-5

Let $\gamma$ be a parameterization by arc length of a curve $C$. Prove that,

if there are: $X_0 \in \mathbb{R}^3$, and $\alpha,\beta \in [a,b] \to \mathbb{R}$, differentiable on $[a, b]$, such that $\alpha(s)N(s) + \beta(s)B(s) +\gamma(s) = X_0$ for all $s \in [a,b]$, then $C$ lies on a sphere.

1 Answers1

0

Given that

$X_0 = \gamma(s) + \alpha(s)N(s) + \beta(s) B(s), \tag 1$

we have

$X_0 - \gamma(s) = \alpha(s)N(s) + \beta(s) B(s), \tag 2$

whence

$\Vert X_0 - \gamma(s) \Vert ^2 = (X_0 - \gamma(s)) \cdot X_0 - \gamma(s))$ $= (\alpha(s)N(s) + \beta(s) B(s)) \cdot (\alpha(s)N(s) + \beta(s) B(s)) = \alpha^2(s) + \beta^2(s), \tag 3$

since

$N(s) \cdot N(s) = 1 = B(s) \cdot B(s) \tag 4$

and

$N(s) \cdot B(s) = 0. \tag 5$

We next differentiate (1) with respect to $s$, observing that $X_0 \in \Bbb R^3$ is fixed:

$0 = \dfrac{d}{ds}(X_0)$ $= \dot{\gamma}(s) + \dot{\alpha}(s)N(s) + \alpha(s) \dot N(s) + \dot{\beta}(s) B(s) + \beta(s) \dot B(s), \tag 6$

and recalling that

$\dot \gamma(s) = T(s), \tag 7$

and the Frenet-Serret equations

$\dot N(s) = -\kappa(s) T(s) + \tau(s) B(s), \tag 8$

$\dot B(s) = -\tau(s) N(s), \tag 9$

we transform (6) into

$0 = \dfrac{d}{ds}(X_0)$ $= T(s) + \dot{\alpha}(s)N(s) + \alpha(s) (-\kappa(s) T(s) + \tau(s) B(s)) + \dot{\beta}(s) B(s) - \beta(s) \tau(s) N(s)$ $ = T(s) + \dot{\alpha}(s)N(s) - \alpha(s) \kappa(s) T(s) + \alpha(s) \tau(s) B(s) + \dot{\beta}(s) B(s) - \beta(s) \tau(s) N(s)$ $= (1 - \alpha(s) \kappa(s)) T(s) + (\dot \alpha(s) - \beta(s) \tau(s))N(s) + (\dot \beta(s) + \alpha(s) \tau(s))B(s); \tag {10}$

by virtue of the orthogonality of $T(s)$, $N(s)$ and $B(s)$ we may infer that

$\alpha(s) \kappa(s) = 1, \tag{11}$

$\dot \alpha(s) = \beta(s) \tau(s), \tag{12}$

$\dot \beta(s) = -\alpha(s) \tau(s); \tag{13}$

thus,

$\alpha(s) \dot \alpha(a) + \beta(s) \dot \beta(s) = \alpha(s) \beta(s) \tau(s) - \beta(s) \alpha(s) \tau(s) = 0, \tag{14}$

which is easily seen to imply that

$\dfrac{d}{ds}(\alpha^2(s) + \beta^2(s)) = 0, \tag{15}$

and we conclude that $\alpha^2(s) + \beta^2(s)$ is constant; returning to (3), we see that this implies that $\gamma(s)$ lies in the sphere of radius $\sqrt{\alpha^2(s) + \beta^2(s)}$ centered at $X_0$.

A Few Further Remarks: First, we observe that (11) may be written

$\alpha(s) = \dfrac{1}{\kappa(s)}, \tag{16}$

which shows not only that

$\alpha(s) > 0, \tag{17}$

but also that $\alpha(s)$ is the radius curvature of $\gamma(s)$ at $s$. Second, (16) yields

$\dot \alpha(s) = -\dfrac{\dot \kappa(s)}{\kappa^2(s)}, \tag{18}$

and combining this with (12) we find

$\beta(s) \tau(s) = -\dfrac{\dot \kappa(s)}{\kappa^2(s)}, \tag{19}$

or, in the event $\tau(s) \ne 0$,

$\beta(s) = -\dfrac{\dot \kappa(s)}{\kappa^2(s) \tau(s)} = \dfrac{d}{ds} \left ( \dfrac{1}{\kappa(s)} \right )\left ( \dfrac{1}{\tau(s)} \right ). \tag{20}$

In light of (16) and (20) we see that (2) may be expressed in the form

$X_0 - \gamma(s) = \dfrac{1}{\kappa(s)} N(s) - \dfrac{\dot \kappa(s)}{\kappa^2(s) \tau(s)} B(s)$ $= \dfrac{1}{\kappa(s)} N(s) + \dfrac{d}{ds} \left ( \dfrac{1}{\kappa(s)} \right )\left ( \dfrac{1}{\tau(s)} \right )B(s), \tag{21}$

which in accord with (3) shows that the radius of the sphere, $\Vert \gamma(s) - X_0 \Vert$, satisfies

$\Vert \gamma(s) - X_0 \Vert^2 = \dfrac{1}{\kappa^2(s)} + \dfrac{\dot \kappa^2(s)}{\kappa^4(s) \tau^2(s)}. \tag{22}$

This formula has appeared elsewhere on this site, see for example this question.

From (20),

$\dot \beta(s) = -\dfrac{\ddot \kappa(s)(\kappa^2(s) \tau(s)) - \dot \kappa(s) (\kappa^2(s) \tau(s))'}{\kappa^4(s) \tau^2(s)}$ $= -\dfrac{\ddot \kappa(s)(\kappa^2(s) \tau(s)) - \dot \kappa(s) (2\kappa(s) \dot \kappa(s) \tau(s) + \kappa^2(s) \dot \tau(s))}{\kappa^4(s) \tau^2(s)}$ $= -\dfrac{\ddot \kappa(s)(\kappa^2(s) \tau(s)) - 2\kappa(s) \dot \kappa^2(s) \tau(s) - \kappa^2(s) \dot \kappa(s) \dot \tau(s)}{\kappa^4(s) \tau^2(s)}. \tag{23}$

We note that (11) implies that

$\alpha(s) > 0; \tag{24}$

if

$\beta(s) = 0, \tag{25}$

then (12) yields

$\dot \alpha(s) = 0, \tag{26}$

that is, $\alpha(s)$ is a positive real constant; in light of (25) we also have

$\dot \beta(s) = 0, \tag{27}$

and then via (13) we see that

$\tau(s) = 0, \tag{28}$

which is sufficient that $\gamma(s)$ lie in a plane (see my answer to this question, click here); now (2) becomes

$X_0 - \gamma(s) = \alpha(s)N(s), \tag {29}$

from which, in accord with (3),

$\Vert X_0 - \gamma(s) \Vert^2 = \alpha^2(s), \tag{30}$

or

$\Vert X_0 - \gamma(s) \Vert = \alpha(s) = \dfrac{1}{\kappa(s)}, \tag{31}$

whence since $\alpha(s)$ and $\kappa(s)$ are constant, indicates that $\gamma(s)$ lies in the sphere or radius $\alpha(s)$ centered at $X_0$; when combined with the fact that $\gamma(s)$ lies in a plane, allows the conclusion that $\gamma(s)$ in fact lies in a circle of radius $\alpha(s)$ centered at $X_0$.

We have seen above (ca. (28)) that $\gamma(s)$ lies in a plane;

Need to re-save in .txt file!

End of Remarks.

Robert Lewis
  • 72,871