1

Let $F:\mathbb{R}^3 \to \mathbb{R}^2 $ be continuously differentiable and let $p\in\mathbb{R}^3 $ for which $F(p)=0$ . Assume $rank DF|_p =2 $ and denote by $E$ the set $E=(x\in \mathbb{R}^3 | F(x)=0) $ .

Implicit function theorem guarantees the existence of a neighberhood $B$ of $p$ such that $B\cap E$ is a smooth curve (/path).

If $DF|_p = \begin{bmatrix} 1 & -1 & 2 \\[0.3em] 3 & 0 & 1 \end{bmatrix} $ how can one calculate a tangent line to the curve $B\cap E$ I just mentioned ? (at $p$ obviously).

Thanks !

criticism
  • 127

1 Answers1

0

If $\gamma(t)$ is some differentiable parametrization of $E \cap B$, then $F(\gamma(t)) = 0$. By the chain rule, $0 = \frac{d}{dt}F(\gamma(t)) = DF(\gamma'(t)) = 0$ for all $t$ for which $\gamma(t)$ is defined. Thus $\gamma'(t) \in \ker DF$. This implies that $\gamma'(t)$ can, up to a scalar multiple, be had by solving the linear system $DF(\gamma'(t)) = 0$. With $DF_p$ as given, we see that with $\gamma'(t_0) = (u, v, w)^T$, where $\gamma(t_0) = p$, we have

$u - v +2w = 0, \tag{1}$

and

$3u + w = 0; \tag{2}$

the solution to (1), (2) is the one-dimensional family of vectors $w(-1/3, 5/3, 1)^T$; this in turn implies the requisite tangent line is $\hat p + w(-1/3, 5/3, 1)$, where we take $\hat p$ to mean the ordered triple $(p_1, p_2, p_3)$ of coordinates of the point $p$.

Hope this helps. Cheerio,

and as always,

Fiat Lux!!!

Robert Lewis
  • 71,180
  • A few questions, if I may:
    1. I guess that in your first line, you mean "parameterization of $B\cap E$ , right?2. Can you explain your use of the chain rule by which you arrived to $DF(\gamma '(t))=0$?

    I think I understood the rest. Thanks a lot !

    – criticism Jan 18 '14 at 08:52
  • @ criticism 46: of course your right about $E \cap B$; so edited. I also added a minor edit in re. the chain rule business, does that clarify things? If not, let me know via comment and I'll be glad to further explain. Regards. – Robert Lewis Jan 18 '14 at 09:42
  • @criticism 46: chain rule shows that $\frac{d}{dt}F_k(\gamma) = \sum \frac{\partial F_k}{\partial x_j}\gamma_j'$etc. Need more? Let me know! – Robert Lewis Jan 18 '14 at 09:55
  • @ criricism: please check the comments for more info. I may have mispelled your handle and my remarks might not have reached you! – Robert Lewis Jan 18 '14 at 10:15
  • I think I got it. Thanks a lot ! (I hope it will be ok if I'll reply here again if I'll have any more questions about this. Thanks) – criticism Jan 18 '14 at 16:01
  • @ criticism: yes, of course! – Robert Lewis Jan 18 '14 at 17:08