I am interested in the kernel of $\Phi: U_1 \oplus U_2 \oplus U_3 \rightarrow V$, $(u_1, u_2, u_3) \mapsto u_1+u_2+u_3$ for $V$ a $K$-Vectorspace and $U_1, U_2, U_3$ arbitrary subspaces of $V$.

Is there a nice way to describe the kernel?

So far I can only show $\max(\dim((U_l+U_k)\cap U_j))\leq\dim \ker\Phi \leq \min (\dim((U_{l_1}+U_{k_1})\cap U_{j_1} \oplus (U_{l_2}+U_{k_2})\cap U_{j_2})) $

And of course that $\ker \Phi$ is a subspace of $((U_2+U_3)\cap U_1) \oplus ((U_1+U_3)\cap U_2) \oplus ((U_1+U_2)\cap U_3)$.

Thank you!


Let's think about the simpler case $\Psi:U_1\oplus U_2\to V$ first. Then $\ker\Psi=(u_1,u_2):u_2=-u_1$ is isomorphic to $U_1\cap U_2$, since given any $u_1\in U_1\cap U_2$ there's a unique corresponding kernel element $(u_1,-u_1)$.

So, now we generalize to your $\Phi$. $\Phi=0$ just if $u_1+u_2=-u_3$, so certainly we'll only find kernel elements when $u_3\in U_1+U_2$. In this case there must exist a solution $\Phi(u_1,u_2,u_3)=0$, but it's not unique even for a fixed $u_3$: we can also take $\Phi(u_1+u_1',u_2+u_2',u_3)=0$ whenever $\Phi(u_1',u_2',0)=0.$ This happens, as above, whenever $u_2'=-u_1',$ so we get another solution for each element of $U_1\cap U_2$. Thus the kernel of $\Phi$ is isomorphic to $U_3\cap (U_1+U_2)\oplus U_1\cap U_2$.

It's then awfully tempting to formulate a generalization to arbitrarily many subspaces-but I'll leave that step to you.

  • 1
    $\begingroup$ Thanks for the answer. Now that you say it, there's even a more schematic way to see that: $\dim \ker \Phi = \dim(U_1 \oplus U_2 \oplus U_2) - \dim (U_1+U_2+U_3)$ hence the result. Works also for generalization. $\endgroup$ – user110071 Nov 20 '13 at 6:48

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.