MISN-0-485

FOURIER INTEGRALS: PART II

Math Physics
Project PHYSNET Physics Bldg. Michigan State University East Lansing, MI

FOURIER INTEGRALS: PART II by R. D. Young, Dept. of Physics, Illinois State Univ. 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 3. Supplementary Notes: I a. The Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 b. Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 4. Supplementary Notes: II a. The Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 b. Representations of (x) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 c. Properties of the Dirac delta function . . . . . . . . . . . . . . . . . . . . 8 d. Problems on the Dirac delta function . . . . . . . . . . . . . . . . . . . . 9 d. Supplemental Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9

ID Sheet: MISN-0-485 THIS IS A DEVELOPMENTAL-STAGE PUBLICATION OF PROJECT PHYSNET The goal of our project is to assist a network of educators and scientists in transferring physics from one person to another. We support manuscript processing and distribution, along with communication and information systems. We also work with employers to identify basic scientic skills as well as physics topics that are needed in science and technology. A number of our publications are aimed at assisting users in acquiring such skills. Our publications are designed: (i) to be updated quickly in response to eld tests and new scientic developments; (ii) to be used in both classroom and professional settings; (iii) to show the prerequisite dependencies existing among the various chunks of physics knowledge and skill, as a guide both to mental organization and to use of the materials; and (iv) to be adapted quickly to specic user needs ranging from single-skill instruction to complete custom textbooks. New authors, reviewers and eld testers are welcome. PROJECT STAFF Andrew Schnepp Eugene Kales Peter Signell Webmaster Graphics Project Director

Title: Fourier Integrals: Part II Author: R. D. Young, Dept. of Physics, Illinois State Univ. Version: 2/1/2000 Length: 2 hr; 15 pages Input Skills: 1. Vocabulary: Fourier integral expansion of f (x), Fourier transform of f (x), Fourier cosine transform and inverse cosine transform, Fourier integral theorem. 2. Unknown: assume (MISN-0-484). Output Skills (Knowledge): K1. Dene or explain each of the terms and concepts as follows: Parsevals identities for Fourier transforms, convolution of f (x) and g(x), convolution theorem for Fourier transforms, Dirac delta function. Output Skills (Rule Application): R1. Evaluate innite integrals using Parsevals identity for Fourier transforms. R2. Solve integral equations using the convolution theorem. R3. Solve problems involving the Dirac delta function. External Resources (Required): 1. G. Arfken, Mathematical Methods for Physicist, Academic Press (1995). 2. Schaums Outline: Murray Spiegel, Theory and Problems of Advanced Mathematics for Scientists and Engineers, McGraw-Hill Book Co. (1971). Evaluation: Stage B0

ADVISORY COMMITTEE D. Alan Bromley E. Leonard Jossem A. A. Strassenburg Yale University The Ohio State University S. U. N. Y., Stony Brook

Views expressed in a module are those of the module author(s) and are not necessarily those of other project participants. c 2001, Peter Signell for Project PHYSNET, Physics-Astronomy Bldg., Mich. State Univ., E. Lansing, MI 48824; (517) 355-3784. For our liberal use policies see: http://www.physnet.org/home/modules/license.html.

MISN-0-485 9/29/99, pss; 1/4/2000, pss

MISN-0-485

3. Supplementary Notes: I
The Convolution Theorem 3a. The Theorem. The convolution theorem is stated in Spiegel, page 203, for arbitrary functions as long as all of the integrals exist. If the functions involved have certain symmetry properties the convolution theorem takes on alternative forms. Two examples are included in Problem 15.5.1 of Arfken which you are asked to prove. Thus: 1) If f and g are both odd functions, the convolution theorem reads 1 2

FOURIER INTEGRALS: PART II by R. D. Young, Dept. of Physics, Illinois State Univ. 1. Introduction
This unit is a continuation of the preceding unit. The new topics include Parsevals identity for Fourier transforms, the convolution theorem, and the Dirac delta function.

f (u)g(x u) du =

Fs ()Gs () cos x d
0

(1)

2. Procedures
1. Read section 15.5 of Arfken. Read pages 202-203 of Spiegel. 2. Underline in the text or write out the denitions and explanations of the terms and concepts of Output Skill K1 using an explanatory equation if necessary. One or two sentences should be sucient. 3. Read part I of the Supplementary Notes, on the Convolution Theorem. 4. Read these Solved Problems in Spiegel: 8.7 (Evaluation of innite integrals using Parsevals Identity) 8.9 (Solving integral equations using the convolution theorem) 5. Solve these Supplementary Problems in Spiegel: 8.23 (Evaluation of innite integrals using Parsevals Identity: Use the result of problem 8.16.) 8.25 (Convolution Theorem: Use the result of problem 8.16 and possibly the trig identity relating sin2 to cos.) Solve this problem in Arfken: 15.5.1 (Convolution Theorem) 6. Read part II of the Supplementary Notes, on the Dirac delta function. Solve the problems at the end of the Supplementary Notes.

which is the rst result of Problem 15.5.1 in Arfken. The convolution of f and g [h(x) = f g], on the other hand, must be an even function. To see this consider 1 h(x) = 2 1 f (u)g(x u) du = 2 1 = 2 1 = 2 so h(x) = +h(x). Thus 1 2h(x) = 2 2

f (u)g(x + u) du

f (u)g(x u) du f (u)g(x u) du

Fs ()Gs () cos x d
0

and 2
0

h(x) cos x dx =

dFs ()Gs ()
0 0

dx cos x cos x .

(2)

But it can be shown that 2

cos x cos x dx = ( )

(3)

MISN-0-485

MISN-0-485 Now, from the denition of the functions, Hc () = 2 1 2

where ( ) is the Dirac delta function. (See the next section of the Supplementary Notes. Then, try to prove Eq. 3). Thus, the alternative to Eq. 18 of Spiegel results from Eqs. 2 and 3: Hc () = Fs ()Gs () . 2) If f and g are both even functions, the convolution theorem reads 1 2

(4)

1 cos x dx = x2 + b2 b 1 b e . 2b

cos bx dx x2 + 1

Hc () = Likewise, (5) Gc () =

f (u)g(x u) du =

Fc ()Gc () cos x d
0

which is the second result of Problem 15.5.1 in Arfken. Again the convolution of f and g (h = f g) is an even function. Using the same analysis as above gives Hc () = Fc ()GC () as the alternative to Eq. 18 of Spiegel. 3b. Example. The solution of Solved Problem 8.9 of Spiegel is actually not quite correct since Fourier cosine transforms must be used. A proper solution rests on the use of Eqs. 5 and 6 above. To carry out this correct solution rewrite the integral equations as

cos x dx = x2 + b2

21 a

cos x dx x2 + 1

Gc () = Thus

1 a e . 2a

(6)

1 a (ba) e . Fc () = Hc ()/Gc () = 2 b 2 Fc () cos u d 0 1 a (ba) = e cos u d b 0 (b a) a . = b u2 + (b a)2

Now f (u) =

f (u) du 1 = 2 , 0 < a < b. 2 + a2 (u x) x + b2

The right-hand side is an even function of x. Therefore, the left-hand side must be an even function of x. Thus can only be true if f (u) is even since

f (u) du = (u + x)2 + a2 f (u) du = (u x)2 + a2

f (u) du = (u + x)2 + a2 f (u) du . (u x)2 + a2

4. Supplementary Notes: II
The Dirac delta function 4a. The Function. eq. (15.20) of Arfken, f (x) =

Consider eq. (4) on page 201 of Spiegel or

Thus, if g(u) = 1/(u2 + a2 ) and the convolution of f and g (dened as h = f g) is given by 1 1 , h(x) = x2 + b2 2 the Convolution Theorem for these functions reads: Hc () = Fc ()Gc () .
7

1 2 dx

dx

df (x )e(xx ) de(xx ) f (x )

1 2

dx K(x x )f (x ) .
8

MISN-0-485

MISN-0-485 Here is a graph of sin N x/x:

I have purposefully written this equation in a way which shows that f(x) can be written as an integral over all x with f (x) appearing in the integrand. Thus, the function multiplying f (x), that is K(x x ), in the integrand must have some rather strange properties. K(x x ) must be very small for x = x so that contributions to the integrand from this portion of the integration region are small while K(x x ) must be large when x = x so that contributions to the integral from the integrand when x = x are enhanced. The function K(x x ) is given a special name and notation. K(xx ) is written (xx ) and called the Dirac delta function. The dirac delta function has these two properties, by denition: 1. (x) = 0 if x = 0. 2.
b a

__ sin Nx px N _ p

x __ -4 p N _ -p N p _ N 2p __ N 3p __ N 4p __ N

(x) dx = 1 when a < x < b.

The zeroes occur at xn/N while the function has a maximum value N/ at x = 0. The function approaches zero as x . As N , the height of the central peak goes to innity, and the zeroes of the function approach x = 0 and get innitesimally close together. Thus limN = if x = 0; = a rapidly oscillating quantity (if x = 0) which when multiplied by a smooth function and integrated gives zero. Also, 1 2

Obviously, (x x ) satises these properties: 1. (x x ) = 0 if x = x . 2.

(x x ) dx = 1 when a < x < b .

The proof is accomplished by letting x x x so that dx dx and a = a + x and b = b + x . 4b. Representations of (x). A representation of the Dirac delta function is an explicit algebraic formula which satises the two criteria above. Such representations include some type of limiting process either explicitly or implicitly. A representation of the Dirac delta function is (x) = Suppose x = 0, then (x) = 1 limN 2
N

sin N x x

1 2

dx

dex = limN = limN

ex d .

sin N x dx x

ex d
N N N

1 sin N x = limN . x
9

1 ex limN 2 x

1 = 1. The interchange of the limiting process and the integration although not valid if the integral is an ordinary Riemann integral is allowed when the Dirac delta function is involved. It is obvious then that the Dirac delta function is not a function in the ordinary sense as used in introductory calculus. It is in fact a more general quantity, that is, a distribution. The mathematical step of interchanging the limit and integration above can be justied by appealing to the theory of distributions as developed by Laurent Schwartz. Essentially, the idea is to ignore the innite jump
10

MISN-0-485

MISN-0-485

discontinuity in (x) at x = 0 when performing calculations. For example, consider this alternative representation of the Dirac delta function: (x) = Examine these diagrams: 1 d2 |x|, < x < + . 2 dx2

The innite jump discontinuity at x = 0 is ignored when performing the integral. 4c. Properties of the Dirac delta function. 1) x(x) = 0 for < x < +. Proof:

|x| +1 x

d |x| dx

x(x) = 0 if x = 0 since (x) = 0 if x = 0. x(x) = 0 if x = 0 since 0 (0) = 0 ignoring the innite jump discontinuity at x = 0 in (x).

x -1

2) f (x)(x a) = f (a)(x a) Proof: f (x)(x a) = [f (x) f (a)](x a) + f (a)(x a). But, (x a) = 0 if x = a, and [(f (x) f (a)] = 0 if x = a. Thus, f (x)(x a) = f (a)(x a) .

and

d2 |x| dx 2
3)

Note:

f (x)(x a) dx = f (a)

(x a) dx = f (a).

f (x) (x a) dx = f (a) where the prime means dierentiation with respect to x. Proof: d [f (x)(x a)] = f (x)(x a) + f (x) (x a) . dx So, d [f (x)(x a)] = f (a)(x a) + f (x) (x a) . dx Integrating,

where I have separated the 0 value of d2 / dx2 |x| for x = 0 from the x-axis for clarity. Obviously, 1 d2 |x| = 0 if = 0. 2 dx2 Also, 1 2 =
N N

d [f (x)(xa)] dx = dx

f (a)(xa) dx+

f (x) (xa) dx,

d2 d 1 |x| |x| dx = limN dx2 2 dx

but d [f (x)(x a)] = limN [(f (x)(x a)]N = limN 0 = 0 . N dx So,

1 limN [1 (1)] = limN 1 = 1 . 2

MISN-0-485 0 = f (a) +

9 f (x) (x a) dx .

4) (x) = +(x) . Proof: (x) = (x) = 0 if x = 0. Also,

(x) dx =

(+x) dx =

(x) dx = 1 .

4d. Problems on the Dirac delta function. 1) Prove that if a is real and a = 0, then (ax) = 2) Show that: [(x a)(x b)] = 1 [(x a) + (x b)] . |a b| 1 (x) . |a|

3) Show that each of these is an alternative representation of the Dirac delta function: 2 1 1 (x) = lim 0 ex / (x) = 1 lim
0

x2 +

4d. Supplemental Reading. (Optional) For more discussion on the Dirac delta function, read section 1.15 of Arfken.

Acknowledgments
The author would like to thank Illinois State University for support in the construction of this lesson. Preparation of this module was supported in part by the National Science Foundation, Division of Science Education Development and Research, through Grant #SED 74-20088 to Michigan State University.

13

14

15