localization of a ring

This entry is about the general notion of localization of a possible noncommutative ring. For the more restrictive but more traditional notion of localization of a commutative ring see there.




Given a (possibly noncommutative) unital ring RR there are many situations when certain elements or matrices can be inverted in a universal way obtaining a new “localized” ring S 1RS^{-1}R equipped with a localization homomorphism RS 1RR\to S^{-1}R under which all elements in SS are mapped to multiplicatively invertible elements (units). The latter property must be modified for Cohn localization at multiplicative set of matrices.

We can typically invert elements in a left or right Ore subset SRS\subset R or much more generally some multiplicative set or matrices (Cohn localization) etc. There are also some specific localizations like Martindale localizations in ring theory.

Localization “at” and “away from”

The common terminology in algebra is as follows.

For SS a set of primes, “localize at SS” means “invert what is not divisible by SS”; so for pp prime, localizing “at pp” means considering only pp-torsion.

Adjoining inverses [S 1][S^{-1}] is pronounced “localized away from SS”. Inverting a prime pp is localizing away from pp, which means ignoring pp-torsion.

See also lecture notes such as (Gathmann) and see at localization of a space for more discussion of this.

Evidently, this conflicts with more-categorial uses of “localized”; “inverting weak equivalences” is called localization, by obvious analogy, and is written as “localizing at weak equivalences”. This is confusing! It’s also weird: since a ring is a one-object Ab-enriched category with morphisms “multiply-by”, the localization-of-the-category RR “at pp” (or its Ab-enriched version, if saying that is necessary) really means the localization-of-the-ring R “away from p”.



The localization of a ring RR at a multiplicative subset SS is the commutative ring whose underlying set is the set of equivalence classes on R×SR \times S under the equivalence relation

(r 1,s 1)(r 2,s 2)uS(r 1s 2r 2s 1)u=0R. (r_1, s_1) \sim (r_2, s_2) \;\;\Leftrightarrow\;\; \exists u \in S \; (r_1 s_2- r_2 s_1) u = 0 \;\in R \,.

Write rs 1r s^{-1} for the equivalence class of (r,s)(r,s). On this set, addition and multiplication is defined by

r 1s 1 1+r 2s 2 1(r 1s 2+r 2s 1)(s 1s 2) 1 r_1 s_1^{-1} + r_2 s_2^{-1} \coloneqq (r_1 s_2 + r_2 s_1) (s_1 s_2)^{-1}
(r 1s 1 1)(r 2s 2 1)r 1r 2(s 1s 2) 1. (r_1 s_1^{-1})(r_2 s_2^{-1}) \coloneqq r_1 r_2 (s_1 s_2)^{-1} \,.

(e.g. Stacks Project, def. 10.9.1)


Basics of localizations of commutative rings are reviewed for instance in

  • Andreas Gathmann, Localization (pdf)

Discusison of the general concept in noncommutative geometry is in

  • Zoran Škoda, Noncommutative localization in noncommutative geometry, London Math. Society Lecture Note Series 330 (pdf), ed. A. Ranicki; pp. 220–313, math.QA/0403276.

Revised on July 25, 2014 10:12:20 by Urs Schreiber (