6

Is there any example showing that the composition of morphisms is not necessarily associative?

isomorphismes
  • 4,758
Jojo
  • 71
  • 8
    Composition of morphisms is always associative. – Qiaochu Yuan Nov 14 '12 at 22:28
  • In fact, it is an axiom of category that composition of morphisms is associative. – tomasz Nov 14 '12 at 22:50
  • 4
    If you would like to know examples of "non-associative categories", specializing to one object these are sets with a non-associative binary operation and a unit element, which include non-associative algebras. There is a list on wikipedia: http://en.wikipedia.org/wiki/Non-associative_algebra – Martin Brandenburg Nov 15 '12 at 11:46
  • 1
    If you meant morphisms in the sense of category theory, then ([tag:category-theory]) tag would be good for this question. – Martin Sleziak Jul 23 '13 at 18:29

1 Answers1

5

I don't know how much category theory you know, but I guess you can look up any terms you don't recognize.

In a cateogry, composition is associative per definition. However, when generalizing categories to higher categories ($n$-categories), it is sometimes useful not to demand associativity, but only "weak associativity". Weak associativity is the same as associativity up to isomorphism in the layer above. That is, for any triple $f,g,h$ of morphisms where $gf$ and $hg$ are defined, there is an "isomorphism of morphisms" $F: (hg)f \rightarrow h(gf)$.

For example, if we define a path in a topological space $X$ to be a continuous function $\alpha:[0,1]\rightarrow X$, there is a well defined operation of "composition" of paths which is weakly associative, but not associative. In this case we get a 2-category where composition is associative up to homotopy.

Espen Nielsen
  • 3,983
  • 18
  • 32
  • Of course, we have to take homotopy classes of homotopies in order to have an associative composition for those... – Zhen Lin Nov 14 '12 at 22:44
  • 2
    Yes, this chain goes on to infinity, and is the basis for the homotopy hypothesis, that $\infty$-groupoids are essentially the same as topological spaces. (See for example http://ncatlab.org/nlab/show/homotopy+hypothesis) – Espen Nielsen Nov 14 '12 at 22:49