# Gaining intuition for how submodules behave

I'm studying elementary commutative algebra this semester, largely following Atiyah-MacDonald. I often find myself in a situation where I'm interested in whether some property of an R-module M is inherited by its submodules (e.g. the property of being finitely generated over R) and I feel like I am lacking the necessary intuition to decide whether something is the case or not. In the case of finite-generation, for example, I think that my mental picture of a finitely-generated R-module is still too close to that of a finite-dimensional vector space to be able to intuit the fact that this simply shouldn't hold true in general. So here's my question - when you run across some property for a module and you want to know whether this is inherited by its submodules, how do you begin to think about the problem? Do you have a standard stock of counterexamples (or a procedure of sorts to concoct a counterexample)? Or do you have a more nuanced way of informally thinking about modules that captures more of their behavior? As I only have a semester of commutative algebra under my belt (think Atiyah-MacDonald), I'd appreciate answers that tend towards the more elementary end of the subject, although if you think that it's not possible to gain a good feel for the way that modules behave without diving deeper, I'd like to hear that too.

-
Have you read the entirety of A-M and especially done 95% of the exercises? There's a lot in those exercises (including all sorts of examples and counterexamples), so it ought to do the job for you. One also gains a lot of intuition from thinking geometrically, but that requires going deeper by way of quasi-coherent sheaves on schemes (or more "classically", coherent sheaves on algebraic varieties as in Serre's FAC paper) as analogues of vector bundles on manifolds. So that is something you'll learn later. – BCnrd May 17 '10 at 13:11
In the particular case of the property you have in mind, the relevant class is that of Noetherian modules. You will gain an intuition about those in chapters 6 to 8, if I recall correctly. I think the problem is just that Atiyah and MacDonald postpone the Noether property too much; it should really be introduced early as one of the main tools in commutative algebra. – Andrea Ferretti May 17 '10 at 15:24
I totally agree with BCnrd. The question is unspecific and is not really an approproate question, but rather asks for partial summaries of the basics of module theory ... – Martin Brandenburg May 18 '10 at 0:21

Doing all exercises in Atiyah-MacDonald, like BCnrd suggested, is surely the ideal way to learn about this and much more. Let me offer a couple of practical tips to get you started:

A surprisingly effective example to keep in mind when you deal with any question about submodules of a module $M$ is to take $M=R$. Then the submodules of $R$ are just the ideals of $R$, which are concrete enough to check your intuition, but still possess a very rich structure so that not much is lost.

Also, since many properties of modules fail to pass to submodules in higher dimension, it usually suffices to consider some small example, say $R= k[x,y]$.

As an example, let says you are trying to understand the following question: Over what Noetherian ring $R$ is a submodule of any free module free? (this is of course true for vector spaces).

If you take $M=R$, it follows that all ideals $I$ have to be free. If $R=k[x]$, this is true, and already an interesting exercise, but if $R=k[x,y]$, just take $I=(x,y)$. $I$ is not free because the generators have a non-zero relation: $xy-yx=0$. This example also suggests that all ideals in $R$ have to be principal, otherwise similar counter-examples can be found. So you naturally gets to principal ideal rings.

If you want to play with it a bit more, since $R/I$ fits into an exact sequence:

$$0 \to I \to R \to R/I \to 0$$

This says that $R/I$ has projective dimension at most $1$ for any ideal $I$. This leads you to some serious restriction on $R$, which will point you to the right condition, from a different perspective.

You can replace "free" by "locally free" and play the same game, it will naturally leads you to all sort of interesting things worth learning about commutative rings, for examples projective modules or Quillen-Suslin theorem, etc.

(There are, of course, other ways to approach this particular question, my point is by considering $M=R$ you can already get quite far). I hope you will have some fun!

-
Thanks - this is very helpful. – Nick Salter May 17 '10 at 15:57
This is a simultaneously friendly, useful and deep answer. Congratulations, Hailong! – Georges Elencwajg Aug 26 '11 at 8:45
Dear Georges, thank you for your very kind words. – Hailong Dao Aug 27 '11 at 15:17

My not-so-expert opinion would be that "What properties of M does a submodule of M inherit" may not be a general enough question to have a helpful answer. There are a few such properties that come up frequently (especially the Noetherian property that you mention), and it's worth understanding each on its own. (For example, proposition 6.3.1 on page 75 of A-M is vital.) I agree with you that thinking up examples of modules, to give yourself some intuition, is useful.

To that end, I'd suggest taking a look at the question "What representative examples of modules should I keep in mind?" Unfortunately, many of the answers to that question may be a little too pathological for your taste, and the accepted answer (to which BCnrd also refers) requires the motivation of algebraic geometry.

I like the suggestion in Andrew Critch's answer to "find interesting rings." A module over $R$ is often a tool to study $R$, rather than a fundamental object in its own right. Many "interesting" rings might seem perfectly natural objects of study, even to a beginning student: for example, a polynomial ring, or a polynomial ring modulo an ideal, $k[x_1,\ldots,x_n]/I$.

To get comfortable working with these rings (and modules over them), I'd suggest learning some algebraic geometry; that way you can think of a ring like $\mathbb{C}[x, y]/(y^2-x^3)$ as the curve $y^2=x^3$ in the plane, rather than some kind of fancy made-up thing. Meanwhile, I agree with BCnrd that doing the exercises in A-M (or taking an algebra course and doing the assignments) is a good way to start getting more familiar.

ETA: Failed to mention that the most important examples of modules over a ring $R$ are ideals in $R$. Luckily, Hailong Dao's excellent answer has this covered, so rather than remark further, I'll just say, read that.

-
There are lots of good suggestions here: +1. About the suggestion to learn algebraic geometry: it's a good suggestion, but not an easy one to follow. In particular, I think you have to study commutative algebra and algebraic geometry side by side for a good while -- more than one semester -- in order for the close interaction between the two to have a positive effect on your understanding.... – Pete L. Clark May 17 '10 at 15:07
...As an example, in my first commutative algebra course there were several exercises dealing with the ring $\mathbb{C}[x,y,z,w]/(xw-yz)$. I found these pretty mysterious. I think I knew enough "geometry" at the time to see that this was the ring of functions vanishing on locus of $2 \times 2$ matrices of determinant zero and that this was not a manifold because of the singular point at the origin. But that was not enough knowledge to help me. Now I know that this is an affine toric variety -- which explains why it is normal despite not being regular -- and so forth. – Pete L. Clark May 17 '10 at 15:20