1: \begin{abstract}

2: Perturbation theory for non-Abelian gauge theories at finite temperature

3: is plagued by infrared divergences caused by magnetic soft modes 

4: $\sim g^2T$, which correspond to the fields of a 3d Yang-Mills theory. 

5: We revisit a gauge invariant resummation scheme 

6: to solve this problem by self-consistent mass generation using an auxiliary scalar field, 

7: improving over previous attempts in two respects. First, we generalise earlier

8: SU($2$) treatments to SU($N$).

9: Second, we obtain a gauge independent two-loop gap equation, correcting

10: an error in the literature. The resulting two-loop 

11: approximation to the magnetic mass represents a $\sim 15\%$ correction to the

12: leading one-loop value, indicating a reasonable convergence of the resummation.

13: \end{abstract}

14: