2

Let $F$ be a subring of $\Bbb R$ and let $A$ and $B$ be two non-empty sets. Suppose that the polynomial rings $F[(X_{\alpha})_{\alpha \in A\;}] \cong F[(X_{\beta})_{\beta \in B\;}]$ are isomorphic as rings. Does it follow that $A$ and $B$ have the same cardinality?

Some possible variations:

  1. We could see what happens if we assume that we have an $F$-modules isomorphism (or an $F$-algebras) between the two polynomial rings. In that case, I think that we have two free $F$-modules, and as $F$ is a commutative ring with unity, it can be shown (e.g. by quotienting by a maximal ideal) that the cardinalities are indeed the same.
  2. It would be interesting to see what happens if we assume that $F$is a subfield and $F((X_{\alpha})_{\alpha \in A\;}) \cong F((X_{\beta})_{\beta \in B\;})$, either a) as fields, or b) as $F$-algebras. This is maybe related to one of my previous questions. As for $F$-algebras, I believe that an argument of transcendence degree over $F$ is sufficient. The result is shown to be wrong if we accepted to take subfields of $\Bbb C$. I don't know if my main question is also wrong if we accept to take subrings of $\Bbb C$.

The result is true for $F=\Bbb Z$. The result of my main question and of $2a)$ is also true for $F=\Bbb Q$ : if $\Bbb Q[(X_{\alpha})_{\alpha \in A\;}] \cong \Bbb Q[(X_{\beta})_{\beta \in B\;}]$ as rings, then they are isomorphic fraction fields, which must have the same transcendence degree over $\Bbb Q$ (because a field isomorphism is a $\Bbb Q$-algebra isomorphism), so we are done. This could probably generalized to extensions of $\Bbb Q$ with finite transcendence degree, I think.

Interesting examples can be found here. Obviously we can easily build counter-examples if we can take any ring for $F$, like $F=\Bbb Z[x_1, \dots, x_n, \dots]$, in which case we have $F[x] \cong F[x,y]$. However, I don't think that $F$ embeds in $\Bbb R$ as a subring (it does as an additive subgroup, I think, because $\Bbb R \cong \Bbb Q^{(2^{\aleph_0})}$ as abelian groups). Here is also a related (but different) question. On MO, this similar question is interesting.

Thank you!

Watson
  • 23,793
  • It is not clear to me when you say "isomorphic" whether you mean "isomorphic as F-algebras" (correct notion) or "isomorphic as rings" – Juan Fran Aug 18 '16 at 13:43
  • 1
    In your last paragraph, $F=\mathbb{Z}[x_1, x_2, \ldots]$ does embed as a subring of $\mathbb{R}$, by sending the $x_i$ to pairwise distinct elements of a transcendence basis of the field $\mathbb{R}$ over $\mathbb{Q}$. – Pierre-Guy Plamondon Aug 18 '16 at 14:11
  • @Pierre-GuyPlamondon: so this is a counter-example to my main question? – Watson Aug 18 '16 at 14:23
  • @Pierre-GuyPlamondon : do you think that the "main result" is however true for $F=\Bbb R$ ? I have no simple argument as for $\Bbb Q$ because transcendence degree over $\Bbb R$ is not preserved by field isomorphism. – Watson Aug 18 '16 at 14:35
  • 1
    Yes, the result should be true whenever $F$ is a field. If $A$ and $B$ are finite, then their cardinal is the Krull dimensions of the polynomial rings, which is invariant under isomorphism. If they are infinite, then an isomorphism would send $F$ to $F$, since it is the set of invertible elements (plus zero), and would send each $X_\alpha$ to a polynomial involving finitely many of the $X_\beta$. Thus $|A|\geq |B|$; using the inverse isomorphism, you would get the inverse inequality. – Pierre-Guy Plamondon Aug 18 '16 at 14:59
  • @Someone : Related, if $F$ is a subfield, $A$ empty and $|B| \geq 1$: http://math.stackexchange.com/questions/2514/ – Watson Aug 19 '16 at 14:32
  • Actually my ring $F$ at the end embeds in $\Bbb R$. It suffices to take countably many algebraically independent (over $\Bbb Q$) elements $y_1, \dots, y_n, \dots \in \Bbb R$ and then $F \cong \Bbb Z[y_1, \dots] \hookrightarrow \Bbb R$. – Watson Nov 05 '16 at 10:19

1 Answers1

1
  • If you mean an isomorphism of rings, then the answer is no. For example, if $\{x_i\}_{i\in \mathbb{N}}$ is a subset of pairwise distinct elements of a transcendence basis of $\mathbb{R}$ over $\mathbb{Q}$, then $F=\mathbb{Z}[x_i]_{i\in\mathbb{N}}$ is a subring of $\mathbb{R}$ such that $$ F \cong F[X_1] \cong F[X_1, X_2] \cong \ldots $$

  • If you mean an isomorphism of $F$-algebras, then the answer becomes yes. Indeed, assume you have an isomorphism $f:F[X_\alpha]_{\alpha \in A}\to F[X_\beta]_{\beta \in B}$. Let $K$ be the field of fractions of $F$; then $f$ induces an isomorphism $\overline{f}:K[X_\alpha]_{\alpha \in A}\to K[X_\beta]_{\beta \in B}$. Now, $A$ is finite iff the Krull dimension of $K[X_\alpha]_{\alpha \in A}$ is finite; in that case, $|A|$ is the Krull dimension of $K[X_\alpha]_{\alpha \in A}$. Since the Krull dimension is invariant under isomorphism, we get $|A|=|B|$.
    Assume now that $A$ is infinite. By the above, this implies that $B$ is infinite as well. Then $\overline{f}$ sends each $X_\alpha$ to a polynomial involving finitely many of the $X_\beta$. Thus $|A|≥|B|$; using the inverse isomorphism, you would get the inverse inequality.

  • If you mean an isomorphism of rings, but assume further that $F$ is a field, then the answer is yes as well. Indeed, an isomorphism $f$ would send all elements of $F$ to elements of $F$, since $F\setminus\{0\}$ is the set of invertible elements of the polynomial rings. Then the above argument for the second case can still be applied.

  • In the second and third case $f$ induces a $K$-isomorphism between the field of fractions, and they should have the same transcendence degree over $K$. – user26857 Aug 18 '16 at 17:48
  • 1
    True, a nice argument. Just a detail: in the third case, it would not necessarily be a $K$-isomorphism, since its restriction to $K$ would only be an automorphism of $K$ (and not necessarily the identity), but this doesn't affect the conclusion. – Pierre-Guy Plamondon Aug 18 '16 at 17:57
  • a) Using the notations of the first paragraph, is it true that $F = \Bbb Q(x_1, \dots, x_n, \dots) \subset \Bbb R$ satisfy $F \cong F(X) \cong F(X,Y) \cong …$ as fields? (This could answer my previous question : http://math.stackexchange.com/questions/1895129) – Watson Aug 19 '16 at 09:14
  • b) However I think that $\Bbb R(X)$ can't be embedded in $\Bbb R$ as a field, is that right? – Watson Aug 19 '16 at 09:16
  • 1
    a) That is correct. b) I'm not sure right now. It could probably stand as a new question on its own here. – Pierre-Guy Plamondon Aug 19 '16 at 10:45