Definition of semimartingale in Protter's book

520 Views Asked by At

In Protter's book, the definition of semimartingale is conducted by a functional approach:

Definition:(total semimartingale) A process $X$ is a total semimartingale if $X$ is cadlag, adapted and $I_x:S_u \rightarrow L^0$ is continuous, where $S_u$ is the space of the simple predictable process topologized with uniform convergence and $L^0$ is the finite valued random variable topologized with the convergence in probability.

Definition:(semimartingale) A process $X$ is called a semimartingale if, for each $t\in[0,\infty)$, $X^t$ is a total semimartingale.

Following this definition, I want to understand the following corollary 2 on p.55.

Corollary 2: A local martingale with continuous paths is a semimartingale.

The proof in the book claim that this corollary 2 is a immediate consequence of the Corollary 1, which is stated as follows

Corollary 1: Each cadlag, locally square integrable local martingale is a semimartingale.

However, I don't see the connection between a local martingale with continuous paths and locally square integrability.

Further, why does both corollary only induce the process $X$ to be a semi-martingale and not a total semi-martingale? What could go wrong with the total semi-martingale?

1

There are 1 best solutions below

5
On BEST ANSWER

Re Question 1: Let $(X_t)_{t \geq 0}$ be a stochastic process with continuous sample paths. If we define stopping times by

$$\tau_k := \inf\{t \geq 0; |X_t| \geq k\}, \qquad k \in \mathbb{N}$$

then the continuity of the sample paths of $(X_t)_{t \geq 0}$ implies that the following properties hold:

  • $|X_{t \wedge \tau_k}| \leq k$, in particular $(X_{t \wedge \tau_k})_{t \geq 0}$ is square integrable
  • $\tau_k \uparrow \infty$ as $k \to \infty$

It follows from the very definition that $(X_t)_{t \geq 0}$ is locally square integrable.

Re Question 2: Let $(X_t)_{t \geq 0}$ be a Brownian motion and consider

$$H_n(t) := \frac{1}{\sqrt{n}} 1_{[0,n]}(t).$$

Then $H_n \to H := 0$ uniformly but

$$I_X(H_n) = \frac{X_n}{\sqrt{n}}$$

does not converge to $I_X(H)=0$ as $n \to \infty$. This shows that $I_X: S_u \to L^0$ is not continuous.