1

Can somebody help me in proving that for a given two non-zero column vectors $u$ and $v$ in $\mathbb{R}^n$, such that $u$ and $v$ are linearly independent, then the matrix $uu^T+vv^T$ has rank two.

Uday Kumar
  • 111
  • Try to prove: if $uu^{T}+ vv^{T}$ has not rank 2 then $u$ and $v$ are linear dependent. For rank = 0 and rank larger than 2 the proof is obvious. For rank =1, it is enough to show that $u = cv$ por some constant $c$. – Alex Silva Oct 23 '16 at 06:54

3 Answers3

2

$$uu^T+vv^T=\begin{bmatrix}u & v \end{bmatrix} \begin{bmatrix} u^T \\ v^T\end{bmatrix}$$

Clearly, $\begin{bmatrix}u & v \end{bmatrix}$ is a rank $2$ matrix.

Let $\begin{bmatrix}u & v \end{bmatrix}=U\Sigma V^T$ be the singular value decomposition where $U\in \mathbb{R}^{n \times n}, \Sigma \in \mathbb{R}^{n \times 2}$ $V^T \in \mathbb{R}^{2\times 2}$.

The last $n-2$ rows of $\Sigma$ are zero matrices and the first two rows form a diagonal matrix. Let $$\Sigma=\begin{bmatrix} D \\ 0_{(n-2) \times 2} \end{bmatrix}$$

$$uu^T+vv^T=U\Sigma \Sigma^T U^T=U\begin{bmatrix} D^2 & 0_{2 \times (n-2)} \\ 0_{(n-2) \times 2} & 0_{(n-2)\times (n-2)}\end{bmatrix}U^T$$

Rank of $uu^T+vv^T$ is equal to the rank of $D^2$ which is $2$.

Siong Thye Goh
  • 149,520
  • 20
  • 88
  • 149
2

There are two parts, why the rank is at least $2$ and why it is at most $2$.

By choosing a vector $w$ with $u^T(w)\neq0$ and $v^T(w)=0$, which is possible since $u,v$ are linearly independent, we have $(uu^T+vv^T)(w)=u^T(w)u$, a nonzero multiple of $u$, so $u$ is in the image of $uu^T+vv^T$; by a similar argument $v$ is as well, so the rank is at least$~2$.

The rank of a sum of two linear maps can never exceed the sum of their ranks (its image is contained in the sum of its images), so the rank of $uu^T+vv^T$ is at most$~1+1=2$.

Marc van Leeuwen
  • 115,048
0

Let $A=uu^\top + vv^\top$.

Apply Gram-Schmidt on $\{u,v\}$ to get an orthonormal basis $\{w_1,\ldots,w_n\}$ where $\operatorname{span}\{u,v\}=\operatorname{span}\{w_1,w_2\}$.

What is $Aw$ when $w \in \operatorname{span}\{w_3,w_4,\ldots,w_n\}$?

The above hints show the rank is at most $2$. To conclude it is exactly $2$, you need to use linear independence of $u$ and $v$.

angryavian
  • 89,882