Useful mathematics — the cuber's toolbox

The Useful Mathematics note on jaapsch.net argues that cube enthusiasts rarely need abstract group theory — just a handful of visual tricks: two-line notation, cycle decomposition, crossing-number-as-parity, and lcm-as-order. This section rolls those tricks (and ten more) into a single practical reference: type any permutation and see all the structural data live; type A, B and watch σ A σ⁻¹, [A, B], σ^k, ⟨A, B⟩ computed on the spot. Think of it as the essay's pocket handbook.

32.1 Four equivalent views

For a permutation $\sigma$ of $n$ elements, the same object can be written four equivalent ways:

• Two-line: top row 1..n, bottom row σ(1)..σ(n). Cauchy's 1815 original, ideal for table look-ups.
• Cycles: (1 a₁ a₂ …)(b₁ b₂ …)…, listing σ's orbits. Introduced by Cauchy in 1844; the format of choice for multiplication.
• Crossings: draw lines from top row to bottom; count pairwise intersections. Parity of crossings = sgn(σ). The geometric parity test.
• Order: ord(σ) = lcm of cycle lengths. "How many times do I apply σ to get back to identity?"

The 5-cycle (1 2 3 4 5) has order 5 and parity (5−1) = 4, even. A transposition (a b) has order 2 and parity (2−1) = 1, odd. On the cube, R acts as a 4-cycle on corners (odd) and a 4-cycle on edges (odd) — product is even, matching §5's "joint parity = +1" invariant.

32.2 Connection to the cube

The cube alg FR acts as a 5-cycle on corners and a 7-cycle on edges (per Singmaster's Cubic Circular issue 2/3), so order(FR) = lcm(5, 7) = 35. Every alg's order follows this recipe: look at corner/edge cycle lengths and take the lcm. G has exactly 73 distinct element orders (§7.2); maximum is 1260, realised by R U2 D' B D' (in the pattern gallery as "Order-1260").

32.3 Two-line notation — Cauchy's original idea

In 1815, 23-year-old Augustin-Louis Cauchy treated "permutation" as a free-standing mathematical object for the first time in his Mémoire sur le nombre des valeurs qu'une fonction peut acquérir. His two-line notation (made standard in the 1844 Mémoire sur les arrangements) is

Computation rule: the bottom row of $\sigma \cdot \tau$ is obtained by feeding τ's bottom row as indices into σ's bottom row. That is$$(\sigma \tau )(i)=\sigma (\tau (i)).$$Note σ is outer (apply τ first, then σ): this is the mathematician's convention. Cubers reverse it (see 32.5).

32.4 Cycle decomposition — algorithm and shape

Given σ, the cycle decomposition algorithm is one you can do by hand in a minute:

1. Find the smallest unvisited i (start at i = 1).
2. Iterate σ: i → σ(i) → σ(σ(i)) → …, stopping when you return to i. Mark all visited.
3. Repeat until every element has been visited.

The result is σ's disjoint cycle decomposition, unique up to (a) rotating each cycle and (b) reordering disjoint cycles (which trivially commute). The unordered multiset of cycle lengths — the cycle type — is σ's fundamental fingerprint, and determines σ's conjugacy class in $S_n$ (§21.4).

This theorem is the soul of §8 (conjugation) and §21.4 (conjugacy classes), and the algebraic basis of blindsolving setups: "use σ to bring the target piece into position, apply τ, then σ⁻¹ to restore" corresponds exactly to σ τ σ⁻¹, which has the same cycle shape as τ but moved to new locations.

32.5 Composition order — the math-vs-cuber trap

Here is the trap every beginner falls into: the same string "σ τ" denotes different products for a mathematician and for a cuber.

Concrete example: σ = (1 2)(3 4), τ = (1 3). Compute σ τ under both conventions:

• Math: στ = σ∘τ. (στ)(1) = σ(τ(1)) = σ(3) = 4. (στ)(3) = σ(1) = 2. (στ)(2) = σ(2) = 1. (στ)(4) = σ(4) = 3. Result = (1 4 3 2).
• Cuber: στ means "first σ then τ" = τ∘σ. Run through the same four values: result = (1 3 2 4) (the inverse of the above).

What does this essay use? The cuber convention (since the subject is the cube). But the widget below shows both results side by side so you can flip mentally between them.

32.6 Parity — three definitions, one invariant

The sign sgn(σ) ∈ {+1, −1} is the most useful invariant on $S_n$. It admits three equivalent definitions, each useful in a different setting:

1. (i) Crossings: draw the two-line diagram and count line crossings. sgn(σ) = (−1)^#crossings. Geometric, ideal for visual quick checks.
2. (ii) Cycle count: $\mathrm{sgn}(\sigma )=(-1{)}^{n-c(\sigma )}$, where c(σ) is the cycle count (including fixed points). Equivalently sgn(σ) = ∏ (−1)^|cᵢ|−1. Algebraic — best when cycles are known.
3. (iii) Transposition decomposition: write σ = t₁ t₂ … t_k as a product of transpositions. sgn(σ) = (−1)^k, and crucially the parity of k is independent of the decomposition. The fact that "parity is well-defined" is the most non-trivial of the three.

32.7 Order — the lcm rule and Landau's function

The order ord(σ) is the smallest positive integer k with σ^k = e. Formula:

where c_i ranges over σ's disjoint cycles. The reasoning is simple: σ^k rotates each cycle c by k steps, which equals identity iff |c| divides k. All cycles satisfy this simultaneously iff k is a common multiple of every |c_i| iff the smallest such k = lcm.

32.7.1 Landau's function — maximum order in Sₙ

Across all of $S_n$, the largest possible order is Landau's function g(n) (Edmund Landau 1903, Über die Maximalordnung der Permutationen gegebenen Grades). It is the answer to the combinatorial optimisation "partition n = n₁ + n₂ + … + n_r so that lcm(n₁, …, n_r) is maximal".

The extremal partitions are surprisingly delicate: g(5) = 6 from 2 + 3; g(7) = 12 from 3 + 4; g(11) = 30 from 2 + 3 + 5 + 1 (note the spare fixed point). Asymptotically $\ln g(n)\sim \sqrt{n\ln n}$ (Landau).

For the cube: corners cp ∈ S₈, edges ep ∈ S₁₂. g(8) = 15 (from 3 + 5) and g(12) = 60 (from 3 + 4 + 5). With orientations layered on (CO ∈ ℤ/3⁷, EO ∈ ℤ/2¹¹), orders become lcm of "(cycle length × orientation order)" — pushing the maximum on G to 1260 = lcm(4, 5, 7, 9), far above either g(8) or g(12). See §13.

32.8 Conjugation visualiser — the relabel rule

Theorem 32.2 says conjugation is "rename via σ", but the verbal statement is hard to internalise. The widget below lets you type σ and τ, computes σ τ σ⁻¹ live, and shows cycle by cycle how each cycle of τ is relabelled into a cycle of σ τ σ⁻¹.

Cube interpretation: τ is "an alg you already know" (e.g. an edge 3-cycle); σ is "a setup move that brings the target pieces into τ's home position"; σ τ σ⁻¹ then "relocates τ's effect to the new spot". This is the shared grammar of BLD, FMC and ZBLL setup work.

32.9 Commutator visualiser — when [A, B] is a 3-cycle

Recall [A, B] = A B A⁻¹ B⁻¹ measures how badly A and B fail to commute. Its key cube application (§9): when A and B affect almost disjoint regions but share exactly one piece, [A, B] is a clean 3-cycle — that shared piece, plus one companion each from A's and B's region, cycle among themselves while everything else returns home.

Pure permutations show this too: A = (1 2 3 4 5), B = (3 4 5 6 7) — overlap {3, 4, 5}, and [A, B] is a 5-cycle. Switch B to (5 6 7) so overlap shrinks to {5}, and [A, B] reduces to a clean 3-cycle. This is the precise criterion for "commutator = 3-cycle" — the same arithmetic that backs the four atom algs in §9.1.

32.10 Powers — σ^k and the loop home

The cycle structure of σ^k follows a simple rule: a cycle of length L breaks into gcd(k, L) cycles of length L / gcd(k, L) under σ^k. When k is a multiple of L, that cycle splinters into L fixed points. So σ^ord(σ) = e is automatic.

Cube application: applying any alg X repeatedly eventually returns the cube to its start, after ord(X) repetitions. R has order 4 (R⁴ = e); R U has order 105 (5-cycle corners × 7-cycle edges × flips → lcm 105); a typical "R U R'" repeats back in 6. Drag the slider below to feel the cycle.

32.11 Generated subgroups — how big do two elements get?

Given two permutations g₁, g₂ ∈ $S_n$, how large is the subgroup ⟨g₁, g₂⟩? An apparently elementary question with deep consequences. Classical examples:

• ⟨(1 2)⟩ = ℤ/2 (2 elements).
• ⟨(1 2), (2 3)⟩ = $S_3$ (6 elements, since adjacent transpositions generate $S_n$).
• ⟨(1 2 3 4 5), (1 2)⟩ = $S_5$ (120 elements — the classic "one transposition + one n-cycle generate $S_n$").
• ⟨(1 2 3 4 5), (1 3)⟩ = ? Guess first, then check with the widget — the answer is not obvious. (In fact, it still equals $S_5$, because conjugates of (1 3) by powers of the 5-cycle cover all transpositions on {1..5}.)

The widget below uses the most naive BFS (works for n ≤ 9; n ≥ 10 has > 3.6M states, too big). The same idea scales via the Schreier-Sims algorithm (Sims 1970) — GAP can compute |⟨U, R⟩| in milliseconds despite |G| = 4.3 × 10¹⁹.

32.12 Cycle index polynomial — the colouring formula

When a group G acts on n objects, the cycle index polynomial Z_G(z₁, …, z_n) encodes the cycle type of each element on the n-object action, averaged over G:

where c_k(g) is the number of length-k cycles in g's action on the n objects. Pólya's enumeration theorem: the number of distinct c-colourings up to G-equivalence equals Z_G(c, c, …, c).

Worked example: the dihedral group D₄ (8 elements) acting on the 4 vertices of a square. List the cycle structure of each of the 8 elements on those 4 vertices, then average to get Z_D₄. This is the root of the "square necklace" counting problem.

For the cube's outer symmetry O_h (48 elements, 24 rotations + 24 mirror rotations) acting on 6 faces, a similar computation gives Z_Oh(c, …, c) = (c⁶ + 3c⁴ + …) / 48. At c = 6 the count is exactly 30 — the number of essentially distinct 6-colour cubes (§21.7).

32.13 Memorising algs as math objects

One last practical takeaway: don't memorise PLL / OLL / ZBLL by finger sequence — recognise them as mathematical objects.

• PLL = a cycle product on top-layer cp ∘ ep. The 21 PLL cases correspond exactly to the 21 non-trivial combinations of "corner-cycle shape + edge-cycle shape" on the top layer (H, Z, Ua, Ub, …). Look at a scramble's top layer and visualise the connection graph: which 3 corners cycle, which 3 edges. Names = cycle shapes.
• OLL = an orientation vector (co, eo) on the top layer. The 57 OLL cases correspond to all (co ∈ (ℤ/3)⁴, eo ∈ (ℤ/2)⁴) consistent with Σco ≡ 0 mod 3. The cross / antisune / T-shape iconography labels distinct eo vectors; corner-orientation icons label co vectors.
• ZBLL = simultaneous (PLL × OLL) recognition. Roughly 493 cases ↔ all (cp, ep, co) states with ep already oriented and eo = 0. Memorise algebraically: "identify cp cycle type (one of 21 PLL); then co vector (one of 8 fixed-cp OLLs)" — only 21 + 8 ≈ 29 primitive patterns instead of 493 independent finger sequences.

This is why §22's "cycle then orientation" subgroup chain (CF → F2L → OLL → PLL) is as much a memory architecture as a solving architecture: it stratifies 4.3 × 10¹⁹ states so that each layer has only tens of equivalence classes — a count the human brain can hold.