Condition for Differentiable Functional to have Extremum
![]() | This article needs to be linked to other articles. You can help $\mathsf{Pr} \infty \mathsf{fWiki}$ by adding these links. To discuss this page in more detail, feel free to use the talk page. When this work has been completed, you may remove this instance of {{MissingLinks}} from the code. |
Theorem
Let $y, h \in S: \R \to \R$ be real functions.
Let $J \sqbrk y: S \to \R$ be a differentiable functional.
Then a necessary condition for the differentiable functional $J \sqbrk {y; h}$ to have an extremum for $y = \hat y$ is:
- $\bigvalueat {\delta J \sqbrk {y; h} } {y \mathop = \hat y} = 0$
Proof
Suppose $J \sqbrk {y; h}$ attains a minimum for $y = \hat y$.
Then:
- $\Delta J \sqbrk {\hat y; h} \ge 0$
By definition of the differentiable functional:
- $\Delta J \sqbrk {y; h} = \delta J \sqbrk {y; h} + \epsilon \size h$
where:
- $\ds \lim_{\size h \mathop \to 0} \epsilon = 0$
Hence, there exists $\size h$ small enough, that signs of $\Delta J \sqbrk {y; h}$ and $\delta J \sqbrk {y; h}$ match.
Therefore, for $\size h$ small enough it holds that $\delta J \sqbrk {\hat y; h}\ge 0$.
Aiming for a contradiction, suppose $\delta J \sqbrk {y; h_0} \ne 0$ for some $h_0 \in S$.
Then:
- $\forall \alpha > 0: \delta J \sqbrk {y; -\alpha h_0} = -\delta J \sqbrk {y; \alpha h_0}$
Therefore, for $\size h$ however small, $\Delta J \sqbrk {y; h}$ can be made to have any sign.
![]() | This article, or a section of it, needs explaining. In particular: Clarify the part with $\alpha$ You can help $\mathsf{Pr} \infty \mathsf{fWiki}$ by explaining it. To discuss this page in more detail, feel free to use the talk page. When this work has been completed, you may remove this instance of {{Explain}} from the code. |
However, by assumption that $J \sqbrk {y; h}$ has a minimum, for sufficiently small $\size h$ it holds that:
- $\Delta J \sqbrk {\hat y; h} = J \sqbrk {\hat y + h} - J \sqbrk {\hat y} \ge 0$
This automatically fixes the sign of $\delta J \sqbrk {y; h}$.
This is in contradiction with the previous statement.
Hence for all $h$ it holds that:
- $\bigvalueat {\delta J \sqbrk {y; h} } {y \mathop = \hat y} = 0$
$\blacksquare$
Sources
- 1963: I.M. Gelfand and S.V. Fomin: Calculus of Variations ... (previous) ... (next): $\S 1.3$: The Variation of a Functional. A Necessary Condition for an Extremum