0
$\begingroup$

Let $ f : \mathbb{R}^3 \rightarrow \mathbb{R}^3$ defined by $f(v)=\lambda v$ where $\lambda >0$ is constant. Consider a surface $S \subseteq \mathbb{R}^3$ and $S'=f(S)$. Find a relation between $K$ of $S$ and $K'$ of $S'$ where $K$ is the Gaussian curvature.

$\textbf{My attempt : }$ Let a parametrization of $S$, $\varphi(u,v) : U_0 \subseteq \mathbb{R}^2 \rightarrow S$. We have a parametrization of $S'$ defined by : $\phi = f \circ \varphi : U_0 \rightarrow S'$ (because $f$ is a diffeomorphism).

So :

$$\phi(u,v) = (f\circ \varphi) (u,v) = \lambda \varphi (u,v)$$

$$ \implies \phi_u = \lambda \varphi_u \hspace{0.5cm} \text{and} \hspace{0.5cm} \phi_v=\lambda \varphi_v $$ $$ \implies \phi_{uu} = \lambda \varphi_{uu} \hspace{0.5cm} \text{and} \hspace{0.5cm} \phi_{vv}=\lambda \varphi_{vv} \hspace{0.5cm} \text{and} \hspace{0.5cm} \phi_{uv}=\lambda \varphi_{uv} $$

Finally :

$E' = \langle \phi_u , \phi_u \rangle = \lambda^2 \langle \varphi_u , \varphi_u \rangle = \lambda^2 E$ and $L' = \langle \phi_{uu} , n \rangle = \lambda \langle \varphi_{uu} , n \rangle = \lambda L$

Analogous :

$F'=\lambda^2 F, G'= \lambda^2 G,M'=\lambda M,N' =\lambda N$

$\implies K' = \dfrac{L'N'-(M')^2}{E'G' - (F')^2} = \dfrac{\lambda^2(LN-M^2)}{\lambda^4(EG-F^2)}=\dfrac{1}{\lambda^2}K $

It´s right? My question is what would happen if I took an arbitrary parameterization of $S'$?

$\endgroup$
3
$\begingroup$

No need to use ugly coordinate formulas. If $N$ is a normal Gauss field for $S$, then a normal Gauss field $\widetilde{N}$ for $\lambda S$ is given by $\widetilde{N}(\lambda p) = N(p)$. The chain rule then gives that ${\rm d}\widetilde{N}_{\lambda p} = (1/\lambda){\rm d}N_p$. It follows that $\widetilde{K}(\lambda p)=K(p)/\lambda^2$ and $\widetilde{H}(\lambda p)=H(p)/\lambda$.

$\endgroup$
  • $\begingroup$ Oh thanks! , the problem is that even in class we don't see that. But your solution is very short, so I'll read what you've put there. $\endgroup$ – Juan Daniel Valdivia Fuentes Dec 4 '19 at 4:16
  • $\begingroup$ How come you don't see that? What's the definition of $K$ and $H$ that you have seen in class then? $\endgroup$ – Ivo Terek Dec 4 '19 at 4:17
  • $\begingroup$ Let $K_1,K_2$ the principal curvatures then $K=K_1 K_2$ and $H=\dfrac{K_1 + K_2}{2}$ $\endgroup$ – Juan Daniel Valdivia Fuentes Dec 4 '19 at 4:22
  • $\begingroup$ Right, so $K(p)=\det(-{\rm d}N_p)$ and $H(p)={\rm tr}(-{\rm d}N_p)/2$. Then you apply $\det$ and ${\rm tr}$ on both sides of ${\rm d}\widetilde{N}_{\lambda p}=(1/\lambda){\rm d}N_p$ to get the correct formulas. You don't even need to waste time finding $k_1$ and $k_2$. $\endgroup$ – Ivo Terek Dec 4 '19 at 4:25
  • 1
    $\begingroup$ Right, thanks for your help! $\endgroup$ – Juan Daniel Valdivia Fuentes Dec 4 '19 at 4:28

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.