An element m of a module M over a ringR is called a torsion element of the module if there exists a regular element r of the ring that annihilates m, i.e., In an integral domain, every non-zero element is regular, so a torsion element of a module over an integral domain is one annihilated by a non-zero element of the integral domain. Some authors use this as the definition of a torsion element, but this definition does not work well over more general rings. A module M over a ring R is called a torsion module if all its elements are torsion elements, and torsion-free if zero is the only torsion element. If the ring R is an integral domain then the set of all torsion elements forms a submodule of M, called the torsion submodule of M, sometimes denoted T. If R is not commutative, T may or may not be a submodule. It is shown in that R is a right Ore ringif and only if T is a submodule of M for all right R modules. Since right Noetherian domains are Ore, this covers the case when R is a right Noetherian domain. More generally, let M be a module over a ringR and S be a multiplicatively closed subset of R. An element m of M is called an S-torsion element if there exists an element s in S such that s annihilates m, i.e., In particular, one can take for S the set of regular elements of the ring R and recover the definition above. An element g of a group G is called a torsion element of the group if it has finite order, i.e., if there is a positive integerm such that gm = e, where e denotes the identity element of the group, and gm denotes the product of m copies of g. A group is called a torsion group if all its elements are torsion elements, and a if the only torsion element is the identity element. Any abelian group may be viewed as a module over the ring Z of integers, and in this case the two notions of torsion coincide.
Examples
Let M be a free module over any ring R. Then it follows immediately from the definitions that M is torsion-free. In particular, any free abelian group is torsion-free and any vector space over a fieldK is torsion-free when viewed as the module over K.
By contrast with Example 1, any finite group is periodic and finitely generated. Burnside's problem asks whether, conversely, any finitely generated periodic group must be finite.
In the modular group, Γ obtained from the group SL of two by two integer matrices with unit determinant by factoring out its center, any nontrivial torsion element either has order two and is conjugate to the element S or has order three and is conjugate to the element ST. In this case, torsion elements do not form a subgroup, for example, S·ST = T, which has infinite order.
The abelian group Q/Z, consisting of the rational numbers, is periodic, i.e. every element has finite order. Analogously, the module K/K over the ring R = K of polynomials in one variable is pure torsion. Both these examples can be generalized as follows: if R is a commutative domain and Q is its field of fractions, then Q/R is a torsion R-module.
Suppose that R is a principal ideal domain and M is a finitely-generated R-module. Then the structure theorem for finitely generated modules over a principal ideal domain gives a detailed description of the module M up to isomorphism. In particular, it claims that where F is a free R-module of finite rank and T is the torsion submodule of M. As a corollary, any finitely-generated torsion-free module over R is free. This corollary does not hold for more general commutative domains, even for R = K, the ring of polynomials in two variables. For non-finitely generated modules, the above direct decomposition is not true. The torsion subgroup of an abelian group may not be a direct summand of it.
Assume that R is a commutative domain and M is an R-module. Let Q be the quotient field of the ring R. Then one can consider the Q-module obtained from M by extension of scalars. Since Q is a field, a module over Q is a vector space, possibly, infinite-dimensional. There is a canonical homomorphism of abelian groups from M to MQ, and the kernel of this homomorphism is precisely the torsion submodule T. More generally, if S is a multiplicatively closed subset of the ring R, then we may consider localization of the R-module M, which is a module over the localization RS. There is a canonical map from M to MS, whose kernel is precisely the S-torsion submodule of M. Thus the torsion submodule of M can be interpreted as the set of the elements that 'vanish in the localization'. The same interpretation continues to hold in the non-commutative setting for rings satisfying the Ore condition, or more generally for any right denominator setS and right R-module M.
The concept of torsion plays an important role in homological algebra. If M and N are two modules over a commutative ringR, Tor functors yield a family of R-modules Tori. The S-torsion of an R-module M is canonically isomorphic to TorR1 by the long exact sequence of TorR*: The short exact sequence of R-modules yields an exact sequence, hence is the kernel of the localisation map of M. The symbol Tor denoting the functors reflects this relation with the algebraic torsion. This same result holds for non-commutative rings as well as long as the set S is a right denominator set.
