I'm trying to solve this step from a Bourbaki's commutative algebra exercise (Chapter I, Flat Modules, exercise 23 (a) page 47, implication $(\gamma)\Rightarrow(\delta)$):

Let $R$ be a ring, $F$ be a free $R$-module and $K\subseteq F$ be an $R$-submodule. Assume that for every $x\in K$ there exists an $R$-module homomorphism $\upsilon:F\to K$ such that $\upsilon(x)=x$. Then for every finite sequence $x_1,\ldots,x_n$ of elements of $K$ there exists an $R$-module homomorphism $\upsilon:F\to K$ such that $\upsilon(x_i)=x_i$ for every $1\leq i\leq n$.

My try. I'm starting with the case $n=2$. By assumption there exists $\upsilon_1,\upsilon_2:F\to K$ such that $\upsilon_1(x_1)=x_1$ and $\upsilon_2(x_2)=x_2$. I'm failed to build $\upsilon$ as a linear combination of $\upsilon_1$ and $\upsilon_2$. Moreover, I cannot find a useful way to use assumption about $F$ to be a free $R$-module. Any hint will be appreciated.

  • $\begingroup$ Note that another equivalent condition, according to Bourbaki, is that the quotient $R$-module $F / K$ is flat. (Caution: What you call $R$ is not what Bourbaki calls $R$.) $\endgroup$ Aug 16, 2019 at 9:40
  • $\begingroup$ Also, $R$ is supposed to be commutative, right? $\endgroup$ Aug 16, 2019 at 9:40
  • 2
    $\begingroup$ With the equivalence mentioned by @darijgrinberg, this is given as Lemma 2.10(c) of Milne's "Etale Cohomology". $\endgroup$ Aug 16, 2019 at 9:43
  • $\begingroup$ I'm already proved $(\alpha)\Rightarrow(\beta)\Rightarrow(\gamma)$ and $(\delta)\Rightarrow(\alpha)$; thus only remains $(\gamma)\Rightarrow(\delta)$. I assume the ring not commutative, since Bourbaki distinguish between right and left modules. $\endgroup$ Aug 16, 2019 at 9:46
  • $\begingroup$ @AlexWertheim: Nice! And Milne's argument does not even use the freeness of $F$. (Link to the relevant part of the book.) $\endgroup$ Aug 16, 2019 at 9:50

2 Answers 2


Here is a self-contained proof, adapted from the proof in Milne's Étale Cohomology referenced in the other answer. We use induction on $n$, the base case $n=1$ being given. Now suppose $n>1$ and the result is known for $n-1$. We can choose $u:F\to K$ such that $u(x_1)=x_1$ and then using the induction hypothesis we can choose $v:F\to K$ such that $v(x_i-u(x_i))=x_i-u(x_i)$ for $i>1$. Defining $$w(x)=u(x)+v(x-u(x))$$ we then have $w(x_1)=x_1+v(x_1-x_1)=x_1$ and $w(x_i)=u(x_i)+v(x_i-u(x_i))=x_i$ for $i>1$, so $w(x_i)=x_i$ for all $i\geq 1$.


As requested above, I am posting my comment as an answer. The implication $(\delta) \implies (\alpha)$ is addressed (modulo the equivalence of the given hypotheses with the condition that $F/K$ be a flat $R$-module) in Lemma 2.10(c) of Milne's "Etale Cohomology". A link to the relevant portion (kindly provided by Darij Grinberg) can be found here.


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.