Abstract Nonsense

Crushing one theorem at a time

Direct Limit of Modules (Pt. II)

Point of Post: This is a continuation of this post.

\text{ }

Ok, fine, so now we have defined what it means to take a direct limit of a directed system, let’s get our hands a little dirty and find direct limits for each of the examples of directed systems we gave.

\text{ }

As claimed in the introduction I claim that if we take the described sequence R^n and the obvious inclusions \iota_{i,j}:R^i\to R^j  that R^{\oplus\mathbb{N}}\cong\varinjlim R^n with the morphisms \varphi_n:R^n\to R^{\oplus\mathbb{N}} being the obvious inclusions (i.e. padding a n-tuple with infinitely many zeros at the tail of the n-tuple). Indeed, to prove this we first must show that \varphi_j\circ \iota_{i,j}=\varphi_i. To do this we note that

\text{ }

\varphi_j(\iota_{i,j}(x_1,\cdots,x_i))=\varphi_j\underbrace{(x_1,\cdots,x_i,0\cdots,0)}_{\in R^j}=\underbrace{(x_1,\cdots,x_i,0,\cdots)}_{\in R^{\oplus\mathbb{N}}}=\varphi_i(x_1,\cdots,x_i)

\text{ }

we now need to show that given any set of maps g_n:R^n\to N such that g_j\circ f_{i,j}=g_i there exists a unique f:R^{\oplus\mathbb{N}}\to N such that f\circ \varphi_n=g_n. But, this is clear enough. Namely, given some (x_n)\in R^{\oplus\mathbb{N}} choose some N\in\mathbb{N} for which x_n=0 for n\geqslant N and define f((x_n))=g_N((x_1,\cdots,x_N)). Note that since g_j\circ\iota_{i,j}=g_i this map is well-defined (i.e. independent of choice of N). Note that this map satisfies

\text{ }


\text{ }

moreover it’s clear that this map is unique since any element of R^{\oplus\mathbb{N}} is of the form \varphi_n(x_1,\cdots,x_n) for some n\in\mathbb{N}. Thus, the claim follows.

\text{ }

Somewhat surprisingly, given a set \{M_\alpha\}_{\alpha\in\mathcal{A}} of morphisms where we have defined the trivial directed system then the coproduct \displaystyle \bigoplus_{\alpha\in\mathcal{A}}M_\alpha is a direct limit. Indeed, if one notes that since the only f_{\alpha,\beta} map defined is f_{\alpha,\alpha}=\text{id}_{M_\alpha} for each \alpha\in\mathcal{A} we see that we may ignore most of the properties a direct limit (M,\{\varphi_\alpha\}) must follow. Indeed, with this revelation we see that (M,\{\varphi_\alpha\}) is a direct limit precisely when it satisfies the following universal property: given any set of maps g_\alpha:M_\alpha\to N there exists a unique map j:M\to N such that j\circ\varphi_\alpha=g_\alpha. We see that this is precisely the universal characterization for a coproduct of the set \{M_\alpha\}_{\alpha\in\mathcal{A}} of modules. We also see that this gives us some intuition about how “big” a direct limit can be. Indeed, since the relations between the M_\alpha are trivial (obviously) in the trivial directed system we see that the coproduct represents the “freest” possible gluing of the \{M_\alpha\} and so we can expect any other directed system with underlying set of modules \{M_\alpha\} to have a “smaller” direct limit. This is intuition rings fairly true, as shall be made clear soon enough.

\text{ }

Let us now see if we can find a nice direct limit of the directed system consisting of \{C^0(U)\}_{U\in\mathcal{U}_x} and \{\text{res}_{U,V}:C^0(U)\to C^0(V)\}. Intuitively a direct limit of this system should be a collection of functions who fundamentally don’t agree near x. To be particular, since we are taking the limit of this system over a shrinking set of neighborhoods it should be clear (at least intuitively) that if two functions are to agree on a neighborhood of x then they will be “equivalent” in the direct limit. Cashing in on this intuition we can define the set \mathcal{G}_x of germs at x to be the set C^0(X)/\sim where f\sim g if and only if there exists some U\in\mathcal{U}_x with f_{\mid U}=g_{\mid U}. It’s easy to verify that \mathcal{G}_x is a \mathbb{R}-space. It is not hard to see that \varinjlim C^0(U)\cong\mathcal{G}_x where the morphisms C^0(U)\to \mathcal{G}_x are given by f\mapsto [f].

\text{ }

Lastly, we had the example where we considered some fixed module M and defined \mathcal{F} to be its set of finitely generated submodules ordered by inclusion. We noted then that if we consider \{A\}_{A\in\mathcal{F}} and the inclusions \{\iota_{A,B}:A\hookrightarrow B\}_{A,B\in\mathcal{F},\; A\subseteq B} formed a direct system. We claim then that \underset{^\mathcal{F}}{\varinjlim}A\cong M. Indeed, define \varphi_A:A\to M to be the usual inclusion A\hookrightarrow M. We note then that evidently \varphi_B\circ\iota_{A,B}=\varphi_A. Moreover, given any set of mappings g_A:A\to N for every A\in\mathcal{F} with g_B\circ \iota_{A,B}=g_A we can merely define j:M\to N by j(x)=g_{\langle x\rangle}(x) where \langle x\rangle denotes, as per usual, the submodule generated by x and claim that j satisfies the needed properties. To see this we just have to note that j(\varphi_A(x))=j(x)=g_{\langle x\rangle}(x)=g_A\iota_{\langle x\rangle,A}(x)=g_A(x) (since we know that since A is a submodule containing x that \langle x\rangle\hookrightarrow A). Moreover, it’s clear that j is the unique such function, and so the claim follows. Thus, we may state this result as

\text{ }

Theorem: Every module is a direct limit of its finitely generated submodules.

\text{ }

\text{ }

Ok, fine, so for a bunch of specific examples we have seen what direct limits look like. This is good, it gives us intuition about what direct limits are, what they do, and why they might be interesting. It is now in our general interest though to prove that direct limits don’t exist in general, to verify that we aren’t just fantastically lucky pickers and that direct limits always exist given a directed system of modules. Indeed the following not only verifies that direct limits always exist (for given directed systems) but moreover that our intuition about directed limits being, in general, smaller than coproducts (they are, in fact, always a homomorphic image of a coproduct):

\text{ }

Theorem: Let R be a ring, \left(\mathcal{A},\leqslant\right) a preordered set and \left(\{M_\alpha\}_{\alpha\in\mathcal{A}},\{f_{\alpha,\beta}:M_\alpha\to M_\beta\}_{\alpha,\beta\in\mathcal{A},\; \alpha\leqslant\beta}\right) a directed system. Consider the direct sum \displaystyle M=\bigoplus_{\alpha\in\mathcal{A}}M_\alpha and consider the submodule

\text{ }

I=\left\langle \iota_\beta f_{\alpha,\beta}(x_\alpha)-\iota_\alpha(x_\alpha):\alpha\leqslant\beta,\; x_\alpha\in M_\alpha\right\rangle

\text{ }

where \iota_\alpha are the natural injections. Then, \displaystyle \left(M/I,\{\varphi_\alpha\}\right) (where \varphi_\alpha is the composition of \iota_\alpha postcomposed with the canonical projection \pi:M\to M/I) is a direct limit of the directed system.

Proof: We must first show that if \alpha\leqslant\beta then \varphi_\beta\circ f_{\alpha,\beta}=\varphi_\alpha but this is immediate since writing \varphi_\gamma=\pi\circ\iota_\gamma and using linearity we have that

\text{ }


\text{ }

where the last equality holds since \iota_\beta(f_{\alpha,\beta}(x_\alpha))-\iota_\alpha(x_\alpha)\in I. We now need to prove that given any set of maps g_\alpha:M_\alpha\to N there exists a unique map j:M/I\to N such that j\circ\varphi_\alpha=g_\alpha. To do this we define j:M/I\to N by the rule \displaystyle \sum_{\alpha\in\mathcal{A}}(x_\alpha)+I\mapsto \sum_{\alpha\in\mathcal{A}}g_\alpha(x_\alpha). Now, note that it’s not apriori clear that j is well-defined. To do this we merely note that if (x_\alpha)+I=(y_\alpha)+I then their difference (x_\alpha-y_\alpha)\in I but then that means that we can write (x_\alpha-y_\alpha) as a finite sum of the form \displaystyle \sum_{i}a_i((\varphi_\beta(f_{\alpha_i,\beta_i}(x_\alpha))-\varphi_\alpha(x_\alpha)) and so trivially by design we see that the image of this under j is \displaystyle \sum_{\alpha}a_i(g_\beta(f_{\alpha,\beta}(x_\alpha))-g_\alpha(x_\alpha))=0 by the assumption on the g_\alpha,g_\beta‘s. Thus, j is well-defined as desired. We are thus left to prove that this was the only conceivable way to define j if we want it to satisfy the golden properties. But, this is clear for clearly how any such j acts on M/I will be determined by how it acts on each element of the form \varphi_\alpha (since the union over each image of M_\alpha under \varphi_\alpha generates M/I) and we already are forced by the condition on j to define this in the way we already have. It follows from this that (M/I,\{\varphi_\alpha\}) is a direct limit of the direct system as desired.  \blacksquare

\text{ }

This constructed module being an example of a direct limit, and thus the unique direct limit up to isomorphism, shall serve as our standard go-to (the way that the usual direct sum serves as our usual representation for a coproduct). Thus, when we now write \varinjlim M_\alpha we shall mean, unless otherwise stated, the module constructed in the previous theorem.

\text{ }

This theorem, giving us a stereotypical direct product, indulges our understanding that the direct limit can be thought of as a two-step process: 1) gluing together (the coproduct) and 2) identifying things which are “eventually equal” in the sequence (the modding out by I).

\text{ }

\text{ }


[1] Dummit, David Steven., and Richard M. Foote. Abstract Algebra. Hoboken, NJ: Wiley, 2004. Print.

[2] Rotman, Joseph J. Advanced Modern Algebra. Providence, RI: American Mathematical Society, 2010. Print.

[3] Blyth, T. S. Module Theory. Clarendon, 1990. Print.

[4] Lang, Serge. Algebra. Reading, MA: Addison-Wesley Pub., 1965. Print.


November 30, 2011 - Posted by | Algebra, Module Theory, Ring Theory | , , , , , , , ,


  1. […] the unique isomorphism between and satisfies the appropriate identities between and it suffices to prove this for . […]

    Pingback by Direct Limits of Directed Sets « Abstract Nonsense | December 1, 2011 | Reply

  2. […] first example has been done before. Namely, the direct limit over this directed system is isomorphic to since, as is clearly by […]

    Pingback by Direct Limit of Rings (Pt. II) « Abstract Nonsense | December 4, 2011 | Reply

  3. […] methodology for doing this, for making this construction, is precisely the same as it was in the case of modules. The proof that this construction actually is a direct limit is exactly the same in both cases as […]

    Pingback by Direct Limit of Rings (Pt. III) « Abstract Nonsense | December 4, 2011 | Reply

  4. […] went with the case of direct limits, it is probably prudent to show that any inverse system of modules actually admits an inverse […]

    Pingback by Inverse Limits of Modules (Pt. II) « Abstract Nonsense | December 9, 2011 | Reply

  5. […] is our direct limit (note that we are assuming a direct limit of any form, not necessarily the standard, this makes things more clear). The key hint was (as was noted in the motivation) the fact that is […]

    Pingback by Relationship Between Hom and Limits (Modules)(Pt. I) « Abstract Nonsense | December 26, 2011 | Reply

  6. […] is no harm in identifying all tensor products of by with , the same way in which we identify direct and inverse limits with their particular, concrete […]

    Pingback by The Construction of the Tensor Product of Modules (Pt. II) « Abstract Nonsense | January 3, 2012 | Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: