1
$\begingroup$

Problem Statement: Let $f:U\rightarrow \mathbb{R}^{n}$ and $g:U\rightarrow \mathbb{R}^{p}$ be differentiable ($U\subset \mathbb{R}^{m}$ open). Define: $$h(\mathbf{x}):U\setminus \ker{f}\rightarrow \mathbb{R}^{p},$$ $$\ \ \ \ \mathbf{x}\mapsto \frac{g(\mathbf{x})}{\lVert f(\mathbf{x})\rVert},$$ with $\lVert\cdot\rVert$ the euclidean norm in $\mathbb{R}^{p}$. Compute $dh(\mathbf{x})[\mathbf{v}]$, for each $\mathbf{v}\in \mathbb{R}^{m}$

So I am trying to work this problem out, but am a little confused because I would think that I need to use the product/quotient rule to determine $dh$, but we have not learned that in my Analysis course yet. I first took $\phi(\mathbf{x})=\lVert f(\mathbf{x})\rVert=\sqrt{\langle f(\mathbf{x}),f(\mathbf{x})\rangle}$, where $\langle\ ,\ \rangle$ is the standard inner product in $\mathbb{R}^{m}$, then I computed $d\phi(\mathbf{x})[\mathbf{v}]=\frac{1}{\lVert f(\mathbf{x})\rVert}\langle df(\mathbf{x})[\mathbf{v}],f(\mathbf{x})\rangle$. I am not sure where to go from here. Should I simply apply regular the product rule with $g(\mathbf{x})$ and $\frac{1}{\lVert f(\mathbf{x})\rVert}$?

Also, when computing differentials in real analysis, must we be precise in doing so by using the $\epsilon$-$\delta$ definition of derivative? In computing $d\phi(\mathbf{x})[\mathbf{v}]$, I just applied the standard limit definition to find the directional derivative in the direction of $\mathbf{v}\in\mathbb{R}^{m}$.

Any hints to lead me in the right direction are appreciated!

$\endgroup$
1
$\begingroup$

I don't really understand your comment in the second to the last paragraph. Anyway, here's my computation using the definition \begin{align} \lim_{t\rightarrow 0} \frac{h(\mathbf{x}+t\mathbf{v})-h(\mathbf{x})}{t} =&\ \lim_{t\rightarrow 0} \frac{\frac{g(\mathbf{x}+t\mathbf{v})}{\|f(\mathbf{x}+t\mathbf{v})\|}-\frac{g(\mathbf{x})}{\|f(\mathbf{x})\|}}{t} = \lim_{t\rightarrow 0} \frac{g(\mathbf{x}+t\mathbf{v})\|f(\mathbf{x})\|-g(\mathbf{x})\|f(\mathbf{x}+t\mathbf{v})\|}{t\|f(\mathbf{x})\|\|f(\mathbf{x}+t\mathbf{v})\|}\\ =&\ \lim_{t\rightarrow 0} \frac{\left(g(\mathbf{x})+tdg(\mathbf{x})\cdot\mathbf{v}+\mathcal{O}(t^2)\right)\|f(\mathbf{x})\|-g(\mathbf{x})\left(\|f(\mathbf{x})\|+t\frac{f(\mathbf{x}) df(\mathbf{x})\cdot \mathbf{v}}{\|f(\mathbf{x})\|}+\mathcal{O}(t^2)\right)}{t\|f(\mathbf{x})\|\|f(\mathbf{x}+t\mathbf{v})\|}\\ =&\ \frac{\|f(\mathbf{x})\| dg(\mathbf{x}) - \frac{g(\mathbf{x})f(\mathbf{x})}{\|f(\mathbf{x})\|}df(\mathbf{x})}{\|f(\mathbf{x})\|^2}\cdot \mathbf{v} =: dh(\mathbf{x})\cdot \mathbf{v}. \end{align} Note: I have only used chain rule and one variable calculus.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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