Solve the SDE given by: $dX_t = (e^{-\gamma t} - \gamma X_t) dt + 2 e^{-\gamma t/2} \sqrt{X_t} dW_t$.

My attempt

Following the hint of my professor: suppose $X_t = e^{\gamma t} g(W_t)$. Then we can write: $$dX_t = (e^{-\gamma t}((1/2)g''(W_t))-\gamma X_t) dt + (e^{-\gamma t} g'(W_t)) dW_t$$

By identification,

$$1=\dfrac{g''(W_t)}{2}$$ and, $$g'(W_t) = 2 (g(W_t))^{1/2}$$ with initial condition $$g(0) = 1$$

EDIT If we consider the equations above as "non-stochastic" DE's, we can attempt to find the solution to the system. The solution to the first equation is a simple polynomial, but the solution to the second equation (the non-linear ODE) is very long. How can I proceed?


The first equation tells us that $g$ is a polynomial of (at most) second order, i.e. we can make the ansatz

$$g(x) = a+bx+cx^2.$$

Since $g(0) \stackrel{!}{=} 1$, we have $a=1$. Moreover, as

$$g'(x) = b+2cx,$$

the condition

$$g'(x)^2 = 4 g(x) \tag{1}$$

is equivalent to

$$b^2+4bcx + 4c^2 x^2 = 4 + 4bx + 4cx^2.$$


$$c^2 \stackrel{!}{=} c \qquad b \stackrel{!}{=} bc \qquad b^2 \stackrel{!}{=} 4.$$

This implies $c=1$ and $b = \pm 2$. In $(1)$, we have lost the information on the sign of $b$; unfortunately, neither $$g_1(x) := 1+2x+x^2 = (x+1)^2$$ nor $$g_2(x) = 1-2x+x^2 = (x-1)^2$$ satisfies the condition

$$g'(x) = 2 \sqrt{g(x)}$$

(as this would imply in particular $g'(x) \geq 0$, which is - in both cases- not satisfied for $x <-1$). This means that we don't get any solution using this approach.


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.