Extrapolated tiling counts for 2s×(2s+2)2s\times(2s+2) rectangles

About 10 years old · traced to

Let Tn×m(s,k)T_{n\times m}(s,k) denote the number of ways to tile an n×mn\times m rectangle using s×ss\times s squares and 1×11\times1 squares, with exactly kk squares of size s×ss\times s. For s>2s>2, the extrapolated tiling-count conjecture.

T2s×(2s+2)(s,2)=3+18s+7s2.T_{2s\times(2s+2)}(s,2)=3+18s+7s^2. T2s×(2s+2)(s,3)=8+40s.T_{2s\times(2s+2)}(s,3)=8+40s. T2s×(2s+2)(s,4)=36.T_{2s\times(2s+2)}(s,4)=36.

These formulas are extrapolated from the computed database of tilings; their general validity is not established in the supplied text.

References

Primary source

Richard J. Mathar, “Tiling n X m rectangles with 1 X 1 and s X s squares”, arXiv:1609.03964 (2016).

Progress summary

Refreshed
Claimed progress

The formulas remain computational conjectures, while a posted but independently unverified argument claims a general counting proof covering them.

Mathar’s 2016 paper states the three formulas for rectangles with s>2s>2 as Conjecture 2, obtained by extrapolating computed tiling data rather than proving them.

Posted attempt

An unverified argument claims a uniform enumeration for 0≤t<s0\le t<s, with B=(t+22)B=\binom{t+2}{2}, that specializes at t=2t=2 to all three conjectured formulas; it also claims companion formulas for t=1t=1. The argument has not been independently verified.

Current status (as of August 2026): The formulas remain unproved in the published source, while a posted general enumeration gives claimed progress but has no independent verification.

Sources

Solutions 1

ProofThis solution needs a summarySee full solutionHide full solution

In fact, both families of conjectures follow from the following uniform enumeration. Let 0≤t<s0\le t<s and put

B=(t+22).B=\binom{t+2}{2}.

Then

T2s×(2s+t)(s,2)=(s+t+1)2+B((s+1)2−2),T2s×(2s+t)(s,3)=2B(s+t+1)+(s−1)(t+1)(t+2)(2t+3)3,T2s×(2s+t)(s,4)=B2.\begin{aligned} T_{2s\times(2s+t)}(s,2)&=(s+t+1)^2+B((s+1)^2-2),\\ T_{2s\times(2s+t)}(s,3)&=2B(s+t+1)+(s-1)\frac{(t+1)(t+2)(2t+3)}3,\\ T_{2s\times(2s+t)}(s,4)&=B^2. \end{aligned}

A large square is determined by its upper-left corner (r,c)(r,c), where 0≤r≤s0\le r\le s and 0≤c≤s+t0\le c\le s+t. Two such squares are disjoint precisely when

∣r−r′∣≥sor∣c−c′∣≥s.|r-r'|\ge s\quad\text{or}\quad |c-c'|\ge s.

Vertical separation is therefore possible only between boundary rows 00 and ss. The number of unordered column pairs with separation at least ss is

∑d=ss+t(s+t+1−d)=(t+22)=B.\sum_{d=s}^{s+t}(s+t+1-d)=\binom{t+2}{2}=B.

Moreover, since s+t<2ss+t<2s, three pairwise horizontally separated squares cannot occur.

For two squares, vertically separated pairs contribute (s+t+1)2(s+t+1)^2. For each of the BB horizontally separated column pairs there are (s+1)2(s+1)^2 ordered row choices, exactly two of which were already counted vertically. This gives the first formula.

For three squares supported entirely on boundary rows, choose which row carries two squares, its separated column pair, and the opposite-row column. This gives 2B(s+t+1)2B(s+t+1). Otherwise exactly one square occupies an interior row. For its column cc, write

q(c)=#{d∈{0,…,s+t}:∣d−c∣≥s}.q(c)=\#\{d\in\{0,\ldots,s+t\}:|d-c|\ge s\}.

All such columns lie on one side of cc and have mutual separation less than ss, so the remaining two squares must occupy the two different boundary rows. Their columns can be chosen independently in q(c)2q(c)^2 ways. The nonzero values of q(c)q(c) are t+1,t,…,1t+1,t,\ldots,1 at one end and 1,…,t,t+11,\ldots,t,t+1 at the other, whence

∑cq(c)2=2∑j=1t+1j2=(t+1)(t+2)(2t+3)3.\sum_cq(c)^2=2\sum_{j=1}^{t+1}j^2=\frac{(t+1)(t+2)(2t+3)}3.

There are s−1s-1 interior rows, proving the second formula.

For four squares, an interior-row square can coexist with at most two others. Thus all four lie on the boundary rows, with exactly two on each row. The separated column pair on each boundary row can be chosen independently, yielding B2B^2.

Taking t=2t=2, so B=6B=6, gives exactly

T2s×(2s+2)(s,2)=7s2+18s+3,T2s×(2s+2)(s,3)=40s+8,T2s×(2s+2)(s,4)=36T_{2s\times(2s+2)}(s,2)=7s^2+18s+3,\quad T_{2s\times(2s+2)}(s,3)=40s+8,\quad T_{2s\times(2s+2)}(s,4)=36

for every s>2s>2. The same theorem with t=1t=1 also gives 4s2+10s+14s^2+10s+1, 16s+216s+2, and 99, settling the companion formulas for 2s×(2s+1)2s\times(2s+1) rectangles.