1
$\begingroup$

The first matrix A has eigenvectors

(0,1,0), (2,0,1), (1,0,1)

The second matrix B has eigenvectors

(4,1,2), (1,1,0), (1,0,1).

Both sets form a basis for $R^3$.

Now, how do I pick out a basis of eigenvectors that simultaneously diagonalizes A and B?

I've searched on MSE but have only found proof-y discussions of this topic, but I want to find an explicit basis.

Thanks,

EDIT: I tried putting all six vectors in a matrix, each in a row, and row reduced until I got 3 linearly independent vectors. I used this as S, computed the inverse, but then $S^{-1}AS$ is not diagonal anymore...almost diagonal, though - off by one non-zero entry, with all of A's eigenvalues still on the diagonal. So, I'm pretty close, I think.

EDIT 2:

The matrices are:

$$ A= \begin{bmatrix} 0 & 0 & 2 \\ 0 & 1 & 0 \\ -1 & 0 & 3 \\ \end{bmatrix} $$

and

$$ B= \begin{bmatrix} 0 & 4 & 4 \\ -1 & 5 & 1 \\ -2 & 2 & 6 \\ \end{bmatrix} $$

Eigenvalues for A are 1,2 -- with 1 having geometric multiplicity = 2.

Eigenvalues for B are 3, 4 -- with 4 having geometric multiplicity = 2.

$\endgroup$
2
  • 1
    $\begingroup$ I've deleted my answer because I think it only works for symmetric matrices. I'll rethink it and let you know. $\endgroup$
    – joriki
    Aug 10 '15 at 22:27
  • $\begingroup$ Ah, ok, got it. Thanks for the quick response, @joriki. I will do some more work on this before moving on to new problems. And unfortunately, I have not seen discussions on this topic on MSE that talk about finding an explicit basis -- the questions and answers I have found are mainly proving the claim of simultaneous diagonalizability. $\endgroup$
    – User001
    Aug 10 '15 at 22:30
1
$\begingroup$

The matrix $A$ has the eigenspaces (according to your computation) $$ E_1(A)=span\left( \pmatrix{0\\1\\0}, \pmatrix{2\\0\\1}\right), E_2(A)=span \pmatrix{1\\0\\1}, $$ while $B$ has the eigenspaces $$ E_4(B)=span\left( \pmatrix{1\\1\\0}, \pmatrix{1\\0\\1}\right), E_3(B)=span \pmatrix{4\\1\\2}. $$ The goal is to find a basis of common eigenvectors. That is, to find vectors that are in eigenspaces of $A$ and of $B$.

Clearly, $\pmatrix{1\\0\\1} \in E_2(A)\cap E_4(B)$, $\pmatrix{4\\1\\2}\in E_1(A)\cap E_3(B)$, $\pmatrix{2\\1\\1}\in E_1(A)\cap E_4(B)$.

Moreover, these vectors are linearly independent, and as such from a basis of eigenvectors of both $A$ and $B$.

$\endgroup$
2
  • $\begingroup$ Very pretty and simple answer @daw -- thanks so much. I have noticed a restriction, though: in picking the third eigenvector from the two-dimensional eigenspaces, the vector must be (of course) expandable in a linear combination of the two vectors that span the two-dimensional eigenspace -- but in the linear combination, the coefficients must both be non-zero. $\endgroup$
    – User001
    Aug 11 '15 at 22:52
  • $\begingroup$ Otherwise, I am guessing that it is like picking an eigenvector that corresponds to an eigevalue having geometric multiplicity = 1, which then makes both A and B not diagonalizable, since the algebraic and geometric multiplicities must match for all eigenvalues of a matrix for it to be diagonalizable. $\endgroup$
    – User001
    Aug 11 '15 at 22:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.