1

I'm stuck trying to solve $uu_x+u_y=-\frac{1}{2}u$ satisfying $u(x,2x)=x^2$, this solution also needs to be in some parametrised form with $x(r,s), y(r,s),u(r,s)$.

So far I have this:

$\dfrac{dx}{ds}=u \longrightarrow x(r,s)=us+x_0$

$\dfrac{dy}{ds}=1 \longrightarrow y(r,s)=s+y_0$

$\dfrac{du}{ds}=-\dfrac{1}{2}u \longrightarrow u(r,s)=u_0e^{-s/2}$

If we have that $y(0)=0$ we can see that $y=s$, so we can replace each instance of $s$ with $y$. I know I will end up with an implicit function, but I don't know how to apply the $u(x,2x)=x^2$ condition from here. I was thinking about directly plugging in $u(x,2x)=u_0e^{-x}=x^2\Rightarrow u_0=x^2e^x$ and then after letting $x(0)=r$ (the initial curve) and some substitutions end up with $u=(2uy-x)^2e^{-x}e^{-y/2}$ but I don't think that's correct.

I feel like there's a step missing before I start doing that. I have only just been introduced to PDEs, so all of this is new, any help is welcome and I appreciate your patience with me.

As an aside, I know there is another question that is asking the exact same thing that I am, but they introduce some $\Phi$ and $F$ notations which I have never seen before and leaves me more confused than I initially was.

  • 1
    ${{dx}\over {ds}}=u$ does not mean $x=us+x_0$. Remember that $u$ is a function of $s$; You are treating $u$ in this DE as if it were a constant. You should parameterize the IC to ensure $$\big(x(0,t),y(0,t),z(0,t)\big)=\big(t,2t,t^2\big)$$ – Matthew H. Sep 24 '21 at 09:22
  • Of course, that makes much more sense. So in my equations should I have $x_0=t,y_0=2t,z_0=t^2$ instead of my $x_0=s$ and $y_0=0$? – Maksymilian5275 Sep 24 '21 at 09:49

2 Answers2

1

$$2uu_x+2u_y=-u$$ Charpit-Lagrange characteristic ODEs : $$\frac{dx}{2u}=\frac{dy}{2}=\frac{du}{-u}$$ First characteristic from $\frac{dx}{2u}=\frac{du}{-u}$ $$2u+x=c_1$$ Second characteristic from $\frac{dy}{2}=\frac{du}{-u}$ $$u\:e^{y/2}=c_2$$ General solution of the PDE on implicit form $c_2=F(c_1)$ : $$\boxed{u\:e^{y/2}=F\left(2u+x\right)}$$ $F$ is an arbitrary function (to be determined according to some boundary condition).

Condition : $u(x,2x)=x^2$ $$x^2\:e^{2x/2}=F\left(2x^2+x\right)$$ $X=2x^2+x\quad\implies\quad x=\frac14\left(-1+\sqrt{1+8X} \right)$

The sign minus of the square root is excluded (in puting $X=2x^2+x$ into the square root). $$F(X)=\frac{1}{16}\left(-1+\sqrt{1+8X} \right)^2 e^{-\frac14\left(-1+\sqrt{1+8X} \right)} $$

We put $F(X)$ into the above general solution where $X=2u+x$ :

$$\boxed{u\:e^{y/2}=\frac{1}{16}\left(-1+\sqrt{1+8(2u+x)} \right)^2 e^{-\frac14\left(-1+\sqrt{1+8(2u+x)} \right)}}$$ This is the solution on the form of implicit equation which satisfies both the PDE and the condition.

JJacquelin
  • 66,221
  • 3
  • 37
  • 87
  • The initial condition is that $u(x,2x)=x^2$, this is one of the problems I was running into, the fact it's a squared term of $x$ and not a linear one made things slightly more difficult – Maksymilian5275 Sep 26 '21 at 13:48
  • Also, there seems to be a sign error in the first equation on the $u_y$ term, but I can manage that. I want to ask about $c_1$ and $c_2$, I usually see $u$ represented as a function of $r$ and $s$, where $r$ is the inital curve in the $x$-$u$ plane and $s$ tells us how it progresses into the $y$ direction. Are $c_1$ and $c_2$ directly related to $r$ and $s$ in some way? – Maksymilian5275 Sep 26 '21 at 14:30
  • The typo in the PDE is corrected. The boundary condition is updated. If fact the relationship $c_2=F(c_1)$ is a simplified form of $\Phi(c_1,c_2)=0$ with arbitrary fonction $\Phi$ of two variables. In this method it is not taken account of a "direction". No symbol $r,s$ appear. At end, one is supposed to check the final result in puting it into the PDE and to check the consistency with the specified condition. Eventually some unvalid solution might be rejected. – JJacquelin Sep 26 '21 at 14:55
0

Let us apply the method of characteristics, with initial condition $u(x_0, 2x_0) = x_0^2$:

  • $\frac{\text d}{\text d s} y = 1$, letting $y(0) = 2x_0$ we know $y = 2x_0 + s$.
  • $\frac{\text d}{\text d s} u = -\frac12 u$, letting $u(0) = x_0^2$ we know $u = x_0^2 e^{-s/2}$
  • $\frac{\text d}{\text d s} x = u$, letting $x(0) = x_0$ we know $x = x_0 - 2 x_0^2 (e^{-s/2} - 1)$

The latter is rewritten $$ x = x_0 - 2 u\, (1 - e^{y/2-x_0}) . $$ Resolution w.r.t. $x_0$ yields the expression $$ x_0 = W_n\left(-2u e^{y/2-2 u - x} \right) + 2 u + x $$ where $W_n$ is the analytic continuation of the product log function, to be substituted in $ue^{y/2} = x_0^2 e^{x_0}$ to get the final implicit expression.

EditPiAf
  • 20,898
  • I've tried to take this a bit further, but again, no real progress. I tried to rearrange and substitute to eliminate $s$ or $x_0$, neither of which worked. I really am quite lost here. I don't think I can find ways to eliminate either of those variables, which means that I'm going to end up with some implicit function where $u(x,y)$ will contain $s$ and $x_0$, or, $u(x,y)=u(x(x_0,s),y(x_0,s))$, I'm just not sure what to do with the information that I have to end up with that kind of answer. – Maksymilian5275 Sep 25 '21 at 09:20
  • @Maksymillian5275 analytical resolution might involve special functions, see updated post (I used Wolfram alpha to solve for $x_0$). – EditPiAf Sep 27 '21 at 21:07