Every so often I see spherical geometry/trigonometry questions on this site, and this has piqued my curiosity about geometric constructions on the sphere.

Constructions for the equilateral triangle, square and regular pentagon on a sphere are long known, even though for the square and pentagon you can't use the diagonal/side ratio from planar geometry. Instead we can get the square from the diagonals being perpendicular bisectors of each other, and the pentagon yields to the cube/dodecahedron relationship (The face diagonal of a regular dodecahedron are also the edges of inscribed cubes).

But what about the higher Fermat prime sided regular polygons, for which such relationships are not evident? How, if at all, can these polygons be constructed directly on a sphere?

  • $\begingroup$ Are you interested in constructing a regular polygon of a particular size on a sphere, or of any size? When you mention dodecahedron, you will only get a pentagon of a particular size. And what kind of construction are you looking for? Compass and "geodesic"-edge? $\endgroup$ – robjohn Jun 17 '17 at 23:59
  • $\begingroup$ You can then construct concentric pentagons of any other size. The central angles of the dodecahedral face provides the necessary template. Likewise, a regular $n$-gon of one size gives central angles for corresponding polygons of any size. One size implies all others. $\endgroup$ – Oscar Lanzi Jun 18 '17 at 0:05
  • $\begingroup$ Compass and geodesic edges, yes. Think of the straightedge as a string you pull taut and it makes great circular arcs on the sphere. $\endgroup$ – Oscar Lanzi Jun 18 '17 at 0:17
  • $\begingroup$ I swear there was a recent article in an MAA journal about constructions on the sphere; I wrote to the authors. Can't find it now. What is shown is just the minimum, that we can construct all lengths $x$ on the unit sphere such that $\sin x \in E,$ where $E$ is the smallest field containing the rationals that is closed under taking square roots of positive elements. Going for that to th 17-gon is still a big leap, but it is a proof of existence. $\endgroup$ – Will Jagy Jun 18 '17 at 0:26
  • 1
    $\begingroup$ until I find that, here is the hyperbolic plane zakuski.utsa.edu/~jagy/papers/Intelligencer_1995.pdf $\endgroup$ – Will Jagy Jun 18 '17 at 0:39

The quickest description is to say that the constructible angles on the sphere are exactly the same as the constructible angles in the plane.

There are a few oddities. It is not possible to trisect a segment on the sphere, as this would involve trisecting an angle; think of the segment as part of the equator, and the north pole as part of a triangle with that.

It is possible to square the circle on the unit sphere. Not impressive, really. A hemisphere is both a square and a circle. There are a countable set of pairs of circular caps and symmetric quadrilaterals of the same area, where both are constructible.

Anyway, i wrote my article about the hyperbolic plane, not the unit sphere, but this material matches up.


April 2016, me to E. Rykken

Dear Prof. Rykken, I have just read your article with Myles Dworkin on constructions on the sphere, and I have some suggestions. A little late in the day, of course. In preparing my Intelligencer article on constructions in the hyperbolic plane, I did not actually write out the analogous statements on the unit sphere. However, it seemed to me very likely that the angles that can be constructed on the sphere are exactly the same as the angles that can be constructed in the plane. As to your final paragraph, I would expect there to be a countable set of pairs of circles and "squares" on the unit sphere, both constructable, with equal area. However, i would expect these both need be constructed at the same time, that there is no construction to begin with one and reach the other. Anyway, here are a copy of my old article and a page with Marvin's later article that incorporates some of it.


| cite | improve this answer | |
  • $\begingroup$ But how do you do it? If you try to use Ptolemy's construction of the regular pentagon on a sphere it gets distorted; if you try e.g. Erchinger's 17-gon construction on a sphere it gets distorted too. We have a ready workaround for the pentagon but what about 17 or higher Fermat primes? $\endgroup$ – Oscar Lanzi Jun 18 '17 at 0:11

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.