Exact number of flips required to sort a burnt stack of pancakes
This solves a long-standing open case for a specific residue class in the burnt pancake problem, but the overall problem remains partially open and the result is incremental.
The paper resolves the exact number of flips required to sort the worst-case burnt pancake stack for n ≡ 1 mod 4, proving the upper bound ⌊(3n+3)/2⌋ matches the lower bound. For even n, the problem remains open with two possible values, and the lower bound is attained for n=24 and n=26.
In this work, we consider the burnt pancake problem, which is a well-studied problem going back to a work of Gates and Papadimitriou from 1979.The problem is to sort a stack of~$n$ one-sided burnt pancakes of different sizes, by a sequence of flips of the top pancakes, such that at the end of the flipping sequence the pancakes have increasing size and the burnt sides of all pancakes are face-down. The pancakes are denoted by $ 1,2,\dots,n$, and a number is multiplied by $-1$, if the corresponding pancake has burnt side face-up. Let $T(n)$ be the minimum number of flips to sort a special stack of $n$ pancakes, namely $\overline{I_n} := [-1,-2,...,-n]$. The instance $\overline{I_n} $ has strong relevance because of its easy structure and as it has been shown to be a worst-case instance for several small $n$. Heydari and Sudborough gave in 1997 the currently best upper bound of $T(n)$, namely $ \lfloor (3n+3)/2 \rfloor $ for $ n \equiv 3 \bmod 4$, which later has been shown to be exact by a work of Cibulka from 2011. Except these two works, no progress regarding lower and upper bounds has been made until now. In our work, we present that $ \lfloor (3n+3)/2 \rfloor $ is also an upper bound of $T(n)$ for $n \equiv 1 \bmod 4 $, which again matches the lower bound of Cibulka in 2011 and thus is exact. Furthermore, we show that our construction approach for $n \equiv 1 \bmod 4 $ and the one of Heydari and Sudborough for $n \equiv 3 \bmod 4 $ cannot be applied for even $n$. However, as there might be different construction approaches, the case of even $n$ remains an open problem, where two possible values for $T(n)$ are possible, namely $ (3/2)n + 1 $ or $ (3/2)n + 2 $. Finally, we found two values, namely $n=24$, $n=26$, where the lower bound is attained.