Representation of the analytical solution to the Poisson equation?

66 Views Asked by At

I have the following 1-dimensional Poisson's problem with the corresponding boundary conditions:

$$ u_{xx} = -f(x) ; \quad u(1) = g, \quad -u_x(0) = h $$ on a open domain $$ \Omega = (0,1) $$

Is it correct to represent an analytical solution as follows?

First, integrating with respect to $x$ once from $0$ to $y$:

$$ \int_0^y u_{xx} dx = -\int_0^y f(x)dx $$

to obtain: $$ u_x(y) - u_x(0) = -\int_0^y f(x)dx $$ Applying the BC, $$ u_x(y) = -\int_0^y f(x)dx + h $$ Let $$ f(z) = \int_0^y f(x)dx $$

and then integrating the previous equation with respect to $y$ from $x$ to $1$: $$ u(1) - u(x) = \int_x^1 f(z)dy + hy|_x^1 $$ Applying the boundary condition $u(1)=g$ and then putting $f(z)$ back to its original form: $$ u(x) = g + h(1-x) + \int_x^1 \left[ \int_0^y f(x)dx \right]dy $$ I'm doing this since I require the representation of the exact solution to compare with the solution obtained from the Galerkin approximation. Is there anything wrong with the steps I've shown above? Please criticize.

1

There are 1 best solutions below

2
On BEST ANSWER

You can just check it.

$$u(x)=g+h\cdot (1-x)+\int_{x}^1\int_0^y f(t)\mathrm dt \mathrm dy \\ u'(x)=-h+\frac{\mathrm d}{\mathrm dx}\left(\int_{x}^1\int_0^y f(t)\mathrm dt \mathrm dy\right) \\ =-h-\int_0^x f(t)\mathrm dt$$ $$u''(x)=-\frac{\mathrm d}{\mathrm dx}\int_0^x f(t)\mathrm dt=-f(x)$$

And, $$u(1)=g+\underbrace{h\cdot(1-1)}_{=0}+\underbrace{\int_1^1\cdots}_{=0}=g$$ $$u'(0)=-h-\underbrace{\int_{0}^0\cdots}_{=0}$$

Everything looks good.


In regards to the second line.

Let $F(y)=\int_0^y f(t)\mathrm dt$. Then, $$u'(x)=-h+\frac{\mathrm d}{\mathrm dx}\int_x^1 F(y)\mathrm dy \\ =-h-\frac{\mathrm d}{\mathrm dx}\int_1^x F(y)\mathrm dy$$ By the FTC, this is $$u'(x)=-h-F(x) \\ =-h-\int_0^x f(t)\mathrm dt$$ The rest follows.