Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Suppose I have the Jordan normal form of a matrix $A$. The decomposition involves the Jordan matrix $J$ and a similarity matrix $P$ such that $P^{-1}.J.P = A$. My question: is it possible to find the similarity matrix of $A^T$ given that we know that of $A$? Thank you in advance.

share|cite|improve this question

Let $P$ and $J$ correspond to the Jordan decomposition of $A$, i.e. $A = P^{-1} J P$, then $J$ is block-diagonal: $$ J = \begin{pmatrix} J_1 & & \\ & \ddots & \\ & & J_p \end{pmatrix} $$ Let $J_k$ be $n$ by $n$ block: $$ J_k = \begin{pmatrix} \lambda_k & 1 & 0 &\cdots \\ 0 & \lambda_k & 1 & \cdots \\ 0 & \vdots & \ddots & \vdots \\ 0 & & & \lambda_k \end{pmatrix} $$ Note that $$ J_k^T = S^{-1} J_k S $$ where $S$ is anti-diagonal matrix, i.e. $S_{ij} = \delta_{i+j, n+1}$. Thus $A^T = P^T J^T (P^{T})^{-1} = P^T \mathcal{S}^{-1} J \mathcal{S} (P^{T})^{-1}$, where $\mathcal{S}$ is block-diagonal matrix with $S_k$ matrices corresponding to each individial $J_k$ block.

share|cite|improve this answer
Note that $S=S^{-1}$, since the effect of $S$ is to reverse the order of the standard basis, and reversing order twice leaves the order intact. – alex.jordan May 14 '12 at 21:41

Since inverting commutes with transposing, and both are order reversing operations,

$$ \begin{align*} A^T & =(P^{-1}JP)^T\\ & =P^TJ^T(P^{-1})^T\\ & =P^TJ^T(P^T)^{-1} \end{align*}$$

$J^T$ is not in Jordan normal form though. First assume $J$ has just one Jordan block with eigenvalue $\lambda$. Then $J^T$ represents a transformation where the $n$th basis vector is scaled by $\lambda$, and each previous basis vector is scaled by $\lambda$ and sheared $1$ unit in the direction of the next basis vector. So if the order of the basis were reversed, we would have a transformation where the first basis vector is scaled by $\lambda$, and each subsequent basis vector is scaled by $\lambda$ and sheared $1$ unit in the direction of the previous basis vector. That is, we'd have a Jordan block.

So if we change basis by reversing the basis order using a matrix $M$, we have $$J^T=MJ_{A^T}M$$ where $M$ is the antidiagonal matrix consisting of all 1's. ($M$ is its own inverse.) Of course, if $J$ has more than one Jordan block, you will need to reverse the orders of several subsets of basis vectors and $M$ will have blocks of antidiagonal submatrices.

Now, $$A^T=(P^TM)J_{A^T}(M(P^T)^{-1})$$

share|cite|improve this answer
What is the matrix $J_{A^T}$? It may be a silly question, but I'm a bit confused since you didn't define this anywhere. Would this proof be sufficient to say that if A is in Jordan form, then you can say what the jordan form a $A^T$ is? – Calculus08 Mar 6 '14 at 16:21
@Calculus08 $J$ is the Jordan normal form matrix for $A$, and $J_{A^T}$ is the Jordan normal form matrix for $A^T$. – alex.jordan Mar 6 '14 at 17:44

Have you tried computing $(PAP^{-1})^T?$. What you will get will be nearly the Jordan normal form of $A^T$. You just have to reorder the vectors. Which matrix do you need for that?

share|cite|improve this answer

Your Answer


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.