AFTD Auto-Formalizing Theoretical Domains

Research Blogs

No Integer Square Admits a Perfect Admissible Mondrian Partition with Common Tile Area $q^s r p^a$ for $p > r + 1$

low tier2026-10-27OP-280reviewed by Claude Code中文版

Abstract

The Mondrian art problem asks for the minimum defect $M(n)$, defined as the difference between the largest and smallest rectangle areas in a partition of an $n \times n$ square into noncongruent integer rectangles with positive integer side lengths. An open question highlighted in arXiv:2610.19193 asks whether there exists any integer $n \ge 3$ admitting a perfect partition, which corresponds to defect $M(n) = 0$ with all tiles sharing an identical area. Previous results have ruled out perfect partitions where the common tile area is a prime power or a product of two prime powers. In this post, we present a formal proof in Lean 4 establishing that no integer square admits a perfect admissible Mondrian partition with common tile area of the form $q^s r p^a$, where $p, q, r$ are distinct prime numbers satisfying $p > r + 1$, uniformly for all natural exponents $a$ and $s$. The proof proceeds by strong induction on the exponent $s$. When $s = 0$, the tile area reduces to two prime factors, excluded by the two-prime theorem. When $s > 0$, total area divisibility forces $q$ to divide the square side $n$. A superincreasing signed-height obstruction across two geometric chains forces $q$ to divide both dimensions of every rectangle, showing that $q^2$ divides the common area, whence $s \ge 2$. Rescaling the canvas by $q$ produces a smaller perfect partition with common area $q^{s-2} r p^a$, which contradicts the induction hypothesis. These results are machine-checked in Lean 4.

1. Introduction

The Mondrian art problem, originating in recreational mathematics and studied in discrete geometry and combinatorial optimization, considers tiling an $n \times n$ square canvas with axis-aligned integer-sided rectangles. To avoid trivial solutions such as cutting a square into identical strips, two strict admissibility conditions are imposed: the partition must consist of at least two rectangles, and no two rectangles may be congruent, where congruence permits quarter-turn rotations so that an $a \times b$ rectangle is congruent to a $b \times a$ rectangle. The defect of such an admissible partition $P$, denoted $\mathrm{defect}(P)$, is defined as the difference between the maximum and minimum areas among all constituent rectangles: $$\mathrm{defect}(P) = \max_{R \in P} \mathrm{area}(R) - \min_{R \in P} \mathrm{area}(R).$$ The Mondrian defect function $M(n)$ is the minimum defect achieved over all admissible partitions of the $n \times n$ square.

An open question in the study of this problem is whether $M(n) = 0$ can ever occur for any integer $n \ge 3$. As noted in arXiv:2610.19193 (Section 1, below equation (1.1)), it remains an open problem whether $M(n) = 0$ for some $n$, though it is conjectured that no such $n$ exists. A partition achieving defect 0 is termed a perfect Mondrian partition. In such a partition, every tile must have identical area $A$, and because the tiles partition the entire $n \times n$ square, the number of rectangles $|P|$ and the common area $A$ must satisfy the product relation $|P| \cdot A = n^2$.

One line of investigation into the nonexistence of perfect partitions establishes structural arithmetic constraints on the canvas side length $n$ and on the common tile area $A$. On the side length side, earlier formal developments proved that no perfect partition exists if $n$ is a prime power (mondrian_no_perfect_prime_power), a product of two prime powers $q^f p^e$ (mondrian_no_perfect_two_prime_powers), or of the form $q p^e$ (mondrian_no_perfect_prime_mul_prime_pow). Furthermore, large prime descent principles have been formalized: mondrian_large_prime_descent and its biconditional mondrian_perfect_partition_large_prime_iff demonstrate that for $p > c$, an admissible perfect partition of the $c p^e$ square exists if and only if one exists on the $c$ square, which implies that no square of side $6 p^e$ admits a perfect partition for prime $p > 6$ (mondrian_no_perfect_six_mul_prime_pow). Modular counting properties of tiles whose side lengths are coprime to prime factors of $n$ were established in mondrian_perfect_prime_free_side_counts and mondrian_perfect_min_width_layer_count.

On the tile area side, results showed that no perfect partition can have a common tile area that is a pure prime power $p^t$ (mondrian_no_perfect_prime_power_area). This was subsequently generalized to any tile area with two distinct prime factors: mondrian_no_perfect_two_prime_area established that no integer square admits a perfect partition where all tiles have area $q^s p^t$ for distinct primes $p$ and $q$ and arbitrary exponents $s, t \in \mathbb{N}$.

The question naturally arises whether this area-based obstruction can be pushed beyond two prime factors. When a candidate tile area has three prime factors, the number of available factor pairs $(w, h)$ with $w \cdot h = A$ grows, providing more combinatorial choices for rectangle shapes satisfying the total area constraints.

In this post, we discuss the theorem mondrian_no_perfect_three_prime_area_one_factor, formally verified in Lean 4. The theorem proves that no integer square admits a perfect admissible Mondrian partition when the common tile area has the form $q^s r p^a$ for distinct primes $p, q, r$ satisfying $p > r + 1$, uniformly for all natural exponents $a$ and $s$. This extends the area obstruction to three prime factors, combining a superincreasing two-chain signed-height argument with strong induction and geometric descent.

2. Preliminaries

We now formulate the definitions and geometric setting as implemented in Lean 4. Throughout, all coordinates and side lengths are non-negative integers in $\mathbb{N} = \{0, 1, 2, \dots\}$.

Definition 1 (mondrian_rect_width). An axis-aligned rectangle $r$ is represented as a 4-tuple $(x_0, x_1, y_0, y_1) \in \mathbb{N}^4$, corresponding to the Cartesian product $[x_0, x_1] \times [y_0, y_1]$ in the plane. Its width is defined by $$\mathrm{width}(r) = r.2.1 - r.1 = x_1 - x_0.$$ Its height is analogously given by $y_1 - y_0 = r.2.2.2 - r.2.2.1$.

Definition 2 (mondrian_rect_area). The area of a rectangle $r = (x_0, x_1, y_0, y_1)$ is the product of its width and height: $$\mathrm{area}(r) = \mathrm{width}(r) \cdot \mathrm{height}(r) = (x_1 - x_0)(y_1 - y_0).$$

Definition 3 (mondrian_rect_cells). The geometric footprint of a rectangle $r = (x_0, x_1, y_0, y_1)$ is the set of its constituent unit cells, each named by its lower-left integer lattice corner $(i, j)$: $$\mathrm{cells}(r) = \mathrm{Ico}(x_0, x_1) \times^\mathrm{s} \mathrm{Ico}(y_0, y_1) = \{(i, j) \in \mathbb{N} \times \mathbb{N} \mid x_0 \le i < x_1 \land y_0 \le j < y_1\}.$$

Definition 4 (mondrian_rect_congruent). Two rectangles $r$ and $s$ are congruent if they share the same unordered pair of side lengths: $$\mathrm{congruent}(r, s) \iff (\mathrm{width}(r) = \mathrm{width}(s) \land \mathrm{height}(r) = \mathrm{height}(s)) \lor (\mathrm{width}(r) = \mathrm{height}(s) \land \mathrm{height}(r) = \mathrm{width}(s)).$$ This definition explicitly permits rotations by 90 degrees.

Definition 5 (mondrian_is_partition). For a canvas side length $n \in \mathbb{N}$, a list $P$ of rectangles in $\mathbb{N}^4$ is an admissible Mondrian partition of the $n \times n$ square if: 1. Non-triviality: $2 \le P.\mathrm{length}$, ensuring that the partition contains at least two rectangles; 2. Positivity: For every $r \in P$, $r.1 < r.2.1$ and $r.2.2.1 < r.2.2.2$, so both width and height are strictly positive; 3. Pairwise disjointness and noncongruence: $P.\mathrm{Pairwise}$ holds with respect to the relation requiring that any two distinct rectangles $r, s$ in the list have disjoint cell sets ($\mathrm{cells}(r) \cap \mathrm{cells}(s) = \emptyset$) and are noncongruent ($\neg \mathrm{congruent}(r, s)$); 4. Canvas coverage: The union of the cell sets of all rectangles in $P$ equals the full set of unit cells of the $n \times n$ canvas: $$\bigcup_{r \in P} \mathrm{cells}(r) = \mathrm{range}(n) \times^\mathrm{s} \mathrm{range}(n) = \{ (i, j) \in \mathbb{N} \times \mathbb{N} \mid 0 \le i < n \land 0 \le j < n \}.$$

Definition 6 (mondrian_defect). The defect of a rectangle list $P$ is the difference between the maximum and minimum rectangle areas: $$\mathrm{defect}(P) = \max_{r \in P} \mathrm{area}(r) - \min_{r \in P} \mathrm{area}(r),$$ evaluated over the list $P.\mathrm{map}\; \mathrm{mondrian\_rect\_area}$ with default value $0$ for empty lists. A partition $P$ is perfect if $\mathrm{defect}(P) = 0$.

Definition 7 (mondrian_rect_swap). The operation of interchanging the $x$ and $y$ coordinate pairs of an integer rectangle is defined by $$\mathrm{swap}(x_0, x_1, y_0, y_1) = (y_0, y_1, x_0, x_1).$$ Applying this operation to every rectangle in $P$ produces the transposed partition $P.\mathrm{map}\; \mathrm{mondrian\_rect\_swap}$, which preserves admissibility by mondrian_partition_swap.

Remarks on Lean choices:

  1. The non-triviality condition $2 \le P.\mathrm{length}$ in mondrian_is_partition explicitly excludes the trivial tiling of an $n \times n$ canvas by the single rectangle $[0, n] \times [0, n]$, which would otherwise trivially attain defect 0.
  2. The positivity requirement uses strict inequalities $x_0 < x_1$ and $y_0 < y_1$, precluding degenerate rectangles of width 0, height 0, or area 0. Upper coordinate bounds $x_1 \le n$ and $y_1 \le n$ are established from canvas coverage in mondrian_partition_bounds.
  3. The defect definition defaults to $0$ on an empty list, but because mondrian_is_partition enforces $2 \le P.\mathrm{length}$, admissible partitions are never empty.
  4. Congruence is invariant under 90-degree rotations, preventing tiles of dimensions $a \times b$ and $b \times a$ from coexisting in an admissible partition.

3. Main result

Theorem (mondrian_no_perfect_three_prime_area_one_factor). Let $p, q, r$ be three distinct prime numbers satisfying $p > r + 1$. For all natural numbers $a, s, n \in \mathbb{N}$, there is no admissible Mondrian partition $P$ of the $n \times n$ square having defect $0$ and common tile area $$\mathrm{area}(u) = q^s \cdot r \cdot p^a$$ for all $u \in P$.

In Lean 4, mondrian_no_perfect_three_prime_area_one_factor universally quantifies over the prime parameters $p, q, r$, the exponents $a, s \in \mathbb{N}$, the square side length $n \in \mathbb{N}$, and the rectangle list $P$, asserting that the hypotheses of admissibility (mondrian_is_partition), zero defect (mondrian_defect), and uniform tile area imply a contradiction.

What this result adds: Prior to this theorem, the known area nonexistence theorem mondrian_no_perfect_two_prime_area excluded common tile areas having at most two distinct prime factors, $q^s p^t$, for arbitrary exponents $s, t \ge 0$. That result settled instances such as pure prime powers $p^t$ (mondrian_no_perfect_prime_power_area) and products of two prime powers like $2^s 3^t$ or $2^s 5^t$.

Moving from two to three prime factors means that for an area $A = q^s r p^a$, the set of integer divisors allows rectangle dimensions to decouple across prime factors, increasing the number of noncongruent factor pairs $(w, h)$ with $w \cdot h = A$ that can tile the canvas. Theorem mondrian_no_perfect_three_prime_area_one_factor addresses this by showing that under the separation condition $p > r + 1$, no perfect partition can exist, uniformly for all $a, s \ge 0$ and all canvas sizes $n$.

For example:

  • Setting $r = 2$, the separation condition $p > r + 1 = 3$ requires $p \ge 5$. This rules out all common tile areas of the form $q^s \cdot 2 \cdot 5^a$, $q^s \cdot 2 \cdot 7^a$, or $q^s \cdot 2 \cdot 11^a$ for any prime $q \notin \{2, p\}$ and all $a, s \ge 0$.
  • Setting $r = 3$, the condition requires $p \ge 5$, excluding areas of the form $q^s \cdot 3 \cdot p^a$ for all primes $p \ge 5$ and $q \notin \{3, p\}$.
Theorem mondrian_no_perfect_three_prime_area_one_factor (Lean 4, verified)
theorem mondrian_no_perfect_three_prime_area_one_factor (p q r a s n : ℕ)
    (hp : Nat.Prime p) (hq : Nat.Prime q) (hr : Nat.Prime r)
    (hpq : p ≠ q) (hqr : q ≠ r) (hpr : r + 1 < p)
    (P : List (ℕ × ℕ × ℕ × ℕ)) (hP : mondrian_is_partition n P)
    (hz : mondrian_defect P = 0)
    (harea : ∀ u ∈ P, mondrian_rect_area u = q ^ s * r * p ^ a) : False

4. Proof

Proof Idea: The theorem is proved by strong induction on the exponent $s$ of the prime $q$, generalizing over the square side length $n$ and the partition $P$. 1. In the base case $s = 0$, the tile area simplifies to $q^0 \cdot r \cdot p^a = r^1 \cdot p^a$. Since $p > r + 1$, the primes $p$ and $r$ are distinct. Thus the area involves only two distinct primes, which is directly ruled out by mondrian_no_perfect_two_prime_area. 2. For $s > 0$, the total area formula $n^2 = |P| \cdot (q^s r p^a)$ implies that $q \mid n^2$, and since $q$ is prime, $q \mid n$. 3. Using $q \mid n$ and $p > r + 1$, the structural lemma mondrian_three_prime_common_area_sides_divide establishes that $q$ must divide both the width and the height of every rectangle in $P$. This lemma combines the encoding of width-indivisible tiles into two geometric height chains (mondrian_prime_free_height_two_chain_encoding) with a superincreasing signed-height obstruction (mondrian_partition_width_dvd_of_height_two_chains), followed by transposition symmetry (mondrian_partition_swap). 4. Because both side lengths of every tile are divisible by $q$, the area of each tile is divisible by $q^2$. Coprimality of $q$ to $r p^a$ then forces $q^2 \mid q^s$, which implies $s \ge 2$. 5. Dividing all rectangle coordinates by $q$ scales down the partition (mondrian_perfect_partition_div_common_area), producing an admissible partition $Q$ of the $(n/q) \times (n/q)$ square with defect 0 and common tile area $q^{s-2} r p^a$. Since $s - 2 < s$, this contradicts the induction hypothesis.

Supporting Lemmas: We state the lemmas on which the induction rests.

Lemma 1 (mondrian_perfect_area_product). Let $P$ be an admissible Mondrian partition of the $n \times n$ square with defect 0. For every rectangle $u \in P$, $$|P| \cdot \mathrm{area}(u) = n^2.$$ Proof: By mondrian_defect_zero_equal, all rectangles in a partition of defect 0 have identical area. The sum of the areas of every rectangle in $P$ equals the total number of unit cells in the canvas, which is $n^2$. Because the list has length $|P|$ and every member has area $\mathrm{area}(u)$, the sum evaluates to $|P| \cdot \mathrm{area}(u) = n^2$.

Lemma 2 (mondrian_prime_free_height_two_chain_encoding). Let $p, q, r$ be prime numbers with $p \ne r$. If $x \cdot y = q^s \cdot r \cdot p^a$ and $q \nmid x$, then there exists $b \in \mathbb{N}$ such that $$y = q^s \cdot p^b \quad \text{or} \quad y = q^s \cdot r \cdot p^b.$$ Proof: Every divisor $x$ of $q^s \cdot r \cdot p^a$ has the unique factorization $q^i \cdot r^j \cdot p^k$ with $0 \le i \le s$, $0 \le j \le 1$, and $0 \le k \le a$. The assumption $q \nmid x$ forces $i = 0$. Consequently, the entire prime power factor $q^s$ divides $y = (q^s r p^a) / x$. Since $j \in \{0, 1\}$, the complementary factor $y$ equals $q^s p^{a - k}$ if $j = 1$, or $q^s r p^{a - k}$ if $j = 0$. Setting $b = a - k$ proves the lemma.

Lemma 3 (mondrian_partition_width_dvd_of_height_two_chains). Let $P$ be an admissible Mondrian partition of the $n \times n$ square with defect 0. Let $d, p, r, C \in \mathbb{N}$ with $d \mid n$, $C > 0$, $2 \le r$, and $r + 1 < p$. Suppose that for every rectangle $u \in P$, if $d \nmid \mathrm{width}(u)$, then there exists $a \in \mathbb{N}$ such that $$\mathrm{height}(u) = C \cdot p^a \quad \text{or} \quad \mathrm{height}(u) = C \cdot r \cdot p^a.$$ Then $d \mid \mathrm{width}(u)$ for every $u \in P$. Proof: If some tile has width not divisible by $d$, the Lean proof produces a nonempty family of such tiles whose heights satisfy a nontrivial signed relation $\sum \pm h_u = 0$. In a perfect partition all tiles have the same area and are pairwise noncongruent, so distinct tiles have distinct heights. Because $r \ge 2$ and $p \ge r + 2$, the values $C p^a$ and $C r p^a$, listed in increasing order, are superincreasing: each exceeds the sum of all smaller ones. So no nontrivial signed sum of distinct such heights vanishes, and every width is divisible by $d$.

Lemma 4 (mondrian_three_prime_common_area_sides_divide). Let $p, q, r$ be primes with $p > r + 1$. Let $P$ be an admissible partition of the $n \times n$ square with defect 0, where every tile has area $q^s \cdot r \cdot p^a$. If $q \mid n$, then for every rectangle $u \in P$, $$q \mid \mathrm{width}(u) \quad \text{and} \quad q \mid \mathrm{height}(u).$$ Proof: First, suppose for contradiction that some $u \in P$ has $q \nmid \mathrm{width}(u)$. Because $\mathrm{width}(u) \cdot \mathrm{height}(u) = q^s r p^a$, Lemma 2 (mondrian_prime_free_height_two_chain_encoding) applies with $x = \mathrm{width}(u)$ and $y = \mathrm{height}(u)$, guaranteeing that $\mathrm{height}(u)$ belongs to either the chain $q^s p^b$ or $q^s r p^b$. Setting $C = q^s$, $d = q$, and noting $r \ge 2$ (as $r$ is prime) and $p > r + 1$, Lemma 3 (mondrian_partition_width_dvd_of_height_two_chains) shows that every rectangle in $P$ has width divisible by $q$. Second, applying the coordinate swap operation (mondrian_rect_swap) transforms $P$ into a transposed partition $P'$ of the same square by mondrian_partition_swap, preserving defect 0 and common tile area, with widths and heights exchanged. Applying the width divisibility result to $P'$ shows that $q$ divides the height of every rectangle in $P$.

Lemma 5 (mondrian_perfect_partition_div_common_area). Let $q > 0$ with $q \mid n$. Let $P$ be an admissible partition of the $n \times n$ square with defect 0 and common tile area $A = q^2 \cdot B$. If $q$ divides both the width and height of every rectangle in $P$, then scaling all coordinates by $1/q$ produces an admissible partition $Q$ of the $(n/q) \times (n/q)$ square with defect 0 and common tile area $B$. Proof: Each rectangle $(x_0, x_1, y_0, y_1) \in P$ maps to $(x_0/q, x_1/q, y_0/q, y_1/q)$. Because $q$ divides all widths and heights, the new dimensions $(x_1 - x_0)/q$ and $(y_1 - y_0)/q$ are positive integers, and their product is $A / q^2 = B$. Disjointness of unit cells, exact coverage of $[0, n/q] \times [0, n/q]$, and pairwise noncongruence are preserved under uniform coordinate dilation by $1/q$. Hence $Q$ is an admissible partition with uniform tile area $B$, giving defect 0 by mondrian_defect_zero_of_equal.

Lemma 6 (mondrian_no_perfect_two_prime_area). For distinct primes $p$ and $r$ and natural exponents $s$ and $a$, no integer square admits an admissible Mondrian partition with defect 0 and common tile area $r^s \cdot p^a$.

Step-by-Step Proof of the Main Theorem: We now trace the Lean proof of mondrian_no_perfect_three_prime_area_one_factor:

1. Strong induction: The proof begins with strong induction on $s$, generalizing over $n$ and $P$. The induction hypothesis ih assumes that no such partition exists for any exponent $m < s$. 2. Base case ($s = 0$): We branch on whether $s = 0$. If $s = 0$, substitution into the area hypothesis gives $$\mathrm{area}(u) = q^0 \cdot r \cdot p^a = 1 \cdot r \cdot p^a = r^1 \cdot p^a.$$ Since $r + 1 < p$, we have $p \ne r$. Invoking mondrian_no_perfect_two_prime_area with primes $p$ and $r$ and exponents $1$ and $a$ directly refutes this case. 3. Existence of a tile: When $s > 0$, the list $P$ cannot be empty because $2 \le P.\mathrm{length}$ by hP.1. Taking the head of the list provides a tile $z \in P$. 4. Side length divisibility ($q \mid n$): Applying Lemma 1 (mondrian_perfect_area_product) to $z$ yields $|P| \cdot \mathrm{area}(z) = n^2$. Substituting $\mathrm{area}(z) = q^s \cdot r \cdot p^a$, we see that $q$ divides $q^s$, which divides $|P| \cdot \mathrm{area}(z) = n^2$. Because $q$ is prime, $q \mid n^2$ implies $q \mid n$. 5. Tile dimension divisibility: With $q \mid n$, all hypotheses of Lemma 4 (mondrian_three_prime_common_area_sides_divide) are satisfied. Therefore, for every rectangle $u \in P$, $q \mid \mathrm{width}(u)$ and $q \mid \mathrm{height}(u)$. 6. Area divisibility by $q^2$: Specifically for the tile $z$, both its width and height are multiples of $q$. Thus $q^2 \mid \mathrm{width}(z) \cdot \mathrm{height}(z) = \mathrm{area}(z) = q^s \cdot (r \cdot p^a)$. 7. Coprimality of $q^2$ and $r \cdot p^a$: Since $p, q, r$ are distinct primes, $q \ne p$ and $q \ne r$. By primality, $q \nmid p$ and $q \nmid r$. By Euclid's lemma, $q$ cannot divide $r \cdot p^a$. Thus $\gcd(q, r \cdot p^a) = 1$, which implies $\gcd(q^2, r \cdot p^a) = 1$. 8. Exponent bound ($2 \le s$): Because $q^2 \mid q^s \cdot (r \cdot p^a)$ and $q^2$ is coprime to $r \cdot p^a$, Gauss's lemma guarantees that $q^2 \mid q^s$. Since $q > 1$, this divisibility holds if and only if $2 \le s$. 9. Area factorization: Since $s \ge 2$, we rewrite $q^s = q^2 \cdot q^{s - 2}$, which yields $$q^s \cdot r \cdot p^a = q^2 \cdot (q^{s - 2} \cdot r \cdot p^a).$$ 10. Geometric descent and contradiction: Invoking Lemma 5 (mondrian_perfect_partition_div_common_area) produces an admissible partition $Q$ of the $(n/q) \times (n/q)$ square with defect 0 and common tile area $q^{s - 2} \cdot r \cdot p^a$. Because $s \ge 2$, we have $s - 2 < s$. Applying the strong induction hypothesis ih (s - 2) to $Q$ yields a contradiction, completing the proof.

5. Discussion

The theorem mondrian_no_perfect_three_prime_area_one_factor expands the class of tile areas known to be impossible for perfect Mondrian partitions. However, several natural boundaries and open questions remain.

First, the theorem treats tile areas with three prime factors where one of the primes appears with multiplicity exactly one: $A = q^s \cdot r \cdot p^a$. The argument rests on the prime-free encoding lemma mondrian_prime_free_height_two_chain_encoding, where tiles whose width is not divisible by $q$ have their height restricted to exactly two geometric chains ($q^s p^b$ or $q^s r p^b$), because the single factor $r$ can either divide the width or the height. If the exponent of $r$ were allowed to be an arbitrary natural number $b \ge 2$, the divisors would yield $b + 1$ parallel chains $q^s r^j p^k$ with $0 \le j \le b$. The two-chain superincreasing height argument mondrian_partition_width_dvd_of_height_two_chains does not immediately apply to three or more simultaneous chains without additional separation or density hypotheses.

Second, the theorem requires the separation condition $p > r + 1$. The primes $p$ and $q$ enter symmetrically, both with arbitrary exponents, so the theorem applies whenever one of the two other primes exceeds $r + 1$. For $r = 2$ this always holds, since one of two distinct odd primes is at least $5$; every area $q^s \cdot 2 \cdot p^a$ is excluded. Among three-prime areas with a prime of exponent one, the cases left open are those where both other primes are at most $r + 1$, such as $2^s \cdot 3^a \cdot 5$ or $3^s \cdot 5^a \cdot 7$.

Third, in relation to the overarching open problem of whether $M(n) = 0$ is possible for any $n \ge 3$ (arXiv:2610.19193), this result provides another piece of evidence supporting the conjecture that no perfect Mondrian partition exists. By establishing nonexistence for infinite families of tile areas, it shows that any hypothetical counterexample to the conjecture would have to involve either:

  • Highly composite tile areas with at least four distinct prime factors;
  • Areas of the form $q^s \cdot r^b \cdot p^a$ with $b \ge 2$; or
  • Three-prime areas $q^s \cdot r \cdot p^a$ in which neither $p$ nor $q$ exceeds $r + 1$, such as $2^s \cdot 3^a \cdot 5$.

Several concrete next questions arise directly from these findings:

  • Can the signed-height argument mondrian_partition_width_dvd_of_height_two_chains be extended to $k$ geometric chains, thereby ruling out general three-prime areas $q^s r^b p^a$ or four-prime areas $q^s r_1 r_2 p^a$?
  • Can alternative combinatorial or linear programming arguments overcome the $p > r + 1$ barrier to eliminate areas such as $2^s \cdot 3^a \cdot 5$?
  • Does there exist a general descent argument on the number of prime factors that could reduce any perfect partition to a small base case, ultimately confirming the conjecture that $M(n) \ge 1$ for all $n \ge 3$?

Why it is research-level

  • Excludes perfect partitions of every integer square with common tile area q^s*r*p^a for distinct primes p,q,r with p>r+1, uniformly in a and s. This extends the two-prime common-area obstruction by using the superincreasing signed-height argument

Lean declarations

Each one elaborated against Mathlib, and #print axioms returned nothing outside Classical.choice · Quot.sound · propext.

References

  1. arXiv:2610.19193 — A Bounded-Defect Construction for the Mondrian Art Problem

Drafted by gemini-3.8-flash from the knowledgebase, checked against it, and reviewed by Claude Code. Only the Lean is verified; the prose is a reading of it.