1: \begin{proof}

2:  We proceed similarly as in Lemma 2.2 in \cite{MiOel.1}.

3:  The computations are thus only formal although  

4:  they can be made rigorous exactly as therein. 

5:  By hypothesis, the  data satisfies $1-\gamma_0-\mu_{0,w}(0) >0$. 

6:  By continuity this also holds on a  time interval  

7:  $[0,t_0)$ small enough. 

8:  We proceed by contradiction and suppose that at time $t_0$ the mass $\mu_{0,w}(t_0)$ reaches $1-\gamma_0$.

9:  The equation on $\mu_{0,w}$ reads:

10:  $$

11:  \e \dt \mu_{0,w} - \beps (1- \mu_{0,w}) + \int_{\rr} \rhow_w(t,a) \zeta(w(t,a)) da = 0\;.

12:  $$

13: Multiplying it by $-1$ and estimating $\beps(t)\leq\bmax$, one deduces that

14:  $$

15:  \e \dt (1- \gamma_0 - \mu_{0,w})+ \bmax (1 - \gamma_0 - \mu_{0,w} ) + \gamma_0 \bmax - \int_{\rr} \rhow_w(t,a) \zeta(w(t,a)) da \geq 0 \; ,

16:  $$

17:  then  the lower bound on $\zteps$ implies

18:  $$

19:  \e \dt (1- \gamma_0 - \mu_{0,w})+ \bmax (1 - \gamma_0 - \mu_{0,w} ) + \gamma_0 \bmax  \geq \ztmin \mu_{0,w} \;.

20:  $$

21:  We transform the latter right hand side writing

22:  $$

23:  \ztmin \mu_{0,w} = -\ztmin (1-\gamma_0 - \mu_{0,w}) + \ztmin (1-\gamma_0) \; .

24:  $$

25:  Setting $q(t):=(1-\gamma_0 - \mu_{0,w}(t))$, one then has

26:  $$

27:  \e \dt q + (\ztmin+ \bmax) q \geq \ztmin - (\ztmin+ \bmax) \gamma_0 > 0 \; ,

28:  $$

29: the latter estimate being true under the hypothesis that $\gamma_0 < \ztmin/(\ztmin+ \bmax)$. 

30: The conclusion then follows integrating the latter inequality in time

31: $$

32: q(t_0) >  \exp\left( - \frac{(\bmax+\ztmin)t_0}{\e} \right) q(0) > 0,

33: $$

34: under the hypothesis that $\gamma_0 < (1-\mu_{0,w}(0))$. 

35: But this contradicts the assumption that $q(t_0)=0$, which ends the proof.

36: \end{proof}

37: