1: \begin{proof}

2: % Formally this result is an integration by parts:

3: % Testing the equation with $\theta$ and integrating by parts yields

4: % \begin{equation*}

5: %  -\int_{\H} r\theta(r,z,T)f(r,z,T)\, d(r,z)+\int_{\H} r\theta(r,z,0)f(r,z,0)\, d(r,z)+\int\int_{\partial \H}f\,\xi\, v\cdot \hat n dS+\int_{\H} \chi\theta =0\,.

6: % \end{equation*}

7: % 

8: % Thanks to the boundary conditions ($v\cdot \hat n=0$ at $\partial \H$ since  $u^r=0$ at $\partial \H$ and, by definition, $v^r=ru^r$) we have 

9: % \begin{equation*}

10: %  \int_{\H} r\theta(r,z,T)f_T(r,z)\, d(r,z)-\int_{\H} r\theta(r,z,0)f(r,z,0)\, d(r,z)+\int \int_{\H} \chi\theta =0\,.

11: % \end{equation*}

12: %

13: % Rigorously one considers an admissible function $\beta$, regularizes, integrates by parts and eventually let the regularization parameter go to zero.

14: % Define

15: % \begin{equation}\label{e1}

16: %  r\partial_t\trhoe+v\cdot\nabla \trhoe= s_{\eps}

17: % \end{equation}

18: %and

19: %\begin{equation}\label{e2}

20: %  -r\partial_t\tfie-v\cdot\nabla \tfie=g_{\eps}+\chi

21: % \end{equation}

22: %where  $\tfie:=\beta(f)\ast\etae$, $\trhoe:=\beta(\theta)\ast\etae$ and $s_{\eps}, g_{\eps}\stackrel{\eps\rightarrow 0}{\rightarrow} 0$ in $L^1((0,T); L^1_{\rm{loc}}(\H))$. Then testing the equation \eqref{e1} by $\trhoe$, integrating by parts and using \eqref{e2} we have 

23: %\begin{multline}

24: % \int_0^t\int_{\H}(\chi+g_{\eps})\trhoe\, d(r,z)\,dt-\int_0^t\int_{\H} s_{\eps}\tfie\, d(r,z)\,dt\\=\int_{\H} r\tfie(t=0)\trhoe(t=0)\, d(r,z)-\int_{\H} r\tfie(t=T)\trhoe(t=T)\, d(r,z)\,.

25: %\end{multline}

26: %We obtain the claim letting $\varepsilon\rightarrow 0$.

27: %

28: %\improvement{actually to make it really rigorous one should also cut-off in space... Shall we keep or remove this proof?}

29: %\end{proof}

30: