Spreadsheets in Education (eJSiE)

Volume 4 | Issue 1

3-31-2010

Spreadsheet Implementations for Solving

Boundary-Value Problems in Electromagnetics
Mark A. Lau
Universidad del Turabo, [email protected]

Sastry P. Kuruganty
Universidad del Turabo, [email protected]

Follow this and additional works at: http://epublications.bond.edu.au/ejsie

Recommended Citation
Lau, Mark A. and Kuruganty, Sastry P. (2010) Spreadsheet Implementations for Solving Boundary-Value Problems in
Electromagnetics, Spreadsheets in Education (eJSiE): Vol. 4: Iss. 1, Article 1.
Available at: http://epublications.bond.edu.au/ejsie/vol4/iss1/1

This Regular Article is brought to you by the Bond Business School at ePublications@bond. It has been accepted for inclusion in Spreadsheets in
Education (eJSiE) by an authorized administrator of ePublications@bond. For more information, please contact Bond University's Repository
Coordinator.

Article 1

Spreadsheet Implementations for Solving Boundary-Value Problems in

Electromagnetics
Abstract

Electromagnetics is arguably one of the most challenging courses in any electrical engineering curriculum. A
solid foundation in vector calculus and a good intuition based on physical grounds are the normal
requirements for a student to successfully complete this course. This paper presents a simple, yet powerful
approach to introducing boundary-value problems arising in electrostatics. The principles of electrostatics find
numerous applications such as electrostatic machines, lightning rods, gas purification, food purification, laser
printers, and crop spraying, to name a few.
This paper focuses on the use of spreadsheets for solving electrostatic boundary-value problems. Sample
problems that introduce the finite difference and the finite element methods are presented. The geometries
included in the problems are sufficiently nontrivial for hand calculation or analytical solution, but reasonably
manageable using spreadsheets. Although specialized software is available for this purpose, oftentimes such
sophistication tends to obscure the mathematical underpinnings of the solution methods. Spreadsheets offer a
transparent alternative perhaps proximate to hand calculation for students to better appreciate the
numerical methods for solving boundary-value problems.
Keywords

Finite difference method, finite element method, Laplaces equation, boundary-value problems, electrostatics.
Cover Page Footnote

The authors gratefully acknowledge support from the Universidad del Turabo.

This regular article is available in Spreadsheets in Education (eJSiE): http://epublications.bond.edu.au/ejsie/vol4/iss1/1

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

1. Introduction
Many phenomena arising in science and engineering are modeled by
partial differential equations (PDEs). In such cases the quantity of interest
(e.g., temperature, potential, or displacement) is a function that depends
on more than one variable (typically, space variables x, y, z and the
temporal variable t). The heat equation, wave equation, and Laplaces
equation are among the most prominent PDEs that undergraduate
engineering students will encounter. The usual practice is to introduce the
student to the analytical solution of these equations via the method of
separation of variables. Under the assumption of linearity, the method
naturally leads to the formulation of solutions as Fourier series
expansions.
Treatment of PDEs and boundary-value problems (BVPs) may be
found in many standard books [2 4, 7]. Reference 2 provides a very
accessible presentation of the topic, while references 3, 4, and 7 provide a
more concise presentation geared toward compendium courses in
engineering mathematics. This paper will not expound the theories that
provide the mathematical underpinnings of PDEs; instead, the paper
emphasizes numerical solutions to PDEs and suggests implementations
through spreadsheets.
This paper focuses on certain numerical methods for solving PDEs; in
particular, the finite difference and the finite element methods are presented in
the context of problems arising in electrostatics. Much of the development
of these methods will follow those found in electromagnetics books [8].
The examples presented in this paper include geometries that are
sufficiently nontrivial for hand calculation or analytical solution, but
reasonably manageable by using spreadsheets. Although specialized
software is available for this purpose, oftentimes such sophistication tends
to obscure the inner workings of the numerical methods employed in the
solution of PDEs. Spreadsheets offer a transparent alternative perhaps
proximate to hand calculation for students to better appreciate the
numerical methods for solving PDEs and BVPs.
The use of spreadsheets in teaching finite element analysis has been
reported in the literature [9]. Reference 9 presents finite element analysis

Published by ePublications@bond, 2010

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

in the context of a plane truss structure, wherein the objective is the

determination of stresses given certain strains; the successful use of
spreadsheets for introducing finite element analysis in an undergraduate
mechanical engineering class is also discussed therein. A solution to the
heat equation using spreadsheet models and 3D graphics is provided in
[6]. Likewise, a menu-driven spreadsheet for solving electromagnetics
problems can be found in [10]. In this paper, however, solutions to
electromagnetics problems are presented in a manner that makes no use of
macro programming; besides, this paper introduces finite element
analysis via spreadsheets.
To the best of the authors knowledge, spreadsheets are more
prevalent in mechanical and civil engineering. The uses of spreadsheets in
electrical engineering appear to be more limited perhaps in this branch
of engineering the use of specialized software is more consonant with the
dynamic nature of the field. However, educators and students who wish
to stress the importance of fundamentals, rather than effortless and
expedite solutions, may want to consider using spreadsheets as a viable
tool. Spreadsheets can offer a reasonable tradeoff between user-defined
programming and specific-purpose software. It is in this spirit that this
paper is presented, continuing the efforts initiated by the authors in a
power systems course [5].
This paper is organized as follows. Section 2 presents the finite
difference method for solving an electrostatic problem and includes the
corresponding spreadsheet implementation. Section 3 presents the finite
element method along with a spreadsheet implementation of the method.
Section 4 compares the results obtained by each method. Section 5
discusses the pedagogical advantages of the spreadsheet implementations.
Finally, Section 6 gives concluding remarks.
2. The finite difference method
The finite difference method (FDM) is conceptually simple. The problems
to which the method applies are specified by a PDE, a solution region
(geometry), and boundary conditions. Only a brief outline of the method

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

is given in this paper; for more detailed derivations the reader may
consult [8]. The finite difference method entails three basic steps:
(1) Divide the solution region into a grid of nodes. Grid points are
typically arranged in a rectangular array of nodes.
(2) Approximate the PDE and boundary conditions by a set of linear
algebraic equations (the finite difference equations) on grid points
within the solution region.
(3) Solve this set of linear algebraic equations.
The method is illustrated with an example arising in electrostatics.
Consider the charge-free region depicted in Figure 1. The region has
prescribed potentials along its boundaries. The region is divided into a
rectangular grid of nodes, with the numbering of free nodes as indicated
in the figure.

Figure 1: Charge-free region showing prescribed potentials at the boundaries and

rectangular grid of free nodes to illustrate the finite difference method.

Published by ePublications@bond, 2010

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

The potential V = V(x,y) at an interior point (x,y) within the region is

governed by the two-dimensional Laplaces equation

  

 
 

 
 

 0.

(1)

Let the location of an interior grid point be identified by a pair of

integers (i,j), where i and j represent the position along the horizontal and
vertical directions, respectively. For a grid having equal horizontal and
vertical step sizes, the potential is given by the finite difference equation
, 




,
,
,
, .

(2)

In essence, the potential at an interior grid point is the average of its

four closest neighboring grid points that lie along the horizontal and
vertical grid lines that intersect at the point. For example, the finite
difference equation for node 4 in Figure 1 is
 





0

 .

(3)

Similar equations are formulated for the remaining free nodes leading
to a system of linear algebraic equations. This system of equations may be
solved by a variety of methods. In this section the Gauss-Seidel method
is implemented in a spreadsheet to solve this system of equations. The
Gauss-Seidel method is a relatively simple iterative method for solving
systems such as those encountered in the finite difference formulation.
The steps to implementing the spreadsheet for FDM are given below:
(1) The first step in the spreadsheet implementation is to input the
prescribed potentials at the boundaries of the solution region. This
step is shown in Figure 2. The prescribed potentials occupy the cell
range C6:C8 and correspond to the potentials (0 V and 100 V) given
in Figure 1. The user may change the numerical values of these
potentials to accommodate different boundary conditions.

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

Figure 2: Input section of spreadsheet implementation of the finite difference method.

(2) The next step is the implementation of the Gauss-Seidel method for
solving the finite difference equations. There are 6 potentials at
interior grid points that need to be determined (nodes 1 through 6
in Figure 1). These have been labeled V1 through V6 and occupy the
cell range B14:G14 in Figure 3. To start the iterations initial estimates
are given in B15:G15; in this case all initial estimates have been set to
zero.

Figure 3: Screenshot showing the Gauss-Seidel iterations of the difference equations.

Published by ePublications@bond, 2010

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

The cell range B16:G16 in Figure 3 implements the difference

equations which are then solved using the Gauss-Seidel method.
For example, for node 4, cell E16 contains the Microsoft Excel
formula that implements Equation (3)
=SUM($C$6,F15,B16,G15)/4.

Observe that the formula in cell E16 makes absolute references

to a node with prescribed potential; this prescribed potential must
remain constant throughout the Gauss-Seidel iterations. Making
the appropriate absolute references will prevent spurious results
from occurring when formulas are copied to other cells.
(3) The final step is to reproduce the formulas in cell range B16:G16 by
using the Copy command of Microsoft Excel. The formulas have
been copied over the cell range B17:G25. It can be observed from
Figure 3 that convergence is reached after 10 iterations for a
precision index of 10-2.
3. The finite element method
The finite element method (FEM) is a numerical technique for solving
PDEs. FEM was originally applied to problems in structural mechanics.
Unlike FDM, FEM is better suited for solution regions having irregularly
shaped boundaries. The finite element analysis involves four basic steps:
(1) Divide the solution region into a finite number of elements. The
most common elements have triangular or quadrilateral shapes.
The collection of all elements should resemble the original region as
closely as possible.
(2) Derive governing equations for a typical element. This step will
determine the element coefficient matrix.
(3) Assemble all elements in the solution region to obtain the global
coefficient matrix.
(4) Solve the resulting system of equations.

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

Again, the method is illustrated with an example. Consider the same

charge-free region shown in Figure 1. The region is divided into 25 equal
triangular elements as indicated in Figure 4. The elements are identified
by encircled numbers 1 through 25. In this discretization there are 21
global nodes numbered 1 through 21. In the paragraphs that follow, a
brief outline of FEM is provided; for detailed derivations the reader may
consult [8].

Figure 4: Finite element arrangement for electrostatic problem.

For each element e the following quantities are computed

     ,      ,     ,
     ,      ,     ,

Published by ePublications@bond, 2010

(4)

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

where the subscripts refer to the local node numbers 1, 2, and 3 of element
e. For example, in Figure 4, element 20 has global nodes 14, 18, and 17,
which correspond, respectively, to local nodes 1, 2, and 3. Local node
numbering is arbitrary; however, local node numbers must be assigned so
that global nodes associated with an element are traversed in a
counterclockwise sense. The coordinates (in meters) of nodes 14, 18, and
17 are (x1,y1) = (0.4, 0.4), (x2,y2) = (0.4, 0.6), and (x3,y3) = (0.2, 0.6),
respectively. These sets of coordinates yield P1 = 0.0, P2 = 0.2, P3 = 0.2, Q1 =
0.2, Q2 = 0.2, and Q3 = 0.0.
With Pi and Qi (i = 1, 2, 3) for element e thus computed, the entries of
the 3 x 3 element coefficient matrix are then given by
"

! 

#

$ 
  %

&, '  1, 2, 3

(5)

where
+




,     -.

(6)

As an example, for elements 20, 21 and 24, the element coefficient

matrices computed according to Equations (5) and (6) are, respectively,
0.5 0.5
0.0
! .  /0.5
1.0 0.51,
0.0 0.5
0.5

(7.a)

1.0 0.5 0.5

!   /0.5
0.5
0.01,
0.5
0.0
0.5

(7.b)

1.0 0.5 0.5

!   /0.5
0.5
0.01.
0.5
0.0
0.5

(7.c)

and

The global coefficient matrix is then assembled from the element

coefficient matrices. Since there are 21 nodes, the global coefficient matrix
will be a 21 x 21 matrix. In the following, the computations of one diagonal

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

and one off-diagonal entries are shown. For example, node 18, which
corresponds to the !2,2 entry in the global coefficient matrix C, belongs to
elements 20, 21 and 24; since node 18 is assigned local node number 2 in
elements 20 and 24, and local number 3 in element 21 (as seen in the
middle table of Figure 5), the corresponding global coefficient is
.





!2,2  !,
!,
!,  1.0
0.5
0.5  2.0.

(8)

Figure 5: Input section of finite element analysis for electrostatic problem: (left) global
node x and y coordinates; (middle) triangular element global and local node
correspondence; (right) nodes in the finite element mesh having prescribed potentials.

Published by ePublications@bond, 2010

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

For the off-diagonal entry !,2, global link 1418 corresponds to local
link 12 of element 20 and local link 13 of element 21 and hence
.



!,2  !,
!,  0.5  0.5  1.0.

(9)

Defining the vector of potentials 45 and 46 , where the subscripts f and

p refer to nodes with free (unknown) potentials and prescribed potentials,
respectively, the global coefficient matrix is then partitioned accordingly
and the unknown potentials are obtained from

45  !55
!56 46 .

(10)

The spreadsheet implementation of the finite element solution involves

the following steps:
(1) Generate the input data section as shown in Figure 5. The input
data consists of three tables: global node x and y coordinates; global
and local node correspondence for each element; and list of nodes
with prescribed (fixed) potentials.
(2) For each element, compute the values of Pi and Qi from Equation (4)
and obtain the element coefficient matrix from Equations (5) and
(6), as shown in Figure 6. Notice that the rows for elements 4
through 23 were intentionally hidden so as to render a more
manageable table.

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

10

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

Figure 6: Computation of element coefficient matrices ! " .

The table in Figure 6 is constructed as follows:

a. For element 1, global node information is linked to local node
numbers. The link is implemented in the cell range O19:O21 via
the Microsoft Excel VLOOKUP function. More precisely, cells
O19, O20, and O21 contain, respectively, the formulas
=VLOOKUP($M19,$E$19:$H$43,2)
=VLOOKUP($M20,$E$19:$H$43,3)
=VLOOKUP($M21,$E$19:$H$43,4)

which search the item in column M (element number) and return

the global nodes from the lookup table $E$19:$H$43.
b. The VLOOKUP function is invoked once again to retrieve global
node coordinates. In the cell range P19:Q21 the following
formulas implement this task:

Published by ePublications@bond, 2010

=VLOOKUP($O19,$A$19:$C$39,2)

=VLOOKUP($O19,$A$19:$C$39,3)

=VLOOKUP($O20,$A$19:$C$39,2)

=VLOOKUP($O20,$A$19:$C$39,3)

=VLOOKUP($O21,$A$19:$C$39,2)

=VLOOKUP($O21,$A$19:$C$39,3)

11

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

c. Cell ranges R20:T20 and V20:X20 compute Pi and Qi (i = 1, 2, 3)

according to Equation (4) with the Excel formulas
=Q20-Q21

=Q21-Q19

=Q19-Q20

=P21-P20

=P19-P21

=P20-P19

d. Cells U20 and Y20 are optional. These cells simply compute P1 +
P2 + P3 and Q1 + Q2 + Q3, as both sums must equal zero.
e. Cell Z20 implements Equation (6):
=(S20*X20-T20*W20)/2

f. The cell range AA19:AC21 computes the entries of the coefficient

matrix for element 1 from Equation (5). For example, the entry

! is computed by the following Excel formula:
=(R20*S20+V20*W20)/(4*Z20)

g. The formulas in cell range N19:AC21 are copied over the cell
range N22:AC93 to complete the computation of the remaining
element coefficient matrices.
(3) The assembly of the global coefficient matrix is shown in Figure 7.
The numbers in AF18:AZ18 and AE19:AE39 refer to global node
numbers. The coefficients of the global matrix occupy cell range
AF19:AZ39, which contain formulas similar to those in Equations (8)
and (9). For example, the coefficient !2,2 in cell AW36 contains the
formula
=AB77+AC81+AB89,

while that of !,2 in cell AW32 is

=AB76+AC79.

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

12

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

Figure 7: Screenshot showing global coefficient matrix C.

(4) The matrices !55 and !56 are formed by extracting the appropriate
rows and columns from the global coefficient matrix C. In this case,
nodes 8, 9, 10, 13, 14, and 17 are the free nodes, while 1, 2, 3, 4, 5, 6,
7, 11, 12, 15, 16, 18, 19, 20, and 21 are the nodes with prescribed
potentials. The result is shown in Figure 8.
(5) The final solution is obtained by using the matrix capabilities of
Microsoft Excel. Although the matrix capabilities of Excel are fairly
limited compared to those available in Matlab, Mathematica, or
Maple, the ones provided by Excel are quite sufficient for the FEM
implementation proposed in this paper. In particular, the
spreadsheet functions MINVERSE (matrix inverse) and MMULT
(matrix multiplication) are used in the implementation of Equation
(10). This step is shown in Figure 9.

Published by ePublications@bond, 2010

13

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

Figure 8: Matrices !55 and !56 obtained from global coefficient matrix C.

Figure 9: Final calculations section of finite element solution to electrostatic problem.

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

14

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

The implementation of Equation (10) has been broken down

into three parts: (i) computation of the inverse of the !55 matrix

(this has been labeled +  !55
in Figure 9); (ii) computation of an
intermediate vector 7  !56 46 ; and (iii) computation of vector of
potentials at free nodes 45  + 7.
4. Discussion of results
In Sections 2 and 3 spreadsheet implementations of FDM and FEM
were presented. As indicated in Table 1, the potentials at the free nodes
computed by both methods compared fairly well. The node numbers in
the table for FDM correspond to those in Figure 1, while those for FEM
correspond to the node numbers shown in Figure 4. A better agreement
can be obtained if more iterations of the Gauss-Seidel method are
performed; this can be easily accommodated by copying a complete row
of existing formulas, say cell range B25:G25 in Figure 3, to new rows until a
desired level of agreement is achieved.
Table 1: Comparison of results obtained by FDM and FEM.

Finite difference
Node
Potential (V)
1
18.180
2
36.363
3
59.091
4
36.363
5
68.181
6
59.091

Finite element
Node
Potential (V)
8
18.182
9
36.364
10
59.091
13
36.364
14
68.182
17
59.091

In general, FEM can better handle complex geometries and boundary

conditions, while FDM is more suitable for solution regions having a
certain degree of regularity. The tradeoff between the simplicity and the
generality of the methods is evidenced by the amount of programming
required by each method.

Published by ePublications@bond, 2010

15

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

The potential distribution may be conveniently displayed graphically

as depicted in Figure 10. The 2D chart on the left of the figure shows the
potential distribution using conditional formatting for the color scheme, as
demonstrated in [1]. The 3D chart on the right simply shows the potential
distribution, with the heights of the cylinders being proportional to the
value of the calculated potentials.

Figure 10: Graphical display of potential distribution.

5. Pedagogical advantages of spreadsheet implementations

In light of the spreadsheet implementations presented in this paper,
the following remarks could be made:
(i)

Setting up spreadsheets demands precise attention to detail from

the user. Incorrect use of the Copy command or improper
referencing of cells will lead to erroneous results. Attention to
detail is a desirable skill that students should hone. Spreadsheets
offer an environment in which such skill can be honed; the use of
highly specialized software without proper understanding of the
underlying methods may at times impede development of the skill
in the student.

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

16

Lau and Kuruganty: Boundary-Value Problems in Electromagnetics

(ii) Spreadsheets offer a reasonable tradeoff between the sophistication

of specific-purpose software and programming. The amount of
programming in a typical spreadsheet is minimal, often reduced to
formula editing and copying. This approach allows students to
concentrate on analysis and interpretation of results rather than on
time-consuming code debugging.
(iii) The spreadsheet implementations mimic hand calculations. The
notepad-like interface of spreadsheets allows the student to keep
track of results and ascertain convergence. Because of the
resemblance to hand calculations, the spreadsheet approach may
provide the student with a deeper understanding of the numerical
methods, which could be obscured if specific-purpose software is
used without proper knowledge of such methods.
(iv) The spreadsheets implementations may be presented to students to
introduce numerical methods for solving BVPs. Students may be
asked to implement similar spreadsheets to solve other types of
PDEs and BVPs. The ambitious students may even improve upon
the spreadsheets presented in this paper by macro programming or
creative use of other Microsoft Excel functions.
In fairness to specific-purpose software designed for BVPs modeled by
PDEs, it can be argued that such programs serve other purposeful needs,
namely, handling large-scale systems, accommodating highly irregular
geometries, and handling complex boundary conditions. Problems of
considerable size and high complexity may not be handled efficiently by
spreadsheets.
6. Conclusions
This paper presented spreadsheet implementations of two numerical
methods for solving electrostatics problems. The spreadsheet approach is
ideal if the emphasis is on understanding of numerical techniques.
Spreadsheets may be considered as a viable alternative to enhancing
education in other subjects and engineering fields.

Published by ePublications@bond, 2010

17

Spreadsheets in Education (eJSiE), Vol. 4, Iss. 1 [2010], Art. 1

The interested reader may obtain a copy of the Microsoft Excel file that
implements the FDM and FEM solutions by sending an e-mail to
[email protected].
References
[1]

Abramovich, S. and Sugden, S. Spreadsheet conditional formatting: an

untapped resource for mathematics education, eJournal of Spreadsheets in
Education, Vol. 1, Issue 2, Article 3, 2004.
Online: http://epublications.bond.edu.au/ejsie/vol1/iss2/3.

[2]

Asmar, N. Partial differential equations with Fourier series and boundary value
problems, 2nd edition, Prentice Hall, 2004.

[3]

Greenberg, M.
1998.

[4]

Kreyszig, E. Advanced engineering mathematics, 9th edition, John Wiley & Sons,
2006.

[5]

Lau, M. A. and Kuruganty, S. P. Spreadsheet implementations for solving

power-flow problems, eJournal of Spreadsheets in Education, 3(1): 27-45, August
2008. Online: http://epublications.bond.edu.au/ejsie/vol3/iss1/3.

[6]

Neuwirth, E. and Arganbright, D. The active modeler: mathematical modeling with

Microsoft Excel, 141-155, Brooks/Cole, 2004.

[7]

ONeil, P. Advanced engineering mathematics, 6th edition, CL-Engineering, 2006.

[8]

Sadiku, M. Elements of electromagnetics, 4th edition, Oxford University Press,

2006.

[9]

Teh, K. and Morgan, L. The application of Excel in teaching finite element

analysis to final year engineering students, Proceedings of the 2005 ASEE/AaeE
(Australian Association for Engineering Education) Global Colloquium on
Engineering Education, Paper # 50.

Advanced engineering mathematics, 2nd edition, Prentice Hall,

[10] Yamani, A. and Kharab, A. Use of a spreadsheet program in electromagnetics,

IEEE Transactions on Education, 44(3): 292-297, 2001.

http://epublications.bond.edu.au/ejsie/vol4/iss1/1

18