*The following post was kindly contributed by *Stan Wagon*. Stan (Wikipedea) is famous for his books, papers, snow-sculptures, and square-wheels bicycles (see picture below) ! *

A round cake has icing on the top but not the bottom. Cut out a piece in the usual shape

(a sector of a circle with vertex at the center), remove it, turn it upside down, and replace

it in the cake to restore roundness. Do the same with the next piece; i.e., take a piece

with the same vertex angle, but rotated counterclockwise from the first one so that one

boundary edge coincides with a boundary edge of the first piece. Remove it, turn it

upside-down, and replace it. Keep doing this in a counterclockwise direction. The figure

shows the situation after two steps when the central angle is 90°. If θ is the central angle

of the pieces, let f (θ) be the number of steps needed so that, under repeated cutting-andflipping just described, all icing returns to the top of the cake, with f (θ) set to ∞ if this

never happens. For example, f(90°) = 8.

### Test your intuition: What is f(181°)?

Source info:

The original source of this problem was Problem 31.2.8.3 in the 1968 Moscow Math Olympiad [L, p. 90] in a slightly different form. The variant given here is in [Winkler].

[L] 60-odd years of Moscow Mathematical Olympiads, G. Leites editor, G. Halperin, A Tolpygo, P Grozman A Shapovalov, V Prasolov, A Fomenko

[W] Mathematical Mind-Benders by Peter Winkler, published by AK Peters, Wellesley, Mass., 2007.

(see here)

My first “intuitive” answer was none of the above – it was 720. The intuition being that, as in the f(90) case, in order for the icing to show up again the cuts will have to cover the entire cake an even, integral number of times (so that every 1 degree sub-slice is flipped an even number of times). This happens for the first time after the 720th iteration – the 180th iteration does not flip the sub-slices evenly, and the 360th iteration flips every part of the cake exactly 181 times, which is odd.

I almost certainly would have not realized this logic was flawed if it weren’t for the fact that 720 was not one of the available options! Realizing my error, I worked on the question more rigorously, and in the process all but one of the suggested answers seemed to me like very reasonable contenders for the correct answer — and of course, the one that seemed least reasonable during the process was the one that eventually turned out to be correct.

(Since this is about intuition rather than computation, though, my vote was for 360 as the nearest representative for 720.)

Very nice and surprising riddle!

Studying some base cases, the answer seems infinity. As the recurrence for total flipped angle after every 4 moves remains odd .

Setup: Let be relatively prime (we will rotate the circle by ). Let be the cyclic group, and let where acts by translation. Let be the vector with ones follows by zeroes, and let be any generator. We need to determine the order of .

Calculation: projecting to we see that the order of divides the order of which is . We next compute the its powers: , etc where is the action of on . It follows that . But for each standard basis vector we have where is the all-ones vector. It follows that (scalar multiplication in the vector space ) and .

Conclusion: The order of is if this is even and otherwise.

In particular for the order is .

Remark: Note that the argument did not require that the rotation be at the same angle as the size of the wedge that is being reversed.

A-ha: we don’t need the order of — we are considering the action of on and specifically want the first such that (rotating the cake as a whole doesn’t change the fact that it is the right way up). But this would make a divisor of which does not divide (since after 360 steps the cake is flipped).

But this means is a multiple of . So let’s compute small powers.

and is translation by if translates by . Since is -invariant we get that . But then flips the first bits so before step some bits are flipped and some aren’t, and the answer is still .

Can someone point out my mistake?

Your model of how the cake is affected by flipping a piece of cake is inaccurate. It is not enough to consider the rotation of the cake (by the cyclic group) and whether each 1-degree piece is up or down. When a 181-degree piece of cake is flipped, it affects the relative order of the 1-degree pieces, breaking the previous cyclic order. This has to be taken into consideration. The ambient group is actually the semi-direct product of $S_q$ and $V$, a.k.a. the hyperoctahedral group $B_q$.

A nice variation/extension of the problem is: What is g(theta), where g is the number of steps to return the cake to its exact initial state? But the key is understanding f(theta) first.

Pingback: Stan Wagon, TYI 23: Ladies and Gentlemen: The Answer | Combinatorics and more