13

I need to solve Problem 3.5 - 11 p. 164 of the book Partial Differential Equations by Lawrence C. Evans (2nd ed., AMS, 2010):

  1. Show that $$ u(x,t) = \begin{cases} -\dfrac{2}{3}\left(t+\sqrt{3x+t^2}\right); & \text{if } 4x + t^2 >0\\ 0; & \text{if } 4x + t^2<0 \end{cases} $$ is an (unbounded) entropy solution of $u_t + \left(\dfrac{u^2}{2}\right)_x=0$.

Clearly it is easy to see that this is both unbounded and a solution to the given PDE, however I am not sure how to gather any information to say that it satisfies the entropy condition from the solution alone. Any hints would be welcomed.

Edit. I know the definition (Evans, §3.4.3.b p. 150). A weak solution $u\in L^\infty(\Bbb R\times (0,\infty))$ of the initial value problem $u_t + \left(\dfrac{u^2}{2}\right)_x = 0$ with data $u|_{t=0} = g$ is an entropy solution if

$$ u(x+z,t) - u(x,t) \leq C \left(1 + \frac{1}{t}\right) z \tag{ii} $$ for some constant $C>0$ and a.e. $x$, $z \in \Bbb R$, $t>0$, with $z>0$.

Is it the one needed to solve the problem? How to use it?

Ѕᴀᴀᴅ
  • 34,263
DaveNine
  • 2,071
  • 2
    I think you should explain in the question what the 'entropy condition' says (explained here, page 9). This (I think) is not known to most people here. – Winther Nov 24 '14 at 05:00
  • 1
    I think that part of the issue is that I'm not sure which one is appropriate to use here. I am using it as aligned with Evan's PDE terminology, but I'm not sure if the Lax-Oleinik version of entropy is appropriate, i.e. $u(x+z,t)-u(x,t) \leq \frac{Cz}{t}$, for some constant $C$. – DaveNine Nov 24 '14 at 05:07
  • You might find some useful information in these related MSE questions: 1) and 2) – Winther Nov 24 '14 at 05:29
  • 1
    Since $u$ is decreasing with respect to $x$, the entropy condition can be satisfied by taking an arbitrary $C>0$. Did I miss something? – Ѕᴀᴀᴅ Oct 10 '18 at 06:27

2 Answers2

7

$\DeclareMathOperator{\supp}{supp}\def\d{\mathrm{d}}\def\peq{\mathrm{\phantom{=}}{}}$Note that $u_t + u u_x = u_t + \left( \dfrac{u^2}{2} \right)_x = 0$ holds for $x > -\dfrac{t^2}{4}$, and $u = 0$ for $x < -\dfrac{t^2}{4}$. Denote$$ g(x) = u(x, 0) = \begin{cases} -\dfrac{2\sqrt{x}}{\sqrt{3}}; & x > 0\\ 0; & x < 0 \end{cases}. $$

For any test function $v$, suppose $\supp(v) \subseteq \left( -\dfrac{T^2}{4}, \dfrac{T^2}{4} \right) × [0, T)$, then\begin{align*} &\peq \int_0^{+∞} \int_{-∞}^{+∞} uv_t \,\d x\d t = \int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} uv_t \,\d x\d t = \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} \int_0^T uv_t \,\d t\d x\\ &= \int_{-\tfrac{T^2}{4}}^0 \int_{\sqrt{-4x}}^T uv_t \,\d t\d x + \int_0^{\tfrac{T^2}{4}} \int_0^T uv_t \,\d t\d x\\ &= \int_{-\tfrac{T^2}{4}}^0 \left( uv \Biggr|_{t = \sqrt{-4x}}^{t = T} uv_t - \int_{\sqrt{-4x}}^T u_tv \,\d t \right)\d x + \int_0^{\tfrac{T^2}{4}} \left( uv \Biggr|_{t = 0}^{t = T} - \int_0^T u_tv \,\d t \right)\d x\\ &= - \left( \int_{-\tfrac{T^2}{4}}^0 \int_{\sqrt{-4x}}^T u_tv \,\d t\d x + \int_0^{\tfrac{T^2}{4}} \int_0^T u_tv \,\d t\d x \right)\\ &\peq + \int_{-\tfrac{T^2}{4}}^0 uv \Biggr|_{t = \sqrt{-4x}}^{t = T} \,\d x + \int_0^{\tfrac{T^2}{4}} uv \Biggr|_{t = 0}^{t = T} \,\d x\\ &= -\int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} \int_0^T u_tv \,\d t\d x - \int_{-\tfrac{T^2}{4}}^0 uv \Biggr|_{t = \sqrt{-4x}} \,\d x - \int_0^{\tfrac{T^2}{4}} uv \Biggr|_{t = 0} \,\d x\\ &= -\int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} u_tv \,\d x\d t + \int_{-\tfrac{T^2}{4}}^0 2\sqrt{-x} · v(x, \sqrt{-4x}) \,\d x - \int_0^{\tfrac{T^2}{4}} g(x) v(x, 0) \,\d x\\ &= -\int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} u_tv \,\d x\d t + \int_0^T \frac{t^2}{2} · v\left( -\frac{t^2}{4}, t \right) \,\d t - \int_0^{\tfrac{T^2}{4}} g(x) v(x, 0) \,\d x, \end{align*}\begin{align*} &\peq \int_0^{+∞} \int_{-∞}^{+∞} \frac{u^2}{2} · v_x \,\d x\d t = \int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} \frac{u^2}{2} · v_x \,\d x\d t = \int_0^T \int_{-\tfrac{t^2}{4}}^{\tfrac{T^2}{4}} \frac{u^2}{2} · v_x \,\d x\d t\\ &= \int_0^T \left( \frac{u^2}{2} · v\Biggr|_{x = -\tfrac{t^2}{4}}^{x = \tfrac{T^2}{4}} - \int_{-\tfrac{t^2}{4}}^{\tfrac{T^2}{4}} \left( \frac{u^2}{2} \right)_x · v \,\d x \right) \d t\\ &= -\int_0^T \frac{t^2}{2} · v\left( -\frac{t^2}{4}, t \right) \,\d t - \int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} uu_xv \,\d x\d t, \end{align*}$$ \int_{-∞}^{+∞} g(x) v(x, 0) \,\d x = \int_0^{\tfrac{T^2}{4}} g(x) v(x, 0) \,\d x, $$ thus\begin{align*} &\peq \int_0^{+∞} \int_{-∞}^{+∞} \left( uv_t + \frac{u^2}{2} · v_x \right) \,\d x\d t + \int_{-∞}^{+∞} g(x) v(x, 0) \,\d x\\ &= -\int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} u_tv \,\d x\d t - \int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} uu_xv \,\d x\d t\\ &= -\int_0^T \int_{-\tfrac{T^2}{4}}^{\tfrac{T^2}{4}} (u_t + uu_x) v \,\d x\d t = 0. \end{align*} Therefore, $u(x, t)$ is a weak solution.

Finally, since $u(x, t)$ is decreasing with respect to $x$, then$$ u(x + z, t) - u(x, t) \leqslant 0 \leqslant \left( 1 + \frac{1}{t} \right) z. \quad \forall x \in \mathbb{R},\ z > 0,\ t > 0 $$

Ѕᴀᴀᴅ
  • 34,263
  • +1 for effort! It gave me the idea of proceeding the other way round, by constructing the entropy solution from the initial data and comparing with the proposed solution. By the way, the derivation of the Rankine-Hugoniot condition for discontinuous weak solutions may be included in your answer (injecting this identity could simplify the verification of the fact that $u$ is a weak solution). – EditPiAf Oct 12 '18 at 22:14
5

As suggested by @Saad, setting $t=0$ in $u(x,t)$ gives the initial data $$ u(x,0) = g(x) = \left\lbrace \begin{aligned} &{-{2}}\sqrt{{x}/{3}} & &\text{if}\quad x>0, \\ &0 & &\text{if}\quad x<0 . \end{aligned} \right. $$ Instead of proving directly that $u$ is an entropy-condition-satisfying weak solution, let us derive the entropy solution to this initial-value problem and compare with the proposed $u$. We first derive a classical solution by applying the method of characteristics for short times. As long as the method of characteristics is valid, we have $u=g(x-u t)$. According to the expression of $g$, two cases must be considered:

  • for $x$ sufficiently small, $g(x)=0$, so that $u(x,t)=0$;
  • for $x$ sufficiently large, $g(x) = -2\sqrt{x/3}$, so that $u(x,t)={-\frac{2}{3}}\big(t+\sqrt{3x+t^2}\big)$.

Below is a plot of the characteristic curves in the $x$-$t$ plane deduced from the initial data:

characteristics

An intersection of the characteristic lines is observed around the origin, and the method of characteristics fails there. According to the Lax entropy condition, a shock wave is generated. Its position $x_s(t)$ must satisfy the Rankine-Hugoniot condition to ensure that such a wave is a weak solution. Therefore, the shock speed is given by $$ x'_s(t) = \frac{1}{2}\left( 0 - \frac{2}{3}\big(t+\sqrt{3 x_s(t)+t^2}\big) \right) $$ with initial position $x_s(0)=0$, i.e., $x_s(t) = -t^2/4$. By construction, the (unique) entropy solution is therefore $$ u(x,t) = \left\lbrace \begin{aligned} &{-\frac{2}{3}}\big(t+\sqrt{3x+t^2}\big) & &\text{if}\quad x>-t^2/4, \\ &0 & &\text{if}\quad x<-t^2/4 , \end{aligned} \right. $$ which ends the proof.

EditPiAf
  • 20,898