MATH
21 B
Mathematics Math21b Spring 2009
Linear Algebra and Differential Equations
Exhibit: Linear algebra: discrete Fourier transform
Course Head: Oliver Knill
Office: SciCtr 434


Remember the examples, where you had to find the eigenvalues and eigenvectors of a circular matrix like
A=  0 1 0 0 0 0 0 0 0 0 
    0 0 1 0 0 0 0 0 0 0 
    0 0 0 1 0 0 0 0 0 0 
    0 0 0 0 1 0 0 0 0 0
    0 0 0 0 0 1 0 0 0 0 
    0 0 0 0 0 0 1 0 0 0 
    0 0 0 0 0 0 0 1 0 0
    0 0 0 0 0 0 0 0 1 0
    0 0 0 0 0 0 0 0 0 1 
    1 0 0 0 0 0 0 0 0 0 
This matrix has as the eigenvalues the roots of the characteristic polynomial det(A-x) = x10-1. These roots are called roots of unity. We have computed also the eigenvectors. If e(k) = exp(2Pi i k/10) are the eigenvalues, then the eigenvectors the columns of the matrix (remember the multiplication table)



The matrix S diagonalizing a circular matrix. Click for a larger picture.
S = 1     1     1     1     1     1     1     1     1     1 
    e(1)  e(2)  e(3)  e(4)  e(5)  e(6)  e(7)  e(8)  e(9)  e(10)            
    e(2)  e(4)  e(6)  e(8)  e(10) e(12) e(14) e(16) e(18) e(20) 
    e(3)  e(6)  e(9)  e(12) e(15) e(18) e(21) e(24) e(27) e(30) 
    e(4)  e(8)  e(12) e(16) e(20) e(24) e(28) e(32) e(36) e(40) 
    e(5)  e(10) e(15) e(20) e(25) e(30) e(35) e(40) e(45) e(50) 
    e(6)  e(12) e(18) e(24) e(30) e(36) e(42) e(48) e(54) e(60) 
    e(7)  e(14) e(21) e(28) e(35) e(42) e(49) e(56) e(63) e(70) 
    e(8)  e(16) e(24) e(32) e(40) e(48) e(56) e(64) e(72) e(80) 
    e(9)  e(18) e(27) e(36) e(45) e(60) e(63) e(72) e(81) e(90) 
 /sqrt(10)
We divided by sqrt(10) so that each column has length one and so that they form an orthonormal eigen basis of the matrix A. The matrix S produces a transformation which is called the discrete Fourier transform here in the case n=10. It is important in signal processing, sound and image manipulation and compression. Note as in midterm practice problems or homework used frequently that because A is orthogonal, S also diagonalizes AT = A-1 and so L = A+AT-2 which is a discretization of the second derivative D2. In the same way as Fourier series diagonalized D2, the discrete Fourier transform diagonalizes the discrete second derivative L. The eigenvalues are e(k) + e(-k) - 2 = 2 cos(2Pi k/10) - 2.
A[n_]:= Table[ If[Mod[j-i,n]==1,1,0],{i,n},{j,n}]
S[n_]:= Table[ Exp[ 2Pi I (i-1) (j-1)/n],{i,n},{j,n}]/Sqrt[n];
Chop[N[Conjugate[Transpose[S[10]]].A[10].S[10]]]
Please send questions and comments to math21b@fas.harvard.edu
Math21b (Exam Group 1)| Oliver Knill | Spring 2009 | Department of Mathematics | Faculty of Art and Sciences | Harvard University