Let $\textbf{$\gamma$ }(t)$ be a unit-speed curve with $\kappa (t) > 0$ and $\tau (t) \neq 0$ for all $t$. Show that, if $\textbf{$\gamma$}$ is spherical, i.e., if it lies on the surface of a sphere, then $$\frac{\tau }{\kappa }=\frac{d}{ds}\left (\frac{\dot \kappa}{\tau \kappa^2}\right ) \tag 1$$

Conversely, show that if Eq. $(1)$ holds, then $$\rho^2+(\dot \rho \sigma )^2=r^2$$ for some (positive) constant $r$, where $\rho = \frac{1}{\kappa}$ and $\sigma = \frac{1}{\tau}$, and deduce that $\textbf{$\gamma$}$ lies on a sphere of radius $r$. Verify that Eq. $(1)$ holds for Viviani’s curve.



At the second part, I am facing some difficulties at showing that $\textbf{$\gamma$}$ lies on a sphere of radius $r$.

Could you give me some hints how to show it?



I have done everything besides the verification that Eq. $(1)$ holds for Viviani’s curve.

I have done the following:

The Viviani's curve is $$\gamma (t)=\left (\cos^2 t -\frac{1}{2}, \sin t\cos t , \sin t\right )$$

The curvature is given by the formula $$\kappa =\frac{ \| \gamma '' \times \gamma '\|}{\|\gamma '\|^3}$$

I found that it is equal to $$\kappa =\frac{\sqrt{5+3\cos^2 t}}{(1+\cos^2 t)^{\frac{3}{2}}}$$

Its derivative is $$\kappa '=\frac{6 \cos \sin t (\cos^2 t+2)}{\sqrt{5+3\cos^2 t}(1+\cos^2 t)^{\frac{5}{2}}}$$

The torsion is given by the formula $$\tau =\frac{(\gamma ' \times \gamma '' ) \cdot }{\|\gamma ' \times \gamma '' \|^2}$$

I found that it is equal to $$\tau =\frac{6 \cos t}{5+3 \cos^2 t}$$

Then $$\frac{\kappa '}{\tau \kappa^2}=\frac{\sin t(\cos^2 t+2)(1+\cos^2 t)^{\frac{1}{2}}}{\sqrt{5+3 \cos^2 t}}$$

I calculated $$\left ( \frac{\kappa '}{\tau \kappa^2} \right ) '$$ at Wolfram but this is not equal to $$\frac{\tau }{\kappa}$$ What have I done wrong?

  • 2
    $\begingroup$ If $\tau = 0$, what does $\frac{\ddot{\kappa}}{\tau \kappa^2}$ mean? $\endgroup$ – Jason DeVito Oct 15 '15 at 20:00
  • $\begingroup$ It should be $\tau (t) \neq 0, \forall t$. I edited it. @JasonDeVito $\endgroup$ – Mary Star Oct 15 '15 at 20:58
  • 2
    $\begingroup$ I just found this:math.stackexchange.com/questions/455661/…, which I think makes this question a duplicate. $\endgroup$ – Jason DeVito Oct 15 '15 at 21:55
  • 2
    $\begingroup$ Could you clarify exactly what in Ted's answer (and the comments following it) that you didn't understand? $\endgroup$ – Jason DeVito Oct 22 '15 at 20:58
  • 2
    $\begingroup$ Yes, they are an orthonormal basis of $\mathbb{R}^3$. In addition, there are known formulas relating $t',n'$ and $b'$ to $t,n,b$ (called the Frenet-Serre formulas.) $\endgroup$ – Jason DeVito Oct 22 '15 at 21:12

I think this is a lot easier than what you think: Take (1) above: $$\frac{\tau }{\kappa }=\frac{d}{ds}\left (\frac{\dot \kappa}{\tau \kappa^2}\right ) \tag 1$$

Rewrite this in terms of $\sigma$ and $\rho$ $$\frac{\rho}{\sigma }=\frac{d}{ds}\left (\sigma (-\dot \rho)\right ) = -\frac{d}{ds}\left (\dot{\rho} \sigma \right ) \tag 2$$

Now you're almost done.

Take the derivative of the LHS of your expression:

$$\frac{d}{ds}\left(\rho^2 + (\dot{\rho} \sigma)^2 \right) = 2\rho \dot{\rho} + 2(\dot{\rho} \sigma)\frac{d}{ds}(\dot{\rho} \sigma)$$ But we just figured out what that last term is in terms of $\rho$ so substitute: $$\frac{d}{ds}\left(\rho^2 + (\dot{\rho} \sigma)^2 \right) = 2\rho \dot{\rho} + 2(\dot{\rho} \sigma)\left(\frac{-\rho}{\sigma}\right)$$ You can now see that $$\frac{d}{ds}\left(\rho^2 + (\dot{\rho} \sigma)^2 \right) = 0$$

Or in other words $\rho^2 + (\dot{\rho} \sigma)^2$ is a constant (it must be positive, do you see why?). And if you call that constant $r^2$ you are done.

Put $a = \rho N + \sigma \dot{\rho} B$

We have just shown that this vector has a constant length of $r$. You can use the Frenet equations to show that $\dot{a} = -T = -\dot{\gamma}$

Then it is clear that $\gamma + a$ is a constant vector (the center of the circle) and the path $\gamma$ is at a constant distance $r$ from it.

For Viviani's curve put $$ \gamma(t) = \left(\frac{r}{2}(1 + \cos t), \frac{r}{2} \sin(t), r \sin(t/2)\right)$$ Which you can get from Wikipedia or Mathworld (remember we have a constant $r$ for our radius). then $$ \gamma'(t) = \frac{r}{2}\left(-\sin (t), \cos(t), \cos(t/2)\right)$$ $$ \gamma''(t) = -\frac{r}{2}\left(\cos (t), \sin(t), \sin(t/2)/2\right)$$ This is different from what you have. You need to calculate this out to finish.

For example I get $||\gamma'' \times \gamma'|| = \frac{r^2}{8\sqrt{2}}\sqrt{13 + \cos (t)}$

Can you take it from there?

  • 2
    $\begingroup$ OK sorry for the confusion it's been a few years since I did this stuff. I think I got it all correct now. Construct the vector $a$ and see that (according to Frenet) it's derivative is $-T$. This should be all you need for the last bit. $\endgroup$ – amcalde Oct 23 '15 at 17:42
  • 2
    $\begingroup$ Because it works! We just saw that this vector has a constant norm. You use that particular combination of $N$ and $B$ to make the Frenet equations work out the right way so that $\dot{a} = -T$. Then everything works out nicely! $\endgroup$ – amcalde Oct 23 '15 at 18:34
  • 2
    $\begingroup$ That is what you conclude. $\gamma$ is a constant distance away from a constant vector in space. That is what it means to be on the surface of a sphere. $\endgroup$ – amcalde Oct 24 '15 at 10:44
  • 2
    $\begingroup$ The center of the circle is fixed in space (that's $\gamma + a$ for us). It doesn't change. The path $\gamma$ is always at a fixed distance from this center. This means that $|a|$ is constant (and equal to $r$) which we have already shown. This is all you need to show. Your path is at a fixed distance from some fixed point in space. QED. $\endgroup$ – amcalde Oct 24 '15 at 11:11
  • 1
    $\begingroup$ @MaryStar I'll take a look and add to the post. $\endgroup$ – amcalde Oct 26 '15 at 18:50

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.