Orthogonal Matrices and Symmetric Matrices

• November 18th 2012, 02:11 PM
Kyo
Orthogonal Matrices and Symmetric Matrices
Hi, I have two problems that I'm really struggling with:

1) If A is an n x n matrix with real eigenvalues, show that A = B + C where A is symmetric and C is nilpotent. Hint: Use the fact that an n x n upper triangular matrix U with zeroes on the main diagonal satisfies U^n = 0.

What I think/know:

I know that there exists a P, such that P inverse = P transpose (P^T) where P^(T)AP = U ; where U is an upper triangular matrix. I tried taking a power of n of the equation:
A^n = (PUP^(T)) = (A + B)^n , but that doesn't seem to get me anywhere. Please show me how else I can approach the problem and a kick start!

2) Let P denote an n x n matrix.
a) If P is orthogonal, Show |Px| = |x| for every column x in R^n.
b) If |Px| = |x|, show that P is orthogonal.

What I think/know:

I know that since P is orthogonal, the rows and columns of P are orthonormal. I tried multiplying them out as P = [ C1 C2...Cn ] and x = [xi]; but when I took the norm of it, I couldn't get anywhere. Perhaps my block multiplication is wrong - I have a hard time wrapping my head around "seeing" 'n' terms of the columns of P times the column of x. Help please! Also - I assumed that part b) would be similar and maybe once I got a), I could try b) alone. If it seems like I assumed incorrectly, please help as well!
• November 18th 2012, 05:03 PM
Deveno
Re: Orthogonal Matrices and Symmetric Matrices
why not do this:

for i ≥ j define B = (bij) by bij = aij and for i < j, bij = aji.

clearly B is symmetric.

now let C = A - B. on or below the diagonal (i ≥ j), cij = aij - bij = 0.

thus C is upper triangular with all 0's on the diagonal.

example:

$A = \begin{bmatrix}1&4&7\\-5&3&2\\1&0&3 \end{bmatrix}$.

then:

$B = \begin{bmatrix}1&-5&1\\-5&3&0\\1&0&3 \end{bmatrix};\ C = \begin{bmatrix}0&9&6\\0&0&2\\0&0&0 \end{bmatrix}$.

it seems to me we can do this no matter what the eigenvalues of A are.

for 2: another way to write |x| is as the scalar:

√(xTx). thus |Px| = √((Px)TPx) = √(xT(PTP)x)

what is PTP?
• November 18th 2012, 08:31 PM
Kyo
Re: Orthogonal Matrices and Symmetric Matrices
Thank you very much Deveno, your suggestion for #2 was great and I managed to figure out 'a' and 'b' from it!

I am still, however, confused about #1, am I not losing full generality when I define B? And am I "allowed" to do such a thing? Sorry if this feels very fundamental - but I've not taken a class on formal math proofs or logic.

And thanks again in general. You've helped me tons and I don't believe I've said that yet.
• November 18th 2012, 08:57 PM
Deveno
Re: Orthogonal Matrices and Symmetric Matrices
how? we are given a matrix A, we are forming B by considering "half" of A, and "making up the difference" with C.
• November 19th 2012, 08:45 PM
Kyo
Re: Orthogonal Matrices and Symmetric Matrices
Ah, thank you. I get it now.