I know the solution to $y''=1$ with $y(0)=y(1)=0$ should be $y=(x-.5)^2/2$. I need to use Green's function to obtain the same result.

I begin with the definition of Green's function as the solution to the associated equation

$$G'' = \delta(x_0)$$

which I separate into cases $x<x_0$ and $x>x_0$ which have corresponding functions

$$G_1 = c_1+c_2x$$

$$G_2 = c_3+c_4x$$

Applying the boundary condition $G(0)=0=c_1$ we get $G_1 = c_2x$. Applying the other, $G_2(1)=0=c_3+c_4\implies c_4=-c_3$. Enforcing the continuity condition


and enforcing that $G'$ has a jump of 1 at $x_0$

$$-c_3-c_2=1 \implies c_3=-1-c_2$$

Then putting these last two together

$$c_2x_0 = (-1-c_2)(1-x_0)$$

which eventually implies

$$c_2=x_0-1, \qquad c_3=-x_0$$

Now to use what I've found to get the solution to the differential equation is the step where I'm really uncertain, but I think what I'm supposed to do is evaluate the following.

$$\int_0^x(x_0-1)x\ dx_0 + \int_x^1(-x_0)(1-x)\ dx_0$$

which is

$$x(x^2/2-x) + (1-x)\left(-\frac{1-x^2}{2}\right)$$

which I simplify to


which, although close, is not the right answer. Noteworthy is that its derivative is $-1$ and it does not quite satisfy the boundary conditions.

  • $\begingroup$ The function $y=(x-5)^2/2$ or $y=(x-.5)^2/2$ does not satisfy BCs:$ y(0)=y(1)=0$. $\endgroup$ – daulomb Dec 2 '17 at 22:19
  • $\begingroup$ It must be $y=\frac{x^2}{2}-\frac{x}{2}$ according to BCs. $\endgroup$ – daulomb Dec 2 '17 at 22:26

$G_1(x,s)=c_1x+c_2$ when $x\leq s$ with $G_1(0, s)=0$ and $G_2(x,s)=d_1x+d_2$ when $x>s$ with $G_2(1, s)=0$ and form $G(x, s)=G_1(x,s)$ if $x\leq s$ and $G(x, s)=G_2(x,s)$ if $x>s$. Using boundary conditions for $G_1$ and $G_2$ we have $G_1(x, s)=c_1x$ and $G_2(x,s)=d_1(x-1)$. We determine $c_1$ and $d_1$ by using continuity of $G$ and jump of the derivatives at $s$. You eventually find that $G_1(x, s)=x(s-1)$ if $0\leq x\leq s$; $G_2(x,s)=s(x-1)$ if $s<x\leq 1$ and so $y(x)=\int_0^1G(x,s)ds= \int_0^xG_2(x,s)ds+\int_x^1G_1(x,s)ds= \int_0^xs(x-1)ds+\int_x^1x(s-1)ds=\frac{x^2(x-1)}{2}-\frac{x(x-1)^2}{2}=\frac{x(x-1)}{2}$.

This is the same solution as I said in the beginning of the comments. What a boring simle problem. After many efforts and comments:)

  • $\begingroup$ How do you get your result for $G_1(x,s)$? Isn't this supposed to merely be the solution to $G''=0$ with initial condition $G(0)=0$? $\endgroup$ – Addem Dec 2 '17 at 22:36
  • $\begingroup$ it comes from the construction of Green's functions. $\endgroup$ – daulomb Dec 2 '17 at 22:37
  • $\begingroup$ I may not know the construction of Green's function--I just have the definition that it satisfies $\mathcal{L}[G]=\delta(s)$ and is continuous at $s$ and has a jump in the derivative at $s$ equal to 1. Is this the wrong definition? $\endgroup$ – Addem Dec 2 '17 at 22:38
  • $\begingroup$ You are actually solve $G_{xx}(x, s)=\delta(x-s)$, where $\delta$ is Dirac delta. We use the fact that $\delta(x-s)=0$ when $x\neq s$. $\endgroup$ – daulomb Dec 2 '17 at 22:39
  • $\begingroup$ That's what I thought, so when $x<s$ do we not effectively have $G_{xx}=0$ so that $G = c_1+c_2x$? I don't see how the $x^2/2$ gets in. $\endgroup$ – Addem Dec 2 '17 at 22:40

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.