In my class of Differential Geometry, the teacher defined geodesics as follows:

A regular curve on a regular surface, denoted as $\gamma:I\subset\Bbb{R}\to S$, ($S$ is the surface) is a geodesic if, $\forall t\in I$, the vector $\gamma"(t)$ is a normal vector to $S$ at the point $\gamma(t)$.

With this definition, I must prove for an exposition project that the möbius band can have geodesics.

The problem is that saying that a vector is normal to a surface implies orientation, and the möbius band is un-orientable.

So this is my question: How can i define geodesics on an un-orientable surface like the möbius band, and with that, how do i calculate them?

Can't the möbius band have any geodesics at all, because of its un-orientability?

If you can provide me a reference, i would apreciate it.

  • 1
    $\begingroup$ if you could show $D_{C'}C'=C''$ for any curve $C$ on any surface and after you project that on the tangent space of the surface, you will get $$\nabla_{C'}C'=C''-(N\cdot C'')N.$$ $\endgroup$ – janmarqz Nov 19 '19 at 3:12
  • 1
    $\begingroup$ Here $D$ is the std covariant derivative of $\mathbb R^3$. $\endgroup$ – janmarqz Nov 19 '19 at 3:18
  • 1
    $\begingroup$ So, an equivalent criterion for a curve to be a geodesic is $\nabla_{C'}C'=\vec 0$ $\endgroup$ – janmarqz Nov 19 '19 at 3:25

Once you have stablished that $\nabla_{\gamma'}\gamma'=\gamma''-(\gamma\cdot\gamma'')N$ then being $\gamma$ a geodesic with your definition we get $\nabla_{\gamma'}\gamma'=0$. You can consider that $\gamma=\Phi\circ\alpha$, where $\Phi$ is the local parametrization for the surface and $\alpha$ is a curve in the domain of $\Phi$, hence $\gamma'=J\Phi\circ\alpha'$. If $\alpha(t)=v(t)e_1+w(t)e_2$ then $\alpha'=v'e_1+w'e_2$ and for $\gamma'$ we get $$\gamma'=v'\partial_1+w'\partial_2,$$ where $\partial_1=J\Phi\ e_1$ and $\partial_2=J\Phi\ e_2$ is the tangent space base.

So, by the properties of a covariant derivative we calculate: \begin{eqnarray*} \nabla_{\gamma'}\gamma'&=&\nabla_{v'\partial_1+w'\partial_2}(v'\partial_1+w'\partial_2)\\ &=&v'\nabla_{\partial_1}(v'\partial_1+w'\partial_2)+w'\nabla_{\partial_2}(v'\partial_1+w'\partial_2)\\ &=&v'\nabla_{\partial_1}(v'\partial_1)+v'\nabla_{\partial_1}(w'\partial_2)+w'\nabla_{\partial_2}(v'\partial_1)+w'\nabla_{\partial_2}(w'\partial_2)\\ &=&(v''+v'^2\Gamma^1{}_{11}+2v'w'\Gamma^1{}_{12}+w'^2\Gamma^1{}_{22})\partial_1+ (w''+v'^2\Gamma^2{}_{11}+2v'w'\Gamma^2{}_{12}+w'^2\Gamma^2{}_{22})\partial_2. \end{eqnarray*} Therefore for $\nabla_{\gamma'}\gamma'$ to be null we need the coefficients comply $$v''+v'^2\Gamma^1{}_{11}+2v'w'\Gamma^1{}_{12}+w'^2\Gamma^1{}_{22}=0,$$ and $$w''+v'^2\Gamma^2{}_{11}+2v'w'\Gamma^2{}_{12}+w'^2\Gamma^2{}_{22}=0.$$ These two ordinary-second-order-non-linear-homogenous equations, under the suitable conditions on the gammas, guarantee that solutions for the two functions $v=v(t)$ and $w=w(t)$ would exists.

For a parametrization of a Möbius band, like: \begin{eqnarray*} x&=&(2+v\cos(w/2))\cos w\\ y&=&(2+v\cos(w/2))\sin w\\ z&=&v\sin(w/2) \end{eqnarray*} where $-0.7<v<0.7$ and $0\le w<2\pi$ is likely to fulfill the conditions.

Excuse me for being too sketchy but i am available 24/7, here at MSE to any discussions or questioning. : )


Normality is not dependent on orientation.

In $\mathbb R^3$, if I give you a plane $P$ and a vector $V$ based at a point of $P$, I can tell you whether $V$ is normal to $P$ without mentioning any orientation of $P$: $V$ is normal to $P$ if and only if $V \cdot W = 0$ for all vectors $W$ parallel to $P$. All I've used to formulate this definition is the (standard) inner product on the vector space $\mathbb R^3$.

You can now apply this principle at the point $\gamma(t)$, using the tangent plane $P = T_{\gamma(t)} S$ and the vector $V = \gamma''(t)$.

  • $\begingroup$ but $\gamma"(t)$ being normal to S doesn't also mean that $\gamma"(t)=k*(N(\gamma(t)))$, where the vector field N is given by the orientation? $\endgroup$ – Armando Rosas Nov 19 '19 at 1:23
  • 3
    $\begingroup$ All your care about is that $\gamma''(t)$ lies on the line spanned by $\pm N(\gamma(t))$. The normal line is well-defined, even for a nonorientable surface. $\endgroup$ – Ted Shifrin Nov 19 '19 at 2:16
  • $\begingroup$ . . . nice . . . $\endgroup$ – janmarqz Nov 19 '19 at 18:29

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.