# Covariance of stochastic process

I need to calculate the expected value $$\mathbb{E}[\eta(s) \cdot \eta( \tilde{s} )]$$, where $$\eta$$ is an Ornstein-Uhlenbeck process defined by the stochastic differential equation $$\begin{equation*} \tau_{\eta} \frac{d \eta(t)}{dt} = - \eta(t) + \sqrt{2(\tau + \tau_{\eta})} \sigma \xi(t). \end{equation*}$$ $$\tau, \tau_{\eta}, \sigma \in \mathbb{R}$$ are constants and $$\xi(t)$$ is an Gaussian white noise process $$\xi(t) = \frac{dW(t)}{dt}$$. \

The needed expected value is equal to the Covariance $$Cov(\eta(s), \eta(\tilde{s}))$$, if he Ornstein-Uhlenbeck processes have mean zero. I also calculated the explicit solution of the SDE \begin{align*} \eta(s) = \exp \left( - \frac{s}{\tau_{\eta}} \right) \left[ \eta_0 + \int_0^s \frac{1}{\tau_{\eta}} \exp \left( \frac{x}{\tau_{\eta}} \right) \sqrt{2(\tau + \tau_{\eta}) } \sigma dW_x \right], \end{align*} where $$W_x$$ is a Brownian motion. I inserted the solution of the stochastic differential equation in the expected value with two different time variables $$s, \tilde{s}$$. Because the increments of the Brownian motion have mean zero, the mixed terms are zero. This leads to \begin{align*} \mathbb{E} \left[ \eta(s) \cdot \eta(\tilde{s}) \right] &= \mathbb{E} \left[ \exp \left( - \frac{s + \tilde{s}}{\tau_{\eta}} \right) \eta_0^2 \right] + \mathbb{E} \left[ n_0^2 + \int_0^s \int_0^{\tilde{s}} \frac{1}{\tau_{\eta}^2} \exp \left( \frac{x + y}{\tau_{\eta}} \right) \cdot 2(\tau + \tau_{\eta}) \sigma^2 dW_x dW_y \right] \end{align*} How can I calculate the expected value of the last double integral. It probably has something to do with the Covariance of the 2 Brownian motions $$dW_x, dW_y$$.

• I think that calculating ${d \over d\tilde{s}} \mathbb{E}[\eta(s) \cdot \eta( \tilde{s} )]$ will give you a closed equation in terms of the covariance. Jul 21, 2021 at 12:32

The explicit solution of your SDE is of the form $$\eta(s)=e^{-f(s)}\Big[\eta_0+\int_0^sae^{f(x)}\,dW_x\Big]\,.$$ The expectation $$E[\eta(s)\,\eta(\tilde s)]$$ is easily obtained from $$E[\int_0^se^{f(x)}\,dW_x]=0$$ and $$E\Big[\int_0^se^{f(x)}\,dW_x\cdot\int_0^\tilde{s}e^{f(x)}\,dW_x\Big]=\int_0^{\min(s,\tilde s)}e^{2f(x)}\,dx\,.$$