1: \begin{definition}\label{defYM} Setting 

2: $\hat{f}_r=\hat{f}\left(r,\cdot,U_r,\sqrt{\frac{d\langle N \rangle}{dV}}(r)\right)$, we define  $M$ 

3: as the cadlag version  of the martingale $t \mapsto  \mathbbm{E}\left[\xi + \int_0^T \hat{f}_rdV_r\middle|\mathcal{F}_t\right]-\mathbbm{E}\left[\xi + \int_0^T \hat{f}_rdV_r\middle|\mathcal{F}_0\right]$.

4:  

5: It admits a cadlag version taking into account

6:  Theorem 4 in Chapter IV of \cite{dellmeyerB}, since the stochastic basis fulfills the usual conditions.

7: We denote by $Y$ the cadlag process defined by 

8: $Y_t = \xi+\int_t^T\hat{f}\left(r,\cdot,U_r,\sqrt{\frac{d\langle N \rangle}{dV}}(r)\right)dV_r - (M_T - M_t)$.

9: This will be called the {\bf cadlag reference process} and we will

10:  often omit its

11: dependence to $(\dot U,N)$.

12: \end{definition}

13: