You are on page 1of 4

Contents: 10/12

Discussion: 2/3 lacks concluding statement


Technical: 3/3 minor typo error
Formats: 2/2
Approximation of transcendental functions using Legendre

TOTAL: 17/20 polynomials


Jonecis A. Dayap

Content checklist
Abstract
History: lacks context of
discovery and This expository paper illustrates some useful properties of Legendre polynomials
motivation
Construction: ok in approximating continuous functions, particularly, transcendental functions. It
Form: ok
Illustration: ok is shown that the condition of the orthogonality of these polynomials leads to this
Pros: ok 2n+1 1
R
Cons: none specified equation a n = 2 −1 f (x)Pn (x)dx so that the function f (x) is best approximated
The claim "best"
by the sum of Legendre polynomials a0 P0 (x)+a1 P1 (x)+...+an Pn (x) over an interval
is unfounded.
Applications: ok [−1, 1] in L2 function space..
References: ok
Keywords: Transcendental functions, Legendre polynomials, orthogonality.

1 Introduction

Transcendental function is an analytic function that doesn’t satisfy a polynomial


such a There are
equation that is, explicit , finite formula is unavailable [1]. Approximating this function many forms of
transcendentals.
using polynomials is a basic building block for most of the numerical techniques. It is ob-
served that polynomials, especially those of low order, are easy to manipulate and provide
computational speed with minimal penalty in representing general nonlinear relationships
the class of
[3]. One amongst these approximating polynomials is Legendre polynomials.
Legendre polynomials are named after French mathematician Adrien-Marie Legendre
who discovered them in 1782. They are a complete set of orthogonal polynomials, with
rich mathematical properties, and plenty of applications [5]. These polynomials are widely
employed in determining the wave functions of electrons within the orbits of an atom [8]
and in determining the potential functions within the spherically symmetric geometry
[6].
Orthogonal polynomials such as Legendre polynomials play a central role in the solu-
tion of least-squares problems. Legendre polynomials have been applied in approximation
theory [2]. Aside from using these polynomials as tools for the approximation of func-
tions that are difficult to compute, they are also essential of numerical integration and
approximating the solution of differential and integral equations [4]. Moreover, these ok!
polynomials have excellent error properties in the approximation of a globally smooth

1
function. However, the convergence rate of these polynomials deteriorates in a large
degree n [9].

2 Construction of Legendre polynomials

Legendre polynomials can be defined in numerous ways. This paper, however, provides
only one approach that is by construction as an orthogonal system.
All Legendre polynomials together form an orthogonal polynomial system with respect
to the weight function ω(x) = 1 over the interval [−1, 1]. That is, Pn is a polynomial of
degree n, such that Z 1
Pk (x)Pn (x)dx = 0 if k 6= n. (1)
−1
This determines the polynomials completely up to an overall scale factor, which is fixed
by the standardization Pn (1) = 1. This implies that P0 (x) must be equal to 1 and P1 (x)
must be orthogonal to P0 . Similarly, P2 (x) is determined by requiring orthogonality to
P0 and P1 , and so on. Pn is fixed by requiring orthogonality to all Pk with m < n. This
process is called Gram-Schmidt orthogonalization. The sequence of Legendre polynomials ok!
forms a complete orthogonal basis of L2 ([−1, 1]).
the
One of the explicit, compact expressions for Legendre Polynomials is given by Ro-
drigues’ formula in the form
1 dn 2
Pn (x) = (x − 1)n (2)
2n n! dxn
and the recurrence relation can be written as

(n + 1)Pn+1 (x) = (2n + 1)xPn (x) − nPn−1 (x). (3)

The first six Legendre polynomials are

P0 (x) = 1 P2 (x) = 21 (3x2 − 1) P4 (x) = 81 (35x4 − 30x2 + 3)


P1 (x) = x P3 (x) = 12 (5x3 − 3x) P5 (x) = 81 (63x5 − 70x3 + 15x).
C([-1,1]) for consistency of notation
Let C(−1, 1) be the set of all continuous functions defined on [−1, 1]. The Legen-
dre polynomials Pn (x) are orthogonal polynomials with respect to the inner product on
C([−1, 1]) given by Z 1
hf, gi = f (x)g(x)dx (4)
−1
and utilizes the L2 function space. This means, the Legendre polynomial approximation
is best in the sense of having the least L2 error.
With the lack of actual comparison,
it is immaterial to say "best".
It is better to describe at 2
"to have a property of minimizing L2-error"
3 Approximating Transcendental function using Legendre polynomials

In this section, transcendental function, particularly exponential function is approximated


using Legendre polynomials. The condition of orthogonality of Legendre Polynomials can
be written as 
1 0 if k 6= n
Z 
Pk (x)Pn (x)dx =
−1 2
if k 6= n.


2k+1

This property leads to this equation


Z 1
2n + 1
an = f (x)Pn (x)dx (5)
2 −1

that determines the values a0 , a1 , ..., an so that f (x) ≈ a0 P0 (x) + a1 P1 (x) + ... + an Pn (x)
a the
is the best polynomial approximation on [−1, 1]. To illustrate this, see example 1.
by a
Example 1. Approximate f (x) = ex using Legendre Polynomials. of degree at most 3
Solution:
Consider f (x) ≈ P (x) = a0 P0 (x) + a1 P1 (x) + ... + an Pn (x) on [−1, 1]. For this problem,
the first four Legendre polynomials were considered in approximating this function. Now,
using equation (5), we have
R1 R1
P (x) = P0 (x) 2(0)+1
2 −1
f (x)P0 (x)dx + P1 (x) 2(1)+1
2 −1
f (x)P1 (x)dx
R1 R1
+P2 (x) 2(2)+1
2 −1
f (x)P2 (x)dx + P3 (x) 2(3)+1
2 −1
f (x)P3 (x)dx

1
R1 R1 R1
= 2 −1
ex dx + x( 32 −1
ex xdx + 12 (3x2 − 1)( 52 −1
ex ( 21 (3x2 − 1))dx
R1
+ 21 (5x3 − 3x)( 72 −1
ex ( 12 (5x3 − 3x))dx

e2 −1 5(2e2 −14) 1 7(74−10e2 ) 1


= 2e
+ 3e x + 4e
( 2 (3x2 − 1)) + 4e
( 2 (5x3 − 3x))

e2 −1 5(2e2 −14) 7(74−10e2 )


= 2e
+ 3e x + 8e
(3x2 − 1) + 8e
(5x3 − 3x)
a
Hence, the best polynomial approximation of the function f (x) = ex is given by
e2 −1 5(2e2 −14) 7(74−10e2 )
P (x) = 2e
+ 3e x + 8e
(3x2 − 1) + 8e
(5x3 − 3x).

Figure 1 shows the graph of the function f (x) - (blue graph) and the approximated
polynomial P (x) - (green graph). It is clear in the graph shown in Figure 1 that the first
four Legendre polynomials can provide a polynomial that best fit over the interval [-1,1].
This means that the computed 2-norm error is approximately equal to 0 over this interval
[-1,1].

3
Figure 1: Graph of f (x) and P (x)
Some concluding statements would have helped.
References

[1] Carrier, G. F., Krook, M., & Pearson, C. E. (2005). Functions of a complex variable:
theory and technique. Society for Industrial and Applied Mathematics.

[2] Cheney, E. W. (1966). Introduction to Approximation Theory, McGraw-Hill, New


York, MR0222517 (36:5568).

[3] Cohen, M. A., & Tan, C. O. (2012). A polynomial approximation for arbitrary func-
tions. Applied Mathematics Letters, 25(11), 1947-1952.

[4] Davis, P. J., & Rabinowitz, P. (2007). Methods of numerical integration. Courier
Corporation.

[5] Dunster, T. M. (2009). Legendre and Related Functions. Rainbow over Woolsthorpe
Manor, 351.

[6] Greiner, W. (2012). Classical electrodynamics. Springer Science & Business Media.

[7] Liu, Y. (2013). Application of Legendre polynomials in solving Volterra integral equa-
tions of the second kind. Appl. Math., 3(5), 157-159.

[8] Phillips, A. C. (2013). Introduction to quantum mechanics. John Wiley & Sons.

[9] Wang, H., & Xiang, S. (2012). On the convergence rates of Legendre approximation.
Mathematics of Computation, 81(278), 861-877.

You might also like