The origin of this question comes from a colleague, who asked "Can you
have two nonisomorphic rings R and S, such that the additive group of
R is isomorphic to the additive group of S, and the multiplicative
semigroup of R is isomorphic to the multiplicative semigroup of S"?
My answer, after some thought, was "yes", with the following example.
Let F be a field, and let R=F[x], S=F[x,y]. The isomorphism of (R,+)
and (S,+) can be done explicitly, as they are vector spaces of F of
dimension aleph_0, and one can give an explicit bijection by ordering
the monomials in R by degree, and the monomials in S by total degree
and lexicographically among those of the same degree.
To prove the isomorphism between (R,*) and (S,*), I argued as follows:
the cardinality of the set of monic irreducible polynomials in F[x] is
|F|*aleph_0: for infinite F, there are at least |F| monic irreducibles
(the linear polynomials), and since |F[x]|=|F|, there are at most |F|
of them. For finite F, there are at least aleph_0 monic irreducibles
(at least one for each positive integer n, given by the finite
extensions and the primitive element theorem), and |F[x]|=aleph_0, so
you get aleph_0 again. Since |F[x,y]|=|F|*aleph_0 as well, this gives
that the cardinality of the set of irreducibles in S is the same. If M
is the free commutative monoid on a set with |F|*aleph_0 elements,
then (R,*) and (S,*) are both isomorphic to F x M, since they are both
UFDs with isomorphic (in fact, identical) group of units.
That R and S are not isomorphic follows from any number of arguments
(R is a PID, S is not; the Krull dimension of S is one more than that
of R, etc).
Now, the isomorphism between additive structures can be given
constructively/explicitly, without having to invoke the Axiom of
Choice. Can the same be done with the isomorphism of the
multiplicative structures? I’m not sure if I’m invoking AC above; I
may be in arguing some of the cardinalities or the inequalities
between cardinalities.
If not in general, can it be done for some specific fields, say F=Q,
where we have explicit well-orderings of Q; or C, where the set of
monic irreducibles of F[x] is easy?
(I think that replacing F with any UFD will also give counterexamples
to the original question, but then we may need choice to select
representatives from the equivalence classes of irreducible elements,
so let’s stick to fields where there is an obvious choice of
representative).
–
Arturo Magidin