Given $\Omega:=[x_a,x_b]\times[y_a,y_b]\subset\mathbb{R}^2$, consider the problem $$ \max_{f\in \mathcal{C}^2(\Omega, \mathbb{R})} \iint_\Omega \left[f_y(x,y)\int_{x_a}^{x}f(z,y)\,dz\right]\, dx\, dy $$ subject to a constraint of the form $$ \iint_\Omega \left[\int_{x_a}^{x}f_y(z,y)\, dz -\int_{x_a}^{x_b} z\,f_y(z,y)\, dz\right]^2\, dx \, dy= c \in\mathbb{R} $$ where the subscript $y$ denotes the partial derivative w.r.t. the second variable, i.e., $y$.

I thought the first step would have been to write down the Euler-Lagrange equations, i.e., given the Langragian $\mathcal{L}$, $$ \frac{\partial}{\partial y}\left(\frac{\partial \mathcal{L}}{\partial f_y}\right) = \frac{\partial \mathcal{L}}{\partial f} $$

but the Lagrangian is pretty involved and I got stuck. Do you have any hint?


You can re-write: $$ F(x,y)=\int_{x_a}^xf(z,y)dz $$ It is easy to check that $F_x(x,y)=f(x,y)$. Your problem becomes:
$$ \int \int_{\Omega} F_{xy}(x,y)F(x,y)dxdy $$
Dealing with the constrain, it seems a lot more difficult. You can use differentiation under the integral sign to show that $$ \int_{x_a}^x f_y(z,y)dz=F_y(x,y) $$ As for the second term, you can use the integration by parts formula, which is (for $F'=f$): $$ \int_a^b xf(x)dx=xF(x)|_a^b-\int_a^b F(x)dx $$ Now we can get rid of that "annoying" $z$ in front of $f_y$: $$ \int_{x_a}^{x_b}zf_y(z,y)dz=\int_{x_a}^{x_b}zF_{zy}(z,y)dz=zF_y(z,y)|_{x_a}^{x_b}-\int_{x_a}^{x_b} F_y(z,y)dz $$ This seems all pretty good by now and your constraint becomes: $$ \int \int_{\Omega}(F_y(x,y)-zF_y(z,y)|_{x_a}^{x_b}+\int_{x_a}^{x_b} F_y(z,y)dz)^2dxdy=c $$
It might be possible to simplify the constrain even more by using Fubini, since all of the terms seem to involve a $\partial_y$, but that is a really lengthy calculation.
Now use that your function $f \in C^2$, which means that $F \in C^2$. It also makes exchanging the order of derivatives possible. You can now apply higher-order Euler-Lagrange equations to find the suitable PDE. See the Wikipedia page for a full formula (https://en.wikipedia.org/wiki/Euler–Lagrange_equation).
In your case, your Euler-Lagrange equation becomes (I used the different notations for a reason): $$ F_{xy}-\partial_{xy}F=0 $$ which is a Null-Lagrangian. That means it holds trivially for a all smooth functions. Now you just have to deal with the constraint. Note, however, you have only found a a candidate for a critical point so far.

  • $\begingroup$ I actually wanted to write the Lagrangian by including the constraint (via Lagrange multiplier), and I think the way you wrote the constraint doesn't look very good to deal with, yet $\endgroup$ – William Tomblin Feb 4 at 10:53

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.