Can you solve the Mondrian squares riddle Gordon Hamilton

Dutch artist Piet Mondrian’s abstract,
rectangular paintings

inspired mathematicians
to create a two-fold challenge.

First, we must completely cover a square
canvas with non-overlapping rectangles.

All must be unique, so if we use a 1x4,
we can’t use a 4x1 in another spot,

but a 2x2 rectangle would be fine.

Let’s try that.

Say we have a canvas measuring 4x4.

We can’t chop it directly in half,

since that would give us
identical rectangles of 2x4.

But the next closest option

  • 3x4 and 1x4 - works.

That was easy, but we’re not done yet.

Now take the area
of the largest rectangle,

and subtract the area of the smallest.

The result is our score,

and the goal is to get
as low a score as possible.

Here, the largest area is 12
and the smallest is 4,

giving us a score of 8.

Since we didn’t try to go
for a low score that time,

we can probably do better.

Let’s keep our 1x4

while breaking the 3x4
into a 3x3 and a 3x1.

Now our score is 9 minus 3, or 6.

Still not optimal, but better.

With such a small canvas,
there are only a few options.

But let’s see what happens
when the canvas gets bigger.

Try out an 8x8;
what’s the lowest score you can get?

Pause here if you want
to figure it out yourself.

Answer in: 3

Answer in: 2

Answer in: 1

To get our bearings,
we can start as before:

dividing the canvas roughly in two.

That gives us a 5x8 rectangle
with area 40

and a 3x8 with area 24,

for a score of 16.

That’s pretty bad.

Dividing that 5x8 into a 5x5
and a 5x3 leaves us with a score of 10.

Better, but still not great.

We could just keep dividing
the biggest rectangle.

But that would leave us
with increasingly tiny rectangles,

which would increase the range
between the largest and smallest.

What we really want

is for all our rectangles to fall
within a small range of area values.

And since the total area
of the canvas is 64,

the areas need to add up to that.

Let’s make a list
of possible rectangles and areas.

To improve on our previous score,

we can try to pick a range
of values spanning 9 or less

and adding up to 64.

You’ll notice that some values
are left out

because rectangles like 1x13
or 2x9 won’t fit on the canvas.

You might also realize

that if you use one of the rectangles
with an odd area like 5, 9, or 15,

you need to use another odd-value
rectangle to get an even sum.

With all that in mind,
let’s see what works.

Starting with area 20 or more
puts us over the limit too quickly.

But we can get to 64 using
rectangles in the 14-18 range,

leaving out 15.

Unfortunately, there’s no way
to make them fit.

Using the 2x7 leaves a gap

that can only be filled
by a rectangle with a width of 1.

Going lower, the next range
that works is 8 to 14,

leaving out the 3x3 square.

This time, the pieces fit.

That’s a score of 6.

Can we do even better?

No.

We can get the same score
by throwing out the 2x7 and 1x8

and replacing them
with a 3x3, 1x7, and 1x6.

But if we go any lower down the list,

the numbers become so small

that we’d need a wider range
of sizes to cover the canvas,

which would increase the score.

There’s no trick or formula here
– just a bit of intuition.

It’s more art than science.

And for larger grids,

expert mathematicians aren’t sure whether
they’ve found the lowest possible scores.

So how would you divide a 4x4,

10x10,

or 32x32 canvas?

Give it a try
and post your results in the comments.

荷兰艺术家 Piet Mondrian 的抽象
矩形画

启发了
数学家创造了双重挑战。

首先,我们必须
用不重叠的矩形完全覆盖一个正方形画布。

一切都必须是唯一的,所以如果我们使用 1x4,
我们不能在其他地方使用 4x1,

但是 2x2 矩形就可以了。

让我们试试看。

假设我们有一个 4x4 的画布。

我们不能直接把它切成两半,

因为那样会得到
相同的 2x4 矩形。

但下一个最接近的选项

  • 3x4 和 1x4 - 有效。

这很容易,但我们还没有完成。

现在取
最大矩形

的面积,减去最小矩形的面积。

结果就是我们的分数

,目标是
尽可能地降低分数。

在这里,最大的区域是 12
,最小的是 4,

给了我们 8 分。

由于我们当时没有尝试
去低分,

我们可能会做得更好。

让我们保持 1x4,

同时将 3x4
分解为 3x3 和 3x1。

现在我们的分数是 9 减去 3 或 6。

仍然不是最佳的,但更好。

有了这么小的画布,
只有几个选择。

但是让我们看看
当画布变大时会发生什么。

尝试 8x8;
你能得到的最低分数是多少?

如果您想
自己弄清楚,请在此处暂停。

答案:3

答案:2

答案:1

为了弄清楚方向,
我们可以像以前一样开始:

将画布大致分成两部分。

这给了我们一个面积为 40 的 5x8 矩形

和一个面积为 24 的 3x8 矩形

,得分为 16。

这非常糟糕。

将 5x8 划分为 5x5
和 5x3 使我们得到 10 分。

更好,但仍然不是很好。

我们可以继续
分割最大的矩形。

但这会给我们
留下越来越小的矩形,

这将增加
最大和最小之间的范围。

我们真正想要的

是让我们所有的矩形都
落在一个小的面积值范围内。

由于画布的总面积
为 64

,因此需要将这些面积相加。

让我们
列出可能的矩形和区域。

为了提高我们之前的分数,

我们可以尝试选择一个
不超过 9

并且加起来为 64 的值的范围。

您会注意到有些值
被遗漏了,

因为像 1x13
或 2x9 这样的矩形不适合画布。

您可能还意识到

,如果您使用其中一个
具有奇数面积(如 5、9 或 15)的矩形,

则需要使用另一个奇值
矩形来获得偶数和。

考虑到所有这些,
让我们看看什么是有效的。

从 20 区或更多区域开始,
我们很快就超过了极限。

但是我们可以使用
14-18 范围内的矩形得到 64,

而忽略 15。

不幸的是,没有
办法让它们适合。

使用 2x7 会留下

一个只能
由宽度为 1 的矩形填充的间隙。

再往下走,下一个有效范围
是 8 到 14,

不包括 3x3 正方形。

这一次,碎片合适。

那是6分。

我们能做得更好吗?

不。

我们可以
通过丢弃 2x7 和 1x8


用 3x3、1x7 和 1x6 替换它们来获得相同的分数。

但如果我们再往下看

,数字会变得如此之小

,以至于我们需要更大范围
的尺寸来覆盖画布,

这会增加分数。

这里没有技巧或公式
——只是一点直觉。

它比科学更艺术。

对于更大的网格,

专业数学家不确定
他们是否找到了可能的最低分数。

那么如何划分 4x4、10x10

或 32x32 的画布呢?

试一试,
并在评论中发布您的结果。