Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I am given a square matrix $A$, and I need to prove that if c is its eigenvalue, then it is also an eigenvalue of its transpose. How should I approach this? Clearly $Av$=$cv$, but I am not sure how to bring transpose into the equation.

share|improve this question
    
If $M$ is diagonalizable it seems easy enough since $M = Q^{-1}\Lambda Q$ implies $M^t = Q^t \Lambda^t (Q^{t})^{-1}$, so $M$ and $M^t$ share the same multiset of eigenvalues. Maybe some well chosen theorem or observation extend this for other matrices? –  Myself Mar 4 '11 at 2:26
    
@Myself: What you want is Jordan canonical form. –  Soarer Mar 4 '11 at 3:55

4 Answers 4

up vote 11 down vote accepted

A simple way would be to look at $\left | A^T-cI \right |$. $$\left | A^T-cI \right | = \left| (A-cI)^T \right| = \left| (A-cI) \right|$$

EDIT

$\left| B \right|$ denotes the determinant of $B$. First note that for any matrix $B$, $\left| B \right| = \left| B^T \right|$.

(This is true since you get the same determinant if you find the determinant along the row or column.)

$c$ is an eigenvalue of $A$ iff $\left| A - cI\right| = 0$.

Note that $\left| A^T - cI \right| = \left| A^T - cI^T \right| = \left| (A - cI)^T \right| = \left| A - cI\right| = 0$.

Hence, $c$ is an eigenvalue of $A^T$.

share|improve this answer
    
Wow, this makes perfect sense -- how did I not realize that... Thanks a bunch. –  LinAlgStudent Mar 4 '11 at 2:24
    
@LinAlgStudent: Right. However you could look for a more direct proof. What I have done above is a bit round about. –  user17762 Mar 4 '11 at 2:28
    
Thanks. Going your way, I assume that by || you mean determinant. I am actually not quite clear how to prove that the determinants of those two matrices will be equal to each other (and equal to zero). –  LinAlgStudent Mar 4 '11 at 2:34

Another way to look at it is to see that it's equivalent to the following statement: $A$ is regular iff $A^T$ is regular. That follows from the fact that the row rank is equal to the column rank. But I'm not sure how easy it is to show that directly. Anyone?

share|improve this answer
    
It is actually interesting to think why intuitively the left and right eigenvalues must be the same. –  user17762 Mar 4 '11 at 2:44
    
because they are similar. –  Soarer Mar 4 '11 at 3:52
    
@Soarer: How does that help with the intuition? –  user17762 Mar 4 '11 at 3:58
    
Sorry for the imprecision, I mean that $A$ and $A^T$ are similar matrices. –  Soarer Mar 4 '11 at 4:04
    
@Soarer: What is the intuition for them to be similar? –  user17762 Mar 4 '11 at 4:36

Here is another way, that is a bit overboard: Suppose $A:V\rightarrow V$ is a linear operator. Then by definition, $A^t$ is the operator on the dual space $V^*$ given by $$\left(A^t(w)\right)(v)=w\left(A(v)\right)$$ where $w\in V^*$ and $v\in V$.

Suppose $A$ is invertible and $Ax=\lambda x$. Let $W=\text{span}(x)$ and then decompose $V$ into $V=W\oplus W^c$ where $W^c$ is also invariant under $A$. Let $w$ be any non-zero functional in the annihilator of $W^c$. Then lets look at $A^t(w)$. We have $$(A^t(w))(v)=w(A(v))$$ for every vector $v$, but it suffices to only consider $v\in W$ or $v\in W^c$. In the previous case, $w(A(v))=w(\lambda v)=\lambda w(v)$, and in the second case $w(A(v))=0$ since $W^c$ is invariant under $A$. Thus $$A^t(w)=\lambda w$$ so we see the transpose will have the same eigenvalues.

share|improve this answer

Let $\lambda_1,v_1$ and $\lambda_2,v_2$ be eigenpairs of $A$ and $A^T$ respectively.

$$Av_1 = \lambda_1 v_1$$

$$A^Tv_2 = \lambda_2 v_2$$

Therefore,

$$v_2^TAv_1 = \lambda_2 v_2^T v_1$$

$$\lambda_1 v_2^Tv_1 = \lambda_2 v_2^T v_1$$

Therefore,

  1. $v_2^T v_1 = 0$ and $\lambda_1 \neq \lambda_2$.
  2. $\lambda_1 = \lambda_2$
share|improve this answer
    
So what have you shown in the case where $v_2^T v_1=0$? –  Myself Mar 4 '11 at 4:07

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.