# Why is the cyclic decomposition of a primary torsion module not unique?

It is known that a finitely generated torsion module $M$ over a principal ideal domain $R$ can be decomposed into a direct sum of primary modules, $$M=M_{p_1}\oplus\cdots\oplus M_{p_n}.$$ Futhermore, the primary submodules $M_{p_i}$ can be decomposed as a direct sum of cyclic submodules, but this decomposition is not unique.

Is there a standard example exhibiting why the cyclic decomposition of a primary module is not unique?

-

Sure, consider an abelian group of order $4$, written additively, which is the sum $$\langle a \rangle \oplus \langle b \rangle,$$ with $a$ and $b$ of order $2$. Then you also have $$\langle a+b \rangle \oplus \langle b \rangle.$$ So you see that this is just the fact that in general a vector space (over $\mathbf{Z}_{2}$, in this case) does not have a unique basis.