You are on page 1of 4

18.

03(4) at ESG

Spring, 2003

Practice Problems I - Properties of Hermitian Matrices


For scalars we often consider the complex conjugate, denoted z in our notation. For matrices, we often consider the Hermitian Conjugate of a matrix, which is the transpose of the matrix of complex conjugates, and will be denoted by A (its a physics thing). Specically, if the entries of A are ajk , the entries of A are akj . If A is an m n matrix (m rows, n columns), A is an n m matrix. In longer, more explicit form, a11 a21 . . . am1 a12 a22 . . . am2 .. . a1n a11 a21 a22 . . . a2n .. . am1

if

A=

amn

a2n ,

a12 = A . . .

a1n

amn

am2 .

A square matrix is Hermitian if A = A and Skew-Hermitian if A = A . (a) To see how this works in a simple case, let x C m be a column vector (an m 1 matrix) of complex elements xj , j = 1 . . . m, and dene the square of the norm of x to be
m

|x | = x1 x1 + x2 x2 + xm xm =
j=1

xj xj .

Show that this can be expressed as |x | = x x .


2

The rest of this practice problem will consider only 2 2 matrix operators and 2 1 column vectors and their 1 2 Hermitian cojugates. If a 2 2 matrix is Hermitian, well denote this matrix as H= a c 1 c , b

where a and b are real and c is complex (real, imaginary or neither). Note that we could have put the overline representing scalar complex conjugation in the lower left instead of the upper right. (b) Show that the eigenvalues of H are real. Do this two ways, rst by direct computation of the roots of p() = det (H I) and then by considering the product f H f for f an eigenvector of H. The point of doing this two ways is that the rst method sort of craps out for larger matrices, while the second is valid for Hermitian matrices of any size. (c) Show that if the eigenvalues of H are equal, then H is a real scalar multiple of I, and hence not worth special consideration (every vector an eigenvector). Show that if the eigenvalues are not equal, then the eigenvectors are orthogonal; in this context, two vectors x and y are orthogonal if x y = 0 (this is consistent with part (a) above.) Again, do this two ways, rst by direct computation (there are a few tricks to look for) and then by considering the product f H f where f and f
2 1 1 2

are eigenvectors corresponding to dierent eigenvalues. As in (b) above, the second method is valid for Hermitian matrices of any size. (d) Show that the sum of any two Hermitian matrices is Hermitian (if they are the same size, of course), H2 is Hermitian, but for two arbitrary Hermitian matrices (if they are the same size, of course), the product is not necessarily Hermitian. Speculate about H3 and higher powers, invoking the Cayley-Hamilton Theorem (or demonstrating the validity of the theorem directly). (e) You knew this was coming: Find exp (H t). In doing parts (b) and (c) using direct computation, youve found that the general form for the eigenvalues and eigenvectors is a bit of a mess and diagonalization would be unpleasant, so heres what well do. Express H= a+b d I+ c 2 c d

where d = (a b)/2. The rst matrix on the right above commutes with the second matrix (Im tempted to say clearly, since the rst is a scalar multiple of the identity, but I wont), so exp (H t) = e(a+b)t/2 exp 2 d c c d t .

Find

c d , express your delight in a seemly manner, and go from there. A sugc d gestion is to simplify the form by introducing the parameter 2 = d2 + c c.

II - Manipulating Solutions
(A) It is known that two functions y1 and y2 satisfy a linear dierential equation of the form y + P (t) y + Q(t) y = 0. Given y1 and y2 , can you determine P (t) and Q(t), and if so, what are they in terms of y1 , y2 , and their derivatives? (B) Find, if possible, P and Q when y1 = et , y2 = t et . (C) Find, if possible, P and Q when y1 = cos(2t), y2 = 1 2 cos2 t.

III - Dierential Operators as Matrices


Consider the set of scalar functions x1 = cos2 t, Let yj = x2 = sin2 t, x3 = sin 2t.

dxj , j = 1, 2, 3, and denote dt y1 y2 , y= y3

x1 x = x2 . x3

Find a constant matrix D such that y = D x . Find the rank of D and hence the dimension of the space spanned by the yi , and explain briey. Find D2 and D3 , and identify the constant scalar c such that D3 = c D.

IV - Rotation, Rotation, Rotation


(a) Find all real 2 2 matrices B such that if x has real components, |B x | = |x | for any x , where |x | is the Euclidean norm. These are the 2 2 rotation matrices. (b) Show that the matrix that represents T : R2 R2 , T rotation matrix. (Note: if you nd this really easy, youre right.) 3 x y = x y is a

(c) Find the determinant of an arbitrary matrix B of the form you found in part (a). (d) Describe the rotations in three dimensions represented by 0 1 1 0 G1 = 0 0 0 0, 1 0 0 G2 = 1 0 1 1 0, 0 0 1 0 0 G3 = 0 0 1. 0 1 0

Pick any two of the above matrices and show that they dont commute. Find exp (G1 t) , exp (G2 t) , exp (G3 t) .

(There are several ways to do this, of course. Hints available on request). (e) Explain why there is no rotation matrix F such that x x F y = y . z z If you cant nd the algebra proof, explain in words.

You might also like