7 Simul Flip Theory

Here we present the proof of the above solution, by formulating the problem of solving Rubik's clock as solving a linear algebra equation.
We will explain how to formulate the problem as a matrix equation, with the example of filling only a few row entries in the matrix.
The entries could also be computed using a computer program.

Problem formulation

Denote initial values of the clock as a0, b0, c0, d0, ...

There are 14 operations in 7 simul flip method. In step 1, we execute left dial (operation 1), then right dial (operation 2). In step 2, we execute left dial (operation 3), then right dial (operation 4) ...
We annotate the clockwise turn count on each operation. If we turn more than 180 degrees clockwise, annotate as a negative.
For example, if we turn left dial at first step by 60 degrees counterclockwise, then x1 = -2.
If we turn right dial at first step by 120 degrees clockwise, then x2 = 4.

Let's consider the effect of applying x1 move(s) onto the clock value.
See the visualilzation below of how would one clockwise move affect the clocks.
We know that clock a will be incremented by 1, clock b will be incremented by 1, etc.

Therefore, after applying x1 moves(s), the clock become:

After applying all the moves, clock a will become a0 + a linear combination of x1, x2, ...
We could either derive these linear combinations by manually tracing the effect of each move, or with a computer program.
Eventually, we want to find a solution of x1, x2, ..., x14, such that each clock become 0.

If you are familiar with linear algebra, this could be represented as an equation:

$$ \begin{bmatrix} a_0 \\ b_0 \\ c_0 \\ \vdots \\ n_0 \end{bmatrix} + T \cdot \begin{bmatrix} x_1 \\ x_2 \\ x_3 \\ \vdots \\ x_{14} \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \\ 0 \\ \vdots \\ 0 \end{bmatrix} $$

Notice T is a 14x14 matrix, and it happens to be invertible.

$$ \Rightarrow \begin{bmatrix} x_1 \\ x_2 \\ x_3 \\ \vdots \\ x_{14} \end{bmatrix} = -T^{-1} \cdot \begin{bmatrix} a_0 \\ b_0 \\ c_0 \\ \vdots \\ n_0 \end{bmatrix} $$

If we denote $$ S = -T^{-1} $$, then $$ x_1 = S_{1,1}a_0 + S_{1, 2}b_0 + \cdots + S_{1, 14}n_0 $$ In other word, each $$x_i$$ is a linear combination of $$a_0, b_0, \cdots, n_0$$ The solution after cleaning up will be the results presented in 7sf+ and speed BLD

Why a no-flip 7-simul needs at least 5 memos

The flip methods on this site memorize very little: 6 values for 7 Simul Flip, and only 2 for the 2-letter memo variant. A natural question is whether a no-flip 7-simul — one that never turns the clock over, so the solver only ever looks at the front face — can also get away with fewer memos than the 5 of Bowen's 7 Simul. The answer is no, and it is a theorem, not just a failed search: no choice of pin order, wheels, or matching rules can beat 5.

Theorem. Consider any solving scheme satisfying the following three assumptions:

Then $$k \ge 5$$.

Proof. By contradiction: suppose 4 memos suffice, i.e. $$k = 4$$ works. A clock state is described by 14 coordinates: the nine front dials, and the five back non-corner dials $$u, l, c, r, d$$. (The four back corner dials are not independent: each is the negative of a front corner dial, so they carry no extra information.)

Restrict to the scrambles whose front face is all 0. There are $$12^5$$ of them, one for each choice of the five back coordinates — but only $$12^4$$ possible combinations of the 4 memos. By the pigeonhole principle, two different back faces must share exactly the same memo.

Those two scrambles receive identical turns: with the same memo and the same front readings — the front starts identical (all 0) and afterwards changes only through the turns themselves, and a wheel turn only ever adds its amount to a fixed set of dials, never copying one dial's value into another — the two executions get the same turn amount at every step.

But the two back faces were different, and identical turns change them identically, so they still differ on the back at the end: at most one of them can end solved. This contradicts A1, which says the scheme solves every scramble. Hence $$k \ge 5$$. ∎

Scope. The proof never uses the pin order, the wheels, the matching targets, the linearity of the memos, or even the number of turns — it applies to any execution that reads only the front face. Rotating the clock in the plane, without flipping it, does not escape the bound either: the same nine dials are still the only ones read. Two things do escape it, and both amount to buying back-face information during execution: flipping the clock (or peeking at the back) — which is exactly what the flip methods do, and how the 2-letter memo gets down to 2 — and choosing the pin order itself at inspection time, since that choice is inspection information and would have to be counted toward $$k$$.

The bound is tight: Bowen's 7 Simul achieves exactly 5 memos, so the minimum for a no-flip 7-simul is exactly 5 — while the flip variant escapes the bound by reading the back face, down to 2 memos.

Further reading

Here are some websites with other fun theories: