# Derivative of a functional formula

+ 5 like - 0 dislike
419 views

I would like the derive the following equation (eq. (25) in [1]):

${d F[\psi(\theta)] \over d \theta} = \int {\delta F[\psi] \over \delta \psi(x; \theta)} {\partial \psi(x; \theta) \over \partial \theta} d x$

For example $\psi(x; \theta) = \varphi(x) \cos(\theta) + \phi(x) \sin(\theta)$. The functional $F[\psi]$is for example $F[\psi] = \int \left({d \psi(x)\over d x}\right)^2 + \psi(x) V(x) d x$

Besides the fact that it works, and that it is used in [1], I found at least the following, hand-waving, explanation:

${d \over d \theta} F(\psi_1, \psi_2, \psi_3, ...) = \sum_i {\partial F \over \partial \psi_i} {d \psi_i \over d \theta}$

Which is a discrete form of the above formula. But I would like to derive it using the definition of the functional derivative, or some other more rigorous way.

[1] Jiang, H., & Yang, W. (2004). Conjugate-gradient optimization method for orbital-free density functional calculations. The Journal of Chemical Physics, 121(5), 2030–2036. doi:10.1063/1.1768163

+ 5 like - 0 dislike

I will sketch just the idea omitting several (actually important) mathematical details.

If $F=F[\psi]$ is a functional, we define, if exists, its functional derivative $\frac{\delta F}{\delta \psi}$ as the function (a distribution, more generally) such that

$$\left.\frac{dF[\psi + h \phi]}{dh}\right|_{h=0}= \int \frac{\delta F}{\delta \psi}(x) \phi(x) dx$$

for every test function $\phi$. Now suppose that $\psi$ parametrically depends on the parameter $\theta$. We have

$$\frac{dF[\psi_\theta]}{d\theta} =\lim_{h\to 0} \frac{1}{h}\left(F[\psi_{\theta + h}] - F[\psi_\theta]\right)$$

$$=\lim_{h\to 0} \frac{1}{h}\left(F[\psi_\theta + h \partial_\theta \psi + O_\theta(h^2)] - F[\psi_\theta]\right)$$

$$= \lim_{h\to 0} \frac{1}{h}\left(F[\psi_\theta + h \partial_\theta \psi] - F[\psi_\theta]\right) \:.$$

$$= \left.\frac{dF[\psi_\theta + h \partial_\theta \psi_\theta]}{dh}\right|_{h=0}\:.$$

Applying the definition of functional derivative given above, we have that

$$\frac{dF[\psi_\theta]}{d\theta} = \left.\frac{dF[\psi_\theta + h \partial_\theta \psi_\theta]}{dh}\right|_{h=0}= \int \frac{\delta F}{\delta \psi_\theta}(x) \partial_\theta\psi_\theta (x) dx\:.$$

There are several open mathematical issues in the outlined procedure (for instance dropping $O_\theta(h^2)$ in the formula above is not so easy as it could seem at first glance). Nevertheless, everything goes right (it can be proved by direct inspection) when dealing with functionals $F[\psi]$ of integral form  like the one you consider and assuming to work with a domain of suitably smooth and rapidly vanishing at infinity functions.

answered May 14, 2014 by (2,085 points)
edited May 14, 2014

I think this is exactly the derivation that I was looking for! Thanks a lot. I will go over it slowly later today and if all works out, accept this as the correct answer.

I went over this and this derivation is correct. You can actually do it even simpler like this:

${d F[\psi(\theta)] \over d \theta} =\left.{d F[\psi(\theta+\epsilon)] \over d \epsilon}\right|_{\epsilon=0} = \left.{d F[\psi(\theta)+\epsilon {d \psi(\theta) \over d \theta} + O(\epsilon^2)] \over d \epsilon}\right|_{\epsilon=0} = \left.{d F[\psi(\theta)+\epsilon {d \psi(\theta) \over d \theta}] \over d \epsilon}\right|_{\epsilon=0} = \int {\delta F[\psi] \over \delta \psi} {d \psi(\theta) \over d \theta} d x$

where we used the definition of a variation with $\delta \psi = {d \psi(\theta) \over d \theta}$:

$\delta F = \left. {d \over d\epsilon} F[\psi+\epsilon \delta\psi] \right|_{\epsilon=0} = \int {\delta F \over \delta \psi} \delta\psi d x$

 Please use answers only to (at least partly) answer questions. To comment, discuss, or ask for clarification, leave a comment instead. To mask links under text, please type your text, highlight it, and click the "link" button. You can then enter your link URL. Please consult the FAQ for as to how to format your post. This is the answer box; if you want to write a comment instead, please use the 'add comment' button. Live preview (may slow down editor)   Preview Your name to display (optional): Email me at this address if my answer is selected or commented on: Privacy: Your email address will only be used for sending these notifications. Anti-spam verification: If you are a human please identify the position of the character covered by the symbol $\varnothing$ in the following word:p$\hbar$ysicsOver$\varnothing$lowThen drag the red bullet below over the corresponding character of our banner. When you drop it there, the bullet changes to green (on slow internet connections after a few seconds). To avoid this verification in future, please log in or register.