Differentiating Laplace transform of $dX=cXdt+\sqrt{X}dW$

277 Views Asked by At

Consider a solution $X$ of the stochastic differential equation

$$dX=cXdt+\sqrt{X}dW$$

For $\mathcal{L}(\alpha,t)=E[\exp(-\alpha X_t)]$ I want to show, that

$$\frac{d}{dt}\mathcal{L}(\alpha,t)=(c\alpha-\frac{1}{2}\alpha^2)\frac{d}{d\alpha}\mathcal{L}(\alpha,t)$$

What I did so far is following:

$$\begin{align} (c\alpha-\frac{1}{2}\alpha^2)\frac{d}{d\alpha}\mathcal{L}(\alpha,t) &=(c\alpha-\frac{1}{2}\alpha^2)E[(-X_t)\exp(-\alpha X_t)] =E[(\frac{1}{2}\alpha^2-c\alpha)X_t\exp(-\alpha X_t)]\\ \end{align}$$

By using the function $h(x,y)=\exp(-xy)$ in Ito's lemma, I find

$$\begin{align}\exp(-\alpha X_t) &=h(\alpha,X_t)\\ &=h(\alpha,0)+\int_0^t(-\alpha)\exp(-\alpha X_s)dX_s+\frac{1}{2}\int_0^t\alpha^2\exp(-\alpha X_s)d[X,X]_s\\ &=1+\int_0^t(-\alpha)cX_s\exp(-\alpha X_s)ds+\int_0^t(-\alpha)\sqrt{X_s}\exp(-\alpha X_s)dW_s\\ &\quad\quad+\frac{1}{2}\int_0^t\alpha^2X_s\exp(-\alpha X_s)ds\\ &=1+\int_0^t(\frac{1}{2}\alpha^2-c\alpha)X_s\exp(-\alpha X_s)ds+\int_0^t(-\alpha)\sqrt{X_s}\exp(-\alpha X_s)dW_s \end{align}$$

Now, is it true, that

$$\frac{d}{dt}\mathcal{L}(\alpha,t)=E[\frac{d}{dt}\exp(-\alpha X_t)]=E[(\frac{1}{2}\alpha^2-c\alpha)X_t\exp(-\alpha X_t)]\quad?$$

I find $$\frac{d}{dt}1=0$$

and

$$\frac{d}{dt}\int_0^t(\frac{1}{2}\alpha^2-c\alpha)X_s\exp(-\alpha X_s)ds=(\frac{1}{2}\alpha^2-c\alpha)X_t\exp(-\alpha X_t)$$

but why is

$$\frac{d}{dt}\int_0^t(-\alpha)\sqrt{X_s}\exp(-\alpha X_s)dW_s=0\quad ?$$

Thank you for your attention!

1

There are 1 best solutions below

1
On BEST ANSWER

You have

$$\mathbb{E}\int_0^t(-\alpha)\sqrt{X_s}\exp(-\alpha X_s)dW_s=0$$

because this is a Wiener integral. Therefore the derivative is also zero. Does this answer the question?