One can view view Lagrange's Theorem $$|G/H|=|G|/|H|$$ and the Rank-Nullity Theorem $$\dim(V/U)=\dim(V)-\dim(U)$$ as directly analogous. Does anyone know a high-level explanation of this analogy? I suspect it may involve the word "valuation" or the word "Artinian".
-
9$\begingroup$ They are the same theorem, when working with vector spaces over a finite field. $\endgroup$– S. Carnahan ♦Commented Oct 18, 2013 at 16:17
-
2$\begingroup$ I believe it involves the word "logarithm". $\endgroup$– Igor RivinCommented Oct 18, 2013 at 19:26
1 Answer
As Igor Rivin suggests, the explanation involves the word "logarithm". In fact, if $\mathfrak C$ is the category of groups, you can define an invariant of the category \begin{align*} \log\lvert{-}\rvert:Ob(\mathfrak C)&\to \mathbb R_{\geq 0}\cup\{\infty\}\\ M&\mapsto \log \lvert M\rvert \end{align*} This invariant is additive, in the sense that, given a surjective morphism $\phi:M\to N$ in the category, $\log\lvert M\rvert=\log\lvert\ker(\phi)\rvert+\log\lvert N\rvert$.
In any category you can give analogously the definition of additive invariant. If your category is Abelian the usual thing is to say that an invariant $i:Ob(\mathfrak C)\to \mathbb R_{\geq 0}\cup\{\infty\}$ is additive provided $i(B)=i(A)+i(C)$ for any short exact sequence $0\to A\to B\to C\to 0$.
The dimension of vector spaces and the composition length of modules are examples of such additive invariants.
There is another property that the above invariants satisfy, that is, they commute with direct unions, that is, if $$M=\underset{\to}{\bigcup}M_\alpha$$ then, $i(M)=\sup_\alpha i(M_\alpha)$. Such an invariant is called upper continuous. In general, additive and upper continuous invariants are called length functions (as they generalize the usual composition length).
There are some attempts in the literature to classify in some sense all the additive invariants of some category, typically a category of modules. See for example:
Northcott, D. G.; Reufel, M. A generalization of the concept of length. Quart. J. Math. Oxford Ser. (2) 16 (1965) 297-–321. MR195905
Vámos, P. Additive functions and duality over Noetherian rings. Quart. J. Math. Oxford Ser. (2) 19 (1968) 43–-55. MR223434
-
$\begingroup$ This is all good, but I think that the Lagrange's theorem (in the category of finite groups) is a bit more subtle, because it is more general that there being a homomorphism $M\to N$.. Note that $M/N$ in this context is the set of cosets of $N$, so it's just a set, not a group. $\endgroup$ Commented Mar 6, 2016 at 17:00
-
$\begingroup$ You are right, let's say that I view an analogy between the two things as long as Lagrange's Theorem is applied to a quotient over a normal subgroup. It is in fact true that the general statement is more subtle, but this difference has no analogy in the category of vector spaces (nor in any abelian category, where every subobject is the kernel of some morphism). $\endgroup$ Commented Mar 6, 2016 at 21:23
-