1: \begin{abstract}

2: \noindent The Golomb-Welch conjecture deals with the existence of perfect $e$%

3: -error correcting Lee codes of word length $n,$ $PL(n,e)$ codes. Although

4: there are many papers on the topic, the conjecture is still far from being

5: solved. In this paper we initiate the study of an invariant connected to

6: abelian groups that enables us to reformulate the conjecture, and then to

7: prove the non-existence of linear $PL(n,2)$ codes for $n\leq 12$. Using this

8: new approach we also construct the first quasi-perfect Lee codes for

9: dimension $n=3,$ and show that, for fixed $n$, there are only finitely many

10: such codes over $%

11: %TCIMACRO{\U{2124} }%

12: %BeginExpansion

13: \mathbb{Z}

14: %EndExpansion

15: .$

16: \end{abstract}