1
$\begingroup$

Let

  • $(\Omega,\mathcal A,\operatorname P)$ be a probability space
  • $(\mathcal F_t)_{t\ge 0}$ be a filtration of $\mathcal A$
  • $(U,\langle\;\cdot\;,\;\cdot\;\rangle)$ be a separable Hilbert space
  • $\left(B^{(n)}\right)_{n\in\mathbb N}$ be independent real-valued Brownian motions with respect to $\mathcal F$ on $(\Omega,\mathcal A,\operatorname P)$
  • $\mathfrak L(X)$ be the set of bounded, linear operators on a normed vector space $X$

I've found two different definitions of a cylindrical Brownian motion and don't understand why they are consistent:

Definition 1: Let

  • $Q\in\mathfrak L(U)$ be nonnegative and symmetric
  • $U_0:=Q^{\frac 12}(U)$, $$\langle u,v\rangle_0:=\langle Q^{-\frac 12}u,Q^{-\frac 12}v\rangle\;\;\;\text{for }u,v\in U_0$$ where $Q^{-\frac 12}$ is the pseudo inverse of $Q^{\frac 12}$ and $(e_n)_{n\in\mathbb N}$ be an orthonormal basis of $U_0$
  • $(U_1,\langle\;\cdot\;,\;\cdot\;\rangle_1)$ be a separable Hilbert space and $$\iota:(U_0,\langle\;\cdot\;,\;\cdot\;\rangle_0)\to(U_1,\langle\;\cdot\;,\;\cdot\;\rangle_1)$$ be a Hilbert-Schmidt embedding, e.g. $(U_1,\langle\;\cdot\;,\;\cdot\;\rangle_1):=(U,\langle\;\cdot\;,\;\cdot\;\rangle)$ and $$\iota u:=\sum_{n\in\mathbb N}\alpha_n\langle u,e_n\rangle_0e_n\;\;\;\text{for }u\in U_0\tag 1$$ for some $(\alpha_n)_{n\in\mathbb N}\subseteq(0,\infty)$ with $\sum_{n\in\mathbb N}\alpha_n^2<\infty$

Then $$W_t:=\sum_{n\in\mathbb N}B_t^{(n)}\iota e_n\;\;\;\text{for }n\in\mathbb N\text{ and }t\ge 0\tag 2$$ is called a cylindrical $Q$-Brownian motion with respect to $\mathcal F$ on $(\Omega,\mathcal A,\operatorname P)$.

[Note that (2) is convergent in $L^2(\Omega; U_1)$.]

Definition 2: Let

  • $(\tilde e_n)_{n\in\mathbb N}$ be an orthonormal basis of $U$

Then $$W_t:=\sum_{n\in\mathbb N}B_t^{(n)}\tilde e_n\;\;\;\text{for }n\in\mathbb N\text{ and }t\ge 0\tag 3$$ is called a cylindrical Brownian motion or space-time white noise with respect to $\mathcal F$ on $(\Omega,\mathcal A,\operatorname P)$.

[Note that (3) is convergent in $L^2(\Omega;\tilde U_1)$ if $U_1\supseteq U$ is a separable Hilbert space and there is a Hilbert-Schmidt inclusion $\tilde\iota:U\to\tilde U_1$.]


Definition 2 seems to be a special case of Definition 1, but how do we obtain it?

If we choose $Q:=\operatorname{id}_U$, then $(U_0,\langle\;\cdot\;,\;\cdot\;\rangle_0)=(U,\langle\;\cdot\;,\;\cdot\;\rangle)$). Now, we could choose $(U_1,\langle\;\cdot\;,\;\cdot\;\rangle_1):=(U,\langle\;\cdot\;,\;\cdot\;\rangle)$ and $\iota$ to be $(1)$ (with $\langle\;\cdot\;,\;\cdot\;\rangle_0$ replaced by $\langle\;\cdot\;,\;\cdot\;\rangle$). However, that doesn't yield that $(2)$ and $(3)$ are equal (does it?).


So, how do we obtain a space-time white noise $(3)$ from $(2)$?

$\endgroup$
0
$\begingroup$

Let $(\tilde e_n)_{n\in\mathbb N}$ be an orthonormal basis of $U$ with $$Q\tilde e_n=\lambda_n\tilde e_n\;\;\;\text{for all }n\in\mathbb N$$ for some $(\lambda_n)_{n\in\mathbb N}\subseteq[0,\infty)$ $\Rightarrow$ $$e_n:=Q^{\frac 12}\tilde e_n=\lambda_n^{\frac 12}\tilde e_n\;\;\;\text{for }n\in\mathbb N\tag 4$$ is an orthonormal basis of $U_0$. Let $(U_1,\langle\;\cdot\;,\;\cdot\;\rangle_1)$ be a separable Hilbert space with $U_0\subseteq U_1$ such that the inclusion $$\iota:(U_0,\langle\;\cdot\;,\;\cdot\;\rangle_0)\to(U_1,\langle\;\cdot\;,\;\cdot\;\rangle_1)$$ is Hilbert-Schmidt. Then $$W_t:=\sum_{n\in\mathbb N}B_t^{(n)}\iota e_n\stackrel{(4)}=\sum_{n\in\mathbb N}\lambda_n^{\frac 12}B_t^{(n)}\tilde e_n\;\;\;\text{for }t\ge 0\tag 5$$ is a cylindrical $Q$-Brownian motion with respect to $\mathcal F$ on $(\Omega,\mathcal A,\operatorname P)$ according to Definition 1. Choose $Q:=\operatorname{id}_U$ $\Rightarrow$ $(U_0,\langle\;\cdot\;,\;\cdot\;\rangle_0)=(U,\langle\;\cdot\;,\;\cdot\;\rangle)$ and $(\tilde e_n)_{n\in\mathbb N}$ can be any orthonormal basis of $U$ (since $\lambda_n=1$ for all $n\in\mathbb N$) and we obtain that $$W_t=\sum_{n\in\mathbb N}B_t^{(n)}\tilde e_n\;\;\;\text{for all }t\ge 0\tag 6$$ is a cylindrical Brownian motion or space-time white noise with respect to $\mathcal F$ on $(\Omega,\mathcal A,\operatorname P)$ according to Definition 2.

$\endgroup$

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.