9

The question asks for the Taylor expansion of a functional. Thus, given a real functional $f(g(x))$, what is the Taylor expansion about a function $h(x)$. What if the function is multi-variate, e.g., $f(x_1,x_2,g(x_1,x_2))$?

I've searched the web for an answer to this, and haven't come up with anything definite. If someone could help, I'd be grateful.

quid
  • 42,135
Jorge
  • 289

2 Answers2

12

This is a physicist answer and is thus not expressed in the better mathematical "vocab", but the Taylor expansion around $x_0$ of a functional $f$ of a function $g(x)$, i.e., of $f[g(x)]$ noted $f[g]$ for simplicity is:

$f[g]=f[g_0]+\int dx \frac{\delta f[g_0]}{\delta g(x)}\Delta g(x)+\frac{1}{2!}\int dx dx^\prime \frac{\delta^2f[g_0]}{\delta g(x^\prime)\delta g(x)}\Delta g(x^\prime)\Delta g(x) + ... $

with $g_0=g(x_0)$ and $\Delta g(x)=g(x)-g(x_0)$

max
  • 427
  • 3
    What is the intuition in the "quadratic" term being $\frac{\delta^2 f}{\delta g(x') \delta g(x)}$ I would've expected to see: $\frac{\delta^2 f}{\delta^2 g(x) }$. Do you have a link of some kind on this? – Sidharth Ghoshal Feb 02 '19 at 18:23
  • I don't understand why $g_0=g(x_0)$ ect is correct. Because this would suggest that $g$ is a function, and $g_0$ is a value ($g$ evaluated at a given $x_0$), which would give $f[g_0]$ no meaning. Have I missed something? As it would seem to me $g_0(x)$ is the whole function that is taken as the expansion point. – Jack Wetherell Apr 29 '20 at 23:03
  • 1
    I think @max wanted to say that $g_0 = g_0(x)$ is the function wrt which the expansion is calculated – Diego Gomez Aug 19 '20 at 14:56
  • arent the units wrong? For instance, from the 1st order term, (integral of delta_f dx) must have the same units as f. How does delta_f have units of 1/x? – Mondo Duke Aug 27 '22 at 06:32
4

I think this paper may be helpful to you.

"Taylor-series expansion of density functionals," by Matthias Ernzerhof. Phys. Rev. A 50, 4593 – Published 1 December 1994

Its abstract reads:

A Taylor-series expansion of the density functional F[n] is constructed from the given expansion for the energy E[v]. The formalism is used to investigate the noninteracting kinetic-energy functional for one-dimensional systems. Furthermore, a formal analysis of the Taylor expansion of the density functional for an interacting electron system at finite temperature is given and the relation between density-functional theory and the field-theoretical approach to the many-particle problem is enlightened.

You can also find a chain rule to work through the high order functional derivatives that you will need.

quid
  • 42,135