Prop:Let $(X_t)$ be a $\mathbb{R}^d$ valued stochastic process with transition probability $P_t(x,dy)$. We assume there exsist probabilities on $\mathbb{R}^d$ $\{m_t\}_{t\geq 0}$ s.t., $P_t (x,B)=m_t(B-x)$, $m_0=\delta _0$(Dirac measure), $m_t*m_s=m_{t+s}$, $m_t(B)=m_t(-B)$, $m_t$ convergent to $\delta _0$ weakly as $t\to 0$. Then $X_t$ has independent increments and stationary increments.

I know it is Markov process becasue of regularity of Dirichlet form. However, I cannot prove independent and stationary increments. How can I prove it?


Denote by $\mathcal{F}_t := \sigma(X_s; s \leq t)$ the canonical filtration and set $$\mathbb{P}^x(X_t \in B) := P_t(x,B).$$ You have already shown that $(X_t)_{t \geq 0}$ is a Markov process, and therefore we have $$\mathbb{E}^x (f(X_t) \mid \mathcal{F}_s) = \mathbb{E}^{X_s} (f(X_{t-s})) \tag{1}$$ for any $s \leq t$ and any bounded measurable function $f$. This gives

\begin{align*} \mathbb{E}^x(e^{i \xi (X_t-X_s)} \mid \mathcal{F}_s) &= e^{-i \xi X_s} \mathbb{E}^x(e^{i \xi X_t} \mid \mathcal{F}_s) \\ &\stackrel{(1)}{=} e^{-i \xi X_s} \mathbb{E}^{X_s} e^{i \xi X_{t-s}} \\ &= h(X_{t-s}) \tag{2}\end{align*} for any $\xi \in \mathbb{R}^d$ where $$h(y) := \mathbb{E}^{y}(e^{i \xi (X_{t-s}-y)}).$$ Since \begin{align*} \mathbb{P}^y(X_{r} \in B) = P_r(y,B) = m_r(B-y) &= P_r(0,B-y) \\ &= \mathbb{P}^0(X_r \in B-y) \\ &= \mathbb{P}^0(y+X_r \in B) \end{align*} for any $r \geq 0$ and $y \in \mathbb{R}^d$, it follows that $$\mathbb{E}^y f(X_r) = \mathbb{E}^0 f(y+X_r)$$ for any bounded measurable function $f$. Consequently, we get $$h(y) = \mathbb{E}^0(e^{i \xi X_{t-s}}). \tag{3}$$ Plugging this into $(2)$ shows that $$\mathbb{E}^x (e^{i \xi (X_t-X_s)} \mid \mathcal{F}_s) = \mathbb{E}^0 e^{i \xi X_{t-s}}. \tag{4}$$ This identity allows us to conclude that $(X_t)_{t \geq 0}$ has stationary and independent increments:

  • stationarity of increments: Taking the expectation of both sides in $(4)$ yields $$\mathbb{E}^xe^{i \xi (X_t-X_s)} = \mathbb{E}^0 e^{i \xi X_{t-s}}. \tag{5}$$ For $s=0$ this gives $$\mathbb{E}^x e^{i \xi (X_t-x)} = \mathbb{E}^0 e^{i \xi X_t} \tag{6}$$ for any $t \geq 0$. Hence, $$\mathbb{E}^x e^{i \xi (X_t-X_s)} \stackrel{(5)}{=} \mathbb{E}^0 e^{i \xi X_{t-s}} \stackrel{(6)}{=} \mathbb{E}^x e^{i \xi (X_{t-s}-x)}, \qquad s \leq t \tag{7}$$ which shows that $X_t-X_s \sim X_{t-s}-X_0$, i.e. $(X_t)_{t \geq 0}$ has stationary increments.

  • independence of increments: Fix $s \leq t$. By the tower property, we have $$\mathbb{E}^x e^{i \eta X_s + i \xi (X_t-X_s)} = \mathbb{E}^x \bigg[ \mathbb{E}^x \big( e^{i \eta X_s + i \xi (X_t-X_s)} \mid \mathcal{F}_s \big) \bigg]$$ and so \begin{align*} \mathbb{E}^x e^{i \eta X_s + i \xi (X_t-X_s)} &= \mathbb{E}^x \bigg[ e^{i \eta X_s} \mathbb{E}^x \big( e^{i \xi (X_t-X_s)} \mid \mathcal{F}_s \big) \bigg] \\ &\stackrel{(4)}{=} \mathbb{E}^x( e^{i \eta X_s}) \mathbb{E}^0 (e^{i \xi X_{t-s}}) \\ &\stackrel{(7)}{=} \mathbb{E}^x( e^{i \eta X_s}) \mathbb{E}^x (e^{i \xi (X_t-X_s)})\end{align*} and therefore $X_t-X_s$ and $X_s$ are independent, see e.g. here for details.

Remark: The idea of the proof is taken from the monograph Lévy Matters III by Böttcher, Schilling & Wang (Theorem 2.6).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.