Module – 3 : Multi-dimensional Steady-State Heat Conduction Problems:

Solution by Separation of Variables

Lecture – 13 : Uniqueness of Solution and Introduction to Separation of Variables
1. Uniqueness of Solution of Heat Conduction Problems

2. Introduction to Separation of Variables

Uniqueness of Solution of
Heat Conduction Problems
Uniqueness of Solution for 3D Heat Equation
A well posed heat conduction problem has generally unique solution.
Prove that the solution of the following linear 3D Heat Conduction Problem is unique.
𝜕𝑇(𝐫, 𝑡) Without any loss of
= 𝛻 2 𝑇(𝐫, 𝑡) On the region R, t > 0
𝜕𝑡 generalization, we
consider α = 1
𝑇 𝐫, 𝑡 = 𝑓(𝐫) On the boundary
𝑇 𝐫, 𝑡 = 0 = 𝐹(𝐫) On the region R, t = 0
If possible, let T1(r, t), T2 (r, t) be two different solutions. Define: 𝑇(𝐫, 𝑡) = 𝑇1 (𝐫, 𝑡) − 𝑇2 (𝐫, 𝑡).
Then 𝑇 𝐫, 𝑡 satisfies:
𝜕𝑇(𝐫, 𝑡)
= 𝛻 2 𝑇(𝐫, 𝑡) On the region R, t > 0
𝜕𝑡
𝑇 𝐫, 𝑡 = 0 On the boundary

𝑇 𝐫, 𝑡 = 0 = 0 On the region R, t = 0

1 V (t) ≥ 0 since the integrand

Define: 𝑉 𝑡 = 𝑇 2 (𝐫, 𝑡)𝑑𝑉 ≥ 0
2 𝑉 𝑇 2 𝐫, 𝑡 ≥ 0 for all (r, t).
Uniqueness of Solution for 3D Heat Equation
1
We have defined: 𝑉 𝑡 = 𝑇 2 (𝐫, 𝑡)𝑑𝑉 ≥ 0
2 𝑉

𝑑𝑉(𝑡) 𝜕𝑇(𝐫, 𝑡) Now substitute 𝜕𝑇(𝐫, 𝑡)

⇒ = 𝑇(𝐫, 𝑡) 𝑑𝑉 = 𝛻 2 𝑇(𝐫, 𝑡)
𝑑𝑡 𝑉 𝜕𝑡 from Heat Eq. : 𝜕𝑡
Recall
𝑑𝑉(𝑡) Divergence 𝐅 ∙ 𝐧 𝑑𝑠 = 𝛻 ∙ 𝐅 𝑑𝑉
⇒ = 𝑇 𝐫, 𝑡 𝛻 2 𝑇(𝐫, 𝑡)𝑑𝑉 (A) Theorem 𝑆 𝐷
𝑑𝑡 𝑉

2 Use T(r, t) ≡ T
Now use: 2 2
𝑇 𝛻 𝑇 ∙ 𝐧 𝑑𝑠 = 𝑇𝛻 𝑇 + 𝛻𝑇 ∙ 𝛻 𝑇 𝑑𝑉 = 𝑇𝛻 𝑇 + 𝛻 𝑇 𝑑𝑉
𝑆 𝑉 𝑉

2
⇒ 𝑇 𝛻𝑇 ∙ 𝐧 𝑑𝑠 = 𝑇𝛻 2 𝑇 𝑑𝑉 + 𝛻 𝑇 𝑑𝑉 (B)
𝑆 𝑉 𝑉

𝑑𝑉(𝑡) 2
Combining (A) and (B), we get: = 𝑇 𝛻 𝑇 ∙ 𝐧 𝑑𝑠 − 𝛻𝑇 𝑑𝑉
𝑑𝑡 𝑆 𝑉
 Uniqueness of Solution for 3D Heat Equation
𝑑𝑉(𝑡)
1
We have obtained: = 𝑇 𝛻 𝑇 ∙ 𝐧 𝑑𝑠 −
2
𝛻𝑇 𝑑𝑉 𝑉 𝑡 = 𝑇 2 (𝐫, 𝑡)𝑑𝑉 ≥ 0
𝑑𝑡 2 𝑉
𝑆 𝑉

Given BC: On the boundary S, 𝑇 = 0. So that the first integral on the RHS vanishes.
Thus,
𝑑𝑉(𝑡) 2
=− 𝛻𝑇 𝑑𝑉 ≤ 0
𝑑𝑡 𝑉

At t = 0: 𝑉 𝑡=0 = 𝑇 2 𝐫, 0 𝑑𝑉 = 0 Recall the IC: 𝑇 𝐫, 𝑡 = 0 = 0

𝑉

Thus we get V(0) = 0, V(t) ≥ 0 and dV/dt ≤ 0

So, V(t) is a non-negative, non-increasing function that starts at zero.

Thus V(t) must be zero for all time t, so that 𝑇 𝐫, 𝑡 must be identically zero
throughout the volume V for all time, implying the two solutions are the
same, T1 = T2.
Thus, the solution to the 3D heat problem is unique.
 Uniqueness of Solution for 3D Heat Equation
Now consider the Boundary Condition is: Specified Flux
𝜕𝑇(𝐫, 𝑡) Without any loss of
= 𝛻 2 𝑇(𝐫, 𝑡) On the region R, t > 0
𝜕𝑡 generalization, we
consider α = 1
𝛻𝑇 𝐫, 𝑡 = 𝑞𝑠′′ On the boundary
𝑇 𝐫, 𝑡 = 0 = 𝐹(𝐫) On the region R, t = 0

Again, consider that if possible, let T1(r, t), T2 (r, t) be two different solutions.
Define: 𝑇(𝐫, 𝑡) = 𝑇1 (𝐫, 𝑡) − 𝑇2 (𝐫, 𝑡). Then 𝑇 𝐫, 𝑡 satisfies:

𝜕𝑇(𝐫, 𝑡)
= 𝛻 2 𝑇(𝐫, 𝑡) On the region R, t > 0
𝜕𝑡
𝛻 𝑇 𝐫, 𝑡 = 0 On the boundary

𝑇 𝐫, 𝑡 = 0 = 0 On the region R, t = 0
 Uniqueness of Solution for 3D Heat Equation
Now use the same definition of V and follow the same approach to get:

𝑑𝑉(𝑡) 2 𝑑𝑉(𝑡) 2
= 𝑇 𝛻 𝑇 ∙ 𝐧 𝑑𝑠 − 𝛻𝑇 𝑑𝑉 ⇒ =− 𝛻𝑇 𝑑𝑉 ≤ 0
𝑑𝑡 𝑆 𝑉 𝑑𝑡 𝑉
Due to BC: 𝛻 𝑇 𝐫, 𝑡 = 0

At t = 0: 𝑉 𝑡=0 = 𝑇 2 𝐫, 0 𝑑𝑉 = 0 Recall the IC: 𝑇 𝐫, 𝑡 = 0 = 0

𝑉

Thus we again get V(0)=0, V(t) ≥ 0 and dV/dt ≤ 0

So, V(t) is a non-negative, non-increasing function that starts at zero.

Thus V(t) must be zero for all time t, so that 𝑇 𝐫, 𝑡 must be identically zero
throughout the volume V for all time, implying the two solutions are the
same, T1 = T2.
Thus, the solution to the 3D heat problem is unique.
 Solution of Homogeneous
Heat Conduction Problems by
Separation of Variables
Separation of Variables: Generalization
Boundary-value problems of heat conduction: The boundary-value problem of heat conduction will be
 Homogeneous problems referred to as a homogeneous problem when both the
 Nonhomogeneous problems differential equation and the boundary conditions are
homogeneous.
A linear differential equation or a linear boundary condition is said to be homogeneous if it
is not altered when the dependent variable in the equation is multiplied by a constant.

A homogeneous condition is one in which all nonzero terms in the expression contain
the dependent variable or its derivative.
Multi-dimensional steady-state heat conduction problems and unsteady-state heat
conduction problems are governed by a partial differential equation. A solution must
satisfy the partial differential equation and all the boundary conditions.
Example: A two-dimensional steady-state heat conduction problem is governed by a
second-order partial differential equation and requires four boundary conditions.

When separation among variables is possible, the method of separation of

variables is most effective and straightforward to apply for the solution of
the homogeneous boundary-value problems of heat conduction.
 Separation of Variables: Generalization
Consider a stationary, homogeneous, isotropic solid is initially at a temperature T(r, t) = F(r). For times t > 0,
heat is dissipated by convection from the bounding surfaces into a medium at constant temperature T = 0.
The homogeneous problem can be formulated as: where ∂/∂ni denotes differentiation
1 𝜕𝑇(𝐫, 𝑡) along the outward-drawn normal
Governing Heat
Equation:
= 𝛻 2 𝑇(𝐫, 𝑡) in region R, t > 0 to the bounding surface Si.
𝛼 𝜕𝑡
𝜕𝑇(𝑟𝑖 , 𝑡) Each boundary surface Si fits the
Boundary 𝑘𝑖 + ℎ𝑖 𝑇 𝑟𝑖 , 𝑡 = 0 On each boundary Si, t > 0 coordinate surface of the chosen
condition: 𝜕𝑛𝑖 𝑖 = 1,2, … . , 𝑝 orthogonal coordinate system.
Initial condition: 𝑇 𝐫, 𝑡 = 𝐹(𝐫) in region R, t = 0

Assume a separation in the form: 𝑇 𝐫, 𝑡 = Ψ(𝐫) ∙ Γ(𝑡)

1 1 𝑑Γ(𝑡)
Substitute this equation into the heat equation (PDE): 𝛻2Ψ 𝑟 = = −𝜆2
Ψ 𝑟 𝛼Γ(𝑡) 𝑑𝑡
(13.1)
In Eq. (13.1), the term on the left is a function of space coordinates (r) alone. The
term on the right is a function of time (t) alone. This is possible only if each side is
equal to the same constant, say –λ2 (separation constant).
 Separation of Variables: Generalization
1 1 𝑑Γ(𝑡)
𝛻2Ψ 𝑟 = = −𝜆2
Ψ 𝑟 𝛼Γ(𝑡) 𝑑𝑡
𝑑Γ(𝑡) 2
The function Γ(t) satisfies the ODE: + 𝛼𝜆2 Γ(𝑡) Its solution is taken as: Γ 𝑡 : 𝑒 −𝛼𝜆 𝑡
𝑑𝑡
The negative sign chosen for λ2 and the positive nature of λ2 implies that the solution asymptotically
approaches to zero as time increases indefinitely.

This result is expected from the physical nature of the problem, that is, for a homogeneous boundary
condition of the third kind it implies that a solid dissipates heat from its bounding surfaces by convection
into a surrounding at zero temperature.

Since the solid initially has a prescribed temperature distribution and for times t > 0 it
looses heat to the surrounding continuously but it has no gains to make up for the
losses, then the temperature of the solid will eventually approach to the temperature of
the surrounding (i.e., zero) as time goes to infinity.
Separation of Variables: Generalization
1 2
1 𝑑Γ(𝑡)
𝛻 Ψ 𝑟 = = −𝜆2
Ψ 𝑟 𝛼Γ(𝑡) 𝑑𝑡

The spatial variable function Ψ(𝐫) satisfies the following

characteristic-value (eigenvalue) problem:
𝛻 2 Ψ 𝐫 + 𝜆2 Ψ 𝐫 = 0 in domain R  This equation is called the
Helmholtz equation.
𝜕Ψ(𝐫)
𝑘𝑖 + ℎ𝑖 Ψ(𝐫) = 0 on boundary Si  In general, it is a PDE in the three
𝜕𝑛𝑖
(Homogeneous spatial variables.
Boundary Conditions)

The Helmholtz equation can be solved by Separation of Variables,

provided that its separation into a set of ODEs is possible.

A simple separation of the Helmholtz equation into ordinary differential

equations is possible in 11 orthogonal coordinate systems.
Separation of Variables: Generalization
𝛻 2 Ψ 𝐫 + 𝜆2 Ψ 𝐫 = 0 in domain R
𝜕Ψ(𝐫) Helmholtz equation
𝑘𝑖 + ℎ𝑖 Ψ(𝐫) = 0 on boundary Si
𝜕𝑛𝑖
The Helmholtz equation can be separated into ordinary differential equations
in 11 orthogonal coordinate systems.
Separation of Variables: Generalization
𝛻 2 Ψ 𝐫 + 𝜆2 Ψ 𝐫 = 0 in domain R
Helmholtz
𝜕Ψ(𝐫) equation
𝑘𝑖 + ℎ𝑖 Ψ(𝐫) = 0 on boundary Si
𝜕𝑛𝑖
The above system has nontrivial solutions only for certain values of the
separation constant λ = λm, called eigenvalues.

The corresponding nontrivial solutions are called eigenfunctions: Ψ 𝜆𝑚 , 𝐫 = Ψ𝑚 (𝐫)

Assuming the eigenfunctions and the eigenvalues λm are determined, the

complete solution of the temperature function T(r, t) can be obtained as:
∞ The summation is taken over all
−𝛼𝜆2𝑚 𝑡 discrete spectrum of eigenvalues λm
𝑇 𝐫, 𝑡 = 𝐶𝑚 Ψ𝑚 𝐫 𝑒
𝑚=1 for the given problem.

Note that for three-dimensional problems (in finite regions) the summation
in above expression for temperature is a triple infinite series.
Separation of Variables: Generalization
∞
2 The solution contains the
𝑇 𝐫, 𝑡 = 𝐶𝑚 Ψ𝑚 𝐫 𝑒 −𝛼𝜆𝑚 𝑡
unknown coefficients Cm.
𝑚=1

The above solution should satisfy the initial condition of the problem: 𝑇 𝐫, 𝑡 = 0 = 𝐹(𝐫)
∞

Therefore, by substituting t = 0, 𝐹 𝐫 = 𝐶𝑚 Ψ𝑚 𝐫 This is Fourier series.

𝑚=1

If the eigenfunctions Ψ𝑚 𝐫 constitute an orthogonal set in the region

considered, the unknown coefficients Cm are determined by making use of the
orthogonality property of eigenfunctions Ψ𝑚 𝐫 :

Ψ𝑚 𝐫 Ψ𝑛 𝐫 𝑑𝐫 = 0, for 𝑚 ≠ 𝑛
To determine Cm, 𝑅

∞
 Multiply both sides of F 𝐫 = 𝑚=1 𝐶𝑚 Ψ𝑚 𝐫 by Ψ𝑚 𝐫

 Integrate it over the region and make use of the orthogonality condition
Separation of Variables: Generalization
Ψ𝑚 𝐫 𝐹 𝐫 𝑑𝐫 where N is called the
𝑅 2
𝐶𝑚 = 𝑁= Ψ𝑚 𝐫 𝑑𝐫 norm of the eigenfunction
𝑁
𝑅

Having determined the coefficients Cm, the complete solution of the homogeneous boundary-value problem
of heat conduction equation is given in the form
∞ ∞ 2
2 𝑚=1 Ψ𝑚 𝐫 𝑒 −𝛼𝜆𝑚 𝑡
𝑇 𝐫, 𝑡 = 𝐶𝑚 Ψ𝑚 𝐫 𝑒 −𝛼𝜆𝑚 𝑡 ⇒ 𝑇 𝐫, 𝑡 = Ψ𝑚 𝐫 𝐹 𝐫 𝑑𝐫
𝑚=1
𝑁
𝑅
Sometimes the eigenfunctions are so adjusted that the norm becomes unity.
This is done if we define the normalized eigenfunctions 𝐾 𝜆𝑚 , 𝐫 as: Ψ𝑚 𝐫
𝐾 𝜆𝑚 , 𝐫 =
𝑁
The solution then becomes:
∞
2
𝑇 𝐫, 𝑡 = 𝐾 𝜆𝑚 , 𝐫 𝑒 −𝛼𝜆𝑚 𝑡 𝐾 𝜆𝑚 , 𝐫 𝐹 𝐫 𝑑𝐫
𝑚=1 𝑅

where the summation is taken over all eigenvalues λm of the problem.

1. Uniqueness of Solution of Heat Conduction Problems

2. Introduction to Separation of Variables

“Heat Conduction” by Sadik Kakac, Yaman Yener, Carolina P. Naveira-Cotta,
5th Ed, CRC Press

“Heat Conduction” by David W. Hahn, M. Necati Ozisik,

3rd Ed, John Wiley and Sons, Inc

“Heat Conduction” by M. Necati Ozisik, 2nd Ed, John Wiley

and Sons, Inc

“Heat Conduction” by Latif Jiji, 3rd Ed, Springer

Module – 3 : Multi-dimensional Steady-State Heat Conduction Problems:
Solution by Separation of Variables
Lecture – 14 : Problems on Rectangular Coordinate System
1. Separation of Variables: Rectangular Coordinate System
Separation of Variables
 The method of separation of variables has been widely used in the solution of heat
conduction problems.

 Homogeneous problems (homogeneous PDE, homogeneous BC) are readily handled with
this method.

 The multidimensional steady-state heat conduction problems with no heat generation can
also be solved with this method if only one of the boundary conditions is nonhomogeneous

 Problems involving more than one nonhomogeneous boundary condition can be

split up into simpler problems using the principle of superposition.
Separation of Variables: Requirements
Following are the requirements for solution of a partial differential equation (PDE), specifically the heat
equation, by the method of separation of variables:
1. Homogeneous PDE

2. For steady-state problems, all homogeneous boundary conditions with the

exception of a single nonhomogeneous boundary condition

3. For transient problems, all homogeneous boundary conditions and a

nonhomogeneous initial condition

4. Linear PDE and linear boundary conditions

When the above conditions are not satisfied, we will try the following to realize the
necessary conditions:
 The principle of superposition
 Shifting of the temperature scale
 Making use of symmetry
Separation of Variables: Example
Consider a solid bar of rectangular cross section. Three sides are The two-dimensional steady-state heat
maintained at a constant temperature T1, while the temperature of conduction problem for T(x, y):
fourth side is given as f1(x). Assume constant thermo-physical
properties and there is no internal heat sources or sinks. There 𝜕2𝑇 𝜕2𝑇
+ =0 0 < 𝑥 < 𝐿, 0<𝑦<𝑊
are no temperature gradients in the z-direction (i.e., either it is 𝜕𝑥 2 𝜕𝑦 2
very long in the z-direction or its surfaces perpendicular to the Laplace equation in two dimensions.
z-direction at the two ends are perfectly insulated).
𝑦 𝐁𝐂𝟏: 𝑇 𝑥 = 0, 𝑦 = 𝑇1 𝐁𝐂𝟐: 𝑇 𝑥 = 𝐿, 𝑦 = 𝑇1
∞
𝐁𝐂𝟑: 𝑇 𝑥, 𝑦 = 0 = 𝑇1 𝐁𝐂𝟒: 𝑇 𝑥, 𝑦 = 𝑊 = 𝑓1 (𝑥)

𝑊 𝑇 = 𝑇1 𝑇 = 𝑇1

𝑥
∞
𝐿 𝐿
𝑇 = 𝑇1

5
Separation of Variables: Example
For notational simplicity: 𝑇 𝑥, 𝑦 ≡ 𝑇 The PDE is homogeneous, but all
the boundary conditions are non-
𝜕2𝑇 𝜕2𝑇 homogenous.
+ =0 0 < 𝑥 < 𝐿, 0<𝑦<𝑊
𝜕𝑥 2 𝜕𝑦 2
Requirements for Separation of
𝐁𝐂𝟏: 𝑇 𝑥 = 0, 𝑦 = 𝑇1 𝐁𝐂𝟐: 𝑇 𝑥 = 𝐿, 𝑦 = 𝑇1 Variables are not met.

𝐁𝐂𝟑: 𝑇 𝑥, 𝑦 = 0 = 𝑇1 𝐁𝐂𝟒: 𝑇 𝑥, 𝑦 = 𝑊 = 𝑓1 (𝑥)

Three sides are kept at the same temperature T1. This type of non-homogeneity is readily
removed by linearly shifting the temperature scale, namely defining a new temperature:

𝜃 𝑥, 𝑦 = 𝑇 𝑥, 𝑦 − 𝑇1

𝜕2𝜃 𝜕2𝜃 Now all the

+ =0 0 < 𝑥 < 𝐿, 0<𝑦<𝑊 requirements for
𝜕𝑥 2 𝜕𝑦 2
separation of
𝐁𝐂𝟏: 𝜃 𝑥 = 0, 𝑦 = 0 𝐁𝐂𝟐: 𝜃 𝑥 = 𝐿, 𝑦 = 0 variables are met.
𝐁𝐂𝟑: 𝜃 𝑥, 𝑦 = 0 = 0 𝐁𝐂𝟒: 𝜃 𝑥, 𝑦 = 𝑊 = 𝑓1 (𝑥) − 𝑇1 ≡ 𝐹(𝑥)
 Separation of Variables: Example
𝜕2𝜃 𝜕2𝜃 Try the product solution: 𝜃 𝑥, 𝑦 = 𝑋 𝑥 𝑌(𝑦)
+ =0 0 < 𝑥 < 𝐿, 0<𝑦<𝑊
𝜕𝑥 2 𝜕𝑦 2 Substituting this into the PDE and
𝐁𝐂𝟏: 𝜃 𝑥 = 0, 𝑦 = 0 dividing both sides by XY yields:
𝐁𝐂𝟐: 𝜃 𝑥 = 𝐿, 𝑦 = 0
𝐁𝐂𝟑: 𝜃 𝑥, 𝑦 = 0 = 0 𝐁𝐂𝟒: 𝜃 𝑥, 𝑦 = 𝑊 = 𝑓1 (𝑥) − 𝑇1 ≡ 𝐹(𝑥) 1 𝑑 2 𝑋(𝑥) 1 𝑑2 𝑌(𝑦)
+ =0
𝑋(𝑥) 𝑑𝑥 2 𝑌(𝑦) 𝑑𝑦 2
Selection Rule for the Separation Constant (λ):
1 𝑑 2 𝑌(𝑦) 1 𝑑2 𝑋 𝑥
The sign of λ2 is chosen so that a Sturm–Liouville problem is ⇒ 2
=− 2
= ±𝜆2
obtained in the direction with the homogeneous boundary conditions. 𝑌(𝑦) 𝑑𝑦 𝑋 𝑥 𝑑𝑥
function of y only function of x only
Recall that the Sturm–Liouville ODE has solutions in terms of
orthogonal eigenfunctions such as sin, cos, and the Bessel functions. λ2 = separation constant
The hyperbolic functions and modified Bessel functions, on the other
hand, will not be orthogonal functions.

For the given problem, the x-direction has the homogeneous boundary
conditions. Therefore, we want the x–direction function ODE to be the
Sturm–Liouville system.
 Separation of Variables: Example
2 2
1 𝑑 2 𝑌(𝑦) 1 𝑑2 𝑋 𝑥
𝜕 𝑇 𝜕 𝑇 ⇒ 2
=− 2
= ±𝜆2
+ =0 0 < 𝑥 < 𝐿, 0<𝑦<𝑊 𝑌(𝑦) 𝑑𝑦 𝑋 𝑥 𝑑𝑥
𝜕𝑥 2 𝜕𝑦 2
𝐁𝐂𝟏: 𝜃 𝑥 = 0, 𝑦 = 0 𝐁𝐂𝟐: 𝜃 𝑥 = 𝐿, 𝑦 = 0 What about solutions for λ2 = 0?
𝐁𝐂𝟑: 𝜃 𝑥, 𝑦 = 0 = 0 𝐁𝐂𝟒: 𝜃 𝑥, 𝑦 = 𝑊 = 𝑓1 (𝑥) − 𝑇1 ≡ 𝐹(𝑥) 𝑌 𝑦 = 𝐴2 𝑦 + 𝐵2
𝑋 𝑥 = 𝐴1 𝑥 + 𝐵1
⇒ 𝜃 𝑥, 𝑦 = 𝑋 𝑥 𝑌 𝑦 = (𝐴1 𝑥 + 𝐵1 )(𝐴2 𝑦 + 𝐵2 )

Now, 𝑌 𝑦 = 𝐴2 𝑦 + 𝐵2 ≠ 0, as it would lead to

 For time-dependent problems, the solutions for
λ = 0 are usually inconsistent with transient 𝜃 𝑥, 𝑦 = 0
behavior. which does not satisfy BC4

 For steady–state problems, if it is determined

that the eigenvalue λn can have a value of zero,
then the unique solution for λ = 0 will have to be
included into the analysis.
 Separation of Variables: Example
The x-direction is the 𝑑2 𝑋
1 𝑑 2 𝑌(𝑦) 1 𝑑2 𝑋 𝑥 2
+ 𝜆2 𝑋 = 0
2 homogeneous direction 𝑑𝑥
= − = +𝜆
𝑌(𝑦) 𝑑𝑦 2 𝑋 𝑥 𝑑𝑥 2 Sturm–Liouville 𝐁𝐂𝟏: 𝑋 𝑥 = 0 = 0
system
𝐁𝐂𝟏: 𝜃 𝑥 = 0, 𝑦 = 0 𝐁𝐂𝟐: 𝜃 𝑥 = 𝐿, 𝑦 = 0
𝐁𝐂𝟐: 𝑋 𝑥 = 𝐿 = 0
𝐁𝐂𝟑: 𝜃 𝑥, 𝑦 = 0 = 0 𝐁𝐂𝟒: 𝜃 𝑥, 𝑦 = 𝑊 = 𝑓1 (𝑥) − 𝑇1 ≡ 𝐹(𝑥)
𝑑2 𝑌
2
− 𝜆2 𝑌 = 0
𝑑𝑦

Separation of Boundary Conditions: 𝐁𝐂𝟑: 𝑌 𝑦 = 0 = 0

Use of 𝜃 𝑥, 𝑦 = 𝑋 𝑥 𝑌(𝑦) reduces the two-dimensional homogeneous

boundary conditions of the problem to one-dimensional conditions.

For example, 𝐁𝐂𝟏: 𝜃 𝑥 = 0, 𝑦 = 0 Note: BC1, BC2, BC3

consider: are separable. But non-
⇒𝑋 𝑥=0 𝑌 𝑦 =0
homogeneous BC4 is
Since Y(y) is arbitrary, X(x = 0) = 0 non-separable.
Separation of Variables: Example 𝑑2 𝑌
𝑑2 𝑋 − 𝜆2 𝑌 = 0
1 𝑑 2 𝑌(𝑦) 1 𝑑2 𝑋 𝑥 2 Sturm–Liouville (B) 𝑑𝑦 2
= − = +𝜆 2 + 𝜆 𝑋 = 0
𝑌(𝑦) 𝑑𝑦 2 𝑋 𝑥 𝑑𝑥 2 𝑑𝑥 2 system
(A) 𝐁𝐂𝟑: 𝑌 𝑦 = 0 = 0
𝐁𝐂𝟏: 𝑋 𝑥 = 0 = 0

𝐁𝐂𝟐: 𝑋 𝑥 = 𝐿 = 0 Also, negative integers do

General solution for Eq (A): not produce any new result.
If λ=5π/L or λ=-5π/L, the results
𝑋 𝑥 = 𝐶1 cos(𝜆𝑥) + 𝐶2 sin(𝜆𝑥) Boundary condition BC1 yields C1 = 0
differ by only a minus sign,
Boundary condition BC2 yields: 𝑋 𝑥 = 𝐿 = 0 = 𝐶2 sin(𝜆𝐿) which may be absorbed by the
arbitrary constant.
Now C2 = 0 gives the trivial solution X(x) = 0; hence we
define the eigenvalues from:
𝑛𝜋
sin 𝜆𝐿 = 0 ⇒ 𝜆𝐿 = 𝑛𝜋 ⇒ 𝜆𝑛 = , 𝑛 = 0,1,2,3, … .
𝐿
This yields the desired Note that n = 0 will be
𝑋𝑛 𝑥 = 𝐶2 sin(𝜆𝑛 𝑥) excluded because it
orthogonal eigenfunctions:
leads to T(x, y) = 0.
 Separation of Variables: Example
𝑑2 𝑌 This ODE (Eq. B) yields the solution which may be written in two equivalent forms:
2
− 𝜆2 𝑌 = 0
𝑑𝑦 𝑌 𝑦 = 𝐶3 𝑒 𝜆𝑦 + 𝐶4 𝑒 −𝜆𝑦 General Rule to Simplify Algebra:
(B)
 For problems on finite domain, the
𝐁𝐂𝟑: 𝑌 𝑦 = 0 = 0 hyperbolic solutions may be used.
𝑌 𝑦 = 𝐶3 cosh(𝜆𝑦) + 𝐶4 sinh(𝜆𝑦)
 For problems on semi-infinite domain,
the exponential form may be used.

Use boundary condition (BC3): 𝑌 𝑦 = 0 = 0 = 𝐶3 cosh(0) + 𝐶4 sinh(0)

⟹ 𝐶3 = 0

At this point, the differential equation and all the 3 homogeneous boundary conditions
have been satisfied. We obtain the product solution of the separated functions:

𝜃 𝑥, 𝑦 = 𝑋 𝑥 𝑌 𝑦 = 𝐶2 sin(𝜆𝑛 𝑥)𝐶4 sinh(𝜆𝑛 𝑦)

⇒ 𝜃𝑛 𝑥, 𝑦 = 𝐶𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦) This yields a solution for the original

PDE for each eigenvalue.
(Here, 𝐶𝑛 = 𝐶2 𝐶4 )
Separation of Variables: Example
The product solution that satisfy the differential equation 𝜃𝑛 𝑥, 𝑦 = 𝐶𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
and all the 3 homogeneous boundary conditions :
∞ Note that n = 0 is not
The most general solution can be written included as the
𝜃 𝑥, 𝑦 = 𝐶𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
as the sum of all solutions, as given by: corresponding
𝑛=1
eigenfunction for λ0 is
The only step that remains is to define the constant Cn. We now use the
zero and this leads to
nonhomogeneous boundary condition BC4.
zero temperature.
𝐁𝐂𝟒: 𝜃 𝑥, 𝑦 = 𝑊 = 𝑓1 (𝑥) − 𝑇1 ≡ 𝐹(𝑥)
∞

𝜃 𝑥, 𝑦 = 𝑊 = 𝐹 𝑥 = 𝐶𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑊)
𝑛=1
∞ where
⇒ 𝜃 𝑥, 𝑦 = 𝑊 = 𝐹(𝑥) = 𝑎𝑛 sin 𝜆𝑛 𝑥 , 𝑎𝑛 = 𝐶𝑛 sinh(𝜆𝑛 𝑊)
𝑛=1

This is Fourier sine series expansion of F(x) in terms of orthogonal functions over
the interval 0 ≤ x ≤ L. Note that the weighting function is unity here.
 Separation of Variables: Example Use orthogonal property to
Find the constant Cn ∞ eliminate all terms of the series
𝜃 𝑥, 𝑦 = 𝑊 = 𝐹(𝑥) = 𝑎𝑛 sin 𝜆𝑛 𝑥 , 𝑎𝑛 = 𝐶𝑛 sinh(𝜆𝑛 𝑊) except for the term with m = n.
𝑛=0
𝐿 𝐿 ∞ 𝐿
2
𝐿
⇒ 𝐹 𝑥 sin 𝜆𝑚 𝑥 𝑑𝑥 = 𝑎𝑛 sin 𝜆𝑛 𝑥 sin 𝜆𝑚 𝑥 𝑑𝑥 = 𝑎𝑚 sin(𝜆𝑛 𝑥) 𝑑𝑥 = 𝑎𝑚 𝑚≠𝑛
𝑥=0 𝑥=0 𝑛=1 𝑥=0 2
𝐿 𝐿
2 2 We have used
⇒ 𝑎𝑚 = 𝐹(𝑥) sin 𝜆𝑚 𝑥 𝑑𝑥 ⇒ 𝐶𝑚 = 𝐹(𝑥) sin 𝜆𝑚 𝑥 𝑑𝑥
𝐿 𝑥=0 𝐿 sinh(𝜆𝑚 𝑊) 𝑥=0 𝑎𝑛 = 𝐶𝑛 sinh(𝜆𝑛 𝑊)
𝐿
2
Replacing arbitrary index m with n: ⇒ 𝐶𝑛 = 𝐹(𝑥) sin 𝜆𝑛 𝑥 𝑑𝑥
𝐿 sinh(𝜆𝑛 𝑊) 𝑥=0
∞
Substituting Cn: 𝜃 𝑥, 𝑦 = 𝐶𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
𝑛=1

∞ 𝐿
2 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
𝜃 𝑥, 𝑦 = 𝐹(𝑥) sin 𝜆𝑛 𝑥 𝑑𝑥
𝐿 sinh(𝜆𝑛 𝑊) 𝑥=0
𝑛=1
 Separation of Variables: Example
We have obtained:
∞
2 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦) 𝐿 𝑛𝜋
𝜃 𝑥, 𝑦 = 𝐹(𝑥) sin 𝜆𝑛 𝑥 𝑑𝑥 where, 𝜆𝑛 = , 𝑛 = 1,2,3, … .
𝐿 sinh(𝜆𝑛 𝑊) 𝐿
𝑛=1 𝑥=0
∞ 𝐿
2 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
𝐒𝐮𝐛𝐬𝐭𝐢𝐭𝐮𝐭𝐞: 𝑓1 𝑥 − 𝑇1 ≡ 𝐹 𝑥 : ⇒ 𝜃 𝑥, 𝑦 = 𝑓1 (𝑥) − 𝑇1 sin 𝜆𝑛 𝑥 𝑑𝑥
𝐿 sinh(𝜆𝑛 𝑊) 𝑥=0
𝑛=1
Now, if f1(x) = T0 = constant:
∞ 𝐿
2 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
𝜃 𝑥, 𝑦 = 𝑇0 − 𝑇1 sin 𝜆𝑛 𝑥 𝑑𝑥
𝑇 = 𝑇0
𝐿 sinh(𝜆𝑛 𝑊) 𝑥=0
𝑛=1
𝑊 ∞ 𝐿
𝑇 − 𝑇1 2 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
𝑇 = 𝑇1 𝑇 = 𝑇1 ⇒ = sin 𝜆𝑛 𝑥 𝑑𝑥
𝑇0 − 𝑇1 𝐿 sinh(𝜆𝑛 𝑊) 𝑥=0
𝑛=1

𝐿
𝑇 = 𝑇1
 Separation of Variables: Example
∞ 𝐿
𝑇 − 𝑇1 2 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦) 𝑛𝜋
If f1(x) = T0 = constant: ⇒ = sin 𝜆𝑛 𝑥 𝑑𝑥 𝜆𝑛 = , 𝑛 = 1,2,3, … .
𝑇0 − 𝑇1 𝐿 sinh(𝜆𝑛 𝑊) 𝑥=0 𝐿
𝑛=1

𝑛𝜋 𝐿
𝐿 𝐿
𝑛𝜋 cos 𝑥 𝐿
sin 𝜆𝑛 𝑥 𝑑𝑥 = sin 𝑥 𝑑𝑥 = − 𝑛𝜋 𝐿 =− cos 𝑛𝜋 − cos(0)
𝑥=0 𝑥=0 𝐿 𝑛𝜋
𝐿 𝑥=0
𝐿
= 1 − (−1)𝑛
𝑛𝜋
Therefore,
𝑇 = 𝑇0 ∞
𝑊 𝑇 − 𝑇1 2 𝐿 1 − (−1)𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
=
𝑇0 − 𝑇1 𝐿 𝑛𝜋 sinh(𝜆𝑛 𝑊)
𝑇 = 𝑇1 𝑇 = 𝑇1 𝑛=1
∞
𝑇 − 𝑇1 2 1 − (−1)𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
⇒ =
𝑇0 − 𝑇1 𝜋 𝑛 sinh(𝜆𝑛 𝑊)
𝑛=1
𝐿
𝑇 = 𝑇1
 Separation of Variables: Example
The temperature is maintained at 0°C along the three surfaces of the rectangular bar shown in the Figure,
while the fourth surface at y = W is held at 100°C. If L = 2W, calculate the centerline temperature under steady-
state conditions.
∞
𝑇(𝑥, 𝑦) − 𝑇1 2 1 − (−1)𝑛 sin(𝜆𝑛 𝑥) sinh(𝜆𝑛 𝑦)
=
𝑇0 − 𝑇1 𝜋 𝑛 sinh(𝜆𝑛 𝑊)
𝑇 = 𝑇0 𝑛=1
𝐿 𝑊 𝐿
𝑊 𝐿 = 2𝑊, 𝑥= , and 𝑦= =
T1 = 0C, T0 = 100C 2 2 4
𝑇 = 𝑇1 𝑇 = 𝑇1 ∞
𝐿 𝐿 2 × 100 1 − (−1)𝑛 sin(𝑛𝜋/2) sinh(𝑛𝜋/4)
⇒𝑇 , =
2 4 𝜋 𝑛 sinh(𝑛𝜋/2)
𝑛=1
𝐿
𝑇 = 𝑇1 𝐿 𝐿
⇒𝑇 , = 48.061 − 3.987 + 0.502 − … .
2 4
𝐿 𝐿
⇒𝑇 , = 44.576 C
2 4
1. Separation of Variables: Rectangular Coordinate System
“Heat Conduction” by Sadik Kakac, Yaman Yener, Carolina P. Naveira-Cotta,
5th Ed, CRC Press

“Heat Conduction” by David W. Hahn, M. Necati Ozisik,

3rd Ed, John Wiley and Sons, Inc

“Heat Conduction” by M. Necati Ozisik, 2nd Ed, John Wiley

and Sons, Inc

“Heat Conduction” by Latif Jiji, 3rd Ed, Springer

Module – 3 : Multi-dimensional Steady-State Heat Conduction Problems:
Solution by Separation of Variables
Lecture – 15 : Problems on Rectangular Coordinate System (Cont’d)
1. Principle of Superposition

2. Separation of Variables: Non-homogeneous Boundary Conditions

3. Separation of Variables: Non-homogeneous Differential Equation

 Separation of Variables: Superposition
The method of separation of variables is readily applicable to two dimensional steady-state
linear problems consisting of a homogeneous differential equation subject to three
homogeneous and one nonhomogeneous boundary conditions.

Nonhomogeneities in a heat conduction problem may result due to

 Nonhomogeneous differential equation, and/or
 Nonhomogeneous boundary conditions

When the conditions for SOV are not satisfied, we will try the following to
realize the necessary conditions:
 The principle of superposition
 Shifting of the temperature scale
 Making use of symmetry
Fourier Series in Finite Interval (0, L)

Reference: “Heat Conduction” by Sadik Kakac, Yaman

Yener, Carolina P. Naveira-Cotta, 5th Ed, CRC Press
Fourier Series in Finite Interval (0, L)

Reference: “Heat Conduction” by Sadik Kakac, Yaman

Yener, Carolina P. Naveira-Cotta, 5th Ed, CRC Press
Fourier Series in Finite Interval (0, L)

Reference: “Heat Conduction” by Sadik Kakac, Yaman

Yener, Carolina P. Naveira-Cotta, 5th Ed, CRC Press
 Separation of Variables: Example-2 Non-homogeneous BC

Problem Formulation: 𝜕2𝑇 𝜕2𝑇

+ =0 𝐁𝐂𝟏: 𝑇 0, 𝑦 = 𝑓(𝑦) 𝐁𝐂𝟐: 𝑇 𝑥 → ∞, 𝑦 = 𝑇∞
𝑦 𝜕𝑥 2 𝜕𝑦 2
ℎ, 𝑇∞ 𝜕𝑇 𝑥, 0 𝜕𝑇 𝑥, 𝑏
𝐁𝐂𝟑: =0 𝐁𝐂𝟒: −𝑘 = ℎ 𝑇 𝑥, 𝑏 − 𝑇∞
𝜕𝑦 𝜕𝑦
𝑓(𝑦)
𝑇(𝑥, 𝑦) 𝑏
symmetry in y-direction
𝑥 ∞ Conditions for SOV are not satisfied
𝑞=0 𝑏
k=constant Define: 𝜃 𝑥, 𝑦 = 𝑇 𝑥, 𝑦 − 𝑇∞
𝐁𝐂𝟏: 𝜃 0, 𝑦 = 𝑓 𝑦 − 𝑇∞ ≡ 𝐹(𝑦)
ℎ, 𝑇∞ 2
𝜕 𝜃 𝜕 𝜃 2
+ =0
Consider the long (semi-infinite) two- 𝜕𝑥 2 𝜕𝑦 2 𝐁𝐂𝟐: 𝜃 𝑥 → ∞, 𝑦 = 0
dimensional fin which exchanges heat by
𝜕𝜃 𝑥, 0
convection with a surrounding fluid of constant 𝐁𝐂𝟑: =0
temperature T∞. Determine steady-state T(x, y). 𝜕𝑦
Assume: 𝜕𝜃 𝑥, 𝑏
𝐁𝐂𝟒: −𝑘 = ℎ𝜃(𝑥, 𝑏)
 Constant heat transfer coefficient h 𝜕𝑦
 Constant thermo-physical properties Conditions for SOV are satisfied
 Separation of Variables: Example-2 Note: 𝜃 𝑥, 𝑦 = 𝑇 𝑥, 𝑦 − 𝑇∞
2 2
𝜕 𝜃 𝜕 𝜃
Problem Formulation: + =0 𝐁𝐂𝟏: 𝜃 0, 𝑦 = 𝑓 𝑦 − 𝑇∞ ≡ 𝐹(𝑦)
𝑦 𝜕𝑥 2 𝜕𝑦 2
ℎ, 𝑇∞ 𝐁𝐂𝟐: 𝜃 𝑥 → ∞, 𝑦 = 0
𝜕𝜃 𝑥, 0 𝜕𝜃 𝑥, 𝑏
𝑓(𝑦) 𝐁𝐂𝟑: = 0 𝐁𝐂𝟒: −𝑘 = ℎ𝜃(𝑥, 𝑏)
𝑇(𝑥, 𝑦) 𝑏 𝜕𝑦 𝜕𝑦
𝑥 ∞ Try the product solution: 𝜃 𝑥, 𝑦 = 𝑋 𝑥 𝑌(𝑦)
𝑞=0 𝑏
k=constant Substituting this into the PDE and dividing both sides by XY yields:
ℎ, 𝑇∞ 1 𝑑 2 𝑋(𝑥) 1 𝑑2 𝑌(𝑦)
+ =0
𝑋(𝑥) 𝑑𝑥 2 𝑌(𝑦) 𝑑𝑦 2

 Homogeneous direction: y 1 𝑑2𝑋 𝑥 1 𝑑2 𝑌(𝑦)

⇒ 2
=− 2
= ±𝜆2
 We should get Sturm-Liouville 𝑋 𝑥 𝑑𝑥 𝑌(𝑦) 𝑑𝑦
problem in y-direction. function of x only function of y only
 Select plus sign for λ2
λ2 = separation constant.
What should be it’s sign?
 Separation of Variables: Example-2
1 𝑑2𝑋 𝑥 1 𝑑2 𝑌(𝑦) 𝐁𝐂𝟏: 𝜃 0, 𝑦 = 𝑓 𝑦 − 𝑇∞ ≡ 𝐹(𝑦) 𝐁𝐂𝟐: 𝜃 𝑥 → ∞, 𝑦 = 0
2
= − = 𝜆
𝑋 𝑥 𝑑𝑥 2 𝑌(𝑦) 𝑑𝑦 2 𝜕𝜃 𝑥, 0 𝜕𝜃 𝑥, 𝑏
𝐁𝐂𝟑: =0 𝐁𝐂𝟒: −𝑘 = ℎ𝜃(𝑥, 𝑏)
Get two separated problems. 𝜕𝑦 𝜕𝑦
General Rule to Simplify Algebra:
𝑑2 𝑋 The y-direction is
− 𝜆2
𝑋=0 𝑑2 𝑌 2  For problems on finite domain, the
𝑑𝑥 2 the homogeneous + 𝜆 𝑌=0 hyperbolic solutions may be used.
direction 𝑑𝑦 2
𝐁𝐂𝟐: 𝑋 𝑥 → ∞ = 0 𝑑𝑌(0)  For problems on semi-infinite domain,
Sturm–Liouville 𝐁𝐂𝟑: =0 the exponential form may be used.
𝑑𝑦
system
Solutions: 𝑑𝑌 𝑏
𝐁𝐂𝟒: 𝑘 + ℎ𝑌 𝑏 = 0
𝑑𝑦
𝑋 𝑥 = 𝐴1 𝑒 −𝜆𝑥 + 𝐴2 𝑒 𝜆𝑥 BC2 requires: A2 = 0

𝑌 𝑦 = 𝐵1 cos 𝜆𝑦 + 𝐵2 sin(𝜆𝑦) BC3 requires B2 = 0

Now B1 = 0 gives the trivial solution Y(y) = 0;

hence we obtain eigenfunctions as: 𝜙𝑛 𝑦 = cos 𝜆𝑛 𝑦
Eigenvalues may be obtained by satisfying BC4
Separation of Variables: Example-2 𝑑𝑌 𝑏
𝑌 𝑦 = 𝐵1 cos 𝜆𝑛 𝑦 𝐁𝐂𝟒: 𝑘 + ℎ𝑌 𝑏 = 0
Solutions: 𝑋 𝑥 = 𝐴1 𝑒 −𝜆𝑥 𝑑𝑦

Eigenfunctions: 𝜙𝑛 𝑦 = cos 𝜆𝑛 𝑦 Eigenvalues may be obtained by satisfying BC4

Use BC4: −𝑘𝐵1 𝜆𝑛 sin 𝜆𝑛 𝑦 + ℎ𝐵1 cos 𝜆𝑛 𝑦 = 0 ⇒ −𝑘𝜆𝑛 sin 𝜆𝑛 𝑏 + ℎ cos 𝜆𝑛 𝑏 = 0

(at y = b)
⇒ 𝑘𝜆𝑛 sin 𝜆𝑛 𝑏 = ℎ cos(𝜆𝑛 𝑏)
ℎ Eigenvalues may be obtained from the roots of this transcendental
⇒ 𝜆𝑛 tan 𝜆𝑛 𝑏 = , 𝑛 = 1,2,3, … . equation. Graphical or numerical solution is required to find λn.
𝑘

Multiply both sides by b:

ℎ𝑏 Graphical
⇒ 𝜆𝑛 𝑏 tan 𝜆𝑛 𝑏 = solution
𝑘
𝐵𝑖 𝜆𝑛 𝑏
⇒ tan 𝜆𝑛 𝑏 = 𝐨𝐫 cot 𝜆𝑛 𝑏 =
𝜆𝑛 𝑏 𝐵𝑖
Bi = Biot number based on
half-thickness of the plate
 𝒅𝟐 𝒚
𝟐
+ 𝝀𝟐 𝒚 𝒙 = 𝟎
𝒅𝒙
Separation of Variables: Example-2 𝜶𝟏 𝒚 𝟎 + 𝜷𝟏
𝒅𝒚(𝟎)
=𝟎
𝒅𝒙
Use of Table to find eigenvalues and eigenfunctions: 𝒅𝒚(𝑳)
𝜶𝟐 𝒚 𝑳 + 𝜷𝟐 =𝟎
𝒅𝒙

Boundary Boundary Characteristic Norm Eigenfunctions

Condition at x = 0 Condition at x = L functions, ϕn(x) 𝑳 are positive
𝑵𝒏 = 𝝓𝒏 (𝒙) 𝟐 𝒅𝒙 roots of
𝟎
Second kind Third kind cos(𝜆𝑛 𝑥) 1 𝜆 tan(𝜆𝐿) = 𝐻2
𝜆 𝐿 + sin(𝜆𝑛 𝐿 cos(𝜆𝑛 𝐿))
(𝛼1 = 0, 𝛽1 ≠ 0) (𝛼1 ≠ 0, 𝛽1 ≠ 0) 2𝜆𝑛 𝑛

𝑑2 𝑌 𝛼2 ℎ
2
+ 𝜆2 𝑌 = 0 Note that the domain of the current Here 𝐻2 = =
𝑑𝑦 problem is [0, b]. Set L = b 𝛽2 𝑘
𝑑𝑌(0)
𝐁𝐂𝟑: =0 Eigenfunctions: 𝜙𝑛 𝑦 = cos 𝜆𝑛 𝑦
𝑑𝑦
𝑑𝑌 𝑏 Eigenvalues are positive roots of:
𝐁𝐂𝟒: 𝑘 + ℎ𝑌 𝑏 = 0
𝑑𝑦 ℎ
𝜆𝑛 tan 𝜆𝑛 𝑏 = , 𝑛 = 1,2,3, … .
Current Sturm-Liouville Problem 𝑘
Separation of Variables: Example-2
Solutions: 𝑋 𝑥 = 𝐴1 𝑒 −𝜆𝑥 𝑌 𝑦 = 𝐵1 cos 𝜆𝑛 𝑦
∞
Using the principle of superposition, the
most general solution can be written as 𝜃 𝑥, 𝑦 = 𝐶𝑛 𝑒 −𝜆𝑛 𝑥 cos(𝜆𝑛 𝑦) where Cn = A1B1
the sum of all solutions, as given by: 𝑛=1 This solution satisfy the PDE and all the
3 homogeneous boundary conditions.
We now determine Cn such that the nonhomogeneous
boundary condition BC1 satisfy this solution:

𝐁𝐂𝟏: 𝜃 0, 𝑦 = 𝑓 𝑦 − 𝑇∞ ≡ 𝐹(𝑦)
∞

𝜃 𝑥 = 0, 𝑦 ≡ 𝐹 𝑦 = 𝐶𝑛 𝑒 −𝜆𝑛 (0) cos(𝜆𝑛 𝑦) (Put x = 0)

𝑛=1
∞

⇒ 𝐹(𝑦) = 𝐶𝑛 cos 𝜆𝑛 𝑦
𝑛=1
This is Fourier sine series expansion of F(y) in terms Use orthogonal property
of orthogonal functions over the interval 0 ≤ x ≤ L. to determine Cn
Note that the weighting function is unity here. (or directly use Table)
 Separation of Variables: Example-2
∞ Current domain: [0, b]
Find the constant Cn : 𝐹(𝑦) = 𝐶𝑛 cos 𝜆𝑛 𝑦 Put L = b
𝑛=1
𝑏
1 1
⇒ 𝐶𝑛 = 𝐹(𝑦) cos 𝜆𝑛 𝑦 𝑑𝑦 From Table, the Norm Nn = 𝜆 𝐿 + sin(𝜆𝑛 𝐿 cos(𝜆𝑛 𝐿))
𝑁𝑛 𝑥=0
2𝜆𝑛 𝑛
𝐿
2𝜆𝑛
⇒ 𝐶𝑛 = 𝑓 𝑦 − 𝑇∞ cos 𝜆𝑛 𝑦 𝑑𝑦 We have used: 𝑓 𝑦 − 𝑇∞ ≡ 𝐹(𝑦)
𝜆𝑛 𝑏 + sin(𝜆𝑛 𝑏 cos(𝜆𝑛 𝑏)) 𝑥=0

Substituting Cn in: 𝜃 𝑥, 𝑦 = 𝐶𝑛 𝑒 −𝜆𝑛𝑥 cos(𝜆𝑛 𝑦)

𝑛=1

⇒ 𝜃 𝑥, 𝑦 = 𝑇 𝑥, 𝑦 − 𝑇∞
∞ 𝑏
𝜆𝑛 𝑒 −𝜆𝑛 𝑥 cos(𝜆𝑛 𝑦)
=2 𝑓 𝑦′ − 𝑇∞ cos 𝜆𝑛 𝑦′ 𝑑𝑦′
𝜆𝑛 𝑏 + sin(𝜆𝑛 𝑏 cos(𝜆𝑛 𝑏)) 𝑥=0
𝑛=1

This represents the temperature distribution in the 2D fin.

SOV: Superposition: Nonhomogeneous BC
The method of Separation of Variables (SOV) can be applied to solve problems with
nonhomogeneous boundary conditions using the principle of superposition.

In this approach, a problem is decomposed into as many simpler problems as the number of
nonhomogeneous boundary conditions.

Each simple problem is assigned a single nonhomogeneous boundary condition such that when the
simpler problems are added, they satisfy the conditions on the original problem.

In solving the simpler problems, it may be more convenient to change the

location of the origin and/or the direction of coordinates. However, all
solutions must be transformed to a common origin and coordinates before
they are added together.
Superposition: Nonhomogeneous BC: Example
Example: 𝜕2𝑇 𝜕2𝑇
+ =0 This is a linear problem with
𝜕𝑥 2 𝜕𝑦 2 one homogeneous differential
𝑇 𝑥 = 0, 𝑦 = 𝑓4 𝑦 , 0<𝑦<𝑊 𝑇 𝑥 = 𝐿, 𝑦 = 𝑓3 𝑦 , 0<𝑦<𝑊 equation and four
nonhomogeneous boundary
𝑇 𝑥, 𝑦 = 0 = 𝑓4 𝑦 , 0<𝑥<𝐿 𝑇 𝑥, 𝑦 = 𝑊 = 𝑓1 𝑦 , 0<𝑥<𝐿 conditions.

≡ + + +

Since it is linear, by the principle of superposition, the solution of this problem can
be obtained as a sum of the four solutions.
𝑇 𝑥, 𝑦 = 𝑇1 𝑥, 𝑦 + 𝑇2 𝑥, 𝑦 + 𝑇3 𝑥, 𝑦 + 𝑇4 𝑥, 𝑦
Superposition: Nonhomogeneous BC: Example
Consider two-dimensional steady-state conduction in a plate of
dimension (L х W) with nonhomogeneous boundary conditions in all 4
ℎ, 𝑇∞ sides. Assume no heat generation and constant thermal conductivity.
Find steady-state T(x, y).
The steady-state heat equation without any heat 𝜕 2 𝑇 𝜕 2 𝑇
𝑞0′′ T=g(y) source for constant thermal conductivity: + =0
𝜕𝑥 2 𝜕𝑦 2
Boundary conditions:
𝜕𝑇 0, 𝑦 𝑇 𝐿, 𝑦 = 𝑔(𝑦)
−𝑘 = 𝑞0′′
T=f(x) 𝜕𝑥
𝜕𝑇 𝑥, 𝑊
𝑇 𝑥, 0 = 𝑓(𝑥) −𝑘 = ℎ 𝑇 𝑥, 𝑊 − 𝑇∞
𝜕𝑦
Split the problem into 4 problems where each of the 4 problems contain only a
single nonhomogeneous boundary condition and can be readily solved by
separation of variables.
Assume that the solution T(x, y) is the sum of the solutions to four problems:
𝑇 𝑥, 𝑦 = 𝑇1 𝑥, 𝑦 + 𝑇2 𝑥, 𝑦 + 𝑇3 𝑥, 𝑦 + 𝑇4 𝑥, 𝑦
 Superposition: Nonhomogeneous BC: Example
𝑦 𝑦 𝑦 𝑦
ℎ, 𝑇∞
ℎ, 0 ℎ, 0 ℎ, 0 ℎ, 𝑇∞

𝑞0′′ T=g(y) ≡ + 𝑔(𝑦) + +

𝑥 𝑥 𝑥 𝑥
𝑓(𝑥)
T=f(x)
Four problems with one nonhomogeneous boundary condition
Note the insulated surfaces at x = 0
A problem with four
nonhomogeneous
boundary conditions
𝑇 𝑥, 𝑦 = 𝑇1 𝑥, 𝑦 + 𝑇2 𝑥, 𝑦 + 𝑇3 𝑥, 𝑦 + 𝑇4 𝑥, 𝑦
4
2
𝜕 𝑇 𝜕 𝑇 2 𝜕 2 𝑇𝑛 𝜕 2 𝑇𝑛
+ =0 + =0
𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥 2 𝜕𝑦 2
𝑛=1
 Superposition: Nonhomogeneous BC: Example
𝑦 𝑦 𝑦 𝑦
ℎ, 𝑇∞
ℎ, 0 ℎ, 0 ℎ, 0 ℎ, 𝑇∞

𝑞0′′ T=g(y) ≡ + 𝑔(𝑦) + +

𝑥 𝑥 𝑥 𝑥
𝑓(𝑥)
T=f(x)
𝑇 𝑥, 𝑦 = 𝑇1 𝑥, 𝑦 + 𝑇2 𝑥, 𝑦 + 𝑇3 𝑥, 𝑦 + 𝑇4 𝑥, 𝑦

𝜕𝑇 0, 𝑦 𝜕𝑇1 0, 𝑦 𝜕𝑇2 0, 𝑦 𝜕𝑇3 0, 𝑦 𝜕𝑇4 0, 𝑦

−𝑘 = 𝑞0′′ −𝑘 −𝑘 −𝑘 −𝑘 = 𝑞0′′
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥

We assign the nonhomogeneous 𝜕𝑇2 0, 𝑦

𝜕𝑇1 0, 𝑦
part of this equation to T1(x, y), −𝑘 = 𝑞0′′ −𝑘 =0
𝜕𝑥 𝜕𝑥
leaving homogeneous conditions for
the remaining three problems 𝜕𝑇3 0, 𝑦 𝜕𝑇4 0, 𝑦
−𝑘 =0 −𝑘 =0
(insulated surface – zero flux). Thus, 𝜕𝑥 𝜕𝑥
Superposition: Nonhomogeneous BC: Example
Similarly, the remaining three conditions are subdivided to give:

Note that each of

the four problems
has only one
nonhomogeneous
boundary condition
Handling Nonhomogeneous Differential Equation
The steady-state heat conduction with heat generation within the solid satisfies a Poisson’s
equation subject to prescribed boundary conditions on the surfaces:

2
𝑞
𝛻 𝑇(𝐫) + = 0 in region R, t > 0 The PDE is non-homogeneous
𝑘
(15.1)
𝜕𝑇 𝐫
𝑘𝑖 + ℎ𝑖 𝑇 𝐫 = 𝑓𝑖 on each boundary of R, 𝑖 = 1,2, … . , 𝑝
𝜕𝑛𝑖
where ∂/∂ni denotes differentiation along the outward-drawn normal to the bounding surface si.

Poisson’s equation can be transformed into Laplace’s equation by a change of dependent

variable as shown below:
Ψ 𝐫 = 𝑇 𝐫 − 𝜙(𝐫) (15.2)

where the function ϕ(r) is so chosen that when (Eq. 15.2) is substituted
into system (Eq. 15.1), Poisson’s equation reduces to Laplace’s equation

𝛻 2 𝑇(𝐫) = 0
subject to boundary conditions in terms of the new variable ψ(r).
 Superposition: Nonhomogeneous DE: Example
Consider the rectangular bar shown in cross section. Internal energy is generated in this bar at a constant rate
𝑞 per unit volume (W/m3). There are no temperature gradients in the z-direction. The thermal conductivity of the
material of the bar is constant. Determine the steady-state temperature distribution T(x, y) in the bar.
𝜕2𝑇 𝜕2𝑇 𝑞 𝜕𝑇 0, 𝑦
The governing differential equation: + + =0 𝐁𝐂𝟏: =0 𝐁𝐂𝟐: 𝑇 𝑎, 𝑦 = 0
𝜕𝑥 2 𝜕𝑦 2 𝑘 𝜕𝑥
Linear, non-homogeneous 𝜕𝑇 𝑥, 0 𝐁𝐂𝟒: 𝑇 𝑥, 𝑏 = 0
PDE 𝐁𝐂𝟑: =0
𝜕𝑦

Since the PDE is non-homogeneous, it is not

separable.

Assume the solution to be in the form:

𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑥)
or
𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑦)
Superposition: Nonhomogeneous DE: Example
𝜕2𝑇 𝜕2𝑇 𝑞 𝜕𝑇 0, 𝑦 𝜕𝑇 𝑥, 0
+ + =0 𝐁𝐂𝟏: =0 𝐁𝐂𝟐: 𝑇 𝑎, 𝑦 = 0 𝐁𝐂𝟑: =0 𝐁𝐂𝟒: 𝑇 𝑥, 𝑏 = 0
𝜕𝑥 2 𝜕𝑦 2 𝑘 𝜕𝑥 𝜕𝑦
𝜕 2 Ψ 𝑑2 𝜙 𝜕 2 Ψ 𝑞
Let us assume the solution to be in the form: 𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑥) + + + =0
𝜕𝑥 2 𝑑𝑥 2 𝜕𝑦 2 𝑘
Then the problem can be written as a superposition of the following two simpler problems
where the non-homogeneous term (heat generation term) is included in 1D problem.

𝑑2 𝜙 𝑞 𝜕2Ψ 𝜕2Ψ
+ =0 + =0
𝑑𝑥 2 𝑘 𝜕𝑥 2 𝜕𝑦 2
𝜕Ψ 0, 𝑦 𝜕𝑇 0, 𝑦 𝑑𝜙 0
𝑑𝜙 0 𝐁𝐂𝟏: = − =0−0=0 2D problem.
𝐁𝐂: =0 𝜕𝑥 𝜕𝑥 𝑑𝑥 Solve by
𝑑𝑥
𝐁𝐂𝟐: Ψ 𝑎, 𝑦 = 𝑇 𝑎, 𝑦 − 𝜙 𝑎 = 0 − 0 = 0 Separation
𝐁𝐂: 𝜙 𝑎 = 0 of Variables.
𝜕Ψ 𝑥, 0 𝜕𝑇 𝑥, 0 𝑑𝜙 𝑥
𝐁𝐂𝟑: = − =0−0=0
1D problem. 𝜕𝑦 𝜕𝑦 𝑑𝑦
Integrate twice and
𝐁𝐂𝟒: Ψ 𝑥, 𝑏 = 𝑇 𝑥, 𝑏 − 𝜙 𝑥 = 0 − 𝜙 𝑥 = −𝜙(𝑥)
solve.
Superposition: Nonhomogeneous DE: Example
𝑞𝑎2 𝑥 2
Solution for 1D problem: 𝜙 𝑥 = 1−
2𝑘 𝑎
∞
2𝑞 (−1)𝑛 cos(𝜆𝑛 𝑥) cosh(𝜆𝑛 𝑦) (2𝑛 + 1)𝜋
Solution for 2D problem: Ψ 𝑥, 𝑦 = − , 𝜆𝑛 = , 𝑛 = 1,2,3, … .
𝑎𝑘 𝜆3𝑛 cosh(𝜆𝑛 𝑏)) 2𝑎
𝑛=0

Final solution: The temperature distribution

T(x, y) in the rectangular bar is: 𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑥)

Heat source term 𝒒 Use

Constant 𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑥)
or How to
𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑦) assume the
form of
𝑞 = 𝑞(𝑥) 𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑥)
solution?
𝑞 = 𝑞(𝑦) 𝑇 𝑥, 𝑦 = Ψ 𝑥, 𝑦 + 𝜙(𝑦)
𝑞 = 𝑞(𝑥, 𝑦) In general, the problem cannot be
separated into simpler problems
23
1. Principle of Superposition

2. Separation of Variables: Non-homogeneous Boundary Conditions

3. Separation of Variables: Non-homogeneous Differential Equation

“Heat Conduction” by Sadik Kakac, Yaman Yener, Carolina P. Naveira-Cotta,
5th Ed, CRC Press

“Heat Conduction” by David W. Hahn, M. Necati Ozisik,

3rd Ed, John Wiley and Sons, Inc

“Heat Conduction” by M. Necati Ozisik, 2nd Ed, John Wiley

and Sons, Inc

“Heat Conduction” by Latif Jiji, 3rd Ed, Springer

Module – 3 : Multi-dimensional Steady-State Heat Conduction Problems:
Solution by Separation of Variables
Lecture – 16 : Problems on Cylindrical Coordinate System
1. 2D Heat Conduction Problems in Cylindrical Coordinate System T(r,ϕ,z):
 T(r, z)
 T(r, ϕ)
2. T(r, z) Problem: Separation of Variables: Example  T(ϕ, z) : Not Important

3. T(r, z) Problem: Compute Temperature at a Given Point in Cylinder

4. Another T(r, z) Problem with different BC: Separation of Variables

Separation of Variables: Cylindrical Coordinate System

𝑇(𝑟, 𝜙)
𝜙
𝑟

𝑧 𝑧

Heat conduction equation at steady-state

without any heat source (Laplace Equation):
𝑇(𝑟, 𝜙, 𝑧) 𝑇(𝑟, 𝑧)
𝛻 2 𝑇 𝑟, 𝜙, 𝑧 = 0

2
𝜕 2 𝑇 1 𝜕𝑇 1 𝜕 2 𝑇 𝜕 2 𝑇
𝛻 𝑇= 2+ + + =0
𝜕𝑟 𝑟 𝜕𝑟 𝑟 2 𝜕𝜙 2 𝜕𝑧 2
𝑟 𝑟
 T(r, z)
𝜙  T(r, ϕ)
 T(ϕ, z) : Not Important
Fourier–Bessel Series in the Finite Interval (0, r0)
Reference: “Heat Conduction” by Sadik Kakac, Yaman Yener, Carolina P. Naveira-Cotta

4
Separation of Variables: Cylindrical Coordinate
Our general guidelines established for separation of variables in the rectangular coordinate system
equally apply to the cylindrical coordinate problems.

Therefore, steady-state problems with a homogeneous PDE must always be solved with only a
single nonhomogeneous boundary condition.

If more than one boundary condition is nonhomogeneous, the principle of superposition is used to
treat each non-homogeneity individually.
For a nonhomogeneous PDE (i.e., when heat generation is present), superposition is
used to treat the heat generation within a 1-D ODE, which is coupled to a homogeneous
PDE at one or more boundary conditions.

T(r, z) Problems: Two Possibilities:

 r-directions homogeneous – Get orthogonal Bessel functions in r-directions
[Solution: J0, Y0]
 z-directions homogeneous – Get orthogonal trigonometric functions in z-dir
[Solution: sin, cos]
5
Example: Separation of Variables: T(r, z)
Steady State 2D Problem. Consider a 2D fin of circular cross section, with radius r and length L,
0
𝑧 protruding from a hot wall as shown in the Figure.
𝑇∞
The fin base temperature is specified as f(r). The fin is cooled by a fluid
stream maintained at a constant temperature T∞.

𝐿 𝑇∞
Assume constant thermal conductivity, an infinite heat transfer coefficient
ℎ→∞ and no sources or sinks of internal energy. Determine steady-state T(r, z).
𝑇∞
𝑟
Problem Formulation: 2 2 2
𝜕 𝑇 1 𝜕𝑇 1 𝜕 𝑇 𝜕 𝑇
𝛻2𝑇 = 2 + + 2 + =0
𝜕𝑟 𝑟 𝜕𝑟 𝑟 𝜕𝜙 2 𝜕𝑧 2
𝑇 = 𝑓(𝑟) 𝜕 2 𝑇 1 𝜕𝑇 𝜕 2 𝑇
2
+ + 2 =0 Laplace Equation
𝜕𝑟 𝑟 𝜕𝑟 𝜕𝑧

𝐁𝐂𝟏: 𝑇 0, 𝑧 = finite 𝐁𝐂𝟐: 𝑇 𝑟0 , 𝑧 = 𝑇∞

𝐁𝐂𝟑: 𝑇 𝑟, 0 = 𝑓(𝑟) 𝐁𝐂𝟒: 𝑇 𝑟, 𝐿 = 𝑇∞

Example: Separation of Variables: T(r, z)
Steady State 2D Problem. Problem Formulation:

𝑧 𝜕 2 𝑇 1 𝜕𝑇 𝜕 2 𝑇
2
+ + 2 =0
𝑇∞ 𝜕𝑟 𝑟 𝜕𝑟 𝜕𝑧

𝐁𝐂𝟏: 𝑇 0, 𝑧 = finite 𝐁𝐂𝟐: 𝑇 𝑟0 , 𝑧 = 𝑇∞ Many non-homogeneous

BC. SOV is not directly
𝐿 𝑇∞ 𝐁𝐂𝟑: 𝑇 𝑟, 0 = 𝑓(𝑟) 𝐁𝐂𝟒: 𝑇 𝑟, 𝐿 = 𝑇∞ applicable.

ℎ→∞
𝑇∞ However, the problem can be re-formulated under steady-state
𝑟 conditions in terms of θ(r, z) = T(r, z) – T∞ as follows:

𝑇 = 𝑓(𝑟) 𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃
+ + =0
𝜕𝑟 2 𝑟 𝜕𝑟 𝜕𝑧 2
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0

𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑓 𝑟 − 𝑇∞ ≡ 𝐹(𝑟) 𝐁𝐂𝟒: 𝜃 𝑟, 𝐿 = 0

Example: Separation of Variables: T(r, z)
Steady State 2D Problem.
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0
𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃
+ + =0 𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑓 𝑟 − 𝑇∞ ≡ 𝐹(𝑟) 𝐁𝐂𝟒: 𝜃 𝑟, 𝐿 = 0
𝜕𝑟 2 𝑟 𝜕𝑟 𝜕𝑧 2

𝑑2 𝑅 1 𝑑𝑅 𝑑2 𝑍
Divide by RZ
1 𝑑2 𝑅 1 𝑑𝑅 1 𝑑 2 𝑍
Try: 𝜃 𝑟, 𝑧 = 𝑅 𝑟 𝑍(𝑧) 𝑍 2 +𝑍 +𝑅 2 =0 2
+ + 2
=0
𝑑𝑟 𝑟 𝑑𝑟 𝑑𝑧 𝑅 𝑑𝑟 𝑅𝑟 𝑑𝑟 𝑍 𝑑𝑧
1 𝑑2 𝑅 1 𝑑𝑅 1 𝑑2 𝑍 2 Homogeneous direction: r
+ = − = −𝜆 f(R) f(Z)
𝑅 𝑑𝑟 2 𝑅𝑟 𝑑𝑟 𝑍 𝑑𝑧 2 Get CVP (Bessel Eq) in r-direction.
Select minus sign.
r-direction problem:
z-direction problem:
1 𝑑2 𝑅 1 𝑑𝑅 2
+ + 𝜆 =0 1 𝑑2 𝑍
𝑅 𝑑𝑟 2 𝑅𝑟 𝑑𝑟 − 𝜆2
=0
𝑑 2 𝑅 𝑑𝑅 𝑍 𝑑𝑧 2
⇒𝑟 2+ + 𝜆2 𝑟𝑅 = 0 Bessel
𝑑𝑟 𝑑𝑟 equation 𝑑2 𝑍
𝑑 2
𝑅 𝑑𝑅 of order ⇒ 2 − 𝜆2 𝑍 = 0
⇒𝑟 2
+ 𝑟 + 𝜆2 2
𝑟 𝑅=0 zero 𝑑𝑧
𝑑𝑟 2 𝑑𝑟
Example: Separation of Variables: T(r, z)
Steady State 2D Problem.
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0
𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃
+ + =0 𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑓 𝑟 − 𝑇∞ ≡ 𝐹(𝑟) 𝐁𝐂𝟒: 𝜃 𝑟, 𝐿 = 0
𝜕𝑟 2 𝑟 𝜕𝑟 𝜕𝑧 2
r-direction problem: z-direction problem:
1 𝑑2 𝑅 1 𝑑𝑅 2
2
+ + 𝜆 =0 (Next multiply by rR) 1 𝑑2 𝑍 𝑑2 𝑍
𝑅 𝑑𝑟 𝑅𝑟 𝑑𝑟 − 𝜆2 = 0 ⇒ 2 − 𝜆2 𝑍 = 0
𝑍 𝑑𝑧 2 𝑑𝑧
𝑑 2 𝑅 𝑑𝑅
⇒𝑟 2+ + 𝜆2 𝑟𝑅 = 0 (Next multiply by r)
𝑑𝑟 𝑑𝑟 Z-Solution:
2
𝑑 𝑅 𝑑𝑅 Bessel equation
⇒ 𝑟2 2 + 𝑟 + 𝜆2 𝑟 2 𝑅 = 0 of order zero 𝑍 𝑧 = 𝐵1 sinh 𝜆𝑧 + 𝐵2 cosh(𝜆𝑧)
𝑑𝑟 𝑑𝑟
R-Solution: 𝑅 𝑟 = 𝐴1 𝐽0 𝜆𝑟 + 𝐴2 𝑌0 (𝜆𝑟) 2
𝑑2 𝑦 𝑑𝑦
𝑥 2
+𝑥 + 𝑚2 𝑥 2 − 𝜈 2 𝑦 = 0
𝑑𝑥 𝑑𝑥
General Solution:
𝜃 𝑟, 𝑧 = 𝑅 𝑟 𝑍 𝑧 = 𝐴1 𝐽0 𝜆𝑟 + 𝐴2 𝑌0 (𝜆𝑟) 𝐵1 sinh 𝜆𝑧 +𝐵2 cosh(𝜆𝑧)
Example: Separation of Variables: T(r, z)
Steady State 2D Problem.
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0
𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃
+ + =0 𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑓 𝑟 − 𝑇∞ ≡ 𝐹(𝑟) 𝐁𝐂𝟒: 𝜃 𝑟, 𝐿 = 0
𝜕𝑟 2 𝑟 𝜕𝑟 𝜕𝑧 2

General Solution: 𝜃 𝑟, 𝑧 = 𝐴1 𝐽0 𝜆𝑟 + 𝐴2 𝑌0 (𝜆𝑟) 𝐵1 sinh 𝜆𝑧 +𝐵2 cosh(𝜆𝑧)

Eigenfunctions will be
obtained from R(r)
Apply BC1: A2 = 0
Note: Y0(0) = - ∞

Apply BC2: 0 = 𝐴1 𝐽0 𝜆𝑟0 A1 = 0 leads to trivial solution

Eigenfunction: 𝐽0 (𝜆𝑛 𝑟)
Eigenvalues λn are positive roots of: 𝐽0 𝜆𝑛 𝑟0 = 0, 𝑛 = 1,2,3, … .
Example: Separation of Variables: T(r, z)
Steady State 2D Problem.
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0
𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃
+ + =0 𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑓 𝑟 − 𝑇∞ ≡ 𝐹(𝑟) 𝐁𝐂𝟒: 𝜃 𝑟, 𝐿 = 0
𝜕𝑟 2 𝑟 𝜕𝑟 𝜕𝑧 2

General Solution: 𝜃 𝑟, 𝑧 = 𝐴1 𝐽0 𝜆𝑟 + 𝐴2 𝑌0 (𝜆𝑟) 𝐵1 sinh 𝜆𝑧 +𝐵2 cosh(𝜆𝑧)

Apply BC4: 𝑍 𝑧 = 𝐿 = 0 = 𝐵1 sinh 𝜆𝐿 + 𝐵2 cosh(𝜆𝐿)

⇒ 𝐵1 sinh 𝜆𝐿 + 𝐵2 cosh 𝜆𝐿 = 0 Now, 𝑍 𝑧 = 𝐵1 sinh 𝜆𝑧 +𝐵2 cosh(𝜆𝑧) becomes

⇒ 𝐵1 sinh 𝜆𝐿 = −𝐵2 cosh 𝜆𝐿 𝐵2 cosh 𝜆𝐿

⇒− sinh 𝜆𝑧 + 𝐵2 cosh(𝜆𝑧)
𝐵2 cosh 𝜆𝐿 sinh 𝜆𝐿
⇒ 𝐵1 = −
sinh 𝜆𝐿 𝐵2 sinh 𝜆𝐿 cosh 𝜆𝑧 − 𝐵2 cosh 𝜆𝐿 sinh(𝜆𝑧)
⇒
sinh 𝜆𝐿
𝐵2 sinh 𝜆(𝐿 − 𝑧 )
⇒ ⇒ 𝐶1 sinh 𝜆(𝐿 − 𝑧 )
Using: sinh 𝜆𝐿
𝐵2
sinh A − B = sinh 𝐴 cosh 𝐵 − cosh 𝐴 sinh(𝐵) Here, 𝐶1 =
sinh 𝜆𝐿
Example: Separation of Variables: T(r, z)
Steady State 2D Problem.
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0
𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃
+ + =0 𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑓 𝑟 − 𝑇∞ ≡ 𝐹(𝑟) 𝐁𝐂𝟒: 𝜃 𝑟, 𝐿 = 0
𝜕𝑟 2 𝑟 𝜕𝑟 𝜕𝑧 2

General Solution now becomes: 𝜃 𝑟, 𝑧 = 𝐴1 𝐽0 𝜆𝑟 + 𝐴2 𝑌0 (𝜆𝑟) + 𝐶1 sinh(λ(𝐿 − 𝑧))

Eigenfunction: 𝐽0 (𝜆𝑛 𝑟)
Eigenvalues are roots of: 𝐽0 𝜆𝑛 𝑟0 = 0, 𝑛 = 1,2,3, … . Z(z)

Using principle of superposition, we ∞

write the most general solution as
the sum of all solutions:
𝜃 𝑟, 𝑧 = 𝐶𝑛 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧
𝑛=1
where Cn = A1C1
This solution satisfy the PDE and all the
3 homogeneous boundary conditions.
Example: Separation of Variables: T(r, z) ∞
Steady State 2D Problem.
𝜃 𝑟, 𝑧 = 𝐶𝑛 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧 𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑓 𝑟 − 𝑇∞ ≡ 𝐹(𝑟)
𝑛=1

We now need to find Cn such that the nonhomogeneous boundary condition BC3 is satisfied.
∞

𝐹(𝑟) = 𝐶𝑛 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 This is Fourier–Bessel expansion of

F(r) on the interval (0, r0).
𝑛=1
𝑟0
2
Making use of orthogonality of eigenfunctions: 𝐶𝑛 sinh 𝜆𝑛 𝐿 = 2 2 𝐹(𝑟)𝐽0 𝜆𝑛 𝑟 𝑟𝑑𝑟
𝑟0 𝐽1 𝜆𝑛 𝑟0 0
Use Table
𝑟0
1 𝑟0
𝑟02 2
𝐴𝑛 = 𝑓(𝑟)𝐽𝜈 𝜆𝑛 𝑟 𝑟𝑑𝑟 𝑁𝑛 = 𝐽𝜈2 𝜆𝑛 𝑟 𝑟𝑑𝑟 = 𝐽𝜈+1 𝜆𝑛 𝑟0
𝑁𝑛 0 0 2

The final solution for θ(r, z) can now be written as:

∞ 𝑟0
2 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧
𝜃 𝑟, 𝑧 = 2 𝐹(𝑟′)𝐽0 𝜆𝑛 𝑟 ′ 𝑟′𝑑𝑟′
𝑟0 𝐽12 𝜆𝑛 𝑟0 sinh 𝜆𝑛 𝐿 0
𝑛=1
Example: Separation of Variables: T(r, z)
Steady State 2D Problem: Special Case
∞ 𝑟0
2 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧
The final solution for θ(r, z) : 𝜃 𝑟, 𝑧 = 2 𝐹(𝑟′)𝐽0 𝜆𝑛 𝑟 ′ 𝑟′𝑑𝑟′
𝑟0 𝐽12 𝜆𝑛 𝑟0 sinh 𝜆𝑛 𝐿 0
𝑛=1
𝑧
𝑇∞ Let us consider: 𝑇 = 𝑓 𝑟 = 𝑇0 = constant (Constant Fin Base Temperature)
𝑟0
𝑟0
𝑟 𝑟0
Evaluate: (𝑇0 −𝑇∞ ) 𝐽0 𝜆𝑛 𝑟 𝑟𝑑𝑟 = (𝑇0 − 𝑇∞ ) 𝐽 𝜆 𝑟 = (𝑇0 − 𝑇∞ ) 𝐽1 𝜆𝑛 𝑟0
0 𝜆𝑛 1 𝑛 𝜆𝑛
𝑇∞ 0
𝐿
The temperature distribution becomes:
ℎ→∞
𝑇∞ ∞
2 (𝑇0 − 𝑇∞ ) 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧
𝑟 𝜃 𝑟, 𝑧 =
𝑟0 𝜆𝑛 𝐽1 𝜆𝑛 𝑟0 sinh 𝜆𝑛 𝐿
𝑛=1
∞
𝑇 = 𝑓 𝑟 = 𝑇0 𝜃(𝑟, 𝑧) 2 1 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧
⇒ =
Constant fin
𝜃0 𝑟0 𝜆𝑛 𝐽1 𝜆𝑛 𝑟0 sinh 𝜆𝑛 𝐿
𝑛=1
base temperature
where 𝜃0 = 𝑇0 − 𝑇∞
 How to Compute Temperature at a Given
Point in the Solid Cylinder?
∞
𝜃(𝑟, 𝑧) 2 1 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧 How to
𝑧 = evaluate this?
𝜃0 𝑟0 𝜆𝑛 𝐽1 𝜆𝑛 𝑟0 sinh 𝜆𝑛 𝐿
𝑇∞ 𝑛=1

where: θ0 = T0 – T∞

𝑇∞
Here λn are the positive roots of
𝐿
ℎ→∞ 𝐽0 𝜆𝑛 𝑟0 = 0, 𝑛 = 1,2,3, … .
𝑇∞
𝑟

𝑇 = 𝑓(𝑟)
Example: Separation of Variables: T(r, z)
Consider a 2D fin of circular cross section, with radius r0 and length L. It is given that L/r0 = 2.0
Find the value of θ(r, z)/θ0 at z/L = 0.75, r/r0 = 0.50.
∞
𝜃(𝑟, 𝑧) 2 1 𝐽0 𝜆𝑛 𝑟 sinh 𝜆𝑛 𝐿 − 𝑧 Here λn are the positive roots of
= where: θ0 = T0 – T∞
𝜃0 𝑟0 𝜆𝑛 𝐽1 𝜆𝑛 𝑟0 sinh 𝜆𝑛 𝐿
𝑛=1 𝐽0 𝜆𝑛 𝑟0 = 0, 𝑛 = 1,2,3, … .
𝑧 Define: αn = λnr0
𝑇∞ ∞
𝜃(𝑟/𝑟0 , 𝑧/𝐿) 1 𝐽0 𝛼𝑛 𝑟/𝑟0 sinh 𝛼𝑛 𝐿/𝑟0 1 − 𝑧/𝐿
=2
𝜃0 𝛼𝑛 𝐽1 𝛼𝑛 sinh 𝛼𝑛 𝐿/𝑟0
𝑛=1
∞
𝐿 𝑇∞ 𝜃(0.5,0.75) 1 𝐽0 0.5𝛼𝑛 sinh 𝛼𝑛 2.0 1 − 0.75
⇒ =2
ℎ→∞ 𝜃0 𝛼𝑛 𝐽1 𝛼𝑛 sinh 2𝛼𝑛
𝑛=1
𝑇∞
𝑟 ∞
𝜃(0.5,0.75) 1 𝐽0 0.5𝛼𝑛 sinh 0.5𝛼𝑛
⇒ =2
𝜃0 𝛼𝑛 𝐽1 𝛼𝑛 sinh 2𝛼𝑛
𝑛=1
𝑇 = 𝑓(𝑟)
How to compute this?
Example: Separation of Variables: T(r, z)
∞
𝜃(0.5,0.75) 1 𝐽0 0.5𝛼𝑛 sinh 0.5𝛼𝑛 Here λn are the positive roots of
=2 where αn = λnr0
𝜃0 𝛼𝑛 𝐽1 𝛼𝑛 sinh 2𝛼𝑛
𝑛=1 𝐽0 𝜆𝑛 𝑟0 = 0, 𝑛 = 1,2,3, … .
Zero’s of J0

3

MATLAB: >> besselj(1, 2.4048)

J = besselj(NU,Z)
Computes the Bessel function ans =
of the first kind, J_nu(Z). 0.5192
Separation of Variables: T(r, z): Different BC
Let us now consider the solution of the heat conduction in the same 2D fin, but with the
𝑧 following boundary conditions:
𝑇∞ 𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃
2
+ + 2=0 θ(r, z) = T(r, z) – T∞
𝜕𝑟 𝑟 𝜕𝑟 𝜕𝑧
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 𝑓 𝑧 − 𝑇∞ ≡ 𝐹(𝑧)
𝐿 𝑓(𝑧)
𝐁𝐂𝟑: 𝜃 𝑟, 0 = 0 𝐁𝐂𝟒: 𝜃 𝑟, 𝐿 = 0
Since the boundary conditions in the z direction are homogeneous,
𝑟 we should get Sturm-Liouville type CVP in z-direction.

Try the product solution: 𝜃 𝑟, 𝑧 = 𝑅 𝑟 𝑍(𝑧)

𝑇 = 𝑇∞ Homogeneous direction: z
1 𝑑2 𝑅 1 𝑑𝑅 1 𝑑2 𝑍
⇒ 2
+ =− 2
= ±𝜆2 Get CVP in z-direction. Select
𝑅 𝑑𝑟 𝑅𝑟 𝑑𝑟 𝑍 𝑑𝑧 PLUS sign. Get modified Bessel
General Solution: function and [sin, cos]

𝜃 𝑟, 𝑧 = 𝑅 𝑟 𝑍 𝑧 = 𝐴1 𝐼0 𝜆𝑟 + 𝐴2 𝐾0 (𝜆𝑟) 𝐵1 sin 𝜆𝑧 +𝐵2 cos(𝜆𝑧)

Apply boundary conditions and complete the solution. Homework!
1. 2D Heat Conduction Problems in Cylindrical Coordinate System T(r,ϕ,z):
 T(r, z)
 T(r, ϕ)
2. T(r, z) Problem: Separation of Variables: Example
 T(ϕ, z) : Not Important

3. T(r, z) Problem: Compute Temperature at a Given Point in Cylinder

4. Another T(r, z) Problem with different BC: Separation of Variables

Once the temperature distribution is known, the heat transfer rate across
any area A can be calculated by using Fourier’s law of heat conduction:
𝜕𝑇 where n represents the
𝑞𝑛 = − 𝑘 𝑑𝐴
𝐴 𝜕𝑛 direction normal to dA.
“Heat Conduction” by Sadik Kakac, Yaman Yener, Carolina P. Naveira-Cotta,
5th Ed, CRC Press

“Heat Conduction” by David W. Hahn, M. Necati Ozisik,

3rd Ed, John Wiley and Sons, Inc

“Heat Conduction” by M. Necati Ozisik, 2nd Ed, John Wiley

and Sons, Inc

“Heat Conduction” by Latif Jiji, 3rd Ed, Springer

Module – 3 : Multi-dimensional Steady-State Heat Conduction Problems:
Solution by Separation of Variables
Lecture – 17 : Problems on Cylindrical Coordinate System (Cont’d)
1. 2D Heat Conduction Problems in Cylindrical Coordinate System:
 T(r, z)
 T(r, ϕ)
2. T(r, ϕ) Problem: Separation of Variables: Example  T(ϕ, z) : Not Important

3. Non-homogeneous Differential Equation: Example

Separation of Variables: Cylindrical
Coordinate System
Heat equation at steady-state
𝑇(𝑟, 𝜙) without any heat source:
𝜙
𝑟 𝛻 2 𝑇 𝑟, 𝜙, 𝑧 = 0
𝑧 𝑧

2
𝜕 2 𝑇 1 𝜕𝑇 1 𝜕 2 𝑇 𝜕 2 𝑇
𝛻 𝑇= 2+ + + =0
𝜕𝑟 𝑟 𝜕𝑟 𝑟 2 𝜕𝜙 2 𝜕𝑧 2
𝑇(𝑟, 𝜙, 𝑧) 𝑇(𝑟, 𝑧)

 T(r, z)
 T(r, ϕ)
𝑟 𝑟  T(ϕ, z) : Not Important

𝜙
 Example: Separation of Variables: T(r, ϕ)
Steady State 2D Problem. Consider a long solid cylinder of circular
cross section and radius r0. Assume that the
surface of the cylinder is held at an arbitrary
temperature f(ϕ). Find steady state
temperature distribution T(r, ϕ).
Assume: No heat generation, constant k.

Governing 𝜕 2 𝑇 1 𝜕𝑇 1 𝜕 2 𝑇
PDE: + + = 0, 0 ≤ 𝑟 ≤ 𝑟0 , 0 ≤ 𝜙 ≤ 2𝜋
𝜕𝑟 2 𝑟 𝜕𝑟 𝑟 2 𝜕𝜙 2

BC: 𝑇 0, 𝜙 = finite 𝑇 𝑟0 , 𝜙 = 𝑓(𝜙) r-direction: Non-homogeneous

Periodic BC,
𝑇 𝑟, 𝜙 = 𝑇(𝑟, 𝜙 + 2𝜋) 𝜕𝑇(𝑟, 𝜙) 𝜕𝑇(𝑟, 𝜙 + 2𝜋)
= 2π periodicity
𝜕𝜙 𝜕𝜙
Continuity of temperature and heat flux when traveling one full ϕ-direction:
revolution (2π radians) around the cylinder Homogeneous BC

Try Product Solution: 𝑇 𝑟, 𝜙 = 𝑅(𝑟)𝜓(𝜙)

 Example: Separation of Variables: T(r, ϕ)
𝜕 2 𝑇 1 𝜕𝑇 1 𝜕 2 𝑇
+ + =0 Try the product solution: 𝑇 𝑟, 𝜙 = 𝑅(𝑟)𝜓(𝜙)
𝜕𝑟 2 𝑟 𝜕𝑟 𝑟 2 𝜕𝜙 2
Divide by ψR
2
𝑑 𝑅 𝜓 𝑑𝑅 𝑅 𝑑 𝜓 2 1 𝑑2 𝑅 1 𝑑𝑅 1 𝑑2 𝜓
⇒𝜓 + + 2 =0 + + =0
𝑑𝑟 2 𝑟 𝑑𝑟 𝑟 𝑑𝜙 2 𝑅 𝑑𝑟 2 𝑟𝑅 𝑑𝑟 𝑟 2 𝜓 𝑑𝜙 2
Multiply both sides by r2
𝑟 2 𝑑 2 𝑅 𝑟 𝑑𝑅 1 𝑑2𝜓 2
⇒ + = − = ±𝜆 Choose the sign of the separation constant
𝑅 𝑑𝑟 2 𝑅 𝑑𝑟 𝜓 𝑑𝜙 2
such that the homogeneous ϕ direction
results in a Sturm–Liouville type characteristic
r-direction problem: value problem. Select plus sign.
(Multiplying by R)
ϕ- direction problem:
2
𝑑2 𝑅 𝑑𝑅 2
𝑟 + 𝑟 − 𝜆 𝑅=0 𝑑2𝜓
𝑑𝑟 2 𝑑𝑟 + 𝜆2 𝜓 = 0
𝑑𝜙 2
(Cauchy-Euler Equation or
Equidimensional Equation)
General Solution: 𝑇 𝑟, 𝜙 = 𝐴1 𝑟 𝜆 + 𝐴2 𝑟 −𝜆 𝐵1 sin 𝜆𝜙 + 𝐵2 cos(𝜆𝜙)
 Example: Separation of Variables: T(r, ϕ)
Separation of BC:
𝑇 0, 𝜙 = finite
r-direction problem:
𝑑2 𝑅 𝑑𝑅 Non-homogeneous
2 2 𝑇 𝑟0 , 𝜙 = 𝑓(𝜙)
𝑟 + 𝑟 − 𝜆 𝑅=0 BC
𝑑𝑟 2 𝑑𝑟
(Cauchy-Euler Equation or 𝑇 𝑟, 𝜙 = 𝑇(𝑟, 𝜙 + 2𝜋)
Equidimensional Equation)
𝜕𝑇(𝑟, 𝜙) 𝜕𝑇(𝑟, 𝜙 + 2𝜋)
𝑅 0 = finite =
𝜕𝜙 𝜕𝜙

Ψ - direction problem: This is a special kind of the

2
𝜓 0 = 𝜓(0 + 2𝜋) Sturm–Liouville problem with
𝑑 𝜓
2
+ 𝜆2 𝜓 = 0 𝑑𝜓(0) 𝑑𝜓(0 + 2𝜋) periodic boundary conditions
𝑑𝜙
= of period 2π.
𝑑𝜙 𝑑𝜙
General Solution: 𝑇 𝑟, 𝜙 = 𝐴1 𝑟 𝜆 + 𝐴2 𝑟 −𝜆 𝐵1 sin 𝜆𝜙 + 𝐵2 cos(𝜆𝜙)
 Example: Separation of Variables: T(r, ϕ)
2
𝑑 𝑅 𝑑𝑅 𝑑2𝜓 𝜓 0 = 𝜓(0 + 2𝜋) Solution: ψ(ϕ) =
𝑟2 2 + 𝑟 − 𝜆2 𝑅 = 0 + 𝜆2
𝜓=0
𝑑𝑟 𝑑𝑟 𝑑𝜙 2
𝑑𝜓(0) 𝑑𝜓(0 + 2𝜋) 𝐵1 sin 𝜆𝜙 + 𝐵2 cos(𝜆𝜙)
𝑇 0, 𝜙 = finite 𝑇 𝑟0 , 𝜙 = 𝑓(𝜙) =
𝑑𝜙 𝑑𝜙

General Solution: 𝑇 𝑟, 𝜙 = 𝐴1 𝑟 𝜆 + 𝐴2 𝑟 −𝜆 𝐵1 sin 𝜆𝜙 + 𝐵2 cos(𝜆𝜙)

Determination of eigenvalues: Impose the two periodic BC in ϕ direction.

𝐵1 sin 𝜆𝜙 + 𝐵2 cos 𝜆𝜙 = 𝐵1 sin 𝜆(𝜙 + 2𝜋) + 𝐵2 cos(𝜆(𝜙 + 2𝜋)) 𝜓 0 = 𝜓(0 + 2𝜋)

⇒ [sin 𝜆𝜙 − sin(𝜆 𝜙 + 2𝜋 )]𝐵1 + [cos 𝜆𝜙 − cos(𝜆 𝜙 + 2𝜋 )]𝐵2 = 0

𝐵1 𝜆 cos 𝜆𝜙 − 𝐵2 λ sin 𝜆𝜙 = 𝐵1 𝜆 cos 𝜆(𝜙 + 2𝜋) − 𝐵2 λ sin(𝜆(𝜙 + 2𝜋))

⇒ [cos 𝜆𝜙 − cos(𝜆 𝜙 + 2𝜋 )]𝐵1 − [sin 𝜆𝜙 − sin(𝜆 𝜙 + 2𝜋 )]𝐵2 = 0

𝑑𝜓(0) 𝑑𝜓(0 + 2𝜋)
=
𝑑𝜙 𝑑𝜙
 Example: Separation of Variables: T(r, ϕ)
Determination of eigenvalues: Impose the two BC in ϕ direction.
In order to have nontrivial solutions
[sin 𝜆𝜙 − sin(𝜆 𝜙 + 2𝜋 )]𝐵1 + [cos 𝜆𝜙 − cos(𝜆 𝜙 + 2𝜋 )]𝐵2 = 0
for B1 and B2, the determinant of the
[cos 𝜆𝜙 − cos(𝜆 𝜙 + 2𝜋 )]𝐵1 − [sin 𝜆𝜙 − sin(𝜆 𝜙 + 2𝜋 )]𝐵2 = 0 coefficients must vanish.

sin 𝜆𝜙 − sin(𝜆 𝜙 + 2𝜋 ) cos 𝜆𝜙 − cos(𝜆 𝜙 + 2𝜋 )

⇒ =0
cos 𝜆𝜙 − cos(𝜆 𝜙 + 2𝜋 ) −[sin 𝜆𝜙 − sin(𝜆 𝜙 + 2𝜋 )]

⇒ sin2 𝜆𝜙 + cos 2 𝜆𝜙 + sin2 𝜆(𝜙 + 2𝜋) + cos 2 𝜆(𝜙 + 2𝜋)

−2 cos 𝜆𝜙 cos 𝜆 𝜙 + 2𝜋 + sin 𝜆𝜙 sin 𝜆 𝜙 + 2𝜋 =0
Note:
⇒ 1 + 1 − 2[cos(𝜆 𝜙 + 2𝜋 − 𝜆𝜙] = 0 cos(𝐴 − 𝐵) = cos 𝐴 cos 𝐵 + sin 𝐴 sin(𝐵)

⇒ cos 2𝜆𝜋 = 1 This is possible only if λ is equal to one of the values of

𝜆𝑛 = 𝑛, 𝑛 = 0,1,2, … . These are the eigenvalues for our
problem. Neglect negative values.
Example: Separation of Variables: T(r, ϕ)
Eigenvalues are obtained as: 𝜆𝑛 = 𝑛, 𝑛 = 0,1,2, … .
𝑑2 𝑅 𝑑𝑅 𝑑2𝜓
𝑟2
+ 𝑟 − 𝜆2
𝑅=0 2
+ 𝜆2 𝜓 = 0
𝑑𝑟 2 𝑑𝑟 𝑑𝜙
Solution: 𝑅 𝑟 = 𝐴1 𝑟 𝑛 + 𝐴2 𝑟 −𝑛 Solution: 𝜓 𝜙 = 𝐵1 sin 𝑛𝜙 + 𝐵2 cos(𝑛𝜙)
Check n = 0 case
When n = λ = 0, the DE in R becomes: When n = 0, ψ = B20

2
𝑑2 𝑅 𝑑𝑅 1 𝑑 𝑑𝑅
𝑟 2
+ 𝑟 =0 ⇒ 𝑟 =0 ⇒ 𝑅 = 𝐴10 +𝐴20 ln 𝑟
𝑑𝑟 𝑑𝑟 𝑟 𝑑𝑟 𝑑𝑟
We can write the solutions in R and ψ directions as follows:
∞

𝑅 𝑟 = 𝐴10 + 𝐴20 ln 𝑟 + 𝐴1𝑛 𝑟 𝑛 + 𝐴2𝑛 𝑟 −𝑛

𝑛=1
and ∞

𝜓 𝜙 = 𝐵20 + 𝐵1𝑛 sin 𝑛𝜙 + 𝐵2𝑛 cos(𝑛𝜙)

𝑛=1
Example: Separation of Variables: T(r, ϕ)
We have obtained: ∞ ∞

𝑅 𝑟 = 𝐴10 + 𝐴20 ln 𝑟 + 𝐴1𝑛 𝑟 𝑛 + 𝐴2𝑛 𝑟 −𝑛 and 𝜓 𝜙 = 𝐵20 + 𝐵1𝑛 sin 𝑛𝜙 + 𝐵2𝑛 cos(𝑛𝜙)
𝑛=1 𝑛=1
Using principle of superposition, we write the general solution as:
∞

𝑇 𝑟, 𝜙 = 𝐵20 𝐴10 + 𝐴20 ln 𝑟 + 𝐴1𝑛 𝑟 𝑛 + 𝐴2𝑛 𝑟 −𝑛 𝐵1𝑛 sin 𝑛𝜙 + 𝐵2𝑛 cos(𝑛𝜙)

𝑛=1

Set A2n = 0 so that the solution will satisfy the finite BC: 𝑇 0, 𝜙 = finite
Note that we retain both
We can now write: constants B1n and B2n, which
∞
will always be the case for the
⇒ 𝑇(𝑟, 𝜙) = 𝑏0 + 𝑟 𝑛 𝑎𝑛 sin 𝑛𝜙 + 𝑏𝑛 cos(𝑛𝜙) 2π-periodicity requirement.
𝑛=1

where we have defined: 𝑏0 = 𝐵20 𝐴10 , 𝑎𝑛 = 𝐴1𝑛 𝐵1𝑛 , 𝑏𝑛 = 𝐴1𝑛 𝐵2𝑛

Example: Separation of Variables: T(r, ϕ)
∞

𝑇(𝑟, 𝜙) = 𝑏0 + 𝑟 𝑛 𝑎𝑛 sin 𝑛𝜙 + 𝑏𝑛 cos(𝑛𝜙) where 𝜆𝑛 = 𝑛, 𝑛 = 0,1,2, … . (A)

𝑛=1

Apply non-homogeneous BC: 𝑇 𝑟0 , 𝜙 = 𝑓(𝜙)

∞ This is the complete Fourier
⇒ 𝑓 𝜙 = 𝑏0 + 𝑟0𝑛 𝑎𝑛 sin 𝑛𝜙 + 𝑏𝑛 cos(𝑛𝜙) series representation of f(ϕ)
𝑛=1 on the interval (0, 2π).
The coefficients (b0, an, and bn) in the above equation can be determined as follows:
2𝜋
1 2𝜋 1
𝑏0 = 𝑓 𝜙 𝑑𝜙 𝑎𝑛 = 𝑛 𝑓 𝜙 sin(𝑛𝜙) 𝑑𝜙, 𝑛 = 1,2,3, … .
2𝜋 𝜋𝑟0 0
0
2𝜋
1
𝑏𝑛 = 𝑛 𝑓 𝜙 cos(𝑛𝜙) 𝑑𝜙, 𝑛 = 1,2,3, … .
𝜋𝑟0 0

We get the final form of T(r, ϕ) by substituting these expressions in Eq. (A).
Review: Complete Fourier Series
1 1
Any function of x, say F(x), can be written as: 𝐹 𝑥 = 𝐹 𝑥 − 𝐹(−𝑥) + 𝐹 𝑥 + 𝐹(−𝑥)
2 2
∞ Odd function Even function
𝑛𝜋 𝑛𝜋
⟹ 𝐹 𝑥 = 𝑏0 + 𝑎𝑛 sin 𝑥 + 𝑏𝑛 cos 𝑥 , −𝐿 < 𝑥 < 𝐿
𝐿 𝐿
𝑛=1

The coefficients an, bn, b0 are given as

 Example: Separation of Variables: T(r, ϕ)
Substituting the values of the coefficients an, bn, and b0, the final solution can then be written as:

2𝜋 ∞ 𝑛 2𝜋 2𝜋
1 1 𝑟
𝑇 𝑟, 𝜙 = 𝑓 𝜙′ 𝑑𝜙′ + sin(𝑛𝜙) 𝑓 𝜙′ sin(𝑛𝜙′) 𝑑𝜙′ + cos(𝑛𝜙) 𝑓 𝜙′ cos(𝑛𝜙′) 𝑑𝜙′
2𝜋 0 𝜋 𝑟0 0 0
𝑛=1

2𝜋
1
At r = 0, the above equation reduces to: 𝑇 𝑟 = 0, 𝜙 = 𝑓 𝜙′ 𝑑𝜙′
2𝜋 0

Hence, the centerline temperature is the average of the surface temperature

distribution.
Nonhomogeneous Differential Equation: T(r, ϕ)
Principle of Superposition
Nonhomogeneous Differential Equation: T(r, z)
A solid cylinder of radius r0 and length L generates heat at a volumetric rate of 𝑞. One plane surface
(base) is maintained at T0 while the other is insulated. The cylindrical surface is at a temperature T∞.
Determine the temperature distribution at steady state, T(r, z). Assume constant k.
Principle of
The problem can be formulated under steady-state Superposition
𝑧
condition in terms of θ(r, z) = T(r, z) – T∞ as follows:
Insulated
𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃 𝑞
2
+ + 2+ =0
𝜕𝑟 𝑟 𝜕𝑟 𝜕𝑧 𝑘
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0
𝐿 𝑇∞
𝜕𝜃 𝑟, 𝐿
𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑇0 − 𝑇∞ 𝐁𝐂𝟒: =0
𝜕𝑧
𝑟 Let us assume the solution to be in the form (sum of 2D and 1D):
𝜃 𝑟, 𝑧 = Ψ 𝑟, 𝑧 + 𝜙(𝑧)
𝑇 = 𝑇0 Then the problem can be written as a superposition of the
two simpler problems where the non-homogeneous term
(heat generation term) is included in 1D problem.
Nonhomogeneous Differential Equation: T(r, z)
𝜕 2 𝜃 1 𝜕𝜃 𝜕 2 𝜃 𝑞 Principle of
+ + + =0 Superposition
𝜕𝑟 2 𝑟 𝜕𝑟 𝜕𝑧 2 𝑘
𝐁𝐂𝟏: 𝜃 0, 𝑧 = finite 𝐁𝐂𝟐: 𝜃 𝑟0 , 𝑧 = 0

𝑧
𝜕𝜃 𝑟, 𝐿
𝐁𝐂𝟑: 𝜃 𝑟, 0 = 𝑇0 − 𝑇∞ 𝐁𝐂𝟒: =0
Insulated 𝜕𝑧
Try the solution: 𝜃 𝑟, 𝑧 = Ψ 𝑟, 𝑧 + 𝜙(𝑧)

𝐿 𝑇∞ 2
𝑑 𝜙 𝑞 1 𝜕 𝜕𝜓 𝜕2𝜓 Solve by
+ =0 𝑟 + 2 =0
𝑑𝑧 2 𝑘 𝑟 𝜕𝑟 𝜕𝑟 𝜕𝑧 SOV

𝑟 𝑑𝜙(𝐿) 𝜓 0, 𝑧 = finite 𝜓 𝑟0 , 𝑧 = −𝜙(𝑧)

=0
𝑑𝑧
𝜕𝜓(𝑟, 𝐿)
𝑇 = 𝑇0
𝜙 0 = 𝑇0 − 𝑇∞ 𝜓 𝑟, 0 = 0 =0
𝜕𝑧
Solve by integrating twice
1. 2D Heat Conduction Problems in Cylindrical Coordinate System:
 T(r, z)
 T(r, ϕ)
2. T(r, ϕ) Problem: Separation of Variables: Example  T(ϕ, z) : Not Important

3. Non-homogeneous Differential Equation: Example

“Heat Conduction” by Sadik Kakac, Yaman Yener, Carolina P. Naveira-Cotta,
5th Ed, CRC Press

“Heat Conduction” by David W. Hahn, M. Necati Ozisik,

3rd Ed, John Wiley and Sons, Inc

“Heat Conduction” by M. Necati Ozisik, 2nd Ed, John Wiley

and Sons, Inc

“Heat Conduction” by Latif Jiji, 3rd Ed, Springer

Module – 3 : Multi-dimensional Steady-State Heat Conduction Problems:
Solution by Separation of Variables
Lecture – 18 : Problems on Spherical Coordinate System
1. Steady-State 2D Heat Conduction Problems in Spherical Coordinate
System: Separation of Variables: Example - T(r,θ) Problem

2. SOV: Extension to 3D Problems

3. Additional Examples on Separation of Variables

Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.

z Consider the steady-state temperature distribution in a solid sphere

of radius r0. Assume that the surface of the sphere at r = r0 is
maintained at some arbitrary temperature distribution f(θ), the thermal
P(r,θ, ϕ) conductivity of the material of the sphere is constant, and there are
no internal energy sources or sinks in the sphere. Determine the
steady-state temperature distribution in the sphere, T(r, θ).
Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.
Consider the steady-state temperature distribution in a solid sphere of
radius r0. Assume that the surface of the sphere at r = r0 is maintained at
some arbitrary temperature distribution f(θ), the thermal conductivity of the
material of the sphere is constant, and there are no internal energy sources
or sinks in the sphere. Determine the steady-state temperature distribution
in the sphere, T(r, θ).

2
1 𝜕 2
𝜕𝑇 1 𝜕 𝜕𝑇 1 𝜕 𝑇
Laplace Equation in 𝑟 + sin 𝜃 + =0
Spherical Coordinate: 𝑟 2 𝜕𝑟 𝜕𝑟 𝑟 2 sin 𝜃 𝜕𝜃 𝜕𝜃 𝑟 2 sin2 𝜃 𝜕𝜙 2
𝜕 𝜕𝑇 1 𝜕 𝜕𝑇
Governing Equation: 𝑟2 + sin 𝜃 =0
𝜕𝑟 𝜕𝑟 sin 𝜃 𝜕𝜃 𝜕𝜃

BC: 𝑇 0, 𝜃 = finite 𝑇 𝑟0 , 𝜃 = 𝑓(𝜃) Non-homogeneous BC

𝑇 𝑟, 0 = finite 𝑇 𝑟, 𝜋 = finite
 Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.
𝑇 0, 𝜃 = finite 𝑇 𝑟0 , 𝜃 = 𝑓(𝜃)
Governing Equation: 𝜕 𝜕𝑇 1 𝜕 𝜕𝑇
𝑟2 + sin 𝜃 =0 𝑇 𝑟, 0 = finite 𝑇 𝑟, 𝜋 = finite
𝜕𝑟 𝜕𝑟 sin 𝜃 𝜕𝜃 𝜕𝜃

Assume a product solution in the form: 𝑇 𝑟, 𝜃 = 𝑅(𝑟) Ω(𝜃)

By substituting T = RΩ, the PDE can be separated into two as follows:

𝑑 𝑑𝑅 𝑅 𝑑 𝑑Ω Divide by RΩ 1 𝑑 2
𝑑𝑅 1 1 𝑑 𝑑Ω
Ω 𝑟 2
+ sin 𝜃 =0 𝑟 + sin 𝜃 =0
𝑑𝑟 𝑑𝑟 sin 𝜃 𝑑𝜃 𝑑𝜃 𝑅 𝑑𝑟 𝑑𝑟 Ω sin 𝜃 𝑑𝜃 𝑑𝜃

𝑟 2 𝑑 2 𝑅 2𝑟 𝑑𝑅 1 1 𝑑 𝑑Ω r-direction is non-homogeneous.
⇒ 2
+ = − sin 𝜃 = ±𝜆2 Select PLUS sign to get Legendre
𝑅 𝑑𝑟 𝑅 𝑑𝑟 Ω sin 𝜃 𝑑𝜃 𝑑𝜃 Differential Eq in θ direction.

𝑑 2 𝑅 2 𝑑𝑅 𝜆2 1 𝑑 𝑑Ω
+ − 𝑅=0 sin 𝜃 + 𝜆2 Ω = 0
𝑑𝑟 2 𝑟 𝑑𝑟 𝑟 2 sin 𝜃 𝑑𝜃 𝑑𝜃
(Multiplying both sides by R/r2)
 Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.
𝑇 0, 𝜃 = finite 𝑇 𝑟0 , 𝜃 = 𝑓(𝜃)
1 𝑑 𝑑Ω 𝑑 2 𝑅 2 𝑑𝑅 𝜆2
(A) sin 𝜃 + 𝜆2 Ω = 0 2
+ − 2𝑅 = 0 𝑇 𝑟, 0 = finite 𝑇 𝑟, 𝜋 = finite
sin 𝜃 𝑑𝜃 𝑑𝜃 𝑑𝑟 𝑟 𝑑𝑟 𝑟
Cauchy–Euler equation

The differential equation (A) can be transformed into Legendre’s Differential Equation by redefining the
independent variable as x = cos θ. This gives
𝑑 𝑑Ω
1 − 𝑥2 +𝑛 𝑛+1 Ω=0 where λ2 = n(n + 1)
𝑑𝑥 𝑑𝑥

If n is zero or a positive integer, the solutions of Legendre’s equation, which are finite at
x = 1 (i.e., θ = 0 and θ = π), are the Legendre polynomials. Therefore,

Ω𝑛 𝑥 = 𝐴𝑛 𝑃𝑛 𝑥 , 𝑛 = 0,1,2, … . .
Note that the finiteness of
or
the solution defines the
Ω𝑛 𝜃 = 𝐴𝑛 𝑃𝑛 cos 𝜃 , 𝑛 = 0,1,2, … . . characteristic functions and
Eigenfunctions Eigenvalues the characteristic values
 Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.
1 𝑑 𝑑Ω 𝑑 2 𝑅 2 𝑑𝑅 𝜆2
sin 𝜃 + 𝜆2 Ω = 0 2
+ − 2𝑅 = 0 𝑇 0, 𝜃 = finite 𝑇 𝑟0 , 𝜃 = 𝑓(𝜃)
sin 𝜃 𝑑𝜃 𝑑𝜃 𝑑𝑟 𝑟 𝑑𝑟 𝑟
Cauchy–Euler Equation
Solution: Ω𝑛 𝜃 = 𝐴𝑛 𝑃𝑛 cos 𝜃 ,
𝑛 = 0,1,2, … . .

The solution of the Cauchy–Euler equation can be written, with λ2= n(n + 1), as
𝑅𝑛 𝑟 = 𝐵𝑛 𝑟 𝑛 + 𝐶𝑛 𝑟 −(𝑛+1)

The general solution can be constructed as

∞

𝑇(𝑟, 𝜃) = 𝐴𝑛 𝐵𝑛 𝑟 𝑛 + 𝐶𝑛 𝑟 −(𝑛+1) 𝑃𝑛 cos 𝜃

𝑛=0
Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.
∞ 𝑇 0, 𝜃 = finite 𝑇 𝑟0 , 𝜃 = 𝑓(𝜃)
𝑇(𝑟, 𝜃) = 𝐴𝑛 𝐵𝑛 𝑟 𝑛 + 𝐶𝑛 𝑟 −(𝑛+1) 𝑃𝑛 cos 𝜃
𝑛=0

Apply the boundary condition: 𝑇 0, 𝜃 = finite Cn = 0

∞ ∞

Therefore, 𝑇(𝑟, 𝜃) = 𝐴𝑛 𝐵𝑛 𝑟 𝑛 𝑃𝑛 cos 𝜃 ⇒ 𝑇(𝑟, 𝜃) = 𝐾𝑛 𝑟 𝑛 𝑃𝑛 cos 𝜃 where Kn = AnBn

𝑛=0 𝑛=0

Now impose the second BC (non-homogeneous) in r-direction, 𝑇 𝑟0 , 𝜃 = 𝑓(𝜃)

∞

𝑓(𝜃) = 𝐾𝑛 𝑟0𝑛 𝑃𝑛 cos 𝜃 , 0<𝜃<𝜋

𝑛=0

This is the Fourier–Legendre series expansion of the surface temperature

distribution f(θ).
Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.
∞ ∞
𝑇(𝑟, 𝜃) = 𝐾𝑛 𝑟 𝑛 𝑃𝑛 cos 𝜃 𝑓(𝜃) = 𝐾𝑛 𝑟0𝑛 𝑃𝑛 cos 𝜃 , 0<𝜃<𝜋
𝑛=0 𝑛=0

If we introduce x = cos θ, the expansion f(θ) can also be written as:

∞

𝐹 𝑥 = 𝐾𝑛 𝑟0𝑛 𝑃𝑛 𝑥 , −1 < 𝑥 < 1 where we introduce 𝐹 𝑥 = 𝑓(cos −1 𝑥)

𝑛=0

The coefficients 𝐾𝑛 𝑟0𝑛 can now be determined by:

1
2𝑛 + 1
𝐾𝑛 𝑟0𝑛 = 𝐹 𝑥 𝑃𝑛 𝑥 𝑑𝑥
Recall: 2 −1
1
2𝑛 + 1 or, alternatively,
𝐴𝑛 = 𝑓 𝑥 𝑃𝑛 𝑥 𝑑𝑥
2 −1 𝜋
𝑛
2𝑛 + 1
𝐾𝑛 𝑟0 = 𝑓 𝜃 𝑃𝑛 cos 𝜃 sin 𝜃 𝑑𝜃
2 0
Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere.
1
We have obtained: 𝑛
2𝑛 + 1
𝐾𝑛 𝑟0 = 𝐹 𝑥 𝑃𝑛 𝑥 𝑑𝑥
∞ 2 −1
𝑇(𝑟, 𝜃) = 𝐾𝑛 𝑟 𝑛 𝑃𝑛 cos 𝜃
or, alternatively,
𝑛=0
𝜋
𝑛
2𝑛 + 1
𝐾𝑛 𝑟0 = 𝑓 𝜃 𝑃𝑛 cos 𝜃 sin 𝜃 𝑑𝜃
2 0

Hence, the temperature distribution T(r, θ) in the sphere can be written as:
∞ 𝑛 𝜋
1 𝑟
𝑇(𝑟, 𝜃) = 2𝑛 + 1 𝑃𝑛 cos 𝜃 𝑓 𝜃 ′ 𝑃𝑛 cos 𝜃 ′ sin 𝜃 ′ 𝑑𝜃′
2 𝑟0 0
𝑛=0
Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere: Special Case

The temperature distribution T(r, θ) in the sphere is:

∞ 𝑛 𝜋
1 𝑟
𝑇(𝑟, 𝜃) = 2𝑛 + 1 𝑃𝑛 cos 𝜃 𝑓 𝜃 ′ 𝑃𝑛 cos 𝜃 ′ sin 𝜃 ′ 𝑑𝜃′
2 𝑟0 0
𝑛=0

Let the surface temperature be specified as:

𝜋 where
𝑇0 , 0<𝜃< 𝑇0 , 0<𝑥<1 𝑥 = cos 𝜃
𝑓 𝜃 = 2 𝐹(𝑥) =
𝜋 0, −1 < 𝑥 < 0
0, <𝜃<𝜋 𝐹 𝑥 = 𝑓(cos −1 𝑥)
2

The expansion coefficients are then given by:

1
𝑛
2𝑛 + 1
𝐾𝑛 𝑟0 = 𝑇0 𝑃𝑛 𝑥 𝑑𝑥
2 0
 Example: Separation of Variables: T(r, θ)
Steady State 2D Problem in a Solid Sphere: Special Case Note: P0(x) = 1, P1(x) = x, ….

2𝑛 + 1 1 Integrating for n = 0, 1, 2, …. 1 1
1
𝑛 𝐾0 = 𝑇0 𝑑𝑥 = 𝑇0
𝐾𝑛 𝑟0 = 𝑇0 𝑃𝑛 𝑥 𝑑𝑥
2 0 2 0 2

1 1 1
3 3 51 3
71 3
7
2 2 𝐾3 𝑟0 = 𝑇 5𝑥 − 3𝑥 𝑑𝑥 = − 𝑇0
𝐾1 𝑟0 = 𝑇0 𝑥 𝑑𝑥 = 𝑇0 𝐾2 𝑟0 = 𝑇 3𝑥 − 1 𝑑𝑥 = 0
2 0 4 22 0 0
22 0 0 16

∞
Now, the temperature distribution
can be obtained from: 𝑇(𝑟, 𝜃) = 𝐾𝑛 𝑟 𝑛 𝑃𝑛 cos 𝜃
𝑛=0

3 5
𝑇(𝑟, 𝜃) 1 3 𝑟 7 𝑟 11 𝑟
⇒ = + 𝑃 (cos 𝜃) − 𝑃3 (cos 𝜃) + 𝑃5 (cos 𝜃) + ⋯
𝑇0 2 4 𝑟0 1 16 𝑟0 32 𝑟0
SOV: Extension to Three-
Dimensional Steady-State
Systems
Three-Dimensional Steady-State Systems
Consider solid rectangular parallelepiped

z-direction: Non-homogeneous

Solve the individual

problems as done in 2D
cases
Additional Examples on Separation of Variables
Example-1: Separation of Variables: Superposition
C
The cross section of a long prism is a right-angle isosceles triangle of side L
(AB=AC=L). The side AC exchanges heat by convection with the surroundings
maintained at temperature is T∞ and heat transfer coefficient h. The side AB is
maintained at zero temperature. The hypotenuse BC is insulated. Determine
the steady-state temperature Tc at the midpoint of the hypotenuse BC.
A B
𝑇=0
Solution:

The geometry is not suitable for separation of

variables.

However, due to symmetry the temperature

distribution in the right-angled isosceles
triangle is the same as that in a square with
boundary conditions symmetrical about the
diagonal.
Example-1: Separation of Variables: Superposition
C
Solution:
The temperature
at the center is
given by

A B
𝑇=0
Tc = T(L/2, L/2)
Example-2: Separation of Variables: T(r,ϕ)
Determine the two-dimensional steady-state temperature distribution T(r, ϕ) inside a large plate
with a circular hole of radius r0. The circular surface at r0 is held at a prescribed temperature f(ϕ).

𝜕 2 𝑇 1 𝜕𝑇 1 𝜕 2 𝑇
+ + = 0, 0 ≤ 𝑟 ≤ 𝑟0 , 0 ≤ 𝜙 ≤ 2𝜋
𝜕𝑟 2 𝑟 𝜕𝑟 𝑟 2 𝜕𝜙 2

𝑇 𝑟 → ∞, 𝜙 = finite 𝑇 𝑟0 , 𝜙 = 𝑓(𝜙)

𝑇 𝑟, 𝜙 = 𝑇(𝑟, 𝜙 + 2𝜋) 𝜕𝑇(𝑟, 𝜙) 𝜕𝑇(𝑟, 𝜙 + 2𝜋)

=
𝜕𝜙 𝜕𝜙

Try Product Solution:

𝑇 𝑟, 𝜙 = 𝑅(𝑟)𝜓(𝜙)
General Solution:

𝑇 𝑟, 𝜙 = 𝐴1 𝑟 𝜆 + 𝐴2 𝑟 −𝜆 𝐵1 sin 𝜆𝜙 + 𝐵2 cos(𝜆𝜙)
Homework!
1. Steady-State 2D Heat Conduction Problems in Spherical Coordinate
System: Separation of Variables: Example - T(r,θ) Problem

2. SOV: Extension to 3D Problems

3. Additional Examples on Separation of Variables

“Heat Conduction” by Sadik Kakac, Yaman Yener, Carolina P. Naveira-Cotta,
5th Ed, CRC Press

“Heat Conduction” by David W. Hahn, M. Necati Ozisik,

3rd Ed, John Wiley and Sons, Inc

“Heat Conduction” by M. Necati Ozisik, 2nd Ed, John Wiley

and Sons, Inc

“Heat Conduction” by Latif Jiji, 3rd Ed, Springer