0

Let there be $A,B$ matrices.
Let $C=A+B$
$Span(Col(C))\subseteq Span(Col(A))$ because C is a linear combination of A . $Span(Col(C))\subseteq Span(Col(B))$ because C is a linear combination of B .
Therefore $Span(Col(C))\subseteq Span(Col(A))+Span(Col(B))$ and $Rank(A+B)\leq Rank(A)+Rank(B)$

gbox
  • 13,645

2 Answers2

3

$Span(Col(C))\subset Span(Col(A))+Span(Col(B)) $ then $Rank(C)\leq \dim(Span(Col(A))+Span(Col(B)))$, by using the inequality $\dim(F+G)\leq \dim F+\dim G$ we get $Rank(C)\leq \dim(Span(Col(A)))+\dim(Span(Col(B)))=Rank( A)+Rank(B)$

Hamou
  • 6,915
1

Define $Lx = (Ax, Bx)$. Clearly $L$ is linear and ${\cal R} L \subset {\cal R} A \times {\cal R} B$ (this inclusion could be strict).

Hence $\dim {\cal R} L \le \dim ( {\cal R} A \times {\cal R} B ) = \dim {\cal R} A + \dim {\cal R} B$.

Now let $\phi(y) = y_1+y_2$, where $y = (y_1,y_2)$, again note that $\phi$ is linear. If $S$ is a subspace, it is easy to see that $\dim \phi(S) \le \dim S$ (this is true for any linear operator, not just $\phi$).

Letting $S = {\cal R} L$, and noting that $A+B = \phi \circ L$ gives the desired result.

copper.hat
  • 178,207