1
$\begingroup$

Given the functional: $$ F[\phi] = \int_V \frac{k_B T}{a^3}\phi\ln(\phi) \ ds = \int_V I(\phi) ds $$ I want to find the functional derivative. I believe this would result in: $$ \frac{\delta F}{\delta \phi} = \frac{\partial I}{\partial \phi}=\frac{k_B T}{a^3}[\ln(\phi)+1]$$ However, the paper I am following along has only the first term. Is my calculation correct? Note, in this case I set the functional derivative equal to the partial derivative because the functional doesn't contain any higher derivatives - hence those partials vanish.

$\endgroup$
4
  • 3
    $\begingroup$ Which paper?... $\endgroup$ Commented Jul 27, 2022 at 14:56
  • 7
    $\begingroup$ The last formula you wrote (that is correct) is not the partial derivative of $F$ but the partial derivative of the integrand of F. Notice that in its present form, your question is not about Physics and looks like a homework-like question. $\endgroup$ Commented Jul 27, 2022 at 14:59
  • $\begingroup$ Oh thanks, that is a typo: by the partial derivative of $F$ with respect to $\phi$, I had meant that of the integrand which I should have defined as something else - maybe $I(\phi)$. And no this is not a homework question. I am merely trying to follow the derivation of the gradient dynamics equations for $h,\psi$ as given in a thesis I was advised to review for my own research. Given the form of the question I asked though, perhaps it would've been more appropriate to post to the math stackexchange $\endgroup$
    – Mjoseph
    Commented Jul 27, 2022 at 16:14
  • 1
    $\begingroup$ Never too late to supplant your I(φ) in... $\endgroup$ Commented Jul 27, 2022 at 20:37

1 Answer 1

2
$\begingroup$

Your attempt points in the right direction, but note that the functional derivative is not the partial derivative as you're deriving with respect to a function and not a variable. Nonetheless, they are connected for certain functionals.

Take a compactly supported smooth function $\psi$, then by the definition of the functional derivative: \begin{align*} \int_V\frac{\delta F[\phi]}{\delta\phi}\psi\; ds \stackrel{!}{=}\left[\frac{\mathrm d}{\mathrm d\varepsilon}F[\phi+\varepsilon \psi]\right]_{\varepsilon=0} =\ldots =\int_V\frac{k_\mathrm{B}T}{a^3}(\ln(\phi)+1)\psi\mathrm ds, \end{align*} from which the result follows with the fundamental theorem of the calculus of variations. I think you can fill in the two missing steps, where you just have to put in your expression of the functional, for yourself.

$\endgroup$
2
  • $\begingroup$ Yes, I can fill in the missing steps to obtain the RHS of your answer. Thank you for this response! As a generalization, given some functional with an integrated $F[\phi,\nabla\phi]$, or even possibly higher derivatives, how would this process change? $\endgroup$
    – Mjoseph
    Commented Jul 27, 2022 at 16:19
  • 1
    $\begingroup$ The functionals considered in physics are often given as $F[\phi]=\int L(\phi,\nabla\phi)$ with $L$ being the Lagrange density. Then you have: $$\frac{\delta F[\phi]}{\delta\phi}=\frac{\partial L}{\partial\phi}-\nabla\frac{\partial L}{\partial\nabla\phi}.$$ Putting this equal to zero is the Euler-Lagrange equation. If you add in higher derivatives, you get the more general Euler-Possion equation. $\endgroup$ Commented Jul 27, 2022 at 16:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.