we know that:

T = (1/3)(TT* - 2I)

so T* = [(1/3)(TT* - 2I)]* = (1/3)(TT*)* - (2I)* = (1/3)((T*)*T*) - 2I*

= (1/3)(TT* - 2I) = T.

Printable View

- October 3rd 2011, 05:48 AMDevenoRe: positive definite problem
we know that:

T = (1/3)(TT* - 2I)

so T* = [(1/3)(TT* - 2I)]* = (1/3)(TT*)* - (2I)* = (1/3)((T*)*T*) - 2I*

= (1/3)(TT* - 2I) = T. - October 3rd 2011, 06:07 AMtransgalacticRe: positive definite problem
ok so we got the T*=T part

now in order o find the eigenvalues

so k=2 and k=1 are eigen values and they are positive

but i am not sure if i have done it correctly

because by definition k is an eigen value if T(v)=kv

or if k nullifies the caracteristic polinomial

i cant show neither of those here

i just said intuitivly that 1 ,2 are eigenvalues

??

i can replace T=Av

but it doesnt go anywhere - October 3rd 2011, 07:37 AMDevenoRe: positive definite problem
x^2 - 3x + 2 is a polynomial that T satisfies. so the minimal polynomial m(x) divides this. there are precisely 3 possibilities:

1) m(x) = x - 2

2) m(x) = x - 1

3) m(x) = (x - 2)(x - 1).

1) means T = 2I. 2) means T = I. in both of these cases, T is positive-definite.

if we have case 3), we know that the minimal polynomial for T has the same roots as the characteristic polynomial for T. thus T has eigenvalues 1,2 > 0, so T is positive-definite.

you are correct when you say we don't know what the characteristic polynomial of T is, we aren't given enough information. and you can easily verify that I and 2I both satisfy the given equation, so T might be one of those 2 matrices. but even if it isn't, we still know it HAS to be positive-definite, because we know all the roots of the characteristic polynomial, even though we don't know what the polynomial actually is.