1: \begin{proof}

2:   We know that linear combinations of martingales are martingales. This extends easily to probability mixtures of martingales. Indeed, let $0\le s\le t$ and let $U$ be positive bounded and $\Grond_s$-measurable. Then, by Fubini-Tonelli's theorem :

3:   \begin{align*}

4:     \gesp{Z_t U} &= \gesp{\esp{ M^\omega_t} U} = \esp{\gesp{M^\omega_t U}} \\

5: &= \esp{\gesp{M^\omega_s U}} &\text{($M^\omega$ is a martingale)}\\

6: &= \gesp{\esp{M^\omega_s} U} = \gesp{Z_s U}\,.

7:   \end{align*}

8: Observe that if $\omega_1,\omega_2$ are paths, then we can compute the quadratic covariation

9: \begin{equation*}

10:   \label{par:eq:3}

11: d\crochet{M^{\omega_1},M^{\omega_2}}_t = M^{\omega_1}_t M^{\omega_2}_t \beta^2 \un{\omega_1(t)=\omega_2(t)}\, dt.

12: \end{equation*}

13: Therefore, we have formally:

14: \begin{align*}

15:   d\crochet{Z,Z}_t &= d\crochet{\int \PP(d\omega_1) M^{\omega_1}, \int \PP(d\omega_2) M^{\omega_2}}_t \\

16: &= \int \PP^{\otimes 2}(d\omega_1,d\omega_2)d\crochet{M^{\omega_1},M^{\omega_2}}_t  \\

17: &= \beta^2 Z_t^2  \unsur{Z_t^2} \int \PP^{\otimes 2}(d\omega_1,d\omega_2)M^{\omega_1}_t M^{\omega_2}_t  \un{\omega_1(t)=\omega_2(t)}\, dt \\

18: &=  Z_t^2 \beta^2 \, I_t\, dt.

19: \end{align*}

20: This again can be made rigorous by writing $N_t = Z_t^2 -\beta^2 \intot Z_s^2 I_s \, ds$ as a probability mixture of martingales:

21: \begin{equation*}

22:   \label{par:eq:4}

23:   N_t =  \int \PP^{\otimes 2}(d\omega_1,d\omega_2) (M^{\omega_1}_t M^{\omega_2}_t - \beta^2 \intot M^{\omega_1}_s M^{\omega_2}_s \un{\omega_1(s)=\omega_2(s)}\, ds)\,.

24: \end{equation*}

25: \end{proof}

26: