2

As explained here, the infinitesimal generator of a Brownian motion is $\frac{1}{2}\Delta$. Can someone please provide (for a non-stochastic-student) the proof of finding the infinitesimal generator on a circle (and if possible, also on a sphere) in the Cartesian coordinates? (or introducing a book or article)

Thanks.

Denis
  • 169
  • 11
  • Could you clarify what you mean by "on a circle?" – user6247850 Jun 15 '20 at 15:11
  • I mean the stochastic process with the following SDE: $$ \\left{ \begin{matrix} d{{y}{1}}=-{{y}{2}}d{{B}{t}}-\frac{1}{2}{{y}{1}}dt \ d{{y}{2}}=+{{y}{1}}d{{B}{t}}-\frac{1}{2}{{y}{2}}dt \ \end{matrix} \right.\

    which was discussed here :

    link

    – Denis Jun 15 '20 at 16:17

1 Answers1

3

Starting from $$dY_1=-\frac12 Y_1 dt-Y_2 dB_t, \\ dY_2=-\frac12 Y_2 dt+Y_1 dB_t, $$

and using the general formula for the generator of a diffusion process $dX_t=f(X_t)dt+g(X_t)dB_t$, which reads $$\mathcal{A}=\sum_{i=1}^n f_i(x)\partial_{x_i}+\frac{1}{2}\sum_{i=1}^n\sum_{j=1}^m g_{ij}(x)\partial_{x_i}\partial_{x_j},$$ we have with $n=2$ and $m=1$ $$\mathcal{A}=\frac12\left(-y_1\partial_{y_1}-y_2\partial_{y_2}+y_2^2\partial_{y_1}^2+y_1^2\partial_{y_2}^2\right)=\frac12\partial_{\theta}^ 2.$$ Which is one half times the Laplacian on $S_1$.

In fact, one of the definitions of Brownian motion on a Riemannian manifold $(M,g)$ is that its generator is $\frac12\Delta_g$, where $\Delta_g$ is the Laplacian of $g$. Thus in order to find the generator of Brownian motion on $S_2$ in Cartesian coordinates, take the Laplacian on $S_2$ and transform it to Cartesian coordinates (a slightly tedious calculation).

S.Surace
  • 1,790
  • Thank you very much. Above, we applied the general formula of the generator to the SDE. The general formula is clear, but what is the starting point on $S_2$? On $S_1$ we started with $Y= (cos(B),sin(B))$ and then applied the Ito formula to obtain the SDE. Unlike the circle case where we had only $B$, In case of sphere, don't we have $B_1$ and $B_2$? I mean : $Y=( cos(B_1)sin(B_2), sin(B_1)sin(B_2), cos(B_1))$ ? If so, what is the ito formula(SDE) obtained from $Y$? Can we use the same general formula for generator you wrote above for this SDE? – Denis Jun 18 '20 at 17:15
  • 1
    A Brownian motion on the chart does not lead to a Brownian motion on the manifold in general. You may want to think how you want to define/construct the Brownian motion if starting from the Laplacian is not satisfactory. You may come up with a number of constructions but you have to decide whether your construction is Brownian motion. I would suggest to ask a new question. – S.Surace Jun 18 '20 at 17:39