A simple object of an A\mathcal A is an object S≇0S\not\cong0 such that every U↪SU\hookrightarrow S has either U≅0U\cong0 or is an isomorphism. Equivalently, its only subobjects are 00 and SS, where subobjects are monomorphisms into SS identified up to isomorphism over SS.

Examples and a distinction

In vector spaces, the simple objects are the one-dimensional spaces. In a module category they are precisely . All one-dimensional kk-vector spaces are isomorphic, so there is one isomorphism class of simples, although there are many objects.

A simple object cannot split as a direct sum of two nonzero objects. The converse fails: the abelian group Z/4Z\mathbb Z/4\mathbb Z has the proper nonzero subgroup 2Z/4Z2\mathbb Z/4\mathbb Z, but cannot be written as a direct sum of two nonzero subgroups.

Endomorphisms

A nonzero map between simple objects is an isomorphism: its kernel must vanish and its image must be the target. Thus End⁡(S)\operatorname{End}(S) is a division ring. If the category is kk-linear, Hom spaces are finite-dimensional, and kk is algebraically closed, then End⁡(S)=kid⁡S\operatorname{End}(S)=k\operatorname{id}_S. Indeed, every endomorphism has an eigenvalue in the sense that its minimal polynomial over kk has a root; in a division algebra this forces that endomorphism to be scalar.

References

P. Etingof, S. Gelaki, D. Nikshych, and V. Ostrik, Tensor Categories (2015), §1.5 and Proposition 1.8.4; author-hosted text.