Why is the radius of curvature = 1/(curvature)?

6.4k Views Asked by At

My textbook Thomas' Calculus (14th edition) initially defines curvature as the magnitude of change of direction of tangent with respect to the arc length of the curve (|dT/ds|, where T is the tangent vector and s is the arc length) and later by intuition conclude that κ = 1/ρ (where, κ=curvature,ρ = radius).

Is there a way to prove this equation mathematically ?

Edit: If curvature was defined as the inverse of radius of curvature, then how does the textbook define and derive an expression of curvature first and then provide a relationship between it and the radius of curvature.

1

There are 1 best solutions below

4
On

Without much calculation (but also without dealing with exceptional cases, such as the tangent being vertical, or the second derivative being zero):

By the definition given at the start of the question, and by the chain rule, the curvature of $y = f(x)$ at $x$ is: $$\left\lvert\frac{(d/dx)\tan^{-1}f'(x)}{ds/dx}\right\rvert,$$ where $s$ is arc length. Since $ds/dx = \sqrt{1 + f'(x)^2}$, the curvature depends only on $f'(x)$ and $f''(x)$.

Therefore, if functions $f, g$ have second order contact at $x$, i.e. $f(x) = g(x)$, $f'(x) = g'(x)$, and $f''(x) = g''(x)$, then $f, g$ will have the same curvature at $x$.

If the graph of $g$ is a circle, then its curvature is the same as the magnitude of the rate of change of the direction of the radius with respect to arc length (because the tangent at a point is perpendicular to the radius to that point). But if the radius has length $r$, and makes an angle $\theta$ with a fixed line, then the length of the arc from the point of intersection of the circle with that line is $r\theta$, therefore the curvature is $1/r$.

It is enough, then, to prove that if $f''(x) \ne 0$, there exists a circle $g$ having second order contact with $f$ at $x$.

Proceeding similarly to Hardy, A Course of Pure Mathematics (10th ed. 1952, p. 299), but leaving out the explicit formulae, which we don't need (although they're simple enough, and the expression for $r$ gives the same result as the defining expression in the second paragraph above):

Writing $y = f(x)$, $\dot{y} = f'(x)$, $\ddot{y} = f''(x)$, so that also $y = g(x)$, $\dot{y} = g'(x)$, $\ddot{y} = g''(x)$ (with a suitable function $g$ to be found), we require a circle centre $(p, q)$, radius $r$, such that: \begin{align*} (x - p)^2 + (y - q)^2 & = r^2, \\ (x - p) + (y - q)\dot{y} & = 0, \\ 1 + \dot{y}^2 + (y - q)\ddot{y} & = 0. \end{align*} Because $\ddot{y} \ne 0$, the third equation can be solved for $q$, then the second equation can be solved for $p$, and finally the first equation can be solved for $r$. So the required circle exists. $\square$