Graduate Can a Non-Diagonal Hermitian Matrix be Diagonalized Using Unitary Matrix?

Click For Summary
Every Hermitian matrix is unitarily diagonalizable, but the discussion explores whether a non-diagonal Hermitian matrix can be diagonalized using a unitary matrix that does not consist solely of its eigenvectors. An example is presented with a 2x2 matrix where the unitary matrix U has eigenvalues of ±1, and the diagonal matrix D has distinct eigenvalues not equal to ±1. The identity matrix is deemed trivial as it is already diagonal. The conversation raises questions about specific cases and conditions under which such diagonalization can occur, particularly focusing on the construction of matrices A, U, and D.
LagrangeEuler
Messages
711
Reaction score
22
Every hermitian matrix is unitary diagonalizable. My question is it possible in some particular case to take hermitian matrix ##A## that is not diagonal and diagonalize it
UAU=D
but if ##U## is not matrix that consists of eigenvectors of matrix ##A##. ##D## is diagonal matrix.
 
Physics news on Phys.org
U=U^{-1}
I am not sure of such a specific case.
 
Yes here I am talking about case ##U=U^{-1}##. I am also not sure. But for me it is interesting.
 
As an example in 2X2 marix
U=<br /> \begin{pmatrix}<br /> k &amp; \alpha \\<br /> \beta &amp; -k \\<br /> \end{pmatrix}<br />
where
\alpha \beta = 1-k^2
would produce non diagonal matrix A from diagonal matrix D which has two different eigenvalues with A=UDU.
 
Last edited:
You can give me 2x2 example. But specify ##A##, ##U## and ##D##. Because ##U## still possibly can be formed of eigenvectors of ##A## in your example of ##U##.
 
Eingenvalues of U are ##\pm 1##. You can pick up the cases that eigenvalues of D and A are not 1 nor -1 for your purpose.
 
jedishrfu said:
is the identity m
No, because the identity matrix is diagonal.
 
anuttarasammyak said:
Eingenvalues of U are ##\pm 1##. You can pick up the cases that eigenvalues of D and A are not 1 nor -1 for your purpose.
Why? Eigenvalue can be for instance ##i##.
 
  • #10
My #7 talks abot U of post #5. There for U \lambda^2 =1
UDU=<br /> \begin{pmatrix}<br /> k &amp; \alpha \\<br /> \beta &amp; -k \\<br /> \end{pmatrix}<br /> \begin{pmatrix}<br /> d_1 &amp; 0 \\<br /> 0 &amp; d_2 \\<br /> \end{pmatrix}<br /> \begin{pmatrix}<br /> k &amp; \alpha \\<br /> \beta &amp; -k \\<br /> \end{pmatrix}<br /> =<br /> \begin{pmatrix}<br /> d_2+(d_1-d_2)k^2 &amp; k\alpha (d_1-d_2) \\<br /> k \beta (d_1-d_2) &amp; d_1-(d_1-d_2)k^2 \\<br /> \end{pmatrix}<br /> =A
, and
d_1,d_2 \neq -1,1
to satisfy your further condition.
 
Last edited:

Similar threads

  • · Replies 14 ·
Replies
14
Views
4K
  • · Replies 2 ·
Replies
2
Views
3K
  • · Replies 6 ·
Replies
6
Views
4K
  • · Replies 33 ·
2
Replies
33
Views
1K
  • · Replies 5 ·
Replies
5
Views
2K
  • · Replies 13 ·
Replies
13
Views
3K
  • · Replies 4 ·
Replies
4
Views
1K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 2 ·
Replies
2
Views
3K
  • · Replies 5 ·
Replies
5
Views
3K