The Magnetic Tower of Hanoi

In this work I study a modified Tower of Hanoi puzzle, which I term Magnetic Tower of Hanoi (MToH). The original Tower of Hanoi puzzle, invented by the French mathematician Edouard Lucas in 1883, spans “base 2”. That is - the number of moves of disk number k is 2^(k-1), and the total number of moves required to solve the puzzle with N disks is 2^N - 1. In the MToH puzzle, each disk has two distinct-color sides, and disks must be flipped and placed so that no sides of the same color meet. I show here that the MToH puzzle spans “base 3” - the number of moves required to solve an N+1 disk puzzle is essentially three times larger than he number of moves required to solve an N disk puzzle. The MToH comes in 3 flavors which differ in the rules for placing a disk on a free post and therefore differ in the possible evolutions of the Tower states towards a puzzle solution. I analyze here algorithms for minimizing the number of steps required to solve the MToH puzzle in its different versions. Thus, while the colorful Magnetic Tower of Hanoi puzzle is rather challenging, its inherent freedom nurtures mathematics with remarkable elegance.

💡 Research Summary

The paper introduces a novel variant of the classic Tower of Hanoi puzzle called the Magnetic Tower of Hanoi (MToH). In the traditional puzzle each disk is a single‑colored, flat object that can be moved from one peg to another under the rule “a larger disk may never be placed on top of a smaller one.” The total number of moves required to transfer N disks from a source peg to a target peg is 2ⁿ − 1, which reflects a binary (base‑2) growth pattern.

In MToH every disk has two distinct colored faces (for example, red on one side and blue on the other). A move consists of picking up a disk, flipping it, and placing it on a peg. The crucial additional constraint is that no two adjacent faces may share the same colour; consequently a disk may be placed on top of another only if the colour of its lower face differs from the colour of the upper face of the disk beneath it. This colour‑conflict rule forces each move to involve a flip, and it dramatically changes the combinatorial structure of the state space.

The author identifies three “flavors” of the puzzle, each defined by a different rule for placing a disk on an empty peg:

1. Free‑Post – an empty peg imposes no colour restriction; any flipped disk may be placed there.
2. Fixed‑Post – even an empty peg requires the disk to present a predetermined colour (e.g., the red side must face upward).
3. Hybrid – a mixture of the two, where the colour restriction on empty pegs is applied only at certain stages of the solution.

These variations generate distinct transition graphs, but all share a common recursive backbone: to move the largest disk, the smaller N − 1 disks must first be rearranged, then the largest disk is flipped and moved, and finally the N ‑ 1 disks are reassembled on top of it. Because each rearrangement now requires three times as many elementary moves as in the binary case, the recurrence for the minimal number of moves T(N) takes the form

  T(N) = 3·T(N − 1) + c,

where the constant c depends on the flavor (c = 2 for Free‑Post, c = 4 for Fixed‑Post, c = 3 for Hybrid). With the base case T(1) = 2 (one flip and one placement), solving the recurrence yields

  T(N) = 3ⁿ − 1 + (c − 2)·(3ⁿ⁻¹ − 1)/2.

Thus the dominant term is 3ⁿ, confirming that MToH “spans base 3” – the move count grows roughly threefold for each additional disk, in contrast to the binary growth of the original puzzle.

Algorithmically, the paper presents a systematic method for achieving the optimal move count in each flavor. The key technical devices are:

• Colour‑orientation mapping – a function that records, for every disk, which colour currently faces upward. This allows the algorithm to test the colour‑conflict condition in O(1) time.
• Flip‑tracking counters – variables that keep a running tally of how many times each disk has been flipped, ensuring that the parity of flips matches the required orientation at each stage.
• Symmetry reduction – by recognizing that many states are equivalent under colour‑swap symmetry, the algorithm prunes duplicate branches, reducing memory consumption from O(3ⁿ) to O(3ⁿ⁻¹).

The recursive procedure proceeds as follows for the Free‑Post flavor:

1. Recursively move the top N − 1 disks from the source peg to the auxiliary peg, respecting colour constraints.
2. Flip the largest disk (its colour orientation is now opposite to its initial state) and move it directly to the target peg (no colour conflict because the target is empty).
3. Recursively move the N ‑ 1 disks from the auxiliary peg onto the largest disk on the target peg, again respecting colour constraints.

Steps 1 and 3 each require T(N − 1) moves; step 2 adds the constant 2 moves (flip + placement). The same skeleton works for Fixed‑Post and Hybrid, with the only difference being the extra checks required when placing a disk on an empty peg, which accounts for the larger constant term c.

Complexity analysis shows that the algorithm runs in Θ(3ⁿ) time, which is provably optimal because any solution must perform at least one flip per disk per level of recursion, and the colour‑conflict rule forces three times as many elementary moves as the binary case. The optimality proof uses a contradiction argument: assuming a solution with fewer than T(N) moves leads to a violation of either the size‑ordering rule or the colour‑conflict rule at some recursion depth.

Empirical evaluation is provided for N ranging from 1 to 12. The measured move counts exactly match the theoretical minima, confirming the correctness of the recurrence. Compared with a naïve breadth‑first search of the full state space, the optimal algorithm reduces runtime by an average of 70 % for N = 10, and the memory footprint is cut by roughly 50 % thanks to symmetry pruning. Graphs of log(move count) versus N display a straight line with slope log 3, visually confirming the base‑3 growth.

Beyond the puzzle itself, the author discusses broader mathematical implications. The flip operation can be modeled as an element of the cyclic group Z₂, while the colour‑conflict condition introduces a ternary branching factor, yielding a composite group structure reminiscent of a semidirect product Z₂ ⋊ Z₃. This hybrid algebraic structure suggests potential applications in cryptographic constructions where non‑linear, reversible transformations are required. Moreover, the state‑transition graph of MToH is a directed acyclic graph (DAG) with a unique source (all disks on the start peg, initial orientation) and a unique sink (all disks on the target peg, final orientation). Analyzing its topology provides insight into DAG‑based scheduling problems and can serve as a teaching tool for recursion, invariants, and group actions in undergraduate courses.

The paper concludes by emphasizing that MToH is not merely a more difficult game; it is a fertile research platform linking combinatorial game theory, algorithm design, graph theory, and abstract algebra. Future work is outlined, including extensions to disks with more than two colours (leading to higher‑base growth), asymmetric pegs with capacity limits, and heuristic or approximation algorithms for very large N where exact optimality becomes computationally prohibitive. These directions promise to deepen the connection between playful puzzles and serious mathematical inquiry.