- #1
- 858
- 914
Homework Statement
Let [itex]f\colon\mathbb{R}^m\to\mathbb{R}[/itex]. All partial derivatives of [itex]f[/itex] are defined at point [itex]P_0\colon = (x_1, x_2, ... , x_m)[/itex].
If [itex]f[/itex] has local extremum at [itex]P_0[/itex] prove that [itex]\frac{\partial f}{\partial x_j} (P_0) = 0, j\in \{1, 2, ..., m\}[/itex]
Homework Equations
Fermat's theorem:
Let [itex]f\colon\mathbb{R}^m\to\mathbb{R}[/itex]. If [itex]f[/itex] is differentiable and has local extremum at point [itex]P_0[/itex] then [itex]\nabla f(P_0) = \overrightarrow{0}[/itex]
The Attempt at a Solution
Assume [itex]f[/itex] has local minimum at point [itex]P_0[/itex], then there exists [itex]\varepsilon > 0[/itex] such that [itex]f(P)\geq f(P_0)[/itex] for every [itex]P\in B(P_0,\varepsilon)[/itex].
Let [itex]j\in \{1,2, ..., m\}[/itex]. Observe the function:
[tex]g(t) = f(x_1, x_2, ... x_{j-1}, t, x_{j+1}, ... , x_m)[/tex]
Note that:
1) [itex]g[/itex] is defined within [itex]t\in (x_j -\varepsilon, x_j +\varepsilon)[/itex]
2) the function [itex]g[/itex] has local minimum at the point [itex]t = x_j[/itex]
3) [itex]g[/itex] is differentiable at [itex]t = x_j[/itex], also [itex]g'(x_j) = f_{x_j}'(P_0)[/itex]
Define for every [itex]t\in\mathbb{R}\ \ \ S_t\colon = (x_1, ...,x_{j-1}, t, x_{j+1}, ..., x_m)[/itex] then for every [itex]t\in (x_j -\varepsilon, x_j +\varepsilon)[/itex] it follows that [itex]d(Q_t, P_0) = |t - x_j| < \varepsilon[/itex], therefore [itex]Q_t\in D\subset\mathbb{R}^m[/itex] i.e [itex]g[/itex] is defined at point [itex]t[/itex] and for every [itex]t[/itex]:
[tex]g(t) = f(Q_t)\geq f(P_0) = g(x_j) [/tex] which means [itex]g[/itex] has local minimum at the point [itex]x_j[/itex].
From 2) and 3) - according to Fermat's theorem [itex]f_{x_j}'(P_0) = g'(x_j) = 0 _{\blacksquare}[/itex]
Last edited: