Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Given a linear mapping $T:V\rightarrow W$ between the two vector spaces, $\operatorname{Im}(T)$ and $V/\operatorname{ker}(T)$ are isomorphic. I wonder if $\operatorname{Ker}(T)$ and $W/\operatorname{Im}(T)$ are also isomorphic?

If yes, then we also have $\dim{(W)}=\dim{(\operatorname{Ker}(T))} + \dim{(\operatorname{Im}(T))}$?


share|cite|improve this question
This is not true. Consider the map $T: \mathbf{R} \rightarrow \mathbf{R}^3$ given by $T(x)=(x,0,0).$ – ShawnD Aug 18 '11 at 22:34
This cannot be true since one can replace W by a larger vector space and keep the same linear mapping. Then Ker(T) stays the same but W/Im(T) changes. – Did Aug 18 '11 at 22:35
$W/\text{im}(T)$ is known as the cokernel. The linear map fits in an exact sequence $0 \to \ker{(T)} \to V \to W \to \text{cokernel}(T) \to 0$ – Juan S Aug 19 '11 at 1:31
Take $0\to K$ or $K\to0$, where $K$ is the base field. In general, you have an exact sequence $$0\to\mathrm{Ker}\to V\to W\to\mathrm{Coker}\to0,$$ and the alternate sum of the dimensions is $0$. – Pierre-Yves Gaillard Aug 19 '11 at 2:44
up vote 3 down vote accepted

Another way to see that this cannot be true is to use the correct rank-nullity theorem, i.e., $\dim{(V)} = \dim{(\operatorname{Ker}{(T)})} + \dim{(\operatorname{Im}{(T)})}$. If we assume $\dim{(W)}=\dim{(\operatorname{Ker}{(T)})} + \dim{(\operatorname{Im}{(T)})}$, then we must have $\dim{(W)} = \dim{(V)}$!

Now, if we assume that $V$ and $W$ are vector spaces of the same dimension, and $T:V\to W$ is a linear map, then $V/\operatorname{Ker}{(T)}$ and $\operatorname{Im}{(T)}$ are canonically isomorphic via $[v] \to T(v)$, where $[v]$ is the class of $v\in V$ in $V/\operatorname{Ker}{(T)}$.

Since $$\dim{(W)} = \dim{(\operatorname{Im}{(T)})} + \dim{(W/\operatorname{Im}{(T)})}$$ and by rank-nullity $$\dim{(W)}=\dim{(V)} = \dim{(\operatorname{Ker}{(T)})} + \dim{(\operatorname{Im}{(T)})},$$ we may subtract both equations and obtain $\dim{(W/\operatorname{Im}{(T)})}=\dim{(\operatorname{Ker}{(T)})}$. Hence, $W/\operatorname{Im}{(T)}$ and $\operatorname{Ker}{(T)}$ are vector spaces of the same dimension, therefore isomorphic... but the isomorphism is not canonical! They are isomorphic simply because their dimensions coincide.

Although the term "canonical" does not have a very precise meaning in general, in this entry by canonical I mean "natural", or "coordinate-free". For instance, if $T:V\to W$ is any linear map of $\mathbb{F}$-vector spaces, then the isomorphism $\phi:V/\operatorname{Ker}{(T)}\to \operatorname{Im}{(T)}$ is canonical because I can describe it without having to choose bases for the vector spaces involved. The map is simply $\phi([v])= T(v)$, for any $v\in V$. However, as we have shown above, if $V$ and $W$ have the same dimension, we also have an isomorphism $\psi: W/\operatorname{Im}{(T)} \to \operatorname{Ker}{(T)}$, but we cannot describe this isomorphism unless I define a basis $\{w_1,\ldots,w_n\}$ of $W/\operatorname{Im}{(T)}$ and a basis $\{v_1,\ldots, v_n\}$ of $\operatorname{Ker}{(T)}$, and then, once in coordinates, we choose an arbitrary isomorphism $\mathbb{F}^n \to \mathbb{F}^n$.

share|cite|improve this answer
Thanks! What do "canonical" and"non-canonical"mean in this case and in its most generaility (I have seen them many times)? – Tim Aug 18 '11 at 23:59
Good question! Perhaps the Wikipedia entry for canonical may help here: -- I will also add something to my entry... – Álvaro Lozano-Robledo Aug 19 '11 at 0:07

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.