Or search by topic
Published 2007 Revised 2022
How curved is a curve? How curved is a surface? When is a 'curved surface' flat? We shall only briefly mention curves in the plane and then move on to discuss positive and negative curvature of surfaces.
Imagine tracing out the ellipse shown in the diagram. How does the curvature change as you go around the ellipse? Without applying any mathematics everyone would agree that the tightest bends are at the ends and the least curvature on the track around the ellipse is halfway between these points .
To measure the curvature at a point you have to find the circle of best fit at that point. This is called the osculating (kissing) circle. The curvature of the curve at that point is defined to be the reciprocal of the radius of the osculating circle. We shall not discuss in this article the method for finding this radius accurately which needs calculus. Suffice it to say that this circle has not
only the same tangent at the point (the first derivatives being the same) but also the curve and the osculating circle have the same second derivatives at the point.
Why use the reciprocal in defining curvature? It is natural for the curvature of a straight line to be zero. Imagine straightening out a curve making it into a straight line. In the limit the circle of best fit has infinite radius giving zero curvature.
The following equations are given in case you want to plot the graphs yourself, but ignore them if not and read on.
The equation of this ellipse is {x^2\over 36} + {y^2\over 9}=1.
The area of the strip cut off by planes A and B on the surface of the sphere is 2\pi xdc because this strip has circumference 2\pi x and height dc.\par The area of the band on the cylinder is 2\pi rdy. As we have shown that xdc=rdy it follows that these areas are equal.
Now we study the curvature of the surface of a sphere of unit radius. Take any point N and draw a circle centre N radius r on the surface of the sphere where the radius is an arc in the surface of the sphere subtending an angle of r radians at the centre of the sphere.
The circumference of the circle centre N radius r is 2\pi \sin r. Because \sin r < r for all non-zero values of r we see that this circumference 2\pi \sin r is less that 2\pi r showing that the surface has positive curvature..
By Archimedes' method the area of this circle is 2\pi r(1 -\cos r). You can verify that this is less than \pi r^2, and hence that the curvature of the surface of the sphere is positive, by drawing the graphs of y=\cos x and y = 1 - {x^2\over 2} on the same axes and showing that \cos r > 1 - {r^2\over 2}.
Tie a long rope to a post at this point and, keeping the rope taut, walk around a circle with the post as centre. As you have to walk uphill and downhill as well as around the post you walk farther than you would walk on flat ground showing that the curvature at that point is negative.
At the red spot on the cube the angle deficiency is 90^o, the circumference of a circle of radius r with centre at the red spot on the cube is {3\over 4}\times 2\pi r so the curvature at that point is positive. The curvature is concentrated at the vertices of the cube rather than being distributed over the surface as with a sphere or a rugby ball shaped solid. Although the curvature is
concentrated at 16 points, the block shown with a hole through it is analagous to the torus (or doughnut shaped solid) shown in yellow.
We investigate the curvature at the red spot at the edge of the hole in the block. The angle deficiency at the red spot is -90^o. The circumference of a circle of radius r with centre at the red spot on the block is {5\over 4}\times 2\pi r so the curvature at that point is negative.
To read more on the subject of Gaussian curvature without getting deep into higher mathematics see this article and related articles from a workshp on 'Geometry and The Imagination' led by John Conway, Peter Doyle, Jane Gilman and Bill Thurston.