Hey all I am looking for help on a problem. I will post it, and than I will add what I have tried and my ideas etc. The question has been up now for a few days, I'm sure someone out there can help! I even put a bounty, I have spent a lot of time on this question!

I am interested in calculating the torsion ($\tau$) and curvature ($\kappa$) of the curve $$X(t)=(at,bt^2,ct^3), \quad t \ge 0 $$ and $a$, $b$, and $c$ are all positive constants.

So here is what I am having problems with. It seems like there are so many different formulas for curvature, and there are also the Frenet–Serret formulas so I am having issues deciding how to do it. I was thinking maybe I could reparametrize with respect to arc length, which would give me it in terms of unit length so I could use some of Frenet–Serret formulas, but I am not confident in that.

What I did so far was I calculated $X'(t)=(a,2bt,3ct^2)$ and $|X'(t)|=\sqrt{a^2+4b^2 t^2 +9c^2 t^4}$.

Then I calculated $$X''(t)=(0, 2b, 6ct)$$ and $$|X''(t)|=2\sqrt{b^{2}+9c^{2}t^{2}}.$$

I also know that the unit tangent, $$T(t)={X'(t)\over|X'(t)|}$$ and the unit normal is $$N(t)={T'(t) \over |T'(t)|}$$ and that the binormal is $B= T \times N$.

But I am really not sure how to take it further.

I know that in arc length parametrzation we would have $dT/ds= \kappa N$ and $dN/ds=-\kappa T + \tau B$ and $dB/ds= -\tau N$.

Should I keep it in the form it is and use the equation $$\kappa={|X'(t) \times X''(t)|\over |X'(t)|^3}?$$

A few of the other things I am thinking of is that maybe I could solve for torsion by the following:

I know that $T$, $B$ and $N$ form an orthonormal basis of $\mathbb R^{3}$ so can and when we write $N'=\alpha T + \tau B$ , is the coefficient, $\tau$ the definition of torsion?

Moreover, noting $B= T \times N$, we have $$B'= T'\times N + T \times N' = T \times N' =T \times (\alpha T+\tau B) = \tau T \times B= -\tau N $$ (because $N=B \times T$ so $T \times B=-N)$. I think this is how I understand the derivation for that equation.

Is it only in arc length parametrization that I can use the Frenet equations for example?

However, I think I should be able to do it just using the regular formulas, and not an arc length parametrization as the integral would be tough. I did conform this as well, so in terms of the basics of this question I'd like to do it without arc length parametrization.

I am also interested in seeing a intuitive derivation of the torsion formula (the one with the triple product).

I apologize if I didn't show enough work, this is all I could do but I am very happy to learn it. Thanks a lot to any help!


I am wondering about the validity of what I now have.

In addition to above I computed $X'''(t)=(0,0,6c)$

$X' \times X''= (6bct^2,-6act,2ab)$

$(X' \times X'') \cdot X''' = (12abc)$

$|X' \times X''| = 2 \sqrt {9b^2c^2t^4+9a^2c^2t^2+a^2b^2}$.

Now can I just apply the formulas

$$\tau = \frac{ (X' \times X'') \cdot X'''}{|X' \times X''|^2}.$$

  • $\begingroup$ The easiest way to calculate the curvature $\kappa$ is with $(dT/ds\cdot dT/ds)=\kappa^2(N\cdot N)=\kappa^2$, as has been done in my answer below (with IMO more common notation). The easiest way to calculate the torsion $\tau$ is with $\det(T,dT/ds,d^2T/ds^2)=\kappa^2\tau$, as has been done in the same answer. Are you familiar with the latter formula? $\endgroup$ – Han de Bruijn Mar 14 '15 at 20:07
  • $\begingroup$ Im not familiar with the latter formula, but would be open to learning it. And sorry I am a little confused on the notation, of why you have (dT/ds $\dot$ dT/ds)=$\kappa(N \dot N) $? $\endgroup$ – Quality Mar 14 '15 at 20:09
  • $\begingroup$ Yes, it is only in arc length parametrization that you can use the Frenet equations. But no, you don't need an integral to calculate the arc length, because it's sufficient to know that $dX/ds = (dX/dt)/(ds/dt)$ as has been done in my answer. $\endgroup$ – Han de Bruijn Mar 14 '15 at 20:14
  • $\begingroup$ I've been trying to format an inner product. $\endgroup$ – Han de Bruijn Mar 14 '15 at 20:16
  • $\begingroup$ The formula sequence in your update seems to be a more efficient method to calculate $\tau$ than the way I did it (with the $\det$ formula). $\endgroup$ – Han de Bruijn Mar 15 '15 at 12:10

Those calculations with space curves tend to be tedious; therefore using a computer algebra system is not quite a luxury. Here goes: $$ \vec{r}(t) = \left[ \begin{matrix} a t \\ b t^2 \\ c t^3 \end{matrix} \right] \quad \Longrightarrow \\ \vec{\iota} = \, \stackrel{.}{\vec{r}} \, = \frac{d\vec{r}}{ds} = \frac{d\vec{r}}{dt}\frac{dt}{ds} = \left[ \begin{matrix} a \\ 2 b t \\ 3 c t^2 \end{matrix} \right] / \sqrt{a^2 + 4 b^2 t^2 + 9 c^2 t^4} \quad \Longrightarrow \\ \stackrel{..}{\vec{r}} \, = \frac{d^2 \vec{r}}{ds^2} = \, \stackrel{.}{\vec{\iota}} \, = \frac{d\vec{\iota}}{ds} = \frac{d\vec{\iota}}{dt}\frac{dt}{ds} = \left[ \begin{matrix} -2at(2b^2+9c^2t^2)\\2b(a^2-9c^2t^4)\\6ct(a^2+2b^2t^2) \end{matrix} \right]/(a^2+4b^2t^2+9c^2t^4)^2 \quad \Longrightarrow \\ \stackrel{...}{\vec{r}} \, = \, \stackrel{..}{\vec{\iota}} \, = \frac{d\stackrel{.}{\vec{\iota}}}{dt} \frac{dt}{ds} = \left[ \begin{matrix} 2a(2a^2b^2-24t^2b^4-162t^4b^2c^2+27a^2c^2t^2-405t^6c^4) \\ -8bt(27a^2c^2t^2-81t^6c^4+4a^2b^2) \\ 6c(a^4-6a^2b^2t^2-63a^2c^2t^4-8b^4t^4-90b^2t^6c^2) \end{matrix} \right] / (a^2+4b^2t^2+9c^2t^4)^{7/2} $$ With $\rho^2 = (\stackrel{.}{\vec{\iota}} \cdot \stackrel{.}{\vec{\iota}})$ - inner product - we find for the curvature $\rho$ : $$ \rho(t) = 2\sqrt{\frac{a^2b^2+9a^2c^2t^2+9t^4b^2c^2}{(a^2+4b^2t^2+9c^2t^4)^3}} $$ With the (hopefully) well-known formula $\;\det\left(\stackrel{.}{\vec{r}},\stackrel{..}{\vec{r}},\stackrel{...}{\vec{r}}\right) = \rho^2 \tau \;$ we find herefrom for the torsion $\tau$ : $$ \tau(t) = \frac{3 a b c}{a^2b^2+9a^2c^2t^2+9t^4b^2c^2} $$

  • $\begingroup$ I am a bit confused on why for example r' you include it over the sort and not just have (a 2bt, 3ct^2)? $\endgroup$ – Quality Mar 14 '15 at 20:17
  • $\begingroup$ @LearningMath: The vector $r$ with the dot above it is the derivative to the arc length $s$. That is the derivative to $t$ divided by the derivative of the arc length $=ds/dt$ or multiplied by $dt/ds$. $\endgroup$ – Han de Bruijn Mar 14 '15 at 20:22
  • $\begingroup$ Thanks, I think I am a little confused about deriving with respect to something else etc. Can the question just be solved by using X' X'', and X''' as a regular parameter, and applying the formulas $\endgroup$ – Quality Mar 14 '15 at 20:23
  • $\begingroup$ @LearningMath: If I understand what you mean, then I think the answer is NO: you have to differentiate to the arc length in the end. $\endgroup$ – Han de Bruijn Mar 14 '15 at 20:29
  • $\begingroup$ @LearningMath: Formula for the torsion of a regular curve paraemeterized by arc length $\endgroup$ – Han de Bruijn Mar 14 '15 at 20:43

Let curvature $\kappa$ be defined by $\displaystyle \kappa = \frac{\big \|\mathbf T'(t) \big \|}{\big \| \mathbf r'(t)\big \|} $.

Since $\mathbf r(t) = (at, bt^2, ct^3)$,

$\mathbf r'(t)= (a, 2bt, 2ct^2)$ ,

$\|\mathbf r'(t) \| = \sqrt{a^2+4b^2t^2+9c^2t^4}$,

so $\mathbf T(t)= \frac{(a, 2bt, 2ct^2)}{\sqrt{a^2+4b^2t^2+9c^2t^4}}$

and $\mathbf T'(t)= \frac{(0, 2b, 4ct)}{\sqrt{a^2+4b^2t^2+9c^2t^4}}$.

and $\| \mathbf T'(t) \|= \frac{\sqrt{4b^2+16c^2t^2}}{\sqrt{a^2+4b^2t^2+9c^2t^4}}$

Now $\displaystyle \kappa = \frac{\big \|\mathbf T'(t) \big \|}{\big \| \mathbf r'(t)\big \|} = \frac{\sqrt{4b^2+16c^2t^2}}{a^2+4b^2t^2+9c^2t^4}$

For torsion do a similar thing but use the formula:

$\tau = \frac{\big | \mathbf r'(t) \ \mathbf r''(t) \ \mathbf r'''(t) \big| }{\kappa^2} $

where the numerator is the scalar triple product.


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.