5

I'm interested in defining a topology on the ring $R[[X_i]]$ of formal power series in $(X_i)_{i\in I}$, where $R$ is a topological ring and $I$ is a (possibly infinite) index set. The wiki article discusses several options for this, and it seems that the most natural topology satisfies

$(x_n)_{n\in\infty}$ converges iff for every monomial $X^\alpha$ (i.e. $\alpha$ is a finite multiset of indexes in $I$), $([X^\alpha]x_n)_{n\in\infty}$ converges in the topology on $R$.

(The other main option discussed as a natural topology for $R[[X_i]]$ is equivalent to this one where $R$'s topology is ignored and replaced by the discrete topology.) My question is:

  • How is this topology defined in terms of open sets? Can it be described as a special case of another topology, i.e. Krull topology or product topology? (The wiki answers this question in the univariate case but I'm more interested in the multivariate case.)

    • More generally, this can be seen as a question of how to translate a specification of a topology in terms of convergence to an explicit definition from open sets.
  • Is this topology metrizable (assuming $R$ is)? Again, the wiki discusses this in the case when $R$ is discrete and there is only one variable via $d(x,y)=2^{-k}$ where $k$ is the smallest nonzero coefficient of $x-y$, but leaves the multivariate case for the reader.

  • I would suggest splitting your two questions, linking one to the other. It's possible to give a very general answer to the first question, not so for the second. – Najib Idrissi Feb 02 '15 at 16:04
  • @NajibIdrissi Well, I'd prefer to keep this question focused on formal power series themselves, the side note about convergence can be dealt with elsewhere, if it is deemed too broad. Answer however you feel appropriate, I might even write an answer myself. – Mario Carneiro Feb 02 '15 at 16:06
  • Since $R[[x]]$ is the completion of $R[x]$ with respect to the $(x)$-adic topology, one can more generally ask what is the topology on the completion of a ring (which is the topology on the completion of the underlying group), see 1, 2. – Elías Guisado Villalgordo Feb 28 '24 at 09:58

1 Answers1

3

The given topology is equivalent to the product topology, under the natural map $R[[X_i]]\to R^{{\cal M}(I)}$ (where ${\cal M}(I)$ is the set of all finite multisets of $I$, isomorphic to the set of monomials in $R[[X_i]]$) that maps a power series to its coefficient function, because a series in the product topology also converges iff all of its projections converge, which here is the same as saying that all of its coefficients converge.

If $R$ itself has a Krull ($I$-adic) topology with generating ideal $J$, there is a natural generating ideal for an $I$-adic topology given by $$G=(J,(X_i)_{i\in I})=\{x\in R[[X_i]]:[1]x\in J\},$$ the set of all power series with constant term in $J$, but only in certain cases does this generate the same topology as the first one. Note that in this case $G^n$ is the set of power series where each degree-$k$ coefficient is in $J^{n-k}$ for all $k\le n$, and by definition this defines a neighborhood base of $0$. An equivalent characterization of the product topology here takes as neighborhood base of $0$ the sets $$U_S=\{x\in R[[X_i]]:\forall\alpha\in S, [X^\alpha]x\in{\cal O}_R\}$$ for each finite $S$, where ${\cal O}_R$ is the topology on $R$. Since for each $S$ there is an $n$ such that $U_S\subseteq G^n$, we have that the $G$-adic topology is finer than the product topology, and if $I$ is finite each $G^n$ is a basic open set, so for discrete $R$ this topology matches the product topology.

When $I$ is infinite $G^n$ will not be a basic open set, and indeed is not open, so it generates a strictly finer topology. This shows in the convergence of series like $\sum_{i=1}^\infty X_i$ (supposing that $I=\Bbb N$). In the product topology this is a convergent series (since every monomial term converges), but in the $G$-adic topology we require that there is an $n$ such that $\sum_{i=n}^\infty X_i\in G^2$, i.e. all degree-$1$ coefficients must stabilize after some $n$, and this is not true.

Regarding the question of metrizability, note that ${\cal M}(I)$ is countable iff $I$ is countable. Then since the product topology $X^S$ is metrizable iff $X$ is metrizable and $S$ is countable (proof), it follows that $R[[X_i]]$ is metrizable under the given topology iff $I$ is countable. I don't think there is a "natural" metric, though, in the case when $I$ is countable, unless you already have well-ordering on $I$ on which to define a bijection $f:\Bbb N\to{\cal M}(I)$, from which you can define $$d(x,y)=\sum_{n\in\Bbb N}\frac{d([X^{f(n)}]x,[X^{f(n)}]y)}{2^{-n}}.$$