My Evans textbook presents a nonlinear first-order PDE \begin{cases}F(Du,u,x)=0 & \text{in }U \\ u=g & \text{on } \Gamma \subseteq \partial U\end{cases} where $x \in U \subset \mathbb{R}^n$. Also, $u : \bar{U} \longrightarrow \mathbb{R}$ is the unknown, $u=u(x)$.

Finally, $F=F(p,z,x)=F(p_1,\ldots,p_n,z,x_1,\ldots,x_n)$ and $g$ are assumed to be smooth functions, for $p \in \mathbb{R}^n, z \in \mathbb{R}, x \in U$.

We also set \begin{cases}D_pF = (F_{p_1},\ldots,F_{p_n}) \\ D_z F=F_z \\ D_xF=(F_{x_1},\ldots,F_{x_n}) \end{cases}

The textbook eventually goes on to differentiate this wtih respect to $x_i$: $$\sum_{j=1}^n F_{p_j}(Du,u,x)u_{x_j x_i} + F_z(Du,u,x)u_{x_i}+F_{x_i}(Du,u,x)=0$$

I have two questions about this differentiated equation.

  • Is the chain rule for several variables is used, which is why we have $\sum_{j=1}^n F_{p_j}+F_z+F_{x_i}$?
  • Why does the first term $F_{p_j}$ have a summation but the last term $F_{x_i}$ does not?

$\bullet\,$ Of course, yes. For instance, taking $i=1$ you get $$ \begin{align} \bigl(F(Du,u,x)\bigr)_{x_1}=\frac{\partial\,}{\partial x_1}\Bigl(F\bigl(u_{x_1}(x),\dots,u_{x_n}(x)\,,u(x),x_1\,,\dots,x_n\bigr)\Bigr)\\ =\sum_{j=1}^n F_{p_j}(p_1\,,\dots,p_n,u,x)\frac{\partial p_j}{\partial x_1} + F_z(Du,z,x)\frac{\partial z}{\partial x_1}+F_{x_1}(Du,u,x_1\,,\dots,x_n) \end{align} $$ where $p_j=u_{x_j}(x_1\,,\dots,x_n)$ while $z=u(x_1\,,\dots,x_n)$.

$\bullet\,$ Strictly according to the multivariate chain rule. The reason is the same why $$ \Bigl(f\bigl(g_1(x),\dots,g_n(x)\bigr)\Bigr)_{x_1}=\sum_{j=1}^n f_{g_j}(g_1\,,\dots,g_n)\cdot g_{x_1}(x) $$ involves summation, while $\bigl(h(x_1\,,\dots,x_n)\bigr)_{x_1}=h_{x_1}(x)$ does not.


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.