$ \kappa = \cos s, \tau = \sin s $ and passing through (1,0,0), TNB triad identity matrix?

previous link

When numerically computed it looks like a catenoid surface of revolution for all appearances. ( unit curvature at extremes and like an symptote at symmetrical plane). Its parametrization is not mentioned in Differential Geometry books that I could come across.


  1. Can it at least be proved that the line lies on a surface of revolution?

  2. In general for a surface of revolution, what functional relation/s can exist between $ \kappa, \tau $ ?


Initial values for (T,N,B) taken as an identity matrix to fix rotation.


The lines given by achille hui are described on an equiangular rotational hyperboloid inclined at $\pi/4$ to one axis, result of initial point $(0,0,0)$ and starting $T,N,B$ an identity matrix.

May I request someone to extend it further or integrate the locus incorporating constants for generality ? like

$$ (a \kappa, a \tau) = ( \cos s/b , \sin s/b) $$

An image of a hyperboloid of revolution for $ (a= 1/6, b =2 ) $:

enter image description here

Motivation of this post was in fact to readily plot these lines on a surface of revolution using scalar curvatures properly.


In hindsight it slightly puzzles (me)... Although we started off with two intrinsic curvature invariants in 3-space, when combined it forces the arc to belong to 1- sheet hyperboloid $ of \, revolution $ in $ \mathbb R^2 $ that was neither specified in the input nor expected (by me). Is there some explanation or insight for it to so happen?

  • 1
    $\begingroup$ What do you mean? There are uncountably many such space curves with different functions $\kappa$, $\tau$ satisfying the equation $\kappa^2 + \tau^2 =1$ $\endgroup$ – Robert Israel Jan 7 '16 at 21:58

Instead of a curve passing through $(1,0,0)$, choose the coordinate system such that at time $t = 0$, the curve start at $(0,0,0)$ with $\vec{T}, \vec{N}, \vec{B}$ in the direction of $x, y, z$ respectively.

It turns out for $\kappa(s) = \cos(s), \tau(s) = \sin(s)$, one can convert the ODE for Fernet-Serret frame to an ODE over quaternion and solve it explicitly. The derivation is pretty messy and I won't reproduce it here. Feel free to throw the expression of space curve below to a CAS and verify it works.

In any event, the space curve $\vec{X}(s) = (x(s),y(s),z(s))$ is given by the formula:

$$\begin{cases} x(s) &= \frac{3}{\sqrt{2}}\sin(\sqrt{2}s)\cos(s) - 2\cos(\sqrt{2}s)\sin(s)\\ y(s) &= -\frac12(3\cos(\sqrt{2}s) + 1)\cos(s) - \sqrt{2}\sin(\sqrt{2}s)\sin(s) + 2\\ z(s) &= +\frac12(3\cos(\sqrt{2}s)-1)\cos(s) + \sqrt{2}\sin(\sqrt{2}s)\sin(s) - 1 \end{cases} $$

With a little bit of algebra (or an CAS), one can verify this curve lies on a one-sheet hyperboloid:

$$x^2 - 2(y-2)(z+1) = 4$$


For the general case $\kappa(s) = \rho\cos(\alpha s), \tau(s) = \rho\sin(\alpha s)$ where $\rho, \alpha$ are constants, if I didn't make any mistake, the space curve is given by

$$\begin{cases} x(s) &= \frac{\rho^2}{2\beta} S(s)\\ y(s) &= y_0 -\frac{\rho^2}{2\beta^2}\left( \frac{2}{\rho}\cos(\alpha s) + \rho C(s)\right)\\ z(s) &= z_0 -\frac{\rho^2}{2\beta^2}\left( \frac{2}{\alpha}\cos(\alpha s) - \alpha C(s)\right) \end{cases} $$ where $\beta = \sqrt{\alpha^2+\rho^2}$, $y_0 = \frac{2}{\rho}$, $z_0 = \frac{\rho^2-2\alpha^2}{a\rho^2}$ and $\displaystyle\; \begin{cases} C(s) &= \frac{\cos((\beta-\alpha)s)}{(\beta-\alpha)^2} + \frac{\cos((\beta+\alpha)s)}{(\beta+\alpha)^2}\\ S(s) &= \frac{\sin((\beta-\alpha)s)}{(\beta-\alpha)^2} + \frac{\sin((\beta+\alpha)s)}{(\beta+\alpha)^2}\\ \end{cases} $.

Furthermore, the space curve lies on the one-sheet hyperboloid:

$$\rho^2\beta^2 x^2 + \rho^2(\alpha(z-z_0) - \rho(y-y_0))^2 - \alpha^2(\rho(z-z_0) + \alpha(y-y_0))^2 = \frac{4\alpha^2\beta^2}{\rho^2}$$

  • $\begingroup$ In the equation for $y(s)$, should it be $3\cos(\sqrt2s)$ ? $\endgroup$ – bubba Jan 9 '16 at 2:58
  • $\begingroup$ @bubba, fixed. it should be $s$. $\endgroup$ – achille hui Jan 9 '16 at 3:26
  • $\begingroup$ Nice and important result this : $ \kappa(s) = \rho\cos(\alpha s), \tau(s) = \rho\sin(\alpha s) $..a one sheet hyperboloid of revolution. May be also connected with hyperbolic geometry. $\endgroup$ – Narasimham Jan 10 '16 at 10:05
  • $\begingroup$ Can $ \kappa(s) = \rho\sin(\alpha s), \tau(s) = \rho\cos(\alpha s) $ lead to a similar result? $\endgroup$ – Narasimham Feb 12 '16 at 18:15
  • $\begingroup$ @achillehui math.stackexchange.com/questions/1721396/… Do you think I am right about relevance of Vishesh's question here? $\endgroup$ – Narasimham Mar 31 '16 at 11:20

Since curvature and torsion are invariant under Euclidean transformations, you can take any curve with given $\kappa(s), \tau(s)$ and revolve around any line, obtaining a surface of revolution made up of curves with the same $\kappa(s)$ and $\tau(s)$. But given an initial point on the curve where, to keep things simple, $\kappa$ and $\tau$ are nonzero, the initial $(T,N,B)$ could point in any three orthogonal directions, so the curves don't all lie on a surface, they fill up a region of space.

  • $\begingroup$ Yes, forgot mentioning the assumed initial $(T,N,B) $ identity matrix also, as the fundamental theorem freezes the rigid/intrinsic equation defined object with the two Euclidean motions arrested. I like to look at the simplest/elegant parametrization in closed form, am really indifferent to any convenient initial conditions that may be assumed. $\endgroup$ – Narasimham Jan 8 '16 at 21:15

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.