See my previous question here, Intersection of images and union of kernels.
Let $A$ be a ring (not necessarily commutative), let $M$ be an $A$-module, and let $u: M \to M$ be an $A$-module morphism. Put $$\text{Image}(u^\infty) := \bigcap_{k=1}^\infty \text{Image}(u^k),\text{ resp. }\text{Ker}(u^\infty) = \bigcup_{k=1}^\infty \text{Ker}(u^k).$$Are $\text{Image}(u^\infty)$ and $\text{Ker}(u^\infty)$ necessarily $A$-submodules in $M$?
My question this time around is, do we moreover have that$$M = \text{Image}(u^\infty) \oplus \text{Ker}(u^\infty)?$$
Since we have$$M \supseteq \text{Im}(u) \supseteq \text{Im}(u^2) \supseteq \dots$$and $M$ is of finite length, we must have some $k \in \mathbb{N}$ such that for all $k' \ge k$,$$\text{Im}(u^{k'}) = \text{Im}(u^k) = \text{Im}(u^\infty).$$Since we also have$$\text{Ker}(u) \subseteq \text{Ker}(u^2) \subseteq \dots \subseteq M,$$we get that for all $k' \ge k$,$$\text{Ker}(u^{k'}) = \text{Ker}(u^k) = \text{Ker}(u^\infty).$$