I recently realized that the number of unimodal cyclic permutations of ${1,2,\dots,n}$ matched the number of primitive binary necklaces where complements are considered equivalent, given by the formula $$\frac{1}{2n}\sum_{d \mid n, d\text{ odd}}\mu(d)2^{n/d}$$ as described in A000048. I do not have a proof of this, but it seems to hold for all $1 \le n \le 20$. My question is twofold:
- Why does this formula apply for primitive binary necklaces where complements are equivalent? In particular, I know that one can derive that the number of primitive binary strings of length $n$ is $\sum_{d \mid n}\mu(d)2^{n/d}$ using Möbius inversion. Dividing through by $n$ accounts for rotational symmetry. Halving the count accounts for complement symmetry, but there are some cases where a rotation might map a necklace to its complement. But then how does d being odd resolve this?
- Does there exists a bijection between unimodal cyclic permutations and primitive binary necklaces where complements are equivalent? If not, then how can one prove my original assertion that the number of unimodal cyclic permutations does indeed follow the above formula? A simple proof requiring little higher math, if possible, would be appreciated.
As an example of what I mean by "unimodal cyclic permutations", the three such permutations for $\{1,2,3,4,5\}$ are $(2,4,5,3,1)$, $(2,3,4,5,1)$, and $(3,5,4,2,1)$, which form the cycles $(1\:2\:4\:3\:5)$, $(1\:2\:3\:4\:5)$, and $(1\:3\:4\:2\:5)$, respectively. The $3$ corresponding primitive binary necklaces, expressed in Lyndon word form, would be $00001$, $00011$, and $00101$.