1

I'm having trouble with the following question:

A large cube is made up of 15625 small cubes. If the large cube is cut by as a plane so that the cross-section is a regular hexagon, how many small cubes are cut?

I also looked it up on google but nothing came up. I don't know if my approach is correct or not, but I thought I could find the number of un-cut small cubes in the question if I could find the pattern to the number of un-cut small cubes in cubes made up of 8, 27, 64 (and so on) small cubes. So I drew cubes like such and tried to count the unaffected small cubes, but it was really difficult when it got to a cube made up of 27 small cubes.

Sorry for my wording as English is not my native language. Could anyone give me a hint to solve this? Thanks!

Hexagonal cross-section of a cube

  • 2
    Are we given that the cross-section is a regular hexagon? – saulspatz Jul 28 '20 at 06:59
  • Yes, I edited my post. Sorry about that – chocolatte Jul 28 '20 at 08:31
  • I think the trick here is to forget about cubes and just think about tessellating a hexagon, with hexagons scaled down by a factor of 25. The number of hexagons required should be the number of cubes. – tkf Jul 28 '20 at 08:50
  • 1
    @tkf It is not just hexagons. The cutting plane between each layer of the cube cuts through the hexagon in a straight line. So the hexagon is cut by a three sets of parallel lines. This results in a tiling of hexagons with triangles in between (a 3.6.3.6 tiling). Now you have to count the hexagons and triangles. – Jaap Scherphuis Jul 28 '20 at 08:57
  • Good point - I thought the triangles would just be at the edges. – tkf Jul 28 '20 at 09:02

2 Answers2

3

Let's define some coordinates first. Say the cube is $[0, n]^3$ (so vertices at $\{(x, y, z) \in \mathbb R^3 : x, y, z \in \{0, n\}\}$), where $n \in \mathbb N$ is the number of subdivisions you want (in your case $25$, since $25^3 = 15625$). The smaller cubes have their vertices on integer lattice points whose coordinates are between $0$ and $n$. Specifically, for each lattice point $(a, b, c)$ with $a, b, c \in \{0, \dotsc, n-1\}$ there is a small cube $$ [a, a+1] \times [b, b+1] \times [c, c+1] \,. $$ We'll call $(a, b, c)$ the "low corner" of this cube, and $(a+1, b+1, c+1)$ the "high corner".

In order to find a cross section of the large cube which is a regular hexagon, we take the intersection with the plane $x + y + z = 1.5n$. This plane divides the cube in half, intersecting the cube's "frame" at the midpoints of six edges. I believe this is the type of cross section you had in mind.

In order to count how many small cubes are cut, we just need to count how many small cubes have their "low corner" $(a, b, c)$ and "high corner" $(a+1, b+1, c+1)$ on opposite sides of this plane. This occurs when $$ a + b + c \leq 1.5n \leq (a+1) + (b+1) + (c+1) \,,$$ or equivalently $$ a + b + c \in [1.5n-3, 1.5n] \,. $$ Since $a+b+c$ is an integer, all we need to do is count the number of lattice points $(a, b, c)$ with $a, b, c \in \{0, \dotsc, n-1\}$ such that $$ a + b + c \in \{\lceil1.5n\rceil-3, \dotsc, \lfloor1.5n\rfloor\} \,. $$ This means the number of small cubes cut is $$ C(n) = \sum_{s=\lceil1.5n\rceil-3}^{\lfloor1.5n\rfloor} S_3(s; n) \,, \tag{$*$} $$ where $S_3(s; n)$ is defined as the number of ways to sum $3$ nonnegative integers less than $n$ to get $s$. (Define $S_k(s; n)$ similarly for other $k$.)


This $S_k(s; n)$ might have some well-known name, but I am not aware of it. Either way, we can derive it by applying the approach from this answer. First of all, the total number of ways of summing $k$ nonnegative integers to get $s$ is $$ S_k(s) = \begin{cases} \binom{s+k-1}{k-1} &:\, s \geq 0 \\ 0 &:\, \text{otherwise} \end{cases} \,. $$ Among these, the number of ways where the first number is $\geq n$ is $$ S_k(s-n) \,, $$ and likewise for the number of ways where the second number is $\geq n$, etc. The number of ways where the first two (or indeed, any particular two) numbers are $\geq n$ is $$ S_k(s-2n) \,. $$ In general, if we require $j$ of the numbers to be $\geq n$, the number of ways of summing to get $s$ is $$ S_k(s-jn) \,. $$ Using Inclusion-Exclusion, the total number of ways of summing to get $s$ where at least one of the summands is $\geq n$, is $$ \sum_{j=1}^k (-1)^{j-1} \binom{k}{j} S_k(s-jn) \,, $$ and thus \begin{align*} S_k(s; n) &= S_k(s) - \sum_{j=1}^k (-1)^{j-1} \binom{k}{j} S_k(s-jn) \\ &= \sum_{j=0}^k (-1)^j \binom{k}{j} S_k(s-jn) \,. \end{align*}


Now, let's apply this result back to $(\ast)$. We wanted to compute $S_3(s; n)$, but only cared about cases where $s \leq \lfloor1.5n\rfloor < 2n$. Assuming $s < 2n$, the formula for $S_3(s; n)$ simplifies to $$ S_3(s) - 3 S_3(s-n) \,, $$ so the final answer for the number of small cubes cut is \begin{align*} C(n) &= \sum_{s=\lceil1.5n\rceil-3}^{\lfloor1.5n\rfloor} \big(S_3(s) - 3 S_3(s-n)\big) \\ &= \sum_{s=\lceil1.5n\rceil-3}^{\lfloor1.5n\rfloor} S_3(s) - 3 \sum_{s=\lceil0.5n\rceil-3}^{\lfloor0.5n\rfloor} S_3(s) \\ &= \sum_{s=\max(0, \lceil1.5n\rceil-3)}^{\lfloor1.5n\rfloor} \binom{s+2}{2} - 3 \sum_{s=\max(0, \lceil0.5n\rceil-3)}^{\lfloor0.5n\rfloor} \binom{s+2}{2} \\ &= \frac12 \sum_{s=\max(2, \lceil1.5n\rceil-1)}^{\lfloor1.5n\rfloor+2} s(s-1) - \frac32 \sum_{s=\max(2, \lceil0.5n\rceil-1)}^{\lfloor0.5n\rfloor+2} s(s-1) \,. \end{align*}

Edit (from comments): if $n$ is odd and $\geq 5$, we can instead simplify the expression to \begin{align*} C(n) &= \sum_{s=(3n-5)/2}^{(3n-1)/2} S_3(s) - 3 \sum_{s=(n-5)/2}^{(n-1)/2} S_3(s) \\ &= \sum_{s=(3n-1)/2}^{(3n+3)/2} \frac{s(s-1)}{2} - 3 \sum_{s=(n-1)/2}^{(n+3)/2} \frac{s(s-1)}{2} \\ &= \frac{(3n-1)(3n-3) + (3n+1)(3n-1) + (3n+3)(3n+1)}{8} \\ &\quad - 3\cdot \frac{(n-1)(n-3) + (n+1)(n-1) + (n+3)(n+1)}{8} \\ &= \frac{27n^2+5}{8} - \frac{9n^2+15}{8} \\ &= \frac{9n^2-5}{4} \,. \end{align*} In fact, this expression also holds for $n=1$ and $n=3$ (see below).


In the original problem, $n$ was $25$. Plugging this in, we get an answer of $$ C(25) = (9\cdot25^2 - 5)/4 = \boxed{1405} \,. $$ As a sanity check, let's also compute $C(1)$, $C(2)$, and $C(3)$: \begin{align*} C(1) &= \frac12 \sum_{s=2}^3 s(s-1) - \frac32 \sum_{s=2}^2 s(s-1) = 1 \,, \\ C(2) &= \frac12 \sum_{s=2}^5 s(s-1) - \frac32 \sum_{s=2}^3 s(s-1) = 8 \,, \\ C(3) &= \frac12 \sum_{s=4}^6 s(s-1) - \frac32 \sum_{s=2}^3 s(s-1) = 19 \,. \end{align*} These numbers match what I calculated by hand.

eccheng
  • 511
  • I think your proofs looks right to me. Thank you so much for explaining in such detail. – chocolatte Jul 28 '20 at 09:55
  • 1
    That seems to come to $(9n^2-5)/4$ for odd $n$. Does that match your other values? – Empy2 Jul 28 '20 at 09:59
  • @Empy2 Good point! I'll add it to the answer. I have given up on simplifying because of the ugliness introduced by even $n$, but your form is certainly nice enough that it deserves to be included. – eccheng Jul 28 '20 at 10:21
  • Very well written answer. Empirically, it seems that when $n$ is even, the formula simplifies to $3n^2-4$ – saulspatz Jul 28 '20 at 19:20
1

A short simple solution was provided by @JaapScherphuis right at the start, in the comments. However it seems to have been overlooked, so I will elaborate his answer:

We cut the large cube into $(2k+1)^3$ smaller cubes in the obvious way (so $k=12$ is the original case posed). The layers of the cube are evenly spaced so they slice the cross sectional hexagon evenly into $n=2k+1$ slices parallel to each edge of the regular hexagon.

Note that as $2k+1$ is odd, there are parallel cuts either side of the main diagonals. Thus if you mark out where parallel cuts start along an edge of the hexagon, the separations along the edge are equal, except for the last which is half the usual separation to the adjacent vertex. This means that as you move along the base edge, the diagonals alternate between going up left, and up right.

$k=1:$

enter image description here

The number of cubes sliced is simply the number of regions this hexagon is divided into.

Label the layers of the hexagon from the bottom to middle (but one) $r_0,r_1,\cdots,r_{k-1}$. The bottom layer $r_0$ has $k+1$ hexagons, $k$ triangles below and $k+2$ triangles above. In general $r_i$ has $k+1+i$ hexagons, $k+i$ triangles below and $k+i+2$ triangles above. The middle layer has $2k+1$ hexagons, $2k$ triangles below and $2k$ triangles above. Counting the top half as well, we get the number of cubes sliced as: $$ 6k+1+2\sum_{i=0}^{k-1} 3(k+i+1)=9k^2+9k+1=\frac{9n^2-5}{4}. $$

tkf
  • 11,563