Chapter 3: Heat equation

We shall derive heat equation from the principle of conservation of energy

and the fact that heat flows from hot regions to cold regions.
1. Derivation:(Conduction of Heat in a Rod)
Consider a wire or rod of length L which is made of some heat-conducting
material and is insulated on the outside, except possibly over the ends at
x = 0 and x = L. Let u(x, t) denote the temperature at x at time t. u(x, t) is
assumed to be constant on each cross section at each time. By the principle
of conservation of energy (heat energy),

the Figure 2.2: A thin rod of length L

Rate of change of heat inside the segment PQ (between x and x + ∆x) is

equal to the net heat flux across the boundaries and the total heat generated
inside PQ. If c is specific heat of the rod, ρ is the density of the rod, A is the
cross-section area of the rod, k is thermal conductivity of the rod and f(x, t)
is the external heat source, then we calculate these terms as follows:
Total amount of heat inside the segment PQ at time t
𝑥+∆𝑥
𝐻 = ∫𝑥 𝑐𝜌𝐴𝑢(𝜏, 𝑡)𝑑𝜏

Rate of change of heat inside 𝑃𝑄

𝑑𝐻 𝑑 𝑥+∆𝑥 𝑥+∆𝑥
= ∫𝑥 𝑐𝜌𝐴𝑢(𝜏, 𝑡)𝑑𝜏 = 𝑐𝜌𝐴 ∫𝑥 𝑢𝑡 (𝜏, 𝑡)𝑑𝜏
𝑑𝑡 𝑑𝑡

Net flux of heat across the boundaries = 𝑘𝐴[𝑢𝑥 (𝑥 + ∆𝑥, 𝑡) − 𝑢𝑥 (𝑥, 𝑡)]
Heat generated due to external heat source inside 𝑃𝑄
𝑥+∆𝑥
= 𝐴 ∫𝑥 𝑓(𝜏, 𝑡)𝑑𝜏
By the principle of conservation of energy, we write
𝑑 𝑥+∆𝑥 𝑥+∆𝑥
∫ 𝑐𝜌𝐴𝑢(𝜏, 𝑡)𝑑𝜏 = 𝑐𝜌𝐴 ∫ 𝑢𝑡 (𝜏, 𝑡)𝑑𝜏
𝑑𝑡 𝑥 𝑥
 𝑥+∆𝑥
= 𝑘𝐴[𝑢𝑥 (𝑥 + ∆𝑥, 𝑡) − 𝑢𝑥 (𝑥, 𝑡)] + 𝐴 ∫ 𝑓(𝜏, 𝑡)𝑑𝜏 … . . (3.1)
𝑥

Applying Mean Value Theorem for integral, we obtain

𝑐𝜌𝐴𝑢𝑡 (𝜉1 , 𝑡)Δ𝑥 = 𝑘𝐴[𝑢𝑥 (𝑥 + ∆𝑥, 𝑡) − 𝑢𝑥 (𝑥, 𝑡)] + 𝐴𝑓(𝜉2 , 𝑡)Δ𝑥
𝑤ℎ𝑒𝑟𝑒 𝜉1 , 𝜉2 ∈ (𝑥, 𝑥 + ∆𝑥) 𝑎𝑛𝑑
𝑘 [𝑢𝑥 (𝑥 + ∆𝑥, 𝑡) − 𝑢𝑥 (𝑥, 𝑡)] 1
ℎ𝑒𝑛𝑐𝑒 𝑢𝑡 (𝜉1 , 𝑡) = + 𝑓(𝜉2 , 𝑡).
𝑐𝜌 ∆𝑥 𝑐𝜌
Now, letting ∆x → 0, we arrive at
𝑘
𝑢𝑡 (𝑥, 𝑡) = 𝛼 2 𝑢𝑥𝑥 (𝑥, 𝑡) + 𝐹(𝑥, 𝑡) … … (3.2) 𝑤ℎ𝑒𝑟𝑒 𝛼 2 = is called
𝑐𝜌
1
the thermal diffusivity of the rod and 𝐹(𝑥, 𝑡) = 𝑓(𝑥, 𝑡) is called the heat
𝑐𝜌
source density.
2. Conduction of Heat in Solids(Derivation):
We consider a domain D∗ bounded by a closed surface B∗ .
Let u (x, y, z, t) be the temperature at a point (x, y, z) at time t. If the
temperature is not constant, heat flows from places of higher temperature
to places of lower temperature. Fourier’s law states that the rate of flow
is proportional to the gradient of the temperature. Thus the velocity of
the heat flow in an isotropic body is
𝑣⃗ = −𝐾 𝑔𝑟𝑎𝑑 𝑢 … … (3.3)
where K is a constant, called the thermal conductivity of the body.
Let D be an arbitrary domain bounded by a closed surface B in D∗. Then
the amount of heat leaving D per unit time is

∬ 𝑣𝑛 𝑑𝑠 𝑤ℎ𝑒𝑟𝑒 𝑣𝑛 = 𝑣⃗. 𝑛⃗⃗

𝐵
is the component of 𝑣⃗ in the direction of the outer unit normal 𝑛⃗⃗ of B.
Thus, by Gauss’ divergence theorem

∬ 𝑣𝑛 𝑑𝑠 = ∭ 𝑑𝑖𝑣 𝑣⃗ 𝑑𝑥 𝑑𝑦 𝑑𝑧 = − ∭ 𝐾 𝑑𝑖𝑣( 𝑔𝑟𝑎𝑑 𝑢 )𝑑𝑥 𝑑𝑦 𝑑𝑧

𝐵 𝐷 𝐷

= −𝐾 ∭ ∇2 𝑢 𝑑𝑥 𝑑𝑦 𝑑𝑧
𝐷
……………(3.4)

Fig 2.3
 But the amount of heat in D is given by

𝐻 = ∭ 𝜎𝜌 𝑢 𝑑𝑥 𝑑𝑦 𝑑𝑧
𝐷
𝑤ℎ𝑒𝑟𝑒 𝜌 , 𝜎 are the density of the material of the body and its specific
heat respectively. Assuming that integration and differentiation are
interchangeable, the rate of decrease of heat in D is
𝑑𝐻 𝜕𝑢
= − ∭ 𝜎𝜌 𝑑𝑥 𝑑𝑦 𝑑𝑧
𝑑𝑡 𝐷 𝜕𝑡
Since the rate of decrease of heat in D must be equal to the amount of
heat leaving D per unit time, we have
𝜕𝑢
− ∭ 𝜎𝜌 𝑑𝑥 𝑑𝑦 𝑑𝑧 = ∬ 𝑣𝑛 𝑑𝑠 = −𝐾 ∭ ∇2 𝑢 𝑑𝑥 𝑑𝑦 𝑑𝑧
𝐷 𝜕𝑡 𝐵
𝐷
𝜕𝑢
𝑜𝑟 ∭ (𝜎𝜌 − 𝐾∇2 𝑢)𝑑𝑥 𝑑𝑦 𝑑𝑧 = 0 … … (3.5)
𝐷 𝜕𝑡
for an arbitrary D in D∗ . We assume that the integrand is continuous. If
we suppose that the integrand is not zero at a point (𝑥0 , 𝑦0 , 𝑧0 ) in D,
then, by continuity, the integrand is not zero in a small region
surrounding the point (𝑥0 , 𝑦0 , 𝑧0 ). Continuing in this fashion we extend
the region encompassing D. Hence the integral must be nonzero. This
contradicts (3.5.5). Thus, the integrand is zero everywhere, that is,
𝜕𝑢 𝜕𝑢 𝐾
𝐾∇2 𝑢 = 𝜎𝜌 𝑜𝑟 = 𝑘∇2 𝑢 𝑤ℎ𝑒𝑟𝑒 𝑘 = … … . . (3.6)
𝜕𝑡 𝜕𝑡 𝜎𝜌
This is known as the heat equation in solid.

3. Solution of Heat Conduction Problem in a finite Rod:

We consider a homogeneous rod of length l. The rod is sufficiently thin
so that the heat is distributed equally over the cross section at time t. The
surface of the rod is insulated, and therefore, there is no heat loss through
the boundary. The temperature distribution of the rod is given by the
solution of the initial boundary-value problem
𝑢𝑡 = 𝑘𝑢𝑥𝑥 , 0 < 𝑥 < 𝑙 ,𝑡 > 0
𝑢(0, 𝑡) = 0 = 𝑢(𝑙, 𝑡) , 𝑡 ≥ 0
𝑢(𝑥, 0) = 𝑓(𝑥) , 0 ≤ 𝑥 ≤ 𝑙 … … . . (3.1)
If we assume a solution in the form 𝑢(𝑥, 𝑡) = 𝑋(𝑥)𝑇(𝑡) ≠ 0
 Equation (3.1) yields
𝑋 ′′ (𝑥) 𝑇 ′ (𝑡)
𝑋(𝑥)𝑇 ′ (𝑡) = 𝑘𝑋 ′′ (𝑥)𝑇(𝑡) ⇒ = = 𝜆(𝑠𝑎𝑦)
𝑋(𝑥) 𝑘𝑇(𝑡)
Hence, X and T must satisfy
X′′ = 𝜆 X and T ′ = 𝜆 kT From the boundary conditions, we have
u (0, t) = X (0) T (t)=0= u (l, t) = X (l) T (t)
Thus, X (0) = 0, X (l)=0 as 𝑇(𝑡) ≠ 0 . Now

Case 1: 𝜆 > 0 so taking 𝜆 = α2 solution of X is

𝑋(𝑥) = acosh α 𝑥 + 𝑏 sinh α 𝑥
Initial condition, X (0) = 0, X (l)=0 give a=0 , b=0 as sinh α 𝑙 ≠ 0
Hence solution is trivial .
Case 2 :𝜆 = 0 solution for X is
𝑋(𝑥) = 𝐴 + 𝐵𝑥 initial condition , X (0) = 0, X (l)=0 give A=0, B=0 .
Hence again solution is trivial.
Case 3: 𝜆 < 0 , 𝑇𝑎𝑘𝑖𝑛𝑔 𝜆 = −𝛼 2 Solution for X is
𝑋(𝑥) = 𝐴 cos 𝛼𝑥 + 𝐵 sin 𝛼𝑥,
𝑋(0) = 0 = 𝑋(𝑙) 𝑔𝑖𝑣𝑒 𝐴 = 0 𝑎𝑛𝑑 𝐵 sin 𝛼𝑙 = 0
𝑎𝑠 𝐵 ≠ 0 ⇒ sin 𝛼𝑙 = 0 ⇒ 𝛼𝑙 = 𝑛𝜋 , 𝑛 = 1,2,3 … …
𝑛𝜋𝑥 𝑛𝜋
𝑋𝑛 (𝑥) = 𝐵𝑛 sin 𝛼𝑛 𝑥 = 𝐵𝑛 sin( ) as 𝛼𝑛 = 𝑛 = 1,2,3 ….
𝑙 𝑙

Considering the equation in 𝑇 𝑖. 𝑒. 𝑇 ′ (𝑡) = −𝛼 2 𝑘𝑇 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 𝑖𝑠

𝑛𝜋 2
−𝛼 2 𝑘𝑡 −( ) 𝑘𝑡
𝑇(𝑡) = 𝐶𝑒 ⇒ 𝑇𝑛 (𝑡) = 𝐶𝑛 𝑒 𝑙

Hence non trivial solution of the equation (3.1) is

𝑛𝜋 2 𝑛𝜋𝑥
−( ) 𝑘𝑡
𝑢𝑛 (𝑥, 𝑡) = 𝑋𝑛 (𝑥)𝑇𝑛 (𝑡) = 𝑎𝑛 𝑒 𝑙 sin ( ) 𝑤ℎ𝑒𝑟𝑒 𝑎𝑛 = 𝐵𝑛 𝐶𝑛 … . . (3.2)
𝑙
By superposition principle(??) we obtain a formal series solution
∞
𝑛𝜋 2 𝑛𝜋𝑥
−( ) 𝑘𝑡
𝑢(𝑥, 𝑡) = ∑ 𝑎𝑛 𝑒 𝑙 sin ( ) … … . (3.3)
𝑙
𝑛=1
∞
𝑛𝜋𝑥
𝑁𝑜𝑤 𝑢(𝑥, 0) = 𝑓(𝑥) = ∑ 𝑎𝑛 sin ( ) 0≤𝑥≤𝑙
𝑙
𝑛=1

This holds true if f (x) can be represented by a Fourier sine series with Fourier
coefficients
 𝑙
2 𝑛𝜋𝑥
𝑎𝑛 = ∫ 𝑓(𝑥) sin ( ) 𝑑𝑥 … … . (3.4)
𝑙 𝑙
0

Hence (3.3) is the formal solution of heat equation in a rod with coefficient
given in (3.4). This statement will be true if 𝑓(𝑥) is twice differentiable and
𝑓 ′′′ (𝑥) is piecewise continuous. In fact this ensures that the series (3.3) can be
differentiated term by term twice with respect to x.
4. With time independent boundary condition:
Let the temperatures at the ends of the rod is not zero but constants. Hence the
problem is:
𝑢𝑡 = 𝑘𝑢𝑥𝑥 , 0 < 𝑥 < 𝑙 ,𝑡 > 0
𝑢(0, 𝑡) = 𝑎 , 𝑢(𝑙, 𝑡) = 𝑏 , 𝑡 ≥ 0
𝑢(𝑥, 0) = 𝑓(𝑥) , 0 ≤ 𝑥 ≤ 𝑙 … … . . (4.1)
We give a linear transformation
𝑢(𝑥, 𝑡) = 𝑣(𝑥, 𝑡) + 𝐴 + 𝐵𝑥 𝑤ℎ𝑒𝑟𝑒 𝑣(0, 𝑡) = 𝑣(𝑙, 𝑡) = 0
⇒ 𝑎 = 𝑢(0, 𝑡) = 𝑣(0, 𝑡) + 𝐴 ⇒ 𝐴 = 𝑎
𝑏−𝑎
𝑎𝑔𝑎𝑖𝑛 𝑏 = 𝑢(𝑙, 𝑡) = 𝑣(𝑙, 𝑡) + 𝐴 + 𝐵𝑙 ⇒ 𝐵 =
𝑙
Hence (4.1) becomes
𝑣𝑡 = 𝑘𝑣𝑥𝑥 , 0 < 𝑥 < 𝑙 ,𝑡 > 0
𝑣(0, 𝑡) = 0 , 𝑢(𝑙, 𝑡) = 0 , 𝑡 ≥ 0
𝑏−𝑎
𝑣(𝑥, 0) = 𝑓(𝑥) − 𝑎 − 𝑥, 0 ≤ 𝑥 ≤ 𝑙 … … . . (4.1)
𝑙
Whose solution is clearly
∞
𝑛𝜋 2 𝑛𝜋𝑥
−( ) 𝑘𝑡
𝑣(𝑥, 𝑡) = ∑ 𝑎𝑛 𝑒 𝑙 sin ( )
𝑙
𝑛=1

Where
𝑙
2 𝑏−𝑎 𝑛𝜋𝑥
𝑎𝑛 = ∫{𝑓(𝑥) − 𝑎 − 𝑥} sin ( ) 𝑑𝑥 … … . (4.2 )
𝑙 𝑙 𝑙
0

In terms of original variables

∞
𝑏−𝑎 𝑛𝜋 2 𝑛𝜋𝑥
−( ) 𝑘𝑡
𝑢(𝑥, 𝑡) = 𝑎 + 𝑥 + ∑ 𝑎𝑛 𝑒 𝑙 sin ( ) … … . (4.3)
𝑙 𝑙
𝑛=1
5. With time dependent boundary condition: if the boundary conditions
are time dependent then we can transform the equation into
inhomogeneous heat equation which can be solved by using Duhamel’s
principle (beyond our scope)
Let the temperatures at the ends of the rod is a function of time t. Hence
the problem can be expressed as:
𝑢𝑡 = 𝑘𝑢𝑥𝑥 , 0 < 𝑥 < 𝑙 , 𝑡 > 0
𝑢(0, 𝑡) = 𝑎(𝑡) , 𝑢(𝑙, 𝑡) = 𝑏(𝑡) , 𝑡 ≥ 0
𝑢(𝑥, 0) = 𝑓(𝑥) , 0 ≤ 𝑥 ≤ 𝑙 … … . . (5.1)
Like the previous article We give a linear transformation
𝑢(𝑥, 𝑡) = 𝑣(𝑥, 𝑡) + 𝐴 + 𝐵𝑥 𝑤ℎ𝑒𝑟𝑒 𝑣(0, 𝑡) = 𝑣(𝑙, 𝑡) = 0
⇒ 𝑎(𝑡) = 𝑢(0, 𝑡) = 𝑣(0, 𝑡) + 𝐴 ⇒ 𝐴 = 𝑎(𝑡)
𝑏(𝑡)−𝑎(𝑡)
𝑎𝑔𝑎𝑖𝑛 𝑏(𝑡) = 𝑢(𝑙, 𝑡) = 𝑣(𝑙, 𝑡) + 𝐴 + 𝐵𝑙 ⇒ 𝐵 =
𝑙
Hence the boundary condition of (5.1) become
𝑣(0, 𝑡) = 𝑣(𝑙, 𝑡) = 0 and for initial condition
𝑏(0)−𝑎(0)
𝑓(𝑥) = 𝑢(𝑥, 0) = 𝑣(𝑥, 0) + 𝑎(0) + 𝑥
𝑙
𝑏(0)−𝑎(0)
𝑣(𝑥, 0) = 𝑓(𝑥) − 𝑎(0) − 𝑥
𝑙
But 𝑢𝑡 = 𝑘𝑢𝑥𝑥 will give
𝑏′ (𝑡)−𝑎′ (𝑡)
𝑣𝑡 + 𝑎′ (𝑡) + 𝑥 = 𝑘𝑣𝑥𝑥 which is of the form
𝑙
𝑏′ (𝑡)−𝑎′ (𝑡)
𝑣𝑡 = 𝑘𝑣𝑥𝑥 + ℎ(𝑥, 𝑡) where ℎ(𝑥, 𝑡) = −𝑎′ (𝑡) − 𝑥
𝑙
Which is a heat equation with heat source i.e., inhomogeneous heat
conduction equation. (This can be solved using Duhamel principle)
Uniqueness of the solution of heat conduction equation
(Inhomogeneous):
Let 𝑢1 (𝑥, 𝑡) 𝑎𝑛𝑑 𝑢2 (𝑥, 𝑡) are two solution of the inhomogeneous heat
conduction equation i.e.
𝑢𝑡 − 𝑘𝑢𝑥𝑥 = 𝐹(𝑥, 𝑡) , 0 < 𝑥 < 𝑙 , 𝑡 > 0
Satisfying the initial condition:𝑢(𝑥, 0) = 𝑓(𝑥) , 0 ≤ 𝑥 ≤ 𝑙
And boundary condition 𝑢(0, 𝑡) = 𝑢(𝑙, 𝑡) = 0 , 𝑡 ≥ 0
Then 𝑣(𝑥, 𝑡) = 𝑢1 (𝑥, 𝑡) − 𝑢2 (𝑥, 𝑡) satisfy
𝑣𝑡 − 𝑘𝑣𝑥𝑥 = 0 , 0 < 𝑥 < 𝑙 , 𝑡 > 0
With 𝑣(𝑥, 0) = 0 , 0 ≤ 𝑥 ≤ 𝑙
And 𝑣(0, 𝑡) = 𝑣(𝑙, 𝑡) = 0 , 𝑡 ≥ 0
1 𝑙
Let us define another function 𝐸(𝑡) =
2𝑘
∫0 𝑣 2 (𝑥, 𝑡)𝑑𝑥 .
Therefore 𝐸(𝑡) ≥ 0 as integrand is positive. Then differentiating w.r.t. t
𝑑𝐸 1 𝑙 𝑙 𝑙
we get = ∫0 𝑣 𝑣𝑡 𝑑𝑥 = ∫0 𝑣 𝑣𝑥𝑥 𝑑𝑥 = 𝑣𝑣𝑥 |𝑙0 − ∫0 𝑣𝑥2 𝑑𝑥
𝑑𝑡 𝑘
Since 𝑣(0, 𝑡) = 𝑣(𝑙, 𝑡) = 0 we get
𝑑𝐸 𝑙
= − ∫0 𝑣𝑥2 𝑑𝑥 ≤ 0 . Hence 𝐸(𝑡) is decreasing.
𝑑𝑡
Also 𝐸(0) = 0 𝑎𝑠 𝑣(𝑥, 0) = 0 . Therefore 𝐸(𝑡) ≤ 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑡 . But by
our assumption 𝐸(𝑡) ≥ 0 . Hence
𝐸(𝑡) = 0 ∀ 𝑡 > 0 ⟹ 𝑣(𝑥, 𝑡) = 0 𝑖𝑛 0 ≤ 𝑥 ≤ 𝑙 , 𝑡 ≥ 0
Therefore 𝑣(𝑥, 𝑡) = 𝑢1 (𝑥, 𝑡) − 𝑢2 (𝑥, 𝑡) = 0 ⟹ 𝑢1 (𝑥, 𝑡) = 𝑢2 (𝑥, 𝑡)
Hence solution (if exists ) is unique.