3

I was given the following differential equation: $$\left\{\begin{matrix}u_t+2u_x=0, \ \ x>0, \ \ t>0, \\ u(x,0)=\arctan(x), \ \ x>0, \hspace{0.15cm} { }\\ u(0,t)=\dfrac{t}{1+t^2}, \ \ t>0. \hspace{0.8cm} { }\end{matrix}\right.$$ But I think that it has no solution. If I'm right, maybe it is very obvious, but I'll give a proof.

Let's consider the reduced problem: $$\left\{\begin{matrix}u_t+2u_x=0, \ \ x>0, \ \ t>0, \\ u(x,0)=\arctan(x), \ \ x>0. \hspace{0.15cm} { }\end{matrix}\right. $$ It is not difficult to check that the function $\widetilde{u}(x,t)=\arctan(x-2t)$ is a solution to this problem. But it is a Cauchy problem, so $\widetilde{u}$ is unique. However, it is clear that the complete solution to the problem (taking $u(x,0)$ and $u(0,t)$), let's call it $u(x,t)$, is also a solution to the reduced problem. By uniqueness of the Cauchy problem, $u(x,t)=\widetilde{u}(x,t)$. But this is absurd, because $\widetilde{u}(0,t)=\arctan(-2t)\not=u(0,t)=t/(1+t^2)$. Then, the complete problem can't have solution.

Finally, am I right? If not, what can be wrong in my proof? Thanks for the help.

Edit (30/06/22): if possible, the solution I am looking for must be a classical solution.

Yester
  • 414

2 Answers2

2

The subtle error in the logic above lies in the fact that the solution you find in the reduced problem does not define a solution in the entirety of $(x,t)\in\mathbb{R}^2$, but only in a subset of it, defined by the equation $x>2t$. The reason why this happens is because not all characteristic curves of the PDE above pass through $\mathcal{C}_1=(x,0), x>0$. When combined with $\mathcal{C}_2=(0,t), t>0$, however, it is now the case that all characteristic curves cut through the initial surface and hence define a solution for any point in $\mathbb{R}^2$.

For a complete solution, note that the characteristic passing through an arbitrary point $(x_0,t_0)$ is given by the equation

$$x-2t=x_0-2t_0$$

When $x_0-2t_0\geq 0$ it is the case that the characteristic hits the initial value surface $\partial \Omega=\mathcal{C}_1\star\mathcal{C}_2$ at the point $(x_0-2t_0,0)$ while for $x_0-2t_0< 0$ the intersection point is $(0,t_0-x_0/2)$. Then, given the constancy of $u$ on the characteristic curves, the solution is simply given by

$$u(x_0,t_0)=\begin{Bmatrix}\arctan( x_0-2t_0),&x_0-2t_0\geq 0\\\frac{t_0-x_0/2}{1+(t_0-x_0/2)^2},&x_0-2t_0<0\end{Bmatrix}$$

Note that the solution is continuous on the break line $x=2t$, as dictated by the continuity of the initial condition on the boundary curve.

DinosaurEgg
  • 10,775
  • "Note that the solution is continuous..." and at the same time not differentiable, and hence not really a classical solution. – Artem Jun 29 '22 at 22:44
  • Never implied it was differentiable! However in this simple problem if the boundary condition was discontinuous (aka had something like a jump at the origin) the solution obtained would not be continuous as well. – DinosaurEgg Jun 30 '22 at 04:32
  • 1
    For a function to be a solution of a first order PDE continuity is not enough. The classical definition is that this function ,must be $C^1$. Yours is not along the characteristic $x=2t$ and therefore, strictly speaking, is not a solution. – Artem Jun 30 '22 at 12:32
  • Thank you, @Artem, I forgot to mention that, if possible, the solution I'm looking for should be classic and I didn't check that in DinosaurEgg's answer. – Yester Jun 30 '22 at 15:47
  • The solution yielded by the method of characteristics is certainly weak, however it seems to be unique, since there is no ambiguity in the determination procedure. Maybe someone more well versed in uniqueness proofs can show that there is no classical solution to this problem by showing that in the more general weak formulation of the problem, the above is unique. – DinosaurEgg Jul 01 '22 at 06:31
1

$$u_t+2u_x=0, \quad x>0, \quad t>0, $$ The general solution is : $$u(x,t)=F(x-2t)$$ with arbitrary function $F$ ( as long as no condition is takent into account ).

Initial condition :

$$u(x,0)=\arctan(x), \quad x>0 \implies F(x)=\arctan(x), \quad x>0 $$

Boundary condition : $$u(0,t)=\dfrac{t}{1+t^2}, \quad t>0 \quad\implies \quad F(-2t)=\dfrac{t}{1+t^2}, \quad t>0$$

Let $X=-2t$ $$F(X)=\frac{-X/2}{1+(-X/2)^2}\quad X<0$$ Thus the function $F$ is a piecewise function : $$F(X)=\begin{cases} =X\qquad X>0\\ =\frac{-X/2}{1+(-X/2)^2}\quad X<0 \end{cases}$$

Now the function $F$ is determined. We put it into the above general solution where $X=x-2t$ :

$$u(x,t)=\begin{cases} =x-2t\qquad x>2t\\ =\frac{t-\frac{x}{2}}{1+(t-\frac{x}{2})^2}\quad x<2t \end{cases}$$ This is the particular solution which satisfies both the PDE and the conditions.

Not forgetting that $x>0$ and $t>0$ as specified in the wording of the problem.

enter image description here

JJacquelin
  • 66,221
  • 3
  • 37
  • 87