n ≥ 3 .
Let u be bounded solution of eq. (1)
Define w = u − u
Δw = 0
and w is bounded. (Difference of two bounded functions)

By Liouville’s theorem
w = constant
or u − u = −C
u = u +C
Hence the result.

Remark. The above results does not hold for n = 2. Since
1
Φ (x ) = − log x is not bounded as x → ∞ .
2π
2.5.5 Analyticity property
Theorem. Let u is harmonic in U ⊂ R n then u is analytic.
Proof. Let x 0 be any point in U. We now show that u can be represented
by a convergent power series in the neighbourhood of x 0 . The Taylor
series of u about x 0
D α u ( x0 ) ( x − x0 )
α
∑
α α!
(1)
converges.
1
Let r : = dist ( x 0 , ∂U )
4
1
Let M : = u L1 B x ,2r (2)
α (n ) r n ( ( 0 ))
<∞
46
For each x ∈ B ( x 0 , r ) , B ( x , r ) ⊂ B ( x 0 ,2r ) ⊂ U
By estimates of derivatives
Ck
D α u ( x0 ) ≤ u L1 B x ,r
r n +k ( ( 0 ))
(2 nk )
n +1 k
where Ck = for each α = k

α (n )
so
(2 nk )
n +1 k
D u ( x0 )
α
≤ u
L∞ ( B ( x 0 ,r ) ) α (n ) r n +k L1 ( B ( x 0 ,r ) )
α
⎛ 2n +1 n ⎞ α
≤M⎜ ⎟ α (using 1) (3)
⎝ r ⎠
By Sterling’s formula
1
k+
k 2 1
lt =
k →0 k ! e k 2π
⇒ k k ≤ C k ! e k where C is a constant
α α
or α ≤C α !e (4)
Also by Multinomial theorem
α!
nk = ∑
α=k α!
α ! ≤ nα α ! (5)
Using (4) and (5) in equation (3)

α
⎛ 2n +1 n ⎞
D u ( x0 )
α α α
L∞ ( B ( x 0 ,r ) )
≤ M ⎜ ⎟ C e n α!
⎝ r ⎠
α
⎛ 2n +1 n 2e ⎞
≤ MC ⎜ ⎟ α!
⎝ r ⎠
We claim that the power series in (1) converges provided
47
r
x − x0 < n +2
2 n 3e
The remainder term after N term is
D α u ( x0 + t ( x − x0 )) ( x − x0 )
α
∞
RN ( x ) = ∑
α =N α!
for some 0 < t < 1 .

N N
⎛ 2n +1 n 2 e ⎞ ⎛ r ⎞
Rn ( x ) ≤ CM ∑ ⎜ ⎟ ⎜ n +2 3 ⎟
α =N ⎝ r ⎠ ⎝2 n e ⎠
N
⎛ 1 ⎞
≤C M ∑⎜ ⎟
α = N ⎝ 2n ⎠
C M
≤ → 0 as N → ∞
2N
Hence series is convergent. So u (x) is analytic in neighbourhood of x 0 .
But x 0 is arbitrary point of U. So u is analytic in U.
2.5.6 Harnack’s Inequality
Harnack’s inequality shows that the values of non-negative harmonic

functions within open connected subset of U, are comparable.
Statement For each connected open set V⊂⊂U, there exists a positive
constant C, depending upon V, such that
sup u ≤ C inf u (1)

V V
for all non-negative harmonic function u in U.

The equation (1) is equivalent to
1
u (y ) ≤ u ( x ) ≤ C u (y ) ∀x , y ∈ V
C
1
Proof. Let r : = dist (V , ∂U )
4
48
Choose x , y ∈ V , x −y ≤ r
Then
1
u (x ) = ∫ u dz = ∫ u dz
α (n ) ( 2r )
n
B ( x ,2r ) B ( x ,2r )
1
≥ ∫ u dz (Q B ( x ,2r ) ⊃ B ( x ,r ) )
α (n ) ( 2r )
n
B ( y ,r )
1
=
2n ∫
B (y ,r )
u dz
1
= u (y )
2n
or
2n u ( x ) ≥ u ( y ) (2)
Interchanging the role of x and y
2n u ( y ) ≥ u ( x ) (3)

1
2n u ( y ) ≥ u ( x ) ≥ u (y ) x, y ∈V
2n
Since V is connected, V is compact, so V can be covered by a chain of
finite number of balls {Bi }i =1 such that Bi ∩ B j ≠ φ for i ≠ j each of radius

N
r
. So
2
N
⎛ 1 ⎞
u ( x ) ≥ ⎜ n ⎟ u (y ) ( Since x , y ∈ V so x , y ∈ ball Bi )
⎝2 ⎠
1
u (x ) ≥ u (y )
C
Similarly, C u ( y ) ≥ u ( x )
or
49
1
u (y ) ≤ u ( x ) ≤ C u (y )
C
for all x , y ∈ V .
Consider the inequality
u ( x ) ≤ C u (y )
Let {u ( x i )}i =1 ∈ U be a sequence in n.b.d. of u (x).

k
So
u ( x1 ) ≤ C u ( y )
u ( xn ) ≤ C u ( y )
l .u.b .{u ( x i )} ≤ u.b. {u ( x i )}
sup {u ( x i )} ≤ C u ( y )
or sup u ( x ) ≤ C u ( y )
x ∈V
⇒ C u ( y ) ≥ sup u ( x )
x ∈V
{
Let u ( y j ) }
k
∈ U be sequence in the n.b.d. u (y)
j =1
1
u ( y1 ) ≥ sup u ( x )
C
1
u ( y2 ) ≥ sup u ( x )
C
1
u ( ym ) ≥ sup u ( x )
C
g.l.b. {u (y )} ≥ l.b. {u (y )}
j j
1
inf
y ∈V
{u (y )} ≥ sup u ( x )
C x ∈V
or
50
C inf {u ( y )} ≥ sup u ( x )
y ∈V x ∈V
Hence the result.

51
Lesson 3
Green’s Function
Structure
3.1 Introduction
3.2 Derivation of Green’s function
3.3 Characteristics of Green’s function
3.4 Energy methods
3.5 Heat equation

3.5.2 Solution of Initial Value Problem
3.5.3 Non-homogeneous heat equation
3.6 Self –Assessment Questions

3.1 Introduction
We now consider an important tool to solve the boundary value
problem
−Δu = f in U
u=g on ∂U
where U ⊂ R n is open and bounded and ∂U is C1, which is; Green’s
function. We obtain the derivation of Green’s function and discuss its

characteristics .Later on , we find the solution of heat equation.
52
3.2 Derivation of Green’s function
Consider the boundary value problem

−Δu = f in U (1)
u=g on ∂U (2)
Solution Let u ∈ C 2 (U ) and fix x ∈ U , choose ∈> 0 such that
B ( x ,∈) ⊂ U .
In the region V∈ = U − B ( x ,∈) , applying Green’s formula to
u ( y ) and Φ ( y − x )
∫ ⎡⎣u (y ) ΔΦ (y − x ) − Φ (y − x ) Δu (y ) dy ⎤⎦
V∈
⌠ ⎡ ∂Φ ∂u ( y ) ⎤
= ⎮ ⎢u ( y ) (y − x ) − Φ (y − x ) ⎥ ds ( y )
⌡ ⎣ ∂ν ∂ν ⎦
∂V ∈
where ν is the outward unit normal to ∂V∈ .Hence
− ∫ Φ ( y − x ) Δu ( y ) dy =
V∈
⌠ ⎡ ∂Φ ∂u ( y ) ⎤
⎮ ⎢u ( y ) (y − x ) − Φ (y − x ) ⎥ ds ( y ) (3)
⌡ ⎣ ∂ν ∂ ν ⎦
∂U +∂B ( x ,∈)
(Q ΔΦ (y − x ) = 0 for x ≠ y)
Now
⌠ ∂u ( y )
⎮ Φ (y − x ) ds ( y ) ≤ Du Φ (y − x ) ∫ ds (y )
∂ν L∞ ( ∂B ( x ,∈) )
⌡
∂B ( x ,∈)
1
≤C n −2
nα (n ) ∈n −1
∈
→0 as ∈→ 0 . (4)
Again
53
⌠ ∂Φ ⌠ ∂Φ
⎮ u (y ) (y − x ) ds (y ) = ⎮ u (y + x ) (y ) ds (y )
⌡ ∂ν ⌡ ∂ν
∂B ( x ,∈) ∂B ( 0,∈)
Using
1 y
D Φ (y ) = − y≠0
n α (n ) y n
−y
ν =
y
⌠ 1
ν = ⎮ u (y + x ) ds ( y )
⌡ nα (n ) ∈n −1
∂B ( 0,∈)
1
u ( y ) ds ( y )
nα (n ) ∈n −1 ∂B (∫x ,∈)
=
= ∫ u ( y ) ds ( y )
∂B ( x ,∈)
→ u (x ) as ∈→ 0 (5)
Using (4) and (5) in equation (3) and making ∈→ 0
− ∫ Φ ( y − x ) Δ ( y ) dy
U
⌠ ⎡ ∂Φ ∂u ⎤
=⎮ ⎢u ( y ) ∂ν ( y − x ) − Φ ( y − x ) ∂ν ⎥ ds ( y ) + u(x )
⌡ ⎣ ⎦
∂U
(y ≠ x )
Hence
⌠ ⎡ ∂u ∂Φ ⎤
u (x ) = ⎮ ⎢Φ ( y − x ) ∂ν − u ( y ) ∂ν ( y − x ) ⎥ ds ( y )
⌡ ⎣ ⎦
∂U
− ∫ Φ ( y − x ) Δu ( y ) dy (6)
U
Equation (6) is valid for any point x ∈ U and any function u ∈ C 2 ( u ) .
Equ. (6) gives the solution of problem defined by equ. (1) and (2) provided
54
∂u
that u ( y ) , are known on the boundary ∂U and the value of Δu in U.
∂ν
∂u
But is unknown to us along the boundary. For it, we define a
∂ν
correction term formula. φ x ( y ) given by the solution of
Δφ x = 0 in U
φ x = Φ ( y − x ) on ∂U (7)
Applying Green’s theorem to φ x ( y )
∫ ⎡⎣u (y ) Δφ − φ x Δu ( y ) ⎤⎦ dy
x
⌠ ⎡ ∂φ x ∂u ⎤
= ⎮ ⎢u ( y ) − φx ⎥ ds
⌡ ⎣ ∂ν ∂ν ⎦
∂U
Thus
⌠ ⎡ ∂φ x ∂u ⎤
− ∫ φ x Δu ( y ) dy = ⎮ ⎢ ( )
u y − φx ⎥ dx (by equ. 7) (8)
U ⌡ ⎣ ∂ν ∂ν ⎦
∂U
Adding equ. (6) and (8)
u ( x ) = − ∫ ⎡⎣Φ ( y − x ) − φ x ( y ) ⎤⎦ Δu ( y ) dy
U
∂ ⎡
−⌠
⎮ ⎣ Φ ( y − x ) − φ x ( y ) ⎤⎦ u ( y ) dy (9)
⌡ ∂ν
∂U
We define the Green’s function G ( x , y )
G ( x , y ) : = Φ (y − x ) − φ x (y ) x,y ∈U , x ≠ y (10)
for the region U.

From equ. (9) and (10)
∂G
u ( x ) = − ∫ G ( x , y ) Δu ( y ) dy − ⌠
⎮ u (y ) ( x , y ) dS (y ) (11)
U
⌡ ∂ν
∂U
55
where
∂G
( x , y ) = DyG ( x , y ) .ν (y )
∂ν
is normal derivative of G w.r.t. y.
∂u
Equ. (11) is independent of .
∂ν
Hence the boundary value problem given by equ. (1) and (2) can be solved
in term of Green’s function and the solution is given by equ. (11).
Equ. (11) is known as Representation Formula for Green’s function.
Remark: (i) By definition of Green’s function G ( x , y ) , for given x,
−ΔG ( x , y ) = δ ( x ) in U
G=0 on ∂U
where δ ( x ) is Dirac Delta function.
(ii) It is not so easy to construct G (x, y) for arbitrary region. We can

construct for simple geometries.
3.3 Characteristics of Green’s function

Theorem. Show that for all x , y ∈ U , x ≠ y G ( x , y ) is symmetric i.e.
G ( x , y ) = G ( y, x )
Proof. Fix x, y ∈ U (x ≠ y ) .
Define v ( z ) : = G ( x , z )
= Φ (z − x ) − φ x (z ) z ∈U , z ≠ x
w ( z ) : = G ( y, z )
= Φ (z − y ) − φy (z ) z ∈U , z ≠ y (1)
So Δv ( z ) = 0 in U
56
Similarly Δw ( z ) = 0 in U ( z ≠ x, y ) . (2)
On ∂U
v (z ) = 0
w (z ) = 0 (3)
Applying Green’s formula on the region
V = U − ⎡⎣B ( x ,∈) ∪ B ( y,∈) ⎤⎦ for sufficiently small ∈ > 0
⌠⎛ ∂v ∂w ⎞
∫ (wΔv − v Δw ) dz = ⎮⌡ ⎜⎝ w ∂ν −v ⎟ ds ( z )
∂ν ⎠
V
∂V
( ν is outward drawn normal.)
⌠ ⎛ ∂v ∂w ⎞
= ⎮ ⎜w −v ⎟ ds ( z ) x, y ∈ U , x ≠ y
⌡ ⎝ ∂ν ∂ν ⎠
∂U + B ( x ,∈) +∂B ( y ,∈)
⌠ ⎛ ∂v ∂w ⎞ ⌠ ⎛ ∂v ∂w ⎞
⇒ ⎮ ⎜w −v ⎟ dz + ⎮ ⎜w −v ⎟ ds ( z ) =0
⌡ ⎝ ∂ν ∂ν ⎠ ⌡ ⎝ ∂ν ∂ν ⎠
∂B ( x ,∈) ∂B ( y ,∈)
or
⌠ ⎛ ∂w ∂v ⎞ ⌠ ⎛ ∂v ∂w ⎞
⎮ ⎜v −w ⎟ ds ( z ) = ⎮ ⎜w −v ⎟ ds ( y ) (4)
⌡ ⎝ ∂ν ∂ν ⎠ ⌡ ⎝ ∂ν ∂ν ⎠
∂B ( y ,∈) ∂B ( x ,∈)
Let us compute
⌠ ⎛ ∂v ∂w ⎞
⎮ ⎜w −v ⎟ ds
⌡ ⎝ ∂ν ∂ν ⎠
∂B ( x ,∈)
As w is smooth near x, so
⌠ ∂w
⎮ ds ( z ) ≤ Dw ∫ ds
⌡ ∂ν ∂B ( x ,∈)
∂B ( x ,∈)
≤ C ∈n −1
→ 0 as ∈→ 0 (5)
57
⌠ ∂v
⎮ w ds ( z )
⌡ ∂ν
∂B ( x ,∈)
⌠ ∂ ⎡
= ⎮ w (z ) Φ ( z − x ) − φ x ( z ) ⎤⎦ ds ( z )
⌡ ∂ν ⎣
∂B ( x ,∈)
⌠ ∂Φ ( z − x ) ⌠ ∂φ x
= ⎮ w (z ) ds ( z ) − ⎮ w (z ) ( z ) ds ( z )
⌡ ∂ν ⌡ ∂ν
∂B ( x ,∈) ∂B ( x ,∈)
⌠ ∂Φ ( z − x )
= ⎮ w (z ) ds ( z ) − ∫ Δφ x w ( z ) dz
⌡ ∂ν B ( x ,∈)
∂B ( x ,∈)
∂Φ ( z − x )
=
⌠
⎮
⌡
w (z )
∂ν
ds ( z ) − 0 (Q φ x
is smooth in U )
∂B ( x ,∈)
(6)
Now
⌠ ∂Φ ( z − x ) ⌠ ∂Φ ( z )
⎮ w (z ) ds ( z ) = ⎮ w (z + x ) ds ( z )
⌡ ∂ν ⌡ ∂ν
∂B ( x ,∈) ∂B ( 0,∈)
1
w ( z ) ds ( z )
nα (n ) ∈n −1 ∂B (∫x ,∈)
=
= ∫ w ( z ) ds ( z )
∂B ( x ,∈)
→ w (x ) as ∈→ 0 (7)
Combining equ. (5), (6) and (7) and taking limit ∈→ 0
⌠ ⎛ ∂v ∂w ⎞
lt ⎮ ⎜w −v ⎟ ds ( z ) → w ( x )
∈→ 0 ⌡ ⎝ ∂ν ∂ν ⎠
∂B ( x1∈1)
⌠ ⎛ ∂w ∂v ⎞
Similarly, lt ⎮ ⎜v −w ⎟ ds ( z ) → v ( y )
∈→ 0 ⌡ ⎝ ∂ν ∂ν ⎠
∂B ( y ,∈)
Hence equation (4) gives
58
w ( x ) = v (y )
⇒ G ( x , y ) = G ( y, x ) .
Hence the result.
Def. Let x ∈ R +n , then the reflection of a point x w.r.t. ∂R +n is the point
x = ( x1, x 2 , ..., xn −1, − xn )

Example. Solve the boundary value problem
Δu = 0 in R +n
u=g on ∂Rn+
with the help of Green’s function.
Sol. Let x , y ∈ R +n , x ≠ y .
By definition, G ( x , y ) = Φ ( y − x ) − φ x ( y )
We choose the corrector term
φ x (y ) = Φ (y − x ) (1)
where x is reflection of x w.r.t. ∂R +n .
Clearly Δφ x = 0 in R +n
1
Now, Φ ( y − x ) = n −2
n ≥3
n (n − 2 ) α (n ) y − x
∂Φ y1 − x1
(y − x ) = −
n α (n ) y − x
n
∂y1
∂ 2Φ
=−
1
+
(y1 − x1 ) 2
n +2
n α (n ) y − x α (n ) y − x
2 n
∂y1
------
∂ 2Φ
=−
1
+
(yn + xn ) 2
n −2
n α (n ) y − x α (n ) y − x
2 n
∂yn
59
Adding
ΔΦ ( y − x ) = 0 .
On ∂R +n y − x = (y − x )
So Φ (y − x ) = Φ (y − x )
Hence both conditions are satisfied.

So, Green’s function
G ( x , y ) = Φ (y − x ) − Φ (y − x )
is well defined.
So using the representation formula
∂G
u (x ) = 0 − ⌠
⎮ g (y ) ( x , y ) ds (y )
⌡ ∂ν
∂R n
+
∂G ∂G
( x , y ) = DG .vˆ = − ( x,y )
∂ν ∂yn
∂G ∂Φ ( y − x ) ∂Φ
= − (y − x )
∂yn ∂yn ∂yn
⎡ yn − xn yn + xn ⎤
= −⎢ − ⎥
⎢⎣ nα (n ) y − x nα (n ) y − x ⎥⎦
n n
=
2xn
n α (n ) x − y
n (on ∂R n
+, y−x = y−x )
2xn ⌠ g (y )
u (x ) = ⎮ ds ( y ) x ∈ R +n
n α (n ) ⌡ x − y n
∂Rn
+
This is the required solution and is known as Poisson’s formula.
Def. Let x ∈ R n − {0}
x
The point x = 2
x
60
is called the dual point of x w.r.t. ∂B ( 0,1) .
Example. Solve the boundary value problem
Δu = 0 in B ( 0,1)
u=g on ∂B ( 0,1) (*)
Sol. Fix any point x ∈ B 0 ( 0,1) and y ≠ x
The Green’s function is given by
G ( x , y ) = Φ (y − x ) − φ x (y ) (1)
We choose φ x ( y ) = Φ ( x ( y − x ) ) . (2)
where x dual of x w.r.t. ∂B ( 0,1)
As we know Φ ( y − x ) is harmonic. So is Φ ( y − x ) for y ≠ x
2 −n
Similarly x Φ ( y − x ) is harmonic for y ≠ x
or Φ ( x ( y − x ) ) is harmonic for y ≠ x
So, Δφ x = 0 in B (0, 1)
On ∂B ( 0,1) :
φ ( x ) = Φ ( x (y − x ) )
But
⎧⎛ 2 2⎫
2 2 2⎪⎜ x1 ⎞ ⎛ xn ⎞ ⎪
x y−x = x ⎨ y1 − 2 ⎟ + ... + ⎜ yn − 2 ⎟ ⎬
⎪⎜⎝ x ⎟⎠ ⎜
⎝ x ⎟⎠ ⎪
⎩ ⎭
⎧
2⎪ 2 1 2x .y ⎫⎪
= x ⎨y + 2 − 2 ⎬
⎪⎩ x x ⎪⎭
= x 2 + 1 − 2x .y (Q y = 1)
2 2
= x + y − 2x y
61
2
= x −y
So φ ( x ) = Φ ( x ( y − x ) ) = Φ ( y − x )
Hence both conditions of φ x ( y ) are satisfied.
So
G ( x , y ) = Φ (y − x ) − Φ ( x (y − x ) ) (3)
is well defined.
Hence solution of problem (*) is given by
∂G
u (x ) = − ⌠
⎮ g (y ) ds ( y ) (4)
⌡ ∂ν
∂B ( 0,1)
Now on ∂B ( 0,1)
∂G ∂G
= .ν ,
∂ν ∂y
ν being the unit normal.

∂G y
=
∂y y
∂G
=∑
∂yi
yi (Q y = 1)
i
2
∂G xi − yi yi x − xi
= +
∂yi nα (n ) x − y n
n α (n ) x − y
n
2
yi x − yi
=
n α (n ) x − y
n
∂G
=−
1− x ( 2
)
∂ν n α (n ) x − y
n
So equation (4) gives
62
2
⌠ 1− x
u (x ) = ⎮ g (y ) ds ( y )
⎮ n α (n ) x − y
n
⌡
∂B ( 0,1)
This is the required solution.
3.4 Energy methods

There exists at most one solution u ∈ C 2 (U ) of
−Δu = f in U
u=g on ∂U (*)
where U is a open, bounded set.

Proof. Let u be another solution of problem (*)
Let w = u −u
then Δw = 0 in U
w=0 on ∂U
Consider
∫ wΔw dx = ⌠⌡ w (wxi )xi dx

U U
= − ∫ wxi wxi dx + ∫ wxi w ν dS ,

U ∂U
(ν being the unit normal).

2
= − ∫ Dw dx + 0
U
2
⇒ Dw =0 in U
⇒ Dw = 0 in U
⇒ w = Constant in U
63
But w = 0 on ∂U
Hence w = 0 in U
⇒ u =u
Hence uniqueness of solution.

Def. We define the energy functional for Poisson’s equation Δu = − f by
the expression
⌠ ⎡1 ⎤
I [w ] = ⎮ ⎢ Dw − wf ⎥ dx
2
⌡ ⎣2 ⎦
U
where w ∈ A and A is the admissible set
{
A = w ∈ C 2 (U ) w = g on ∂U }
Theorem. Let u ∈ C 2 (U ) be a solution of Poisson’s equation. Then
I [u ] = min I [w ] (1)
w∈A
Conversely if u ∈ A satisfies (1) then u is a solution of boundary value
problem.
−Δu = f in U
u=g on ∂U (2)
Proof. Let w ∈ A and u be a solution of Poisson’s equation. So

−Δu = f in U
0= ∫ ( −Δu − f ) (u − w ) dx
U
= − ∫ Δu ( u − w ) dx − ∫ f (u − w ) dx
U U
0 = ∫ Du. D ( u − w ) dx − ∫ (u − w ) D u.ν dS − ∫ f (u − w ) dx
U ∂U U
64
= ∫ ( Du . Du − fu ) dx − 0 − ∫ ( Du . Dw − f w ) dx
U U
⌡ (
⇒ ⌠ Du − f u dx =
2
) U
∫ ( Du . Dw − fw ) dx
U
⌡
2
( ⌠ ⎡1
⌡ ⎣2
2 1
2
2
) ⎤
⇒ ⌠ Du − fu dx ≤ ⎮ ⎢ Du + Dw − fw ⎥ dx
⎦
U
(By Cauchy- Schwatz’s inequality)
⎡1 2 ⎤ ⌠ ⎡1 2 ⎤
i.e. ⎢⎣ 2 Du − fu ⎥⎦ dx ≤ ⎮ ⎢
⌡ ⎣2
Dw − fw ⎥ dx
⎦
I [u ] ≤ I [ w ]
Since u ∈ A , so
I [u ] = min I [w ]
w∈A
Conversely
Suppose I [u ] = min I [w ]
w∈A
For any v ∈ Cc∞ (U )
define i (τ ) = I [u + τ v ] τ ∈R
So i (τ ) attains minimum for τ = 0
i ′ (τ ) = 0 for τ = 0
⌠ ⎡1 ⎤
i (τ ) = ⎮ ⎢ Du + τ Dv − ( u + τ v ) f ⎥ dx
2
⌡ ⎣2 ⎦
U
⌠ ⎡1
( ) ⎤
= ⎮ ⎢ Du + τ 2 Dv + τ Du Dv − ( u + τ v ) f ⎥ dx
⌡ ⎣2
2 2
⎦
U
i ′ (0) = ∫ [Du . Dv − vf ] dx
U
65
0 = − ∫ v Δu dx + ∫ D u .ν dS − ∫ vf dx
U ∂U U
0= ∫ [ −Δu − f ]v dx + 0 (Q v ∈ C
∞
C (U ) )
U
This is true for each function v ∈ Cc∞ (U )
So Δu = − f in U
So u is a solution of Poisson’s equation.
3.5 Heat Equation

The linear partial differential equation
ut − Δu = 0
where
x ∈U ⊂ Rn
u : U × [ −, ∞ ) → R
is known as homogeneous Heat equation or Diffusion equation.

The equation
ut − Δu = f ( x , t )
where f : U × [0, ∞ ) → R
is known as non-homogeneous heat equation.

Consider the homogeneous heat equation
ut − Δu = 0 (1)
We seek a solution of equation (1) of the form
1 ⎛x ⎞
u ( x,t ) = v⎜ ⎟
tα ⎝ t β ⎠
where α , β are to be determined.
66
1
u ( x,t ) = v (y ) (2)
tα
where y = x / t β (3)
Differentiating w.r.t. t, x
α β y Dv
ut = − α +1
v (y ) −
t t α +1
1
Δu = α + 2β
Δv
t
Using in equation (1) and simplifying
1
α v ( y ) + β y Dv + 2 β −1
Δv = 0 (4)
t
To make equation (4) independent of t, we put β = 1/2, so that
α v ( y ) + β y . Dv + Δv = 0 (5)
We seek a radial solution of equation (5) as
v ( y ) : = w (r ) where r = y (6)
From equations (5) and (6)
⎛ n −1⎞
Δv ( y ) = w′′ + ω ′ (r ) ⎜ ⎟
⎝ r ⎠
Using in equation (6),
⎛ r n −1⎞
w′′ + ⎜ + ⎟ w′ + α w = 0
⎝2 r ⎠
To make it exact differential, we put α = n /2 and multiply by r n −1 .This
gives
r nw ( )′ = 0
(r n −1 ′
w′ +
2
)
Integrating
r nw
r n −1w′ + = c,
2
67
where c is a constant.
Assuming r → 0, w, w′ → 0, c → 0
wr
Hence w′ + =0
2
w′ r
⇒ =−
w 2
Integrating
2 /4
w = b e −r
where b is a constant.
2
So v ( y ) = be
− y /4
b − x 2 /4t
Hence u ( x , t ) = n /2
e
t
is solution of equation (1).

To find b, we normalize the solution.
∫n u ( x , t ) dx = 1
R
b − x 2 /4t
t n /2
∫4 e dx = 1
R
b
(2 )
n
n /2
πt =1
t
1
or b=
( 4π )n /2
Hence fundamental solution is
⎧ 1 2
− x /4t
⎪ e t >0 x ∈ Rn
u ( x , t ) = ⎨ ( 4π t )
n /2
⎪
⎩0 t <0
Note: The fundamental solution of equation (1) is singular at (0, 0).
3.5.2 Solution of Initial Value Problem
68
Assume that g ∈ C R n ∩ L∞ R n ( ) ( )
and define
2
− x −y
1 ⌠
u ( x,t ) = ⎮e
4t g ( y ) dy (1)
( 4π t )n /2 ⌡
Rn
= ∫n Φ ( x − y ) g (y ) dy
R
where Φ ( x ) is fundamental solution of heat equation.
Then
(i) (
u ∈ C ∞ R n × ( 0, ∞ ) )
(ii) ut ( x , t ) − Δu ( x , t ) = 0 x ∈ Rn , t > 0
(iii) lt u ( x,t ) = g x 0 ( ) for each x 0 ∈ R n

( x ,t ) → ( x 0 ,0)
x ∈ Rn , t > 0
Proof.
2
−x
1
(i) Since the function e 4t is infinitely differentiable with uniform
t n /2
bounded derivatives of all order on R n × [δ , ∞ ) for δ > 0
So u ∈ C ∞ R n × ( 0, ∞ ) ( )
(ii) ut = ∫n Φt ( x − y, t ) g (y ) dy
R
Δu = ∫n ΔΦ ( x − y, t ) g (y ) dy
R
Therefore, ut − Δu = 0 since Φ ( x − y ) is a solution of heat
equation.
69
(iii) Fix x 0 ∈ R n . Since g is continuous ,given ∈> 0 , there exists a δ > 0
such that
( )
g ( y ) − g x 0 <∈ whenever y − x 0 < δ (2)
y ∈Rn
δ
Then if x − x 0 <
2
⌡ ( )
u ( x , t ) − g x 0 = ⌠ Φ ( x − y, t ) ⎡g ( y ) − g x 0 ⎤ dy
⎣ ⎦ ( )
Rn
⎛ ⎞
⎜Q Φ ( x − y, t ) dx = 1⎟
⎜ ∫n ⎟
⎝ R ⎠
≤ ⌠
⌡ ⎣ ( )
Φ ( x − y, t ) ⎡g ( y ) − g x 0 ⎤ dy +
⎦
B ( x0 ,δ )
⌠
⌡ ⎣ ( )
Φ ( x − y, t ) ⎡g ( y ) − g x 0 ⎤ dy
⎦
Rn − B ( x 0 ,δ )
≤ ⌠
⌡ ( )
Φ ( x − y, t ) g ( y ) − g x 0 dy
B ( x 0 ,δ )
+ ⌠
⌡ ( )
Φ ( x − y, t ) g ( y ) − g x 0 dy
Rn − B ( x 0 ,δ )
u ( x,t ) − g x 0 ≤ I + J ( ) (3)
where I = ⌠
⌡ ( )
Φ ( x − y, t ) g ( y ) − g x 0 dy
(
B x 0 ,δ )
J = ⌠
⌡ ( )
Φ ( x − y, t ) g ( y ) − g x 0 dy (4)
(
B x 0 ,δ )
70
I ≤∈ ∫ Φ ( x − y, t ) dy
(
R n − B x 0 ,δ )
I ≤∈ (5)
δ
Further if x − x0 <
2
and δ ≤ y − x0
then y − x0 ≤ y − x + x − x0
≤ y − x + x − x0
≤ y − x + δ /2
y − x0
≤ y−x +
2
1
or y−x ≥ y − x0
2
Hence
J = ⌠
⌡
Φ ( x − y, t ) g ( y ) − g x 0 dy ( )
(
R n − B x 0 ,δ )
≤2 g ∫ Φ ( x − y, t ) dy
(
L∞ ⎛⎜ R n − B x 0 ,δ ⎞⎟ )
⎝ ⎠
(
R n − B x 0 ,δ )
2
− x −y
c ⌠
≤ n /2 ⎮ e 4t dy
t ⌡
(
R n − B x 0 ,δ )
2
y −x0
c ⌠ −
J ≤ ⎮ e 16t dy
t n /2 ⎮
⌡
(
R n − B x 0 ,δ )
71
∞
⌠ ⎧ −r 2 ⎫
c ⎮ ⎪ 16t n −1 ⎪
≤ n /2 ⎨e r ⎬ dr (By Cor. of coarea formula)
t ⎮⎪ ⎪⎭
⌡⎩
δ
→0 as t → 0 (6)
Using (5) and (6) in (3)
u ( x,t ) − g x 0 ( ) <∈
So u ( x,t ) → g x 0 ( ) as ( x , t ) → ( x 0 ,0 )
Hence the result.
3.5.3 Non-homogeneous heat equation

Consider the non-homogeneous heat equation
ut − Δu = f ( x , t ) on R n × ( 0, ∞ )
u =0 on R n × {t = 0}
then
t 2
− x −y
⌠ 1 ⌠ 4(t − s )
u ( x,t ) = ⎮ n /2 ⎮
e f ( y, s ) dy ds (1)
⌡ ⎡⎣4π (t − s ) ⎤⎦ ⌡
0 n R
x ∈ Rn , t > 0
t
⌠
=⎮
⌡
∫n Φ ( x − y, t − s ) f (y, s ) dy ds (2)
0 R
( )
where f ∈ C12 R n × [0, ∞ ) and has compact support then
(i) (
u ∈ C12 R n × ( 0, ∞ ) )
(ii) ut ( x , t ) − Δu ( x , t ) = f ( x , t )
72
(iii) lt u ( x,t ) = 0 for each point x 0 ∈ R n
(
( x ,t ) → x 0 ,0 )
x ∈ Rn, t > 0
Proof. (i) Since Φ has a singularity at (0, 0) we cannot differentiate under

the integral sign. Substituting the variable x – y = 0 , t – s = φ and again
converting to original variable

t
⌠
u ( x,t ) = ⎮ ∫n Φ (y, s ) f ( x − y, t − s ) dy ds
⌡
0 R
(
Since f ∈ C 2 R n × [0, ∞ ) ) and Φ ( y, s ) is smooth near s = t > 0, we
compute
t
⌠
ut ( x , t ) = ⎮ ∫n Φ (y, s ) f t ( x − y, t − s ) dy ds
⌡
0 R
+ ∫n Φ (y, t ) f ( x − y,0 ) dy
R
(By Leibnitz’s Rule)

t
⌠⌠
∂ 2u ∂2
( x , t ) = ⎮ ⎮ Φ (y, s ) f ( x − y, t − s ) dy ds
∂xi ∂x j ⎮⌡ ∂xi ∂x j
⌡ n
0 R
Thus
ut , D 2u ∈ C 2 R n × ( 0, ∞ ) ( )
(ii) ut ( x , t ) − Δu ( x , t )
t
⌠⌠ ⎛ ∂ ⎞
= ⎮ ⎮ Φ ( y, s ) ⎜ − Δ x ⎟ f ( x − y, t − s ) dy ds
⎮⌡ ⎝ ∂t ⎠
⌡ n
0 R
+ ∫n Φ (y, t ) f ( x − y,0 ) dy
R
73
t
⌠⌠ ⎛ ∂ ⎞
= ⎮ ⎮ Φ ( y, s ) ⎜ − − Δy ⎟ f ( x − y, t − s ) dy ds + K
⎮⌡ ⎝ ∂s ⎠
⌡ n
0 R
where
K = ∫n Φ (y, t ) f ( x − y,0 ) dy (3)

R
∈
⌠⌠ ⎛ ∂ ⎞
ut − Δu ( x , t ) = ⎮ ⎮ Φ ( y, s ) ⎜ − − Δy ⎟ f ( x − y, t − s ) dy ds
⎮⌡ ⎝ ∂s ⎠
⌡ n
0 R
t
⌠⌠ ⎛ ∂ ⎞
+⎮ ⎮ Φ ( y, s ) ⎜ − − Δy ⎟ f ( x − y, t − s ) dy ds
⎮⌡ ⎝ ∂s ⎠
⌡ n
∈R
= I∈ + J ∈ + K (4)
where
∈
⌠⌠ ⎛ ∂ ⎞
I∈ = ⎮ ⎮ Φ (y, s ) ⎜ − − Δy ⎟ f ( x − y, t − s ) dy ds
⎮⌡ ⎝ ∂s ⎠
⌡ n
0 R
t
⌠⌠ ⎛ ∂ ⎞
J ∈ = ⎮ ⎮ Φ ( y,s ) ⎜ − − Δy ⎟ f ( x − y, t − s ) dy ds
⎮⌡ ⎝ ∂s ⎠
⌡ Rn
∈
Now,
∈
⌠⌠ ⎛ ∂ ⎞
I ∈ = ⎮ ⎮ Φ ( y, s ) ⎜ − − Δy ⎟ f ( x − y, t − s ) dy ds
⎮⌡ ⎝ ∂s ⎠
⌡ n
0 R
∈
⎛ ⎞⌠
≤ ⎜ ft D2 f ⎟ ⎮ ∫ Φ ( y, s ) dy ds
⎜
⎝
L∞ R n( )+ ( )
L∞ R n ⎟⌡ n
⎠0R
74
∈
≤ C ∫ ds
0
≤C ∈ (5)
t
⌠⌠ ⎛ ∂ ⎞
J ∈ = ⎮ ⎮ Φ ( y, s ) ⎜ − − Δy ⎟ f ( x − y, t − s ) dy ds
⎮⌡ ⎝ ∂s ⎠
⌡ n
∈R
t
t
⌠⌠ ⎡ ∂ ⎤
⎛ ⎞
= ⎮ ⎮ ⎢⎜ − Δy ⎟ Φ ( y, s ) ⎥ f ( x − y, t − s ) dy ds − ∫n Φ (y, s ) f ( x − y, t − s ) dy
⎮ ⌡ ⎣⎝ ∂s ⎠ ⎦
⌡ n R ∈
∈R
(surface integrals are zero since f has compact support)
=− ∫n Φ (y, t ) f ( x − y, 0 ) + ∫n Φ (y,∈) f ( x − y, t − ∈) dy
R R
= ∫n Φ (y,∈) f ( x − y, t − ∈) dy − K (6)
R
Using (5) and (6) in (4)
ut − Δu ( x , t ) = ∫n Φ (y,∈) f ( x − y, t − ∈) dy − K + C ∈ +K
R
Taking limit as ∈→ 0
ut − Δu ( x , t ) = lt ∫n Φ (y,∈) f ( x − y, t − ∈) dy
∈→ 0
R
= lt
∈→ 0
∫n Φ ( x − y, ∈) f (y, t − ∈) dy
R
= f ( x,t )
(using the result Lt then ∫n Φ ( x − y, t ) g (y ) dy → g ( x0 ) )

( x ,t )→( x0 ,0)
R
75
t
⌠
(iii) u ( x,t ) = ⎮ ∫n Φ (y, s ) f ( x − y, t − s ) dy ds
⌡
0 R
t
⌠
u
( )≤
L∞ R n
f
( ) R∫n Φ (y, s ) dy ds
L∞ R n ⎮
⌡
0
t
= f ∫ ds
0
= f t
Taking limit as t → 0
lt u ( x , t ) = 0 for each x ∈ R n .
t →0
Def. We define the parabolic cylinder
U T : = U × ( 0,T ]
where U ⊂ R n is open and bounded.
The parabolic boundary of U T
ΓT : = U T − U T
Remark. U T is interpreted as the parabolic interior of U × ( 0,T ] including
the top U × {t = T } . But ΓT includes bottom and vertical sides of U × [0,T )
3.6 Self –Assessment Questions

Q. Find the solution of boundary value problem
Δu = 0 in B 0 ( 0, r )
u=g on ∂B ( 0, r )
g (y )
2
r2 − x ⌠
Ans. u ( x ) = ⎮ ds ( y )
n α (n ) r ⌡ x −y
n
∂B ( 0,r )
76
77
Lesson 4
Solution of Wave Equation
Structure
4.1 Introduction
4.2 Solution of 1-D wave equation
4.3 Kirchoff’s formula
4.5 Solution of wave equation for n ≥3

4.5.1 Solution for odd n
4.5.2 Solution for even n
4.6 Solution of Non-homogeneous wave equation
4.7 Energy Methods
4.8 Self Assessment Questions
4.9 Suggested Readings
4.1 Introduction
In this lesson, we seek the solution of wave equation. The
homogeneous wave equation
utt − Δu = 0
where t > 0 , x ∈ U ⊂ R n is open and
u : U × [0, ∞ ) → R .
The non-homogeneous wave equation.
78
utt − Δu = f (n , t )
where f : U × [0, ∞ ) → R is a prescribed function.

First we find the solution of wave equation in the one dimensional
case. Consider the initial value problem
utt − u xx = 0 in R × ( 0, ∞ ) (1)
u=g ut = h on R × {t = 0} (2)
where g and h are prescribed functions.

Factorizing equation (1)
⎛∂ ∂ ⎞⎛ ∂ ∂ ⎞
⎜ + ⎟⎜ − ⎟u=0 (3)
⎝ ∂t ∂x ⎠ ⎝ ∂t ∂x ⎠
Let
⎛∂ ∂ ⎞
v ( x,t ) : = ⎜ − ⎟ u ( x,t ) (4)
⎝ ∂t ∂x ⎠
From (3) and (4)
⎛∂ ∂ ⎞
⎜ + ⎟ v ( x,t ) = 0 x ∈ R, t > 0
⎝ ∂t ∂x ⎠
vt + v x = 0 (5)
which is a transport equation with constant coefficients whose solution is
v ( x,t ) = a ( x − t ) (6)
where
a ( x ) : = v ( x ,0 ) (7)
Using equation (6) in equation (4)
ut − u x = a ( x − t ) in R × ( 0, ∞ )
= f (x, t).
79
This is a non-homogeneous transport equation whose solution is
t
u ( x,t ) = ∫ f ( x + t − s, s ) ds + b ( x + t )
0
where
b ( x ) = u ( x ,0 )
or
t
u ( x , t ) = ∫ a ( x + t − 2s ) ds + b ( x + t )
0
Changing the variable x + t − 2s = y

x +t
1
u ( x,t ) =
2 ∫ a (y ) dy + b ( x + t )
x −t
Using equation (2)
g (x ) = b (x )
so
x +t
1
u ( x,t ) = g ( x + t ) + a ( y ) dy
2 x∫−t
To find a (x) : We have
ut ( x ,0 ) − u x ( x ,0 ) = v ( x ,0 )
h (x ) − g′ (x ) = a (x ) (using 2)
x +t
1
u ( x,t ) = g ( x + t ) + ⎡h ( y ) − g ′ ( y ) ⎤⎦ dy
2 x∫−t ⎣
or
x +t
1 1
u ( x,t ) = ⎡⎣g ( x + t ) − g ( x − t ) ⎤⎦ + h ( y ) dy
2 x∫−t
x ∈ R, t > 0 (8)
2
This is required solution of wave equation.Equation (8) is known as D

Alembert’s formula.
80
Note. The general solution of 1-D wave equation
(ut + ux ) (ut − ux ) = 0
is the sum of general solution of ut + u x = 0 and ut − u x = 0
i.e. u ( x , t ) = F ( x + t ) + G ( x − t )
To find the solution of wave equation over R n (n ≥ 2) , we first prove a
lemma.
Def. We define
U ( x; r ,t ) = ∫ u ( y, t ) ds ( y )
∂B ( x ,r )
G ( x; r ,t ) = ∫ g ( y ) ds ( y )
∂B ( x ,r )
H ( x; r ,t ) = ∫ h ( y ) ds ( y )
∂B ( x ,r )
Lemma. Fix x ∈ R n , satisfying
utt − Δu = 0 in R n × ( 0, ∞ ) (1)
u = g , ut = h on R n × {t = 0} (2)
then
U ∈ C m ( R + × [0, ∞ ) ) and
n −1
U tt − U rr − Ur = 0 in R + × ( 0, ∞ ) (3)
r
U =G , Ut = H on R + × {t = 0} (4)
Equation (3) is known as Euler Poisson Darboux Equation.

Proof. We know
U ( x ;r , t ) = ∫ u ( y, t ) ds ( y )
∂B ( x ,r )
Shifting to unit Ball B ( 0,1)
81
U ( x ;r , t ) = ∫ u ( x + rz ) ds ( z )
∂B ( 0,1)
Ur = ∫ Du ( x + rz ) .z ds ( z )
∂B ( 0,1)
y−x
= ∫ Du ( y ) . ds ( y )
∂B ( x ,r )
r
= ∫ Du ( y ) .v ds ( y )
∂B ( x ,r )
(where v is unit outward normal).
∂u
= ∫ ds ( y )
∂B ( x ,r )
∂v
1 ⌠ ∂u
= ⎮ ds ( y )
nα (n ) r n −1 ⌡ ∂v
∂B ( x ,r )
1
nα ( x ) r n −1 B ( ∫x ,r )
= Δu dy (By Green ‘s formula)
r
n B ( ∫x ,r )
= Δu dy
Hence
r
U r ( x; r ,t ) =
n B ( ∫x ,r )
Δu dy (5)
Again differentiating w.r.t. r
1∂ ⎡⎢ 1 ⎤
U rr ( x ; r , t ) = ⎥
nα (n ) ∂r ⎢ r n −1 B ( ∫x ,r )
Δu dy
⎥
⎣ ⎦
1−n 1
=
n α (n ) r n ∫ Δu dy +
nα (n ) r n −1 ∂B (∫x ,r )
Δu ds
B ( x ,r )
82
⎛1 ⎞
= ⎜ − 1⎟ ∫ Δu dy + ∫ Δu ds (6)
⎝n ⎠ B ( x ,r ) ∂B ( x ,r )
From equation (5) and (6), we observe that
lt U r ( x ; r , t ) = 0
r →0
⎛1 ⎞
lt U rr = Δu + ⎜ − 1⎟ Δu
r →0 ⎝n ⎠
1
= Δu ( x , t )
n
So
U ∈ C m ( R + × [0, ∞ ) )
By equation (5)
r
U r ( x; r ,t ) =
n B ( ∫x ,r )
Δu dy
r
n B ( ∫x ,r )
= utt dy
1
nα (n ) r n −1 B ( ∫x ,r )
= utt dy
1
r n −1U r =:
nα (n ) B ( ∫x ,r )
utt dy
1d ⎡⎢ ⎤
U rr + (n − 1) r
n −1 n −2
utt dy ⎥
nα (n ) dr ⎢B ( ∫x ,r )
r Ur =
⎥
⎣ ⎦
1
nα (n ) ∂B (∫x ,r )
= utt ds
= r n −1 ∫ utt ds
∂B ( x ,r )
or
83
U rr +
(n − 1) U = U tt (7)
r
r
which is required equation.
Also u = g on R n × {t = 0}
∫ u ( y,0 ) dS ( y ) = ∫ g ( y ) dS ( y )
∂B ( x ,r ) ∂B ( x ,r )
Dividing by nα (n )r n −1
U ( x ,0 ) = G ( x )
Similarly we can show
U t ( x ,0 ) = H ( x ) for R + × {t = 0}
4.3 Kirchoff’s formula
utt − Δu = 0 in R 3 × ( 0, ∞ ) (1)
u ( x ) = g ( x ) , ut = h on R 3 × {t = 0} (2)
Sol. First we prove that
U% tt − U% rr = 0 in R + × ( 0, ∞ ) (3)
U% = G% , U% t = H% on R + × {t = 0}
U% = 0 on {r = 0} × ( 0, ∞ ) (4)
where
U% : = rU
G% : = rG
H% : = rH
We know Euler Poisson Darboux Equation for n=3 is
84
2
U tt − U rr − U r = 0 in R + × ( 0, ∞ ) (5)
r
U =G , U t = H on R + × {t = 0} (6)
U tt : = rU tt
⎧ 2 ⎫
= r ⎨U rr + U r ⎬ (using 5)
⎩ r ⎭
= rU rr + 2rU r
= (rU r + U )r
( )r
= U% r
= U% rr (7)
So U% satisfies the 1-D wave equation.
Also U% (r ,0 ) = rU (r ,0 )
= rG (r )
= G%
Similarly U% t (r ,0 ) = H% (r )
Hence, by D Alembert’s formula, we have 0 ≤ r ≤ t

t +r
1 1
U% ( x , r ; t ) = ⎡⎣G% (t + r ) − G% (t − r ) ⎤⎦ + ∫ H% ( y ) dy (8)
2 2 t −r
Now
u ( x , t ) = lt U ( x , r ; t ) (by def.)
r →0
U% ( x , r ; t )
= lt
r →0 r
⎧⎪ ⎡ G% (t + r ) − G% (t − r ) ⎤ 1 t +r ⎫
% ( y ) dy ⎪⎬
= lt ⎨ ⎢
2r
⎥+ ∫ H
r →0 ⎢
⎪⎩ ⎣ ⎦⎥ 2r t −r ⎪⎭
= G% ′ (t ) + H% (t )
85
∂ ⎡ ⎤
⎢t g ds ⎥ + t ∫ h ds
⎢ ∂B (∫x ,t )
= (9)
∂t ⎥ ∂B ( x ,t )
⎣ ⎦
But
∂ ⎡⎢ ⎤ ∂ ⎡ ⎤
⎥= ⎢ ( ) ( ) ⎥
∂t ⎢ ∂B (∫x ,t ) ⎥ ∂t ⎢ ∂B (∫0,1)
gds g x + tz ds z
⎥
⎣ ⎦ ⎣ ⎦
= ∫ Dg ( x + tz ) . z ds ( z )
∂B ( 0,1)
⎛y − x ⎞
= ∫ Dg ( y ) . ⎜ ⎟ ds ( y )
∂B ( x ,t ) ⎝ t ⎠
So
u ( x,t ) = ∫ gds + ∫ Dg ( y ) ( y − x )ds ( y ) + ∫ t h ( y ) ds ( y )

∂B ( x ,t ) ∂B ( x ,t ) ∂B ( x ,t )
= ∫ ⎡⎣g + t h ( y ) + D g ( y ) . ( y − x ) ⎤⎦ ds ( y ) (10)
∂B ( x ,t )
This is required solution.Equation (10) is known as Kirchoff’s formula.
4.4 Solution of 2-D Wave Equation
Now we find the solution of wave equation by the method of descent.

Consider initial value problem
utt − Δu = 0 in R 2 × ( 0, ∞ ) (1)
u=g , ut = h on R 2 × {t = 0} (2)
Sol. We regard it as a problem for n = 3 in which the third spatial
variable x 3 does not appear.Let us write
u ( x1, x 2 , x 3 , t ) : = u ( x1, x 2 , t ) (3)
So equation (1) and (2) are modified to
86
utt − Δu = 0 in R 3 × ( 0, ∞ ) (4)
u = g , ut = h on R 3 × {t = 0} (5)
where
g ( x1, x 2 , x 3 ) = g ( x1, x 2 )
h ( x1, x 2 , x 3 ) = g ( x1, x 2 )
If x = ( x1, x 2 ) ∈ R 2 then x ∈ R 3
The solution of initial value problem defined in equation (4) and (5) is
given by Kirchoff’s formula i.e.
∂ ⎛ ⎞
u ( x,t ) = ⎜t g ds ⎟ + t ∫ h ds
⎜ ∂B (∫x ,t )
(6)
∂t ⎟ ∂B ( x ,t )
⎝ ⎠
where B ( x , t ) denotes the ball in R 3 with centre x and radius t > 0 and
d s denotes the two-dimensional surface measure on ∂B ( x , t ) .
Now
1
∫ g dS =
4π t 2 ∫ g dS
∂B ( x ,t ) ∂B ( x ,t )
1/2
⌠ ⎡ ⎛ d γ ⎞2 ⎤
2 ⎮ g ( y ) ⎢1 + ⎜
= ⎟ ⎥ dy
4π t 2 ⎮ ⎢⎣ ⎝ dy ⎠ ⎥⎦
⌡
B ( x ,t )
where factor ‘2’ is taken as B ( x , t ) consists of two hemisphere and
γ (y ) = t 2 − y − x is the parametric equation of any y ∈ B ( x , t )

2
1 ⌠ t
∫ g ds = ⎮ g (y ) dy
2π t 2 ⎮ 2 2
∂B ( x ,t ) ⌡ t − y−x
B ( x ,t )
87
⎡ ⎤
t ⎢ 1 ⌠ t ⎥
= ⎢ 2 ⎮ g (y ) dy ⎥
2 ⎢π t ⎮ 2
⎥
⌡ t2 − y − x
⎢⎣ B ( x ,t ) ⎥⎦
t g (y )
= ∫
2 B ( x ,t ) t 2 − y − x 2
dy (7)
Similarly,
t h (y )
∫ h dS = ∫
2 B ( x ,t ) t 2 − y − x 2
dy (8)
∂B ( x ,t )
Using (7) and (8) in equation (6)
⎛ ⎞
1 ∂ 2⎜ ⌠ g ( y ) ⎟
u ( x,t ) = ⎜ t ⎮ dy ⎟
2 ∂t ⎜ ⎮ 2 2 ⎟
⎜ B (⌡x ,t ) t − y − x ⎟
⎝ ⎠
t2 h (y )
+
2 ∫ 2 2
dy (9)
B ( x ,t ) t − y−x
⎛ ⎞ ⎛ ⎞
∂ ⎜t 2 g (y ) ⎟ = ∂ ⎜t g ( x + tz ) ⎟
∂t ⎜⎜ ∫ 2 2
dy
⎟⎟ ∂t ⎜⎜ ∫ 2
dz
⎟⎟
⎝ B ( x ,t ) t − y − x ⎠ ⎝ B ( 0,1) 1 − z ⎠
g ( x + tz ) D.g ( x + tz ) z
= ∫ 2
dz + t ∫ 2
dz
B ( 0,1) 1− z B ( 0,1) 1− z
g ( y ) dy Dg ( y ) . ( y − x ) dy
=t ∫ 2
+t ∫ 2
B ( x ,t ) t2 − y − x B ( x ,t ) t2 − y − x
Hence equation (9) gives
t g ( y ) dy t Dg ( y )( y − x ) dy
u ( x,t ) = ∫ + ∫
2 B ( x ,t ) t 2 − y − x 2 2 B ( x ,t ) t2 − y − x
2
88
t2 h ( y ) dy
2 B (∫x ,t ) t 2 − y − x 2
+
1 t g ( y ) + t 2h ( y ) + Dy ( y ) . ( y − x )
u ( x,t ) =
2 B (∫x ,t )
dy (10)
2 2
t − y−x
where x ∈ R 2 .Eq. (10) is required solution Equation (10) is known as
Poisson’s Formula.
4.5 Solution of wave equation for n ≥3

To find the solution of wave equation for n > 3 we derive some identities.
Suppose φ : R → R be C k +1 for k = 1,2,...

k −1 k
d2 ⎛ 1 d ⎞ ⎛ 1 d ⎞ ⎛ 2k dφ ⎞
I. ⎜ ⎟
dr 2 ⎝ r dr ⎠
(r φ (r )
2k −1
) =⎜ ⎟ ⎜r
⎝ r dr ⎠ ⎝
⎟
dr ⎠
k −1 k −1 d j φ (r )
II.
⎛1 d ⎞
⎜ ⎟
⎝ r dr ⎠
(r φ (r ) =
2k −1
) ∑ j =0
β kj r j +1
dr j
where
β kj ( j = 0,1,2,..., k − 1) are independent of r.
III. β0k = 1.3.5.... ( 2k − 1)
Proof
I. We prove it by induction. For k = 1 .We have to show
d2 ⎛ 1 d ⎞ ⎛ 2 dφ ⎞
dr 2
( r φ ( r ) ) = ⎜ ⎟ ⎜r ⎟
⎝ r dr ⎠ ⎝ dr ⎠
d
= ⎡r φ ′ (r ) + φ (r ) ⎤⎦
dr ⎣
= r φ ′′ (r ) + 2φ ′ (r )
1⎡ 2
= r φ ′′ (r ) + 2r φ ′ (r ) ⎤⎦
r⎣
1 d ⎡ 2
= r φ ′ (r ) ⎤⎦ = R.H.S.
r dr ⎣
89
Suppose result holds for k. So
k −1 k
d2 ⎛ 1 d ⎞ ⎛ 1 d ⎞ ⎛ 2k dφ ⎞
⎜ ⎟
dr 2 ⎝ r dr ⎠
(r 2k −1
φ (r ) ) =⎜ ⎟ ⎜r
⎝ r dr ⎠ ⎝
⎟
dr ⎠
(a)
We have to prove for k +1 i.e.

k k +1
d2 ⎛ 1 d ⎞ ⎛ 2k + 2 dφ ⎞
⎜ ⎟
dr 2 ⎝ r dr ⎠
( ⎛1 d ⎞
r 2k +1φ (r ) = ⎜ )⎟
⎝ r dr ⎠
⎜r
⎝
⎟
dr ⎠
Now
k
d2 ⎛ 1 d ⎞
⎜ ⎟
dr 2 ⎝ r dr ⎠
(r φ (r )
2k +1
)
k −1
d2 ⎛ 1 d ⎞
= ⎜ ⎟
dr 2 ⎝ r dr ⎠
⎡1 d 2k +1
⎢⎣ r dr r { ⎤
φ (r ) ⎥
⎦
}
k −1
d2 ⎛ 1 d ⎞ ⎡1 ⎤
⎢⎣ r ( 2k + 1) r φ (r ) + r φ ′ (r ) ⎦⎥
2k 2k
= ⎜ ⎟
dr 2 ⎝ r dr ⎠
k −1
d2 ⎛ 1 d ⎞
= ⎜ ⎟ ⎣( 2k + 1) r
⎡ 2k −1
φ (r ) + r 2k −1 {φ ′ (r ) r }⎦⎤ ( Using (a))
dr 2 ⎝ r dr ⎠
k k
⎛ 1 d ⎞ ⎡ 2k ′ ⎤ ⎛ 1 d ⎞ ⎡ 2k d
= ( 2k + 1) ⎜ ⎟ ⎣r φ (r ) ⎦ + ⎜ ⎟ ⎢r {r φ ′}⎤⎥
⎝ r dr ⎠ ⎝ r dr ⎠ ⎣ dr ⎦
k k
⎛ 1 d ⎞ ⎡ 2k ′ ⎤ ⎛ 1 d ⎞ ⎡ 2k ′ 2k +1 ′′⎤
= ( 2k + 1) ⎜ ⎟ r φ (r ) ⎦ + ⎜ ⎟ r φ +r φ ⎦
⎝ r dr ⎠ ⎣ ⎝ r dr ⎠ ⎣
k
⎛1 d ⎞ ⎡
⎟ ⎣( 2k + 2) r φ ′ + r φ ′′ (r )⎤⎦
2k 2k +1
=⎜
⎝ r dr ⎠
k
⎛1 d ⎞ 1 ⎡
=⎜ ⎟ ⎣ ( 2k + 2) r 2k +1φ ′ + r 2k + 2φ ′′ (r )⎤⎦
⎝ r dr ⎠ r
k
⎛ 1 d ⎞ 1 d ⎡ 2k + 2 ′⎤
=⎜ ⎟ r φ⎦
⎝ r dr ⎠ r dr ⎣
k +1
⎛1 d ⎞ ⎡ 2k + 2φ ′⎤
=⎜ ⎟ ⎣r ⎦
⎝ r dr ⎠
Hence the result holds for k + 1.
90
So result is true for all k = 1, 2, …
II. Try yourself.
III. Try yourself.
Def. Assume n is odd ,say n = 2 k + 1 , (k ≥ 1) . We define
k −1
⎛1 d ⎞
U% (r , t ) : = ⎜ ⎟
⎝ r dr ⎠
(r U ( x ;r , t )
2k −1
)
k −1
⎛1 d ⎞
G% (r , t ) : = ⎜ ⎟
⎝ r dr ⎠
(r G ( x ;r )
2k −1
)
k −1
⎛1 d ⎞
H% (r , t ) : = ⎜ ⎟
⎝ r dr ⎠
(r 2k −1
H ( x ;r ) )
U% (r ,0 ) = G% (r ) ,U% t (r ,0 ) = H% (r )
Lemma. U% satisfies the 1-D wave equation.
U% tt − U% rr = 0 in R + × ( 0, ∞ ) (1)
U% = G% ; U% t = H% on R + × {t = 0}
U% = 0 on {r = 0} × ( 0, ∞ ) (2)
k −1
⎛ ∂2 ⎞ ⎛1 ∂ ⎞
Proof. U% rr = ⎜⎜ 2 ⎟⎟ ⎜ ⎟
⎝ r ∂r ⎠
(r 2k −1
U )
⎝ ∂r ⎠
k
⎛ 1 ∂ ⎞ ⎛ 2k ∂U ⎞
=⎜ ⎟ ⎜r ⎟ (by identity I)
⎝ r ∂r ⎠ ⎝ ∂r ⎠
k −1
⎛1 ∂ ⎞ ⎡1 ∂ ⎛ 2k ∂U ⎞ ⎤
=⎜ ⎟ ⎢ r ∂r ⎜ r ⎟
⎝ r ∂r ⎠ ⎣ ⎝ ∂r ⎠ ⎥⎦
k −1
⎛1 ∂ ⎞ ⎡ 2k 2k −1 ⎤
=⎜ ⎟ ⎢⎣ r r U r + r 2k −1U rr ⎥
⎝ r ∂r ⎠ ⎦
k −1
⎛1 ∂ ⎞ ⎡r 2k −1U rr + 2k r 2k − 2U r ⎤
=⎜ ⎟ ⎣ ⎦
⎝ r ∂r ⎠
91
⎛1 ∂ ⎞
=⎜
k −1
⎡ 2k −1 ⎧ (n − 1) U ⎫⎤
⎟ ⎢r ⎨U rr + r ⎬⎥
⎝ r ∂r ⎠ ⎣⎢ ⎩ r ⎭⎦⎥
k −1
⎛1 ∂ ⎞ ⎡r 2k −1U tt ⎤
=⎜ ⎟ ⎣ ⎦
⎝ r ∂r ⎠
= U% tt .
Also
k −1
⎛1 ∂ ⎞
U% = ⎜ ⎟
⎝ r ∂r ⎠
(r 2k −1
U )
d jU
= ∑ β kj r j +1 (by identity II)
j dr j
U% ( 0, t ) = 0 .
By definition U% (r ,0 ) = G%
U% (r ,0 ) = H% on R + × {t = 0}
Hence the lemma.
4.5.1 Solution for odd n (n ≥ 3 )
utt − Δu = 0 in R n × ( 0, ∞ ) (1)
u = g , ut = h on R n × {t = 0} (2)
Solution. By lemma, U% satisfies the 1-D wave equation and the initial
condition. Therefore, by D Alembert’s formula, on half-line 0 ≤ r < t

t +r
1 1
U% (r , t ) = ⎡⎣G% (t + r ) − G% (t − r ) ⎤⎦ + ∫ H% ( y ) dy (3)
2 2 t −r
for all r ∈ R , t ≥ 0
k −1
⎛1 ∂ ⎞
U% (r , t ) = ⎜ ⎟
⎝ r ∂r ⎠
(r 2k −1
U ( x;r , t ) )
92
∂U
= β0k r U + β1k r 2 + ...
∂r
U% (r , t )
⇒ = U + O (r )
β0k r
Taking limit as r → 0
U% (r , t )
lt = lt U ( x , r ; t ) = lt ∫ u ( y ) dS ( y )
r →0 β0k r r →0 r →0
∂B ( x ,r )
= u ( x,t )
So
1 ⎡ G% (t + r ) − G% (t − r ) 1 t +r ⎤
u ( x , t ) = k lt ⎢ + ∫ % ( y ) dy ⎥
H (using 3)
β0 r →0 ⎣⎢ 2r 2r t −r ⎦⎥
1
= ⎡G% ′ (t ) + H% (t ) ⎤
β0k ⎣ ⎦
Since n = 2k + 1
β0k = 1.3....2k − 1
= 1.3. ... (n − 2) (Q n = 2k + 1)
= γ n (say)
Hence,
⎡ n −3
⎛ ⎞
1 ⎢∂ ⎛1 ∂ ⎞ 2
u ( x,t ) = ⎜ ⎟ ⎜ t n −2
∫ gds ⎟
γ n ⎢ ∂t ⎝ t ∂t ⎠ ⎜ ( x ,t )
⎟
⎢⎣ ⎝ ∂B ⎠
n −3
⎛ ⎞⎤
⎛1 ∂ ⎞
∫ h dS ⎟⎟⎥⎥
2
+⎜ ⎟ ⎜ t n −2 (4)
⎝ t ∂t ⎠ ⎜ ∂B ( x ,t )
⎝ ⎠ ⎥⎦
is required solution for odd n.

Note. Putting n = 3, we obtain Kirchoff’s formula.
4.5.2 Solution for even n
93
Suppose that n is even i.e. n ≥ 2 . 2m = n + 2 (say),so m ≥ 2 .
We again use the method of Descent.

utt − Δu = 0 in R n × ( 0, ∞ ) (1)
u=g , ut = h on R n × {t = 0} (2)
Sol. Since n is even, n + 1 is odd.

Suppose
u ( x1, x 2 , ..., xn +1, t ) : = u ( x1, x 2 ,..., xn , t ) (3)
is the solution of wave equation in R n +1 × ( 0, ∞ ) i.e.
utt − Δu = 0 in R n +1 × ( 0, ∞ ) (4)
with initial condition.
u =g , ut = h on R n +1 × {t = 0} (5)
where
g ( x1, x 2 ,..., xn +1 ) : = g ( x1, x 2 ,..., xn )
h ( x1, x 2 ,..., xn +1 ) : = h ( x1, x 2 ,..., xn ) .
The solution of equation (4) subject to (5) is
⎧ n −2
⎛ ⎞
n −2
⎛ ⎞⎫
1 ⎪ ∂ ⎛1 ∂ ⎞ 2 ⎛ 1 ∂ ⎞ 2 ⎪
u ( x,t ) = ⎨ ⎜ ⎟ ⎜t n +1
∫ gds ⎟ + ⎜ ⎟ ⎜t n −1
∫ hds ⎟ ⎬ (6)
γ n +1 ⎪ ∂t ⎝ t ∂t ⎠ ⎜ ∂B ( x ,t )
⎟ ⎝ t ∂t ⎠ ⎜ ∂B ( x ,t )
⎟⎪
⎩ ⎝ ⎠ ⎝ ⎠⎭
where B ( x , t ) denotes the ball in R n +1 with centre x and radius t and
ds denotes the n-dimensional surface measure on ∂B ( x , t )
Now
1/2
2 ⌠ g y ⎡1 + D γ y 2 ⎤
∫ gds = ⎮ ( ) ⎣⎢ ( ) ⎦⎥ dy
∂B ( x ,t ) (n + 1) α (n + 1) t n ⌡
B ( x ,t )
94
where the factor ‘2 ‘is due to the fact that the surface area consists of
two hemispheres and ∂B ( x , t ) I ( yn +1 ≥ 0 ) has the equation
γ (y ) = t 2 − y − x , y ∈ B ( x , t ) and ∂B ( x , t ) I ( yn +1 ≤ 0 ) is the graph of -

2
γ (y ) .
⌠ ⎡ ⎤
2 ⎮ g (y ) ⎢ t ⎥ dy
∫ gds =
(n + 1) α (n + 1) t n ⎮ ⎢ 2 2 ⎥
∂B ( x ,t )
⌡ ⎣⎢ t − y − x ⎦⎥
B ( x ,t )
2α (n ) t g ( y ) dy
(n + 1) α (n + 1) B (∫x ,t )
= (7)
2 2
t − y−x
Similarly
2α (n ) t h (y )
∫ hds =
(n + 1) α (n + 1) B (∫x ,t ) 2
dy (8)
∂B ( x ,t ) t2 − y − x
Using equation (7) and (8) in equation (6)
⎡ n −2 ⎛ ⎞
2 α (n ) ⎢ ∂ ⎛1 ∂ ⎞ 2 ⎜tn gdy ⎟
u ( x,t ) = ∫
γ n +1 (n + 1) α (n + 1) ⎢ ∂t ⎝⎜ t ∂t ⎠⎟ ⎜⎜ 2 2 ⎟⎟
⎢⎣ ⎝ B ( x ,t ) t − y − x ⎠
n −2
⎛ ⎞⎤
⎛1 ∂ ⎞ 2
⎜tn h dy ⎟⎥
+⎜ ⎟
⎝ t ∂t ⎠ ⎜ ∫ 2 2∫
⎟⎥
⎝ B ( x ,t ) t − y − x ⎠ ⎥⎦
But
2α (n ) 1
=
γ n +1 (n + 1) α (n + 1) 2.4.... (n − 2) n
1
= (say) (9)
γn
Hence
95
⎡ n −2 ⎛ ⎞
1 ⎢ ∂ ⎛1 ∂ ⎞ 2 ⎜ n gdy ⎟
u ( x,t ) = ⎜ ⎟ t ∫
⎢
γ n ∂t ⎝ t ∂t ⎠ ⎜
⎜ 2 2 ⎟⎟
⎢⎣ ⎝ B ( x ,t ) t − y − x ⎠
n −2
⎛ ⎞⎤
⎛1 ∂ ⎞ 2
⎜tn h dy ⎟⎥
+⎜ ⎟
⎝ t ∂t ⎠ ⎜ ∫ 2 2∫
⎟⎥
⎝ B ( x ,t ) t − y − x ⎠ ⎥⎦
is required solution, where n is even.
Note. For n = 2, γ 2 = 2 , we get the Poisson’s formula.
4.6 Solution of Non-homogeneous wave equation
utt − Δu = f in R n × ( 0, ∞ ) (1)
u = 0 , ut = 0 on R n × {t = 0} (2)
where f ∈ C ⎢⎣n /2 ⎥⎦ +1 R n × ( 0, ∞ ) ; ( )
⎡ ⎤
⎡n
⎣
/ 2 ⎤⎦ denotes the greatest integer
function, then solution of equation (1) subject to (2) is

t
u ( x , t ) = ∫ u ( x , t ; s ) ds x ∈ Rn t ≥ 0 (3)
0
where u ( x , t ; s ) is a solution of
utt ( x , t ; s ) − Δu ( x , t ; s ) = 0 in R n × ( s, ∞ )
u ( x , t ; s ) = 0; ut ( x , t ; s ) = f ( x , t ; s ) on R n × {t = s} (4)
Sol. To show that equation (3) is a solution of equation (1) subject to (2)
we need to show
(i) (
u ∈ C 2 R n × [0, ∞ ) )
(ii) utt − Δu = f ( x , t ) in R n × ( 0, ∞ )
(iii) lt u ( x,t ) = 0
( x ,t ) → ( x 0 ,0 )
96
lt ut ( x , t ) = 0
(n ,t ) → ( x 0 ,0 )
for each point x 0 ∈ R n .
⎡n ⎤
Proof. (i) ⎢ ⎥ denotes the greatest integer function.
⎣2⎦
⎡n ⎤ n −1 n +1
If n is even ⎢ ⎥ + 1 = +1 =
⎣2⎦ 2 2
⎡n ⎤ n
If n is odd ⎢⎣ 2 ⎥⎦ + 1 = 2 + 1
From previous article,
(
u ( x , t ; s ) ∈ C 2 R n × (δ , ∞ ) for each δ ≥ 0)
so (
u ∈ C 2 R n × [0, ∞ ) )
t
(ii) u ( x , t ) : = ∫ u ( x , t ; s ) ds
0
Differentiating w.r.t. t
t
ut ( x , t ) : = ∫ ut ( x , t ; s ) ds + u ( x , t ; t )
0
t
= ∫ ut ( x , t ; s ) ds (by 4)
0
Again differentiating w.r.t. t

t
utt ( x , t ) : = ∫ utt ( x , t ; s ) ds + ut ( x , t ; t )
0
t
= ∫ utt ( x , t ; s ) ds + f ( x , t ) (by 4) (5)
0
t
Δu ( x , t ) := ∫ Δu ( x , t ; s ) ds
0
97
t
= ∫ utt ( x , t ; s ) ds (by 3) (6)
0
t
utt ( x , t ) − Δu ( x , t ) = ∫ ⎡⎣utt ( x , t ; s ) − Δu ( x , t ; s ) ⎤⎦ ds + f ( x , t )
0
= f ( x,t )
(iii) Also u ( x ,0 ) = 0
ut ( x ,0 ) = 0
The solution of non-homogeneous wave equation is given by equation (3)

Q. Find the solution of
utt − u xx = f (x , t ) in R × ( 0, ∞ )
u = 0 , ut = 0 on R × {t = 0}
Sol. The solution of homogeneous wave equation in
utt − u xx = 0 in R × ( 0, ∞ )
u = g , ut = h on R × {t = 0}
is given by D Alembert’s formula.
1⎡
n +t ⎤
1
u ( x,t ) = ⎢g ( x + t ) + g ( x − t ) + ∫ h ( y ) dy ⎥
2 ⎢⎣ 2 n −t ⎥⎦
Hence
x +t −s
1
u ( x,t; s ) = f ( y, s ) dy
2 x −∫t + s
(Replacing t by t – s)
Hence
t
x +t −s
1⌠
u ( x , t ) = ⎮ ∫ f ( y, s ) dy ds
2 ⌡ x −t + s
0
Replacing t – s by s ,We find
98
t
x +s
1⌠
u ( x , t ) = ⎮ ∫ f ( y, t − s ) dy ds
2 ⌡ x −s
0
is the required solution.

4.7 Energy Methods
Let U ⊂ R n be a bounded, open set with a smooth boundary ∂U and
U T = U × ( 0,T ]
ΓT = U T − U T where T > 0
there exists at most one function u ∈ C 2 (U T ) of the initial value problem.
utt − Δu = f in U T (1)
u = g on ΓT ; U t = h on U × {t = 0} (2)
Proof. Let u be another solution of equation (1). We take
w ( x,t ) = u − u ,
So
wtt − Δw = 0 in U T (3)
w = 0 on Γt ; wt = 0 on U × {t = 0} (4)
Define
1⌠⎡ 2
e (t ) : = wt + Dw ⎤ dx
2
0 ≤t ≤T
2⌡⎣ ⎦
U
Differentiating w.r.t. t
e& (t ) = ∫ [wt wtt + Dw Dwt ] dx

U
(Integrating the 2nd integral by parts)
∫ ⎡⎣wt wtt − D w wt ⎤⎦ dx + ∫ wt Dw vˆ dS
2
=
U ∂U
99
= ∫ wt [wtt − Δw ] dx + 0 (using 4)
U
=0 (by 3)
So e (t) = Constant for all t.
1⌠⎡ 2
⎮ ⎢wt ( x ,0 ) + D w ( x ,0 ) ⎤⎥ dx
2
But e ( 0 ) =
2⌡⎣ ⎦
U
= 0.
So e (t) is zero for all t.
i.e. Dw ≡ wt = 0 within U T
Since w = 0 on U × {t = 0}
w = u − u = 0 in U T
u = u in U T
Def. Let u ∈ C 2 be a solution of
utt − Δu = 0 in R n × ( 0, ∞ )
Fix x 0 ∈ R n , t0 > 0 .
Consider the set
C= {( x , t ) 0 ≤ t ≤ t0 ; x − x 0 ≤ t0 − t }
which defines a cone.
Theorem. If u ≡ ut ≡ 0 on B ( x 0 , t0 ) × {t = 0} then u = 0 within cone C.
Proof. We define
1 ⌠ ⎡u 2 ( x , t ) + Du ( x , t ) 2 ⎤ dx
e (t ) = ⎮ 0 ≤ t ≤ t0
2 ⌡ ⎣⎢ t ⎦⎥
B ( x0 ,t0 −t )
Differentiating w.r.t. t.
e& (t ) = ∫ ⎡⎣ut utt ( x , t ) + ( D u D ut ) ⎤⎦ dx

B ( x0 ,t0 −t )
100
1 ⌠ ⎡u 2 + Du 2 ⎤ ds
− (By Cor of coarea formula)
2 ⌡ ⎣ t ⎦
∂B ( x 0 ,t0 −t )
Integrating by parts (2nd term of 1st integral)
= ∫ [ut utt − ut Δu ] dx
B ( x0 ,t0 −t )
⌠ ∂u 1 ⌠ ⎡u 2 + Du 2 ⎤ dS
+ ⎮ ut ds −
⌡ ∂v 2 ⌡ ⎣ t ⎦
∂B ( x0 ,t0 −t ) ∂B0 ( x 0 , t0 −t )
⌠ ⎡ ∂u 1 2 1 2⎤
=0+ ⎮ ⎢⎣ut ∂v − 2 ut − 2 Du ⎥⎦ dx
⌡
∂B ( x0 ,t0 −t )
⌠ ⎡ 2 2 1 2 1 2⎤
≤ ⎮ ⎢⎣ut + Du − 2 ut − 2 Du ⎥⎦ dx
⌡
∂B ( x0 ,t0 −t )
(by Cauding Schwartz Inequality)

≤ 0.
So e (t) is a decreasing function of t
e (t ) ≤ e ( 0 )
But e (0) = 0
e (t ) ≤ 0
hence e(t)=0 (Q e (t ) is a sum of square quantities )
ut = D u = 0 within C
⇒ u is constant within C
Hence u = 0 within C (Q u = 0 for t = 0 )

Q. Find the solution of
utt − Δu = f ( x , t ) in R 3 × ( 0, ∞ )
u = 0 , ut = 0 on R 3 × {t = 0}
101
1 ⌠ f ( y, t − y − x ) dy
Ans. u ( x , t ) = ⎮
4π ⌡ y−x
B ( x ,t )

102
Lesson 5
Other Techniques to Represent Solution
Structure
5.1 Introduction
5.2 Separation of variables
5.3 Similarity solutions
5.4 Connecting non-linear partial differential equations to linear
partial differential equations

5.4.1 Cole-Hopf transformation
5.4.2 Potential function
5.5 Transform methods
5.5.1 Fourier Transforms
5.5.2 Laplace Transforms
5.1 Introduction
There are several other techniques to solve the linear and non-
linear partial differential equations. e.g. Separation of variables
,Similarity solutions, Connecting non-linear partial differential equations
to linear partial differential equations, Transform methods. Here we will
discuss them.
103
5.2 Separation of variables
In this method, we assume a solution given by sum or product of

undetermined functions and form ordinary differential equations, which
are solved. This technique is well understood by examples.
Exp. Consider the boundary value problem in heat equation
ut − Δu = 0 in U × ( 0, ∞ ) (1)
u = 0 on ∂U × [0, ∞ ) (2)
u = g on U × {t = 0}
where g : U → R is given.
Sol. Let the solution of equ. (1) be given by
u ( x , t ) = v (t ) w ( x ) x ∈U , t ≥ 0 (3)
From (1) and (3)
v ′w ( x ) − Δw v (t ) = 0
Dividing by w ( x ) v (t )
v ′ (t ) Δw ( x )
= (4)
v (t ) w (x )
L.H.S. of equ. (4) is a function of t only and R.H.S. is a function of x only.

Equ. (4) is true if each side is equal to some constant, say, μ .
v ′ (t ) Δw ( x )
=μ=
v (t ) w (x )
⇒ v ′ (t ) − μ v (t ) = 0 (5)
and
Δw ( x ) − μ w ( x ) = 0 (6)
Considering equ. (5) and integrating
v = C eμ t (7)
104
where C is a constant.
Taking equ. (6), comparing with the
−Δw = λ w in U ⎫
⎬ (8)
w=0 on ∂U ⎭
then λ is eigen value and w ( ≠ 0 ) is the corresponding eigen function.So
μ = −λ is eigen value of equ. (6) and w is corresponding eigen function.
Hence solution of problem defined by equ. (1) and (2) is
u ( x , t ) = Ce − λ t w ( x ) (9)
where C is a constant to be determined from the initial condition at t = 0,

which gives
g = Cw
so u = Ce − λt w
where g = Cw
is required solution.
Particular case:
(a) If λ1, λ2 , ..., λm are eigen values of problem (8) and w1, w2 , ..., wm
are the corresponding eigen functions and c1, c 2 , ..., c m are constants
then solution of equ. (1)

m
u ( x , t ) = ∑ c i e − λi t wi ( x )
i =1
m
provided ∑d w
i =1
i i =g.
(b) Let λ1, λ2 , ... be a countable set of eigen values with corresponding
eigen function w1, w2 , ... so that

∞
u = ∑ c i e − λi t wi ( x )
i =1
105
∞
provided that ∑c w (x ) = g
i =1
i i in U.
Exp. Find the solution of the non-linear porous medium equation
ut − Δ ( u γ ) = 0 in R n × ( 0, ∞ ) (1)
where u ≥ 0 and γ >1 is a constant.
Sol. We seek a solution of equ. (1) of the type
u ( x , t ) = v (t ) w ( x ) (2)
From (1) and (2)
w ( x ) v ′ (t ) − ( Δw γ ) v γ = 0
Dividing by wv γ
v ′ (t ) Δw γ
= (3)
vγ w
L.H.S. is a function of t only and R.H.S. is a function of x only. Equ. 3 is
true if each side is equal to some constant say μ .
v ′ (t )
=μ
vγ
v −γ +1
= μt + λ ,
−γ + 1
where λ is a constant.
v 1−γ = (1 − γ ) μt + λ ,
1
v = ⎡⎣(1 − γ ) μt + λ ⎤⎦1−γ (4)
Δw γ = μ w (5)
α
Suppose w = x is solution of equ. (5) where α is a constant to be
determined.
αγ
Δw γ = Δ x
106
αγ [n + αγ − 2]
α y −2
= x
Using in equ. (5)
αγ (n + αγ − 2) = μ x
α γ −2 α
x (6)
In order to hold equ. (6) in Rn, we must have
2
α = αγ − 2 ⇒ α= and
γ −1
μ = αγ (n + αγ − 2) > 0 (7)
So solution of equ. (1) is

1
u = ⎡⎣(1 − γ ) μt + λ ⎤⎦1−γ x .
α
(8)
where α , μ are given by equ. (7)
Remark. In equ. (8) u is singular when
(1 − γ ) μt + λ = 0 or
λ
t= = t * (say) ,t* is called the critical time.
(γ − 1) μ
5.3 Similarity solution
Certain symmetries of partial differential equation help to convert

them in ordinary differential equations.
Def. Plane Travelling Wave:
A solution u ( x , t ) of the partial differential equations of two
variables x , t ∈ R of the form
u ( x,t ) = v ( x − σ t ) x ∈ R, t ∈ R
represents a travelling wave with speed σ and velocity profile v.
Generalization. A solution u ( x , t ) of a partial differential equations in
n + 1 variables x = ( x1, x 2 , ..., xn ) ∈ R n , t ∈ R having the form
107
u ( x , t ) = v ( y.x − σ t )
is called a plane wave with wave front normal to y ∈ R n .
Exponential solution: The exponential solution of partial differential

equations is
u ( x , t ) = e i (y x +wt )
where w ∈ , y = ( y1, y2 , ..., yn ) ∈ R n , w being frequency and {yi }i =1

n
the
wave number.
Exp. The heat equation
ut − Δu = 0
has the exponential solution.
u =e
(
i yx +i y t
2
)
2
−y t
=e ⎡⎣e iyx ⎤⎦
2 2 2
−y t −y t −y t
e cos yx and e sin yx are solutions of equ.(1).Here the term e
corresponds the dissipation of energy.

Exp. The wave equation
utt − Δu = 0
has the exponential solution
u =e(
i y x ± y t)
Since w is real, no dissipation effects occur.

Exp. The dispersive equation
ut + u xxx = 0 in R × ( 0, ∞ )
has the exponential solution
u =e
(
i y x + y 3t )
No dissipation of energy. Also the velocity of propagation depends on

frequency. Hence dispersion takes place.
108
Exp. Barenblaltt’s solution
Consider the porous medium equation
ut − Δu γ = 0 in R n × ( 0, ∞ ) (1)
where u ≥ 0 and γ >1 is a constant.
Sol. We seek a solution of equ. (1) of the form
1 ⎛x ⎞
u ( x,t ) = v⎜ ⎟ x ∈ Rn , t > 0
tα ⎝ t β ⎠
where α , β are unknowns.
1
u ( x,t ) = v (y )
tα
where y = x / t β (2)

1
α v ( y ) + β y. Dv ( y ) + αγ + 2 β −α −1
vγ = 0 (3)
t
To convert equ. (3) into an equation independent of t, we must have
αγ + 2β − α − 1 = 0 or
1 − 2β
α= (4)
γ −1
Hence equ. (3) gives
αv + β y.Dv + Δv γ = 0 (5)
We seek a radial solution of equ. (5)

Let it be
v ( y ) = w (r ) where r = y (6)
α w + β w′ (r ) r + ( w γ )′′ + (n − 1) ( w γ )′ r n −2 = 0
where dash denotes derivative w.r.t. r.
To make it exact differential, multiplying by r n −1 and taking α = n β
109
⎡ ⎤′
β (r n w )′ + ⎢r n −1 ( w γ )′ ⎥ = 0
⎣ ⎦
Integrating and assuming that as r → 0 w, w′ → 0
β r n w + r n −1 ( w γ )′ = 0
(w )′ = −β r w
γ
γ w γ −1 = − β rw
−β
or w γ −2 = r
γ
Again integrating
−β r 2
w γ −1 = +b
2γ ( γ − 1)
where b is a constant.
1
⇒
⎡
w (r ) = ⎢b −
(γ − 1) β r 2 ⎤ γ −1
⎥
⎣ 2γ ⎦
Hence
1
⎡ ( γ − 1) β x ⎤ γ −1
2
v ( y ) = ⎢b − ⎥
⎢⎣ 2γ t 2 β ⎥⎦
1 − 2β
where α = nβ =
γ −1
1
i.e. β=
2 − n + nγ
n
α=
2 − n + nγ
110
5.4 Connecting non-linear partial differential equations to linear
partial differential equations
5.4.1 Cole-Hopf transformation
Consider the initial value problem for a quasi-linear parabolic

equation
ut − a Δu + b Du = 0 in R n × ( 0, ∞ )
2
(1)
u = g on R n × {t = 0} (2)
where a > 0;a,b are constants..
Sol. Let w = φ ( u ) (3)
where u is a smooth solution of equ. (1) and φ : R → R is a smooth
function. We seek φ such that w solves the linear equation.
From (1) and (3)
Dw = φ ′ ( u ) Du
Δw = φ ′ ( u ) Δu + φ ′′ ( u ) Du
2
wt = φ ′ ( u ) ⎡a Δu − b Du ⎤
2
⎣ ⎦
= a ⎡ Δw − φ ′′ ( u ) Du ⎤ − bφ ′ ( u ) Du
2 2
⎣ ⎦
Hence, wt − a Δw = − ⎡⎣aφ ′′ ( u ) + b φ ′ ( u ) ⎤⎦ Du
2
We choose φ such that
aφ ′′ ( u ) + bφ ′ ( u ) = 0 (4)
So we have
wt − a Δw = 0 (5)
To find the solution of equ. (4)
Auxiliary equation is am 2 + bm = 0
roots with m = 0, − b / a
111
Hence
φ ( u ) = e −(b/a )u +C ,where C is a constant.
Neglecting the constant
w ( x , t ) = e −(b/a )u (*)
w ( x ,0 ) = e −(b/a )g (6)
wt − a Δw = 0 in R n × ( 0, ∞ )
b
− g
w =e a
on R n × {t = 0}
which is heat equation having the solution

2
− x −y −b
1 ⌠ g
w ( x,t ) = ⎮e 4at
ea dy x ∈ Rn
( 4π at ) ⌡n
n /2
R
a
or u ( x,t ) = − log w
b
a ⎡ 1
− x −y
2
−b ⎤
u ( x , t ) = − log ⎢ ⌠
⎮ e 4 at
e a
g
dy ⎥ x ∈ Rn , t > 0
⎢ ( 4π at ) ⌡n
n /2
b ⎥
⎣ R ⎦
This is the required solution. Equation (*) is known as Cole Hopf

transformation.
Exp. Find the solution of Burger’s equation with viscosity
ut − a u xx + u u x = 0 in R × ( 0, ∞ )
u=g on R × {t = 0} (1)
Sol. Let us take

x
w ( x,t ) : = ∫ u (y, t ) dy
−∞
x
h (x ) : = ∫ g (y ) dy (2)
−∞
112
so that wx = u, wx ( x ,0 ) = u ( x ,0 ) = g ( x ) = h ′ ( x )
From (1) and (3)
wxt − a wxxx + wx wxx = 0 in R × ( 0, ∞ )
∂ ⎡ 1 2⎤
∂x ⎢⎣wt − a wxx + 2 wx ⎥⎦ =0 in R × ( 0, ∞ )
Thus, problem is converted to
1 2
wt − a wxx + wx = 0 in R × ( 0, ∞ )
2
w ( x ,0 ) = h ( x ) on R × {t = 0} (4)
The equation (4) is a quasi-linear parabolic equation (previous example)

1
with b = .So solution of equ. (4) is
2
⎡ ⎤
2
∞ − x −y h (y )
1 −
w ( x , t ) = −2a log ⎢ ∫e
4at 2a
dy ⎥
⎢ ( 4π at )n /2 ⎥
⎣ −∞
⎦
Differentiating w.r.t. x
∞ 2
− x −y h (y )
⌠ x −y −
⎮ t e dy
4at 2a
⌡
u = wx = −∞
∞ 2
− x −y h (y )
⌠
⎮e 4at
−
2a
dy
⌡
−∞
This is required solution.
5.4.2 Potential function
By use of potential function, non-linear partial differential

equations can be converted to linear partial differential equations.
Exp. Consider the Euler’s equation for inviscid, incompressible flow
ut + u.Du = −Dp + f in R 3 × ( 0, ∞ )
div u = 0 in R 3 × ( 0, ∞ )
113
u = g on R 3 × {t = 0} (1)
where f and g are prescribed functions, u and p are unknowns.
Sol. Let the external body force be derived from potential function h,
such that
f = Dh (2)
Let the velocity u be derived from the potential v s.t.
u = Dv (3)

div u = Δv = 0 (4)
So from equ. (4) we can find v and thus u .
From (1) and (3)
Dvt + Dv D ( Dv ) = −Dp + Dh
or
⎡ 1 2 ⎤
D ⎢v t + Dv + p − h ⎥ = 0
⎣ 2 ⎦
Integrating
1 2
vt + Dv + p = h
2
which is Bernoulli’s equation to get p.
5.5 Transform Methods
5.5.1 Fourier Transforms
We now discuss the transform methods to solve linear and non-linear
partial differentiation equations. First we define Fourier transform over L1
and L2 spaces, respectively.
( )
Def. Let u ∈ L1 R n , we define the Fourier transform of u ( x ) , denoted by
û ( y ) as
114
1
uˆ ( y ) = ∫n e
−i x .y
u ( x ) dx y ∈ Rn
( 2π ) n /2
R
and its inverse Fourier transform
( 1
u (y ) : = ∫n e
i x. y
u ( x ) dx y ∈ Rn
( 2π )
n /2
R
Since e ±i x y = 1 and u ∈ L1 R n ( )
So integral converges for each y.
Plancherel’s theorem
Assume that u ∈ L1 R n ∩ L2 R n ( ) ( ) then

(
uˆ, u ∈ L2 R n ( ) and
(
= u
uˆ
( )
L2 Rn ( )=
L2 Rn
u
L2 Rn( ) (1)
Proof. To prove (1), we prove three results
(i) ∫n v (y ) wˆ (y ) dy = ∫n vˆ ( x ) w ( x ) dx
R R
1
∫n v (y ) ∫n ⎡⎣w ( x ) e
−ix .y
L.H.S. = dx ⎤⎦ dy
( 2π ) n /2
R R
1
∫n w ( x ) ∫n v (y ) e
−ixy
= dy dx
( 2π ) n /2
R R
= ∫n w ( x ) vˆ ( x ) dx
R
Hence the result.
(ii) ( )
If u,v ∈ L1 R n ∩ L2 R n ( )
then ( u * v ) = ( 2π )
n /2
uˆ vˆ
where * denotes the convolution operator.
115
By def.
u *v = ∫n u ( z )v ( x − z ) dz
R
1 ⌠ ⎧⎪ ⎫⎪
( u * v )^ = ⎮ e −i x y ⎨∫ u ( z ) v ( x − z ) ⎬ dx
( 2π )n /2 ⎮
⌡ ⎩⎪R n ⎭⎪
Rn
1 −i y ( x − z )
∫n v ( x − z ) e u ( z ) dz
−i z y
=
( 2π ) n /2
dx ∫n e
R R
= vˆ ( y ) ∫n e
−i z y
u ( z ) dz
R
= vˆ ( y ) ⎡uˆ ( y )( 2π )
n /2 ⎤
⎣ ⎦
= ( 2π ) uˆ ( y ) vˆ ( y )
n /2
(iii) Consider
n
−i x y − t x 2 −ixi yi −t x 2
∫n e dx = ∏ ∫ e i dx
R i =1 R
∞ ⎡ ix y ⎤
−ix y −t x 2 −t ⎢ x 2 + i i ⎥
But ∫ e i i i dxi = ⌠
⎮e ⎣ i t ⎦
dx
R
⌡
−∞
∞ ⎛ iy ⎞2 ⎛ iy ⎞
2
−t ⎜ xi + i ⎟ + t ⎜ i
⌠ ⎟
=⎮ e ⎝ 2t ⎠ ⎝ 2t ⎠ dxi
⌡
−∞
−y 2 /4t ∞
e i 2 ⎛ iy ⎞
= ∫ e − z dz where z = t ⎜ xi + i ⎟
t −∞ ⎝ 2t ⎠
−y 2 /4t
e i
= π
t
Hence
116
n /2
−ix .y −t x
2 ⎛π ⎞ 2
− y /4t
∫ e dx = ⎜ ⎟
⎝t ⎠
e
Rn
Proof of theorem:
Choosing a function for ∈> 0
2
v∈ ( x ) = e
−∈ x
1
vˆ∈ ( y ) = ∫n e
−ixy
v∈ ( x ) dx (Using result (iii), putting t = ∈ )
( 2π ) n /2
R
1 − y 2 /4∈
= e (2)
( 2 ∈) n /2
Hence
−∈ y 2 1 − x 2 /4∈
∫n wˆ (y ) e dy =
( 2 ∈) n /2 ∫n w ( x ) e dx (Using result (i))
R R
( 3)
Taking limit as ∈→ 0
⌠ 1 − x 2 /4∈
⎮wˆ ( y ) dy = lt ∫n w ( x ) e dx
∈→ 0 ( 2 ∈)n /2
⌡ R
Rn
xi2
= ( 2π )n /2 w (0) where = zi2 (4)
4∈
Suppose u ∈ L1 R n ∩ L2 R n ( ) ( )
and set v ( x ) : = u ( −x ) , u is the conjugate of u.
w (x ) : = u * v
= ∫n u ( z ) v ( x − z ) dz
R
ˆ = ( 2π )
n /2
w uˆ vˆ (by result II)
117
1
But vˆ = ∫n e
−ixy
u ( −x ) dx
( 2π ) n /2
R
1
= ∫n e
ixy
u ( x ) dx
( 2π ) n /2
R
1 uuuuuuuuuuuuuv
= ∫ e −i x y u ( x ) dx
( 2π )n /2 Rn
= û ( y )
ˆ = ( 2π )
n /2 2
∴w uˆ (5)
From (4) and (5)
( 2π )n /2 ∫ uˆ
2
dy = ( 2π )
n /2
w (0)
Rn
∫n u ( z ) u ( z ) dz
2
or ∫n uˆ dy =
R R
2
= ∫n u dz (by def.)
R
uˆ
( )=
L2 Rn
u
( )
L2 Rn
(
Similarly u
( )=
L2 Rn
u
( )
L2 Rn
(The result can be obtained by previously argument changing i to –i)

Hence
(
u = uˆ = u
( )
L2 Rn L2 Rn ( ) ( )
L2 Rn
Note
Since u ∈ L2 R n ( ) choose a sequence {u } ∞

k k =1 ( ) ( ) with
⊂ L1 R n ∩ L2 R n
uk → u in L2 R n . ( )
By (1)
118
uk − u j = uˆk − uˆ j = uk − u j
( )
L2 R n ( )
L2 R n ( )
L2 R n
{uˆk }k∞=1 ( )
is a Cauchy sequence in L2 R n which converges to û .
So uˆk → uˆ in L2 R n ( )
Def Fourier Transform of u over L2 R n ( )
( )
Let u ∈ L2 R n then
(
= u
uˆ
( )
L2 Rn ( )=
L2 Rn
u
( )
L2 Rn
(
So uˆ, u ∈ L2 R n ( ) (by above theorem)
(
hence uˆ, v are well defined over L2 R n . ( )
Properties of Fourier transform:
Assume u,v ∈ L2 R n ( )
(i) ∫n u v dx = ∫n uˆ vˆ dy
R R
α
(ii) D α u = (iy ) uˆ
for each multiindex α s.t. D α u ∈ L2 R n ( )
Proof. Let u, v ∈ L2 R n ( ) and α ∈ C n
then
2 2
u + αv = uˆ + αvˆ (Using Plancherel’s theorem)
i.e. ∫n (u + αv )(u + αv ) dx = ∫n (uˆ + αvˆ ) (uˆ + αvˆ ) dy

R R
⇒ ⌠ ⎡ u + αv + u (αv ) + u (α v ) ⎤ dx
2 2
⌡ ⎣ ⎦
Rn
119
) )
( )
= ⌠ ⎡ uˆ + αvˆ + u (α v ) + uˆ α vˆ ⎤ dy
⌡ ⎣
2 2
⎦
Rn
or ∫4 ⎡⎣u (αv ) + u (αv )⎤⎦ dx = ∫n (α uv ˆ ˆ ) dy

ˆ ˆ + α uv (1)
R R
Taking α = 1,i in (1) respectively and subtracting we obtain
∫n uv dx = ∫n (uv
ˆ ˆ ) dy
R R
(ii) If u is smooth and has compact support
1
Dα u = ∫n e
−ix y
D α u dx
( 2π ) n /2
R
α
( −1)
Dα e −ixy u ( x ) dx
n /2 ∫
=
( 2π ) Rn
α
( −1) α α
e −ixy ( −1) (iy ) u ( x ) dx
n /2 ∫
=
( 2π ) Rn
α
= (iy ) uˆ ( y )
Exp. Solve the partial differential equation
−Δu + u = f in R n (1)
where f ∈ C 2 R n . ( )
Sol. Taking Fourier transform of equation (1)
− (iy ) uˆ + uˆ = fˆ ,
2
y ∈ Rn
fˆ
uˆ = (2)
1 + y2
Taking inverse Fourier transform of (2)

v
⎛ fˆ ⎞
u =⎜ 2⎟
⎝1 + y ⎠
u = f * B where
120
1
B=
(1 + y ) 2 ∨
To find B, we know that

∞
1
= e −ta dt
a 0∫
∞
1 −t ⎛⎜1+ y 2 ⎞⎟
so = ⌠e ⎝ ⎠ dt
1+ y
2 ⌡
0
∨
⎛ 1 ⎞
⇒⎜ ⎟ =
1 ⌠ ∞ −t
⎮⌠⎮e
(1+ y 2 )e i x ydt dy
⎜1 + y 2 ⎟ ( 2π )n /2 ⎮⌡
⎝ ⎠ ⌡ 0
Rn
∞
n /2
1 ⌠ −t ⎛ π ⎞ 2
− x /4t
= ⎮e ⎜ t ⎟ e dt
( 2π )n /2 ⌡ ⎝ ⎠
0
∞ 2
x
⌠ e −t −
1 4t dt
= n /2 ⎮ n /2
x ∈ Rn (3)
2 ⎮ t
⌡
0
So,
∞
⌠ x −y
2
⌠ −t −
1 ⎮ f (y ) e 4t
u ( x,t ) = ⎮⎮ dy dt
( 4π )n /2 ⎮⎮ t n /2
⎮ ⌡n
⌡R
0
Here ,B given in equation (3), is called the Bessel ’s potential.

Exp. Find the solution of initial value problem of heat equation
ut − Δu = 0 in R n × ( 0, ∞ ) (1)
u=g on R n × {t = 0} (2)
121
Sol. Taking Fourier transform of equation (1) and (2) w.r.t. the spatial
variable x.
uˆt − (iy ) uˆ = 0
2
for t > 0 (3)
uˆ = gˆ for t = 0 (4)
or
uˆt
= −y 2
uˆ
Integrating
2t
uˆ = Ce −y , where C is a constant.
Since C = ĝ (Using (4)
− y2 t
uˆ = ge
ˆ
Taking inverse Fourier transform
g *F
u=
( 2π )n /2
where,
∨
⎛ −t y 2 ⎞
F = ⎜e ⎟
⎝ ⎠
1 − x 2 /4t
= e
(2t ) n /2
Hence solution is
2
− x −y
1 ⌠
u ( x,t ) = ⎮ g (y ) e 4t dy
( 4π t ) n /2
⌡
Rn
Exp. Solve the Schroödinger’s equation.
iut + Δu = 0 in R n × ( 0, ∞ ) (1)
u=g on R n × {t = 0} (2)
122
where u and g are complex valued functions .
Sol. Equ. (1) can be rewritten as
∂u
+ Δu = 0
∂ ( −it )
which is obtained from heat equation replacing t by it, hence we get

2
i x −y
1 ⌠
u ( x,t ) = ⎮ e 4t g ( y ) dy t ≠0 (3)
( 4π it ) n /2
⌡
Rn
which is required solution.

Remark. From equ. (3), we can obtain the fundamental solution of
Schrodinger equation
2
ix
1
Ψ ( x, y ) : = e 4t x ∈ Rn , t ≠ 0
( 4π it ) n /2
Exp. Find the solution of initial value problem
utt − Δu = 0 in R n × ( 0, ∞ ) (1)
u=g ⎫
⎬ on R n × {t = 0} (2)
ut = 0 ⎭
Sol. Taking Fourier transform of equation (1) w.r.t. x

2
uˆtt + y uˆ = 0 for t > 0 (3)
uˆ = gˆ uˆt = 0 for t = 0 (4)
We seek an exponential solution of equ. (3). Let
uˆ = β e t γ where β , γ ∈ are to be determined.
From (1) and (3)

2 2
γ +y =0
γ = ±i y
iyt −i y t
uˆ = β1e + β 2e
123
Using equation (2), we obtain
β1 − β2 = 0
⇒ β1 = β2 = β (say)
and β = gˆ / 2
Hence
uˆ ( y, t ) =
2
(
gˆ i
e
yt
+e
−i y t
)
Taking inverse Fourier transform
u ( x,t ) =
( 2π )
1
n /2 ⎮
⌡ 2
( )
⌠ gˆ e i y t + e −i y t e ixy dy x ∈ Rn , t ≥ 0
R4
is the required solution.
5.5.2 Laplace Transforms
Laplace transform method is useful for functions defined on R+ i.e.
( 0, ∞ ) if u ∈ L1 ( R + ) , we define the Laplace transform of u,
∞
L ( u ( s ) ) : = ∫ e −st u (t ) dt s≥0
0
We denoted by u .
Exp. Solve the heat equation
vt − Δv = 0 in U × ( 0, ∞ ) (1)
v=f on U × {t = 0} (2)
Sol. Taking Laplace transform of (1) w.r.t. t

∞
Δv ( x , s ) = ∫ e −st Δv ( x , t ) dt
0
∞
= ∫ e −stvt ( x , t ) dt
0
124
∞
∞
= e −st
v ( x,t ) + ∫ s e −stv ( x , t ) dt
0
0
∞
= e −stv ( x , t ) + sv ( x , s )
0
= − f ( x ) + sv ( s )
Hence
−Δv ( s ) + sv ( s ) = f (3)
Equation (3) is called Resolvent equation. The solution of resolvent

equation is the Laplace transform of equation (1).
Exp. Solve the Hamilton Jacobi equation
ut + H ( Du ) = 0 in R n × ( 0, ∞ )
where H is the Hamilton function.

Exp. Find the exponential solution of Schrodinger ’s equation
i ut + Δu = 0 in R n
Q. Solve the telegraph equation.
utt + 2dut − u xx = 0 in R × ( 0, ∞ )
u=g ut = h on R × {t = 0}
for d > 0.
Q Prove that
(i) ( )
If u,v ∈ L1 R n ∩ L2 R n ( ) then
(u * v )n = ( 2π )n /2 uˆ vˆ
(ii) u= (û )∨
125
126

Partial Differential Equations PDF

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Partial Differential Equations PDF

Transféré par

Droits d'auteur :

Formats disponibles

Directorate of Distance Education

Guru Jambheshwar University of Science & Technology

Partial Differential Equations

Lesson 1: Preliminaries 3-20

Lesson 2:Solution of Linear Partial 21-51

Lesson 3:Green’s Function 52- 77

Lesson 4:Solution of Wave Equation 78-102

Lesson 5 :Other Techniques to

R epresent Solution 103-126

Vetted by: Prof. Sarva Jit Singh

1.8 Solutions of Partial Differential Equations

(i) R n = n-dimensional real Euclidean space

(ii) R1 = R = real line

(iv) A point x in Rn is x = ( x1, x 2 ,... xn )

where t is time variable.

(vii) U, V, W denote open subsets of Rn. We write

V⊂⊂U if V ⊂ V ⊂ U and V is compact

i.e. V is compactly contained in U.

(xiii) α (n ) = volume of unit ball B (0, 1) in Rn

nα (n ) = surface area of unit sphere B (0, 1) in Rn

(xv) denotes n dimensional complex plane.

(b) Notations for functions

(ii) u is smooth if u is infinitely differentiable.

u :=v means u is equal to v.

where u = ( u1, u 2 ,..., u m )

(vii) The symbol

denotes the integral of f over (n - 1) dimensional surface Σ in Rn.

denotes line integral of f over the curve C in Rn.

denote the volume integral of S over V ∈ R n and x ∈ V is an

= average of f over ball B (x, r)

= average of f over surface of ball B (x, r)

for some constant C and all x, y ∈ U .We denote

(xii) The convolution of functions f, g is denoted by f * g.

(c) Notations for derivatives

(ii) Multi-index Notation

(a) A vector α of the form α = (α1 , ..., αn ) where each αi is a non-

negative integer, is called a multi-index of order

(b) Given multi-index α , define

(c) If k is a non-negative integer

is called Hessian Matrix.

(iv) Let x , y ∈ R n i.e. x = (x1, …, xn), y = (y1, …, yn)

(d) Function Spaces

(ii) C (U ) = {u ∈ C ( u ) u is uniformly continuous

on bounded subsets of U, for all α ≤ k }

(v) C ∞ (U ) = {u : U → R|u is infinitly differentiable}

(vi) CC (U ) means C (U) has compact support.

Similarly , CCK (U ) means C k (U ) has compact support.

(vii) The function u : U → R is Lebsegue measurable over Lp if

The function u : U → R is Lebsegue measurable over L∞ if

where u = ( u1, u 2 , ..., u m )

Similarly other operators follow.

f = O ( g ) as x → xO provided there exists a constant C such that

for all x sufficiently close to xO

(a) Cauchy’s Inequality

(b) Holder’s Inequality

(d) Cauchy Schwartz Inequality

Let U ⊂ R n be open and bounded, k = {1, 2, ..., }

exists r > 0 and a Ck function ϒ : R n −1 → R such that

Also , ∂U is analytic if ϒ is analytic.

(ii) If U is C 1 , then along ∂U , the outward unit normal at any point

(iii) Let u ∈ C 1 (U ) then normal derivative of u is denoted by

(b) Gauss- Green Theorem