1

Problem:

Let $\phi$ and $\psi : V \to V$ be linear operators on a vector space $V$ of dimension $n$. Show that $$rank(\phi \circ \psi) ≥ rank(\phi) + rank(\psi) − n.$$

My work:

I have already seen this and this questions where the same is answered however I have a different question. I started writing the following proof:

Consider the linear operator $\phi \circ \psi \in \text{Hom}(V,V)$ and let $\dim \text{Ker } \psi = a$ and $\dim \text{Im } \phi = b$, . We want to show $$\dim \text{Im } (\phi \circ \psi) \ge \dim \text{Im } \phi + \dim \text{Im } \psi - n = \dim \text{Im } \phi - \dim \text{Ker } \psi = b - a$$ so we only need to consider the case $b > a$, otherwise we would be finished. Let $(u_1, \ldots, u_a)$ be the basis of $\text{Ker } \psi \subseteq V$ so we can extend it to $(u_1, \ldots, u_a, v_1, \ldots, v_{n-a})$ a basis of $V$. For any $v \in V$ and for some scalars $x_i$, $$(\phi \circ \psi)(v) = (\phi \circ \psi)(u_1x_1 + \ldots + u_ax_a + v_1x_{a+1} + \ldots + v_{n-a}x_{a+n}) = (\phi \circ \psi)(v_1)x_{a+1} + \ldots + (\phi \circ \psi)(v_{n-a})x_{a+n}$$ So $span((\phi \circ \psi)(v_1), \ldots, (\phi \circ \psi)(v_{n-a})) = \text{Im } (\phi \circ \psi)$ and...

As you see it is unfinished since my intention was to prove $\text{Im }(\phi \circ \psi) \ge n - a \ge b - a$ since $\text{Im } \phi \subseteq V$ but for that I needed $(\phi \circ \psi)(v_1), \ldots, (\phi \circ \psi)(v_{n-a})$ to be linearly independent, but I didn't realise that there is no need for them to be linearly independent until I got to that point.

So my doubt is: Is there a way to fix the proof without changing it entirely?

Huge thanks in advance.

Torrente
  • 680
  • 1
  • 2
  • 11

1 Answers1

1

Consider $\ker \phi\vert_{\mathrm{Im}\ \psi}.$ We have $$\ker \phi\vert_{\mathrm{Im}\ \psi}=\ker \phi\cap\mathrm{Im}\ \psi.$$ Then$$\dim \mathrm{Im}\ \phi\circ\psi=\dim \mathrm{Im}\ \phi\vert_{\mathrm{Im}\ \psi}=\dim \mathrm{Im}\ \psi-\dim\ker \phi\cap\mathrm{Im}\ \psi\geq \dim \mathrm{Im} \ \psi-\dim \ker \phi=\dim \mathrm{Im} \ \psi-n+\dim \mathrm{Im}\ \phi.$$ that's done.

Park
  • 172
  • But aren't you assuming $\dim \text{Ker } \psi \ge \dim \text{Ker }(\phi \circ \psi)$? – Torrente Sep 04 '22 at 12:20
  • I appreciate your work but I think that your proof is wrong since $\dim \text{Ker } \psi \le \dim \text{Ker }(\phi \circ \psi)$ given that $\text{Ker } \psi \subseteq \text{Ker }(\phi \circ \psi)$. As well, are you assuming $\dim \text{Im } \phi > \dim \text{Im } \psi$? Could you explain how you conclude that? – Torrente Sep 04 '22 at 13:10
  • @ItsTrex Sorry. I always write \phi as \psi. $\dim ker \phi\cap\mathrm{Im} \psi\leq \dim \ker \phi.$ – Park Sep 04 '22 at 15:17
  • By Rank-nullity theorem, $\dim \mathrm{Im} \phi\vert_{\mathrm{Im}\ \psi}=\dim \mathrm{Im}\ \psi - \ker \phi\vert_{\mathrm{Im}\ \psi}=\dim \mathrm{Im}\ \psi -\dim\ker \phi\cap\mathrm{Im}\ \psi.$ – Park Sep 04 '22 at 15:23