1: \begin{abstract}

2: We reconsider in some detail a construction 

3:  allowing (Borel) convergence of  

4: an alternative perturbative expansion,

5: for specific physical quantities of asymptotically free

6: models. The usual perturbative expansions     

7: (with an explicit mass dependence) 

8: are transmuted into expansions in $1/F$, where 

9: $F \sim 1/g(m)$ for $m \gg \Lambda$ while $F \sim (m/\Lambda)^\alpha$

10: for $m \lsim \Lambda$, 

11: $\Lambda$ being the basic scale and $\alpha$ given by

12: renormalization group coefficients. (Borel) convergence holds 

13: in a range of $F$ which corresponds

14: to reach unambiguously the strong coupling infrared regime near

15: $m\to 0$, which can define certain ``non-perturbative" quantities, such

16: as the mass gap, from a resummation of this alternative expansion. 

17: Convergence properties can be further improved,  

18: when combined with $\delta$ expansion (variationally improved 

19: perturbation) methods. We

20: illustrate these results by re-evaluating, from purely perturbative

21: informations, the $O(N)$ Gross-Neveu model mass gap, known for arbitrary $N$

22: from exact S matrix results.  Comparing different levels of

23: approximations that can be defined within our framework, 

24: we find reasonable agreement with the exact result.

25: \end{abstract}

26: