Orthogonality relations of functions e^(2 pi i n x)

Click For Summary

Homework Help Overview

The discussion revolves around deriving the orthogonality relations of the functions e^(2 pi i n x) for integer values of n, which are periodic functions with a period of 1. Participants explore the conditions under which these functions are orthogonal in the context of square integrable functions.

Discussion Character

  • Exploratory, Assumption checking, Mathematical reasoning

Approaches and Questions Raised

  • Participants discuss the definition of orthogonality and the necessary integral conditions to demonstrate it. There are questions regarding the appropriate limits for integration and the implications of different intervals on the orthogonality results.

Discussion Status

The discussion is active, with participants providing various insights and calculations regarding the orthogonality of the functions. Some have suggested specific intervals for integration, while others are questioning the validity of results obtained on different intervals. There is a recognition of the need for clarity on how to derive the relations in a general context.

Contextual Notes

Participants are considering the implications of using different integration limits, such as [-1/2, 1/2] versus [0, 1], and how these affect the orthogonality conditions. There is also mention of the need to interpret the problem statement regarding deriving orthogonality relations in a general sense.

Logarythmic
Messages
277
Reaction score
0
I know that the functions e^{2 \pi inx} for n \in \mathbb{Z} are a base in the space of functions whith period 1. How do I derive the orthogonality relations for these functions?
 
Physics news on Phys.org
By using the definition of "orthogonality"! For functions in L2 (square integrable functions- \int f^2(x)dx exists), the functions over which the Fourier series are defined, the inner product is \int f(x)g(x)dx. To show that e^{2\pi inx} for n any integer are orthogonal show that
\int e^{2\pi inx}e^{2\pi imx} dx is 0 as long as m \ne n.
 
What limits should I use for integration?
 
I don't get it. The integral is never zero with m different from n. I guess I should use some weight function?
 
Logarythmic said:
What limits should I use for integration?
Looks to me like -1/2 to +1/2 will do nicely.
Logarythmic said:
I don't get it. The integral is never zero with m different from n.
Can you show an example?
 
\int_{- \frac{1}{2}}^{\frac{1}{2}}f_n^*(x)f_m(x)dx = \int_{- \frac{1}{2}}^{\frac{1}{2}}e^{-2 \pi inx}e^{2 \pi imx}dx = \int_{- \frac{1}{2}}^{\frac{1}{2}}e^{2 \pi ix(m-n)}dx =

= \left[ \frac{e^{2 \pi ix(m-n)}}{2 \pi i(m-n)} \right]_{- \frac{1}{2}}^{\frac{1}{2}} = \frac{e^{\pi i (m-n)} - e^{- \pi i (m-n)}}{2 \pi i (m-n)} = \frac{1}{\pi (m-n)} \sin{\pi (m-n)} = 0

Because m and n are integers. But this is only a solution for [-1/2,1/2] ? What if the interval is [0,1]? The I get

I = \frac{e^{2 \pi i (m-n)} - 1}{2 \pi i (m-n)}

and this is only zero for m = n...

So this shows that the functions are orthogonal only on the interval [-1/2,1/2] and this is my solution. Am I right?
 
Last edited:
Logarythmic said:
\int_{- \frac{1}{2}}^{\frac{1}{2}}f_n^*(x)f_m(x)dx = \int_{- \frac{1}{2}}^{\frac{1}{2}}e^{-2 \pi inx}e^{2 \pi imx}dx = \int_{- \frac{1}{2}}^{\frac{1}{2}}e^{2 \pi ix(m-n)}dx =

= \left[ \frac{e^{2 \pi ix(m-n)}}{2 \pi i(m-n)} \right]_{- \frac{1}{2}}^{\frac{1}{2}} = \frac{e^{\pi i (m-n)} - e^{- \pi i (m-n)}}{2 \pi i (m-n)} = \frac{1}{\pi (m-n)} \sin{\pi (m-n)} = 0

Because m and n are integers. But this is only a solution for [-1/2,1/2] ? What if the interval is [0,1]? The I get

I = \frac{e^{2 \pi i (m-n)} - 1}{2 \pi i (m-n)}

and this is only zero for m = n...

So this shows that the functions are orthogonal only on the interval [-1/2,1/2] and this is my solution. Am I right?
Factor an e^{\pi i (m-n)} out of your I. Does that help?
 
Then I is zero if e^{\pi i (m-n)} = 0 or if -\pi i(m-n) = 2.
The first one gives m-n = \frac{2k+1}{2} and the second one gives m - n = \frac{2i}{\pi}?
 
Logarythmic said:
Then I is zero if e^{\pi i (m-n)} = 0 or if -\pi i(m-n) = 2.
The first one gives m-n = \frac{2k+1}{2} and the second one gives m - n = \frac{2i}{\pi}?
OK. This is it.

I = \frac{e^{2 \pi i (m-n)} - 1}{2 \pi i (m-n)}

I = e^{\pi i(m - n)} \left[ {\frac{{e^{\pi i(m - n)} - e^{ - \pi i(m - n)} }}{{2\pi i(m - n)}}} \right] = e^{\pi i(m - n)} \left[ {\frac{{\sin \pi (m - n)}}{{\pi (m - n)}}} \right]

Compare this to your result for the [-1/2, 1/2] limits. In the form you had it, you cannot say I is zero when m = n; it is indeterminate. In this form, you can see the sine term is zero, but (sinx)/x is 1 for x = 0. When m = n the exponential is 1 and (sinx)/x is 1.
 
Last edited:
  • #10
So how do I do this on a general interval [a,b] where b - a = 1? How should I interpret the problem "Derive the orthogonality relations for these functions"?
 
  • #11
Logarythmic said:
So how do I do this on a general interval [a,b] where b - a = 1? How should I interpret the problem "Derive the orthogonality relations for these functions"?
I think the factoring step I did will work for any interval. Give it a try.
 
  • #12
So, if I use a general interval [a,b] with b-a=1, I get

I = e^{\pi i (m-n)(2a+1)} \frac{\sin{\pi (m-n)}}{\pi (m-n)} = \left\{\begin{array}{cc}0,&\mbox{ if }m \ne n\\1, & \mbox{ if } m = n\end{array}\right

and everyone is happy.
 
  • #13
Logarythmic said:
So, if I use a general interval [a,b] with b-a=1, I get

I = e^{\pi i (m-n)(2a+1)} \frac{\sin{\pi (m-n)}}{\pi (m-n)} = \left\{\begin{array}{cc}0,&\mbox{ if }m \ne n\\1, & \mbox{ if } m = n\end{array}\right

and everyone is happy.
I can't speak for everyone, but it works for me.
 
  • #14
Thanks for your help. =)
 

Similar threads

  • · Replies 3 ·
Replies
3
Views
2K
Replies
7
Views
2K
Replies
4
Views
2K
  • · Replies 21 ·
Replies
21
Views
2K
Replies
12
Views
3K
Replies
4
Views
1K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
3K
Replies
1
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K