Nullspace that spans $\mathbb{R}^n$?

My professor said that if for a $n \times n$ matrix $A$, $\text{null}(A) = \mathbb{R}^n$, then $A = 0_{n}$. Why is this true? I understand what its saying – if everything times this matrix is zero, then the matrix has to be zero.

The intuition is simple enough with numbers, but could someone explain why this is true with matrices? Thanks

Solutions Collecting From Web of "Nullspace that spans $\mathbb{R}^n$?"

If $\operatorname{null}(A)=\Bbb R^n$, then for all $x\in \Bbb R^n$ it holds that $Ax=0_{\Bbb R^n}$. In particular for all $i\in \{1, \ldots, n\}$ it is true that $Ae_i=0_{\Bbb R^n}$.

So $Ae_1=0_{\Bbb R^n}\land Ae_2=0_{\Bbb R^n}\land \ldots \land Ae_n=0_{\Bbb R^n}$.

Consider the matrix whose column $i$ is $e_i$, $[e_1\mid \ldots \mid e_n]$.

The above tells you that $0_{n\times n}=[Ae_1\mid \ldots \mid Ae_n]$, but $[Ae_1\mid \ldots \mid Ae_n]=A[e_1\mid \ldots \mid e_n]$.

Now note that $[e_1\mid \ldots \mid e_n]=I_n$ and conclude.

Show that if the matrix $A$ contains one non-zero entry, then you can find a vector $x$ such that $Ax \ne 0$. The easiest vectors to consider are ones that are all zero except for one entry, which is one.

If $A \neq 0$ then some column of $A$ is not the $0$ (column) vector. Say the $i$th column of $A$ is not the $0$ vector. Then $$A e_i = i\text{th column of } A$$
where $e_i$ is the (column) vector with $1$ in the $i$th positions and zero otherwise. Why does this imply that null$(A)$ isn’t all of $\Bbb{R}^n$?

A linear transformation can be visualized as taking a grid over space and skewing/distorting it. The grid lines all have to stay lines, and parallel lines remain parallel (i.e. it respects vector addition and multiplication), but that’s it. So “stretch the z-axis by 2” and “rotate around the y axis by 30$^\circ$” are valid linear transformations, represented by particular matrices.

In particular, squashing $\mathbb{R}^n$ onto some subspace is also a linear transformation (e.g. “squish everything down into the xz plane”). If we end up with a subspace, then that means that some other subspace (in this case the y-axis) got mapped to the origin. When this happens we say that this subspace is part of the nullspace of the matrix. So $\operatorname{null}(A)=\mathbb{R}^n$ is saying “this matrix squishes everything to the origin”. The zero matrix multiplies everything by zero, so it squishes everything to the origin. What your professor has proved is that the only matrix that does this is the zero matrix: the “squish everything to zero” property is enough to pin it down, because nothing else can do this.

In particular, any matrix with a nonzero entry in its $n^{th}$ column can be multiplied by the basis vector $(0,0,0,…1…0)$ with $1$ at position $n$, which will then produce this column vector as output (which by hypothesis is nonzero). So any nonzero matrix leaves something unsquished, and the conclusion follows.