Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Possible Duplicate:
Eigenvalues of sum of anti-commuting matrices

I know two anti-commuting (nxn)-matrices A and B, n -even. I know also that +-a and +-b are real eigenvalues among all eigenvalues of A and B respectively. How to show that the matrix A+B has also at least two real eigenvalues of the form +-\sqrt{a^2+b^2} ? (I know also that (A+B)(A+B)^t=(A+B)^t(A+B)=(a^2+b^2)I, where I is (nxn)-identity matrix and AA^t=A^tA=a^2I, BB^t=B^tB=b^2I and ^t denotes transposition).

share|improve this question
3  
Please could you edit your original question, rather than post a second almost identical one. –  Robin Chapman Oct 11 '10 at 8:04
    
Note: I deleted the duplicate, so the banner on the top is spurious. –  S. Carnahan Jun 14 '12 at 11:25

1 Answer 1

up vote 1 down vote accepted

Suppose for simplicity's sak that $A$ and $B$ are diagonalizable over $\mathbb{R}$ and are non-singular.

Let $V_a$ be the $a$-eigenspace of $A$. Then by anti-commutativity, we find $BV_a\subseteq V_{-a}$ etc. As $A$ and $B^2$ commute then there is an eigenvector $v\in V_a$ with $B^2v=b^2 v$ for some $v$. If we let $w=bv+Bv$ then $Bw=bw$ so $b$ is real (assuming $B$ has real eigenvectors). On the space $W$ spanned by $v$ and $w=b^{-1}Bv$ the linear transformation $A+B$ has matrix $$\left(\begin{array}{rr} a&b\\\\ b&-a\\\\ \end{array}\right)$$ which has an eigenvectors with eigenvalues $\pm\sqrt{a^2+b^2}$.

share|improve this answer
1  
Thanks a lot! But I don't understand how this: AV_a ⊆ V_a follows from anti-commutativity and how you got this: w=b^{−1}Bv. I also don't understand why B^2v=b^2v holds for v\in V_a. Sorry and thank you again. –  Anna Oct 11 '10 at 9:00
    
corrected ! –  Robin Chapman Oct 11 '10 at 18:35

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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