1: \begin{proof} Below, we assume that the initial data is sufficiently smooth, say $\psi_0, n_0 \in H^3(\T)$. In view of Proposition \ref{prop:local}, this yields a 

2: solution $$(\psi, n)\in C([0, T_{\rm max});\mathcal H^3 (\T))\cap C^1((0, T_{\rm max});\mathcal H^1 (\T))$$ 

3: for which all subsequent computations are rigorously justified. Invoking 

4: a standard density argument (see, e.g., \cite{Tao}) combined with the continuous dependence on initial data (and the asserted persistence of regularity), 

5: we can conclude that the result holds for $\mathcal H^1$-solutions.

6: 

7: Multiplying the first equation in \eqref{GP} by $\overline{\psi}$, integrating over $\T$, and taking the real part, we obtain

8: \[

9: \frac{d}{dt}M_{\rm c}(t) = \int_\T (Rn-\alpha)|\psi|^2 dx.

10: \]

11: Similarly, integrating the second equation in \eqref{GP} over $\T$ gives

12: \[

13: \varepsilon\frac{d}{dt} M_{\rm r}(t) = \int_\T P - (R|\psi|^2+\beta)n \, dx.

14: \]

15: Therefore, we have

16: \begin{align*}

17: \frac{d}{dt}M(t) &= \int_\T (Rn-\alpha)|\psi|^2 dx + P - (R|\psi|^2+\beta)n \, dx \\

18: &= \int_\T P dx - \alpha \int_\T |\psi|^2 dx - \beta \int_\T n \, dx \\

19: &\le P |\T| - \gamma M(t),

20: \end{align*}

21: with $\gamma = \min \{\alpha, \beta\}$. Integrating in time, yields

22: \begin{align*}

23: M(t) &\le e^{-\gamma t}M(0) + P |\T| \int_0^t e^{-\gamma(t-s)} ds \\

24: &= e^{-\gamma t}\left(M(0) - \frac{P |\T|}{\gamma} \right) + \frac{P |\T|}{\gamma} .

25: \end{align*}

26: In the case where $\alpha =\beta$ and $\varepsilon=1$, we see that 

27: this inequality actually becomes an equality. 

28: \end{proof}

29: