15

At a workshop it was suggested that it likely remains an open problem whether or not there is a 3- or 2 -piece dissection of a square to an equilateral triangle. Can anyone confirm that this is unresolved?

Four-piece dissections are known, the most famous being Henry Dudeney's century-old gem:

                [Maple animation from this link.]

Joseph O'Rourke
  • 149,182
  • 34
  • 342
  • 933
  • 15
    2 is not possible because the side length of the triangle is more than the diagonal of the square. – Ken Fan Nov 05 '11 at 02:18
  • 6
    I just want to tell that applet to "hold still, dammit"! – Todd Trimble Nov 05 '11 at 13:07
  • 5
    @Todd: Added a stable image (in a different orientation). – Joseph O'Rourke Nov 05 '11 at 13:40
  • 3
    Wow, ask and ye shall receive! Thank you, Joseph! – Todd Trimble Nov 05 '11 at 14:30
  • 4
    A 3-piece dissection of the equilateral triangle would have to create 4 right angles to serve as corners of the square, and there are just a few ways this can be done. At first glance, none of them recombine as a square (though you can get a rectangle). It should be pretty easy to run through the options and rule them all out. – Anton Lukyanenko Nov 06 '11 at 18:34
  • 4
    I comment only to point out that the so-called Dudeney construction may not, in fact, be due to H. Dudeney. For more details, see G.N. Frederickson's "Hinged Dissections: Swinging and Twisting" (pp. 8-10). http://books.google.com/books?id=pW_0EisSP4IC&lpg=PR9&ots=9B2Po0jPx5&dq=dissecting%20%22equilateral%20triangle%20into%20a%20square%22&lr&pg=PA8#v=onepage&q&f=false – Benjamin Dickman Nov 28 '13 at 06:04
  • 1
    @AntonLukyanenko I don't follow your argument. If we dissect the square with corners (0,0), (0,1), (1,0), (1,1) using the curves $y=x(1-x)$ and $y=x(x-1)+1$, then we get three pieces and no right angles. – Timothy Chow Sep 27 '23 at 12:28

1 Answers1

4

In the paper, "Dissection with the Fewest Pieces is Hard, Even to Approximate" (arXiv, doi) by Bosboom et al., they write:

We have known for centuries how to dissect any polygon $P$ into any other polygon $Q$ of equal area, that is, how to cut $P$ into finitely many pieces and re-arrange the pieces to form $Q$. But we know relatively little about how many pieces are necessary. For example, it is unknown whether a square can be dissected into an equilateral triangle using fewer than four pieces. [emphasis added]

The authors seem to be knowledgeable about the subject, so I'd be inclined to trust their claim. As I mentioned in another MO answer, they point out that more generally, it is not known if "$k$ piece dissection," the problem of deciding whether two given polygons (say with rational vertices) admit a common dissection into at most $k$ pieces, is decidable. The trouble is that even if you restrict yourself to cuts that are piecewise-linear, there is no known a priori upper bound on how many "zigs and zags" a cut might have to take.

Timothy Chow
  • 78,129