MATH4141/MATH6651/PHYS5070


FW02


Assignment 4 Due date: Friday, Nov. 8




1. Show that the following matrices are orthogonal:

(a) the elementary reflectorwhere;

(b) the rotation matrix P which differs from the identity matrix in four elements, viz,:

pii = pjj = c

pij = -pji = s

for i not equal to j and c2 + s2 = 1.


2. The Jacobi method uses a rotation matrices of the form given in question 1(b) to carry out similarity transformations of a symmetric matrix A, i.e. a sequence of matrices are generated where A(i+1) = PTA(i)P and A(1) = A.

(a) If = PTAP show that the elements of are identical to those of A except for

a´ki = caki - sakj = a´ik

a´kj = cakj + saki = a´jk

for k not equal to i or j and

a´ii = c2aii + s2ajj - 2csaij

a´jj = c2ajj + s2aii + 2csaij

a´ij = (c2 - s2)aij + cs(aii - ajj) = a´ji

where amn are the elements of A. Note that is symmetric if A is.


(b) The basis of the Jacobi method is to choose c (and s) such that a´ij = 0. If S is the sum of the off diagonal elements of A, i.e.

S =

and S´ is similarly defined for show that S´ = S - 2|aij|2 when a´ij = 0, using the results of part (a). Since

this shows that repeated Jacobi transformations for a sequence of (i, j) values will result in a sequence of matrices whose off diagonal elements tend to zero. The diagonal elements will then tend to the eigenvalues of A.


(c) In order to have a´ij = 0 show that c = [1 + t2]-1/2, s = t c and t can be chosen to be



3. The following IMSL routines implement the Housholder plus QR method to find all the eigenvalues and eigenvectors of a real, symmetric matrix:

i) imsl_d_eig_sym (C version);

ii) DEVCSF (double precision Fortran version).


Write a program to use one of these routines to find all the eigenvalues and eigenvectors of the following matrices:


(a) ;


(b) ;


(c) .


Format your output to display each of the eigenvalues with the corresponding eigenvector.






4. If a matrix A has pseudoinverse A+ prove it has the following properties:

i) AA+A = A

ii) A+AA+ = A+

iii) (AA+)T = AA+

iv) (A+A)T = A+A

Hint: If the singular value decomposition of A is UWVT show first that W has these four properties. These properties are known as the Penrose properties.


5. The amount of a product p of a chemical reaction present at time t is measured. The results are given in the following table:

t

p

0.100

4.066

0.200

4.492

0.300

4.901

0.400

5.296

0.500

5.704


We want to find a linear least-squares fit to these data of the form ct + d = p where c and d are unknowns to be determined. If we put the data in the form of a system of linear equations

Ax = p where A is a 5 by 2 matrix whose first column consists of the values of t and the second column has all elements equal to one, x is the vector of unknowns [c, d]T and p is the vector of p values then the least-squares solution is given by A+ p. Use the IMSL routines DLSGRR or lin_svd_gen to find the pseudoinverse of A. Use this to find the least squares solution to the system of linear equations given above.