Solving a Nonlinear SDE with an Integrating Factor
Summary
The document derives a strong solution for a stochastic differential equation whose drift is inversely proportional to the state and whose diffusion is proportional to the state. One approach applies Itô’s formula to the squared process, turning it into a linear SDE; an integrating factor then yields an explicit expression involving Brownian motion and a time integral. A second approach constructs an exponential process, uses the product rule to cancel stochastic terms, and solves the resulting ordinary differential equation for the transformed state.
Both derivations arrive at the same expression for the square of the process in terms of its initial value, Brownian motion, and an integral over the path. The original motivation is to evaluate a conditional expectation of a positive-part payoff, but the responses stop after obtaining the process solution and do not calculate that expectation. No numerical evaluation or distributional simplification is provided, so further analysis would be needed for a pricing result.
Key ideas
- Squaring the process and applying Itô’s formula produces a linear SDE.
- An integrating factor removes the stochastic term and gives an explicit pathwise expression.
- A second derivation uses an exponential transformation and the product rule to reach the same solution.
- The requested conditional payoff expectation is not evaluated in the document.
Tags
Full text
# What is the strong solution for this SDE
# What is the strong solution for this SDE
I want to calculate $E_t[(X_T-K)^+]$ where $$dX_t=\frac{3}{X_t}dt+2X_t dW_t$$ and $X_0=x$. I don't know how extact the strong solution of this SDE. Indeed I used Ito's lemma but it was not usefule.
Thanks for attantion.
## Answer by Gordon (score 2, accepted)
https://quant.stackexchange.com/a/30837
Note that \begin{align*} d\left(X_t^2\right) &= 2X_t dX_t + d\langle X, X\rangle_t\\ &=(6+4X_t^2)dt + 4X_t^2dW_t, \end{align*} which can be solved using the technique with an integral factor. Specifically, note that \begin{align*} d\left(e^{4t-4W_t}X_t^2 \right) &= X_t^2 d\left(e^{4t-4W_t}\right) + e^{4t-4W_t} d(X_t^2) + \left\langle d\left(e^{4t-4W_t}\right), d(X_t^2) \right\rangle\\ &=e^{4t-4W_t}X_t^2(12dt-4dW_t)\\ &\quad +e^{4t-4W_t}\left[(6+4X_t^2)dt + 4X_t^2dW_t\right]- 16 e^{4t-4W_t}X_t^2 dt\\ &=6e^{4t-4W_t} dt. \end{align*} Then \begin{align*} e^{4t-4W_t}X_t^2 &= x^2+6\int_0^t e^{4s-4W_s} ds. \end{align*} That is, \begin{align*} X_t^2 = e^{4W_t-4t}\left(x^2+6\int_0^t e^{4s-4W_s} ds \right). \end{align*}
> A general note. For an equation of the form \begin{align*} dY_t = (aY_t+b)dt + (cY_t+d) dW_t, \end{align*} we can apply an integral factor of the form $$e^{(-a+\frac{1}{2}c^2)t -cW_t}.$$
## Answer by user16651 (score 3)
https://quant.stackexchange.com/a/30832
You shouldn't say: "Ito's lemma wasn't useful". Set $$Y_t=-2W_t+2t\tag 1$$ Note $W_0=0$ thus $Y_0=0$.We have $$dY_t=2\,dt-2\,dW_t\tag 2$$ Set $Z_t=e^{Y_t}$. By application of Ito's lemma, we have $$dZ_t=e^{Y_t}\,dY_t+\frac{1}{2}e^{Y_t}d[Y_t,Y_t]\tag 3$$ therefore $$dZ_t=4e^{Y_t}dt-2\,e^{Y_t}dW_t=4Z_tdt-2Z_tdW_t\tag 4$$ on the other hand $$d(X_t\,Z_t)=Z_t\,dX_t+X_t\,dZ_t+d[X_t,Z_t]\tag 5$$ thus $$d(X_t\,Z_t)=\frac{3Z_t}{X_t}dt\tag 6$$ in other words $$d(X_t\,Z_t)=\frac{3Z_t^2}{X_tZ_t}dt\tag 7$$ set $R_t=X_tZ_t$. We have $$dR_t=\frac{3Z_t^2}{R_t}dt\tag 8$$ This can be solved as a regular ODE with separable variables : $$R_t\,dR_t=3Z_t^2\, dt$$ and $$\frac{1}{2}R_t^2-\frac{1}{2}R_0^2=3\int_0^{t}Z_s^2ds\\ \frac{1}{2}R_t^2-\frac{1}{2}x^2=3\int_0^{t}Z_s^2ds\tag 9$$ then $$X_t^2=e^{-2Y_t}\left(x^2+6\int_0^{t}e^{2Y_s}ds\right)\tag {10}$$ Finally
> $$\color{red}{X_t^2=e^{4W_t-4t}\left(x^2+6\int_0^{t}e^{-4W_s+4s}ds\right)\tag{11}}$$
Now How can you calculate $\mathbb{E}_t[(X_T-K)^+]?$Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.