Quotient of a module is free implies the module is free

194 Views Asked by At

I'd like to ask your opinion whether this answer is correct. If not, under which conditions the following lemma is true?

Lemma Let $p$ be a prime number and $B$ a noetherian torsion free $\mathbb{Z}_p$-algebra. Let $M$ be a finitely generated $B$-module such that $M/pM$ is a free $B/pB$-module and $M$ is torsion-free as $\mathbb{Z}_p$-module. Then $M$ is a free $B$-module.

Proof Let $e_1,\dots,e_n$ be a $B/pB$-basis of $M/pM$. Let $N=(b_1,\dots,b_n)$ a lifting of this basis. Thus $M=N+pM$. By Nakayama lemma we have $(1+p)M/N=0$. As $1+p$ is invertible in $\mathbb{Z}_p$ it is invertible in $B$. Thus $M=N$. In other words $M=(b_1,\dots,b_n)$. We will prove that the $b_i$'s form a basis. Suppose that there exist $a_1,\dots,a_n$ such that $\sum_{i=1}^{n}a_ib_i=0$. Thus working on $M/pM$ we deduce that $p\mid a_i$ thus for $1\leq i\leq 0$ we have $a_i=pa'_i$ for some $a_i'\in B$. Similarily we show that $a'_i=p a''_i$ for some $a''_i\in B$. We deduce that $a_i\in \bigcap_{n>0}p^nB$. By Artin-Rees lemma $\bigcap_{n>0}p^nB$ is the set of $x\in B$ such that $(1+p)x=0$. The only possibility is that $x=0$. Which concludes the lemma.