I am considering the following ordinary integro-differential equation $$ A g = \sigma^2(y) + \int_\mathbb{R} \nu(y,dz) z^2 $$ where $$ A = b(y) \partial + \frac{1}{2} \sigma^2(y) \partial^2 + \int_\mathbb{R} \nu(y,dz) (e^{z \partial} - 1 - z \partial) $$ $\nu(y,\cdot)$ is a Levy measure for every $y$ and $$ b(y) = - \frac{1}{2} \sigma^2(y) -\int_\mathbb{R} \nu(y,dz) (e^{z} - 1 - z) $$ I use the notation $e^{z \partial}$ to indicate the shift operator $e^{z \partial} f(y) = f(y+z)$. Those with a background in probability will recognize $A$ as the generator of a Levy-type process.

I would like to find conditions on $\sigma$ and $\nu$ such that the classical solution $g$ of the top equation is exponentially bounded. That is, there exists $C,b>0$ such that $$ g(y) \leq C \exp ( b |y| ) . $$ If $\sigma(y)=a$ and $\nu(y,dz)=\mu(dz)$ then it is easy to check that the solution is $g(y) = Q y$ where $Q$ is a constant $$ Q = \frac{a^2 + \int \mu(dz) z^2 }{a^2/2 + \int \mu(dz) (e^z - 1 - z)} . $$ So, I imagine that bounded coefficients would yield a solution that satisfies the exponential growth bound. But, I do not know how to show this. By "bounded" levy measure I mean $\sup_y \nu(y,B) \leq \overline{\nu}(B)$. I am happy to consider stronger conditions if needed.

I was hoping there might be a general result concerning pseudo-differential operators that might show this (the generator $A$ can be thought of as a $\Psi$DO the generates a Markov process). For example, maybe if the symbol of $A$ satisfies certain conditions then I could say something about the solution $g$. But, I am not an expert in this area, so any help would be greatly appreciated.


Your Answer

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

Browse other questions tagged or ask your own question.