I have been looking at this specific case of a Sylvester equation for the square matrix $X$, $$ AX-XA=-A, $$ given a nilpotent square matrix $A$.

For a general Sylvester equation $$ AX + XB = C, $$ we should have a unique solution if and only if $A$ and $-B$ have distinct eigenvalue. So we definitely do not have uniqueness in our specific case.

But what can we say about existence?

What I have so far


I have tried writing our everything using the vectorization and Kronecker product, but I'm not sure who to use this.

Our equation becomes $$ \left( I_N \otimes (-A) + A^T \otimes I_N \right)\mathbf{vec}(X)=-\mathbf{vec}(A), $$ where \begin{equation} \begin{aligned} \left( I_N \otimes (-A) \right)_{n,m} &= -A_{n (mod N), m (mod N)} \\ \left( A^T \otimes I_N \right)_{n,m} &= A_{\lfloor m/N \rfloor, \lfloor n/N \rfloor}\, \delta_{n (mod N), m (mod N)} \\ \mathbf{vec}(X)_{k} &= (X)_{k-\lfloor k/N \rfloor, \lfloor k/N\rfloor+1} \end{aligned} \end{equation}


I have one specific ad hoc solution for a given $(A)_{i,j}= \delta_{i,j+1}$, $$ (X)_{i,j} = j \delta_{i,j} $$


If $A$ is not nilpotent, then no solutions (Jacobson). Let $A$ be $n\times n$ nilpotent. Then we may assume that $A=diag(J_{i_1},\cdots,J_{i_r})$ where $J_k$ is the nilpotent Jordan block of dimension $k$.

The general solution of our equation is $X=X_0+C(A)$ where $C(A)$ is the commutant of $A$ and $X_0$ is a particular solution. Then it remains to obtain some $X_0$; it suffices to obtain a particular solution of


There is a diagonal solution $X_0$, which is

$diag(0,-1,-2,\cdots)$, and we are done.


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.