# Find the common solution of the PDE $u_{xy}+a(x,y)u_x=0$

Find the common solution of $$u_{xy}+a(x,y)u_x=0~~~~~\text{in}~~~~~\Omega:=\left\{(x,y)\in\mathbb{R}^2 : \lvert x-x_0\rvert<d_1, \lvert y-y_0\rvert <d_2\right\},\\ d_1,d_2>0, (x_0,y_0)\in\mathbb{R}^2, a\in C(\Omega)$$

In the meantime, I thought about an answer to this question (see below). Would be great to get a feedback to it.

With kind regards

I add my answer in order to get a feedback from you. Please tell me, if my general solution is right or maybe nonsense, would be very kind of you!

First of all I use the theorem of Schwarz and substitute $z:=u_x$, getting an ODE of order 1 in $y$ (here $x$ is only a parameter): $$z_y=-a(x,y)z~~~~~(1)$$ If one considers $a$ as only dependent on $y$ and $x$ only as a parameter, then $a$ is continious on the intervall $I:=(y_0-d_2,y_0+d_2)$. So the Fundamental Theorem of Calculus says that $$A(x):=\int_{y_0}^{y}a(x,\tau)\ d\tau~~~~~(2)$$ is an antiderivative on $I$. So I continue (1) by separation of variables, getting $$z=C_1(x)\cdot\exp\left(-\int_{y_0}^{y}a(x,\tau)\ d\tau\right).$$ Resubstituting, it follows $$u(x,y)=\int_{x_0}^{x}C_1(\eta)\exp\left(-\int_{y_0}^y a(\eta,\tau)\, d\tau\right)\, d\eta+C_2(y)~~~~~(3)$$ with $C_1\in C^1((x_0-d_1,x_0+d_1))$ and $C_2\in C^1(I)$ arbitrary functions.

Comment: Instead of chosing $y_0$ and $x_0$ as lower borders in (2) and (3)it is of course possible to chose any point in $I$ respectively any point in $(x_0-d_1,x_0+d_1)$.


\begin{align} \partiald{{\rm u}\pars{x,y}}{x} &= \exp\pars{-\int_{y_{0}}^{y}{\rm a}\pars{x,y'}\,\dd y'} + \partiald{{\rm u}\pars{x,y_{0}}}{x} \\ \partiald{{\rm u}\pars{x,y_{0}}}{x} &\equiv \left.\partiald{{\rm u}\pars{x,y'}}{x}\right\vert_{y' = y_{0}} \equiv \tilde{\rm u}\pars{x,y_{0}} \end{align}

$${\rm u}\pars{x,y} = \int_{x_{0}}^{x} \exp\pars{-\int_{y_{0}}^{y}{\rm a}\pars{x',y'}\,\dd y'}\,\dd x' + \int_{x_{0}}^{x}\tilde{\rm u}\pars{x',y_{0}}\,\dd x' + {\rm u}\pars{x_{0},y} \tag{1}$$

In addition, $${\rm u}_{x}\pars{x,y}{\rm u}_{xy}\pars{x,y} + {\rm a}\pars{x,y}{\rm u}_{x}^{2}\pars{x,y}=0\,, \quad\imp\quad \bracks{{1 \over 2}\,\partiald{}{y} + {\rm a}\pars{x,y}}{\rm u}_{x}^{2}\pars{x,y} = 0$$ $$\partiald{}{y}\bracks{% \exp\pars{2\int_{y_{0}}^{y}{\rm a}\pars{x,y'}\,\dd y'} {\rm u}_{x}^{2}\pars{x,y}} = 0$$ and $${\rm u}_{x}^{2}\pars{x,y} = \tilde{\rm u}^{2}\pars{x,y_{0}}\exp\pars{-2\int_{y_{0}}^{y}{\rm a}\pars{x,y'}\,\dd y'}$$

Let's assume we have two solutions $\phi\pars{x,y}$ and $\varphi\pars{x,y}$ of the differential equation for given values of ${\rm u}\pars{x_{0},y}$ and $\tilde{\rm u}\pars{x,y_{0}}$. Then, we have $\delta_{x}^{2}\pars{x,y} = 0$ where $\delta\pars{x,y} \equiv \phi\pars{x,y} - \varphi\pars{x,y}$. We conclude that $$\delta\pars{x,y} = {\rm f}\pars{y}\ \mbox{where}\ {\rm f}\ \mbox{is an }\ {\it arbitrary}\ \mbox{function}.$$ However, $${\rm f}\pars{y} = \phi\pars{x_{0},y} - \varphi\pars{x_{0},y} = {\rm u}\pars{x_{0},y} - {\rm u}\pars{x_{0},y} = 0 \quad\imp\quad \phi\pars{x,y} = \varphi\pars{x,y}$$

We conclude that the solution $\pars{1}$ is the unique solution for a given boundary conditions as defined by $\ds{{\rm u}\pars{x_{0},y}}$ and $\ds{\tilde{\rm u}\pars{x_{0},y} \equiv \left.\partiald{{\rm u}\pars{x,y'}}{y'}\right\vert_{y'\ =\ y_{0}}}$.