Invariant basis number
In mathematics, more specifically in the field of ring theory, a ring has the invariant basis number (IBN) property if all finitely generated free left modules over R have a well-defined rank. In the case of fields, the IBN property becomes the statement that finite dimensional vector spaces have a unique dimension.
Equivalently, this means there do not exist distinct positive integers m and n such that Rm is isomorphic to Rn.
Rephrasing the definition of invariant basis number in terms of matrices, it says that, whenever A is an m by n matrix over R and B is an n by m matrix over R such that AB=1 and BA=1, then m=n. This form reveals that the definition is left-right symmetric, so it makes no difference whether we define IBN in terms of left or right modules; the two definitions are equivalent.
Note that the isomorphisms in the definitions are not ring isomorphisms, they are module isomorphisms.
The main purpose of the invariant basis number condition is that free modules over an IBN ring satisfy an analogue of the dimension theorem for vector spaces: any two bases for a free module over an IBN ring have the same cardinality. Assuming the ultrafilter lemma (a strictly weaker form of the axiom of choice), this result is actually equivalent to the definition given here, and can be taken as an alternative definition.
The rank of a free module Rn over an IBN ring R is defined to be the cardinality of the exponent m of any (and therefore every) R-module Rm isomorphic to Rn. Thus the IBN property asserts that every isomorphism class of free R-modules has a unique rank. The rank is not defined for rings not satisfying IBN. For vector spaces, the rank is also called the dimension. Thus the result above is in short: the rank is uniquely defined for all free R-modules iff it is uniquely defined for finitely generated free R-modules.
Any field satisfies IBN, and this amounts to the fact that finite dimensional vector spaces have a well defined dimension. Moreover, any commutative ring (except in the trivial case where 1 = 0) satisfies IBN, as does any left-Noetherian ring and any semilocal ring.
An example of a ring that does not satisfy IBN is the ring of column finite matrices , the matrices with coefficients in a ring R, with entries indexed by and with each column having only finitely many non-zero entries. That last requirement allows us to define the product of infinite matrices MN, giving the ring structure. A left module isomorphism is given by:
From this isomorphism, it is possible to show (abbreviating ) that S≅Sn for any positive integer n, and hence Sn≅Sm for any two positive integers m and n. There are other examples of non-IBN rings without this property, among them Leavitt algebras as seen in (Abrams 2002).
Abrams, Gene; Ánh, P. N. (2002), "Some ultramatricial algebras which arise as intersections of Leavitt algebras", J. Algebra Appl. 1 (4): 357–363, doi:10.1142/S0219498802000227, ISSN 0219-4988, MR 1950131