This is stated in [Rigid modules over preprojective algebras][1] by Geiss, Leclerc and Schröer, Sections 8.1-8.3. As noted in my comment a proof of the $n=5$ statement can be found in [The module theoretic approach to quasi-hereditary algebras][2]. (These are of course not the original references, I doubt that there is one original reference)
Note also that you can split up the $n\geq 5$ case in $n=5$, which is the tame case, i.e. there is a classification of the infinitely many indecomposable modules, and the $n>5$ case, where a classification is not possible.

EDIT: There is a proof in [Semicanonical bases and preprojective algebras][3] by Geiss, Leclerc, Schröer, Section 9, via the theory of Galois coverings. The Auslander-Reiten quiver of these algebras can also be found there. The statement was long before and is probably part of the folklore in the representation theory of quivers.


  [1]: http://www.math.uni-bonn.de/~schroer/preprints/rigid.pdf
  [2]: http://www.mathematik.uni-bielefeld.de/~ringel/opus/good-d-r.pdf
  [3]: http://arxiv.org/PS_cache/math/pdf/0402/0402448v2.pdf