2D puzzles, 3D puzzles, and Dehn’s Invariant
One of the earliest puzzle types and most common antique puzzle types I run across aside from wire link puzzles is the 2D geometric arrangement puzzles, closely related to puzzles which can form multiple possible shapes. In my collection, aside from Tangrams, I have the checkerboard puzzle:
A classic example I don’t own is the “square the triangle” puzzle, which is available here in a nice package, although it can also be found or made much cheaper. https://www.puzzlemaster.ca/browse/inventors/petergal/8537-puzzle-booklet-square-to-triangle
This version of the puzzle begs the abstract math question “When can two shapes be made from the same set of pieces?” One constraint you will quickly find is that both shapes must have the same area, but is this enough? It turns out that in 2D, (if you constrain yourself to polygons with finite numbers of sides), it is enough. This is solved by first choosing a canonical shape for each area, typically a 1xA rectangle, and developing an algorithm to cut an arbitrary shape with the same area into some number of pieces which can be re-assembled into the standard rectangle. Since the algorithm is clearly invertible, this solves the problem. (For brevity and so that you can try to puzzle it I will leave the algorithm out, although I will link to one source at the bottom, and if there is interest, I can make more posts about this).
This immediately begs two more generalizations. First, you could add more shapes that have to be made by the same pieces, or second you could try the same thing in 3 dimensions, with solid shapes. The first problem is a nice exercise after understanding the above algorithm. The second is much more interesting.
Many puzzles and related topics can be found in the great book by Stewart Coffin (who incidentally graduated from UMass Amherst for electrical engineering) “The Puzzling World of Polyhedral Disections”, available graciously online at https://johnrausch.com/PuzzlingWorld/
The natural guess you would have is that the only constraint is volume. But this turns out not to be true. It turns out we need a second invariant, known as “Dehn’s Invariant” https://en.wikipedia.org/wiki/Dehn_invariant. To formalize this requires an understanding of something called a tensor, (https://en.wikipedia.org/wiki/Tensor_product ) so I will only sketch it here, although at some point I think it would be nice to see if I can write it up more precisely without requiring the use of tensors.
Leaving out details for an easier understanding, a tensor for our purposes will be an abstract object made from a pair of real numbers, represented by (a,b). (In reality, tensors are typically represented by the tensor product, which is an X surrounded by a circle, but to make things easier I’ll leave it at this). This is not the whole story, because two pairs (a,b) and (c,d) represent the same tensor if there is an integer n such that c = a/n and d = b*n. In other words, (f*n, g) is the same tensor as (f, g*n). In our case, we will have another rule like this regarding the second number, which is that the second number will be modulo 2*PI, so that (a,b) is the same tensor as (a,b+2PI).
The tensor object also has addition: (a,b) + (c,d) = (a+c, b+d).
The Dehn invariant of a polyhedron can now be defined. Given a polyhedron, each edge has a length, and it has an angle (the dihedral angle, which is the largest angle that can be formed by taking a 2D slice and measuring the angle, or alternatively is the angle when the edge is sliced perpendicular to the edge). The Dehn invariant is then the sum over all the edges of the tensors (l,a) corresponding to each edge, where l is the length and a is the dihedral angle. (If there is more than one piece of the polygon, sum the Dehn invariants of all the pieces) The Dehn invariant of a cube or any other parallelapiped is zero, but the dehn invariants of the Platonic solids are all different from each other. So cubes and paralelapipeds sharing volumes could create such a puzzle, but no pair of different Platonic solids could.
It is a fun exercise to prove to yourself that the Dehn invariant won’t change if you cut a shape into two pieces using a plane slice, or if you glue two pieces together by gluing a face from one to the other. (Usually with the same shape and size of the two faces, but I don’t think it has to be).
To sum up, two 3D polyhedrons can both be formed from some set of smaller polyhedrons if and only if they share the same volume and they share the same Dehn invariant.
A nice numberphile video https://youtu.be/eYfpSAxGakI explains the above in more detail, although I’m happy to go over any of it if anyone is interested.











Comments
Post a Comment