Let $\phi:\mathbb{R}_+\times \mathbb{R}_+\to\mathbb{R}$ be a function with continuous partial derivatives in both arguments which we denote by $\partial_1$ and $\partial_2$. Consider the following differential equation: \begin{align} &\partial_1[\phi(t,x)+2\phi(t,y)]+\partial_2[\phi(t,x)+2\phi(t,y)] +6\phi(t,0)-2\phi(t,x)\phi(t,y)-\phi(t,y)^2=0\;,\\[1ex] &0\leq t<\infty\;,\;0\leq x<y\;. \end{align} This is not a PDE in the classical sense.

Question: Does the equation have a solution?

  • $\begingroup$ I do not fully understand your notation: why don' t you consider a map $t\mapsto \phi_t$, with $\phi_t=\phi_t(x,y)$, instead? You could try to define your PDE involving the derivatives w.r.t. $x$ and $y$ for $any$ $t\geq 0$. Does it convince you? $\endgroup$ – Avitus Aug 13 '13 at 8:15
  • $\begingroup$ By continuity you can take the limit $x\to y$ and show that necessarily your function $\phi$ must solve also the equation $$ \partial_t \phi + \partial_x\phi - \phi^2 = -2 f$$ where $f(t) = \phi(t,0)$ is independent of $x$. So it may help to consider the solvability of the reduced equation above. $\endgroup$ – Willie Wong Aug 13 '13 at 8:33


$$ F(t,x,y) = \phi_1(t,x) + 2\phi_1(t,y) + \phi_2(t,x) + 2 \phi_2(t,y) + 6 \phi(t,0) - 2\phi(t,x) \phi(t,y) - \phi(t,y)^2 $$

Your equation requires $F \equiv 0$ for $t\in [0,\infty)$ and $0\leq x < y$. For any point in the domain we must have that $\mathrm{d}F = 0$ This is equivalent to

$$ \begin{align} \phi_{11}(t,x) + 2\phi_{11}(t,y) + \phi_{12}(t,x) + 2 \phi_{12}(t,y) + 6 \phi_1(t,0) - 2 \phi_1(t,x) \phi(t,y) - 2 \phi(t,x) \phi_1(t,y) - 2 \phi(t,y)\phi_1(t,y) &= 0 \\ \phi_{12}(t,x) + \phi_{22}(t,x) - 2\phi_2(t,x) \phi(t,y) &= 0\\ \phi_{12}(t,y) + \phi_{22}(t,y) - \phi_2(t,y) \phi(t,x) - \phi_2(t,y) \phi(t,y) &= 0 \end{align} $$

Consider the second of the equations. Again this holds for all points $(t,x,y)$ in the relevant domain. Taking a derivative in $y$ then requires that $$ \phi_2(t,x) \phi_2(t,y) = 0 $$ (note that for this we only need to assume that $\phi$ is twice continuously differentiable and not thrice!) Taking the limit $x \to y$ in the domain, by continuity we must have that $\phi_2 = 0$ everywhere!

This implies that $\phi$ is constant in its second variable, and you equation reduces to just a simple ODE $$ \phi'(t) + 2 \phi(t) - \phi(t)^2 = 0 $$ which we can solve to be $$ \frac12 (\log (2 - \phi) - \log\phi) ) + C = t \implies C e^{2t} = \frac{2-\phi}{\phi} \implies \phi(t) = \frac{2}{C e^{2t} + 1}$$

where the constant $C$ on the far right hand side is determined by $\phi(0)$: $$ C = \frac{2}{\phi(0)} - 1$$

In summary, if we assume $\phi$ is twice continuously differentiable, then $\phi$ must be independent of its second argument and

$$ \phi(t,x) = \frac{2}{\left(\frac{2}{\phi(0,0)} - 1\right) e^{2t} + 1} $$


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.