NPTEL - Computational Electromagnetics – Dr.

Uday Khankhoje, EE IITM

Week 2: Numerical Integration, Quadrature rules, Gauss-Legendre quadrature,
introduction to integral equations, the Line charge problem, the Helmholtz equation for
wave scattering
Instructions: The objective of these questions is to assess your understanding of this week’s
content. You are not expected to memorize any of the questions, rather, you should derive the
answers from first principles based on what you have learnt so far.

1. Consider a function f ( x ) defined on the interval [ a, b]. You are given, say, N points ( xi , f ( xi )
on the graph of the function in this interval. Then, the number of polynomials of degree N
passing through these N points is
A. 0
B. 1
C. N
D. ∞

Solution: D.
A polynomial of degree N has the following form:

c N x N + c N −1 x N −1 + ..... + c2 x2 + c1 x1 + c0

Therefore, there are N + 1 unknowns. Since, we are given N points, we have N equa-
tions in N + 1 variables; therefore, there are infinitely many polynomials.

2. For the N-point Gauss quadrature on the interval [−1, 1], the nodes are chosen to be the
roots of the Legendre polynomial PN ( x ). What are the properties of these polynomials?
R1
A. −1 x k PN ( x )dx = 0 for k = 0, 1, 2, 3, 4, ...., N − 1
R1
B. −1 Q( X ) PN ( x )dx = 0 where Q( x ) is any polynomial of degree less than N.
R1
C. −1 Q( X ) PN ( x )dx = 0 where Q( x ) is any polynomial of degree greater than N.
D. It allows the Gauss-quadrature rule to be accurate upto polynomial order 2N − 1
despite having only N nodes.
E. It allows the Gauss-quadrature rule to be accurate upto polynomial order N − 1
since there are only N nodes.

Solution: A,B,D.

3. What is the order of the error in the approximation of an integral if we integrate the func-
tion using the trapezoidal rule over some interval [ a, b], where b − a = h?
A. O(h)
B. O(h3 )
C. O(h2 )
D. O(h4 )
 Solution: B.

4. The Gauss quadrature that was discussed in the lectures was for the interval [−1, 1]. How
would you make an N point Gauss quadrature on some other interval [ a, b]? (Assume that
you want to find the integral of the function f ( x ) over the interval [ a, b]).
Rb
S1: For the integral a f ( x )dx, we can use a change of variable from x to z such that
(b− a)z Rb (b− a) R 1
x = 2 + a. This converts the integral to a f ( x )dx = 2 −1 F ( z ) dz, where F ( z ) =
(b− a)z
f( 2+ a) Then, we can use an N point Gauss quadrature for this new integral in z.
 
2( x − a )
S2: We can compute the roots of the polynomials: PN (b−a) − 1 , where PN ( x ) is a Leg-
endre polynomial. These points can be used as the nodes of the quadrature rule. The
weights are the integrals of the Lagrange polynomials constructed using these nodes.
A. Both S1 and S2 are true.
B. S1 is true but S2 is false.
C. Both S1 and S2 are false.
D. S1 is false but S2 is true.

Solution: A.

5. Which of the polynomials are orthogonal to each other in the interval x ∈ [−1, 1]?
A. 1, x, 2x2 − 1
B. 1, 2x, 4x2 − 1
C. 1, x, x3
D. 1, x, (3x2 − 1)/3

Solution: A, B, D

6. In the potential problem discussed in the lectures, the metallic cylinder was kept at a con-
stant potential. Then, which of the following are correct? (Assume that the wire is placed
with its axis parallel to the y-axis).
A. Charges appear on the surface of conductor.
B. Charges appear in the bulk of the conductor.
C. To evaluate the potential at some arbitrary point in space, it is necessary to first
find out the distribution of charges on the cylinder.
D. We need to solve a Fredholm integral equation of the second kind in this problem.

Solution: A, C

7. Discretisation of the domain is an important parameter in solving an integral equation. In

the lectures, we discussed a method to arrive at a suitable value of the discrestisation for
the numerical convergence. Suppose, we are happy with a solution that does not change
more than 3% when compared to a more refined solution. Consider the following data:

Page 2
 k x100 − x75 k
1. k x75 k
= 0.10
k x125 − x100 ||
2. k x100 k
= 0.05
k x150 − x125 k
3. k x125 k
= 0.02
k x175 − x150 k
4. k x150 k
= 0.005

Here, x N denotes the solution with discretisation N; k x N − x M k denotes the norm of the
difference in solutions with discretisation N and M. Then, from the given data, the least
value of the discretisation that we should choose to satisfy the error constraint is
A. 150
B. 125
C. 100
D. 75

Solution: B
150 also satisfies the condition too, but 125 is the lesser of the 2.

8. What effect does the use of linear basis functions instead of pulse basis have on the solution
of the electric potential problem discussed in the lectures?
A. If the number of basis functions is kept the same, the resulting size of the matrix
equation is also the same in both cases
B. If the number of basis functions is kept the same, it improves accuracy of the
solution
C. If the number of basis functions is kept the same, the computational cost will also
increase
D. We can get same accuracy but with lesser number of basis functions
E. We can get same accuracy but with larger number of basis functions

Solution: B,C,D

9. In the 2-D transverse magnetic (TM) polarisation defined with respect to the x − y plane,
the field components considered are (as per the class lectures):
A. Ex , Ey , Hx
B. Ez , Hx , Hy
C. Ex , Ey , Hz
D. Ex , Hx , Hy

Solution: B

Page 3
10. In the lectures, we derived an integral equation for the the scattering problem. The integral
was over a surface; for region 1 (outside the scattering object and extending to ∞) and
resulted in the following identity:
Z I
( g1 ∇2 φ1 − φ1 ∇2 g1 )dS = ( g1 ∇φ1 − φ1 ∇ g1 ) · n̂dl
V1 S

Which of the following options were helpful for the conversion from a 2-D surface integral
to a 1D line integral?
A. The identity ∇ · ( g∇φ) = ∇ g · ∇φ + g∇2 φ
B. Stokes theorem
C. Gauss’s theorem
D. The fact that the field φ is 0 at infinity (since the other boundary of this problem
lies at ∞)

Solution: A, C, D

11. The lectures discussed the Helmholtz equation for the scattering problem:

∇2 E(r) − k20 ε(r ) E(r) = jωµ0 J (r)

The fields were assumed to be TM polarised, and they did not vary in the z-direction.
However, if we were to write the equation for scattering problem without any of these
assumptions, then which of the following equations best represent the modified Helmholtz
equation?
A. The equation remains unchanged; it is still correct.
B. ∇ × (∇ × E(r)) − k20 ε(r)E(r) = jωµJ(r)
C. ∇ × (∇ × E(r)) − k20 ε(r)E(r) = − jωµJ(r)
D. ∇ × (∇ × E(r)) − k20 ε(r)E(r) = 0

Solution: C.
This is because, ∇ × ∇ × E cannot be simplified to ∇2 E; instead, it is now ∇(∇ · E) −
∇2 E.

12. In the scattering problem, space is divided into 2 regions; the scattering object and free
space outside it. They are separated by the boundary S. In free space, the field is denoted
by φ1 , and inside the object, it is denoted by φ2 . Then, for φ1 the following equation is true:
(
φ1 (r0 ) r0 eV1 Eq(1)
I
φi (r0 ) − ( g1 (r, r0 )∇φ1 (r0 ) − φ1 (r0 )∇ g1 (r, r0 )) · n̂(r)dl =
S 0 r0 eV2 Eq(2)

Here, φi indicates the incident field. Mark the correct choices.

A. Eq (1) is the statement of Huygens’ principle
B. Eq (2) is the statement of the Extinction theorem
C. The normal vector is a function of r0

Page 4
 D. The normal vector is a function of r
E. The surface integral is over primed co-ordinates

Solution: A, B, D

13. In the scattering problem, suppose, the object is an infinitely long metallic cylinder with its
axis parallel to the z-axis. The cross-section of the cylinder is z-invariant. We assume a TM
polarisation. Then, the final expression for Huygens’ principle has terms which include:
A. g1 ∇φ1
B. φ1 ∇ g1
C. k2 φ1
D. k2 g1
E. None of the above

Solution: A

Page 5