Laplace equation with additional nonlinear term

206 Views Asked by At

In my research I encountered the following nonlinear PDE$$\nabla^2 f + \frac{1}{2} (\nabla f)^2 = 0,$$which is just the Laplace equation with additional term.

Does this equation belong to a class of well-known nonlinear PDEs? Is there a simple way to solve this equation?

1

There are 1 best solutions below

0
On

I have found a solution to the one dimensional version which may help you. The context of your research would also be helpful.

The one dimensional version is:

$$\frac{d^2 f}{d x^2} + \frac{1}{2}\bigg(\frac{df}{dx}\bigg)^2 = 0$$

A reduction of order can be accomplished by definining $u = \frac{df}{dx}$. The equation then becomes first order nonlinear.

$$\frac{du}{dx} + \frac{1}{2}u^2 = 0$$

$$\frac{2du}{u^2} = -dx$$

$$\frac{2}{u} = x + C$$

$C$ is just an arbitrary integration constant.

Thus

$$u = \frac{1}{\frac{1}{2}(x+C)}$$

From before:

$$\frac{df}{dx} = \frac{1}{\frac{1}{2}(x+C)} $$

$$f = 2 \ln(\frac{1}{2}|x+C|) + D$$

Where $D$ is the second arbitrary constant. You can plug this into the 1D equation and check that it works.

Now if we look at the full equation:

$$\frac{d^2 f}{d x^2} + \frac{d^2 f}{d y^2}+\frac{d^2 f}{d z^2} + \frac{1}{2}\bigg(\bigg(\frac{df}{dx}\bigg)^2 +\bigg(\frac{df}{dy}\bigg)^2 + \bigg(\frac{df}{dz}\bigg)^2 \bigg)= 0$$

Separation of variables does not work in this form, but if we approximate the squares of the derivatives by just the derivatives themselves ($q^2 \approx q$) then it is.

$$\frac{d^2 f}{d x^2} + \frac{d^2 f}{d y^2}+\frac{d^2 f}{d z^2} + \frac{1}{2}\bigg(\bigg(\frac{df}{dx}\bigg) +\bigg(\frac{df}{dy}\bigg) + \bigg(\frac{df}{dz}\bigg) \bigg)\approx 0$$

Then using the anzatz: $f(x,y,z) = X(x)Y(y)Z(z)$ we get three decoupled equations whose sum must be 0.

$$\frac{d^2 f}{d x_i^2} + \frac{1}{2}\bigg(\frac{df}{dx_i}\bigg) = \alpha_i$$

with

$$\sum_i \alpha_i = 0$$

$i$ just runs through $x$, $y$, and $z$.

I don't feel like writing the work right now, plus you may not really need this solution, but using the same reduction of order method as above ($u_x = \frac{dX}{dx}$), one can find that the solution for $X(x)$ is

$$X(x) = -2C_x e^{-x} + 2\alpha_x x + D_x$$

Where $C_x$ and $D_x$ are the arbitrary integration constants.

Thus the full solution would be:

$$f \approx \prod_i (-2C_ie^{-x_i} + 2\alpha_i x_i + D_i)$$

Which is just expressing $f = XYZ$. This could be plotted using some kind of 3D colormap if you wanted to. I don't know how to quickly at the moment. A simple thing to plot (just by using WolframAlpha) is $-e^{-x} + x$, which looks similar to the solution of the 1D problem.

I hope this helps you find your solution.