$$ (x-y)\frac{\partial u}{\partial x}+(x+y)\frac{\partial u}{\partial y}=\alpha u $$ where α is a constant, with initial condition $u(x, 0) = x^2$ , $x > 0$. How do I solve this partial differential equation? I have tried to use the method of characteristics, nd have parameterized $x$ and $y$ but I can't find a way to express $y$ in terms of $x$! Help please!

  • $\begingroup$ This PDE is linear! $\endgroup$ – Cesareo May 12 at 18:07

As @Cesareo already noted, the PDE is indeed linear. And you don't need necessarily to express $y$ as a function of $x$ to solve your PDE.

First of all, observe that your PDE is defined in the whole plane $\mathbb{R}^2$, that the coefficients $x-y$ and $x+y$ of $u_x$ and $u_y$ are smooth and vanish simultaneously in $(0,0)$ (i.e., $(0,0)$ is a singular point for your equation). Therefore your PDE has smooth solutions in $\mathbb{R}^2 \setminus \{(0,0)\}$.

In order to find the solution $u$, you need to put your PDE in implicit form:

$$\underbrace{(x-y) u_x + (x+y)u_y - \alpha u}_{=:F(x,y,u,u_x,u_y)} = 0\; ,$$

or $F(x,y,u,p,q)=0$, in which the socalled Monge notation $p=u_x,\ q=u_y$ is used.

Since $F$ is linear, the characteristic system you need to solve reduces to the following linear system of ODEs:

$$\begin{cases} x^\prime (s) = F_p(x(s),y(s),u(s),p(s),q(s)) \\ y^\prime (s) = F_q(x(s),y(s),u(s),p(s),q(s)) \\ u^\prime (s) = p(s)\ F_p(x(s),y(s),u(s),p(s),q(s)) +q(s)\ F_q(x(s),y(s),u(s),p(s),q(s)) \end{cases}$$

which for $F(x,y,u,p,q)=(x-y)p + (x+y)q - \alpha u$ rewrites:

$$\tag{CS} \begin{cases} x^\prime (s) = x(s) - y(s) \\ y^\prime (s) = x(s) + y(s) \\ u^\prime (s) = \alpha u(s) \end{cases}$$

(in the last equation I used the PDE $(x-y)p + (x+y)q=\alpha u$); moreover, you can couple to (CS) initial conditions:

$$\begin{cases} x(0) = x_0\\ y(0) = y_0 \\ u(0) = u_0\end{cases}\; ,$$

with $(x_0,y_0) \in \Gamma$ ($\Gamma \subset \mathbb{R}^2 \setminus \{(0,0)\}$ being a suitable smooth boundary curve) and $u_0=u(x_0,y_0) \in \mathbb{R}$, in order to obtain a Cauchy problem.

It's easily seen that $u(s) = u_0 e^{\alpha s}$ solves the subproblem:

$$\tag{CS-u}\begin{cases} u^\prime (s) = \alpha u(s) \\ u(0) = u_0\end{cases}\; ,$$

hence you only have to solve:

$$\tag{CS-(x,y)}\begin{cases} x^\prime (s) = x(s) - y(s) \\ y^\prime (s) = x(s) + y(s) \\ x(0) = x_0 \\ y(0) = y_0 \end{cases}\; .$$

Multiplying the first equation by $x$, the second by $y$ and subtracting you got:

$$\begin{cases}\frac{\text{d}}{\text{d} s} [x^2 (s) + y^2 (s)] = 2 (x^2(s) + y^2(s)) \\ x^2(0) + y^2(0) = x_0^2 + y_0^2 \end{cases}$$


$$x^2(s) + y^2(s) = (x_0^2 + y_0^2) e^{2s}\; ,$$

which simplifies into:

$$\tag{1} x^2(s) + y^2 (s) = e^{2s}$$

if you choose the unitary circle $x^2 + y^2 = 1$ as boundary curve $\Gamma$.

From (1) you get:

$$s= \frac{1}{2}\ \log (x^2 + y^2)$$

which you can plug into $u(s) = u_0 e^{\alpha s}$ in order to obtain the solution $u(x,y)$:

$$u(x,y) = u_0\ (x^2 + y^2)^{\alpha/2}\; .$$


$$ (x-y)\frac{\partial u}{\partial x}+(x+y)\frac{\partial u}{\partial y}=\alpha u $$ Note that it should be easier to solve it in polar coordinates since the PDE would be $\rho\frac{\partial u}{\partial \rho}+\frac{\partial u}{\partial \theta}=\alpha u$. Nevertheless the solving below is carried out in Cartesian coordinates.

The Charpit-Lagrange system of equations is : $$\frac{dx}{x-y}=\frac{dy}{x+y}=\frac{du}{\alpha u}$$ A first characteristic equation comes from $\frac{dx}{x-y}=\frac{dy}{x+y}$ which solving leads to : $$\ln(x^2+y^2)-2 \tan^{-1}(\frac{y}{x})=c_1$$ A second characteristic equation comes from

$\frac{dx}{x-y}=\frac{dy}{x+y}=\frac{xdx+ydy}{x(x-y)+y(x+y)}=\frac{du}{\alpha u}\quad\implies\quad\frac{d(x^2+y^2)}{x^2+y^2}=2\frac{du}{\alpha u}$ $$u(x^2+y^2)^{-\alpha/2}=c_2$$ The solution of the PDE expressed on the form of implicit equation $c_2=F(c_1)$ is : $\quad u(x^2+y^2)^{-\alpha/2}=F\left(\ln(x^2+y^2)-2 \tan^{-1}(\frac{y}{x})\right)$ .

$F$ is an arbitrary function, to be determined according to some boundary condition (not specified in the wording of the question). $$\boxed{u(x,y)=(x^2+y^2)^{\alpha/2}F\left(\ln(x^2+y^2)-2 \tan^{-1}(\frac{y}{x})\right)}$$

Note that I agree with the particular solution proposed by Pacciu : $u(x,y)=u_0(x^2+y^2)^{\alpha/2}$ belongs to the above general family of solutions. In this particular case $F(X)=u_0$ . But they are an infinity of other solutions. For example, with $F(X)=e^{-\alpha X/2}$ the solution $u(x,y)=\exp\left(\alpha \tan^{-1}(\frac{y}{x}) \right)$ is a solution which satisfies the PDE as well.

Other example : Suppose that the OP specifies the condition $u(x,0)=x^2$ obviously $u(x,y)=u_0(x^2+y^2)^{\alpha/2}$ cannot be a solution of the problem, what ever $u_0$ is. In this particular case the solution is : $u(x,y)=(x^2+y^2)\exp\left(\alpha \tan^{-1}(\frac{y}{x}) \right)$ .

IN ADDITION, after that the condition $u(x,0)=x^2$ was specified by MathematicianP.

$u(x,0)=x^2=(x^2+0^2)^{\alpha/2}F\left(\ln(x^2+0^2)-2 \tan^{-1}(\frac{0}{x})\right) = x^{\alpha}F\left(2\ln(x)\right)$

Let : $\quad X=2\ln(x)\quad;\quad x=e^{X/2}$ .

$x^2=x^{\alpha}F\left(2\ln(x)\right)=e^X=e^{\alpha X/2}F(X) \quad\implies\quad F(X)=\exp\left((1-\frac{\alpha}{2})X\right)$

So, the function $F(X)$ is determined. We put it into the above general solution where $X=\ln(x^2+y^2)-2 \tan^{-1}(\frac{y}{x})$ .

$u(x,y)=(x^2+y^2)^{\alpha/2}\exp\left(\left(1-\frac{\alpha}{2}\right)\left(\ln(x^2+y^2)-2 \tan^{-1}(\frac{y}{x})\right)\right)$

After simplification : $$u(x,y)=(x^2+y^2)\exp\left(\alpha \tan^{-1}(\frac{y}{x}) \right)$$

  • $\begingroup$ I haven't heard about this Charpit Euqations in my class, I have edited the question to include the initial conditions, is there any other way to solve this equation without the charpit equations? $\endgroup$ – MathematicianP May 14 at 16:52
  • $\begingroup$ with condition $u(x,0)=x^2$ the solution is : $$u(x,y)=(x^2+y^2)\exp\left(\alpha \tan^{-1}(\frac{y}{x}) \right)$$ $\endgroup$ – JJacquelin May 14 at 17:18
  • $\begingroup$ The system of equations $$\frac{dx}{x-y}=\frac{dy}{x+y}=\frac{du}{\alpha u}=ds$$ is an other manner to write $$\begin{cases}\frac{dx}{ds}=x-y \\ \frac{dy}{ds}=x+y\\ \frac{du}{ds}=\alpha u \end{cases}$$ $\endgroup$ – JJacquelin May 14 at 17:26

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.