Theorem Let $\gamma\colon [a,b]\rightarrow \mathbb{R}^2$ be a unit speed simple closed curve, with $\gamma'(a)=\gamma'(b)$ and $N$ is the inward-pointing normal. Then $$ \int_{a}^b \kappa_N(s)ds=2\pi. $$

Question: If a circle has curvature $\kappa$ and circumference $C$ then from above theorem, how does it follow that $\kappa C=2\pi$?

Ref: Riemannian Manifolds - John M. Lee, Theorem 1.6 and comments, p.4.

  • $\begingroup$ I edited your post to add the "differential geometry" and "plane curves" tags. Cheers! $\endgroup$ – Robert Lewis Jun 9 '15 at 18:09
  • 2
    $\begingroup$ If $\gamma$ is unit speed, then $t$ is the arc-length parametrization and $\kappa$ is constant on circle! $\endgroup$ – achille hui Jun 9 '15 at 18:13
  • $\begingroup$ Changed the variable to $s$ from $t$ as it is not "any" independent variable.. Also for the simplest case of the circle, $\kappa\ = 1/R$ $\endgroup$ – Narasimham Jun 9 '15 at 18:22

I started preparing this answer before Narasimham made his edits, wherein the original independent variable name $t$ was replaced by $s$. Since I was nearly done when his edits occurred, I chose to respect the original formulation in my answer; nevertheless, the relationship 'twixt $s$ and $t$ should be clear in what follows.

A unit-speed circle in $\Bbb R^2$, centered at $(x_0, y_0)$ and of radius $\rho > 0$, may be parametrically represented as

$\gamma(t) = (x_0 + \rho \cos (\dfrac{t}{\rho}), y_0 + \rho \sin (\dfrac{t}{\rho})); \tag{1}$

taking $a = 0$ and $b = 2\pi$, we see that

$\gamma(a) = \gamma(0) = \gamma(2\pi) = \gamma(b). \tag{2}$


$\gamma'(t) = (-\sin(\dfrac{t}{\rho}), \cos(\dfrac{t}{\rho})), \tag{3}$


$\dfrac{ds}{dt} = \Vert \gamma'(t) \Vert = \sqrt{\sin^2(\dfrac{t}{\rho}) + \cos^2(\dfrac{t}{\rho})} = 1, \tag{4}$

where $s$ is the arc-length along $\gamma(t)$; from (4),

$s = t + s_0, \tag{5}$

which shows that both $s$ and $t$ effectively measure distance traversed along $\gamma(t)$, i.e. that $\gamma(t)$ is in fact a unit speed curve; the offset $s_0$ simply indicates a possible difference in parametrized points on $\gamma(t)$ from which the measurement of distance is taken. (By paremetrized point I mean a point $\gamma(t')$ on the curve together with a specific parameter value $t'$; this usage to disambiguate situations in which the same $\gamma(t')$ is in fact the image under $\gamma$ of more than one value $t'$, as in the present case with $\gamma(t' + 2\pi) = \gamma(t')$ for all $t'$.) If we agree, as we will, that $s$ is measured from $\gamma(0)$ at $t = 0$, then $s_0 = 0$ and $s$ and $t$ agree:

$s = t. \tag{6}$

We see from (4) that $\gamma'(t)$ is in fact the unit tangent vector field along $\gamma(t)$; in the notation of Frenet-Serret, we write

$T(t) = \gamma'(t); \tag{7}$

since $t = s$ we have the Frenet-Serret equation

$\dfrac{dT}{ds} = \dfrac{dT}{dt} = \kappa(t) N(t), \tag{8}$

where $\kappa(t)$ is the curvature, and $N(t)$ is the unit normal vector, to $\gamma(t)$ at the parameter value $t$; that is, at the point $\gamma(t)$ itself. From (3), (7) and (8) we have

$\dfrac{dT}{dt} = \gamma''(t) = (-\dfrac{1}{\rho} \cos (\dfrac{t}{\rho}), -\dfrac{1}{\rho} \sin (\dfrac{t}{\rho})) = \dfrac{1}{\rho}(-\cos(\dfrac{t}{\rho}), -\sin (\dfrac{t}{\rho}))$ $= \dfrac{1}{\rho}N(t) = \kappa(t) N(t). \tag{9}$

Inspecting (9), it is clear that (i.)

$\kappa(t) = \dfrac{1}{\rho} \tag{10}$\

is a constant; and (ii.)

$N(t) = (-\cos(\dfrac{t}{\rho}), -\sin (\dfrac{t}{\rho})) = -(\cos(\dfrac{t}{\rho}), \sin (\dfrac{t}{\rho})) = \dfrac{1}{\rho}((x_0, y_0) - \gamma(t)), \tag{11}$

which shows that $N(t)$ is directed from $\gamma(t)$ towards $(x_0, y_0)$, the center of the circle; thus it is an inward-pointing normal vector. Last but not least, we observe that

$\gamma'(a) = \gamma'(0) = \gamma'(2\pi) = \gamma'(b); \tag{12}$

we see that the criterion of the stated theorem are all met by the circle $\gamma(t)$; furthermore, since $\kappa(t) = \kappa(s) = 1/\rho$ is constant, we have

$\int_a^b \kappa(t) dt = \int_a^b \kappa(s) ds = \kappa \int_a^b ds = \kappa C; \tag{13}$

then by the theorem, the integral

$\int_a^b \kappa(t) dt = \int_a^b \kappa(s) ds = 2\pi; \tag{14}$

thus it follows that

$\kappa C = 2\pi, \tag{15}$

as was requested.

Nota Bene: Establishing the formula (15) by calling on the much more general theorem given in the text of the problem seems a little bit like going after flies with a shotgun, since once we see that the curvature of the circle is a constant $\kappa = 1/\rho$, (15) follows immediately from the formula

$C = 2\pi \rho, \tag{16}$

with which I presume most readers are familiar; nevertheless, the relation (15) does provide perhaps the simplest possible case/application of the stated theorem, and so it is worthwhile to work it out; a collection of concrete, simple examples being in many ways the true grist which is ground by the mathematical mill. End of Note.


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.