7
$\begingroup$

I have seen a lot of papers mentioning that a certain matrix is rank-$1$. What properties does a rank-$1$ matrix have?

I know that if a matrix is rank-$1$ then there are no independent columns or rows in that matrix.

$\endgroup$
1
6
$\begingroup$

Any rank one matrix $A$ can always be written $A=xy^\intercal$ for vectors $x$ and $y$. More precisely...

Proposition: A matrix in $\mathbb{C}^{n\times n}$ has rank one if and only if it can be written as the outer product of two nonzero vectors in $\mathbb{C}^{n}$ (i.e., $A=xy^{\intercal}$).

Proof. This follows from the observation $$ \begin{pmatrix}x_{1}y^{\intercal}\\ x_{2}y^{\intercal}\\ \vdots\\ x_{n}y^{\intercal} \end{pmatrix}=xy^{\intercal}=\begin{pmatrix}y_{1}x & y_{2}x & \cdots & y_{n}x\end{pmatrix}. $$

Corollary: The eigenspace of a rank one matrix in $\mathbb{C}^{n\times n}$ is one dimensional.

Proof. If $A$ is a rank one matrix in $\mathbb{C}^{n\times n}$, the previous result tells us that it can be written in the form $A=xy^{\intercal}$. Next, note that for any vector $w$ in $\mathbb{C}^{n}$, $$ Aw=(xy^{\intercal})w=(y^{\intercal}w)x. $$ Since $(y^{\intercal}w)$ is a scalar, it follows that if $w$ is an eigenvector of $A$, it must be a multiple of $x$. In other words, the eigenspace of $A$ is $$ \operatorname{span}(x)=\left\{ \alpha x\colon\alpha\in\mathbb{C}\right\} . $$

$\endgroup$
2
  • $\begingroup$ Is there also a geometric interpretation of the linear transformation associated with $A = xy^T$? $\endgroup$ – philmcole Mar 28 '18 at 17:35
  • $\begingroup$ @philmcole: according to the answer, $Aw$ is always a scalar multiple of the vector $x$. The scalar multiple depends on the length of $y$ and $w$ and the angle between them (see en.wikipedia.org/wiki/Dot_product). $\endgroup$ – parsiad Mar 28 '18 at 18:52

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.